File size: 6,085 Bytes
ccef021
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
#include <torch/library.h>
#include <ATen/Tensor.h>

#include "registration.h"
#include "torch_binding.h"

// Wrapper functions that adapt the interface for TORCH_LIBRARY

std::tuple<at::Tensor, at::Tensor, std::optional<at::Tensor>, std::optional<at::Tensor>>
sparse_decode_fwd(
    const at::Tensor &q,
    const at::Tensor &kv,
    const at::Tensor &indices,
    const std::optional<at::Tensor> &topk_length,
    const std::optional<at::Tensor> &attn_sink,
    const std::optional<at::Tensor> &tile_scheduler_metadata,
    const std::optional<at::Tensor> &num_splits,
    const std::optional<at::Tensor> &extra_kv,
    const std::optional<at::Tensor> &extra_indices,
    const std::optional<at::Tensor> &extra_topk_length,
    int64_t d_v,
    double sm_scale
) {
    // Create mutable copies for the interface that modifies these
    std::optional<at::Tensor> tile_scheduler_metadata_mut = tile_scheduler_metadata;
    std::optional<at::Tensor> num_splits_mut = num_splits;
    return sparse_attn_decode_interface(
        q, kv, indices, topk_length, attn_sink,
        tile_scheduler_metadata_mut, num_splits_mut,
        extra_kv, extra_indices, extra_topk_length,
        static_cast<int>(d_v), static_cast<float>(sm_scale)
    );
}

std::tuple<at::Tensor, at::Tensor, std::optional<at::Tensor>, std::optional<at::Tensor>>
dense_decode_fwd(
    at::Tensor q,
    const at::Tensor &kcache,
    int64_t head_size_v,
    const at::Tensor &seqlens_k,
    const at::Tensor &block_table,
    double softmax_scale,
    bool is_causal,
    const std::optional<at::Tensor> &tile_scheduler_metadata,
    const std::optional<at::Tensor> &num_splits
) {
    // Create mutable copies for the interface that modifies these
    std::optional<at::Tensor> tile_scheduler_metadata_mut = tile_scheduler_metadata;
    std::optional<at::Tensor> num_splits_mut = num_splits;
    return dense_attn_decode_interface(
        q, kcache, static_cast<int>(head_size_v),
        seqlens_k, block_table,
        static_cast<float>(softmax_scale), is_causal,
        tile_scheduler_metadata_mut, num_splits_mut
    );
}

std::vector<at::Tensor> sparse_prefill_fwd(
    const at::Tensor &q,
    const at::Tensor &kv,
    const at::Tensor &indices,
    double sm_scale,
    int64_t d_v,
    const std::optional<at::Tensor> &attn_sink,
    const std::optional<at::Tensor> &topk_length
) {
    return sparse_attn_prefill_interface(
        q, kv, indices,
        static_cast<float>(sm_scale), static_cast<int>(d_v),
        attn_sink, topk_length
    );
}

void dense_prefill_fwd(
    at::Tensor workspace_buffer,
    at::Tensor q,
    at::Tensor k,
    at::Tensor v,
    at::Tensor cumulative_seqlen_q,
    at::Tensor cumulative_seqlen_kv,
    at::Tensor o,
    at::Tensor lse,
    int64_t mask_mode_code,
    double softmax_scale,
    int64_t max_seqlen_q,
    int64_t max_seqlen_kv,
    bool is_varlen
) {
    FMHACutlassSM100FwdRun(
        workspace_buffer, q, k, v,
        cumulative_seqlen_q, cumulative_seqlen_kv,
        o, lse,
        static_cast<int>(mask_mode_code), static_cast<float>(softmax_scale),
        static_cast<int>(max_seqlen_q), static_cast<int>(max_seqlen_kv),
        is_varlen
    );
}

void dense_prefill_bwd(
    at::Tensor workspace_buffer,
    at::Tensor d_o,
    at::Tensor q,
    at::Tensor k,
    at::Tensor v,
    at::Tensor o,
    at::Tensor lse,
    at::Tensor cumulative_seqlen_q,
    at::Tensor cumulative_seqlen_kv,
    at::Tensor dq,
    at::Tensor dk,
    at::Tensor dv,
    int64_t mask_mode_code,
    double softmax_scale,
    int64_t max_seqlen_q,
    int64_t max_seqlen_kv,
    bool is_varlen
) {
    FMHACutlassSM100BwdRun(
        workspace_buffer, d_o, q, k, v, o, lse,
        cumulative_seqlen_q, cumulative_seqlen_kv,
        dq, dk, dv,
        static_cast<int>(mask_mode_code), static_cast<float>(softmax_scale),
        static_cast<int>(max_seqlen_q), static_cast<int>(max_seqlen_kv),
        is_varlen
    );
}

TORCH_LIBRARY_EXPAND(TORCH_EXTENSION_NAME, ops) {
    // Sparse decode forward
    ops.def(
        "sparse_decode_fwd(Tensor q, Tensor kv, Tensor indices, "
        "Tensor? topk_length, Tensor? attn_sink, "
        "Tensor? tile_scheduler_metadata, Tensor? num_splits, "
        "Tensor? extra_kv, Tensor? extra_indices, Tensor? extra_topk_length, "
        "int d_v, float sm_scale) -> (Tensor, Tensor, Tensor?, Tensor?)"
    );
    ops.impl("sparse_decode_fwd", torch::kCUDA, &sparse_decode_fwd);

    // Dense decode forward
    ops.def(
        "dense_decode_fwd(Tensor q, Tensor kcache, int head_size_v, "
        "Tensor seqlens_k, Tensor block_table, "
        "float softmax_scale, bool is_causal, "
        "Tensor? tile_scheduler_metadata, Tensor? num_splits) -> (Tensor, Tensor, Tensor?, Tensor?)"
    );
    ops.impl("dense_decode_fwd", torch::kCUDA, &dense_decode_fwd);

    // Sparse prefill forward
    ops.def(
        "sparse_prefill_fwd(Tensor q, Tensor kv, Tensor indices, "
        "float sm_scale, int d_v, "
        "Tensor? attn_sink, Tensor? topk_length) -> Tensor[]"
    );
    ops.impl("sparse_prefill_fwd", torch::kCUDA, &sparse_prefill_fwd);

    // Dense prefill forward (SM100)
    ops.def(
        "dense_prefill_fwd(Tensor workspace_buffer, Tensor q, Tensor k, Tensor v, "
        "Tensor cumulative_seqlen_q, Tensor cumulative_seqlen_kv, "
        "Tensor o, Tensor lse, "
        "int mask_mode_code, float softmax_scale, "
        "int max_seqlen_q, int max_seqlen_kv, bool is_varlen) -> ()"
    );
    ops.impl("dense_prefill_fwd", torch::kCUDA, &dense_prefill_fwd);

    // Dense prefill backward (SM100)
    ops.def(
        "dense_prefill_bwd(Tensor workspace_buffer, Tensor d_o, "
        "Tensor q, Tensor k, Tensor v, Tensor o, Tensor lse, "
        "Tensor cumulative_seqlen_q, Tensor cumulative_seqlen_kv, "
        "Tensor dq, Tensor dk, Tensor dv, "
        "int mask_mode_code, float softmax_scale, "
        "int max_seqlen_q, int max_seqlen_kv, bool is_varlen) -> ()"
    );
    ops.impl("dense_prefill_bwd", torch::kCUDA, &dense_prefill_bwd);
}

REGISTER_EXTENSION(TORCH_EXTENSION_NAME)