ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-05-13 17:55:48 +00:00
Code Issues Packages Projects Releases Wiki Activity
Files
b686143624f2d7154c3f5fb235df317100b22e45
composable_kernel/example/ck_tile/42_unified_attention/unified_attention.cpp

312 lines
17 KiB
C++
Raw Blame History

// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2025, Advanced Micro Devices, Inc. All rights reserved.
#include "unified_attention.hpp"
#include "unified_attention_impl.hpp"
#include "mask.hpp"
namespace ck_tile {
std::ostream& operator<<(std::ostream& stream,
const unified_attention_args::data_type_enum& data_type)
{
switch(data_type)
{
case unified_attention_args::data_type_enum::fp16: return stream << "fp16";
case unified_attention_args::data_type_enum::bf16: return stream << "bf16";
default: return stream << "unknown";
}
}
// Helper macro to reduce dispatch boilerplate.
// Dispatches based on DataType, IsMasking, HeadSize, BlockM, NumQPerKV.
#define DISPATCH_UNIFIED_ATTENTION(DType, IsMask, HSize, BM, NQPKV) \
{ \
using kernel_traits = unified_attention_kernel_traits<DType, IsMask, HSize, BM, NQPKV>; \
return unified_attention_kernel_dispatch<kernel_traits>(args, config); \
}
// SWA-aware variant: requires explicit BlockSize (since IsLocal is the 7th template arg).
// HeadSize<=64 -> BlockSize=64; HeadSize=128 -> BlockSize=32. Caller must supply.
#define DISPATCH_UNIFIED_ATTENTION_LOCAL(DType, HSize, BM, NQPKV, BSize) \
{ \
using kernel_traits = unified_attention_kernel_traits<DType, /*IsMasking=*/true, HSize, BM, NQPKV, BSize, /*IsLocal=*/true>; \
return unified_attention_kernel_dispatch<kernel_traits>(args, config); \
}
// SWA-aware decode dispatchers for bs32. These mirror the non-local *_BS32 macros
// but flip IsLocal=true on the 7th template arg, so the kernel uses the
// sliding-window mask AND the per-Q-tile KV-block iteration clipping.
#define DISPATCH_UNIFIED_ATTENTION_DECODE_MEDIUM_BS32_LOCAL(DType, HSize, BM, NQPKV) \
{ \
using kernel_traits = unified_attention_decode_kernel_traits<DType, /*IsMasking=*/true, HSize, BM, NQPKV, 32, /*IsLocal=*/true>; \
return unified_attention_kernel_dispatch<kernel_traits>(args, config); \
}
#define DISPATCH_UNIFIED_ATTENTION_DECODE_BS32_NARROW_LOCAL(DType, HSize, BM, NQPKV) \
{ \
using kernel_traits = unified_attention_decode_bs32_kernel_traits<DType, /*IsMasking=*/true, HSize, BM, NQPKV, 32, /*IsLocal=*/true>; \
return unified_attention_kernel_dispatch_decode<kernel_traits>(args, config); \
}
// Dispatch macros for three tile tiers (default block_size).
#define DISPATCH_UNIFIED_ATTENTION_DECODE_MEDIUM(DType, IsMask, HSize, BM, NQPKV) \
{ \
using kernel_traits = unified_attention_decode_kernel_traits<DType, IsMask, HSize, BM, NQPKV>; \
return unified_attention_kernel_dispatch<kernel_traits>(args, config); \
}
#define DISPATCH_UNIFIED_ATTENTION_DECODE_SMALL(DType, IsMask, HSize, BM, NQPKV) \
{ \
using kernel_traits = unified_attention_decode_small_kernel_traits<DType, IsMask, HSize, BM, NQPKV>; \
return unified_attention_kernel_dispatch_decode<kernel_traits>(args, config); \
}
#define DISPATCH_UNIFIED_ATTENTION_DECODE_TINY(DType, IsMask, HSize, BM, NQPKV) \
{ \
using kernel_traits = unified_attention_decode_tiny_kernel_traits<DType, IsMask, HSize, BM, NQPKV>; \
return unified_attention_kernel_dispatch_decode<kernel_traits>(args, config); \
}
// block_size=32 dispatch macros (6th template arg = 32).
#define DISPATCH_UNIFIED_ATTENTION_DECODE_MEDIUM_BS32(DType, IsMask, HSize, BM, NQPKV) \
{ \
using kernel_traits = unified_attention_decode_kernel_traits<DType, IsMask, HSize, BM, NQPKV, 32>; \
return unified_attention_kernel_dispatch<kernel_traits>(args, config); \
}
#define DISPATCH_UNIFIED_ATTENTION_DECODE_SMALL_BS32(DType, IsMask, HSize, BM, NQPKV) \
{ \
using kernel_traits = unified_attention_decode_small_kernel_traits<DType, IsMask, HSize, BM, NQPKV, 32>; \
return unified_attention_kernel_dispatch_decode<kernel_traits>(args, config); \
}
#define DISPATCH_UNIFIED_ATTENTION_DECODE_BS32_NARROW(DType, IsMask, HSize, BM, NQPKV) \
{ \
using kernel_traits = unified_attention_decode_bs32_kernel_traits<DType, IsMask, HSize, BM, NQPKV, 32>; \
return unified_attention_kernel_dispatch_decode<kernel_traits>(args, config); \
}
enum class tile_tier { large, medium, small, tiny };
static tile_tier select_tile_tier(const unified_attention_args& args)
{
const index_t avg_q = args.num_seqs > 0 ? args.num_tokens / args.num_seqs : args.num_tokens;
const index_t kBlockQ_tiny = 16 / args.num_queries_per_kv;
const index_t kBlockQ_small = 64 / args.num_queries_per_kv;
[[maybe_unused]] const index_t kBlockQ_medium = 128 / args.num_queries_per_kv;
// Decode tiers use a 2D grid (num_kv_heads, num_seqs) that assumes each
// seq has at most kBlockQ tokens. For mixed batches where some seqs have
// many more tokens, we must use the medium tier (1D grid with Q tile iteration).
const index_t max_q = args.max_seqlen_q > 0 ? args.max_seqlen_q : avg_q;
if(avg_q <= kBlockQ_tiny && max_q <= kBlockQ_tiny)
return tile_tier::tiny;
if(avg_q <= kBlockQ_small && max_q <= kBlockQ_small)
return tile_tier::small;
return tile_tier::medium;
}
std::pair<bool, float> unified_attention(const unified_attention_args& args,
const stream_config& config)
{
const bool is_mask = (args.mask_type != static_cast<int>(mask_enum::no_mask));
// Real SWA = "at least one non-trivial window edge". Plain causal lives at
// (left=-1, right=0); without this guard it would hit the IsLocal=true path
// and fail for shape tiers where we have not (yet) instantiated local kernels.
// left >= 0 : finite look-back (e.g. causal SWA, dense SWA, diagonal-only)
// right > 0 : finite look-ahead (bidirectional SWA, anti-causal SWA)
// Note "right >= 0" would mis-classify plain causal (right=0) as SWA.
const bool is_local =
is_mask && (args.window_size_left >= 0 || args.window_size_right > 0);
auto tier = select_tile_tier(args);
// SWA-instance availability matrix (IsLocal=true). Anything not listed here
// returns {false, 0} so the caller (e.g. _try_ck_unified_attention) falls
// back to a backend that handles it (Triton). Falling through to the
// IsLocal=false path would silently ignore window_size_left and produce
// wrong outputs, so we reject explicitly.
//
// shape | page_blk_size | tier we route to | instance
// ---------------+---------------+-----------------------------+---------------------
// d128 MHA | >= 32 | tile_tier::large | d128_*_mask_local
// d64 GQA-8 | >= 64 | tile_tier::large | d64_*_mask_gqa8_local
// d64 GQA-8 | == 32, tiny | tile_tier::tiny | d64_*_mask_gqa8_bs32_narrow_local
// d64 GQA-8 | == 32, med | tile_tier::medium | d64_*_mask_gqa8_bs32_decode_local
// (small+bs32 SWA has no instance yet -> Triton fallback; GPT-OSS shows zero
// such shapes in practice. Bumping it to medium is plausible but wastes a
// full kBlockM=128 tile when only kBlockM=64 is needed -- revisit if the
// workload changes.)
if(is_local)
{
const bool d128_mha = (args.hdim == 128 && args.num_queries_per_kv == 1);
const bool d64_gqa8 = (args.hdim == 64 && args.num_queries_per_kv == 8);
if(d128_mha)
{
if(args.page_blk_size < 32) return {false, 0.f};
tier = tile_tier::large;
}
else if(d64_gqa8)
{
if(args.page_blk_size >= 64)
{
tier = tile_tier::large;
}
else if(args.page_blk_size >= 32)
{
if(tier != tile_tier::tiny && tier != tile_tier::medium)
return {false, 0.f};
// keep selected tier as-is
}
else
{
return {false, 0.f};
}
}
else
{
return {false, 0.f};
}
}
// d128, MHA (num_queries_per_kv == 1)
if(args.hdim == 128 && args.num_queries_per_kv == 1)
{
if(args.data_type == unified_attention_args::data_type_enum::fp16)
{
if(!is_mask) DISPATCH_UNIFIED_ATTENTION(unified_attention_args::data_type_enum::fp16, false, 128, 256, 1)
else if(is_local) DISPATCH_UNIFIED_ATTENTION_LOCAL(unified_attention_args::data_type_enum::fp16, 128, 256, 1, 32)
else DISPATCH_UNIFIED_ATTENTION(unified_attention_args::data_type_enum::fp16, true, 128, 256, 1)
}
else if(args.data_type == unified_attention_args::data_type_enum::bf16)
{
if(!is_mask) DISPATCH_UNIFIED_ATTENTION(unified_attention_args::data_type_enum::bf16, false, 128, 256, 1)
else if(is_local) DISPATCH_UNIFIED_ATTENTION_LOCAL(unified_attention_args::data_type_enum::bf16, 128, 256, 1, 32)
else DISPATCH_UNIFIED_ATTENTION(unified_attention_args::data_type_enum::bf16, true, 128, 256, 1)
}
}
// d64, GQA-8 (num_queries_per_kv == 8)
if(args.hdim == 64 && args.num_queries_per_kv == 8)
{
const bool use_bs32 = (args.page_blk_size < 64);
if(tier == tile_tier::tiny)
{
if(use_bs32) {
// bs32 narrow: 2 warps, 16x16 MFMA, kBlockM=32, kBlockQ=4.
// Avoids 1-warp race condition; 2x less waste than small tier.
if(args.data_type == unified_attention_args::data_type_enum::fp16)
{
if(!is_mask) DISPATCH_UNIFIED_ATTENTION_DECODE_BS32_NARROW(unified_attention_args::data_type_enum::fp16, false, 64, 32, 8)
else if(is_local) DISPATCH_UNIFIED_ATTENTION_DECODE_BS32_NARROW_LOCAL(unified_attention_args::data_type_enum::fp16, 64, 32, 8)
else DISPATCH_UNIFIED_ATTENTION_DECODE_BS32_NARROW(unified_attention_args::data_type_enum::fp16, true, 64, 32, 8)
}
else if(args.data_type == unified_attention_args::data_type_enum::bf16)
{
if(!is_mask) DISPATCH_UNIFIED_ATTENTION_DECODE_BS32_NARROW(unified_attention_args::data_type_enum::bf16, false, 64, 32, 8)
else if(is_local) DISPATCH_UNIFIED_ATTENTION_DECODE_BS32_NARROW_LOCAL(unified_attention_args::data_type_enum::bf16, 64, 32, 8)
else DISPATCH_UNIFIED_ATTENTION_DECODE_BS32_NARROW(unified_attention_args::data_type_enum::bf16, true, 64, 32, 8)
}
} else {
// bs64 tiny: 1 warp, 16x16 MFMA, kBlockM=16, kBlockQ=2.
if(args.data_type == unified_attention_args::data_type_enum::fp16)
{
if(!is_mask) DISPATCH_UNIFIED_ATTENTION_DECODE_TINY(unified_attention_args::data_type_enum::fp16, false, 64, 16, 8)
else DISPATCH_UNIFIED_ATTENTION_DECODE_TINY(unified_attention_args::data_type_enum::fp16, true, 64, 16, 8)
}
else if(args.data_type == unified_attention_args::data_type_enum::bf16)
{
if(!is_mask) DISPATCH_UNIFIED_ATTENTION_DECODE_TINY(unified_attention_args::data_type_enum::bf16, false, 64, 16, 8)
else DISPATCH_UNIFIED_ATTENTION_DECODE_TINY(unified_attention_args::data_type_enum::bf16, true, 64, 16, 8)
}
}
}
else if(tier == tile_tier::small)
{
if(args.data_type == unified_attention_args::data_type_enum::fp16)
{
if(use_bs32) {
if(!is_mask) DISPATCH_UNIFIED_ATTENTION_DECODE_SMALL_BS32(unified_attention_args::data_type_enum::fp16, false, 64, 64, 8)
else DISPATCH_UNIFIED_ATTENTION_DECODE_SMALL_BS32(unified_attention_args::data_type_enum::fp16, true, 64, 64, 8)
} else {
if(!is_mask) DISPATCH_UNIFIED_ATTENTION_DECODE_SMALL(unified_attention_args::data_type_enum::fp16, false, 64, 64, 8)
else DISPATCH_UNIFIED_ATTENTION_DECODE_SMALL(unified_attention_args::data_type_enum::fp16, true, 64, 64, 8)
}
}
else if(args.data_type == unified_attention_args::data_type_enum::bf16)
{
if(use_bs32) {
if(!is_mask) DISPATCH_UNIFIED_ATTENTION_DECODE_SMALL_BS32(unified_attention_args::data_type_enum::bf16, false, 64, 64, 8)
else DISPATCH_UNIFIED_ATTENTION_DECODE_SMALL_BS32(unified_attention_args::data_type_enum::bf16, true, 64, 64, 8)
} else {
if(!is_mask) DISPATCH_UNIFIED_ATTENTION_DECODE_SMALL(unified_attention_args::data_type_enum::bf16, false, 64, 64, 8)
else DISPATCH_UNIFIED_ATTENTION_DECODE_SMALL(unified_attention_args::data_type_enum::bf16, true, 64, 64, 8)
}
}
}
else if(tier == tile_tier::medium)
{
if(args.data_type == unified_attention_args::data_type_enum::fp16)
{
if(use_bs32) {
if(!is_mask) DISPATCH_UNIFIED_ATTENTION_DECODE_MEDIUM_BS32(unified_attention_args::data_type_enum::fp16, false, 64, 128, 8)
else if(is_local) DISPATCH_UNIFIED_ATTENTION_DECODE_MEDIUM_BS32_LOCAL(unified_attention_args::data_type_enum::fp16, 64, 128, 8)
else DISPATCH_UNIFIED_ATTENTION_DECODE_MEDIUM_BS32(unified_attention_args::data_type_enum::fp16, true, 64, 128, 8)
} else {
if(!is_mask) DISPATCH_UNIFIED_ATTENTION_DECODE_MEDIUM(unified_attention_args::data_type_enum::fp16, false, 64, 128, 8)
else DISPATCH_UNIFIED_ATTENTION_DECODE_MEDIUM(unified_attention_args::data_type_enum::fp16, true, 64, 128, 8)
}
}
else if(args.data_type == unified_attention_args::data_type_enum::bf16)
{
if(use_bs32) {
if(!is_mask) DISPATCH_UNIFIED_ATTENTION_DECODE_MEDIUM_BS32(unified_attention_args::data_type_enum::bf16, false, 64, 128, 8)
else if(is_local) DISPATCH_UNIFIED_ATTENTION_DECODE_MEDIUM_BS32_LOCAL(unified_attention_args::data_type_enum::bf16, 64, 128, 8)
else DISPATCH_UNIFIED_ATTENTION_DECODE_MEDIUM_BS32(unified_attention_args::data_type_enum::bf16, true, 64, 128, 8)
} else {
if(!is_mask) DISPATCH_UNIFIED_ATTENTION_DECODE_MEDIUM(unified_attention_args::data_type_enum::bf16, false, 64, 128, 8)
else DISPATCH_UNIFIED_ATTENTION_DECODE_MEDIUM(unified_attention_args::data_type_enum::bf16, true, 64, 128, 8)
}
}
}
else
{
// Large prefill: 8 warps, kBlockM=256 (kBlockQ=32)
// No bs32 variant -- NumIssues < 1 for 8-warp tier with block_size=32.
if(args.data_type == unified_attention_args::data_type_enum::fp16)
{
if(!is_mask) DISPATCH_UNIFIED_ATTENTION(unified_attention_args::data_type_enum::fp16, false, 64, 256, 8)
else if(is_local) DISPATCH_UNIFIED_ATTENTION_LOCAL(unified_attention_args::data_type_enum::fp16, 64, 256, 8, 64)
else DISPATCH_UNIFIED_ATTENTION(unified_attention_args::data_type_enum::fp16, true, 64, 256, 8)
}
else if(args.data_type == unified_attention_args::data_type_enum::bf16)
{
if(!is_mask) DISPATCH_UNIFIED_ATTENTION(unified_attention_args::data_type_enum::bf16, false, 64, 256, 8)
else if(is_local) DISPATCH_UNIFIED_ATTENTION_LOCAL(unified_attention_args::data_type_enum::bf16, 64, 256, 8, 64)
else DISPATCH_UNIFIED_ATTENTION(unified_attention_args::data_type_enum::bf16, true, 64, 256, 8)
}
}
}
std::cerr << "unified_attention: no matching kernel instance for hdim=" << args.hdim
<< " num_queries_per_kv=" << args.num_queries_per_kv
<< " data_type=" << args.data_type << " mask_type=" << args.mask_type << std::endl;
return std::make_pair(false, -1.f);
}
#undef DISPATCH_UNIFIED_ATTENTION_DECODE_BS32_NARROW_LOCAL
#undef DISPATCH_UNIFIED_ATTENTION_DECODE_MEDIUM_BS32_LOCAL
#undef DISPATCH_UNIFIED_ATTENTION_DECODE_BS32_NARROW
#undef DISPATCH_UNIFIED_ATTENTION_DECODE_SMALL_BS32
#undef DISPATCH_UNIFIED_ATTENTION_DECODE_MEDIUM_BS32
#undef DISPATCH_UNIFIED_ATTENTION_DECODE_TINY
#undef DISPATCH_UNIFIED_ATTENTION_DECODE_SMALL
#undef DISPATCH_UNIFIED_ATTENTION_DECODE_MEDIUM
#undef DISPATCH_UNIFIED_ATTENTION_LOCAL
#undef DISPATCH_UNIFIED_ATTENTION
} // namespace ck_tile
Reference in New Issue View Git Blame Copy Permalink