TylerHilbert commited on
Commit
b7a26d0
·
1 Parent(s): 0d7e51f

Progress on CUDA / OpenCL category

Browse files
PyTorchConference2025_GithubRepos.json CHANGED
@@ -320,21 +320,25 @@
320
  "github_topic_closest_fit": "performance-analysis"
321
  },
322
  {
323
- "repo_link": "https://github.com/vosen/ZLUDA",
324
  "repo_name": "ZLUDA",
 
 
325
  "github_about_section": "CUDA on non-NVIDIA GPUs",
326
  "homepage_link": "https://vosen.github.io/ZLUDA/",
327
  "github_topic_closest_fit": "cuda"
328
  },
329
  {
330
- "repo_link": "https://github.com/vtsynergy/CU2CL",
331
  "repo_name": "CU2CL",
 
 
332
  "github_about_section": "A prototype CUDA-to-OpenCL source-to-source translator, built on the Clang compiler framework",
333
- "homepage_link": "http://chrec.cs.vt.edu/cu2cl"
 
334
  },
335
  {
336
- "repo_link": "https://github.com/pocl/pocl",
337
  "repo_name": "pocl",
 
 
338
  "github_about_section": "pocl - Portable Computing Language",
339
  "homepage_link": "https://portablecl.org",
340
  "github_topic_closest_fit": "opencl"
@@ -410,6 +414,7 @@
410
  {
411
  "repo_link": "https://github.com/NVIDIA/nccl",
412
  "repo_name": "nccl",
 
413
  "github_about_section": "Optimized primitives for collective multi-GPU communication",
414
  "homepage_link": "https://docs.nvidia.com/deeplearning/nccl/user-guide/docs/index.html",
415
  "github_topic_closest_fit": "cuda"
@@ -678,6 +683,7 @@
678
  {
679
  "repo_link": "https://github.com/NVIDIA/cuda-python",
680
  "repo_name": "cuda-python",
 
681
  "github_about_section": "CUDA Python: Performance meets Productivity",
682
  "homepage_link": "https://nvidia.github.io/cuda-python/"
683
  },
@@ -696,16 +702,35 @@
696
  "github_topic_closest_fit": "inference"
697
  },
698
  {
699
- "repo_link": "https://github.com/KhronosGroup/SYCL-Docs",
700
  "repo_name": "SYCL-Docs",
701
- "github_about_section": "SYCL Open Source Specification"
 
 
 
702
  },
703
  {
704
- "repo_link": "https://github.com/triSYCL/triSYCL",
705
  "repo_name": "triSYCL",
 
 
706
  "github_about_section": "Generic system-wide modern C++ for heterogeneous platforms with SYCL from Khronos Group",
707
  "github_topic_closest_fit": "opencl"
708
  },
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
709
  {
710
  "repo_link": "https://github.com/pybind/pybind11",
711
  "repo_name": "pybind11",
@@ -734,9 +759,11 @@
734
  "category": "compiler"
735
  },
736
  {
737
- "repo_link": "https://github.com/KhronosGroup/OpenCL-SDK",
738
  "repo_name": "OpenCL-SDK",
739
- "github_about_section": "OpenCL SDK"
 
 
 
740
  },
741
  {
742
  "repo_link": "https://github.com/meta-llama/synthetic-data-kit",
 
320
  "github_topic_closest_fit": "performance-analysis"
321
  },
322
  {
 
323
  "repo_name": "ZLUDA",
324
+ "repo_link": "https://github.com/vosen/ZLUDA",
325
+ "category": "CUDA / OpenCL",
326
  "github_about_section": "CUDA on non-NVIDIA GPUs",
327
  "homepage_link": "https://vosen.github.io/ZLUDA/",
328
  "github_topic_closest_fit": "cuda"
329
  },
330
  {
 
331
  "repo_name": "CU2CL",
332
+ "repo_link": "https://github.com/vtsynergy/CU2CL",
333
+ "category": "CUDA / OpenCL",
334
  "github_about_section": "A prototype CUDA-to-OpenCL source-to-source translator, built on the Clang compiler framework",
335
+ "homepage_link": "http://chrec.cs.vt.edu/cu2cl",
336
+ "github_topic_closest_fit": "opencl"
337
  },
338
  {
 
339
  "repo_name": "pocl",
340
+ "repo_link": "https://github.com/pocl/pocl",
341
+ "category": "CUDA / OpenCL",
342
  "github_about_section": "pocl - Portable Computing Language",
343
  "homepage_link": "https://portablecl.org",
344
  "github_topic_closest_fit": "opencl"
 
414
  {
415
  "repo_link": "https://github.com/NVIDIA/nccl",
416
  "repo_name": "nccl",
417
+ "category": "CUDA / OpenCL",
418
  "github_about_section": "Optimized primitives for collective multi-GPU communication",
419
  "homepage_link": "https://docs.nvidia.com/deeplearning/nccl/user-guide/docs/index.html",
420
  "github_topic_closest_fit": "cuda"
 
683
  {
684
  "repo_link": "https://github.com/NVIDIA/cuda-python",
685
  "repo_name": "cuda-python",
686
+ "category": "CUDA / OpenCL",
687
  "github_about_section": "CUDA Python: Performance meets Productivity",
688
  "homepage_link": "https://nvidia.github.io/cuda-python/"
689
  },
 
702
  "github_topic_closest_fit": "inference"
703
  },
704
  {
 
705
  "repo_name": "SYCL-Docs",
706
+ "repo_link": "https://github.com/KhronosGroup/SYCL-Docs",
707
+ "category": "CUDA / OpenCL",
708
+ "github_about_section": "SYCL Open Source Specification",
709
+ "github_topic_closest_fit": "opencl"
710
  },
711
  {
 
712
  "repo_name": "triSYCL",
713
+ "repo_link": "https://github.com/triSYCL/triSYCL",
714
+ "category": "CUDA / OpenCL",
715
  "github_about_section": "Generic system-wide modern C++ for heterogeneous platforms with SYCL from Khronos Group",
716
  "github_topic_closest_fit": "opencl"
717
  },
718
+ {
719
+ "repo_name": "AdaptiveCpp",
720
+ "repo_link": "https://github.com/AdaptiveCpp/AdaptiveCpp",
721
+ "category": "compiler",
722
+ "github_about_section": "Compiler for multiple programming models (SYCL, C++ standard parallelism, HIP/CUDA) for CPUs and GPUs from all vendors: The independent, community-driven compiler for C++-based heterogeneous programming models. Lets applications adapt themselves to all the hardware in the system - even at runtime!",
723
+ "homepage_link": "https://adaptivecpp.github.io",
724
+ "github_topic_closest_fit": "compiler"
725
+ },
726
+ {
727
+ "repo_name": "oneDPL",
728
+ "repo_link": "https://github.com/uxlfoundation/oneDPL",
729
+ "category": "compiler",
730
+ "github_about_section": "oneAPI DPC++ Library (oneDPL)",
731
+ "homepage_link": "https://software.intel.com/content/www/us/en/develop/tools/oneapi/components/dpc-library.html",
732
+ "github_topic_closest_fit": "compiler"
733
+ },
734
  {
735
  "repo_link": "https://github.com/pybind/pybind11",
736
  "repo_name": "pybind11",
 
759
  "category": "compiler"
760
  },
761
  {
 
762
  "repo_name": "OpenCL-SDK",
763
+ "repo_link": "https://github.com/KhronosGroup/OpenCL-SDK",
764
+ "category": "CUDA / OpenCL",
765
+ "github_about_section": "OpenCL SDK",
766
+ "github_topic_closest_fit": "opencl"
767
  },
768
  {
769
  "repo_link": "https://github.com/meta-llama/synthetic-data-kit",