Commit e7c1b7f3 authored by zhuwenwen's avatar zhuwenwen
Browse files

Merge branch 'v0.5.4-dtk24.04.1'

parents 7462218e 04c62b93
...@@ -273,8 +273,6 @@ __global__ void Code2x8Dequant( ...@@ -273,8 +273,6 @@ __global__ void Code2x8Dequant(
} }
__syncthreads(); __syncthreads();
float res = 0;
int iters = (prob_k / 8 - 1) / (8 * 32) + 1; int iters = (prob_k / 8 - 1) / (8 * 32) + 1;
while (iters--) { while (iters--) {
if (pred && a_gl_rd < a_gl_end) { if (pred && a_gl_rd < a_gl_end) {
......
...@@ -95,6 +95,7 @@ __device__ uint4 dequantize_s4_to_fp16x2(uint32_t const& source) { ...@@ -95,6 +95,7 @@ __device__ uint4 dequantize_s4_to_fp16x2(uint32_t const& source) {
return result; return result;
#endif #endif
__builtin_unreachable(); // Suppress missing return statement warning
} }
} // namespace awq } // namespace awq
......
...@@ -17,14 +17,6 @@ Shang and Dang, Xingyu and Han, Song}, journal={arXiv}, year={2023} ...@@ -17,14 +17,6 @@ Shang and Dang, Xingyu and Han, Song}, journal={arXiv}, year={2023}
namespace vllm { namespace vllm {
namespace awq { namespace awq {
// Pack two half values.
static inline __device__ __host__ unsigned __pack_half2(const half x,
const half y) {
unsigned v0 = *((unsigned short*)&x);
unsigned v1 = *((unsigned short*)&y);
return (v1 << 16) | v0;
}
template <int N> template <int N>
__global__ void __launch_bounds__(64) __global__ void __launch_bounds__(64)
gemm_forward_4bit_cuda_m16nXk32(int G, int split_k_iters, gemm_forward_4bit_cuda_m16nXk32(int G, int split_k_iters,
...@@ -42,11 +34,7 @@ __global__ void __launch_bounds__(64) ...@@ -42,11 +34,7 @@ __global__ void __launch_bounds__(64)
__shared__ half A_shared[16 * (32 + 8)]; __shared__ half A_shared[16 * (32 + 8)];
__shared__ half B_shared[32 * (N + 8)]; __shared__ half B_shared[32 * (N + 8)];
__shared__ half scaling_factors_shared[N];
__shared__ half zeros_shared[N];
int j_factors1 = ((OC + N - 1) / N); int j_factors1 = ((OC + N - 1) / N);
int blockIdx_x = 0;
int blockIdx_y = blockIdx.x % ((M + 16 - 1) / 16 * j_factors1); int blockIdx_y = blockIdx.x % ((M + 16 - 1) / 16 * j_factors1);
int blockIdx_z = blockIdx.x / ((M + 16 - 1) / 16 * j_factors1); int blockIdx_z = blockIdx.x / ((M + 16 - 1) / 16 * j_factors1);
...@@ -60,7 +48,6 @@ __global__ void __launch_bounds__(64) ...@@ -60,7 +48,6 @@ __global__ void __launch_bounds__(64)
static constexpr int row_stride_warp = 32 * 8 / 32; static constexpr int row_stride_warp = 32 * 8 / 32;
static constexpr int row_stride = 2 * 32 * 8 / N; static constexpr int row_stride = 2 * 32 * 8 / N;
bool ld_zero_flag = (threadIdx.y * 32 + threadIdx.x) * 8 < N;
// TODO: Haotian: blockIdx_y / j_factors1 in A loading to support bsz > 16 // TODO: Haotian: blockIdx_y / j_factors1 in A loading to support bsz > 16
bool ld_A_flag = bool ld_A_flag =
(blockIdx_y / j_factors1 * 16 + threadIdx.y * row_stride_warp + (blockIdx_y / j_factors1 * 16 + threadIdx.y * row_stride_warp +
...@@ -145,11 +132,7 @@ __global__ void __launch_bounds__(64) ...@@ -145,11 +132,7 @@ __global__ void __launch_bounds__(64)
uint32_t B_loaded = uint32_t B_loaded =
*(uint32_t*)(B_ptr_local + ax0_ax1_fused_0 * row_stride * (OC / 8)); *(uint32_t*)(B_ptr_local + ax0_ax1_fused_0 * row_stride * (OC / 8));
uint4 B_loaded_fp16 = dequantize_s4_to_fp16x2(B_loaded); uint4 B_loaded_fp16 = dequantize_s4_to_fp16x2(B_loaded);
// uint4 B_loaded_zero = *(uint4*)(zeros_shared + (threadIdx.x % (cta_N /
// 8)) * 8);
// uint4 B_loaded_scale = *(uint4*)(scaling_factors_shared + (threadIdx.x
// % (cta_N / 8)) * 8);
// - zero and * scale // - zero and * scale
// TODO (Haotian): can save 4 assembly instructions if sormulate as deq = // TODO (Haotian): can save 4 assembly instructions if sormulate as deq =
// q * scale - zero * scale. // q * scale - zero * scale.
...@@ -367,17 +350,11 @@ __global__ void __launch_bounds__(64) ...@@ -367,17 +350,11 @@ __global__ void __launch_bounds__(64)
__global__ void __launch_bounds__(64) __global__ void __launch_bounds__(64)
dequantize_weights(int* __restrict__ B, half* __restrict__ scaling_factors, dequantize_weights(int* __restrict__ B, half* __restrict__ scaling_factors,
int* __restrict__ zeros, half* __restrict__ C, int G) { int* __restrict__ zeros, half* __restrict__ C, int G) {
int j_factors1 = 4;
int row_stride2 = 4;
int split_k_iters = 1;
static constexpr uint32_t ZERO = 0x0; static constexpr uint32_t ZERO = 0x0;
half B_shared[32 * (128 + 8)]; half B_shared[32 * (128 + 8)];
half* B_shared_ptr2 = B_shared; half* B_shared_ptr2 = B_shared;
half B_shared_warp[32];
int OC = 512;
int N = blockDim.x * gridDim.x; // 2 int N = blockDim.x * gridDim.x; // 2
int col = (blockIdx.x * blockDim.x + threadIdx.x); int col = (blockIdx.x * blockDim.x + threadIdx.x);
int row = blockIdx.y * blockDim.y + threadIdx.y; int row = blockIdx.y * blockDim.y + threadIdx.y;
......
...@@ -64,8 +64,6 @@ using namespace detail; ...@@ -64,8 +64,6 @@ using namespace detail;
// Row vector broadcast // Row vector broadcast
template< template<
// Row bcast reuses the mbarriers from the epilogue subtile load pipeline, so this must be at least
// ceil_div(StagesC, epi tiles per CTA tile) + 1 to ensure no data races
int Stages, int Stages,
class CtaTileShapeMNK, class CtaTileShapeMNK,
class Element, class Element,
...@@ -73,14 +71,12 @@ template< ...@@ -73,14 +71,12 @@ template<
int Alignment = 128 / sizeof_bits_v<Element> int Alignment = 128 / sizeof_bits_v<Element>
> >
struct Sm90RowOrScalarBroadcast { struct Sm90RowOrScalarBroadcast {
static_assert(Alignment * sizeof_bits_v<Element> % 128 == 0, "sub-16B alignment not supported yet"); static_assert(Stages == 0, "Row broadcast doesn't support smem usage");
static_assert( static_assert(is_static_v<decltype(take<0,2>(StrideMNL{}))>); // batch stride can be dynamic or static
(cute::is_same_v<StrideMNL, Stride<_0,_1, _0>>) || // row vector broadcast, e.g. per-col alpha/bias static_assert(take<0,2>(StrideMNL{}) == Stride<_0,_1>{});
(cute::is_same_v<StrideMNL, Stride<_0,_1,int>>)); // batched row vector broadcast
// Accumulator doesn't distribute row elements evenly amongst threads so we must buffer in smem struct SharedStorage {
struct SharedStorage { array_aligned<Element, size<1>(CtaTileShapeMNK{})> smem;
alignas(16) array_aligned<Element, size<1>(CtaTileShapeMNK{}) * Stages> smem_row;
}; };
// This struct has been modified to have a bool indicating that ptr_row is a // This struct has been modified to have a bool indicating that ptr_row is a
...@@ -100,6 +96,12 @@ struct Sm90RowOrScalarBroadcast { ...@@ -100,6 +96,12 @@ struct Sm90RowOrScalarBroadcast {
return args; return args;
} }
template <class ProblemShape>
static bool
can_implement(ProblemShape const& problem_shape, Arguments const& args) {
return true;
}
template <class ProblemShape> template <class ProblemShape>
static size_t static size_t
get_workspace_size(ProblemShape const& problem_shape, Arguments const& args) { get_workspace_size(ProblemShape const& problem_shape, Arguments const& args) {
...@@ -118,15 +120,15 @@ struct Sm90RowOrScalarBroadcast { ...@@ -118,15 +120,15 @@ struct Sm90RowOrScalarBroadcast {
CUTLASS_HOST_DEVICE CUTLASS_HOST_DEVICE
Sm90RowOrScalarBroadcast(Params const& params, SharedStorage const& shared_storage) Sm90RowOrScalarBroadcast(Params const& params, SharedStorage const& shared_storage)
: params(params), : params(params)
smem_row(const_cast<Element*>(shared_storage.smem_row.data())) { } , smem(const_cast<Element*>(shared_storage.smem.data())) { }
Params params; Params params;
Element* smem_row; Element *smem = nullptr;
CUTLASS_DEVICE bool CUTLASS_DEVICE bool
is_producer_load_needed() const { is_producer_load_needed() const {
return true; return false;
} }
CUTLASS_DEVICE bool CUTLASS_DEVICE bool
...@@ -139,78 +141,76 @@ struct Sm90RowOrScalarBroadcast { ...@@ -139,78 +141,76 @@ struct Sm90RowOrScalarBroadcast {
return (!params.row_broadcast && *(params.ptr_row) == Element(0)); return (!params.row_broadcast && *(params.ptr_row) == Element(0));
} }
template <int EpiTiles, class GTensor, class STensor>
struct ProducerLoadCallbacks : EmptyProducerLoadCallbacks {
CUTLASS_DEVICE
ProducerLoadCallbacks(GTensor&& gRow, STensor&& sRow, Params const& params)
: gRow(cute::forward<GTensor>(gRow)),
sRow(cute::forward<STensor>(sRow)),
params(params) {}
GTensor gRow; // (CTA_M,CTA_N)
STensor sRow; // (CTA_M,CTA_N,PIPE)
Params const& params;
CUTLASS_DEVICE void
begin(uint64_t* full_mbarrier_ptr, int load_iteration, bool issue_tma_load) {
if (params.ptr_row == nullptr) {
return;
}
if (issue_tma_load) {
// Increment the expect-tx count of the first subtile's mbarrier by the row vector's byte-size
constexpr uint32_t copy_bytes = size<1>(CtaTileShapeMNK{}) * sizeof_bits_v<Element> / 8;
cutlass::arch::ClusterTransactionBarrier::expect_transaction(full_mbarrier_ptr, copy_bytes);
// Issue the TMA bulk copy
auto bulk_copy = Copy_Atom<SM90_BULK_COPY_AUTO, Element>{}.with(*full_mbarrier_ptr);
// Filter so we don't issue redundant copies over stride-0 modes
int bcast_pipe_index = (load_iteration / EpiTiles) % Stages;
copy(bulk_copy, filter(gRow), filter(sRow(_,_,bcast_pipe_index)));
}
}
};
template <class... Args> template <class... Args>
CUTLASS_DEVICE auto CUTLASS_DEVICE auto
get_producer_load_callbacks(ProducerLoadArgs<Args...> const& args) { get_producer_load_callbacks(ProducerLoadArgs<Args...> const& args) {
return EmptyProducerLoadCallbacks{};
auto [M, N, K, L] = args.problem_shape_mnkl;
auto [m, n, k, l] = args.tile_coord_mnkl;
Tensor mRow = make_tensor(make_gmem_ptr(params.ptr_row), make_shape(M,N,L), params.dRow);
Tensor gRow = local_tile(mRow, take<0,2>(args.tile_shape_mnk), make_coord(m,n,l)); // (CTA_M,CTA_N)
Tensor sRow = make_tensor(make_smem_ptr(smem_row), // (CTA_M,CTA_N,PIPE)
make_shape(size<0>(CtaTileShapeMNK{}), size<1>(CtaTileShapeMNK{}), Stages),
make_stride(_0{},_1{},size<1>(CtaTileShapeMNK{})));
constexpr int EpiTiles = decltype(size<1>(zipped_divide(make_layout(take<0,2>(args.tile_shape_mnk)), args.epi_tile)))::value;
return ProducerLoadCallbacks<EpiTiles, decltype(gRow), decltype(sRow)>(
cute::move(gRow), cute::move(sRow), params);
} }
template <int EpiTiles, class RTensor, class STensor> template <class GS_GTensor, class GS_STensor, class GS_CTensor, class Tiled_G2S, class SR_STensor, class SR_RTensor, class CTensor, class ThrResidue, class ThrNum>
struct ConsumerStoreCallbacks : EmptyConsumerStoreCallbacks { struct ConsumerStoreCallbacks : EmptyConsumerStoreCallbacks {
CUTLASS_DEVICE CUTLASS_DEVICE
ConsumerStoreCallbacks(RTensor&& tCrRow, STensor&& tCsRow, Params const& params) ConsumerStoreCallbacks(
: tCrRow(cute::forward<RTensor>(tCrRow)), GS_GTensor tGS_gRow_, GS_STensor tGS_sRow_,
tCsRow(cute::forward<STensor>(tCsRow)), GS_CTensor tGS_cRow_, Tiled_G2S tiled_g2s_,
params(params) {} SR_STensor tSR_sRow_, SR_RTensor tSR_rRow_,
CTensor tCcRow_, ThrResidue residue_tCcRow_, ThrNum thr_num_, Params const& params_)
RTensor tCrRow; // (CPY,CPY_M,CPY_N) : tGS_gRow(tGS_gRow_)
STensor tCsRow; // (CPY,CPY_M,CPY_N,EPI_M,EPI_N,PIPE) , tGS_sRow(tGS_sRow_)
, tGS_cRow(tGS_cRow_)
, tiled_G2S(tiled_g2s_)
, tSR_sRow(tSR_sRow_)
, tSR_rRow(tSR_rRow_)
, tCcRow(tCcRow_)
, residue_tCcRow(residue_tCcRow_)
, params(params_) {}
GS_GTensor tGS_gRow; // (CPY,CPY_M,CPY_N)
GS_STensor tGS_sRow; // (CPY,CPY_M,CPY_N)
GS_CTensor tGS_cRow; // (CPY,CPY_M,CPY_N)
Tiled_G2S tiled_G2S;
SR_STensor tSR_sRow; // (CPY,CPY_M,CPY_N,EPI_M,EPI_N)
SR_RTensor tSR_rRow; // (CPY,CPY_M,CPY_N,EPI_M,EPI_N)
CTensor tCcRow; // (CPY,CPY_M,CPY_N,EPI_M,EPI_N)
ThrResidue residue_tCcRow; // (m, n)
ThrNum thr_num;
Params const& params; Params const& params;
CUTLASS_DEVICE void CUTLASS_DEVICE void
previsit(int epi_m, int epi_n, int load_iteration, bool is_producer_load_needed) { begin() {
if (!params.row_broadcast) { if (!params.row_broadcast) {
fill(tCrRow, *(params.ptr_row)); fill(tSR_rRow, *(params.ptr_row));
return; return;
} }
auto synchronize = [&] () { cutlass::arch::NamedBarrier::sync(thr_num, cutlass::arch::ReservedNamedBarriers::EpilogueBarrier); };
Tensor tGS_gRow_flt = filter_zeros(tGS_gRow);
Tensor tGS_sRow_flt = filter_zeros(tGS_sRow);
Tensor tGS_cRow_flt = make_tensor(tGS_cRow.data(), make_layout(tGS_gRow_flt.shape(), tGS_cRow.stride()));
for (int i = 0; i < size(tGS_gRow_flt); ++i) {
if (get<1>(tGS_cRow_flt(i)) >= size<1>(CtaTileShapeMNK{})) {
continue; // OOB of SMEM,
}
if (elem_less(tGS_cRow_flt(i), make_coord(get<0>(residue_tCcRow), get<1>(residue_tCcRow)))) {
tGS_sRow_flt(i) = tGS_gRow_flt(i);
}
else {
tGS_sRow_flt(i) = Element(0); // Set to Zero when OOB so LDS could be issue without any preds.
}
}
synchronize();
}
CUTLASS_DEVICE void
begin_loop(int epi_m, int epi_n) {
if (epi_m == 0) { // Assumes M-major subtile loop if (epi_m == 0) { // Assumes M-major subtile loop
// Filter so we don't issue redundant copies over stride-0 modes if (!params.row_broadcast) return; // Do not issue LDS when row is scalar
// (only works if 0-strides are in same location, which is by construction) Tensor tSR_sRow_flt = filter_zeros(tSR_sRow(_,_,_,epi_m,epi_n));
int bcast_pipe_index = (load_iteration / EpiTiles) % Stages; Tensor tSR_rRow_flt = filter_zeros(tSR_rRow);
copy_aligned(filter(tCsRow(_,_,_,epi_m,epi_n,bcast_pipe_index)), filter(tCrRow)); copy(tSR_sRow_flt, tSR_rRow_flt);
} }
} }
...@@ -221,7 +221,7 @@ struct Sm90RowOrScalarBroadcast { ...@@ -221,7 +221,7 @@ struct Sm90RowOrScalarBroadcast {
CUTLASS_PRAGMA_UNROLL CUTLASS_PRAGMA_UNROLL
for (int i = 0; i < FragmentSize; ++i) { for (int i = 0; i < FragmentSize; ++i) {
frg_row[i] = tCrRow(epi_v * FragmentSize + i); frg_row[i] = tSR_rRow(epi_v * FragmentSize + i);
} }
return frg_row; return frg_row;
...@@ -234,17 +234,41 @@ struct Sm90RowOrScalarBroadcast { ...@@ -234,17 +234,41 @@ struct Sm90RowOrScalarBroadcast {
> >
CUTLASS_DEVICE auto CUTLASS_DEVICE auto
get_consumer_store_callbacks(ConsumerStoreArgs<Args...> const& args) { get_consumer_store_callbacks(ConsumerStoreArgs<Args...> const& args) {
auto [M, N, K, L] = args.problem_shape_mnkl;
auto [m, n, k, l] = args.tile_coord_mnkl;
using ThreadCount = decltype(size(args.tiled_copy));
Tensor sRow = make_tensor(make_smem_ptr(smem_row), // (CTA_M,CTA_N,PIPE) Tensor mRow = make_tensor(make_gmem_ptr(params.ptr_row), make_shape(M,N,L), params.dRow);
make_shape(size<0>(CtaTileShapeMNK{}), size<1>(CtaTileShapeMNK{}), Stages), Tensor gRow = local_tile(mRow(_,_,l), take<0,2>(args.tile_shape_mnk), make_coord(m, n)); // (CTA_M, CTA_N)
make_stride(_0{},_1{},size<1>(CtaTileShapeMNK{}))); Tensor sRow = make_tensor(make_smem_ptr(smem),
Tensor tCsRow = sm90_partition_for_epilogue<ReferenceSrc>( // (CPY,CPY_M,CPY_N,EPI_M,EPI_N,PIPE) make_shape(size<0>(CtaTileShapeMNK{}), size<1>(CtaTileShapeMNK{})), make_shape(_0{}, _1{})); // (CTA_M, CTA_N)
sRow, args.epi_tile, args.tiled_copy, args.thread_idx); //// G2S: Gmem to Smem
Tensor tCrRow = make_tensor_like(take<0,3>(tCsRow)); // (CPY,CPY_M,CPY_N) auto tiled_g2s = make_tiled_copy(Copy_Atom<DefaultCopy, Element>{},
Layout< Shape<_1, ThreadCount>,
constexpr int EpiTiles = decltype(size<1>(zipped_divide(make_layout(take<0,2>(args.tile_shape_mnk)), args.epi_tile)))::value; Stride<_0, _1>>{},
return ConsumerStoreCallbacks<EpiTiles, decltype(tCrRow), decltype(tCsRow)>( Layout<_1>{});
cute::move(tCrRow), cute::move(tCsRow), params); auto thr_g2s = tiled_g2s.get_slice(args.thread_idx);
Tensor tGS_gRow = thr_g2s.partition_S(gRow);
Tensor tGS_sRow = thr_g2s.partition_D(sRow);
//// G2S: Coord
auto cRow = make_identity_tensor(make_shape(size<0>(CtaTileShapeMNK{}), size<1>(CtaTileShapeMNK{})));
Tensor tGS_cRow = thr_g2s.partition_S(cRow);
//// S2R: Smem to Reg
Tensor tSR_sRow = sm90_partition_for_epilogue<ReferenceSrc>(sRow, args.epi_tile, args.tiled_copy, args.thread_idx);
Tensor tSR_rRow = make_tensor_like(take<0,3>(tSR_sRow)); // (CPY,CPY_M,CPY_N)
return ConsumerStoreCallbacks<decltype(tGS_gRow), decltype(tGS_sRow), decltype(tGS_cRow), decltype(tiled_g2s), decltype(tSR_sRow), decltype(tSR_rRow), decltype(args.tCcD), decltype(args.residue_cD), ThreadCount>(
tGS_gRow,
tGS_sRow,
tGS_cRow, tiled_g2s,
tSR_sRow,
tSR_rRow,
args.tCcD,
args.residue_cD,
ThreadCount{},
params);
} }
}; };
...@@ -285,6 +309,12 @@ struct Sm90ColOrScalarBroadcast { ...@@ -285,6 +309,12 @@ struct Sm90ColOrScalarBroadcast {
return args; return args;
} }
template <class ProblemShape>
static bool
can_implement(ProblemShape const& problem_shape, Arguments const& args) {
return true;
}
template <class ProblemShape> template <class ProblemShape>
static size_t static size_t
get_workspace_size(ProblemShape const& problem_shape, Arguments const& args) { get_workspace_size(ProblemShape const& problem_shape, Arguments const& args) {
...@@ -328,20 +358,36 @@ struct Sm90ColOrScalarBroadcast { ...@@ -328,20 +358,36 @@ struct Sm90ColOrScalarBroadcast {
return EmptyProducerLoadCallbacks{}; return EmptyProducerLoadCallbacks{};
} }
template<class GTensor, class RTensor> template<class GTensor, class RTensor, class CTensor, class ProblemShape>
struct ConsumerStoreCallbacks : EmptyConsumerStoreCallbacks { struct ConsumerStoreCallbacks : EmptyConsumerStoreCallbacks {
CUTLASS_DEVICE CUTLASS_DEVICE
ConsumerStoreCallbacks(GTensor&& tCgCol, RTensor&& tCrCol, Params const& params) ConsumerStoreCallbacks(
: tCgCol(cute::forward<GTensor>(tCgCol)), GTensor&& tCgCol,
tCrCol(cute::forward<RTensor>(tCrCol)), RTensor&& tCrCol,
params(params) {} CTensor&& tCcCol,
ProblemShape problem_shape,
Params const& params
):
tCgCol(cute::forward<GTensor>(tCgCol)),
tCrCol(cute::forward<RTensor>(tCrCol)),
tCcCol(cute::forward<CTensor>(tCcCol)),
m(get<0>(problem_shape)),
params(params) {}
GTensor tCgCol; // (CPY,CPY_M,CPY_N,EPI_M,EPI_N) GTensor tCgCol; // (CPY,CPY_M,CPY_N,EPI_M,EPI_N)
RTensor tCrCol; // (CPY,CPY_M,CPY_N,EPI_M,EPI_N) RTensor tCrCol;
CTensor tCcCol; // (CPY,CPY_M,CPY_N,EPI_M,EPI_N)
Params const& params; Params const& params;
int m;
CUTLASS_DEVICE void CUTLASS_DEVICE void
begin() { begin() {
Tensor pred = make_tensor<bool>(shape(tCgCol));
CUTLASS_PRAGMA_UNROLL
for (int i = 0; i < size(pred); ++i) {
pred(i) = get<0>(tCcCol(i)) < m;
}
if (!params.col_broadcast) { if (!params.col_broadcast) {
fill(tCrCol, *(params.ptr_col)); fill(tCrCol, *(params.ptr_col));
return; return;
...@@ -349,7 +395,7 @@ struct Sm90ColOrScalarBroadcast { ...@@ -349,7 +395,7 @@ struct Sm90ColOrScalarBroadcast {
// Filter so we don't issue redundant copies over stride-0 modes // Filter so we don't issue redundant copies over stride-0 modes
// (only works if 0-strides are in same location, which is by construction) // (only works if 0-strides are in same location, which is by construction)
copy_aligned(filter(tCgCol), filter(tCrCol)); copy_if(pred, filter(tCgCol), filter(tCrCol));
} }
template <typename ElementAccumulator, int FragmentSize> template <typename ElementAccumulator, int FragmentSize>
...@@ -381,8 +427,20 @@ struct Sm90ColOrScalarBroadcast { ...@@ -381,8 +427,20 @@ struct Sm90ColOrScalarBroadcast {
mCol, args.tile_shape_mnk, args.tile_coord_mnkl, args.epi_tile, args.tiled_copy, args.thread_idx); mCol, args.tile_shape_mnk, args.tile_coord_mnkl, args.epi_tile, args.tiled_copy, args.thread_idx);
Tensor tCrCol = make_tensor_like(tCgCol); // (CPY,CPY_M,CPY_N,EPI_M,EPI_N) Tensor tCrCol = make_tensor_like(tCgCol); // (CPY,CPY_M,CPY_N,EPI_M,EPI_N)
return ConsumerStoreCallbacks<decltype(tCgCol), decltype(tCrCol)>( // Generate an identity tensor matching the shape of the global tensor and
cute::move(tCgCol), cute::move(tCrCol), params); // partition the same way, this will be used to generate the predicate
// tensor for loading
Tensor cCol = make_identity_tensor(mCol.shape());
Tensor tCcCol = sm90_partition_for_epilogue<ReferenceSrc>( // (CPY,CPY_M,CPY_N,EPI_M,EPI_N)
cCol, args.tile_shape_mnk, args.tile_coord_mnkl, args.epi_tile, args.tiled_copy, args.thread_idx);
return ConsumerStoreCallbacks(
cute::move(tCgCol),
cute::move(tCrCol),
cute::move(tCcCol),
args.problem_shape_mnkl,
params
);
} }
}; };
......
#pragma once #pragma once
#include "cutlass/cutlass.h" #include "cutlass/cutlass.h"
#include <climits>
/** /**
* Helper function for checking CUTLASS errors * Helper function for checking CUTLASS errors
...@@ -10,3 +11,17 @@ ...@@ -10,3 +11,17 @@
TORCH_CHECK(status == cutlass::Status::kSuccess, \ TORCH_CHECK(status == cutlass::Status::kSuccess, \
cutlassGetStatusString(status)) \ cutlassGetStatusString(status)) \
} }
inline uint32_t next_pow_2(uint32_t const num) {
if (num <= 1) return num;
return 1 << (CHAR_BIT * sizeof(num) - __builtin_clz(num - 1));
}
inline int get_cuda_max_shared_memory_per_block_opt_in(int const device) {
int max_shared_mem_per_block_opt_in = 0;
cudaDeviceGetAttribute(&max_shared_mem_per_block_opt_in,
cudaDevAttrMaxSharedMemoryPerBlockOptin,
device);
return max_shared_mem_per_block_opt_in;
}
#include <stddef.h> #include <stddef.h>
#include <torch/all.h> #include <torch/all.h>
#include <ATen/cuda/CUDAContext.h>
// clang-format will break include orders
// clang-format off
#include "cute/tensor.hpp"
#include "cute/atom/mma_atom.hpp"
#include "cutlass/numeric_types.h"
#include "cutlass/util/device_memory.h"
#include "cutlass/cutlass.h" #include "cutlass/cutlass.h"
#include "cutlass/gemm_coord.h"
#include "cutlass/arch/mma_sm75.h"
#include "cutlass/arch/arch.h"
#include "cutlass/arch/mma.h"
#include "cutlass/gemm/device/gemm.h"
#include "cutlass/gemm/device/gemm_universal_adapter.h"
#include "cutlass/epilogue/threadblock/fusion/visitors.hpp"
#include "cutlass/gemm/kernel/default_gemm_universal_with_visitor.h"
#include "broadcast_load_epilogue_c2x.hpp"
#include "common.hpp"
// clang-format on
using namespace cute; #include "scaled_mm_c2x.cuh"
#include "scaled_mm_c2x_sm75_dispatch.cuh"
#include "scaled_mm_c2x_sm80_dispatch.cuh"
#include "scaled_mm_c2x_sm89_fp8_dispatch.cuh"
#include "scaled_mm_c2x_sm89_int8_dispatch.cuh"
/* /*
This file defines quantized GEMM operations using the CUTLASS 2.x API, for This file defines quantized GEMM operations using the CUTLASS 2.x API, for
NVIDIA GPUs with SM versions prior to sm90 (Hopper). NVIDIA GPUs with SM versions prior to sm90 (Hopper).
Epilogue functions can be defined to post-process the output before it is
written to GPU memory.
Epilogues must contain a public type named EVTCompute of type Sm80EVT,
as well as a static prepare_args function that constructs an
EVTCompute::Arguments struct.
*/
namespace {
// Wrappers for the GEMM kernel that is used to guard against compilation on
// architectures that will never use the kernel. The purpose of this is to
// reduce the size of the compiled binary.
// __CUDA_ARCH__ is not defined in host code, so this lets us smuggle the ifdef
// into code that will be executed on the device where it is defined.
template <typename Kernel>
struct enable_sm75_to_sm80 : Kernel {
template <typename... Args>
CUTLASS_DEVICE static void invoke(Args&&... args) {
#if defined __CUDA_ARCH__ && __CUDA_ARCH__ >= 750 && __CUDA_ARCH__ < 800
Kernel::invoke(std::forward<Args>(args)...);
#endif
}
};
template <typename Kernel>
struct enable_sm80_to_sm89 : Kernel {
template <typename... Args>
CUTLASS_DEVICE static void invoke(Args&&... args) {
#if defined __CUDA_ARCH__ && __CUDA_ARCH__ >= 800 && __CUDA_ARCH__ < 890
Kernel::invoke(std::forward<Args>(args)...);
#endif
}
};
template <typename Kernel>
struct enable_sm89_to_sm90 : Kernel {
template <typename... Args>
CUTLASS_DEVICE static void invoke(Args&&... args) {
#if defined __CUDA_ARCH__ && __CUDA_ARCH__ >= 890 && __CUDA_ARCH__ < 900
Kernel::invoke(std::forward<Args>(args)...);
#endif
}
};
/*
This epilogue function defines a quantized GEMM operation similar to
torch._scaled_mm.
A and B may be both either int8 or fp8_e4m3. A can be quantized per-tensor or
per-row. B can be quantized per-tensor or per-column.
Any combination of per-tensor and per-row or column is supported.
A and B must have symmetric quantization (zero point == 0).
So the GEMM operation is D = (a_scales * A) (b_scales * B), where the
scales are applied elementwise with numpy-style broadcasting.
ScaleA and ScaleB define the epilogue functions that apply the scales for
the A and B operands respectively. These scales may be either per-tensor or
per row or column.
*/ */
template <typename ElementD, typename OutputTileThreadMap>
struct ScaledEpilogue {
private:
using Accum = cutlass::epilogue::threadblock::VisitorAccFetch;
using ScaleA = cutlass::epilogue::threadblock::VisitorColOrScalarBroadcast<
OutputTileThreadMap, float, Stride<Int<1>, Int<0>, Int<0>>>;
using ScaleB = cutlass::epilogue::threadblock::VisitorRowOrScalarBroadcast<
OutputTileThreadMap, float, Stride<Int<0>, Int<1>, Int<0>>>;
using Compute0 = cutlass::epilogue::threadblock::VisitorCompute< template <template <typename, typename> typename Epilogue,
cutlass::multiplies, float, float, typename... EpilogueArgs>
cutlass::FloatRoundStyle::round_to_nearest>; void cutlass_scaled_mm_sm75_epilogue(torch::Tensor& out, torch::Tensor const& a,
torch::Tensor const& b,
using EVTCompute0 = EpilogueArgs&&... epilogue_args) {
cutlass::epilogue::threadblock::Sm80EVT<Compute0, ScaleB, Accum>; TORCH_CHECK(a.dtype() == torch::kInt8);
TORCH_CHECK(b.dtype() == torch::kInt8);
using Compute1 = cutlass::epilogue::threadblock::VisitorCompute<
cutlass::multiplies, ElementD, float,
cutlass::FloatRoundStyle::round_to_nearest>;
public:
using EVTCompute =
cutlass::epilogue::threadblock::Sm80EVT<Compute1, ScaleA, EVTCompute0>;
using ArgumentType = typename EVTCompute::Arguments;
static ArgumentType prepare_args(torch::Tensor const& a_scales,
torch::Tensor const& b_scales) {
using ScaleAArgs = typename ScaleA::Arguments;
using ScaleBArgs = typename ScaleB::Arguments;
ScaleBArgs b_args{b_scales.data_ptr<float>(), b_scales.numel() != 1, {}};
ScaleAArgs a_args{a_scales.data_ptr<float>(), a_scales.numel() != 1, {}};
typename EVTCompute0::Arguments evt0_compute_args{b_args};
typename EVTCompute::Arguments evt_compute_args{a_args, evt0_compute_args}; if (out.dtype() == torch::kBFloat16) {
return evt_compute_args; return vllm::cutlass_gemm_sm75_dispatch<int8_t, cutlass::bfloat16_t,
Epilogue>(
out, a, b, std::forward<EpilogueArgs>(epilogue_args)...);
} else {
TORCH_CHECK(out.dtype() == torch::kFloat16);
return vllm::cutlass_gemm_sm75_dispatch<int8_t, cutlass::half_t, Epilogue>(
out, a, b, std::forward<EpilogueArgs>(epilogue_args)...);
} }
};
template <typename Arch, template <typename> typename ArchGuard,
typename ElementAB_, typename ElementD_,
template <typename, typename> typename Epilogue_, typename TileShape,
typename WarpShape, typename InstructionShape, int32_t MainLoopStages>
struct cutlass_2x_gemm {
using ElementAB = ElementAB_;
using ElementD = ElementD_;
using ElementAcc =
typename std::conditional<std::is_same_v<ElementAB, int8_t>, int32_t,
float>::type;
using Operator =
typename std::conditional<std::is_same_v<ElementAB, int8_t>,
cutlass::arch::OpMultiplyAddSaturate,
cutlass::arch::OpMultiplyAdd>::type;
using OutputTileThreadMap =
cutlass::epilogue::threadblock::OutputTileThreadLayout<
TileShape, WarpShape, float, 4, 1 /* epilogue stages */
>;
using Epilogue = Epilogue_<ElementD, OutputTileThreadMap>;
using EVTCompute = typename Epilogue::EVTCompute;
using D = cutlass::epilogue::threadblock::VisitorAuxStore<
OutputTileThreadMap, ElementD, cutlass::FloatRoundStyle::round_to_nearest,
Stride<int64_t, Int<1>, Int<0>>>;
using EVTD = cutlass::epilogue::threadblock::Sm80EVT<D, EVTCompute>;
// clang-format off
using RowMajor = typename cutlass::layout::RowMajor;
using ColumnMajor = typename cutlass::layout::ColumnMajor;
using KernelType =
ArchGuard<typename cutlass::gemm::kernel::DefaultGemmWithVisitor<
ElementAB, RowMajor, cutlass::ComplexTransform::kNone, 16,
ElementAB, ColumnMajor, cutlass::ComplexTransform::kNone, 16,
float, cutlass::layout::RowMajor, 4,
ElementAcc, float, cutlass::arch::OpClassTensorOp,
Arch,
TileShape, WarpShape, InstructionShape,
EVTD,
cutlass::gemm::threadblock::ThreadblockSwizzleStreamK,
MainLoopStages, Operator,
1 /* epilogue stages */
>::GemmKernel>;
// clang-format on
using Op = cutlass::gemm::device::GemmUniversalAdapter<KernelType>;
};
template <typename Gemm, typename... EpilogueArgs>
void cutlass_gemm_caller(torch::Tensor& out, torch::Tensor const& a,
torch::Tensor const& b,
EpilogueArgs&&... epilogue_params) {
using ElementAB = typename Gemm::ElementAB;
using ElementD = typename Gemm::ElementD;
int32_t m = a.size(0);
int32_t n = b.size(1);
int32_t k = a.size(1);
cutlass::gemm::GemmCoord problem_size{m, n, k};
int64_t lda = a.stride(0);
int64_t ldb = b.stride(1);
int64_t ldc = out.stride(0);
using StrideC = Stride<int64_t, Int<1>, Int<0>>;
StrideC c_stride{ldc, Int<1>{}, Int<0>{}};
auto a_ptr = static_cast<ElementAB const*>(a.data_ptr());
auto b_ptr = static_cast<ElementAB const*>(b.data_ptr());
auto c_ptr = static_cast<ElementD*>(out.data_ptr());
typename Gemm::D::Arguments d_args{c_ptr, c_stride};
using Epilogue = typename Gemm::Epilogue;
auto evt_args =
Epilogue::prepare_args(std::forward<EpilogueArgs>(epilogue_params)...);
typename Gemm::EVTD::Arguments epilogue_args{
evt_args,
d_args,
};
typename Gemm::Op::Arguments args{
cutlass::gemm::GemmUniversalMode::kGemmSplitKParallel, // universal mode
problem_size, // problem size
1, // batch count
epilogue_args,
a_ptr,
b_ptr,
nullptr,
nullptr,
0,
0,
0,
0,
lda,
ldb,
ldc,
ldc};
// Launch the CUTLASS GEMM kernel.
typename Gemm::Op gemm_op;
size_t workspace_size = gemm_op.get_workspace_size(args);
cutlass::device_memory::allocation<uint8_t> workspace(workspace_size);
auto stream = at::cuda::getCurrentCUDAStream(a.get_device());
CUTLASS_CHECK(gemm_op.can_implement(args));
cutlass::Status status = gemm_op(args, workspace.get(), stream);
CUTLASS_CHECK(status);
} }
} // namespace
void cutlass_scaled_mm_sm75(torch::Tensor& out, torch::Tensor const& a, void cutlass_scaled_mm_sm75(torch::Tensor& out, torch::Tensor const& a,
torch::Tensor const& b, torch::Tensor const& b,
torch::Tensor const& a_scales, torch::Tensor const& a_scales,
torch::Tensor const& b_scales) { torch::Tensor const& b_scales,
TORCH_CHECK(a.dtype() == torch::kInt8); c10::optional<torch::Tensor> const& bias) {
TORCH_CHECK(b.dtype() == torch::kInt8);
TORCH_CHECK(a_scales.dtype() == torch::kFloat32); TORCH_CHECK(a_scales.dtype() == torch::kFloat32);
TORCH_CHECK(b_scales.dtype() == torch::kFloat32); TORCH_CHECK(b_scales.dtype() == torch::kFloat32);
if (bias) {
using TileShape = typename cutlass::gemm::GemmShape<128, 128, 64>; TORCH_CHECK(bias->dtype() == out.dtype(),
using WarpShape = typename cutlass::gemm::GemmShape<64, 64, 64>; "currently bias dtype must match output dtype ", out.dtype());
using InstructionShape = typename cutlass::gemm::GemmShape<8, 8, 16>; return cutlass_scaled_mm_sm75_epilogue<vllm::ScaledEpilogueBias>(
out, a, b, a_scales, b_scales, *bias);
if (out.dtype() == torch::kBFloat16) {
return cutlass_gemm_caller<cutlass_2x_gemm<
cutlass::arch::Sm75, enable_sm75_to_sm80, int8_t, cutlass::bfloat16_t,
ScaledEpilogue, TileShape, WarpShape, InstructionShape, 2>>(
out, a, b, a_scales, b_scales);
} else { } else {
TORCH_CHECK(out.dtype() == torch::kFloat16); return cutlass_scaled_mm_sm75_epilogue<vllm::ScaledEpilogue>(
return cutlass_gemm_caller<cutlass_2x_gemm<
cutlass::arch::Sm75, enable_sm75_to_sm80, int8_t, cutlass::half_t,
ScaledEpilogue, TileShape, WarpShape, InstructionShape, 2>>(
out, a, b, a_scales, b_scales); out, a, b, a_scales, b_scales);
} }
} }
void cutlass_scaled_mm_sm80(torch::Tensor& out, torch::Tensor const& a, template <template <typename, typename> typename Epilogue,
torch::Tensor const& b, typename... EpilogueArgs>
torch::Tensor const& a_scales, void cutlass_scaled_mm_sm80_epilogue(torch::Tensor& out, torch::Tensor const& a,
torch::Tensor const& b_scales) { torch::Tensor const& b,
EpilogueArgs&&... epilogue_args) {
TORCH_CHECK(a.dtype() == torch::kInt8); TORCH_CHECK(a.dtype() == torch::kInt8);
TORCH_CHECK(b.dtype() == torch::kInt8); TORCH_CHECK(b.dtype() == torch::kInt8);
TORCH_CHECK(a_scales.dtype() == torch::kFloat32);
TORCH_CHECK(b_scales.dtype() == torch::kFloat32);
using TileShape = typename cutlass::gemm::GemmShape<128, 128, 64>;
using WarpShape = typename cutlass::gemm::GemmShape<64, 64, 64>;
using InstructionShape = typename cutlass::gemm::GemmShape<16, 8, 32>;
if (out.dtype() == torch::kBFloat16) { if (out.dtype() == torch::kBFloat16) {
return cutlass_gemm_caller<cutlass_2x_gemm< return vllm::cutlass_gemm_sm80_dispatch<int8_t, cutlass::bfloat16_t,
cutlass::arch::Sm80, enable_sm80_to_sm89, int8_t, cutlass::bfloat16_t, Epilogue>(
ScaledEpilogue, TileShape, WarpShape, InstructionShape, 5>>( out, a, b, std::forward<EpilogueArgs>(epilogue_args)...);
out, a, b, a_scales, b_scales);
} else { } else {
TORCH_CHECK(out.dtype() == torch::kFloat16); TORCH_CHECK(out.dtype() == torch::kFloat16);
return cutlass_gemm_caller<cutlass_2x_gemm< return vllm::cutlass_gemm_sm80_dispatch<int8_t, cutlass::half_t, Epilogue>(
cutlass::arch::Sm80, enable_sm80_to_sm89, int8_t, cutlass::half_t, out, a, b, std::forward<EpilogueArgs>(epilogue_args)...);
ScaledEpilogue, TileShape, WarpShape, InstructionShape, 5>>(
out, a, b, a_scales, b_scales);
} }
} }
void cutlass_scaled_mm_sm89(torch::Tensor& out, torch::Tensor const& a, void cutlass_scaled_mm_sm80(torch::Tensor& out, torch::Tensor const& a,
torch::Tensor const& b, torch::Tensor const& b,
torch::Tensor const& a_scales, torch::Tensor const& a_scales,
torch::Tensor const& b_scales) { torch::Tensor const& b_scales,
using TileShape = typename cutlass::gemm::GemmShape<128, 128, 64>; c10::optional<torch::Tensor> const& bias) {
using WarpShape = typename cutlass::gemm::GemmShape<64, 64, 64>;
using InstructionShape = typename cutlass::gemm::GemmShape<16, 8, 32>;
TORCH_CHECK(a_scales.dtype() == torch::kFloat32); TORCH_CHECK(a_scales.dtype() == torch::kFloat32);
TORCH_CHECK(b_scales.dtype() == torch::kFloat32); TORCH_CHECK(b_scales.dtype() == torch::kFloat32);
if (bias) {
TORCH_CHECK(bias->dtype() == out.dtype(),
"currently bias dtype must match output dtype ", out.dtype());
return cutlass_scaled_mm_sm80_epilogue<vllm::ScaledEpilogueBias>(
out, a, b, a_scales, b_scales, *bias);
} else {
return cutlass_scaled_mm_sm80_epilogue<vllm::ScaledEpilogue>(
out, a, b, a_scales, b_scales);
}
}
template <template <typename, typename> typename Epilogue,
typename... EpilogueArgs>
void cutlass_scaled_mm_sm89_epilogue(torch::Tensor& out, torch::Tensor const& a,
torch::Tensor const& b,
EpilogueArgs&&... epilogue_args) {
if (a.dtype() == torch::kInt8) { if (a.dtype() == torch::kInt8) {
TORCH_CHECK(b.dtype() == torch::kInt8); TORCH_CHECK(b.dtype() == torch::kInt8);
if (out.dtype() == torch::kBFloat16) { if (out.dtype() == torch::kBFloat16) {
return cutlass_gemm_caller<cutlass_2x_gemm< return vllm::cutlass_gemm_sm89_int8_dispatch<int8_t, cutlass::bfloat16_t,
cutlass::arch::Sm89, enable_sm89_to_sm90, int8_t, cutlass::bfloat16_t, Epilogue>(
ScaledEpilogue, TileShape, WarpShape, InstructionShape, 5>>( out, a, b, std::forward<EpilogueArgs>(epilogue_args)...);
out, a, b, a_scales, b_scales);
} else { } else {
assert(out.dtype() == torch::kFloat16); assert(out.dtype() == torch::kFloat16);
return cutlass_gemm_caller<cutlass_2x_gemm< return vllm::cutlass_gemm_sm89_int8_dispatch<int8_t, cutlass::half_t,
cutlass::arch::Sm89, enable_sm89_to_sm90, int8_t, cutlass::half_t, Epilogue>(
ScaledEpilogue, TileShape, WarpShape, InstructionShape, 5>>( out, a, b, std::forward<EpilogueArgs>(epilogue_args)...);
out, a, b, a_scales, b_scales);
} }
} else { } else {
TORCH_CHECK(a.dtype() == torch::kFloat8_e4m3fn); TORCH_CHECK(a.dtype() == torch::kFloat8_e4m3fn);
TORCH_CHECK(b.dtype() == torch::kFloat8_e4m3fn); TORCH_CHECK(b.dtype() == torch::kFloat8_e4m3fn);
if (out.dtype() == torch::kBFloat16) { if (out.dtype() == torch::kBFloat16) {
return cutlass_gemm_caller<cutlass_2x_gemm< return vllm::cutlass_gemm_sm89_fp8_dispatch<
cutlass::arch::Sm89, enable_sm89_to_sm90, cutlass::float_e4m3_t, cutlass::float_e4m3_t, cutlass::bfloat16_t, Epilogue>(
cutlass::bfloat16_t, ScaledEpilogue, TileShape, WarpShape, out, a, b, std::forward<EpilogueArgs>(epilogue_args)...);
InstructionShape, 5>>(out, a, b, a_scales, b_scales);
} else { } else {
TORCH_CHECK(out.dtype() == torch::kFloat16); TORCH_CHECK(out.dtype() == torch::kFloat16);
return cutlass_gemm_caller<cutlass_2x_gemm< return vllm::cutlass_gemm_sm89_fp8_dispatch<cutlass::float_e4m3_t,
cutlass::arch::Sm89, enable_sm89_to_sm90, cutlass::float_e4m3_t, cutlass::half_t, Epilogue>(
cutlass::half_t, ScaledEpilogue, TileShape, WarpShape, out, a, b, std::forward<EpilogueArgs>(epilogue_args)...);
InstructionShape, 5>>(out, a, b, a_scales, b_scales);
} }
} }
} }
void cutlass_scaled_mm_sm89(torch::Tensor& out, torch::Tensor const& a,
torch::Tensor const& b,
torch::Tensor const& a_scales,
torch::Tensor const& b_scales,
c10::optional<torch::Tensor> const& bias) {
TORCH_CHECK(a_scales.dtype() == torch::kFloat32);
TORCH_CHECK(b_scales.dtype() == torch::kFloat32);
if (bias) {
TORCH_CHECK(bias->dtype() == out.dtype(),
"currently bias dtype must match output dtype ", out.dtype());
return cutlass_scaled_mm_sm89_epilogue<vllm::ScaledEpilogueBias>(
out, a, b, a_scales, b_scales, *bias);
} else {
return cutlass_scaled_mm_sm89_epilogue<vllm::ScaledEpilogue>(
out, a, b, a_scales, b_scales);
}
}
#pragma once
#include <stddef.h>
#include <torch/all.h>
#include <ATen/cuda/CUDAContext.h>
// clang-format will break include orders
// clang-format off
#include "cute/tensor.hpp"
#include "cute/atom/mma_atom.hpp"
#include "cutlass/numeric_types.h"
#include "cutlass/cutlass.h"
#include "cutlass/gemm_coord.h"
#include "cutlass/arch/mma_sm75.h"
#include "cutlass/arch/arch.h"
#include "cutlass/arch/mma.h"
#include "cutlass/gemm/device/gemm.h"
#include "cutlass/gemm/device/gemm_universal_adapter.h"
#include "cutlass/epilogue/threadblock/fusion/visitors.hpp"
#include "cutlass/gemm/kernel/default_gemm_universal_with_visitor.h"
#include "broadcast_load_epilogue_c2x.hpp"
#include "common.hpp"
// clang-format on
using namespace cute;
/*
Epilogue functions can be defined to post-process the output before it is
written to GPU memory.
Epilogues must contain a public type named EVTCompute of type Sm80EVT,
as well as a static prepare_args function that constructs an
EVTCompute::Arguments struct.
*/
namespace vllm {
// Wrappers for the GEMM kernel that is used to guard against compilation on
// architectures that will never use the kernel. The purpose of this is to
// reduce the size of the compiled binary.
// __CUDA_ARCH__ is not defined in host code, so this lets us smuggle the ifdef
// into code that will be executed on the device where it is defined.
template <typename Kernel>
struct enable_sm75_to_sm80 : Kernel {
template <typename... Args>
CUTLASS_DEVICE static void invoke(Args&&... args) {
#if defined __CUDA_ARCH__ && __CUDA_ARCH__ >= 750 && __CUDA_ARCH__ < 800
Kernel::invoke(std::forward<Args>(args)...);
#endif
}
};
template <typename Kernel>
struct enable_sm80_to_sm89 : Kernel {
template <typename... Args>
CUTLASS_DEVICE static void invoke(Args&&... args) {
#if defined __CUDA_ARCH__ && __CUDA_ARCH__ >= 800 && __CUDA_ARCH__ < 890
Kernel::invoke(std::forward<Args>(args)...);
#endif
}
};
template <typename Kernel>
struct enable_sm89_to_sm90 : Kernel {
template <typename... Args>
CUTLASS_DEVICE static void invoke(Args&&... args) {
#if defined __CUDA_ARCH__ && __CUDA_ARCH__ >= 890 && __CUDA_ARCH__ < 900
Kernel::invoke(std::forward<Args>(args)...);
#endif
}
};
/*
* This class provides the common ScaleA and ScaleB descriptors for the
* ScaledEpilogue and ScaledEpilogueBias classes.
*/
template <typename ElementD, typename OutputTileThreadMap>
struct ScaledEpilogueBase {
protected:
using Accum = cutlass::epilogue::threadblock::VisitorAccFetch;
using ScaleA = cutlass::epilogue::threadblock::VisitorColOrScalarBroadcast<
OutputTileThreadMap, float, Stride<Int<1>, Int<0>, Int<0>>>;
using ScaleB = cutlass::epilogue::threadblock::VisitorRowOrScalarBroadcast<
OutputTileThreadMap, float, Stride<Int<0>, Int<1>, Int<0>>>;
};
/*
This epilogue function defines a quantized GEMM operation similar to
torch._scaled_mm.
A and B may be both either int8 or fp8_e4m3. A can be quantized per-tensor or
per-row. B can be quantized per-tensor or per-column.
Any combination of per-tensor and per-row or column is supported.
A and B must have symmetric quantization (zero point == 0).
So the GEMM operation is D = (a_scales * A) (b_scales * B), where the
scales are applied elementwise with numpy-style broadcasting.
ScaleA and ScaleB define the epilogue functions that apply the scales for
the A and B operands respectively. These scales may be either per-tensor or
per row or column.
*/
template <typename ElementD, typename OutputTileThreadMap>
struct ScaledEpilogue
: private ScaledEpilogueBase<ElementD, OutputTileThreadMap> {
private:
using SUPER = ScaledEpilogueBase<ElementD, OutputTileThreadMap>;
using Accum = typename SUPER::Accum;
using ScaleA = typename SUPER::ScaleA;
using ScaleB = typename SUPER::ScaleB;
using Compute0 = cutlass::epilogue::threadblock::VisitorCompute<
cutlass::multiplies, float, float,
cutlass::FloatRoundStyle::round_to_nearest>;
using EVTCompute0 =
cutlass::epilogue::threadblock::Sm80EVT<Compute0, ScaleB, Accum>;
using Compute1 = cutlass::epilogue::threadblock::VisitorCompute<
cutlass::multiplies, ElementD, float,
cutlass::FloatRoundStyle::round_to_nearest>;
public:
using EVTCompute =
cutlass::epilogue::threadblock::Sm80EVT<Compute1, ScaleA, EVTCompute0>;
using ArgumentType = typename EVTCompute::Arguments;
static ArgumentType prepare_args(torch::Tensor const& a_scales,
torch::Tensor const& b_scales) {
using ScaleAArgs = typename ScaleA::Arguments;
using ScaleBArgs = typename ScaleB::Arguments;
ScaleBArgs b_args{b_scales.data_ptr<float>(), b_scales.numel() != 1, {}};
ScaleAArgs a_args{a_scales.data_ptr<float>(), a_scales.numel() != 1, {}};
typename EVTCompute0::Arguments evt0_compute_args{b_args};
typename EVTCompute::Arguments evt_compute_args{a_args, evt0_compute_args};
return evt_compute_args;
}
};
template <typename ElementD, typename OutputTileThreadMap>
struct ScaledEpilogueBias
: private ScaledEpilogueBase<ElementD, OutputTileThreadMap> {
private:
using SUPER = ScaledEpilogueBase<ElementD, OutputTileThreadMap>;
using Accum = typename SUPER::Accum;
using ScaleA = typename SUPER::ScaleA;
using ScaleB = typename SUPER::ScaleB;
using Compute0 = cutlass::epilogue::threadblock::VisitorCompute<
cutlass::multiplies, float, float,
cutlass::FloatRoundStyle::round_to_nearest>;
using EVTCompute0 =
cutlass::epilogue::threadblock::Sm80EVT<Compute0, ScaleB, Accum>;
using Compute1 = cutlass::epilogue::threadblock::VisitorCompute<
cutlass::multiply_add, ElementD, float,
cutlass::FloatRoundStyle::round_to_nearest>;
using Bias = cutlass::epilogue::threadblock::VisitorRowBroadcast<
OutputTileThreadMap, ElementD, Stride<Int<0>, Int<1>, Int<0>>>;
public:
using EVTCompute = cutlass::epilogue::threadblock::Sm80EVT<Compute1, ScaleA,
EVTCompute0, Bias>;
using ArgumentType = typename EVTCompute::Arguments;
static ArgumentType prepare_args(torch::Tensor const& a_scales,
torch::Tensor const& b_scales,
torch::Tensor const& bias) {
using ScaleAArgs = typename ScaleA::Arguments;
using ScaleBArgs = typename ScaleB::Arguments;
using BiasArgs = typename Bias::Arguments;
ScaleBArgs b_args{b_scales.data_ptr<float>(), b_scales.numel() != 1, {}};
ScaleAArgs a_args{a_scales.data_ptr<float>(), a_scales.numel() != 1, {}};
BiasArgs bias_args{static_cast<ElementD*>(bias.data_ptr()), {}};
typename EVTCompute0::Arguments evt0_compute_args{b_args};
typename EVTCompute::Arguments evt_compute_args{a_args, evt0_compute_args,
bias_args};
return evt_compute_args;
}
};
template <typename Arch, template <typename> typename ArchGuard,
typename ElementAB_, typename ElementD_,
template <typename, typename> typename Epilogue_, typename TileShape,
typename WarpShape, typename InstructionShape, int32_t MainLoopStages,
typename FP8MathOperator = cutlass::arch::OpMultiplyAdd>
struct cutlass_2x_gemm {
using ElementAB = ElementAB_;
using ElementD = ElementD_;
using ElementAcc =
typename std::conditional<std::is_same_v<ElementAB, int8_t>, int32_t,
float>::type;
using Operator =
typename std::conditional<std::is_same_v<ElementAB, int8_t>,
cutlass::arch::OpMultiplyAddSaturate,
FP8MathOperator>::type;
using OutputTileThreadMap =
cutlass::epilogue::threadblock::OutputTileThreadLayout<
TileShape, WarpShape, float, 4, 1 /* epilogue stages */
>;
using Epilogue = Epilogue_<ElementD, OutputTileThreadMap>;
using EVTCompute = typename Epilogue::EVTCompute;
using D = cutlass::epilogue::threadblock::VisitorAuxStore<
OutputTileThreadMap, ElementD, cutlass::FloatRoundStyle::round_to_nearest,
Stride<int64_t, Int<1>, Int<0>>>;
using EVTD = cutlass::epilogue::threadblock::Sm80EVT<D, EVTCompute>;
// clang-format off
using RowMajor = typename cutlass::layout::RowMajor;
using ColumnMajor = typename cutlass::layout::ColumnMajor;
using KernelType =
ArchGuard<typename cutlass::gemm::kernel::DefaultGemmWithVisitor<
ElementAB, RowMajor, cutlass::ComplexTransform::kNone, 16,
ElementAB, ColumnMajor, cutlass::ComplexTransform::kNone, 16,
float, cutlass::layout::RowMajor, 4,
ElementAcc, float, cutlass::arch::OpClassTensorOp,
Arch,
TileShape, WarpShape, InstructionShape,
EVTD,
cutlass::gemm::threadblock::ThreadblockSwizzleStreamK,
MainLoopStages, Operator,
1 /* epilogue stages */
>::GemmKernel>;
// clang-format on
using Op = cutlass::gemm::device::GemmUniversalAdapter<KernelType>;
};
template <typename Gemm, typename... EpilogueArgs>
inline void cutlass_gemm_caller(torch::Tensor& out, torch::Tensor const& a,
torch::Tensor const& b,
EpilogueArgs&&... epilogue_params) {
using ElementAB = typename Gemm::ElementAB;
using ElementD = typename Gemm::ElementD;
int32_t m = a.size(0);
int32_t n = b.size(1);
int32_t k = a.size(1);
cutlass::gemm::GemmCoord problem_size{m, n, k};
int64_t lda = a.stride(0);
int64_t ldb = b.stride(1);
int64_t ldc = out.stride(0);
using StrideC = Stride<int64_t, Int<1>, Int<0>>;
StrideC c_stride{ldc, Int<1>{}, Int<0>{}};
auto a_ptr = static_cast<ElementAB const*>(a.data_ptr());
auto b_ptr = static_cast<ElementAB const*>(b.data_ptr());
auto c_ptr = static_cast<ElementD*>(out.data_ptr());
typename Gemm::D::Arguments d_args{c_ptr, c_stride};
using Epilogue = typename Gemm::Epilogue;
auto evt_args =
Epilogue::prepare_args(std::forward<EpilogueArgs>(epilogue_params)...);
typename Gemm::EVTD::Arguments epilogue_args{
evt_args,
d_args,
};
typename Gemm::Op::Arguments args{
cutlass::gemm::GemmUniversalMode::kGemmSplitKParallel, // universal mode
problem_size, // problem size
1, // batch count
epilogue_args,
a_ptr,
b_ptr,
nullptr,
nullptr,
0,
0,
0,
0,
lda,
ldb,
ldc,
ldc};
// Launch the CUTLASS GEMM kernel.
typename Gemm::Op gemm_op;
size_t workspace_size = gemm_op.get_workspace_size(args);
auto const workspace_options =
torch::TensorOptions().dtype(torch::kUInt8).device(a.device());
auto workspace = torch::empty(workspace_size, workspace_options);
auto stream = at::cuda::getCurrentCUDAStream(a.get_device());
CUTLASS_CHECK(gemm_op.can_implement(args));
cutlass::Status status = gemm_op(args, workspace.data_ptr(), stream);
CUTLASS_CHECK(status);
}
template <typename Gemm, typename FallbackGemm, typename... EpilogueArgs>
inline void fallback_cutlass_gemm_caller(torch::Tensor& out,
torch::Tensor const& a,
torch::Tensor const& b,
EpilogueArgs&&... args) {
// In some cases, the GPU isn't able to accommodate the
// shared memory requirements of the Gemm. In such cases, use
// the FallbackGemm instead.
static const int max_shared_mem_per_block_opt_in =
get_cuda_max_shared_memory_per_block_opt_in(0);
size_t const gemm_shared_mem_size =
sizeof(typename Gemm::KernelType::SharedStorage);
size_t const fallback_gemm_shared_mem_size =
sizeof(typename FallbackGemm::KernelType::SharedStorage);
if (gemm_shared_mem_size <= max_shared_mem_per_block_opt_in) {
return cutlass_gemm_caller<Gemm>(out, a, b,
std::forward<EpilogueArgs>(args)...);
} else {
TORCH_CHECK(fallback_gemm_shared_mem_size <=
max_shared_mem_per_block_opt_in);
return cutlass_gemm_caller<FallbackGemm>(
out, a, b, std::forward<EpilogueArgs>(args)...);
}
}
} // namespace vllm
#pragma once
#include "scaled_mm_c2x.cuh"
/**
* This file defines Gemm kernel configurations for SM75 based on the Gemm
* shape.
*/
namespace vllm {
template <typename InType, typename OutType,
template <typename, typename> typename Epilogue>
struct sm75_config_default {
// This config is used in 2 cases,
// - M in (256, inf]
// - M in (64, 128]
// Shared memory required by this Gemm 32768
static_assert(std::is_same<InType, int8_t>());
using TileShape = typename cutlass::gemm::GemmShape<128, 128, 64>;
using WarpShape = typename cutlass::gemm::GemmShape<64, 64, 64>;
using InstructionShape = typename cutlass::gemm::GemmShape<8, 8, 16>;
using Cutlass2xGemm =
cutlass_2x_gemm<cutlass::arch::Sm75, enable_sm75_to_sm80, InType, OutType,
Epilogue, TileShape, WarpShape, InstructionShape, 2>;
};
template <typename InType, typename OutType,
template <typename, typename> typename Epilogue>
struct sm75_config_M256 {
// M in (128, 256]
// Shared memory required by this Gemm 65536
static_assert(std::is_same<InType, int8_t>());
using TileShape = typename cutlass::gemm::GemmShape<128, 128, 128>;
using WarpShape = typename cutlass::gemm::GemmShape<64, 64, 64>;
using InstructionShape = typename cutlass::gemm::GemmShape<8, 8, 16>;
using Cutlass2xGemm =
cutlass_2x_gemm<cutlass::arch::Sm75, enable_sm75_to_sm80, InType, OutType,
Epilogue, TileShape, WarpShape, InstructionShape, 2>;
};
template <typename InType, typename OutType,
template <typename, typename> typename Epilogue>
struct sm75_config_M64 {
// M in (32, 64]
// Shared memory required by this Gemm 49152
static_assert(std::is_same<InType, int8_t>());
using TileShape = typename cutlass::gemm::GemmShape<64, 128, 128>;
using WarpShape = typename cutlass::gemm::GemmShape<64, 64, 64>;
using InstructionShape = typename cutlass::gemm::GemmShape<8, 8, 16>;
using Cutlass2xGemm =
cutlass_2x_gemm<cutlass::arch::Sm75, enable_sm75_to_sm80, InType, OutType,
Epilogue, TileShape, WarpShape, InstructionShape, 2>;
};
template <typename InType, typename OutType,
template <typename, typename> typename Epilogue>
struct sm75_config_M32 {
// M in [1, 32]
// Shared memory required by this Gemm 49152
static_assert(std::is_same<InType, int8_t>());
using TileShape = typename cutlass::gemm::GemmShape<32, 128, 64>;
using WarpShape = typename cutlass::gemm::GemmShape<32, 64, 64>;
using InstructionShape = typename cutlass::gemm::GemmShape<8, 8, 16>;
using Cutlass2xGemm =
cutlass_2x_gemm<cutlass::arch::Sm75, enable_sm75_to_sm80, InType, OutType,
Epilogue, TileShape, WarpShape, InstructionShape, 2>;
};
template <typename InType, typename OutType,
template <typename, typename> typename Epilogue,
typename... EpilogueArgs>
inline void cutlass_gemm_sm75_dispatch(torch::Tensor& out,
torch::Tensor const& a,
torch::Tensor const& b,
EpilogueArgs&&... args) {
static_assert(std::is_same<InType, int8_t>());
TORCH_CHECK(a.dtype() == torch::kInt8);
TORCH_CHECK(b.dtype() == torch::kInt8);
using Cutlass2xGemmDefault =
typename sm75_config_default<InType, OutType, Epilogue>::Cutlass2xGemm;
using Cutlass2xGemmM256 =
typename sm75_config_M256<InType, OutType, Epilogue>::Cutlass2xGemm;
using Cutlass2xGemmM128 = Cutlass2xGemmDefault;
using Cutlass2xGemmM64 =
typename sm75_config_M64<InType, OutType, Epilogue>::Cutlass2xGemm;
using Cutlass2xGemmM32 =
typename sm75_config_M32<InType, OutType, Epilogue>::Cutlass2xGemm;
// Due to shared memory requirements, some Gemms may fail to run on some
// GPUs. As the name indicates, the Fallback Gemm is used as an alternative
// in such cases.
// sm75_config_default has the least shared-memory requirements.
using FallbackGemm = Cutlass2xGemmDefault;
uint32_t const m = a.size(0);
uint32_t const mp2 =
std::max(static_cast<uint32_t>(32), next_pow_2(m)); // next power of 2
if (mp2 <= 32) {
// M in [1, 32]
return fallback_cutlass_gemm_caller<Cutlass2xGemmM32, FallbackGemm>(
out, a, b, std::forward<EpilogueArgs>(args)...);
} else if (mp2 <= 64) {
// M in (32, 64]
return fallback_cutlass_gemm_caller<Cutlass2xGemmM64, FallbackGemm>(
out, a, b, std::forward<EpilogueArgs>(args)...);
} else if (mp2 <= 128) {
// M in (64, 128]
return fallback_cutlass_gemm_caller<Cutlass2xGemmM128, FallbackGemm>(
out, a, b, std::forward<EpilogueArgs>(args)...);
} else if (mp2 <= 256) {
// M in (128, 256]
return fallback_cutlass_gemm_caller<Cutlass2xGemmM256, FallbackGemm>(
out, a, b, std::forward<EpilogueArgs>(args)...);
} else {
// M in (256, inf)
return fallback_cutlass_gemm_caller<Cutlass2xGemmDefault, FallbackGemm>(
out, a, b, std::forward<EpilogueArgs>(args)...);
}
}
} // namespace vllm
#pragma once
#include "scaled_mm_c2x.cuh"
/**
* This file defines Gemm kernel configurations for SM80 based on the Gemm
* shape.
*/
namespace vllm {
template <typename InType, typename OutType,
template <typename, typename> typename Epilogue>
struct sm80_config_default {
// This config is used in 2 cases,
// - M in (128, inf)
// - M in (64, 128] and N >= 8192
// Shared Memory required by this Gemm - 81920 bytes
static_assert(std::is_same<InType, int8_t>());
using TileShape = typename cutlass::gemm::GemmShape<128, 128, 64>;
using WarpShape = typename cutlass::gemm::GemmShape<64, 64, 64>;
using InstructionShape = typename cutlass::gemm::GemmShape<16, 8, 32>;
using Cutlass2xGemm =
cutlass_2x_gemm<cutlass::arch::Sm80, enable_sm80_to_sm89, InType, OutType,
Epilogue, TileShape, WarpShape, InstructionShape, 5>;
};
template <typename InType, typename OutType,
template <typename, typename> typename Epilogue>
struct sm80_config_M64 {
// This config is used in 2 cases,
// - M in (32, 64]
// - M in (64, 128] and N < 8192
// Shared Memory required by this Gemm - 122880 bytes
static_assert(std::is_same<InType, int8_t>());
using TileShape = typename cutlass::gemm::GemmShape<64, 128, 128>;
using WarpShape = typename cutlass::gemm::GemmShape<64, 64, 64>;
using InstructionShape = typename cutlass::gemm::GemmShape<16, 8, 32>;
using Cutlass2xGemm =
cutlass_2x_gemm<cutlass::arch::Sm80, enable_sm80_to_sm89, InType, OutType,
Epilogue, TileShape, WarpShape, InstructionShape, 5>;
};
template <typename InType, typename OutType,
template <typename, typename> typename Epilogue>
struct sm80_config_M32 {
// M in (16, 32]
// Shared Memory required by this Gemm - 61440 bytes
static_assert(std::is_same<InType, int8_t>());
using TileShape = typename cutlass::gemm::GemmShape<32, 64, 128>;
using WarpShape = typename cutlass::gemm::GemmShape<32, 64, 64>;
using InstructionShape = typename cutlass::gemm::GemmShape<16, 8, 32>;
using Cutlass2xGemm =
cutlass_2x_gemm<cutlass::arch::Sm80, enable_sm80_to_sm89, InType, OutType,
Epilogue, TileShape, WarpShape, InstructionShape, 5>;
};
template <typename InType, typename OutType,
template <typename, typename> typename Epilogue>
struct sm80_config_M16 {
// M in [1, 16]
// Shared Memory required by this Gemm - 51200 bytes
static_assert(std::is_same<InType, int8_t>());
using TileShape = typename cutlass::gemm::GemmShape<16, 64, 128>;
using WarpShape = typename cutlass::gemm::GemmShape<16, 64, 64>;
using InstructionShape = typename cutlass::gemm::GemmShape<16, 8, 32>;
using Cutlass2xGemm =
cutlass_2x_gemm<cutlass::arch::Sm80, enable_sm80_to_sm89, InType, OutType,
Epilogue, TileShape, WarpShape, InstructionShape, 5>;
};
template <typename InType, typename OutType,
template <typename, typename> typename Epilogue,
typename... EpilogueArgs>
inline void cutlass_gemm_sm80_dispatch(torch::Tensor& out,
torch::Tensor const& a,
torch::Tensor const& b,
EpilogueArgs&&... args) {
static_assert(std::is_same<InType, int8_t>());
TORCH_CHECK(a.dtype() == torch::kInt8);
TORCH_CHECK(b.dtype() == torch::kInt8);
using Cutlass2xGemmDefault =
typename sm80_config_default<InType, OutType, Epilogue>::Cutlass2xGemm;
using Cutlass2xGemmM128BigN =
typename sm80_config_default<InType, OutType, Epilogue>::Cutlass2xGemm;
using Cutlass2xGemmM128SmallN =
typename sm80_config_M64<InType, OutType, Epilogue>::Cutlass2xGemm;
using Cutlass2xGemmM64 =
typename sm80_config_M64<InType, OutType, Epilogue>::Cutlass2xGemm;
using Cutlass2xGemmM32 =
typename sm80_config_M32<InType, OutType, Epilogue>::Cutlass2xGemm;
using Cutlass2xGemmM16 =
typename sm80_config_M16<InType, OutType, Epilogue>::Cutlass2xGemm;
// Due to shared memory requirements, some Gemms may fail to run on some
// GPUs. As the name indicates, the Fallback Gemm is used as an alternative
// in such cases.
// sm80_config_M16 has the least shared-memory requirement. However,
// based on some profiling, we select sm80_config_M32 as a better alternative
// performance wise.
using FallbackGemm =
typename sm80_config_M32<InType, OutType, Epilogue>::Cutlass2xGemm;
uint32_t const m = a.size(0);
uint32_t const mp2 =
std::max(static_cast<uint32_t>(16), next_pow_2(m)); // next power of 2
if (mp2 <= 16) {
// M in [1, 16]
return fallback_cutlass_gemm_caller<Cutlass2xGemmM16, FallbackGemm>(
out, a, b, std::forward<EpilogueArgs>(args)...);
} else if (mp2 <= 32) {
// M in (16, 32]
return fallback_cutlass_gemm_caller<Cutlass2xGemmM32, FallbackGemm>(
out, a, b, std::forward<EpilogueArgs>(args)...);
} else if (mp2 <= 64) {
// M in (32, 64]
return fallback_cutlass_gemm_caller<Cutlass2xGemmM64, FallbackGemm>(
out, a, b, std::forward<EpilogueArgs>(args)...);
} else if (mp2 <= 128) {
// M in (64, 128]
uint32_t const n = out.size(1);
bool const small_n = n < 8192;
if (small_n) {
return fallback_cutlass_gemm_caller<Cutlass2xGemmM128SmallN,
FallbackGemm>(
out, a, b, std::forward<EpilogueArgs>(args)...);
} else {
return fallback_cutlass_gemm_caller<Cutlass2xGemmM128BigN, FallbackGemm>(
out, a, b, std::forward<EpilogueArgs>(args)...);
}
} else {
// M in (128, inf)
return fallback_cutlass_gemm_caller<Cutlass2xGemmDefault, FallbackGemm>(
out, a, b, std::forward<EpilogueArgs>(args)...);
}
}
} // namespace vllm
#pragma once
#include "scaled_mm_c2x.cuh"
#include "cutlass/float8.h"
/**
* This file defines Gemm kernel configurations for SM89 (FP8) based on the Gemm
* shape.
*/
namespace vllm {
template <typename InType, typename OutType,
template <typename, typename> typename Epilogue>
struct sm89_fp8_fallback_gemm {
// Shared Memory required by this Gemm - 61440 bytes
static_assert(std::is_same<InType, cutlass::float_e4m3_t>());
using TileShape = typename cutlass::gemm::GemmShape<64, 128, 64>;
using WarpShape = typename cutlass::gemm::GemmShape<32, 64, 64>;
using InstructionShape = typename cutlass::gemm::GemmShape<16, 8, 32>;
using FP8MathOperator = typename cutlass::arch::OpMultiplyAdd;
using Cutlass2xGemm =
cutlass_2x_gemm<cutlass::arch::Sm89, enable_sm89_to_sm90, InType, OutType,
Epilogue, TileShape, WarpShape, InstructionShape, 5,
FP8MathOperator>;
};
struct sm89_fp8_config_default {
// M in (256, inf)
using WarpShape = typename cutlass::gemm::GemmShape<64, 64, 64>;
using InstructionShape = typename cutlass::gemm::GemmShape<16, 8, 32>;
using FP8MathOperator = typename cutlass::arch::OpMultiplyAddFastAccum;
template <typename InType, typename OutType,
template <typename, typename> typename Epilogue,
typename... EpilogueArgs>
static void dispatch(torch::Tensor& out, torch::Tensor const& a,
torch::Tensor const& b, EpilogueArgs&&... args) {
static_assert(std::is_same<InType, cutlass::float_e4m3_t>());
TORCH_CHECK(a.dtype() == torch::kFloat8_e4m3fn);
using FallbackGemm =
typename sm89_fp8_fallback_gemm<InType, OutType,
Epilogue>::Cutlass2xGemm;
uint32_t const n = out.size(1);
uint32_t const np2 = next_pow_2(n);
if (np2 <= 4096) {
using TileShape = typename cutlass::gemm::GemmShape<128, 128, 64>;
return vllm::fallback_cutlass_gemm_caller<
vllm::cutlass_2x_gemm<cutlass::arch::Sm89, vllm::enable_sm89_to_sm90,
InType, OutType, Epilogue, TileShape, WarpShape,
InstructionShape, 5, FP8MathOperator>,
FallbackGemm>(out, a, b, std::forward<EpilogueArgs>(args)...);
} else if (np2 <= 8192) {
using TileShape = typename cutlass::gemm::GemmShape<256, 128, 64>;
return vllm::fallback_cutlass_gemm_caller<
vllm::cutlass_2x_gemm<cutlass::arch::Sm89, vllm::enable_sm89_to_sm90,
InType, OutType, Epilogue, TileShape, WarpShape,
InstructionShape, 3, FP8MathOperator>,
FallbackGemm>(out, a, b, std::forward<EpilogueArgs>(args)...);
} else {
using TileShape = typename cutlass::gemm::GemmShape<128, 128, 64>;
return vllm::fallback_cutlass_gemm_caller<
vllm::cutlass_2x_gemm<cutlass::arch::Sm89, vllm::enable_sm89_to_sm90,
InType, OutType, Epilogue, TileShape, WarpShape,
InstructionShape, 5, FP8MathOperator>,
FallbackGemm>(out, a, b, std::forward<EpilogueArgs>(args)...);
}
}
};
struct sm89_fp8_config_M256 {
// M in (128, 256]
using WarpShape = typename cutlass::gemm::GemmShape<64, 64, 64>;
using InstructionShape = typename cutlass::gemm::GemmShape<16, 8, 32>;
using FP8MathOperator = typename cutlass::arch::OpMultiplyAddFastAccum;
template <typename InType, typename OutType,
template <typename, typename> typename Epilogue,
typename... EpilogueArgs>
static void dispatch(torch::Tensor& out, torch::Tensor const& a,
torch::Tensor const& b, EpilogueArgs&&... args) {
static_assert(std::is_same<InType, cutlass::float_e4m3_t>());
TORCH_CHECK(a.dtype() == torch::kFloat8_e4m3fn);
using FallbackGemm =
typename sm89_fp8_fallback_gemm<InType, OutType,
Epilogue>::Cutlass2xGemm;
uint32_t const n = out.size(1);
uint32_t const np2 = next_pow_2(n);
if (np2 <= 4096) {
using TileShape = typename cutlass::gemm::GemmShape<64, 128, 128>;
return vllm::fallback_cutlass_gemm_caller<
vllm::cutlass_2x_gemm<cutlass::arch::Sm89, vllm::enable_sm89_to_sm90,
InType, OutType, Epilogue, TileShape, WarpShape,
InstructionShape, 3, FP8MathOperator>,
FallbackGemm>(out, a, b, std::forward<EpilogueArgs>(args)...);
} else {
using TileShape = typename cutlass::gemm::GemmShape<128, 128, 64>;
return vllm::fallback_cutlass_gemm_caller<
vllm::cutlass_2x_gemm<cutlass::arch::Sm89, vllm::enable_sm89_to_sm90,
InType, OutType, Epilogue, TileShape, WarpShape,
InstructionShape, 5, FP8MathOperator>,
FallbackGemm>(out, a, b, std::forward<EpilogueArgs>(args)...);
}
}
};
struct sm89_fp8_config_M128 {
// M in (64, 128]
using WarpShape = typename cutlass::gemm::GemmShape<64, 64, 64>;
using InstructionShape = typename cutlass::gemm::GemmShape<16, 8, 32>;
using FP8MathOperator = typename cutlass::arch::OpMultiplyAddFastAccum;
template <typename InType, typename OutType,
template <typename, typename> typename Epilogue,
typename... EpilogueArgs>
static void dispatch(torch::Tensor& out, torch::Tensor const& a,
torch::Tensor const& b, EpilogueArgs&&... args) {
static_assert(std::is_same<InType, cutlass::float_e4m3_t>());
TORCH_CHECK(a.dtype() == torch::kFloat8_e4m3fn);
using FallbackGemm =
typename sm89_fp8_fallback_gemm<InType, OutType,
Epilogue>::Cutlass2xGemm;
uint32_t const n = out.size(1);
uint32_t const np2 = next_pow_2(n);
if (np2 <= 8192) {
using TileShape = typename cutlass::gemm::GemmShape<64, 128, 128>;
return vllm::fallback_cutlass_gemm_caller<
vllm::cutlass_2x_gemm<cutlass::arch::Sm89, vllm::enable_sm89_to_sm90,
InType, OutType, Epilogue, TileShape, WarpShape,
InstructionShape, 3, FP8MathOperator>,
FallbackGemm>(out, a, b, std::forward<EpilogueArgs>(args)...);
} else if (np2 <= 16384) {
using TileShape = typename cutlass::gemm::GemmShape<128, 128, 64>;
return vllm::fallback_cutlass_gemm_caller<
vllm::cutlass_2x_gemm<cutlass::arch::Sm89, vllm::enable_sm89_to_sm90,
InType, OutType, Epilogue, TileShape, WarpShape,
InstructionShape, 5, FP8MathOperator>,
FallbackGemm>(out, a, b, std::forward<EpilogueArgs>(args)...);
} else {
using TileShape = typename cutlass::gemm::GemmShape<128, 64, 128>;
return vllm::fallback_cutlass_gemm_caller<
vllm::cutlass_2x_gemm<cutlass::arch::Sm89, vllm::enable_sm89_to_sm90,
InType, OutType, Epilogue, TileShape, WarpShape,
InstructionShape, 3, FP8MathOperator>,
FallbackGemm>(out, a, b, std::forward<EpilogueArgs>(args)...);
}
}
};
struct sm89_fp8_config_M64 {
// M in (32, 64]
using InstructionShape = typename cutlass::gemm::GemmShape<16, 8, 32>;
template <typename InType, typename OutType,
template <typename, typename> typename Epilogue,
typename... EpilogueArgs>
static void dispatch(torch::Tensor& out, torch::Tensor const& a,
torch::Tensor const& b, EpilogueArgs&&... args) {
static_assert(std::is_same<InType, cutlass::float_e4m3_t>());
TORCH_CHECK(a.dtype() == torch::kFloat8_e4m3fn);
using FallbackGemm =
typename sm89_fp8_fallback_gemm<InType, OutType,
Epilogue>::Cutlass2xGemm;
uint32_t const n = out.size(1);
uint32_t const np2 = next_pow_2(n);
if (np2 <= 8196) {
using TileShape = typename cutlass::gemm::GemmShape<64, 64, 128>;
using WarpShape = typename cutlass::gemm::GemmShape<32, 64, 64>;
using FP8MathOperator = typename cutlass::arch::OpMultiplyAdd;
return vllm::fallback_cutlass_gemm_caller<
vllm::cutlass_2x_gemm<cutlass::arch::Sm89, vllm::enable_sm89_to_sm90,
InType, OutType, Epilogue, TileShape, WarpShape,
InstructionShape, 5, FP8MathOperator>,
FallbackGemm>(out, a, b, std::forward<EpilogueArgs>(args)...);
} else if (np2 <= 16384) {
using TileShape = typename cutlass::gemm::GemmShape<64, 128, 128>;
using WarpShape = typename cutlass::gemm::GemmShape<64, 64, 64>;
using FP8MathOperator = typename cutlass::arch::OpMultiplyAddFastAccum;
return vllm::fallback_cutlass_gemm_caller<
vllm::cutlass_2x_gemm<cutlass::arch::Sm89, vllm::enable_sm89_to_sm90,
InType, OutType, Epilogue, TileShape, WarpShape,
InstructionShape, 3, FP8MathOperator>,
FallbackGemm>(out, a, b, std::forward<EpilogueArgs>(args)...);
} else {
using TileShape = typename cutlass::gemm::GemmShape<64, 64, 128>;
using WarpShape = typename cutlass::gemm::GemmShape<32, 64, 64>;
using FP8MathOperator = typename cutlass::arch::OpMultiplyAdd;
return vllm::fallback_cutlass_gemm_caller<
vllm::cutlass_2x_gemm<cutlass::arch::Sm89, vllm::enable_sm89_to_sm90,
InType, OutType, Epilogue, TileShape, WarpShape,
InstructionShape, 5, FP8MathOperator>,
FallbackGemm>(out, a, b, std::forward<EpilogueArgs>(args)...);
}
}
};
struct sm89_fp8_config_M32 {
// M in (16, 32]
using InstructionShape = typename cutlass::gemm::GemmShape<16, 8, 32>;
using FP8MathOperator = typename cutlass::arch::OpMultiplyAddFastAccum;
template <typename InType, typename OutType,
template <typename, typename> typename Epilogue,
typename... EpilogueArgs>
static void dispatch(torch::Tensor& out, torch::Tensor const& a,
torch::Tensor const& b, EpilogueArgs&&... args) {
static_assert(std::is_same<InType, cutlass::float_e4m3_t>());
TORCH_CHECK(a.dtype() == torch::kFloat8_e4m3fn);
using FallbackGemm =
typename sm89_fp8_fallback_gemm<InType, OutType,
Epilogue>::Cutlass2xGemm;
uint32_t const n = out.size(1);
uint32_t const np2 = next_pow_2(n);
if (np2 <= 8192) {
using TileShape = typename cutlass::gemm::GemmShape<32, 64, 128>;
using WarpShape = typename cutlass::gemm::GemmShape<16, 64, 64>;
return vllm::fallback_cutlass_gemm_caller<
vllm::cutlass_2x_gemm<cutlass::arch::Sm89, vllm::enable_sm89_to_sm90,
InType, OutType, Epilogue, TileShape, WarpShape,
InstructionShape, 5, FP8MathOperator>,
FallbackGemm>(out, a, b, std::forward<EpilogueArgs>(args)...);
} else if (np2 <= 16384) {
using TileShape = typename cutlass::gemm::GemmShape<32, 128, 128>;
using WarpShape = typename cutlass::gemm::GemmShape<32, 64, 64>;
return vllm::fallback_cutlass_gemm_caller<
vllm::cutlass_2x_gemm<cutlass::arch::Sm89, vllm::enable_sm89_to_sm90,
InType, OutType, Epilogue, TileShape, WarpShape,
InstructionShape, 4, FP8MathOperator>,
FallbackGemm>(out, a, b, std::forward<EpilogueArgs>(args)...);
} else {
using TileShape = typename cutlass::gemm::GemmShape<32, 64, 128>;
using WarpShape = typename cutlass::gemm::GemmShape<16, 64, 64>;
return vllm::fallback_cutlass_gemm_caller<
vllm::cutlass_2x_gemm<cutlass::arch::Sm89, vllm::enable_sm89_to_sm90,
InType, OutType, Epilogue, TileShape, WarpShape,
InstructionShape, 5, FP8MathOperator>,
FallbackGemm>(out, a, b, std::forward<EpilogueArgs>(args)...);
}
}
};
struct sm89_fp8_config_M16 {
// M in [1, 16]
using WarpShape = typename cutlass::gemm::GemmShape<16, 64, 64>;
using InstructionShape = typename cutlass::gemm::GemmShape<16, 8, 32>;
using FP8MathOperator = typename cutlass::arch::OpMultiplyAddFastAccum;
static const int32_t MainLoopStages = 5;
template <typename InType, typename OutType,
template <typename, typename> typename Epilogue,
typename... EpilogueArgs>
static void dispatch(torch::Tensor& out, torch::Tensor const& a,
torch::Tensor const& b, EpilogueArgs&&... args) {
static_assert(std::is_same<InType, cutlass::float_e4m3_t>());
TORCH_CHECK(a.dtype() == torch::kFloat8_e4m3fn);
using FallbackGemm =
typename sm89_fp8_fallback_gemm<InType, OutType,
Epilogue>::Cutlass2xGemm;
uint32_t const n = out.size(1);
uint32_t const np2 = next_pow_2(n);
if (np2 <= 8192) {
using TileShape = typename cutlass::gemm::GemmShape<16, 64, 128>;
return vllm::fallback_cutlass_gemm_caller<
vllm::cutlass_2x_gemm<cutlass::arch::Sm89, vllm::enable_sm89_to_sm90,
InType, OutType, Epilogue, TileShape, WarpShape,
InstructionShape, MainLoopStages,
FP8MathOperator>,
FallbackGemm>(out, a, b, std::forward<EpilogueArgs>(args)...);
} else if (np2 <= 24576) {
using TileShape = typename cutlass::gemm::GemmShape<16, 128, 64>;
return vllm::fallback_cutlass_gemm_caller<
vllm::cutlass_2x_gemm<cutlass::arch::Sm89, vllm::enable_sm89_to_sm90,
InType, OutType, Epilogue, TileShape, WarpShape,
InstructionShape, MainLoopStages,
FP8MathOperator>,
FallbackGemm>(out, a, b, std::forward<EpilogueArgs>(args)...);
} else {
using TileShape = typename cutlass::gemm::GemmShape<32, 64, 128>;
return vllm::fallback_cutlass_gemm_caller<
vllm::cutlass_2x_gemm<cutlass::arch::Sm89, vllm::enable_sm89_to_sm90,
InType, OutType, Epilogue, TileShape, WarpShape,
InstructionShape, MainLoopStages,
FP8MathOperator>,
FallbackGemm>(out, a, b, std::forward<EpilogueArgs>(args)...);
}
}
};
template <typename InType, typename OutType,
template <typename, typename> typename Epilogue,
typename... EpilogueArgs>
inline void cutlass_gemm_sm89_fp8_dispatch(torch::Tensor& out,
torch::Tensor const& a,
torch::Tensor const& b,
EpilogueArgs&&... args) {
static_assert(std::is_same<InType, cutlass::float_e4m3_t>());
TORCH_CHECK(a.dtype() == torch::kFloat8_e4m3fn);
TORCH_CHECK(b.dtype() == torch::kFloat8_e4m3fn);
uint32_t const m = a.size(0);
uint32_t const mp2 =
std::max(static_cast<uint32_t>(32), next_pow_2(m)); // next power of 2
if (mp2 <= 16) {
// M in [1, 16]
return sm89_fp8_config_M16::dispatch<InType, OutType, Epilogue>(
out, a, b, std::forward<EpilogueArgs>(args)...);
} else if (mp2 <= 32) {
// M in (16, 32]
return sm89_fp8_config_M32::dispatch<InType, OutType, Epilogue>(
out, a, b, std::forward<EpilogueArgs>(args)...);
} else if (mp2 <= 64) {
// M in (32, 64]
return sm89_fp8_config_M64::dispatch<InType, OutType, Epilogue>(
out, a, b, std::forward<EpilogueArgs>(args)...);
} else if (mp2 <= 128) {
// M in (64, 128]
return sm89_fp8_config_M128::dispatch<InType, OutType, Epilogue>(
out, a, b, std::forward<EpilogueArgs>(args)...);
} else if (mp2 <= 256) {
// M in (128, 256]
return sm89_fp8_config_M256::dispatch<InType, OutType, Epilogue>(
out, a, b, std::forward<EpilogueArgs>(args)...);
} else {
// M in (256, inf)
return sm89_fp8_config_default::dispatch<InType, OutType, Epilogue>(
out, a, b, std::forward<EpilogueArgs>(args)...);
}
}
} // namespace vllm
#pragma once
#include "scaled_mm_c2x.cuh"
/**
* This file defines Gemm kernel configurations for SM89 (int8) based on the
* Gemm shape.
*/
namespace vllm {
template <typename InType, typename OutType,
template <typename, typename> typename Epilogue>
struct sm89_int8_fallback_gemm {
// Shared mem requirement : 61440
static_assert(std::is_same<InType, int8_t>());
using TileShape = cutlass::gemm::GemmShape<32, 64, 128>;
using WarpShape = cutlass::gemm::GemmShape<16, 64, 64>;
using InstructionShape = typename cutlass::gemm::GemmShape<16, 8, 32>;
static int32_t const MainLoopStages = 5;
using Cutlass2xGemm =
cutlass_2x_gemm<cutlass::arch::Sm89, enable_sm89_to_sm90, InType, OutType,
Epilogue, TileShape, WarpShape, InstructionShape, 5>;
};
struct sm89_int8_config_default {
// M in (256, inf)
using WarpShape = typename cutlass::gemm::GemmShape<64, 64, 64>;
using InstructionShape = typename cutlass::gemm::GemmShape<16, 8, 32>;
template <typename InType, typename OutType,
template <typename, typename> typename Epilogue,
typename... EpilogueArgs>
static void dispatch(torch::Tensor& out, torch::Tensor const& a,
torch::Tensor const& b, EpilogueArgs&&... args) {
static_assert(std::is_same<InType, int8_t>());
TORCH_CHECK(a.dtype() == torch::kInt8);
using FallbackGemm =
typename sm89_int8_fallback_gemm<InType, OutType,
Epilogue>::Cutlass2xGemm;
uint32_t const n = out.size(1);
uint32_t const np2 = next_pow_2(n);
if (np2 <= 4096) {
using TileShape = cutlass::gemm::GemmShape<128, 128, 64>;
return vllm::fallback_cutlass_gemm_caller<
vllm::cutlass_2x_gemm<cutlass::arch::Sm89, vllm::enable_sm89_to_sm90,
InType, OutType, Epilogue, TileShape, WarpShape,
InstructionShape, 5>,
FallbackGemm>(out, a, b, std::forward<EpilogueArgs>(args)...);
} else if (np2 <= 8192) {
using TileShape = cutlass::gemm::GemmShape<256, 128, 64>;
return vllm::fallback_cutlass_gemm_caller<
vllm::cutlass_2x_gemm<cutlass::arch::Sm89, vllm::enable_sm89_to_sm90,
InType, OutType, Epilogue, TileShape, WarpShape,
InstructionShape, 3>,
FallbackGemm>(out, a, b, std::forward<EpilogueArgs>(args)...);
} else if (np2 <= 16384) {
using TileShape = cutlass::gemm::GemmShape<128, 128, 64>;
return vllm::fallback_cutlass_gemm_caller<
vllm::cutlass_2x_gemm<cutlass::arch::Sm89, vllm::enable_sm89_to_sm90,
InType, OutType, Epilogue, TileShape, WarpShape,
InstructionShape, 5>,
FallbackGemm>(out, a, b, std::forward<EpilogueArgs>(args)...);
} else {
using TileShape = cutlass::gemm::GemmShape<256, 128, 64>;
return vllm::fallback_cutlass_gemm_caller<
vllm::cutlass_2x_gemm<cutlass::arch::Sm89, vllm::enable_sm89_to_sm90,
InType, OutType, Epilogue, TileShape, WarpShape,
InstructionShape, 3>,
FallbackGemm>(out, a, b, std::forward<EpilogueArgs>(args)...);
}
}
};
struct sm89_int8_config_M256 {
// M in (128, 256]
using WarpShape = typename cutlass::gemm::GemmShape<64, 64, 64>;
using InstructionShape = typename cutlass::gemm::GemmShape<16, 8, 32>;
template <typename InType, typename OutType,
template <typename, typename> typename Epilogue,
typename... EpilogueArgs>
static void dispatch(torch::Tensor& out, torch::Tensor const& a,
torch::Tensor const& b, EpilogueArgs&&... args) {
static_assert(std::is_same<InType, int8_t>());
TORCH_CHECK(a.dtype() == torch::kInt8);
using FallbackGemm =
typename sm89_int8_fallback_gemm<InType, OutType,
Epilogue>::Cutlass2xGemm;
uint32_t const n = out.size(1);
uint32_t const np2 = next_pow_2(n);
if (np2 <= 4096) {
using TileShape = cutlass::gemm::GemmShape<64, 128, 128>;
return vllm::fallback_cutlass_gemm_caller<
vllm::cutlass_2x_gemm<cutlass::arch::Sm89, vllm::enable_sm89_to_sm90,
InType, OutType, Epilogue, TileShape, WarpShape,
InstructionShape, 3>,
FallbackGemm>(out, a, b, std::forward<EpilogueArgs>(args)...);
} else if (np2 <= 8192) {
using TileShape = cutlass::gemm::GemmShape<128, 128, 64>;
return vllm::fallback_cutlass_gemm_caller<
vllm::cutlass_2x_gemm<cutlass::arch::Sm89, vllm::enable_sm89_to_sm90,
InType, OutType, Epilogue, TileShape, WarpShape,
InstructionShape, 5>,
FallbackGemm>(out, a, b, std::forward<EpilogueArgs>(args)...);
} else if (np2 <= 16384) {
using TileShape = cutlass::gemm::GemmShape<256, 128, 64>;
return vllm::fallback_cutlass_gemm_caller<
vllm::cutlass_2x_gemm<cutlass::arch::Sm89, vllm::enable_sm89_to_sm90,
InType, OutType, Epilogue, TileShape, WarpShape,
InstructionShape, 3>,
FallbackGemm>(out, a, b, std::forward<EpilogueArgs>(args)...);
} else {
using TileShape = cutlass::gemm::GemmShape<128, 128, 64>;
return vllm::fallback_cutlass_gemm_caller<
vllm::cutlass_2x_gemm<cutlass::arch::Sm89, vllm::enable_sm89_to_sm90,
InType, OutType, Epilogue, TileShape, WarpShape,
InstructionShape, 5>,
FallbackGemm>(out, a, b, std::forward<EpilogueArgs>(args)...);
}
}
};
struct sm89_int8_config_M128 {
// M in (64, 128]
using InstructionShape = typename cutlass::gemm::GemmShape<16, 8, 32>;
template <typename InType, typename OutType,
template <typename, typename> typename Epilogue,
typename... EpilogueArgs>
static void dispatch(torch::Tensor& out, torch::Tensor const& a,
torch::Tensor const& b, EpilogueArgs&&... args) {
static_assert(std::is_same<InType, int8_t>());
TORCH_CHECK(a.dtype() == torch::kInt8);
using FallbackGemm =
typename sm89_int8_fallback_gemm<InType, OutType,
Epilogue>::Cutlass2xGemm;
uint32_t const n = out.size(1);
uint32_t const np2 = next_pow_2(n);
if (np2 <= 8192) {
using TileShape = cutlass::gemm::GemmShape<64, 128, 128>;
using WarpShape = cutlass::gemm::GemmShape<64, 64, 64>;
return vllm::fallback_cutlass_gemm_caller<
vllm::cutlass_2x_gemm<cutlass::arch::Sm89, vllm::enable_sm89_to_sm90,
InType, OutType, Epilogue, TileShape, WarpShape,
InstructionShape, 3>,
FallbackGemm>(out, a, b, std::forward<EpilogueArgs>(args)...);
} else if (np2 <= 16384) {
using TileShape = cutlass::gemm::GemmShape<128, 128, 64>;
using WarpShape = cutlass::gemm::GemmShape<64, 64, 64>;
return vllm::fallback_cutlass_gemm_caller<
vllm::cutlass_2x_gemm<cutlass::arch::Sm89, vllm::enable_sm89_to_sm90,
InType, OutType, Epilogue, TileShape, WarpShape,
InstructionShape, 5>,
FallbackGemm>(out, a, b, std::forward<EpilogueArgs>(args)...);
} else {
using TileShape = cutlass::gemm::GemmShape<64, 64, 128>;
using WarpShape = cutlass::gemm::GemmShape<32, 64, 64>;
return vllm::fallback_cutlass_gemm_caller<
vllm::cutlass_2x_gemm<cutlass::arch::Sm89, vllm::enable_sm89_to_sm90,
InType, OutType, Epilogue, TileShape, WarpShape,
InstructionShape, 5>,
FallbackGemm>(out, a, b, std::forward<EpilogueArgs>(args)...);
}
}
};
struct sm89_int8_config_M64 {
// M in (32, 64]
using InstructionShape = typename cutlass::gemm::GemmShape<16, 8, 32>;
template <typename InType, typename OutType,
template <typename, typename> typename Epilogue,
typename... EpilogueArgs>
static void dispatch(torch::Tensor& out, torch::Tensor const& a,
torch::Tensor const& b, EpilogueArgs&&... args) {
static_assert(std::is_same<InType, int8_t>());
TORCH_CHECK(a.dtype() == torch::kInt8);
using FallbackGemm =
typename sm89_int8_fallback_gemm<InType, OutType,
Epilogue>::Cutlass2xGemm;
uint32_t const n = out.size(1);
uint32_t const np2 = next_pow_2(n);
if (np2 <= 8192) {
using TileShape = cutlass::gemm::GemmShape<64, 64, 128>;
using WarpShape = cutlass::gemm::GemmShape<32, 64, 64>;
return vllm::fallback_cutlass_gemm_caller<
vllm::cutlass_2x_gemm<cutlass::arch::Sm89, vllm::enable_sm89_to_sm90,
InType, OutType, Epilogue, TileShape, WarpShape,
InstructionShape, 5>,
FallbackGemm>(out, a, b, std::forward<EpilogueArgs>(args)...);
} else {
using TileShape = cutlass::gemm::GemmShape<64, 128, 128>;
using WarpShape = cutlass::gemm::GemmShape<64, 64, 64>;
return vllm::fallback_cutlass_gemm_caller<
vllm::cutlass_2x_gemm<cutlass::arch::Sm89, vllm::enable_sm89_to_sm90,
InType, OutType, Epilogue, TileShape, WarpShape,
InstructionShape, 3>,
FallbackGemm>(out, a, b, std::forward<EpilogueArgs>(args)...);
}
}
};
struct sm89_int8_config_M32 {
// M in (16, 32]
using InstructionShape = typename cutlass::gemm::GemmShape<16, 8, 32>;
template <typename InType, typename OutType,
template <typename, typename> typename Epilogue,
typename... EpilogueArgs>
static void dispatch(torch::Tensor& out, torch::Tensor const& a,
torch::Tensor const& b, EpilogueArgs&&... args) {
static_assert(std::is_same<InType, int8_t>());
TORCH_CHECK(a.dtype() == torch::kInt8);
using FallbackGemm =
typename sm89_int8_fallback_gemm<InType, OutType,
Epilogue>::Cutlass2xGemm;
uint32_t const n = out.size(1);
uint32_t const np2 = next_pow_2(n);
if (np2 <= 8192) {
using TileShape = cutlass::gemm::GemmShape<32, 64, 128>;
using WarpShape = cutlass::gemm::GemmShape<16, 64, 64>;
return vllm::fallback_cutlass_gemm_caller<
vllm::cutlass_2x_gemm<cutlass::arch::Sm89, vllm::enable_sm89_to_sm90,
InType, OutType, Epilogue, TileShape, WarpShape,
InstructionShape, 5>,
FallbackGemm>(out, a, b, std::forward<EpilogueArgs>(args)...);
} else {
using TileShape = cutlass::gemm::GemmShape<32, 128, 128>;
using WarpShape = cutlass::gemm::GemmShape<32, 64, 64>;
return vllm::fallback_cutlass_gemm_caller<
vllm::cutlass_2x_gemm<cutlass::arch::Sm89, vllm::enable_sm89_to_sm90,
InType, OutType, Epilogue, TileShape, WarpShape,
InstructionShape, 4>,
FallbackGemm>(out, a, b, std::forward<EpilogueArgs>(args)...);
}
}
};
struct sm89_int8_config_M16 {
// M in [1, 16]
using WarpShape = typename cutlass::gemm::GemmShape<16, 64, 64>;
using InstructionShape = typename cutlass::gemm::GemmShape<16, 8, 32>;
template <typename InType, typename OutType,
template <typename, typename> typename Epilogue,
typename... EpilogueArgs>
static void dispatch(torch::Tensor& out, torch::Tensor const& a,
torch::Tensor const& b, EpilogueArgs&&... args) {
static_assert(std::is_same<InType, int8_t>());
TORCH_CHECK(a.dtype() == torch::kInt8);
using FallbackGemm =
typename sm89_int8_fallback_gemm<InType, OutType,
Epilogue>::Cutlass2xGemm;
uint32_t const n = out.size(1);
uint32_t const np2 = next_pow_2(n);
if (np2 <= 8192) {
using TileShape = cutlass::gemm::GemmShape<16, 64, 128>;
return vllm::fallback_cutlass_gemm_caller<
vllm::cutlass_2x_gemm<cutlass::arch::Sm89, vllm::enable_sm89_to_sm90,
InType, OutType, Epilogue, TileShape, WarpShape,
InstructionShape, 5>,
FallbackGemm>(out, a, b, std::forward<EpilogueArgs>(args)...);
} else {
using TileShape = cutlass::gemm::GemmShape<16, 128, 128>;
return vllm::fallback_cutlass_gemm_caller<
vllm::cutlass_2x_gemm<cutlass::arch::Sm89, vllm::enable_sm89_to_sm90,
InType, OutType, Epilogue, TileShape, WarpShape,
InstructionShape, 4>,
FallbackGemm>(out, a, b, std::forward<EpilogueArgs>(args)...);
}
}
};
template <typename InType, typename OutType,
template <typename, typename> typename Epilogue,
typename... EpilogueArgs>
inline void cutlass_gemm_sm89_int8_dispatch(torch::Tensor& out,
torch::Tensor const& a,
torch::Tensor const& b,
EpilogueArgs&&... args) {
static_assert(std::is_same<InType, int8_t>());
TORCH_CHECK(a.dtype() == torch::kInt8);
TORCH_CHECK(b.dtype() == torch::kInt8);
uint32_t const m = a.size(0);
uint32_t const mp2 =
std::max(static_cast<uint32_t>(32), next_pow_2(m)); // next power of 2
if (mp2 <= 16) {
// M in [1, 16]
return sm89_int8_config_M16::dispatch<InType, OutType, Epilogue>(
out, a, b, std::forward<EpilogueArgs>(args)...);
} else if (mp2 <= 32) {
// M in (16, 32]
return sm89_int8_config_M32::dispatch<InType, OutType, Epilogue>(
out, a, b, std::forward<EpilogueArgs>(args)...);
} else if (mp2 <= 64) {
// M in (32, 64]
return sm89_int8_config_M64::dispatch<InType, OutType, Epilogue>(
out, a, b, std::forward<EpilogueArgs>(args)...);
} else if (mp2 <= 128) {
// M in (64, 128]
return sm89_int8_config_M128::dispatch<InType, OutType, Epilogue>(
out, a, b, std::forward<EpilogueArgs>(args)...);
} else if (mp2 <= 256) {
// M in (128, 256]
return sm89_int8_config_M256::dispatch<InType, OutType, Epilogue>(
out, a, b, std::forward<EpilogueArgs>(args)...);
} else {
// M in (256, inf)
return sm89_int8_config_default::dispatch<InType, OutType, Epilogue>(
out, a, b, std::forward<EpilogueArgs>(args)...);
}
}
} // namespace vllm
...@@ -18,8 +18,6 @@ ...@@ -18,8 +18,6 @@
#include "cute/atom/mma_atom.hpp" #include "cute/atom/mma_atom.hpp"
#include "cutlass/numeric_types.h" #include "cutlass/numeric_types.h"
#include "cutlass/util/device_memory.h"
#include "cutlass/gemm/device/gemm_universal_adapter.h" #include "cutlass/gemm/device/gemm_universal_adapter.h"
#include "cutlass/gemm/kernel/gemm_universal.hpp" #include "cutlass/gemm/kernel/gemm_universal.hpp"
#include "cutlass/epilogue/collective/collective_builder.hpp" #include "cutlass/epilogue/collective/collective_builder.hpp"
...@@ -44,11 +42,6 @@ using namespace cute; ...@@ -44,11 +42,6 @@ using namespace cute;
namespace { namespace {
uint32_t next_pow_2(uint32_t const num) {
if (num <= 1) return num;
return 1 << (CHAR_BIT * sizeof(num) - __builtin_clz(num - 1));
}
// A wrapper for the GEMM kernel that is used to guard against compilation on // A wrapper for the GEMM kernel that is used to guard against compilation on
// architectures that will never use the kernel. The purpose of this is to // architectures that will never use the kernel. The purpose of this is to
// reduce the size of the compiled binary. // reduce the size of the compiled binary.
...@@ -64,6 +57,24 @@ struct enable_sm90_or_later : Kernel { ...@@ -64,6 +57,24 @@ struct enable_sm90_or_later : Kernel {
} }
}; };
/*
* This class provides the common ScaleA and ScaleB descriptors for the
* ScaledEpilogue and ScaledEpilogueBias classes.
*/
template <typename ElementAcc, typename ElementD, typename EpilogueDescriptor>
struct ScaledEpilogueBase {
protected:
using Accum = cutlass::epilogue::fusion::Sm90AccFetch;
using ScaleA = cutlass::epilogue::fusion::Sm90ColOrScalarBroadcast<
0 /*Stages*/, typename EpilogueDescriptor::TileShape, float,
Stride<Int<1>, Int<0>, Int<0>>>;
using ScaleB = cutlass::epilogue::fusion::Sm90RowOrScalarBroadcast<
0 /*Stages*/, typename EpilogueDescriptor::TileShape, float,
Stride<Int<0>, Int<1>, Int<0>>>;
};
/* /*
This epilogue function defines a quantized GEMM operation similar to This epilogue function defines a quantized GEMM operation similar to
torch.scaled_mm_. torch.scaled_mm_.
...@@ -81,21 +92,13 @@ struct enable_sm90_or_later : Kernel { ...@@ -81,21 +92,13 @@ struct enable_sm90_or_later : Kernel {
per row or column. per row or column.
*/ */
template <typename ElementAcc, typename ElementD, typename EpilogueDescriptor> template <typename ElementAcc, typename ElementD, typename EpilogueDescriptor>
struct ScaledEpilogue { struct ScaledEpilogue
: private ScaledEpilogueBase<ElementAcc, ElementD, EpilogueDescriptor> {
private: private:
using Accum = cutlass::epilogue::fusion::Sm90AccFetch; using SUPER = ScaledEpilogueBase<ElementAcc, ElementD, EpilogueDescriptor>;
using Accum = typename SUPER::Accum;
using ScaleA = cutlass::epilogue::fusion::Sm90ColOrScalarBroadcast< using ScaleA = typename SUPER::ScaleA;
0 /*Stages*/, typename EpilogueDescriptor::TileShape, float, using ScaleB = typename SUPER::ScaleB;
Stride<Int<1>, Int<0>, Int<0>>>;
using ScaleBDescriptor =
cutlass::epilogue::collective::detail::RowBroadcastDescriptor<
EpilogueDescriptor, float>;
using ScaleB = cutlass::epilogue::fusion::Sm90RowOrScalarBroadcast<
ScaleBDescriptor::Stages, typename EpilogueDescriptor::TileShape,
typename ScaleBDescriptor::Element, Stride<Int<0>, Int<1>, Int<0>>>;
using Compute0 = cutlass::epilogue::fusion::Sm90Compute< using Compute0 = cutlass::epilogue::fusion::Sm90Compute<
cutlass::multiplies, float, float, cutlass::multiplies, float, float,
...@@ -125,6 +128,50 @@ struct ScaledEpilogue { ...@@ -125,6 +128,50 @@ struct ScaledEpilogue {
} }
}; };
template <typename ElementAcc, typename ElementD, typename EpilogueDescriptor>
struct ScaledEpilogueBias
: private ScaledEpilogueBase<ElementAcc, ElementD, EpilogueDescriptor> {
private:
using SUPER = ScaledEpilogueBase<ElementAcc, ElementD, EpilogueDescriptor>;
using Accum = typename SUPER::Accum;
using ScaleA = typename SUPER::ScaleA;
using ScaleB = typename SUPER::ScaleB;
using Compute0 = cutlass::epilogue::fusion::Sm90Compute<
cutlass::multiplies, float, float,
cutlass::FloatRoundStyle::round_to_nearest>;
using EVTCompute0 =
cutlass::epilogue::fusion::Sm90EVT<Compute0, ScaleB, Accum>;
using Compute1 = cutlass::epilogue::fusion::Sm90Compute<
cutlass::multiply_add, ElementD, float,
cutlass::FloatRoundStyle::round_to_nearest>;
using Bias = cutlass::epilogue::fusion::Sm90RowBroadcast<
0 /*Stages*/, typename EpilogueDescriptor::TileShape, ElementD,
Stride<Int<0>, Int<1>, Int<0>>, 128 / sizeof_bits_v<ElementD>, false>;
public:
using EVTCompute =
cutlass::epilogue::fusion::Sm90EVT<Compute1, ScaleA, EVTCompute0, Bias>;
using ArgumentType = typename EVTCompute::Arguments;
static ArgumentType prepare_args(torch::Tensor const& a_scales,
torch::Tensor const& b_scales,
torch::Tensor const& bias) {
using ScaleA_Args = typename ScaleA::Arguments;
using ScaleB_Args = typename ScaleB::Arguments;
using Bias_Args = typename Bias::Arguments;
ScaleA_Args a_args{a_scales.data_ptr<float>(), a_scales.numel() != 1, {}};
ScaleB_Args b_args{b_scales.data_ptr<float>(), b_scales.numel() != 1, {}};
Bias_Args bias_args{static_cast<ElementD*>(bias.data_ptr())};
return ArgumentType{a_args, {b_args}, bias_args};
}
};
template <typename ElementAB_, typename ElementD_, template <typename ElementAB_, typename ElementD_,
template <typename, typename, typename> typename Epilogue_, template <typename, typename, typename> typename Epilogue_,
typename TileShape, typename ClusterShape, typename KernelSchedule, typename TileShape, typename ClusterShape, typename KernelSchedule,
...@@ -194,12 +241,12 @@ void cutlass_gemm_caller(torch::Tensor& out, torch::Tensor const& a, ...@@ -194,12 +241,12 @@ void cutlass_gemm_caller(torch::Tensor& out, torch::Tensor const& a,
int64_t ldb = b.stride(1); int64_t ldb = b.stride(1);
int64_t ldc = out.stride(0); int64_t ldc = out.stride(0);
using StrideA = Stride<int64_t, Int<1>, Int<0>>; using StrideA = Stride<int64_t, Int<1>, int64_t>;
using StrideB = Stride<int64_t, Int<1>, Int<0>>; using StrideB = Stride<int64_t, Int<1>, int64_t>;
using StrideC = typename Gemm::StrideC; using StrideC = typename Gemm::StrideC;
StrideA a_stride{lda, Int<1>{}, Int<0>{}}; StrideA a_stride{lda, Int<1>{}, 0};
StrideB b_stride{ldb, Int<1>{}, Int<0>{}}; StrideB b_stride{ldb, Int<1>{}, 0};
StrideC c_stride{ldc, Int<1>{}, Int<0>{}}; StrideC c_stride{ldc, Int<1>{}, Int<0>{}};
using GemmKernel = typename Gemm::GemmKernel; using GemmKernel = typename Gemm::GemmKernel;
...@@ -225,24 +272,26 @@ void cutlass_gemm_caller(torch::Tensor& out, torch::Tensor const& a, ...@@ -225,24 +272,26 @@ void cutlass_gemm_caller(torch::Tensor& out, torch::Tensor const& a,
CUTLASS_CHECK(gemm_op.can_implement(args)); CUTLASS_CHECK(gemm_op.can_implement(args));
size_t workspace_size = gemm_op.get_workspace_size(args); size_t workspace_size = gemm_op.get_workspace_size(args);
cutlass::device_memory::allocation<uint8_t> workspace(workspace_size); auto const workspace_options =
torch::TensorOptions().dtype(torch::kUInt8).device(a.device());
auto workspace = torch::empty(workspace_size, workspace_options);
auto stream = at::cuda::getCurrentCUDAStream(a.get_device()); auto stream = at::cuda::getCurrentCUDAStream(a.get_device());
cutlass::Status status = gemm_op.run(args, workspace.get(), stream); cutlass::Status status = gemm_op.run(args, workspace.data_ptr(), stream);
CUTLASS_CHECK(status); CUTLASS_CHECK(status);
} }
template <typename InType, typename OutType, template <typename InType, typename OutType,
template <typename, typename, typename> typename Epilogue, int32_t M> template <typename, typename, typename> typename Epilogue>
struct sm90_fp8_config { struct sm90_fp8_config_default {
// M in (128, inf)
static_assert(std::is_same<InType, cutlass::float_e4m3_t>()); static_assert(std::is_same<InType, cutlass::float_e4m3_t>());
using KernelSchedule = using KernelSchedule =
cutlass::gemm::KernelTmaWarpSpecializedPingpongFP8FastAccum; cutlass::gemm::KernelTmaWarpSpecializedPingpongFP8FastAccum;
using EpilogueSchedule = typename cutlass::epilogue::TmaWarpSpecialized; using EpilogueSchedule = typename cutlass::epilogue::TmaWarpSpecialized;
using TileShape = Shape<_128, _128, _128>; using TileShape = Shape<_128, _128, _128>;
using ClusterShape = Shape<_2, _1, _1>; using ClusterShape = Shape<_2, _1, _1>;
using Cutlass3xGemm = using Cutlass3xGemm =
cutlass_3x_gemm<InType, OutType, Epilogue, TileShape, ClusterShape, cutlass_3x_gemm<InType, OutType, Epilogue, TileShape, ClusterShape,
KernelSchedule, EpilogueSchedule>; KernelSchedule, EpilogueSchedule>;
...@@ -250,14 +299,14 @@ struct sm90_fp8_config { ...@@ -250,14 +299,14 @@ struct sm90_fp8_config {
template <typename InType, typename OutType, template <typename InType, typename OutType,
template <typename, typename, typename> typename Epilogue> template <typename, typename, typename> typename Epilogue>
struct sm90_fp8_config<InType, OutType, Epilogue, 128> { struct sm90_fp8_config_M128 {
// M in (64, 128]
static_assert(std::is_same<InType, cutlass::float_e4m3_t>()); static_assert(std::is_same<InType, cutlass::float_e4m3_t>());
using KernelSchedule = using KernelSchedule =
cutlass::gemm::KernelTmaWarpSpecializedPingpongFP8FastAccum; cutlass::gemm::KernelTmaWarpSpecializedPingpongFP8FastAccum;
using EpilogueSchedule = typename cutlass::epilogue::TmaWarpSpecialized; using EpilogueSchedule = typename cutlass::epilogue::TmaWarpSpecialized;
using TileShape = Shape<_64, _128, _128>; using TileShape = Shape<_64, _128, _128>;
using ClusterShape = Shape<_2, _1, _1>; using ClusterShape = Shape<_2, _1, _1>;
using Cutlass3xGemm = using Cutlass3xGemm =
cutlass_3x_gemm<InType, OutType, Epilogue, TileShape, ClusterShape, cutlass_3x_gemm<InType, OutType, Epilogue, TileShape, ClusterShape,
KernelSchedule, EpilogueSchedule>; KernelSchedule, EpilogueSchedule>;
...@@ -265,7 +314,8 @@ struct sm90_fp8_config<InType, OutType, Epilogue, 128> { ...@@ -265,7 +314,8 @@ struct sm90_fp8_config<InType, OutType, Epilogue, 128> {
template <typename InType, typename OutType, template <typename InType, typename OutType,
template <typename, typename, typename> typename Epilogue> template <typename, typename, typename> typename Epilogue>
struct sm90_fp8_config<InType, OutType, Epilogue, 64> { struct sm90_fp8_config_M64 {
// M in [1, 64]
static_assert(std::is_same<InType, cutlass::float_e4m3_t>()); static_assert(std::is_same<InType, cutlass::float_e4m3_t>());
using KernelSchedule = using KernelSchedule =
cutlass::gemm::KernelTmaWarpSpecializedPingpongFP8FastAccum; cutlass::gemm::KernelTmaWarpSpecializedPingpongFP8FastAccum;
...@@ -278,6 +328,78 @@ struct sm90_fp8_config<InType, OutType, Epilogue, 64> { ...@@ -278,6 +328,78 @@ struct sm90_fp8_config<InType, OutType, Epilogue, 64> {
KernelSchedule, EpilogueSchedule>; KernelSchedule, EpilogueSchedule>;
}; };
template <typename InType, typename OutType,
template <typename, typename, typename> typename Epilogue>
struct sm90_int8_config_default {
// For M > 128 and any N
static_assert(std::is_same<InType, int8_t>());
using KernelSchedule =
typename cutlass::gemm::KernelTmaWarpSpecializedPingpong;
using EpilogueSchedule = typename cutlass::epilogue::TmaWarpSpecialized;
using TileShape = Shape<_128, _128, _128>;
using ClusterShape = Shape<_2, _1, _1>;
using Cutlass3xGemm =
cutlass_3x_gemm<InType, OutType, Epilogue, TileShape, ClusterShape,
KernelSchedule, EpilogueSchedule>;
};
template <typename InType, typename OutType,
template <typename, typename, typename> typename Epilogue>
struct sm90_int8_config_M128 {
// For M in (64, 128] and any N
static_assert(std::is_same<InType, int8_t>());
using KernelSchedule =
typename cutlass::gemm::KernelTmaWarpSpecializedPingpong;
using EpilogueSchedule = typename cutlass::epilogue::TmaWarpSpecialized;
using TileShape = Shape<_64, _128, _128>;
using ClusterShape = Shape<_2, _1, _1>;
using Cutlass3xGemm =
cutlass_3x_gemm<InType, OutType, Epilogue, TileShape, ClusterShape,
KernelSchedule, EpilogueSchedule>;
};
template <typename InType, typename OutType,
template <typename, typename, typename> typename Epilogue>
struct sm90_int8_config_M64 {
// For M in (32, 64] and any N
static_assert(std::is_same<InType, int8_t>());
using KernelSchedule = typename cutlass::gemm::KernelTmaWarpSpecialized;
using EpilogueSchedule = typename cutlass::epilogue::TmaWarpSpecialized;
using TileShape = Shape<_64, _64, _256>;
using ClusterShape = Shape<_1, _1, _1>;
using Cutlass3xGemm =
cutlass_3x_gemm<InType, OutType, Epilogue, TileShape, ClusterShape,
KernelSchedule, EpilogueSchedule>;
};
template <typename InType, typename OutType,
template <typename, typename, typename> typename Epilogue>
struct sm90_int8_config_M32_NBig {
// For M in [1, 32] and N >= 8192
static_assert(std::is_same<InType, int8_t>());
using KernelSchedule = typename cutlass::gemm::KernelTmaWarpSpecialized;
using EpilogueSchedule = typename cutlass::epilogue::TmaWarpSpecialized;
using TileShape = Shape<_64, _128, _256>;
using ClusterShape = Shape<_1, _4, _1>;
using Cutlass3xGemm =
cutlass_3x_gemm<InType, OutType, Epilogue, TileShape, ClusterShape,
KernelSchedule, EpilogueSchedule>;
};
template <typename InType, typename OutType,
template <typename, typename, typename> typename Epilogue>
struct sm90_int8_config_M32_NSmall {
// For M in [1, 32] and N < 8192
static_assert(std::is_same<InType, int8_t>());
using KernelSchedule = typename cutlass::gemm::KernelTmaWarpSpecialized;
using EpilogueSchedule = typename cutlass::epilogue::TmaWarpSpecialized;
using TileShape = Shape<_64, _64, _256>;
using ClusterShape = Shape<_1, _8, _1>;
using Cutlass3xGemm =
cutlass_3x_gemm<InType, OutType, Epilogue, TileShape, ClusterShape,
KernelSchedule, EpilogueSchedule>;
};
} // namespace } // namespace
template <typename InType, typename OutType, template <typename InType, typename OutType,
...@@ -291,11 +413,12 @@ void cutlass_gemm_sm90_fp8_dispatch(torch::Tensor& out, torch::Tensor const& a, ...@@ -291,11 +413,12 @@ void cutlass_gemm_sm90_fp8_dispatch(torch::Tensor& out, torch::Tensor const& a,
TORCH_CHECK(b.dtype() == torch::kFloat8_e4m3fn); TORCH_CHECK(b.dtype() == torch::kFloat8_e4m3fn);
using Cutlass3xGemmDefault = using Cutlass3xGemmDefault =
typename sm90_fp8_config<InType, OutType, Epilogue, 0>::Cutlass3xGemm; typename sm90_fp8_config_default<InType, OutType,
Epilogue>::Cutlass3xGemm;
using Cutlass3xGemmM64 = using Cutlass3xGemmM64 =
typename sm90_fp8_config<InType, OutType, Epilogue, 64>::Cutlass3xGemm; typename sm90_fp8_config_M64<InType, OutType, Epilogue>::Cutlass3xGemm;
using Cutlass3xGemmM128 = using Cutlass3xGemmM128 =
typename sm90_fp8_config<InType, OutType, Epilogue, 128>::Cutlass3xGemm; typename sm90_fp8_config_M128<InType, OutType, Epilogue>::Cutlass3xGemm;
uint32_t const m = a.size(0); uint32_t const m = a.size(0);
uint32_t const mp2 = uint32_t const mp2 =
...@@ -316,49 +439,111 @@ void cutlass_gemm_sm90_fp8_dispatch(torch::Tensor& out, torch::Tensor const& a, ...@@ -316,49 +439,111 @@ void cutlass_gemm_sm90_fp8_dispatch(torch::Tensor& out, torch::Tensor const& a,
} }
} }
void cutlass_scaled_mm_sm90(torch::Tensor& out, torch::Tensor const& a, template <typename InType, typename OutType,
torch::Tensor const& b, template <typename, typename, typename> typename Epilogue,
torch::Tensor const& a_scales, typename... EpilogueArgs>
torch::Tensor const& b_scales) { void cutlass_gemm_sm90_int8_dispatch(torch::Tensor& out, torch::Tensor const& a,
TORCH_CHECK(a_scales.dtype() == torch::kFloat32); torch::Tensor const& b,
TORCH_CHECK(b_scales.dtype() == torch::kFloat32); EpilogueArgs&&... args) {
static_assert(std::is_same<InType, int8_t>());
TORCH_CHECK(a.dtype() == torch::kInt8);
TORCH_CHECK(b.dtype() == torch::kInt8);
using Cutlass3xGemmDefault =
typename sm90_int8_config_default<InType, OutType,
Epilogue>::Cutlass3xGemm;
using Cutlass3xGemmM128 =
typename sm90_int8_config_M128<InType, OutType, Epilogue>::Cutlass3xGemm;
using Cutlass3xGemmM64 =
typename sm90_int8_config_M64<InType, OutType, Epilogue>::Cutlass3xGemm;
using Cutlass3xGemmM32NBig =
typename sm90_int8_config_M32_NBig<InType, OutType,
Epilogue>::Cutlass3xGemm;
using Cutlass3xGemmM32NSmall =
typename sm90_int8_config_M32_NSmall<InType, OutType,
Epilogue>::Cutlass3xGemm;
uint32_t const n = out.size(1);
bool const is_small_n = n < 8192;
uint32_t const m = a.size(0);
uint32_t const mp2 =
std::max(static_cast<uint32_t>(32), next_pow_2(m)); // next power of 2
if (mp2 <= 32) {
// m in [1, 32]
if (is_small_n) {
return cutlass_gemm_caller<Cutlass3xGemmM32NSmall>(
out, a, b, std::forward<EpilogueArgs>(args)...);
} else {
return cutlass_gemm_caller<Cutlass3xGemmM32NBig>(
out, a, b, std::forward<EpilogueArgs>(args)...);
}
} else if (mp2 <= 64) {
// m in (32, 64]
return cutlass_gemm_caller<Cutlass3xGemmM64>(
out, a, b, std::forward<EpilogueArgs>(args)...);
} else if (mp2 <= 128) {
// m in (64, 128]
return cutlass_gemm_caller<Cutlass3xGemmM128>(
out, a, b, std::forward<EpilogueArgs>(args)...);
} else {
// m in (128, inf)
return cutlass_gemm_caller<Cutlass3xGemmDefault>(
out, a, b, std::forward<EpilogueArgs>(args)...);
}
}
template <template <typename, typename, typename> typename Epilogue,
typename... EpilogueArgs>
void cutlass_scaled_mm_sm90_epilogue(torch::Tensor& out, torch::Tensor const& a,
torch::Tensor const& b,
EpilogueArgs&&... epilogue_args) {
if (a.dtype() == torch::kInt8) { if (a.dtype() == torch::kInt8) {
TORCH_CHECK(b.dtype() == torch::kInt8); TORCH_CHECK(b.dtype() == torch::kInt8);
using TileShape = Shape<_128, _128, _128>;
using ClusterShape = Shape<_1, _2, _1>;
using KernelSchedule =
typename cutlass::gemm::KernelTmaWarpSpecializedPingpong;
using EpilogueSchedule = typename cutlass::epilogue::TmaWarpSpecialized;
if (out.dtype() == torch::kBFloat16) { if (out.dtype() == torch::kBFloat16) {
return cutlass_gemm_caller<cutlass_3x_gemm< return cutlass_gemm_sm90_int8_dispatch<int8_t, cutlass::bfloat16_t,
int8_t, cutlass::bfloat16_t, ScaledEpilogue, TileShape, ClusterShape, Epilogue>(
KernelSchedule, EpilogueSchedule>>(out, a, b, a_scales, b_scales); out, a, b, std::forward<EpilogueArgs>(epilogue_args)...);
} else { } else {
TORCH_CHECK(out.dtype() == torch::kFloat16); TORCH_CHECK(out.dtype() == torch::kFloat16);
return cutlass_gemm_sm90_int8_dispatch<int8_t, cutlass::half_t, Epilogue>(
return cutlass_gemm_caller< out, a, b, std::forward<EpilogueArgs>(epilogue_args)...);
cutlass_3x_gemm<int8_t, cutlass::half_t, ScaledEpilogue, TileShape,
ClusterShape, KernelSchedule, EpilogueSchedule>>(
out, a, b, a_scales, b_scales);
} }
} else { } else {
TORCH_CHECK(a.dtype() == torch::kFloat8_e4m3fn); TORCH_CHECK(a.dtype() == torch::kFloat8_e4m3fn);
TORCH_CHECK(b.dtype() == torch::kFloat8_e4m3fn); TORCH_CHECK(b.dtype() == torch::kFloat8_e4m3fn);
if (out.dtype() == torch::kBFloat16) { if (out.dtype() == torch::kBFloat16) {
return cutlass_gemm_sm90_fp8_dispatch< return cutlass_gemm_sm90_fp8_dispatch<cutlass::float_e4m3_t,
cutlass::float_e4m3_t, cutlass::bfloat16_t, ScaledEpilogue>( cutlass::bfloat16_t, Epilogue>(
out, a, b, a_scales, b_scales); out, a, b, std::forward<EpilogueArgs>(epilogue_args)...);
} else { } else {
TORCH_CHECK(out.dtype() == torch::kFloat16); TORCH_CHECK(out.dtype() == torch::kFloat16);
return cutlass_gemm_sm90_fp8_dispatch<cutlass::float_e4m3_t, return cutlass_gemm_sm90_fp8_dispatch<cutlass::float_e4m3_t,
cutlass::half_t, ScaledEpilogue>( cutlass::half_t, Epilogue>(
out, a, b, a_scales, b_scales); out, a, b, std::forward<EpilogueArgs>(epilogue_args)...);
} }
} }
} }
void cutlass_scaled_mm_sm90(torch::Tensor& c, torch::Tensor const& a,
torch::Tensor const& b,
torch::Tensor const& a_scales,
torch::Tensor const& b_scales,
c10::optional<torch::Tensor> const& bias) {
TORCH_CHECK(a_scales.dtype() == torch::kFloat32);
TORCH_CHECK(b_scales.dtype() == torch::kFloat32);
if (bias) {
TORCH_CHECK(bias->dtype() == c.dtype(),
"currently bias dtype must match output dtype ", c.dtype());
return cutlass_scaled_mm_sm90_epilogue<ScaledEpilogueBias>(
c, a, b, a_scales, b_scales, *bias);
} else {
return cutlass_scaled_mm_sm90_epilogue<ScaledEpilogue>(c, a, b, a_scales,
b_scales);
}
}
#endif #endif
...@@ -6,28 +6,49 @@ ...@@ -6,28 +6,49 @@
void cutlass_scaled_mm_sm75(torch::Tensor& c, torch::Tensor const& a, void cutlass_scaled_mm_sm75(torch::Tensor& c, torch::Tensor const& a,
torch::Tensor const& b, torch::Tensor const& b,
torch::Tensor const& a_scales, torch::Tensor const& a_scales,
torch::Tensor const& b_scales); torch::Tensor const& b_scales,
c10::optional<torch::Tensor> const& bias);
void cutlass_scaled_mm_sm80(torch::Tensor& c, torch::Tensor const& a, void cutlass_scaled_mm_sm80(torch::Tensor& c, torch::Tensor const& a,
torch::Tensor const& b, torch::Tensor const& b,
torch::Tensor const& a_scales, torch::Tensor const& a_scales,
torch::Tensor const& b_scales); torch::Tensor const& b_scales,
c10::optional<torch::Tensor> const& bias);
void cutlass_scaled_mm_sm89(torch::Tensor& c, torch::Tensor const& a, void cutlass_scaled_mm_sm89(torch::Tensor& c, torch::Tensor const& a,
torch::Tensor const& b, torch::Tensor const& b,
torch::Tensor const& a_scales, torch::Tensor const& a_scales,
torch::Tensor const& b_scales); torch::Tensor const& b_scales,
c10::optional<torch::Tensor> const& bias);
#if defined CUDA_VERSION && CUDA_VERSION >= 12000 #if defined CUDA_VERSION && CUDA_VERSION >= 12000
void cutlass_scaled_mm_sm90(torch::Tensor& c, torch::Tensor const& a, void cutlass_scaled_mm_sm90(torch::Tensor& c, torch::Tensor const& a,
torch::Tensor const& b, torch::Tensor const& b,
torch::Tensor const& a_scales, torch::Tensor const& a_scales,
torch::Tensor const& b_scales); torch::Tensor const& b_scales,
c10::optional<torch::Tensor> const& bias);
#endif #endif
bool cutlass_scaled_mm_supports_fp8(int64_t cuda_device_capability) {
// CUTLASS FP8 kernels need at least
// CUDA 12.0 on SM90 systems (Hopper)
// CUDA 12.4 on SM89 systems (Lovelace)
#if defined CUDA_VERSION
if (cuda_device_capability >= 90) {
return CUDA_VERSION >= 12000;
} else if (cuda_device_capability >= 89) {
return CUDA_VERSION >= 12040;
}
#endif
return false;
}
void cutlass_scaled_mm(torch::Tensor& c, torch::Tensor const& a, void cutlass_scaled_mm(torch::Tensor& c, torch::Tensor const& a,
torch::Tensor const& b, torch::Tensor const& a_scales, torch::Tensor const& b, torch::Tensor const& a_scales,
torch::Tensor const& b_scales) { torch::Tensor const& b_scales,
c10::optional<torch::Tensor> const& bias) {
int32_t major_capability; int32_t major_capability;
int32_t minor_capability; int32_t minor_capability;
cudaDeviceGetAttribute(&major_capability, cudaDevAttrComputeCapabilityMajor, cudaDeviceGetAttribute(&major_capability, cudaDevAttrComputeCapabilityMajor,
...@@ -50,6 +71,11 @@ void cutlass_scaled_mm(torch::Tensor& c, torch::Tensor const& a, ...@@ -50,6 +71,11 @@ void cutlass_scaled_mm(torch::Tensor& c, torch::Tensor const& a,
b.stride(1) % 16 == 0); // 16 Byte Alignment b.stride(1) % 16 == 0); // 16 Byte Alignment
TORCH_CHECK(a_scales.is_contiguous() && b_scales.is_contiguous()); TORCH_CHECK(a_scales.is_contiguous() && b_scales.is_contiguous());
if (bias) {
TORCH_CHECK(bias->numel() == b.size(1) && bias->is_contiguous() &&
bias->dim() == 1);
}
at::cuda::OptionalCUDAGuard const device_guard(device_of(a)); at::cuda::OptionalCUDAGuard const device_guard(device_of(a));
if (version_num >= 90) { if (version_num >= 90) {
...@@ -57,19 +83,19 @@ void cutlass_scaled_mm(torch::Tensor& c, torch::Tensor const& a, ...@@ -57,19 +83,19 @@ void cutlass_scaled_mm(torch::Tensor& c, torch::Tensor const& a,
// Guard against compilation issues for sm90 kernels // Guard against compilation issues for sm90 kernels
#if defined CUDA_VERSION && CUDA_VERSION >= 12000 #if defined CUDA_VERSION && CUDA_VERSION >= 12000
cutlass_scaled_mm_sm90(c, a, b, a_scales, b_scales); cutlass_scaled_mm_sm90(c, a, b, a_scales, b_scales, bias);
#else #else
cutlass_scaled_mm_sm80(c, a, b, a_scales, b_scales); cutlass_scaled_mm_sm80(c, a, b, a_scales, b_scales, bias);
#endif #endif
} else if (version_num == 89) { } else if (version_num == 89) {
// Ada Lovelace // Ada Lovelace
cutlass_scaled_mm_sm89(c, a, b, a_scales, b_scales); cutlass_scaled_mm_sm89(c, a, b, a_scales, b_scales, bias);
} else if (version_num >= 80) { } else if (version_num >= 80) {
// Ampere // Ampere
cutlass_scaled_mm_sm80(c, a, b, a_scales, b_scales); cutlass_scaled_mm_sm80(c, a, b, a_scales, b_scales, bias);
} else { } else {
// Turing // Turing
TORCH_CHECK(version_num >= 75); TORCH_CHECK(version_num >= 75);
cutlass_scaled_mm_sm75(c, a, b, a_scales, b_scales); cutlass_scaled_mm_sm75(c, a, b, a_scales, b_scales, bias);
} }
} }
...@@ -526,6 +526,7 @@ __inline__ __device__ Tout convert(const Tin& x) { ...@@ -526,6 +526,7 @@ __inline__ __device__ Tout convert(const Tin& x) {
} }
#endif #endif
assert(false); assert(false);
return {}; // Squash missing return statement warning
} }
template <typename Tout, typename Tin, Fp8KVCacheDataType kv_dt> template <typename Tout, typename Tin, Fp8KVCacheDataType kv_dt>
...@@ -536,6 +537,7 @@ __inline__ __device__ Tout scaled_convert(const Tin& x, const float scale) { ...@@ -536,6 +537,7 @@ __inline__ __device__ Tout scaled_convert(const Tin& x, const float scale) {
} }
#endif #endif
assert(false); assert(false);
return {}; // Squash missing return statement warning
} }
// The following macro is used to dispatch the conversion function based on // The following macro is used to dispatch the conversion function based on
......
...@@ -7,6 +7,8 @@ ...@@ -7,6 +7,8 @@
#include "cuda_compat.h" #include "cuda_compat.h"
#include "dispatch_utils.h" #include "dispatch_utils.h"
#include "../../reduction_utils.cuh"
namespace vllm { namespace vllm {
__device__ __forceinline__ float atomicMaxFloat(float* addr, float value) { __device__ __forceinline__ float atomicMaxFloat(float* addr, float value) {
...@@ -21,10 +23,16 @@ __device__ __forceinline__ float atomicMaxFloat(float* addr, float value) { ...@@ -21,10 +23,16 @@ __device__ __forceinline__ float atomicMaxFloat(float* addr, float value) {
#define FP8_E4M3_MAX std::numeric_limits<c10::Float8_e4m3fn>::max() #define FP8_E4M3_MAX std::numeric_limits<c10::Float8_e4m3fn>::max()
template <typename scalar_t> template <bool is_scale_inverted>
__device__ __forceinline__ c10::Float8_e4m3fn scaled_fp8_conversion( __device__ __forceinline__ c10::Float8_e4m3fn scaled_fp8_conversion(
const scalar_t val, const float inverted_scale) { float const val, float const scale) {
float x = static_cast<float>(val) * inverted_scale; float x = 0.0f;
if constexpr (is_scale_inverted) {
x = val * scale;
} else {
x = val / scale;
}
float r = fmax(-FP8_E4M3_MAX, fmin(x, FP8_E4M3_MAX)); float r = fmax(-FP8_E4M3_MAX, fmin(x, FP8_E4M3_MAX));
return static_cast<c10::Float8_e4m3fn>(r); return static_cast<c10::Float8_e4m3fn>(r);
} }
...@@ -40,7 +48,7 @@ __global__ void segmented_max_reduction(float* __restrict__ scale, ...@@ -40,7 +48,7 @@ __global__ void segmented_max_reduction(float* __restrict__ scale,
const scalar_t* __restrict__ input, const scalar_t* __restrict__ input,
int64_t num_elems) { int64_t num_elems) {
__shared__ float cache[1024]; __shared__ float cache[1024];
int i = blockDim.x * blockIdx.x + threadIdx.x; int64_t i = blockDim.x * blockIdx.x + threadIdx.x;
// First store maximum for all values processes by // First store maximum for all values processes by
// the current thread in cache[threadIdx.x] // the current thread in cache[threadIdx.x]
...@@ -87,6 +95,70 @@ typedef struct __align__(4) { ...@@ -87,6 +95,70 @@ typedef struct __align__(4) {
} }
float8x4_t; float8x4_t;
template <typename scalar_t>
__device__ float thread_max_vec(scalar_t const* __restrict__ input,
int64_t const num_elems, int const tid,
int const step) {
// Vectorized input/output to better utilize memory bandwidth.
vec4_t<scalar_t> const* vectorized_in =
reinterpret_cast<vec4_t<scalar_t> const*>(input);
int64_t const num_vec_elems = num_elems >> 2;
float absmax_val = 0.0f;
#pragma unroll 4
for (int64_t i = tid; i < num_vec_elems; i += step) {
vec4_t<scalar_t> in_vec = vectorized_in[i];
absmax_val = max(absmax_val, fabs(in_vec.x));
absmax_val = max(absmax_val, fabs(in_vec.y));
absmax_val = max(absmax_val, fabs(in_vec.z));
absmax_val = max(absmax_val, fabs(in_vec.w));
}
// Handle the remaining elements if num_elems is not divisible by 4
for (int64_t i = num_vec_elems * 4 + tid; i < num_elems; i += step) {
absmax_val = max(absmax_val, fabs(input[i]));
}
return absmax_val;
}
template <typename scalar_t, bool is_scale_inverted>
__device__ void scaled_fp8_conversion_vec(c10::Float8_e4m3fn* __restrict__ out,
scalar_t const* __restrict__ input,
float const scale,
int64_t const num_elems,
int const tid, int const step) {
// Vectorized input/output to better utilize memory bandwidth.
vec4_t<scalar_t> const* vectorized_in =
reinterpret_cast<vec4_t<scalar_t> const*>(input);
float8x4_t* vectorized_out = reinterpret_cast<float8x4_t*>(out);
int64_t const num_vec_elems = num_elems >> 2;
#pragma unroll 4
for (int64_t i = tid; i < num_vec_elems; i += step) {
vec4_t<scalar_t> in_vec = vectorized_in[i];
float8x4_t out_vec;
out_vec.x = scaled_fp8_conversion<is_scale_inverted>(
static_cast<float>(in_vec.x), scale);
out_vec.y = scaled_fp8_conversion<is_scale_inverted>(
static_cast<float>(in_vec.y), scale);
out_vec.z = scaled_fp8_conversion<is_scale_inverted>(
static_cast<float>(in_vec.z), scale);
out_vec.w = scaled_fp8_conversion<is_scale_inverted>(
static_cast<float>(in_vec.w), scale);
vectorized_out[i] = out_vec;
}
// Handle the remaining elements if num_elems is not divisible by 4
for (int64_t i = num_vec_elems * 4 + tid; i < num_elems; i += step) {
out[i] = scaled_fp8_conversion<is_scale_inverted>(
static_cast<float>(input[i]), scale);
}
}
template <typename scalar_t> template <typename scalar_t>
__global__ void scaled_fp8_quant_kernel(c10::Float8_e4m3fn* __restrict__ out, __global__ void scaled_fp8_quant_kernel(c10::Float8_e4m3fn* __restrict__ out,
const scalar_t* __restrict__ input, const scalar_t* __restrict__ input,
...@@ -97,38 +169,68 @@ __global__ void scaled_fp8_quant_kernel(c10::Float8_e4m3fn* __restrict__ out, ...@@ -97,38 +169,68 @@ __global__ void scaled_fp8_quant_kernel(c10::Float8_e4m3fn* __restrict__ out,
// Invert the scale so that we can use multiplications to avoid expensive // Invert the scale so that we can use multiplications to avoid expensive
// division. // division.
const float inverted_scale = 1.0f / (*scale); const float inverted_scale = 1.0f / (*scale);
scaled_fp8_conversion_vec<scalar_t, true>(
out, input, inverted_scale, num_elems, tid, blockDim.x * gridDim.x);
}
// Vectorized input/output to better utilize memory bandwidth. template <typename scalar_t>
const vec4_t<scalar_t>* vectorized_in = __global__ void dynamic_per_token_scaled_fp8_quant_kernel(
reinterpret_cast<const vec4_t<scalar_t>*>(input); c10::Float8_e4m3fn* __restrict__ out, float* __restrict__ scale,
float8x4_t* vectorized_out = reinterpret_cast<float8x4_t*>(out); scalar_t const* __restrict__ input, float const* __restrict__ scale_ub,
const int hidden_size) {
float const min_scaling_factor = 1.0f / (FP8_E4M3_MAX * 512.f);
int num_vec_elems = num_elems >> 2; int const tid = threadIdx.x;
int const token_idx = blockIdx.x;
#pragma unroll 4 scalar_t const* __restrict__ token_input = &input[token_idx * hidden_size];
for (int i = tid; i < num_vec_elems; i += blockDim.x * gridDim.x) { c10::Float8_e4m3fn* __restrict__ token_output = &out[token_idx * hidden_size];
vec4_t<scalar_t> in_vec = vectorized_in[i];
float8x4_t out_vec;
out_vec.x = scaled_fp8_conversion(in_vec.x, inverted_scale); // For vectorization, token_input and token_output pointers need to be
out_vec.y = scaled_fp8_conversion(in_vec.y, inverted_scale); // aligned at 8-byte and 4-byte addresses respectively.
out_vec.z = scaled_fp8_conversion(in_vec.z, inverted_scale); bool const can_vectorize = hidden_size % 4 == 0;
out_vec.w = scaled_fp8_conversion(in_vec.w, inverted_scale);
vectorized_out[i] = out_vec; float absmax_val = 0.0f;
if (can_vectorize) {
absmax_val = thread_max_vec(token_input, hidden_size, tid, blockDim.x);
} else {
for (int i = tid; i < hidden_size; i += blockDim.x) {
float const x = static_cast<float>(token_input[i]);
absmax_val = max(absmax_val, fabs(x));
}
} }
// Handle the remaining elements if num_elems is not divisible by 4 float const block_absmax_val_maybe = blockReduceMax(absmax_val);
for (int i = num_vec_elems * 4 + tid; i < num_elems; __shared__ float token_scale;
i += blockDim.x * gridDim.x) { if (tid == 0) {
out[i] = scaled_fp8_conversion(input[i], inverted_scale); if (scale_ub) {
token_scale = min(block_absmax_val_maybe, *scale_ub);
} else {
token_scale = block_absmax_val_maybe;
}
// token scale computation
token_scale = max(token_scale / FP8_E4M3_MAX, min_scaling_factor);
scale[token_idx] = token_scale;
}
__syncthreads();
// Note that we don't use inverted scales so we can match FBGemm impl.
if (can_vectorize) {
scaled_fp8_conversion_vec<scalar_t, false>(
token_output, token_input, token_scale, hidden_size, tid, blockDim.x);
} else {
for (int i = tid; i < hidden_size; i += blockDim.x) {
token_output[i] = scaled_fp8_conversion<false>(
static_cast<float>(token_input[i]), token_scale);
}
} }
} }
} // namespace vllm } // namespace vllm
void static_scaled_fp8_quant(torch::Tensor& out, // [..., d] void static_scaled_fp8_quant(torch::Tensor& out, // [..., d]
torch::Tensor& input, // [..., d] torch::Tensor const& input, // [..., d]
torch::Tensor& scale) // [1] torch::Tensor const& scale) // [1]
{ {
int64_t num_tokens = input.numel() / input.size(-1); int64_t num_tokens = input.numel() / input.size(-1);
int64_t num_elems = input.numel(); int64_t num_elems = input.numel();
...@@ -144,9 +246,9 @@ void static_scaled_fp8_quant(torch::Tensor& out, // [..., d] ...@@ -144,9 +246,9 @@ void static_scaled_fp8_quant(torch::Tensor& out, // [..., d]
}); });
} }
void dynamic_scaled_fp8_quant(torch::Tensor& out, // [..., d] void dynamic_scaled_fp8_quant(torch::Tensor& out, // [..., d]
torch::Tensor& input, // [..., d] torch::Tensor const& input, // [..., d]
torch::Tensor& scale) // [1] torch::Tensor& scale) // [1]
{ {
int64_t num_tokens = input.numel() / input.size(-1); int64_t num_tokens = input.numel() / input.size(-1);
int64_t num_elems = input.numel(); int64_t num_elems = input.numel();
...@@ -163,3 +265,28 @@ void dynamic_scaled_fp8_quant(torch::Tensor& out, // [..., d] ...@@ -163,3 +265,28 @@ void dynamic_scaled_fp8_quant(torch::Tensor& out, // [..., d]
scale.data_ptr<float>(), num_elems); scale.data_ptr<float>(), num_elems);
}); });
} }
void dynamic_per_token_scaled_fp8_quant(
torch::Tensor& out, // [..., d]
torch::Tensor const& input, // [..., d]
torch::Tensor& scales, std::optional<at::Tensor> const& scale_ub) {
TORCH_CHECK(input.is_contiguous());
TORCH_CHECK(out.is_contiguous());
int const hidden_size = input.size(-1);
int const num_tokens = input.numel() / hidden_size;
dim3 const grid(num_tokens);
dim3 const block(std::min(hidden_size, 1024));
const at::cuda::OptionalCUDAGuard device_guard(device_of(input));
const cudaStream_t stream = at::cuda::getCurrentCUDAStream();
VLLM_DISPATCH_FLOATING_TYPES(
input.scalar_type(), "dynamic_per_token_scaled_fp8_quant_kernel", [&] {
vllm::dynamic_per_token_scaled_fp8_quant_kernel<scalar_t>
<<<grid, block, 0, stream>>>(
out.data_ptr<c10::Float8_e4m3fn>(), scales.data_ptr<float>(),
input.data_ptr<scalar_t>(),
scale_ub.has_value() ? scale_ub->data_ptr<float>() : nullptr,
hidden_size);
});
}
/*
* Modified by Neural Magic
* Copyright (C) Marlin.2024 Elias Frantar
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
/*
* Adapted from https://github.com/IST-DASLab/marlin
*/
#include "../gptq_marlin/marlin.cuh"
#include "../gptq_marlin/marlin_dtypes.cuh"
using namespace marlin;
#define STATIC_ASSERT_SCALAR_TYPE_VALID(scalar_t) \
static_assert(std::is_same<scalar_t, half>::value || \
std::is_same<scalar_t, nv_bfloat16>::value, \
"only float16 and bfloat16 is supported");
template <typename T>
inline std::string str(T x) {
return std::to_string(x);
}
namespace fp8_marlin {
#if defined(__CUDA_ARCH__) && __CUDA_ARCH__ < 800
template <typename scalar_t, // compute dtype, half or nv_float16
const int num_bits, // number of bits used for weights
const int threads, // number of threads in a threadblock
const int thread_m_blocks, // number of 16x16 blocks in the m
// dimension (batchsize) of the
// threadblock
const int thread_n_blocks, // same for n dimension (output)
const int thread_k_blocks, // same for k dimension (reduction)
const int stages, // number of stages for the async global->shared
// fetch pipeline
const int group_blocks = -1 // number of consecutive 16x16 blocks
// with a separate quantization scale
>
__global__ void Marlin(
const int4* __restrict__ A, // fp16 input matrix of shape mxk
const int4* __restrict__ B, // 4bit quantized weight matrix of shape kxn
int4* __restrict__ C, // fp16 output buffer of shape mxn
const int4* __restrict__ scales_ptr, // fp16 quantization scales of shape
// (k/groupsize)xn
int num_groups, // number of scale groups per output channel
int prob_m, // batch dimension m
int prob_n, // output dimension n
int prob_k, // reduction dimension k
int* locks // extra global storage for barrier synchronization
) {}
} // namespace fp8_marlin
torch::Tensor fp8_marlin_gemm(torch::Tensor& a, torch::Tensor& b_q_weight,
torch::Tensor& b_scales, torch::Tensor& workspace,
int64_t num_bits, int64_t size_m, int64_t size_n,
int64_t size_k) {
TORCH_CHECK_NOT_IMPLEMENTED(false,
"marlin_gemm(..) requires CUDA_ARCH >= 8.0");
return torch::empty({1, 1});
}
#else
// m16n8k16 tensor core mma instruction with fp16 inputs and fp32
// output/accumulation.
template <typename scalar_t>
__device__ inline void mma(const typename ScalarType<scalar_t>::FragA& a_frag,
const typename ScalarType<scalar_t>::FragB& frag_b,
typename ScalarType<scalar_t>::FragC& frag_c) {
const uint32_t* a = reinterpret_cast<const uint32_t*>(&a_frag);
const uint32_t* b = reinterpret_cast<const uint32_t*>(&frag_b);
float* c = reinterpret_cast<float*>(&frag_c);
if constexpr (std::is_same<scalar_t, half>::value) {
asm volatile(
"mma.sync.aligned.m16n8k16.row.col.f32.f16.f16.f32 "
"{%0,%1,%2,%3}, {%4,%5,%6,%7}, {%8,%9}, {%10,%11,%12,%13};\n"
: "=f"(c[0]), "=f"(c[1]), "=f"(c[2]), "=f"(c[3])
: "r"(a[0]), "r"(a[1]), "r"(a[2]), "r"(a[3]), "r"(b[0]), "r"(b[1]),
"f"(c[0]), "f"(c[1]), "f"(c[2]), "f"(c[3]));
} else if constexpr (std::is_same<scalar_t, nv_bfloat16>::value) {
asm volatile(
"mma.sync.aligned.m16n8k16.row.col.f32.bf16.bf16.f32 "
"{%0,%1,%2,%3}, {%4,%5,%6,%7}, {%8,%9}, {%10,%11,%12,%13};\n"
: "=f"(c[0]), "=f"(c[1]), "=f"(c[2]), "=f"(c[3])
: "r"(a[0]), "r"(a[1]), "r"(a[2]), "r"(a[3]), "r"(b[0]), "r"(b[1]),
"f"(c[0]), "f"(c[1]), "f"(c[2]), "f"(c[3]));
} else {
STATIC_ASSERT_SCALAR_TYPE_VALID(scalar_t);
}
}
// Instruction for loading a full 16x16 matrix fragment of operand A from shared
// memory, directly in tensor core layout.
template <typename scalar_t>
__device__ inline void ldsm4(typename ScalarType<scalar_t>::FragA& frag_a,
const void* smem_ptr) {
uint32_t* a = reinterpret_cast<uint32_t*>(&frag_a);
uint32_t smem = static_cast<uint32_t>(__cvta_generic_to_shared(smem_ptr));
asm volatile("ldmatrix.sync.aligned.m8n8.x4.shared.b16 {%0,%1,%2,%3}, [%4];\n"
: "=r"(a[0]), "=r"(a[1]), "=r"(a[2]), "=r"(a[3])
: "r"(smem));
}
// Fast FP8ToFp16/FP8ToBf16: Efficiently dequantize 8bit fp8_e4m3 values to fp16
// bf16 Reference:
// - FP16:
// https://github.com/NVIDIA/FasterTransformer/blob/release/v5.3_tag/src/fastertransformer/cutlass_extensions/include/cutlass_extensions/interleaved_numeric_conversion.h#L53-L85
// - BF16:
// https://github.com/NVIDIA/FasterTransformer/blob/release/v5.3_tag/src/fastertransformer/cutlass_extensions/include/cutlass_extensions/interleaved_numeric_conversion.h#L125-L175
template <typename scalar_t>
__device__ inline typename ScalarType<scalar_t>::FragB dequant_8bit(int q) {
STATIC_ASSERT_SCALAR_TYPE_VALID(scalar_t);
}
template <>
__device__ inline typename ScalarType<half>::FragB dequant_8bit<half>(int q) {
// Constants for FP8 (E4M3) and FP16 formats
constexpr int FP8_EXPONENT = 4, FP8_MANTISSA = 3, FP16_EXPONENT = 5;
constexpr int RIGHT_SHIFT = FP16_EXPONENT - FP8_EXPONENT;
// Calculate MASK for extracting mantissa and exponent
constexpr int MASK1 = 0x80000000;
constexpr int MASK2 = MASK1 >> (FP8_EXPONENT + FP8_MANTISSA);
constexpr int MASK3 = MASK2 & 0x7fffffff;
constexpr int MASK = MASK3 | (MASK3 >> 16);
// Final MASK value: 0x7F007F00
// Extract and shift FP8 values to FP16 format
int Out1 = (q & 0x80008000) | ((q & MASK) >> RIGHT_SHIFT);
int Out2 = ((q << 8) & 0x80008000) | (((q << 8) & MASK) >> RIGHT_SHIFT);
// Construct and apply exponent bias
constexpr int BIAS_OFFSET =
(1 << (FP16_EXPONENT - 1)) - (1 << (FP8_EXPONENT - 1));
const half2 bias_reg = __float2half2_rn(float(1 << BIAS_OFFSET));
// Convert to half2 and apply bias
typename ScalarType<half>::FragB frag_b;
// Note: reverse indexing is intentional because weights are permuted
frag_b[1] = __hmul2(*reinterpret_cast<const half2*>(&Out1), bias_reg);
frag_b[0] = __hmul2(*reinterpret_cast<const half2*>(&Out2), bias_reg);
return frag_b;
}
template <>
__device__ inline typename ScalarType<nv_bfloat16>::FragB
dequant_8bit<nv_bfloat16>(int q) {
// Constants for FP8 (E4M3) and BF16 formats
constexpr int FP8_EXPONENT = 4, FP8_MANTISSA = 3, BF16_EXPONENT = 8;
constexpr int RIGHT_SHIFT = BF16_EXPONENT - FP8_EXPONENT;
// Calculate MASK for extracting mantissa and exponent
constexpr int MASK1 = 0x80000000;
constexpr int MASK2 = MASK1 >> (FP8_EXPONENT + FP8_MANTISSA);
constexpr int MASK3 = MASK2 & 0x7fffffff;
constexpr int MASK = MASK3 | (MASK3 >> 16);
// Final MASK value: 0x7F007F00
// Extract and shift FP8 values to BF16 format
int Out1 = (q & 0x80008000) | ((q & MASK) >> RIGHT_SHIFT);
int Out2 = ((q << 8) & 0x80008000) | (((q << 8) & MASK) >> RIGHT_SHIFT);
// Construct and apply exponent bias
constexpr int BIAS_OFFSET =
(1 << (BF16_EXPONENT - 1)) - (1 << (FP8_EXPONENT - 1));
// Add 127 (float exponent bias) to BIAS_OFFSET and shift to float exponent
// position
constexpr uint32_t BIAS = (BIAS_OFFSET + 127) << 23;
const nv_bfloat162 bias_reg =
__float2bfloat162_rn(*reinterpret_cast<const float*>(&BIAS));
// Convert to bfloat162 and apply bias
typename ScalarType<nv_bfloat16>::FragB frag_b;
// Note: reverse indexing is intentional because weights are permuted
frag_b[1] = __hmul2(*reinterpret_cast<const nv_bfloat162*>(&Out1), bias_reg);
frag_b[0] = __hmul2(*reinterpret_cast<const nv_bfloat162*>(&Out2), bias_reg);
return frag_b;
}
// Multiply dequantized values by the corresponding quantization scale; used
// only for grouped quantization.
template <typename scalar_t>
__device__ inline void scale(typename ScalarType<scalar_t>::FragB& frag_b,
typename ScalarType<scalar_t>::FragS& frag_s,
int i) {
using scalar_t2 = typename ScalarType<scalar_t>::scalar_t2;
scalar_t2 s =
ScalarType<scalar_t>::num2num2(reinterpret_cast<scalar_t*>(&frag_s)[i]);
frag_b[0] = __hmul2(frag_b[0], s);
frag_b[1] = __hmul2(frag_b[1], s);
}
// Given 2 floats multiply by 2 scales (halves)
template <typename scalar_t>
__device__ inline void scale_float(float* c,
typename ScalarType<scalar_t>::FragS& s) {
scalar_t* s_ptr = reinterpret_cast<scalar_t*>(&s);
c[0] = __fmul_rn(c[0], ScalarType<scalar_t>::num2float(s_ptr[0]));
c[1] = __fmul_rn(c[1], ScalarType<scalar_t>::num2float(s_ptr[1]));
}
// Wait until barrier reaches `count`, then lock for current threadblock.
__device__ inline void barrier_acquire(int* lock, int count) {
if (threadIdx.x == 0) {
int state = -1;
do
// Guarantee that subsequent writes by this threadblock will be visible
// globally.
asm volatile("ld.global.acquire.gpu.b32 %0, [%1];\n"
: "=r"(state)
: "l"(lock));
while (state != count);
}
__syncthreads();
}
// Release barrier and increment visitation count.
__device__ inline void barrier_release(int* lock, bool reset = false) {
__syncthreads();
if (threadIdx.x == 0) {
if (reset) {
lock[0] = 0;
return;
}
int val = 1;
// Make sure that all writes since acquiring this barrier are visible
// globally, while releasing the barrier.
asm volatile("fence.acq_rel.gpu;\n");
asm volatile("red.relaxed.gpu.global.add.s32 [%0], %1;\n"
:
: "l"(lock), "r"(val));
}
}
template <typename scalar_t, // compute dtype, half or nv_float16
const int num_bits, // number of bits used for weights
const int threads, // number of threads in a threadblock
const int thread_m_blocks, // number of 16x16 blocks in the m
// dimension (batchsize) of the
// threadblock
const int thread_n_blocks, // same for n dimension (output)
const int thread_k_blocks, // same for k dimension (reduction)
const int stages, // number of stages for the async global->shared
// fetch pipeline
const int group_blocks = -1 // number of consecutive 16x16 blocks
// with a separate quantization scale
>
__global__ void Marlin(
const int4* __restrict__ A, // fp16 input matrix of shape mxk
const int4* __restrict__ B, // 4bit quantized weight matrix of shape kxn
int4* __restrict__ C, // fp16 output buffer of shape mxn
const int4* __restrict__ scales_ptr, // fp16 quantization scales of shape
// (k/groupsize)xn
int num_groups, // number of scale groups per output channel
int prob_m, // batch dimension m
int prob_n, // output dimension n
int prob_k, // reduction dimension k
int* locks // extra global storage for barrier synchronization
) {
// Each threadblock processes one "stripe" of the B matrix with (roughly) the
// same size, which might involve multiple column "slices" (of width 16 *
// `thread_n_blocks`). Stripes are defined as shown in the 3x3 matrix 5 SM
// example:
// 0 1 3
// 0 2 3
// 1 2 4
// While this kind of partitioning makes things somewhat more complicated, it
// ensures good utilization of all SMs for many kinds of shape and GPU
// configurations, while requiring as few slow global cross-threadblock
// reductions as possible.
using Dtype = ScalarType<scalar_t>;
using scalar_t2 = typename ScalarType<scalar_t>::scalar_t2;
using FragA = typename ScalarType<scalar_t>::FragA;
using FragB = typename ScalarType<scalar_t>::FragB;
using FragC = typename ScalarType<scalar_t>::FragC;
using FragS = typename ScalarType<scalar_t>::FragS;
constexpr int pack_factor = 32 / num_bits;
// For larger GEMMs we run multiple batchsize 64 versions in parallel for a
// better partitioning with less reductions
int parallel = 1;
if (prob_m > 16 * thread_m_blocks) {
parallel = prob_m / (16 * thread_m_blocks);
prob_m = 16 * thread_m_blocks;
}
int k_tiles = prob_k / 16 / thread_k_blocks;
int n_tiles = prob_n / 16 / thread_n_blocks;
int iters = div_ceil(k_tiles * n_tiles * parallel, gridDim.x);
int slice_row = (iters * blockIdx.x) % k_tiles;
int slice_col_par = (iters * blockIdx.x) / k_tiles;
int slice_col = slice_col_par;
int slice_iters; // number of threadblock tiles in the current slice
int slice_count =
0; // total number of active threadblocks in the current slice
int slice_idx; // index of threadblock in current slice; numbered bottom to
// top
// We can easily implement parallel problem execution by just remapping
// indices and advancing global pointers
if (slice_col_par >= n_tiles) {
A += (slice_col_par / n_tiles) * 16 * thread_m_blocks * prob_k / 8;
C += (slice_col_par / n_tiles) * 16 * thread_m_blocks * prob_n / 8;
locks += (slice_col_par / n_tiles) * n_tiles;
slice_col = slice_col_par % n_tiles;
}
// Compute all information about the current slice which is required for
// synchronization.
auto init_slice = [&]() {
slice_iters =
iters * (blockIdx.x + 1) - (k_tiles * slice_col_par + slice_row);
if (slice_iters < 0 || slice_col_par >= n_tiles * parallel) slice_iters = 0;
if (slice_iters == 0) return;
if (slice_row + slice_iters > k_tiles) slice_iters = k_tiles - slice_row;
slice_count = 1;
slice_idx = 0;
int col_first = iters * div_ceil(k_tiles * slice_col_par, iters);
if (col_first <= k_tiles * (slice_col_par + 1)) {
int col_off = col_first - k_tiles * slice_col_par;
slice_count = div_ceil(k_tiles - col_off, iters);
if (col_off > 0) slice_count++;
int delta_first = iters * blockIdx.x - col_first;
if (delta_first < 0 || (col_off == 0 && delta_first == 0))
slice_idx = slice_count - 1;
else {
slice_idx = slice_count - 1 - delta_first / iters;
if (col_off > 0) slice_idx--;
}
}
if (slice_col == n_tiles) {
A += 16 * thread_m_blocks * prob_k / 8;
C += 16 * thread_m_blocks * prob_n / 8;
locks += n_tiles;
slice_col = 0;
}
};
init_slice();
// A sizes/strides
// stride of the A matrix in global memory
int a_gl_stride = prob_k / 8;
// stride of an A matrix tile in shared memory
constexpr int a_sh_stride = 16 * thread_k_blocks / 8;
// delta between subsequent A tiles in global memory
constexpr int a_gl_rd_delta_o = 16 * thread_k_blocks / 8;
// between subsequent accesses within a tile
int a_gl_rd_delta_i = a_gl_stride * (threads / a_gl_rd_delta_o);
// between shared memory writes
constexpr int a_sh_wr_delta = a_sh_stride * (threads / a_gl_rd_delta_o);
// between shared memory tile reads
constexpr int a_sh_rd_delta_o = 2 * ((threads / 32) / (thread_n_blocks / 4));
// within a shared memory tile
constexpr int a_sh_rd_delta_i = a_sh_stride * 16;
// overall size of a tile
constexpr int a_sh_stage = a_sh_stride * (16 * thread_m_blocks);
// number of shared write iterations for a tile
constexpr int a_sh_wr_iters = div_ceil(a_sh_stage, a_sh_wr_delta);
// B sizes/strides
int b_gl_stride = 16 * prob_n / (pack_factor * 4);
constexpr int b_sh_stride = ((thread_n_blocks * 16) * 16 / pack_factor) / 4;
constexpr int b_thread_vecs = num_bits == 4 ? 1 : 2;
constexpr int b_sh_stride_threads = b_sh_stride / b_thread_vecs;
int b_gl_rd_delta_o = b_gl_stride * thread_k_blocks;
int b_gl_rd_delta_i = b_gl_stride * (threads / b_sh_stride_threads);
constexpr int b_sh_wr_delta = threads * b_thread_vecs;
constexpr int b_sh_rd_delta = threads * b_thread_vecs;
constexpr int b_sh_stage = b_sh_stride * thread_k_blocks;
constexpr int b_sh_wr_iters = b_sh_stage / b_sh_wr_delta;
// Scale sizes/strides without act_order
int s_gl_stride = prob_n / 8;
constexpr int s_sh_stride = 16 * thread_n_blocks / 8;
// Scale size/strides with act_order
constexpr int tb_k = 16 * thread_k_blocks;
constexpr int g_idx_stage = 0;
// constexpr int act_s_row_stride = 1;
// int act_s_col_stride = act_s_row_stride * num_groups;
int act_s_col_stride = 1;
int act_s_col_warp_stride = act_s_col_stride * 8;
int tb_n_warps = thread_n_blocks / 4;
int act_s_col_tb_stride = act_s_col_warp_stride * tb_n_warps;
// Global A read index of current thread.
int a_gl_rd = a_gl_stride * (threadIdx.x / a_gl_rd_delta_o) +
(threadIdx.x % a_gl_rd_delta_o);
a_gl_rd += a_gl_rd_delta_o * slice_row;
// Shared write index of current thread.
int a_sh_wr = a_sh_stride * (threadIdx.x / a_gl_rd_delta_o) +
(threadIdx.x % a_gl_rd_delta_o);
// Shared read index.
int a_sh_rd =
a_sh_stride * ((threadIdx.x % 32) % 16) + (threadIdx.x % 32) / 16;
a_sh_rd += 2 * ((threadIdx.x / 32) / (thread_n_blocks / 4));
int b_gl_rd = b_gl_stride * (threadIdx.x / b_sh_stride_threads) +
(threadIdx.x % b_sh_stride_threads) * b_thread_vecs;
b_gl_rd += b_sh_stride * slice_col;
b_gl_rd += b_gl_rd_delta_o * slice_row;
int b_sh_wr = threadIdx.x * b_thread_vecs;
int b_sh_rd = threadIdx.x * b_thread_vecs;
// For act_order
int slice_k_start = tb_k * slice_row;
int slice_k_start_shared_fetch = slice_k_start;
int slice_n_offset = act_s_col_tb_stride * slice_col;
// No act_order
int s_gl_rd = s_sh_stride * slice_col + threadIdx.x;
int s_sh_wr = threadIdx.x;
bool s_sh_wr_pred = threadIdx.x < s_sh_stride;
// We scale a `half2` tile in row-major layout for column-wise quantization.
int s_sh_rd =
8 * ((threadIdx.x / 32) % (thread_n_blocks / 4)) + (threadIdx.x % 32) % 4;
// Precompute which thread should not read memory in which iterations; this is
// needed if there are more threads than required for a certain tilesize or
// when the batchsize is not a multiple of 16.
bool a_sh_wr_pred[a_sh_wr_iters];
#pragma unroll
for (int i = 0; i < a_sh_wr_iters; i++)
a_sh_wr_pred[i] = a_sh_wr_delta * i + a_sh_wr < a_sh_stride * prob_m;
// To ensure that writing and reading A tiles to/from shared memory, the
// latter in fragment format, is fully bank conflict free, we need to use a
// rather fancy XOR-based layout. The key here is that neither reads nor
// writes of the 16-byte `int4` blocks of 8 consecutive threads involve the
// same shared memory banks. Further, it seems (based on NSight-Compute) that
// each warp must also write a consecutive memory segment?
auto transform_a = [&](int i) {
int row = i / a_gl_rd_delta_o;
return a_gl_rd_delta_o * row + (i % a_gl_rd_delta_o) ^ row;
};
// Since the computation of this remapping is non-trivial and, due to our main
// loop unrolls, all shared memory accesses are static, we simply precompute
// both transformed reads and writes.
int a_sh_wr_trans[a_sh_wr_iters];
#pragma unroll
for (int i = 0; i < a_sh_wr_iters; i++)
a_sh_wr_trans[i] = transform_a(a_sh_wr_delta * i + a_sh_wr);
int a_sh_rd_trans[b_sh_wr_iters][thread_m_blocks];
#pragma unroll
for (int i = 0; i < b_sh_wr_iters; i++) {
#pragma unroll
for (int j = 0; j < thread_m_blocks; j++)
a_sh_rd_trans[i][j] =
transform_a(a_sh_rd_delta_o * i + a_sh_rd_delta_i * j + a_sh_rd);
}
// Since B-accesses have non-constant stride they have to be computed at
// runtime; we break dependencies between subsequent accesses with a tile by
// maintining multiple pointers (we have enough registers), a tiny
// optimization.
const int4* B_ptr[b_sh_wr_iters];
#pragma unroll
for (int i = 0; i < b_sh_wr_iters; i++)
B_ptr[i] = B + b_gl_rd_delta_i * i + b_gl_rd;
extern __shared__ int4 sh[];
// Shared memory storage for global fetch pipelines.
int4* sh_a = sh;
int4* sh_b = sh_a + (stages * a_sh_stage);
int4* sh_g_idx = sh_b + (stages * b_sh_stage);
int4* sh_s = sh_g_idx + (stages * g_idx_stage);
// Register storage for double buffer of shared memory reads.
FragA frag_a[2][thread_m_blocks];
I4 frag_b_quant[2][b_thread_vecs];
FragC frag_c[thread_m_blocks][4][2];
FragS frag_s[2][4];
// Zero accumulators.
auto zero_accums = [&]() {
#pragma unroll
for (int i = 0; i < thread_m_blocks * 4 * 2 * 4; i++)
reinterpret_cast<float*>(frag_c)[i] = 0;
};
int sh_first_group_id = -1;
int sh_num_groups = -1;
constexpr int sh_max_num_groups = 32;
auto fetch_scales_to_shared = [&](bool is_async, int first_group_id,
int last_group_id) {
sh_first_group_id = first_group_id;
sh_num_groups = last_group_id - first_group_id + 1;
if (sh_num_groups < sh_max_num_groups) {
sh_num_groups = sh_max_num_groups;
}
if (sh_first_group_id + sh_num_groups > num_groups) {
sh_num_groups = num_groups - sh_first_group_id;
}
int row_offset = first_group_id * s_gl_stride;
if (is_async) {
for (int i = 0; i < sh_num_groups; i++) {
if (threadIdx.x < s_sh_stride) {
cp_async4_pred(&sh_s[(i * s_sh_stride) + threadIdx.x],
&scales_ptr[row_offset + (i * s_gl_stride) +
slice_n_offset + threadIdx.x]);
}
}
} else {
for (int i = 0; i < sh_num_groups; i++) {
if (threadIdx.x < s_sh_stride) {
sh_s[(i * s_sh_stride) + threadIdx.x] =
scales_ptr[row_offset + (i * s_gl_stride) + slice_n_offset +
threadIdx.x];
}
}
}
};
// Asynchronously fetch the next A, B and s tile from global to the next
// shared memory pipeline location.
auto fetch_to_shared = [&](int pipe, int a_off, bool pred = true) {
if (pred) {
int4* sh_a_stage = sh_a + a_sh_stage * pipe;
#pragma unroll
for (int i = 0; i < a_sh_wr_iters; i++) {
cp_async4_pred(
&sh_a_stage[a_sh_wr_trans[i]],
&A[a_gl_rd_delta_i * i + a_gl_rd + a_gl_rd_delta_o * a_off],
a_sh_wr_pred[i]);
}
int4* sh_b_stage = sh_b + b_sh_stage * pipe;
#pragma unroll
for (int i = 0; i < b_sh_wr_iters; i++) {
#pragma unroll
for (int j = 0; j < b_thread_vecs; j++) {
cp_async4(&sh_b_stage[b_sh_wr_delta * i + b_sh_wr + j], B_ptr[i] + j);
}
B_ptr[i] += b_gl_rd_delta_o;
}
}
// Insert a fence even when we are winding down the pipeline to ensure that
// waiting is also correct at this point.
cp_async_fence();
};
// Wait until the next thread tile has been loaded to shared memory.
auto wait_for_stage = [&]() {
// We only have `stages - 2` active fetches since we are double buffering
// and can only issue the next fetch when it is guaranteed that the previous
// shared memory load is fully complete (as it may otherwise be
// overwritten).
cp_async_wait<stages - 2>();
__syncthreads();
};
// Load the next sub-tile from the current location in the shared memory pipe
// into the current register buffer.
auto fetch_to_registers = [&](int k, int pipe) {
int4* sh_a_stage = sh_a + a_sh_stage * pipe;
#pragma unroll
for (int i = 0; i < thread_m_blocks; i++)
ldsm4<scalar_t>(frag_a[k % 2][i],
&sh_a_stage[a_sh_rd_trans[k % b_sh_wr_iters][i]]);
int4* sh_b_stage = sh_b + b_sh_stage * pipe;
#pragma unroll
for (int i = 0; i < b_thread_vecs; i++) {
frag_b_quant[k % 2][i] = *reinterpret_cast<I4*>(
&sh_b_stage[b_sh_rd_delta * (k % b_sh_wr_iters) + b_sh_rd + i]);
}
};
bool is_same_group[stages];
int same_group_id[stages];
auto init_same_group = [&](int pipe) {
is_same_group[pipe] = false;
same_group_id[pipe] = 0;
return;
};
// Execute the actual tensor core matmul of a sub-tile.
auto matmul = [&](int k) {
// We have the m dimension as the inner loop in order to encourage overlapping
// dequantization and matmul operations.
#pragma unroll
for (int j = 0; j < 4; j++) {
FragB frag_b0;
FragB frag_b1;
int* frag_b_quant_ptr = reinterpret_cast<int*>(frag_b_quant[k % 2]);
int b_quant_0 = frag_b_quant_ptr[j * 2 + 0];
int b_quant_1 = frag_b_quant_ptr[j * 2 + 1];
frag_b0 = dequant_8bit<scalar_t>(b_quant_0);
frag_b1 = dequant_8bit<scalar_t>(b_quant_1);
#pragma unroll
for (int i = 0; i < thread_m_blocks; i++) {
mma<scalar_t>(frag_a[k % 2][i], frag_b0, frag_c[i][j][0]);
mma<scalar_t>(frag_a[k % 2][i], frag_b1, frag_c[i][j][1]);
}
}
};
// Since we slice across the k dimension of a tile in order to increase the
// number of warps while keeping the n dimension of a tile reasonable, we have
// multiple warps that accumulate their partial sums of the same output
// location; which we have to reduce over in the end. We do in shared memory.
auto thread_block_reduce = [&]() {
constexpr int red_off = threads / b_sh_stride_threads / 2;
if (red_off >= 1) {
int red_idx = threadIdx.x / b_sh_stride_threads;
constexpr int red_sh_stride = b_sh_stride_threads * 4 * 2;
constexpr int red_sh_delta = b_sh_stride_threads;
int red_sh_rd = red_sh_stride * (threadIdx.x / b_sh_stride_threads) +
(threadIdx.x % b_sh_stride_threads);
// Parallel logarithmic shared memory reduction. We make sure to avoid any
// unnecessary read or write iterations, e.g., for two warps we write only
// once by warp 1 and read only once by warp 0.
#pragma unroll
for (int m_block = 0; m_block < thread_m_blocks; m_block++) {
#pragma unroll
for (int i = red_off; i > 0; i /= 2) {
if (i <= red_idx && red_idx < 2 * i) {
#pragma unroll
for (int j = 0; j < 4 * 2; j++) {
int red_sh_wr =
red_sh_delta * j + (red_sh_rd - red_sh_stride * i);
if (i < red_off) {
float* c_rd =
reinterpret_cast<float*>(&sh[red_sh_delta * j + red_sh_rd]);
float* c_wr = reinterpret_cast<float*>(&sh[red_sh_wr]);
#pragma unroll
for (int k = 0; k < 4; k++)
reinterpret_cast<FragC*>(frag_c)[4 * 2 * m_block + j][k] +=
c_rd[k] + c_wr[k];
}
sh[red_sh_wr] =
reinterpret_cast<int4*>(&frag_c)[4 * 2 * m_block + j];
}
}
__syncthreads();
}
if (red_idx == 0) {
#pragma unroll
for (int i = 0; i < 4 * 2; i++) {
float* c_rd =
reinterpret_cast<float*>(&sh[red_sh_delta * i + red_sh_rd]);
#pragma unroll
for (int j = 0; j < 4; j++)
reinterpret_cast<FragC*>(frag_c)[4 * 2 * m_block + i][j] +=
c_rd[j];
}
}
__syncthreads();
}
}
};
// Since multiple threadblocks may process parts of the same column slice, we
// finally have to globally reduce over the results. As the striped
// partitioning minimizes the number of such reductions and our outputs are
// usually rather small, we perform this reduction serially in L2 cache.
auto global_reduce = [&](bool first = false, bool last = false) {
// We are very careful here to reduce directly in the output buffer to
// maximize L2 cache utilization in this step. To do this, we write out
// results in FP16 (but still reduce with FP32 compute).
constexpr int active_threads = 32 * thread_n_blocks / 4;
if (threadIdx.x < active_threads) {
int c_gl_stride = prob_n / 8;
int c_gl_wr_delta_o = 8 * c_gl_stride;
int c_gl_wr_delta_i = 4 * (active_threads / 32);
int c_gl_wr = c_gl_stride * ((threadIdx.x % 32) / 4) +
4 * (threadIdx.x / 32) + threadIdx.x % 4;
c_gl_wr += (2 * thread_n_blocks) * slice_col;
constexpr int c_sh_wr_delta = active_threads;
int c_sh_wr = threadIdx.x;
int row = (threadIdx.x % 32) / 4;
if (!first) {
// Interestingly, doing direct global accesses here really seems to mess up
// the compiler and lead to slowdowns, hence we also use async-copies even
// though these fetches are not actually asynchronous.
#pragma unroll
for (int i = 0; i < thread_m_blocks * 4; i++) {
cp_async4_pred(
&sh[c_sh_wr + c_sh_wr_delta * i],
&C[c_gl_wr + c_gl_wr_delta_o * (i / 2) +
c_gl_wr_delta_i * (i % 2)],
i < (thread_m_blocks - 1) * 4 || 8 * (i / 2) + row < prob_m);
}
cp_async_fence();
cp_async_wait<0>();
}
#pragma unroll
for (int i = 0; i < thread_m_blocks * 4; i++) {
if (i < (thread_m_blocks - 1) * 4 || 8 * (i / 2) + row < prob_m) {
if (!first) {
int4 c_red = sh[c_sh_wr + i * c_sh_wr_delta];
#pragma unroll
for (int j = 0; j < 2 * 4; j++) {
reinterpret_cast<float*>(
&frag_c)[4 * 2 * 4 * (i / 4) + 4 * j + (i % 4)] +=
Dtype::num2float(reinterpret_cast<scalar_t*>(&c_red)[j]);
}
}
if (!last) {
int4 c;
#pragma unroll
for (int j = 0; j < 2 * 4; j++) {
reinterpret_cast<scalar_t*>(&c)[j] =
Dtype::float2num(reinterpret_cast<float*>(
&frag_c)[4 * 2 * 4 * (i / 4) + 4 * j + (i % 4)]);
}
C[c_gl_wr + c_gl_wr_delta_o * (i / 2) + c_gl_wr_delta_i * (i % 2)] =
c;
}
}
}
}
};
// Write out the reduce final result in the correct layout. We only actually
// reshuffle matrix fragments in this step, the reduction above is performed
// in fragment layout.
auto write_result = [&]() {
int c_gl_stride = prob_n / 8;
constexpr int c_sh_stride = 2 * thread_n_blocks + 1;
int c_gl_wr_delta = c_gl_stride * (threads / (2 * thread_n_blocks));
constexpr int c_sh_rd_delta =
c_sh_stride * (threads / (2 * thread_n_blocks));
int c_gl_wr = c_gl_stride * (threadIdx.x / (2 * thread_n_blocks)) +
(threadIdx.x % (2 * thread_n_blocks));
c_gl_wr += (2 * thread_n_blocks) * slice_col;
int c_sh_wr =
(4 * c_sh_stride) * ((threadIdx.x % 32) / 4) + (threadIdx.x % 32) % 4;
c_sh_wr += 32 * (threadIdx.x / 32);
int c_sh_rd = c_sh_stride * (threadIdx.x / (2 * thread_n_blocks)) +
(threadIdx.x % (2 * thread_n_blocks));
int c_gl_wr_end = c_gl_stride * prob_m;
// We first reorder in shared memory to guarantee the most efficient final
// global write patterns
auto write = [&](int idx, float c0, float c1, FragS& s) {
scalar_t2 res =
Dtype::nums2num2(Dtype::float2num(c0), Dtype::float2num(c1));
((scalar_t2*)sh)[idx] = res;
};
if (threadIdx.x / 32 < thread_n_blocks / 4) {
#pragma unroll
for (int i = 0; i < thread_m_blocks; i++) {
#pragma unroll
for (int j = 0; j < 4; j++) {
int wr = c_sh_wr + 8 * j;
write(wr + (4 * c_sh_stride) * 0 + 0, frag_c[i][j][0][0],
frag_c[i][j][0][1], frag_s[j / 2][2 * (j % 2) + 0]);
write(wr + (4 * c_sh_stride) * 8 + 0, frag_c[i][j][0][2],
frag_c[i][j][0][3], frag_s[j / 2][2 * (j % 2) + 0]);
write(wr + (4 * c_sh_stride) * 0 + 4, frag_c[i][j][1][0],
frag_c[i][j][1][1], frag_s[j / 2][2 * (j % 2) + 1]);
write(wr + (4 * c_sh_stride) * 8 + 4, frag_c[i][j][1][2],
frag_c[i][j][1][3], frag_s[j / 2][2 * (j % 2) + 1]);
}
c_sh_wr += 16 * (4 * c_sh_stride);
}
}
__syncthreads();
#pragma unroll
for (int i = 0;
i < div_ceil(16 * thread_m_blocks, threads / (2 * thread_n_blocks));
i++) {
if (c_gl_wr < c_gl_wr_end) {
C[c_gl_wr] = sh[c_sh_rd];
c_gl_wr += c_gl_wr_delta;
c_sh_rd += c_sh_rd_delta;
}
}
};
// Start global fetch and register load pipelines.
auto start_pipes = [&]() {
#pragma unroll
for (int i = 0; i < stages - 1; i++) {
fetch_to_shared(i, i, i < slice_iters);
}
zero_accums();
wait_for_stage();
init_same_group(0);
fetch_to_registers(0, 0);
a_gl_rd += a_gl_rd_delta_o * (stages - 1);
slice_k_start_shared_fetch += tb_k * (stages - 1);
};
if (slice_iters) {
start_pipes();
}
// Main loop.
while (slice_iters) {
// We unroll over both the global fetch and the register load pipeline to
// ensure all shared memory accesses are static. Note that both pipelines
// have even length meaning that the next iteration will always start at
// index 0.
#pragma unroll
for (int pipe = 0; pipe < stages;) {
#pragma unroll
for (int k = 0; k < b_sh_wr_iters; k++) {
fetch_to_registers(k + 1, pipe % stages);
if (k == b_sh_wr_iters - 2) {
fetch_to_shared((pipe + stages - 1) % stages, pipe,
slice_iters >= stages);
pipe++;
wait_for_stage();
init_same_group(pipe % stages);
}
matmul(k);
}
slice_iters--;
if (slice_iters == 0) {
break;
}
}
a_gl_rd += a_gl_rd_delta_o * stages;
slice_k_start += tb_k * stages;
slice_k_start_shared_fetch += tb_k * stages;
// Process results and, if necessary, proceed to the next column slice.
// While this pattern may not be the most readable, other ways of writing
// the loop seemed to noticeably worse performance after compilation.
if (slice_iters == 0) {
cp_async_wait<0>();
bool last = slice_idx == slice_count - 1;
// For per-column scales, we only fetch them here in the final step before
// write-out
if (s_sh_wr_pred) {
cp_async4(&sh_s[s_sh_wr], &scales_ptr[s_gl_rd]);
}
cp_async_fence();
thread_block_reduce();
cp_async_wait<0>();
__syncthreads();
if (threadIdx.x / 32 < thread_n_blocks / 4) {
reinterpret_cast<int4*>(&frag_s)[0] = sh_s[s_sh_rd + 0];
reinterpret_cast<int4*>(&frag_s)[1] = sh_s[s_sh_rd + 4];
}
// For 8-bit channelwise, we apply the scale before the global reduction
// that converts the fp32 results to fp16 (so that we avoid possible
// overflow in fp16)
if (threadIdx.x / 32 < thread_n_blocks / 4) {
#pragma unroll
for (int i = 0; i < thread_m_blocks; i++) {
#pragma unroll
for (int j = 0; j < 4; j++) {
scale_float<scalar_t>(reinterpret_cast<float*>(&frag_c[i][j][0][0]),
frag_s[j / 2][2 * (j % 2) + 0]);
scale_float<scalar_t>(reinterpret_cast<float*>(&frag_c[i][j][0][2]),
frag_s[j / 2][2 * (j % 2) + 0]);
scale_float<scalar_t>(reinterpret_cast<float*>(&frag_c[i][j][1][0]),
frag_s[j / 2][2 * (j % 2) + 1]);
scale_float<scalar_t>(reinterpret_cast<float*>(&frag_c[i][j][1][2]),
frag_s[j / 2][2 * (j % 2) + 1]);
}
}
}
if (slice_count > 1) { // only globally reduce if there is more than one
// block in a slice
barrier_acquire(&locks[slice_col], slice_idx);
global_reduce(slice_idx == 0, last);
barrier_release(&locks[slice_col], last);
}
if (last) // only the last block in a slice actually writes the result
write_result();
slice_row = 0;
slice_col_par++;
slice_col++;
init_slice();
if (slice_iters) {
a_gl_rd = a_gl_stride * (threadIdx.x / a_gl_rd_delta_o) +
(threadIdx.x % a_gl_rd_delta_o);
#pragma unroll
for (int i = 0; i < b_sh_wr_iters; i++)
B_ptr[i] += b_sh_stride - b_gl_rd_delta_o * k_tiles;
if (slice_col == 0) {
#pragma unroll
for (int i = 0; i < b_sh_wr_iters; i++) B_ptr[i] -= b_gl_stride;
}
// Update slice k/n for scales loading
s_gl_rd = s_sh_stride * slice_col + threadIdx.x;
start_pipes();
}
}
}
}
#define __CALL_IF(NUM_BITS, THREAD_M_BLOCKS, THREAD_N_BLOCKS, \
THREAD_K_BLOCKS, GROUP_BLOCKS, NUM_THREADS) \
else if (num_bits == NUM_BITS && thread_m_blocks == THREAD_M_BLOCKS && \
thread_n_blocks == THREAD_N_BLOCKS && \
thread_k_blocks == THREAD_K_BLOCKS && \
group_blocks == GROUP_BLOCKS && num_threads == NUM_THREADS) { \
cudaFuncSetAttribute( \
Marlin<scalar_t, NUM_BITS, NUM_THREADS, THREAD_M_BLOCKS, \
THREAD_N_BLOCKS, THREAD_K_BLOCKS, pipe_stages, GROUP_BLOCKS>, \
cudaFuncAttributeMaxDynamicSharedMemorySize, max_shared_mem); \
Marlin<scalar_t, NUM_BITS, NUM_THREADS, THREAD_M_BLOCKS, \
THREAD_N_BLOCKS, THREAD_K_BLOCKS, pipe_stages, GROUP_BLOCKS> \
<<<blocks, NUM_THREADS, max_shared_mem, stream>>>( \
A_ptr, B_ptr, C_ptr, s_ptr, num_groups, prob_m, prob_n, prob_k, \
locks); \
}
typedef struct {
int thread_k;
int thread_n;
int num_threads;
} thread_config_t;
typedef struct {
int max_m_blocks;
thread_config_t tb_cfg;
} exec_config_t;
thread_config_t small_batch_thread_configs[] = {
// Ordered by priority
// thread_k, thread_n, num_threads
{128, 128, 256},
{64, 128, 128},
{128, 64, 128},
};
thread_config_t large_batch_thread_configs[] = {
// Ordered by priority
// thread_k, thread_n, num_threads
{64, 256, 256},
{64, 128, 128},
{128, 64, 128},
};
int get_scales_cache_size(thread_config_t const& th_config, int prob_m,
int prob_n, int prob_k, int num_bits,
int group_size) {
int tb_n = th_config.thread_n;
// Get max scale groups per thread-block
// Fixed for channelwise
int tb_groups = 1;
int tb_scales = tb_groups * tb_n * 2;
return tb_scales * pipe_stages;
}
bool is_valid_cache_size(thread_config_t const& th_config, int max_m_blocks,
int prob_m, int prob_n, int prob_k, int num_bits,
int scales_cache_size, int max_shared_mem) {
int pack_factor = 32 / num_bits;
// Get B size
int tb_k = th_config.thread_k;
int tb_n = th_config.thread_n;
int b_size = (tb_k * tb_n / pack_factor) * 4;
// Get A size
int m_blocks = div_ceil(prob_m, 16);
int tb_max_m = 16;
while (true) {
if (m_blocks >= max_m_blocks) {
tb_max_m *= max_m_blocks;
break;
}
max_m_blocks--;
if (max_m_blocks == 0) {
TORCH_CHECK(false, "Unexpected m_blocks = ", m_blocks);
}
}
int a_size = (tb_max_m * tb_k) * 2;
float pipe_size = (a_size + b_size) * pipe_stages;
TORCH_CHECK(max_shared_mem / 2 > scales_cache_size); // Sanity
return pipe_size < 0.95f * (max_shared_mem - scales_cache_size);
}
bool is_valid_config(thread_config_t const& th_config, int max_m_blocks,
int prob_m, int prob_n, int prob_k, int num_bits,
int group_size, int max_shared_mem) {
// Sanity
if (th_config.thread_k == -1 || th_config.thread_n == -1 ||
th_config.num_threads == -1) {
return false;
}
// Verify K/N are divisible by thread K/N
if (prob_k % th_config.thread_k != 0 || prob_n % th_config.thread_n != 0) {
return false;
}
// Verify min for thread K/N
if (th_config.thread_n < min_thread_n || th_config.thread_k < min_thread_k) {
return false;
}
// num_threads must be at least 128 (= 4 warps)
if (th_config.num_threads < 128) {
return false;
}
// Determine cache for scales
int scales_cache_size = get_scales_cache_size(th_config, prob_m, prob_n,
prob_k, num_bits, group_size);
// Check that pipeline fits into cache
if (!is_valid_cache_size(th_config, max_m_blocks, prob_m, prob_n, prob_k,
num_bits, scales_cache_size, max_shared_mem)) {
return false;
}
return true;
}
exec_config_t determine_thread_config(int prob_m, int prob_n, int prob_k,
int num_bits, int group_size,
int max_shared_mem) {
int max_m_blocks = 4;
while (max_m_blocks > 0) {
if (prob_m <= 16) {
for (auto th_config : small_batch_thread_configs) {
if (is_valid_config(th_config, max_m_blocks, prob_m, prob_n, prob_k,
num_bits, group_size, max_shared_mem)) {
return exec_config_t{max_m_blocks, th_config};
}
}
} else {
for (auto th_config : large_batch_thread_configs) {
if (is_valid_config(th_config, max_m_blocks, prob_m, prob_n, prob_k,
num_bits, group_size, max_shared_mem)) {
return exec_config_t{max_m_blocks, th_config};
}
}
}
max_m_blocks--; // Process less M blocks per invocation to reduce cache
// usage
}
return exec_config_t{0, {-1, -1, -1}};
}
#define CALL_IF(NUM_BITS, N_BLOCKS, K_BLOCKS, NUM_THREADS) \
__CALL_IF(NUM_BITS, 1, N_BLOCKS, K_BLOCKS, -1, NUM_THREADS) \
__CALL_IF(NUM_BITS, 2, N_BLOCKS, K_BLOCKS, -1, NUM_THREADS) \
__CALL_IF(NUM_BITS, 3, N_BLOCKS, K_BLOCKS, -1, NUM_THREADS) \
__CALL_IF(NUM_BITS, 4, N_BLOCKS, K_BLOCKS, -1, NUM_THREADS)
template <typename scalar_t>
void marlin_mm_f16i4(const void* A, const void* B, void* C, void* s, int prob_m,
int prob_n, int prob_k, void* workspace, int num_bits,
int num_groups, int group_size, int dev,
cudaStream_t stream, int thread_k, int thread_n, int sms,
int max_par) {
TORCH_CHECK(num_bits == 8, "num_bits must be 8. Got = ", num_bits);
TORCH_CHECK(prob_m > 0 && prob_n > 0 && prob_k > 0, "Invalid MNK = [", prob_m,
", ", prob_n, ", ", prob_k, "]");
int tot_m = prob_m;
int tot_m_blocks = div_ceil(tot_m, 16);
int pad = 16 * tot_m_blocks - tot_m;
if (sms == -1) {
cudaDeviceGetAttribute(&sms, cudaDevAttrMultiProcessorCount, dev);
}
int max_shared_mem = 0;
cudaDeviceGetAttribute(&max_shared_mem,
cudaDevAttrMaxSharedMemoryPerBlockOptin, dev);
TORCH_CHECK(max_shared_mem > 0);
// Set thread config
exec_config_t exec_cfg;
if (thread_k != -1 && thread_n != -1) {
// User-defined config
exec_cfg =
exec_config_t{4, thread_config_t{thread_k, thread_n, default_threads}};
} else {
// Auto config
exec_cfg = determine_thread_config(prob_m, prob_n, prob_k, num_bits,
group_size, max_shared_mem);
}
TORCH_CHECK(
exec_cfg.max_m_blocks > 0 &&
is_valid_config(exec_cfg.tb_cfg, exec_cfg.max_m_blocks, prob_m,
prob_n, prob_k, num_bits, group_size, max_shared_mem),
"Invalid thread config: max_m_blocks = ", exec_cfg.max_m_blocks,
", thread_k = ", exec_cfg.tb_cfg.thread_k,
", thread_n = ", exec_cfg.tb_cfg.thread_n,
", num_threads = ", exec_cfg.tb_cfg.num_threads, " for MKN = [", prob_m,
", ", prob_k, ", ", prob_n, "] and num_bits = ", num_bits,
", group_size = ", group_size, ", max_shared_mem = ", max_shared_mem);
int num_threads = exec_cfg.tb_cfg.num_threads;
thread_k = exec_cfg.tb_cfg.thread_k;
thread_n = exec_cfg.tb_cfg.thread_n;
int thread_k_blocks = thread_k / 16;
int thread_n_blocks = thread_n / 16;
int blocks = sms;
TORCH_CHECK(prob_n % thread_n == 0, "prob_n = ", prob_n,
" is not divisible by thread_n = ", thread_n);
TORCH_CHECK(prob_k % thread_k == 0, "prob_k = ", prob_k,
" is not divisible by thread_k = ", thread_k);
int group_blocks = -1;
const int4* A_ptr = (const int4*)A;
const int4* B_ptr = (const int4*)B;
int4* C_ptr = (int4*)C;
const int4* s_ptr = (const int4*)s;
int* locks = (int*)workspace;
// Main loop
for (int i = 0; i < tot_m_blocks; i += exec_cfg.max_m_blocks) {
int thread_m_blocks = tot_m_blocks - i;
prob_m = tot_m - 16 * i;
int par = 1;
if (thread_m_blocks > exec_cfg.max_m_blocks) {
// Note that parallel > 1 currently only works for inputs without any
// padding
par = (16 * thread_m_blocks - pad) / (16 * exec_cfg.max_m_blocks);
if (par > max_par) par = max_par;
prob_m = (16 * exec_cfg.max_m_blocks) * par;
i += exec_cfg.max_m_blocks * (par - 1);
thread_m_blocks = exec_cfg.max_m_blocks;
}
// Define kernel configurations
if (false) {
}
CALL_IF(8, 32, 2, 256)
CALL_IF(8, 16, 4, 256)
CALL_IF(8, 8, 8, 256)
CALL_IF(8, 8, 4, 128)
CALL_IF(8, 4, 8, 128)
else {
TORCH_CHECK(false, "Unsupported shapes: MNK = [" + str(prob_m) + ", " +
str(prob_n) + ", " + str(prob_k) + "]" +
", num_groups = " + str(num_groups) +
", group_size = " + str(group_size) +
", thread_m_blocks = " + str(thread_m_blocks) +
", thread_n_blocks = " + str(thread_n_blocks) +
", thread_k_blocks = " + str(thread_k_blocks));
}
A_ptr += 16 * thread_m_blocks * (prob_k / 8) * par;
C_ptr += 16 * thread_m_blocks * (prob_n / 8) * par;
}
}
} // namespace fp8_marlin
torch::Tensor fp8_marlin_gemm(torch::Tensor& a, torch::Tensor& b_q_weight,
torch::Tensor& b_scales, torch::Tensor& workspace,
int64_t num_bits, int64_t size_m, int64_t size_n,
int64_t size_k) {
// Verify num_bits
TORCH_CHECK(num_bits == 8, "num_bits must be 8. Got = ", num_bits);
int pack_factor = 32 / num_bits;
// Verify A
TORCH_CHECK(a.size(0) == size_m, "Shape mismatch: a.size(0) = ", a.size(0),
", size_m = ", size_m);
TORCH_CHECK(a.size(1) == size_k, "Shape mismatch: a.size(1) = ", a.size(1),
", size_k = ", size_k);
// Verify B
TORCH_CHECK(size_k % marlin::tile_size == 0, "size_k = ", size_k,
" is not divisible by tile_size = ", marlin::tile_size);
TORCH_CHECK((size_k / marlin::tile_size) == b_q_weight.size(0),
"Shape mismatch: b_q_weight.size(0) = ", b_q_weight.size(0),
", size_k = ", size_k, ", tile_size = ", marlin::tile_size);
TORCH_CHECK(b_q_weight.size(1) % marlin::tile_size == 0,
"b_q_weight.size(1) = ", b_q_weight.size(1),
" is not divisible by tile_size = ", marlin::tile_size);
int actual_size_n = (b_q_weight.size(1) / marlin::tile_size) * pack_factor;
TORCH_CHECK(size_n == actual_size_n, "size_n = ", size_n,
", actual_size_n = ", actual_size_n);
// Verify device and strides
TORCH_CHECK(a.device().is_cuda(), "A is not on GPU");
TORCH_CHECK(a.is_contiguous(), "A is not contiguous");
TORCH_CHECK(b_q_weight.device().is_cuda(), "b_q_weight is not on GPU");
TORCH_CHECK(b_q_weight.is_contiguous(), "b_q_weight is not contiguous");
TORCH_CHECK(b_scales.device().is_cuda(), "b_scales is not on GPU");
TORCH_CHECK(b_scales.is_contiguous(), "b_scales is not contiguous");
// Alloc buffers
const at::cuda::OptionalCUDAGuard device_guard(device_of(a));
auto options = torch::TensorOptions().dtype(a.dtype()).device(a.device());
torch::Tensor c = torch::empty({size_m, size_n}, options);
// thread_k: `k` size of a thread_tile in `weights` (can usually be left as
// auto -1)
int thread_k = -1;
// thread_n: `n` size of a thread_tile in `weights` (can usually be left as
// auto -1)
int thread_n = -1;
// sms: number of SMs to use for the kernel (can usually be left as auto -1)
int sms = -1;
// Detect groupsize and act_order
int num_groups = -1;
int group_size = -1;
int b_rank = b_scales.sizes().size();
TORCH_CHECK(b_rank == 2, "b_scales rank = ", b_rank, " is not 2");
TORCH_CHECK(b_scales.size(1) == size_n, "b_scales dim 1 = ", b_scales.size(1),
" is not size_n = ", size_n);
// Channelwise only for FP8
TORCH_CHECK(b_scales.size(0) == 1)
num_groups = b_scales.size(0);
// Verify workspace size
TORCH_CHECK(size_n % marlin::min_thread_n == 0, "size_n = ", size_n,
", is not divisible by min_thread_n = ", marlin::min_thread_n);
int min_workspace_size = (size_n / marlin::min_thread_n) * marlin::max_par;
TORCH_CHECK(workspace.numel() >= min_workspace_size,
"workspace.numel = ", workspace.numel(),
" is below min_workspace_size = ", min_workspace_size);
int dev = a.get_device();
if (a.scalar_type() == at::ScalarType::Half) {
fp8_marlin::marlin_mm_f16i4<half>(
a.data_ptr<at::Half>(), b_q_weight.data_ptr(), c.data_ptr<at::Half>(),
b_scales.data_ptr<at::Half>(), size_m, size_n, size_k,
workspace.data_ptr(), num_bits, num_groups, group_size, dev,
at::cuda::getCurrentCUDAStream(dev), thread_k, thread_n, sms,
marlin::max_par);
} else if (a.scalar_type() == at::ScalarType::BFloat16) {
fp8_marlin::marlin_mm_f16i4<nv_bfloat16>(
a.data_ptr<at::BFloat16>(), b_q_weight.data_ptr(),
c.data_ptr<at::BFloat16>(), b_scales.data_ptr<at::BFloat16>(), size_m,
size_n, size_k, workspace.data_ptr(), num_bits, num_groups, group_size,
dev, at::cuda::getCurrentCUDAStream(dev), thread_k, thread_n, sms,
marlin::max_par);
} else {
TORCH_CHECK(false, "fp8_marlin_gemm only supports bfloat16 and float16");
}
return c;
}
#endif
...@@ -475,6 +475,7 @@ __inline__ __device__ uint8_t scaled_vec_conversion<uint8_t, __nv_bfloat16>( ...@@ -475,6 +475,7 @@ __inline__ __device__ uint8_t scaled_vec_conversion<uint8_t, __nv_bfloat16>(
__NV_SATFINITE, fp8_type); __NV_SATFINITE, fp8_type);
return (uint8_t)res; return (uint8_t)res;
#endif #endif
__builtin_unreachable(); // Suppress missing return statement warning
} }
// float -> fp8 // float -> fp8
...@@ -508,6 +509,7 @@ __inline__ __device__ Tout convert(const Tin& x) { ...@@ -508,6 +509,7 @@ __inline__ __device__ Tout convert(const Tin& x) {
} }
#endif #endif
assert(false); assert(false);
__builtin_unreachable(); // Suppress missing return statement warning
} }
template <typename Tout, typename Tin, Fp8KVCacheDataType kv_dt> template <typename Tout, typename Tin, Fp8KVCacheDataType kv_dt>
...@@ -520,6 +522,7 @@ __inline__ __device__ Tout scaled_convert(const Tin& x, const float scale) { ...@@ -520,6 +522,7 @@ __inline__ __device__ Tout scaled_convert(const Tin& x, const float scale) {
} }
#endif #endif
assert(false); assert(false);
__builtin_unreachable(); // Suppress missing return statement warning
} }
// The following macro is used to dispatch the conversion function based on // The following macro is used to dispatch the conversion function based on
......
#include "marlin.cuh"
#if defined(__CUDA_ARCH__) && __CUDA_ARCH__ < 800
namespace marlin {
template <int const num_threads, int const num_bits, bool const has_perm>
__global__ void awq_marlin_repack_kernel(
uint32_t const* __restrict__ b_q_weight_ptr, uint32_t* __restrict__ out_ptr,
int size_k, int size_n) {}
} // namespace marlin
torch::Tensor awq_marlin_repack(torch::Tensor& b_q_weight, torch::Tensor& perm,
int64_t size_k, int64_t size_n,
int64_t num_bits) {
TORCH_CHECK_NOT_IMPLEMENTED(
false, "marlin_repack_from_gptq(..) requires CUDA_ARCH >= 8.0");
return torch::empty({1, 1});
}
#else
namespace marlin {
template <int const num_threads, int const num_bits>
__global__ void awq_marlin_repack_kernel(
uint32_t const* __restrict__ b_q_weight_ptr, uint32_t* __restrict__ out_ptr,
int size_k, int size_n) {
constexpr int pack_factor = 32 / num_bits;
int k_tiles = size_k / tile_k_size;
int n_tiles = size_n / tile_n_size;
int block_k_tiles = div_ceil(k_tiles, gridDim.x);
int start_k_tile = blockIdx.x * block_k_tiles;
if (start_k_tile >= k_tiles) {
return;
}
int finish_k_tile = min(start_k_tile + block_k_tiles, k_tiles);
// Wait until the next thread tile has been loaded to shared memory.
auto wait_for_stage = [&]() {
// We only have `stages - 2` active fetches since we are double buffering
// and can only issue the next fetch when it is guaranteed that the previous
// shared memory load is fully complete (as it may otherwise be
// overwritten).
cp_async_wait<repack_stages - 2>();
__syncthreads();
};
extern __shared__ int4 sh[];
constexpr int tile_n_ints = tile_n_size / pack_factor;
constexpr int stage_n_threads = tile_n_ints / 4;
constexpr int stage_k_threads = tile_k_size;
constexpr int stage_size = stage_k_threads * stage_n_threads;
auto fetch_to_shared = [&](int pipe, int k_tile_id, int n_tile_id) {
if (n_tile_id >= n_tiles) {
cp_async_fence();
return;
}
int first_n = n_tile_id * tile_n_size;
int first_n_packed = first_n / pack_factor;
int4* sh_ptr = sh + stage_size * pipe;
if (threadIdx.x < stage_size) {
int k_id = threadIdx.x / stage_n_threads;
int n_id = threadIdx.x % stage_n_threads;
int first_k = k_tile_id * tile_k_size;
cp_async4(&sh_ptr[k_id * stage_n_threads + n_id],
reinterpret_cast<int4 const*>(
&(b_q_weight_ptr[(first_k + k_id) * (size_n / pack_factor) +
first_n_packed + (n_id * 4)])));
}
cp_async_fence();
};
auto repack_tile = [&](int pipe, int k_tile_id, int n_tile_id) {
if (n_tile_id >= n_tiles) {
return;
}
int warp_id = threadIdx.x / 32;
int th_id = threadIdx.x % 32;
if (warp_id >= 4) {
return;
}
int tc_col = th_id / 4;
int tc_row = (th_id % 4) * 2;
constexpr int tc_offsets[4] = {0, 1, 8, 9};
int cur_n = warp_id * 16 + tc_col;
int cur_n_packed = cur_n / pack_factor;
int cur_n_pos = cur_n % pack_factor;
constexpr int sh_stride = tile_n_ints;
constexpr uint32_t mask = (1 << num_bits) - 1;
int4* sh_stage_ptr = sh + stage_size * pipe;
uint32_t* sh_stage_int_ptr = reinterpret_cast<uint32_t*>(sh_stage_ptr);
// Undo interleaving
int cur_n_pos_unpacked;
if constexpr (num_bits == 4) {
constexpr int undo_pack[8] = {0, 4, 1, 5, 2, 6, 3, 7};
cur_n_pos_unpacked = undo_pack[cur_n_pos];
} else {
constexpr int undo_pack[4] = {0, 2, 1, 3};
cur_n_pos_unpacked = undo_pack[cur_n_pos];
}
uint32_t vals[8];
#pragma unroll
for (int i = 0; i < 4; i++) {
int cur_elem = tc_row + tc_offsets[i];
int packed_src_0 = sh_stage_int_ptr[cur_n_packed + sh_stride * cur_elem];
int packed_src_1 = sh_stage_int_ptr[cur_n_packed + (8 / pack_factor) +
sh_stride * cur_elem];
vals[i] = (packed_src_0 >> (cur_n_pos_unpacked * num_bits)) & mask;
vals[4 + i] = (packed_src_1 >> (cur_n_pos_unpacked * num_bits)) & mask;
}
constexpr int tile_size = tile_k_size * tile_n_size / pack_factor;
int out_offset = (k_tile_id * n_tiles + n_tile_id) * tile_size;
// Result of:
// https://github.com/NVIDIA/FasterTransformer/blob/main/src/fastertransformer/cutlass_extensions/include/cutlass_extensions/interleaved_numeric_conversion.h
if constexpr (num_bits == 4) {
constexpr int pack_idx[8] = {0, 2, 4, 6, 1, 3, 5, 7};
uint32_t res = 0;
#pragma unroll
for (int i = 0; i < 8; i++) {
res |= vals[pack_idx[i]] << (i * 4);
}
out_ptr[out_offset + th_id * 4 + warp_id] = res;
} else {
constexpr int pack_idx[4] = {0, 2, 1, 3};
uint32_t res1 = 0;
uint32_t res2 = 0;
#pragma unroll
for (int i = 0; i < 4; i++) {
res1 |= vals[pack_idx[i]] << (i * 8);
res2 |= vals[4 + pack_idx[i]] << (i * 8);
}
out_ptr[out_offset + th_id * 8 + (warp_id * 2) + 0] = res1;
out_ptr[out_offset + th_id * 8 + (warp_id * 2) + 1] = res2;
}
};
auto start_pipes = [&](int k_tile_id, int n_tile_id) {
#pragma unroll
for (int pipe = 0; pipe < repack_stages - 1; pipe++) {
fetch_to_shared(pipe, k_tile_id, n_tile_id + pipe);
}
wait_for_stage();
};
#pragma unroll
for (int k_tile_id = start_k_tile; k_tile_id < finish_k_tile; k_tile_id++) {
int n_tile_id = 0;
start_pipes(k_tile_id, n_tile_id);
while (n_tile_id < n_tiles) {
#pragma unroll
for (int pipe = 0; pipe < repack_stages; pipe++) {
fetch_to_shared((pipe + repack_stages - 1) % repack_stages, k_tile_id,
n_tile_id + pipe + repack_stages - 1);
repack_tile(pipe, k_tile_id, n_tile_id + pipe);
wait_for_stage();
}
n_tile_id += repack_stages;
}
}
}
} // namespace marlin
#define CALL_IF(NUM_BITS) \
else if (num_bits == NUM_BITS) { \
cudaFuncSetAttribute( \
marlin::awq_marlin_repack_kernel<marlin::repack_threads, NUM_BITS>, \
cudaFuncAttributeMaxDynamicSharedMemorySize, max_shared_mem); \
marlin::awq_marlin_repack_kernel<marlin::repack_threads, NUM_BITS> \
<<<blocks, marlin::repack_threads, max_shared_mem, stream>>>( \
b_q_weight_ptr, out_ptr, size_k, size_n); \
}
torch::Tensor awq_marlin_repack(torch::Tensor& b_q_weight, int64_t size_k,
int64_t size_n, int64_t num_bits) {
// Verify compatibility with marlin tile of 16x64
TORCH_CHECK(size_k % marlin::tile_k_size == 0, "size_k = ", size_k,
" is not divisible by tile_k_size = ", marlin::tile_k_size);
TORCH_CHECK(size_n % marlin::tile_n_size == 0, "size_n = ", size_n,
" is not divisible by tile_n_size = ", marlin::tile_n_size);
TORCH_CHECK(num_bits == 4 || num_bits == 8,
"num_bits must be 4 or 8. Got = ", num_bits);
int const pack_factor = 32 / num_bits;
// Verify B
TORCH_CHECK(b_q_weight.size(0) == size_k,
"b_q_weight.size(0) = ", b_q_weight.size(0),
" is not size_k = ", size_k);
TORCH_CHECK((size_n / pack_factor) == b_q_weight.size(1),
"Shape mismatch: b_q_weight.size(1) = ", b_q_weight.size(1),
", size_n = ", size_n, ", pack_factor = ", pack_factor);
// Verify device and strides
TORCH_CHECK(b_q_weight.device().is_cuda(), "b_q_weight is not on GPU");
TORCH_CHECK(b_q_weight.is_contiguous(), "b_q_weight is not contiguous");
TORCH_CHECK(b_q_weight.dtype() == at::kInt, "b_q_weight type is not kInt");
// Alloc buffers
const at::cuda::OptionalCUDAGuard device_guard(device_of(b_q_weight));
auto options = torch::TensorOptions()
.dtype(b_q_weight.dtype())
.device(b_q_weight.device());
torch::Tensor out = torch::empty(
{size_k / marlin::tile_size, size_n * marlin::tile_size / pack_factor},
options);
// Get ptrs
uint32_t const* b_q_weight_ptr =
reinterpret_cast<uint32_t const*>(b_q_weight.data_ptr());
uint32_t* out_ptr = reinterpret_cast<uint32_t*>(out.data_ptr());
// Get dev info
int dev = b_q_weight.get_device();
cudaStream_t stream = at::cuda::getCurrentCUDAStream(dev);
int blocks;
cudaDeviceGetAttribute(&blocks, cudaDevAttrMultiProcessorCount, dev);
int max_shared_mem = 0;
cudaDeviceGetAttribute(&max_shared_mem,
cudaDevAttrMaxSharedMemoryPerBlockOptin, dev);
TORCH_CHECK(max_shared_mem > 0);
if (false) {
}
CALL_IF(4)
CALL_IF(8)
else {
TORCH_CHECK(false, "Unsupported repack config: num_bits = ", num_bits);
}
return out;
}
#endif
...@@ -19,8 +19,9 @@ ...@@ -19,8 +19,9 @@
* Adapted from https://github.com/IST-DASLab/marlin * Adapted from https://github.com/IST-DASLab/marlin
*/ */
#include "gptq_marlin.cuh" #include "marlin.cuh"
#include "gptq_marlin_dtypes.cuh" #include "marlin_dtypes.cuh"
#include "core/scalar_type.hpp"
#define STATIC_ASSERT_SCALAR_TYPE_VALID(scalar_t) \ #define STATIC_ASSERT_SCALAR_TYPE_VALID(scalar_t) \
static_assert(std::is_same<scalar_t, half>::value || \ static_assert(std::is_same<scalar_t, half>::value || \
...@@ -32,7 +33,7 @@ inline std::string str(T x) { ...@@ -32,7 +33,7 @@ inline std::string str(T x) {
return std::to_string(x); return std::to_string(x);
} }
namespace gptq_marlin { namespace marlin {
#if defined(__CUDA_ARCH__) && __CUDA_ARCH__ < 800 #if defined(__CUDA_ARCH__) && __CUDA_ARCH__ < 800
...@@ -59,23 +60,27 @@ __global__ void Marlin( ...@@ -59,23 +60,27 @@ __global__ void Marlin(
const int4* __restrict__ A, // fp16 input matrix of shape mxk const int4* __restrict__ A, // fp16 input matrix of shape mxk
const int4* __restrict__ B, // 4bit quantized weight matrix of shape kxn const int4* __restrict__ B, // 4bit quantized weight matrix of shape kxn
int4* __restrict__ C, // fp16 output buffer of shape mxn int4* __restrict__ C, // fp16 output buffer of shape mxn
int4* __restrict__ C_tmp, // fp32 tmp output buffer (for reduce)
const int4* __restrict__ scales_ptr, // fp16 quantization scales of shape const int4* __restrict__ scales_ptr, // fp16 quantization scales of shape
// (k/groupsize)xn // (k/groupsize)xn
const int* __restrict__ g_idx, // int32 group indices of shape k const int* __restrict__ g_idx, // int32 group indices of shape k
int num_groups, // number of scale groups per output channel int num_groups, // number of scale groups per output channel
int prob_m, // batch dimension m int prob_m, // batch dimension m
int prob_n, // output dimension n int prob_n, // output dimension n
int prob_k, // reduction dimension k int prob_k, // reduction dimension k
int* locks // extra global storage for barrier synchronization int* locks, // extra global storage for barrier synchronization
bool use_fp32_reduce // whether to use fp32 global reduce
) {} ) {}
} // namespace gptq_marlin } // namespace marlin
torch::Tensor gptq_marlin_gemm(torch::Tensor& a, torch::Tensor& b_q_weight, torch::Tensor gptq_marlin_gemm(torch::Tensor& a, torch::Tensor& b_q_weight,
torch::Tensor& b_scales, torch::Tensor& g_idx, torch::Tensor& b_scales, torch::Tensor& b_zeros,
torch::Tensor& perm, torch::Tensor& workspace, torch::Tensor& g_idx, torch::Tensor& perm,
int64_t num_bits, int64_t size_m, int64_t size_n, torch::Tensor& workspace,
int64_t size_k, bool is_k_full) { vllm::ScalarTypeTorchPtr const& b_q_type,
int64_t size_m, int64_t size_n, int64_t size_k,
bool is_k_full, bool has_zp) {
TORCH_CHECK_NOT_IMPLEMENTED(false, TORCH_CHECK_NOT_IMPLEMENTED(false,
"marlin_gemm(..) requires CUDA_ARCH >= 8.0"); "marlin_gemm(..) requires CUDA_ARCH >= 8.0");
return torch::empty({1, 1}); return torch::empty({1, 1});
...@@ -264,6 +269,114 @@ dequant_8bit<nv_bfloat16>(int q) { ...@@ -264,6 +269,114 @@ dequant_8bit<nv_bfloat16>(int q) {
return frag_b; return frag_b;
} }
// Zero-point dequantizers
template <typename scalar_t>
__device__ inline typename ScalarType<scalar_t>::FragB dequant_4bit_zp(int q) {
STATIC_ASSERT_SCALAR_TYPE_VALID(scalar_t);
}
template <>
__device__ inline typename ScalarType<half>::FragB dequant_4bit_zp<half>(
int q) {
const int LO = 0x000f000f;
const int HI = 0x00f000f0;
const int EX = 0x64006400;
// Guarantee that the `(a & b) | c` operations are LOP3s.
int lo = lop3<(0xf0 & 0xcc) | 0xaa>(q, LO, EX);
int hi = lop3<(0xf0 & 0xcc) | 0xaa>(q, HI, EX);
const int SUB = 0x64006400;
const int MUL = 0x2c002c00;
const int ADD = 0xd400d400;
typename ScalarType<half>::FragB frag_b;
frag_b[0] = __hsub2(*reinterpret_cast<half2*>(&lo),
*reinterpret_cast<const half2*>(&SUB));
frag_b[1] = __hfma2(*reinterpret_cast<half2*>(&hi),
*reinterpret_cast<const half2*>(&MUL),
*reinterpret_cast<const half2*>(&ADD));
return frag_b;
}
template <>
__device__ inline typename ScalarType<nv_bfloat16>::FragB
dequant_4bit_zp<nv_bfloat16>(int q) {
static constexpr uint32_t MASK = 0x000f000f;
static constexpr uint32_t EX = 0x43004300;
// Guarantee that the `(a & b) | c` operations are LOP3s.
int lo = lop3<(0xf0 & 0xcc) | 0xaa>(q, MASK, EX);
q >>= 4;
int hi = lop3<(0xf0 & 0xcc) | 0xaa>(q, MASK, EX);
typename ScalarType<nv_bfloat16>::FragB frag_b;
static constexpr uint32_t MUL = 0x3F803F80;
static constexpr uint32_t ADD = 0xC300C300;
frag_b[0] = __hfma2(*reinterpret_cast<nv_bfloat162*>(&lo),
*reinterpret_cast<const nv_bfloat162*>(&MUL),
*reinterpret_cast<const nv_bfloat162*>(&ADD));
frag_b[1] = __hfma2(*reinterpret_cast<nv_bfloat162*>(&hi),
*reinterpret_cast<const nv_bfloat162*>(&MUL),
*reinterpret_cast<const nv_bfloat162*>(&ADD));
return frag_b;
}
template <typename scalar_t>
__device__ inline typename ScalarType<scalar_t>::FragB dequant_8bit_zp(int q) {
STATIC_ASSERT_SCALAR_TYPE_VALID(scalar_t);
}
template <>
__device__ inline typename ScalarType<half>::FragB dequant_8bit_zp<half>(
int q) {
static constexpr uint32_t mask_for_elt_01 = 0x5250;
static constexpr uint32_t mask_for_elt_23 = 0x5351;
static constexpr uint32_t start_byte_for_fp16 = 0x64646464;
uint32_t lo = prmt<start_byte_for_fp16, mask_for_elt_01>(q);
uint32_t hi = prmt<start_byte_for_fp16, mask_for_elt_23>(q);
static constexpr uint32_t I8s_TO_F16s_MAGIC_NUM = 0x64006400;
typename ScalarType<half>::FragB frag_b;
frag_b[0] = __hsub2(*reinterpret_cast<half2*>(&lo),
*reinterpret_cast<const half2*>(&I8s_TO_F16s_MAGIC_NUM));
frag_b[1] = __hsub2(*reinterpret_cast<half2*>(&hi),
*reinterpret_cast<const half2*>(&I8s_TO_F16s_MAGIC_NUM));
return frag_b;
}
template <>
__device__ inline typename ScalarType<nv_bfloat16>::FragB
dequant_8bit_zp<nv_bfloat16>(int q) {
typename ScalarType<nv_bfloat16>::FragB frag_b;
float fp32_intermediates[4];
uint32_t* fp32_intermediates_casted =
reinterpret_cast<uint32_t*>(fp32_intermediates);
static constexpr uint32_t fp32_base = 0x4B000000;
fp32_intermediates_casted[0] = __byte_perm(q, fp32_base, 0x7650);
fp32_intermediates_casted[1] = __byte_perm(q, fp32_base, 0x7652);
fp32_intermediates_casted[2] = __byte_perm(q, fp32_base, 0x7651);
fp32_intermediates_casted[3] = __byte_perm(q, fp32_base, 0x7653);
fp32_intermediates[0] -= 8388608.f;
fp32_intermediates[1] -= 8388608.f;
fp32_intermediates[2] -= 8388608.f;
fp32_intermediates[3] -= 8388608.f;
uint32_t* bf16_result_ptr = reinterpret_cast<uint32_t*>(&frag_b);
bf16_result_ptr[0] = __byte_perm(fp32_intermediates_casted[0],
fp32_intermediates_casted[1], 0x7632);
bf16_result_ptr[1] = __byte_perm(fp32_intermediates_casted[2],
fp32_intermediates_casted[3], 0x7632);
return frag_b;
}
// Multiply dequantized values by the corresponding quantization scale; used // Multiply dequantized values by the corresponding quantization scale; used
// only for grouped quantization. // only for grouped quantization.
template <typename scalar_t> template <typename scalar_t>
...@@ -277,6 +390,17 @@ __device__ inline void scale(typename ScalarType<scalar_t>::FragB& frag_b, ...@@ -277,6 +390,17 @@ __device__ inline void scale(typename ScalarType<scalar_t>::FragB& frag_b,
frag_b[1] = __hmul2(frag_b[1], s); frag_b[1] = __hmul2(frag_b[1], s);
} }
template <typename scalar_t>
__device__ inline void sub_zp(typename ScalarType<scalar_t>::FragB& frag_b,
typename ScalarType<scalar_t>::scalar_t2& frag_zp,
int i) {
using scalar_t2 = typename ScalarType<scalar_t>::scalar_t2;
scalar_t2 zp =
ScalarType<scalar_t>::num2num2(reinterpret_cast<scalar_t*>(&frag_zp)[i]);
frag_b[0] = __hsub2(frag_b[0], zp);
frag_b[1] = __hsub2(frag_b[1], zp);
}
// Same as above, but for act_order (each K is multiplied individually) // Same as above, but for act_order (each K is multiplied individually)
template <typename scalar_t> template <typename scalar_t>
__device__ inline void scale4(typename ScalarType<scalar_t>::FragB& frag_b, __device__ inline void scale4(typename ScalarType<scalar_t>::FragB& frag_b,
...@@ -404,6 +528,7 @@ template <typename scalar_t, // compute dtype, half or nv_float16 ...@@ -404,6 +528,7 @@ template <typename scalar_t, // compute dtype, half or nv_float16
const int stages, // number of stages for the async global->shared const int stages, // number of stages for the async global->shared
// fetch pipeline // fetch pipeline
const bool has_act_order, // whether act_order is enabled const bool has_act_order, // whether act_order is enabled
const bool has_zp, // whether zero-points are enabled
const int group_blocks = -1 // number of consecutive 16x16 blocks const int group_blocks = -1 // number of consecutive 16x16 blocks
// with a separate quantization scale // with a separate quantization scale
> >
...@@ -411,14 +536,18 @@ __global__ void Marlin( ...@@ -411,14 +536,18 @@ __global__ void Marlin(
const int4* __restrict__ A, // fp16 input matrix of shape mxk const int4* __restrict__ A, // fp16 input matrix of shape mxk
const int4* __restrict__ B, // 4bit quantized weight matrix of shape kxn const int4* __restrict__ B, // 4bit quantized weight matrix of shape kxn
int4* __restrict__ C, // fp16 output buffer of shape mxn int4* __restrict__ C, // fp16 output buffer of shape mxn
int4* __restrict__ C_tmp, // fp32 tmp output buffer (for reduce)
const int4* __restrict__ scales_ptr, // fp16 quantization scales of shape const int4* __restrict__ scales_ptr, // fp16 quantization scales of shape
// (k/groupsize)xn // (k/groupsize)xn
const int4* __restrict__ zp_ptr, // 4bit packed zero-points of shape
// (k/groupsize)x(n/pack_factor)
const int* __restrict__ g_idx, // int32 group indices of shape k const int* __restrict__ g_idx, // int32 group indices of shape k
int num_groups, // number of scale groups per output channel int num_groups, // number of scale groups per output channel
int prob_m, // batch dimension m int prob_m, // batch dimension m
int prob_n, // output dimension n int prob_n, // output dimension n
int prob_k, // reduction dimension k int prob_k, // reduction dimension k
int* locks // extra global storage for barrier synchronization int* locks, // extra global storage for barrier synchronization
bool use_fp32_reduce // whether to use fp32 global reduce
) { ) {
// Each threadblock processes one "stripe" of the B matrix with (roughly) the // Each threadblock processes one "stripe" of the B matrix with (roughly) the
// same size, which might involve multiple column "slices" (of width 16 * // same size, which might involve multiple column "slices" (of width 16 *
...@@ -437,6 +566,7 @@ __global__ void Marlin( ...@@ -437,6 +566,7 @@ __global__ void Marlin(
using FragB = typename ScalarType<scalar_t>::FragB; using FragB = typename ScalarType<scalar_t>::FragB;
using FragC = typename ScalarType<scalar_t>::FragC; using FragC = typename ScalarType<scalar_t>::FragC;
using FragS = typename ScalarType<scalar_t>::FragS; using FragS = typename ScalarType<scalar_t>::FragS;
using FragZP = typename ScalarType<scalar_t>::FragZP;
constexpr int pack_factor = 32 / num_bits; constexpr int pack_factor = 32 / num_bits;
...@@ -471,6 +601,8 @@ __global__ void Marlin( ...@@ -471,6 +601,8 @@ __global__ void Marlin(
int slice_idx; // index of threadblock in current slice; numbered bottom to int slice_idx; // index of threadblock in current slice; numbered bottom to
// top // top
int par_id = 0;
// We can easily implement parallel problem execution by just remapping // We can easily implement parallel problem execution by just remapping
// indices and advancing global pointers // indices and advancing global pointers
if (slice_col_par >= n_tiles) { if (slice_col_par >= n_tiles) {
...@@ -478,6 +610,7 @@ __global__ void Marlin( ...@@ -478,6 +610,7 @@ __global__ void Marlin(
C += (slice_col_par / n_tiles) * 16 * thread_m_blocks * prob_n / 8; C += (slice_col_par / n_tiles) * 16 * thread_m_blocks * prob_n / 8;
locks += (slice_col_par / n_tiles) * n_tiles; locks += (slice_col_par / n_tiles) * n_tiles;
slice_col = slice_col_par % n_tiles; slice_col = slice_col_par % n_tiles;
par_id = slice_col_par / n_tiles;
} }
// Compute all information about the current slice which is required for // Compute all information about the current slice which is required for
...@@ -508,6 +641,7 @@ __global__ void Marlin( ...@@ -508,6 +641,7 @@ __global__ void Marlin(
C += 16 * thread_m_blocks * prob_n / 8; C += 16 * thread_m_blocks * prob_n / 8;
locks += n_tiles; locks += n_tiles;
slice_col = 0; slice_col = 0;
par_id++;
} }
}; };
init_slice(); init_slice();
...@@ -566,6 +700,13 @@ __global__ void Marlin( ...@@ -566,6 +700,13 @@ __global__ void Marlin(
int tb_n_warps = thread_n_blocks / 4; int tb_n_warps = thread_n_blocks / 4;
int act_s_col_tb_stride = act_s_col_warp_stride * tb_n_warps; int act_s_col_tb_stride = act_s_col_warp_stride * tb_n_warps;
// Zero-points sizes/strides
int zp_gl_stride = (prob_n / pack_factor) / 4;
constexpr int zp_sh_stride = ((16 * thread_n_blocks) / pack_factor) / 4;
constexpr int zp_tb_groups = s_tb_groups;
constexpr int zp_sh_stage = has_zp ? zp_tb_groups * zp_sh_stride : 0;
int zp_gl_rd_delta = zp_gl_stride;
// Global A read index of current thread. // Global A read index of current thread.
int a_gl_rd = a_gl_stride * (threadIdx.x / a_gl_rd_delta_o) + int a_gl_rd = a_gl_stride * (threadIdx.x / a_gl_rd_delta_o) +
(threadIdx.x % a_gl_rd_delta_o); (threadIdx.x % a_gl_rd_delta_o);
...@@ -605,6 +746,19 @@ __global__ void Marlin( ...@@ -605,6 +746,19 @@ __global__ void Marlin(
int s_sh_wr = threadIdx.x; int s_sh_wr = threadIdx.x;
bool s_sh_wr_pred = threadIdx.x < s_sh_stride; bool s_sh_wr_pred = threadIdx.x < s_sh_stride;
// Zero-points
int zp_gl_rd;
if constexpr (has_zp) {
if constexpr (group_blocks == -1) {
zp_gl_rd = zp_sh_stride * slice_col + threadIdx.x;
} else {
zp_gl_rd = zp_gl_stride * ((thread_k_blocks * slice_row) / group_blocks) +
zp_sh_stride * slice_col + threadIdx.x;
}
}
int zp_sh_wr = threadIdx.x;
bool zp_sh_wr_pred = threadIdx.x < zp_sh_stride;
// We use a different scale layout for grouped and column-wise quantization as // We use a different scale layout for grouped and column-wise quantization as
// we scale a `half2` tile in column-major layout in the former and in // we scale a `half2` tile in column-major layout in the former and in
// row-major in the latter case. // row-major in the latter case.
...@@ -616,6 +770,18 @@ __global__ void Marlin( ...@@ -616,6 +770,18 @@ __global__ void Marlin(
s_sh_rd = 8 * ((threadIdx.x / 32) % (thread_n_blocks / 4)) + s_sh_rd = 8 * ((threadIdx.x / 32) % (thread_n_blocks / 4)) +
(threadIdx.x % 32) % 4; (threadIdx.x % 32) % 4;
// Zero-points have the same read layout as the scales
// (without column-wise case)
constexpr int num_col_threads = 8;
constexpr int num_row_threads = 4;
constexpr int num_ints_per_thread = 8 / pack_factor;
int zp_sh_rd;
if constexpr (has_zp) {
zp_sh_rd = num_ints_per_thread * num_col_threads *
((threadIdx.x / 32) % (thread_n_blocks / 4)) +
num_ints_per_thread * ((threadIdx.x % 32) / num_row_threads);
}
// Precompute which thread should not read memory in which iterations; this is // Precompute which thread should not read memory in which iterations; this is
// needed if there are more threads than required for a certain tilesize or // needed if there are more threads than required for a certain tilesize or
// when the batchsize is not a multiple of 16. // when the batchsize is not a multiple of 16.
...@@ -664,14 +830,17 @@ __global__ void Marlin( ...@@ -664,14 +830,17 @@ __global__ void Marlin(
int4* sh_a = sh; int4* sh_a = sh;
int4* sh_b = sh_a + (stages * a_sh_stage); int4* sh_b = sh_a + (stages * a_sh_stage);
int4* sh_g_idx = sh_b + (stages * b_sh_stage); int4* sh_g_idx = sh_b + (stages * b_sh_stage);
int4* sh_s = sh_g_idx + (stages * g_idx_stage); int4* sh_zp = sh_g_idx + (stages * g_idx_stage);
int4* sh_s = sh_zp + (stages * zp_sh_stage);
// Register storage for double buffer of shared memory reads. // Register storage for double buffer of shared memory reads.
FragA frag_a[2][thread_m_blocks]; FragA frag_a[2][thread_m_blocks];
I4 frag_b_quant[2][b_thread_vecs]; I4 frag_b_quant[2][b_thread_vecs];
FragC frag_c[thread_m_blocks][4][2]; FragC frag_c[thread_m_blocks][4][2];
FragS frag_s[2][4]; // No act-order FragS frag_s[2][4]; // No act-order
FragS act_frag_s[2][4][4]; // For act-order FragS act_frag_s[2][4][4]; // For act-order
int frag_qzp[2][num_ints_per_thread]; // Zero-points
FragZP frag_zp; // Zero-points in fp16
// Zero accumulators. // Zero accumulators.
auto zero_accums = [&]() { auto zero_accums = [&]() {
...@@ -777,6 +946,28 @@ __global__ void Marlin( ...@@ -777,6 +946,28 @@ __global__ void Marlin(
} }
} }
} }
if constexpr (has_zp && group_blocks != -1) {
int4* sh_zp_stage = sh_zp + zp_sh_stage * pipe;
if constexpr (group_blocks >= thread_k_blocks) {
// Only fetch zero-points if this tile starts a new group
if (pipe % (group_blocks / thread_k_blocks) == 0) {
if (zp_sh_wr_pred) {
cp_async4(&sh_zp_stage[zp_sh_wr], &zp_ptr[zp_gl_rd]);
}
zp_gl_rd += zp_gl_rd_delta;
}
} else {
for (int i = 0; i < zp_tb_groups; i++) {
if (zp_sh_wr_pred) {
cp_async4(&sh_zp_stage[i * zp_sh_stride + zp_sh_wr],
&zp_ptr[zp_gl_rd]);
}
zp_gl_rd += zp_gl_rd_delta;
}
}
}
} }
} }
// Insert a fence even when we are winding down the pipeline to ensure that // Insert a fence even when we are winding down the pipeline to ensure that
...@@ -784,6 +975,12 @@ __global__ void Marlin( ...@@ -784,6 +975,12 @@ __global__ void Marlin(
cp_async_fence(); cp_async_fence();
}; };
auto fetch_zp_to_shared = [&]() {
if (zp_sh_wr_pred) {
cp_async4(&sh_zp[zp_sh_wr], &zp_ptr[zp_gl_rd]);
}
};
// Wait until the next thread tile has been loaded to shared memory. // Wait until the next thread tile has been loaded to shared memory.
auto wait_for_stage = [&]() { auto wait_for_stage = [&]() {
// We only have `stages - 2` active fetches since we are double buffering // We only have `stages - 2` active fetches since we are double buffering
...@@ -932,8 +1129,82 @@ __global__ void Marlin( ...@@ -932,8 +1129,82 @@ __global__ void Marlin(
} }
}; };
auto fetch_zp_to_registers = [&](int k, int full_pipe) {
// This code does not handle group_blocks == 0,
// which signifies act_order.
// has_zp implies AWQ, which doesn't have act_order,
static_assert(!has_zp || group_blocks != 0);
if constexpr (has_zp) {
int pipe = full_pipe % stages;
if constexpr (group_blocks == -1) {
for (int i = 0; i < num_ints_per_thread; i++) {
frag_qzp[k % 2][i] = (reinterpret_cast<int*>(sh_zp))[zp_sh_rd + i];
}
} else if constexpr (group_blocks >= thread_k_blocks) {
int4* sh_zp_stage =
sh_zp + zp_sh_stage * ((group_blocks / thread_k_blocks) *
(pipe / (group_blocks / thread_k_blocks)));
for (int i = 0; i < num_ints_per_thread; i++) {
frag_qzp[k % 2][i] =
(reinterpret_cast<int*>(sh_zp_stage))[zp_sh_rd + i];
}
} else {
int warp_id = threadIdx.x / 32;
int n_warps = thread_n_blocks / 4;
int warp_row = warp_id / n_warps;
int cur_k = warp_row * 16;
cur_k += k_iter_size * (k % b_sh_wr_iters);
int k_blocks = cur_k / 16;
int cur_group_id = 0;
// Suppress bogus and persistent divide-by-zero warning
#pragma nv_diagnostic push
#pragma nv_diag_suppress divide_by_zero
cur_group_id = k_blocks / group_blocks;
#pragma nv_diagnostic pop
int4* sh_zp_stage = sh_zp + zp_sh_stage * pipe;
sh_zp_stage += cur_group_id * zp_sh_stride;
for (int i = 0; i < num_ints_per_thread; i++) {
frag_qzp[k % 2][i] =
(reinterpret_cast<int*>(sh_zp_stage))[zp_sh_rd + i];
}
}
}
};
// Execute the actual tensor core matmul of a sub-tile. // Execute the actual tensor core matmul of a sub-tile.
auto matmul = [&](int k) { auto matmul = [&](int k) {
if constexpr (has_zp) {
FragB frag_zp_0;
FragB frag_zp_1;
if constexpr (num_bits == 4) {
int zp_quant = frag_qzp[k % 2][0];
int zp_quant_shift = zp_quant >> 8;
frag_zp_0 = dequant_4bit_zp<scalar_t>(zp_quant);
frag_zp_1 = dequant_4bit_zp<scalar_t>(zp_quant_shift);
} else {
int zp_quant_0 = frag_qzp[k % 2][0];
int zp_quant_1 = frag_qzp[k % 2][1];
frag_zp_0 = dequant_8bit_zp<scalar_t>(zp_quant_0);
frag_zp_1 = dequant_8bit_zp<scalar_t>(zp_quant_1);
}
frag_zp[0] = frag_zp_0[0];
frag_zp[1] = frag_zp_0[1];
frag_zp[2] = frag_zp_1[0];
frag_zp[3] = frag_zp_1[1];
}
// We have the m dimension as the inner loop in order to encourage overlapping // We have the m dimension as the inner loop in order to encourage overlapping
// dequantization and matmul operations. // dequantization and matmul operations.
#pragma unroll #pragma unroll
...@@ -944,16 +1215,32 @@ __global__ void Marlin( ...@@ -944,16 +1215,32 @@ __global__ void Marlin(
int b_quant = frag_b_quant[k % 2][0][j]; int b_quant = frag_b_quant[k % 2][0][j];
int b_quant_shift = b_quant >> 8; int b_quant_shift = b_quant >> 8;
frag_b0 = dequant_4bit<scalar_t>(b_quant); if constexpr (has_zp) {
frag_b1 = dequant_4bit<scalar_t>(b_quant_shift); frag_b0 = dequant_4bit_zp<scalar_t>(b_quant);
frag_b1 = dequant_4bit_zp<scalar_t>(b_quant_shift);
} else {
frag_b0 = dequant_4bit<scalar_t>(b_quant);
frag_b1 = dequant_4bit<scalar_t>(b_quant_shift);
}
} else { } else {
int* frag_b_quant_ptr = reinterpret_cast<int*>(frag_b_quant[k % 2]); int* frag_b_quant_ptr = reinterpret_cast<int*>(frag_b_quant[k % 2]);
int b_quant_0 = frag_b_quant_ptr[j * 2 + 0]; int b_quant_0 = frag_b_quant_ptr[j * 2 + 0];
int b_quant_1 = frag_b_quant_ptr[j * 2 + 1]; int b_quant_1 = frag_b_quant_ptr[j * 2 + 1];
frag_b0 = dequant_8bit<scalar_t>(b_quant_0); if constexpr (has_zp) {
frag_b1 = dequant_8bit<scalar_t>(b_quant_1); frag_b0 = dequant_8bit_zp<scalar_t>(b_quant_0);
frag_b1 = dequant_8bit_zp<scalar_t>(b_quant_1);
} else {
frag_b0 = dequant_8bit<scalar_t>(b_quant_0);
frag_b1 = dequant_8bit<scalar_t>(b_quant_1);
}
}
// Apply zero-point to frag_b0
if constexpr (has_zp) {
sub_zp<scalar_t>(frag_b0, frag_zp[j], 0);
} }
// Apply scale to frag_b0 // Apply scale to frag_b0
...@@ -967,6 +1254,11 @@ __global__ void Marlin( ...@@ -967,6 +1254,11 @@ __global__ void Marlin(
} }
} }
// Apply zero-point to frag_b1
if constexpr (has_zp) {
sub_zp<scalar_t>(frag_b1, frag_zp[j], 1);
}
// Apply scale to frag_b1 // Apply scale to frag_b1
if constexpr (has_act_order) { if constexpr (has_act_order) {
scale4<scalar_t>(frag_b1, act_frag_s[k % 2][0][j], scale4<scalar_t>(frag_b1, act_frag_s[k % 2][0][j],
...@@ -1048,7 +1340,7 @@ __global__ void Marlin( ...@@ -1048,7 +1340,7 @@ __global__ void Marlin(
// finally have to globally reduce over the results. As the striped // finally have to globally reduce over the results. As the striped
// partitioning minimizes the number of such reductions and our outputs are // partitioning minimizes the number of such reductions and our outputs are
// usually rather small, we perform this reduction serially in L2 cache. // usually rather small, we perform this reduction serially in L2 cache.
auto global_reduce = [&](bool first = false, bool last = false) { auto global_reduce_fp16 = [&](bool first = false, bool last = false) {
// We are very careful here to reduce directly in the output buffer to // We are very careful here to reduce directly in the output buffer to
// maximize L2 cache utilization in this step. To do this, we write out // maximize L2 cache utilization in this step. To do this, we write out
// results in FP16 (but still reduce with FP32 compute). // results in FP16 (but still reduce with FP32 compute).
...@@ -1109,6 +1401,53 @@ __global__ void Marlin( ...@@ -1109,6 +1401,53 @@ __global__ void Marlin(
} }
}; };
// Globally reduce over threadblocks that compute the same column block.
// We use a tmp C buffer to reduce in full fp32 precision.
auto global_reduce_fp32 = [&](bool first = false, bool last = false) {
constexpr int tb_m = thread_m_blocks * 16;
constexpr int tb_n = thread_n_blocks * 16;
constexpr int c_size = tb_m * tb_n * sizeof(float) / 16;
constexpr int active_threads = 32 * thread_n_blocks / 4;
bool is_th_active = threadIdx.x < active_threads;
int par_offset = c_size * n_tiles * par_id;
int slice_offset = c_size * slice_col;
constexpr int num_floats = thread_m_blocks * 4 * 2 * 4;
constexpr int th_size = num_floats * sizeof(float) / 16;
int c_cur_offset = par_offset + slice_offset;
if (!is_th_active) {
return;
}
if (!first) {
float* frag_c_ptr = reinterpret_cast<float*>(&frag_c);
#pragma unroll
for (int k = 0; k < th_size; k++) {
sh[threadIdx.x] =
C_tmp[c_cur_offset + active_threads * k + threadIdx.x];
float* sh_c_ptr = reinterpret_cast<float*>(&sh[threadIdx.x]);
#pragma unroll
for (int f = 0; f < 4; f++) {
frag_c_ptr[k * 4 + f] += sh_c_ptr[f];
}
}
}
if (!last) {
int4* frag_c_ptr = reinterpret_cast<int4*>(&frag_c);
#pragma unroll
for (int k = 0; k < th_size; k++) {
C_tmp[c_cur_offset + active_threads * k + threadIdx.x] = frag_c_ptr[k];
}
}
};
// Write out the reduce final result in the correct layout. We only actually // Write out the reduce final result in the correct layout. We only actually
// reshuffle matrix fragments in this step, the reduction above is performed // reshuffle matrix fragments in this step, the reduction above is performed
// in fragment layout. // in fragment layout.
...@@ -1189,6 +1528,12 @@ __global__ void Marlin( ...@@ -1189,6 +1528,12 @@ __global__ void Marlin(
} }
fetch_scales_to_shared(true, g_idx[slice_k_start], g_idx[last_g_idx]); fetch_scales_to_shared(true, g_idx[slice_k_start], g_idx[last_g_idx]);
} }
if constexpr (has_zp && group_blocks == -1) {
if (i == 0) {
fetch_zp_to_shared();
}
}
fetch_to_shared(i, i, i < slice_iters); fetch_to_shared(i, i, i < slice_iters);
} }
...@@ -1197,6 +1542,7 @@ __global__ void Marlin( ...@@ -1197,6 +1542,7 @@ __global__ void Marlin(
init_same_group(0); init_same_group(0);
fetch_to_registers(0, 0); fetch_to_registers(0, 0);
fetch_scales_to_registers(0, 0); fetch_scales_to_registers(0, 0);
fetch_zp_to_registers(0, 0);
a_gl_rd += a_gl_rd_delta_o * (stages - 1); a_gl_rd += a_gl_rd_delta_o * (stages - 1);
slice_k_start_shared_fetch += tb_k * (stages - 1); slice_k_start_shared_fetch += tb_k * (stages - 1);
}; };
...@@ -1217,6 +1563,7 @@ __global__ void Marlin( ...@@ -1217,6 +1563,7 @@ __global__ void Marlin(
for (int k = 0; k < b_sh_wr_iters; k++) { for (int k = 0; k < b_sh_wr_iters; k++) {
fetch_to_registers(k + 1, pipe % stages); fetch_to_registers(k + 1, pipe % stages);
fetch_scales_to_registers(k + 1, pipe); fetch_scales_to_registers(k + 1, pipe);
fetch_zp_to_registers(k + 1, pipe);
if (k == b_sh_wr_iters - 2) { if (k == b_sh_wr_iters - 2) {
fetch_to_shared((pipe + stages - 1) % stages, pipe, fetch_to_shared((pipe + stages - 1) % stages, pipe,
slice_iters >= stages); slice_iters >= stages);
...@@ -1325,7 +1672,11 @@ __global__ void Marlin( ...@@ -1325,7 +1672,11 @@ __global__ void Marlin(
if (slice_count > 1) { // only globally reduce if there is more than one if (slice_count > 1) { // only globally reduce if there is more than one
// block in a slice // block in a slice
barrier_acquire(&locks[slice_col], slice_idx); barrier_acquire(&locks[slice_col], slice_idx);
global_reduce(slice_idx == 0, last); if (use_fp32_reduce) {
global_reduce_fp32(slice_idx == 0, last);
} else {
global_reduce_fp16(slice_idx == 0, last);
}
barrier_release(&locks[slice_col], last); barrier_release(&locks[slice_col], last);
} }
if (last) // only the last block in a slice actually writes the result if (last) // only the last block in a slice actually writes the result
...@@ -1354,6 +1705,7 @@ __global__ void Marlin( ...@@ -1354,6 +1705,7 @@ __global__ void Marlin(
} else { } else {
s_gl_rd = s_sh_stride * slice_col + threadIdx.x; s_gl_rd = s_sh_stride * slice_col + threadIdx.x;
zp_gl_rd = zp_sh_stride * slice_col + threadIdx.x;
} }
start_pipes(); start_pipes();
...@@ -1363,22 +1715,24 @@ __global__ void Marlin( ...@@ -1363,22 +1715,24 @@ __global__ void Marlin(
} }
#define __CALL_IF(NUM_BITS, THREAD_M_BLOCKS, THREAD_N_BLOCKS, \ #define __CALL_IF(NUM_BITS, THREAD_M_BLOCKS, THREAD_N_BLOCKS, \
THREAD_K_BLOCKS, HAS_ACT_ORDER, GROUP_BLOCKS, NUM_THREADS) \ THREAD_K_BLOCKS, HAS_ACT_ORDER, HAS_ZP, GROUP_BLOCKS, \
NUM_THREADS) \
else if (num_bits == NUM_BITS && thread_m_blocks == THREAD_M_BLOCKS && \ else if (num_bits == NUM_BITS && thread_m_blocks == THREAD_M_BLOCKS && \
thread_n_blocks == THREAD_N_BLOCKS && \ thread_n_blocks == THREAD_N_BLOCKS && \
thread_k_blocks == THREAD_K_BLOCKS && \ thread_k_blocks == THREAD_K_BLOCKS && \
has_act_order == HAS_ACT_ORDER && group_blocks == GROUP_BLOCKS && \ has_act_order == HAS_ACT_ORDER && has_zp == HAS_ZP && \
num_threads == NUM_THREADS) { \ group_blocks == GROUP_BLOCKS && num_threads == NUM_THREADS) { \
cudaFuncSetAttribute( \ cudaFuncSetAttribute( \
Marlin<scalar_t, NUM_BITS, NUM_THREADS, THREAD_M_BLOCKS, \ Marlin<scalar_t, NUM_BITS, NUM_THREADS, THREAD_M_BLOCKS, \
THREAD_N_BLOCKS, THREAD_K_BLOCKS, pipe_stages, HAS_ACT_ORDER, \ THREAD_N_BLOCKS, THREAD_K_BLOCKS, pipe_stages, HAS_ACT_ORDER, \
GROUP_BLOCKS>, \ HAS_ZP, GROUP_BLOCKS>, \
cudaFuncAttributeMaxDynamicSharedMemorySize, max_shared_mem); \ cudaFuncAttributeMaxDynamicSharedMemorySize, max_shared_mem); \
Marlin<scalar_t, NUM_BITS, NUM_THREADS, THREAD_M_BLOCKS, \ Marlin<scalar_t, NUM_BITS, NUM_THREADS, THREAD_M_BLOCKS, \
THREAD_N_BLOCKS, THREAD_K_BLOCKS, pipe_stages, HAS_ACT_ORDER, \ THREAD_N_BLOCKS, THREAD_K_BLOCKS, pipe_stages, HAS_ACT_ORDER, \
GROUP_BLOCKS><<<blocks, NUM_THREADS, max_shared_mem, stream>>>( \ HAS_ZP, GROUP_BLOCKS> \
A_ptr, B_ptr, C_ptr, s_ptr, g_idx_ptr, num_groups, prob_m, prob_n, \ <<<blocks, NUM_THREADS, max_shared_mem, stream>>>( \
prob_k, locks); \ A_ptr, B_ptr, C_ptr, C_tmp_ptr, s_ptr, zp_ptr, g_idx_ptr, \
num_groups, prob_m, prob_n, prob_k, locks, use_fp32_reduce); \
} }
typedef struct { typedef struct {
...@@ -1517,6 +1871,27 @@ bool is_valid_config(thread_config_t const& th_config, int max_m_blocks, ...@@ -1517,6 +1871,27 @@ bool is_valid_config(thread_config_t const& th_config, int max_m_blocks,
return true; return true;
} }
int determine_reduce_max_m(int prob_m, int max_par) {
constexpr int tile_m_size = 16;
if (prob_m <= tile_m_size) {
return tile_m_size;
} else if (prob_m <= tile_m_size * 2) {
return tile_m_size * 2;
} else if (prob_m <= tile_m_size * 3) {
return tile_m_size * 3;
} else if (prob_m <= tile_m_size * 4) {
return tile_m_size * 4;
} else {
int cur_par = min(div_ceil(prob_m, tile_m_size * 4), max_par);
return tile_m_size * 4 * cur_par;
}
}
exec_config_t determine_thread_config(int prob_m, int prob_n, int prob_k, exec_config_t determine_thread_config(int prob_m, int prob_n, int prob_k,
int num_bits, int group_size, int num_bits, int group_size,
bool has_act_order, bool is_k_full, bool has_act_order, bool is_k_full,
...@@ -1548,44 +1923,77 @@ exec_config_t determine_thread_config(int prob_m, int prob_n, int prob_k, ...@@ -1548,44 +1923,77 @@ exec_config_t determine_thread_config(int prob_m, int prob_n, int prob_k,
return exec_config_t{0, {-1, -1, -1}}; return exec_config_t{0, {-1, -1, -1}};
} }
#define CALL_IF(NUM_BITS, N_BLOCKS, K_BLOCKS, NUM_THREADS) \ #define GPTQ_CALL_IF(NUM_BITS, N_BLOCKS, K_BLOCKS, NUM_THREADS) \
__CALL_IF(NUM_BITS, 1, N_BLOCKS, K_BLOCKS, true, 0, NUM_THREADS) \ __CALL_IF(NUM_BITS, 1, N_BLOCKS, K_BLOCKS, true, false, 0, NUM_THREADS) \
__CALL_IF(NUM_BITS, 2, N_BLOCKS, K_BLOCKS, true, 0, NUM_THREADS) \ __CALL_IF(NUM_BITS, 2, N_BLOCKS, K_BLOCKS, true, false, 0, NUM_THREADS) \
__CALL_IF(NUM_BITS, 3, N_BLOCKS, K_BLOCKS, true, 0, NUM_THREADS) \ __CALL_IF(NUM_BITS, 3, N_BLOCKS, K_BLOCKS, true, false, 0, NUM_THREADS) \
__CALL_IF(NUM_BITS, 4, N_BLOCKS, K_BLOCKS, true, 0, NUM_THREADS) \ __CALL_IF(NUM_BITS, 4, N_BLOCKS, K_BLOCKS, true, false, 0, NUM_THREADS) \
\ \
__CALL_IF(NUM_BITS, 1, N_BLOCKS, K_BLOCKS, false, -1, NUM_THREADS) \ __CALL_IF(NUM_BITS, 1, N_BLOCKS, K_BLOCKS, false, false, -1, NUM_THREADS) \
__CALL_IF(NUM_BITS, 1, N_BLOCKS, K_BLOCKS, false, 2, NUM_THREADS) \ __CALL_IF(NUM_BITS, 1, N_BLOCKS, K_BLOCKS, false, false, 2, NUM_THREADS) \
__CALL_IF(NUM_BITS, 1, N_BLOCKS, K_BLOCKS, false, 4, NUM_THREADS) \ __CALL_IF(NUM_BITS, 1, N_BLOCKS, K_BLOCKS, false, false, 4, NUM_THREADS) \
__CALL_IF(NUM_BITS, 1, N_BLOCKS, K_BLOCKS, false, 8, NUM_THREADS) \ __CALL_IF(NUM_BITS, 1, N_BLOCKS, K_BLOCKS, false, false, 8, NUM_THREADS) \
\ \
__CALL_IF(NUM_BITS, 2, N_BLOCKS, K_BLOCKS, false, -1, NUM_THREADS) \ __CALL_IF(NUM_BITS, 2, N_BLOCKS, K_BLOCKS, false, false, -1, NUM_THREADS) \
__CALL_IF(NUM_BITS, 2, N_BLOCKS, K_BLOCKS, false, 2, NUM_THREADS) \ __CALL_IF(NUM_BITS, 2, N_BLOCKS, K_BLOCKS, false, false, 2, NUM_THREADS) \
__CALL_IF(NUM_BITS, 2, N_BLOCKS, K_BLOCKS, false, 4, NUM_THREADS) \ __CALL_IF(NUM_BITS, 2, N_BLOCKS, K_BLOCKS, false, false, 4, NUM_THREADS) \
__CALL_IF(NUM_BITS, 2, N_BLOCKS, K_BLOCKS, false, 8, NUM_THREADS) \ __CALL_IF(NUM_BITS, 2, N_BLOCKS, K_BLOCKS, false, false, 8, NUM_THREADS) \
\ \
__CALL_IF(NUM_BITS, 3, N_BLOCKS, K_BLOCKS, false, -1, NUM_THREADS) \ __CALL_IF(NUM_BITS, 3, N_BLOCKS, K_BLOCKS, false, false, -1, NUM_THREADS) \
__CALL_IF(NUM_BITS, 3, N_BLOCKS, K_BLOCKS, false, 2, NUM_THREADS) \ __CALL_IF(NUM_BITS, 3, N_BLOCKS, K_BLOCKS, false, false, 2, NUM_THREADS) \
__CALL_IF(NUM_BITS, 3, N_BLOCKS, K_BLOCKS, false, 4, NUM_THREADS) \ __CALL_IF(NUM_BITS, 3, N_BLOCKS, K_BLOCKS, false, false, 4, NUM_THREADS) \
__CALL_IF(NUM_BITS, 3, N_BLOCKS, K_BLOCKS, false, 8, NUM_THREADS) \ __CALL_IF(NUM_BITS, 3, N_BLOCKS, K_BLOCKS, false, false, 8, NUM_THREADS) \
\ \
__CALL_IF(NUM_BITS, 4, N_BLOCKS, K_BLOCKS, false, -1, NUM_THREADS) \ __CALL_IF(NUM_BITS, 4, N_BLOCKS, K_BLOCKS, false, false, -1, NUM_THREADS) \
__CALL_IF(NUM_BITS, 4, N_BLOCKS, K_BLOCKS, false, 2, NUM_THREADS) \ __CALL_IF(NUM_BITS, 4, N_BLOCKS, K_BLOCKS, false, false, 2, NUM_THREADS) \
__CALL_IF(NUM_BITS, 4, N_BLOCKS, K_BLOCKS, false, 4, NUM_THREADS) \ __CALL_IF(NUM_BITS, 4, N_BLOCKS, K_BLOCKS, false, false, 4, NUM_THREADS) \
__CALL_IF(NUM_BITS, 4, N_BLOCKS, K_BLOCKS, false, 8, NUM_THREADS) __CALL_IF(NUM_BITS, 4, N_BLOCKS, K_BLOCKS, false, false, 8, NUM_THREADS)
#define AWQ_CALL_IF(NUM_BITS, N_BLOCKS, K_BLOCKS, NUM_THREADS) \
__CALL_IF(NUM_BITS, 1, N_BLOCKS, K_BLOCKS, false, true, -1, NUM_THREADS) \
__CALL_IF(NUM_BITS, 1, N_BLOCKS, K_BLOCKS, false, true, 2, NUM_THREADS) \
__CALL_IF(NUM_BITS, 1, N_BLOCKS, K_BLOCKS, false, true, 4, NUM_THREADS) \
__CALL_IF(NUM_BITS, 1, N_BLOCKS, K_BLOCKS, false, true, 8, NUM_THREADS) \
\
__CALL_IF(NUM_BITS, 2, N_BLOCKS, K_BLOCKS, false, true, -1, NUM_THREADS) \
__CALL_IF(NUM_BITS, 2, N_BLOCKS, K_BLOCKS, false, true, 2, NUM_THREADS) \
__CALL_IF(NUM_BITS, 2, N_BLOCKS, K_BLOCKS, false, true, 4, NUM_THREADS) \
__CALL_IF(NUM_BITS, 2, N_BLOCKS, K_BLOCKS, false, true, 8, NUM_THREADS) \
\
__CALL_IF(NUM_BITS, 3, N_BLOCKS, K_BLOCKS, false, true, -1, NUM_THREADS) \
__CALL_IF(NUM_BITS, 3, N_BLOCKS, K_BLOCKS, false, true, 2, NUM_THREADS) \
__CALL_IF(NUM_BITS, 3, N_BLOCKS, K_BLOCKS, false, true, 4, NUM_THREADS) \
__CALL_IF(NUM_BITS, 3, N_BLOCKS, K_BLOCKS, false, true, 8, NUM_THREADS) \
\
__CALL_IF(NUM_BITS, 4, N_BLOCKS, K_BLOCKS, false, true, -1, NUM_THREADS) \
__CALL_IF(NUM_BITS, 4, N_BLOCKS, K_BLOCKS, false, true, 2, NUM_THREADS) \
__CALL_IF(NUM_BITS, 4, N_BLOCKS, K_BLOCKS, false, true, 4, NUM_THREADS) \
__CALL_IF(NUM_BITS, 4, N_BLOCKS, K_BLOCKS, false, true, 8, NUM_THREADS)
template <typename scalar_t> template <typename scalar_t>
void marlin_mm_f16i4(const void* A, const void* B, void* C, void* s, void marlin_mm(const void* A, const void* B, void* C, void* C_tmp, void* s,
void* g_idx, void* perm, void* a_tmp, int prob_m, void* zp, void* g_idx, void* perm, void* a_tmp, int prob_m,
int prob_n, int prob_k, void* workspace, int num_bits, int prob_n, int prob_k, void* workspace,
bool has_act_order, bool is_k_full, int num_groups, vllm::ScalarType const& q_type, bool has_act_order,
int group_size, int dev, cudaStream_t stream, int thread_k, bool is_k_full, bool has_zp, int num_groups, int group_size,
int thread_n, int sms, int max_par) { int dev, cudaStream_t stream, int thread_k, int thread_n,
TORCH_CHECK(num_bits == 4 || num_bits == 8, int sms, int max_par, bool use_fp32_reduce) {
"num_bits must be 4 or 8. Got = ", num_bits); if (has_zp) {
TORCH_CHECK(
q_type == vllm::kU4 || q_type == vllm::kU8,
"q_type must be u4 or u8 when has_zp = True. Got = ", q_type.str());
} else {
TORCH_CHECK(
q_type == vllm::kU4B8 || q_type == vllm::kU8B128,
"q_type must be uint4b8 or uint8b128 when has_zp = False. Got = ",
q_type.str());
}
TORCH_CHECK(prob_m > 0 && prob_n > 0 && prob_k > 0, "Invalid MNK = [", prob_m, TORCH_CHECK(prob_m > 0 && prob_n > 0 && prob_k > 0, "Invalid MNK = [", prob_m,
", ", prob_n, ", ", prob_k, "]"); ", ", prob_n, ", ", prob_k, "]");
// TODO: remove alias when we start supporting other 8bit types
int num_bits = q_type.size_bits();
int tot_m = prob_m; int tot_m = prob_m;
int tot_m_blocks = div_ceil(tot_m, 16); int tot_m_blocks = div_ceil(tot_m, 16);
int pad = 16 * tot_m_blocks - tot_m; int pad = 16 * tot_m_blocks - tot_m;
...@@ -1664,7 +2072,9 @@ void marlin_mm_f16i4(const void* A, const void* B, void* C, void* s, ...@@ -1664,7 +2072,9 @@ void marlin_mm_f16i4(const void* A, const void* B, void* C, void* s,
const int4* A_ptr = (const int4*)A; const int4* A_ptr = (const int4*)A;
const int4* B_ptr = (const int4*)B; const int4* B_ptr = (const int4*)B;
int4* C_ptr = (int4*)C; int4* C_ptr = (int4*)C;
int4* C_tmp_ptr = (int4*)C_tmp;
const int4* s_ptr = (const int4*)s; const int4* s_ptr = (const int4*)s;
const int4* zp_ptr = (const int4*)zp;
const int* g_idx_ptr = (const int*)g_idx; const int* g_idx_ptr = (const int*)g_idx;
const int* perm_ptr = (const int*)perm; const int* perm_ptr = (const int*)perm;
int4* a_tmp_ptr = (int4*)a_tmp; int4* a_tmp_ptr = (int4*)a_tmp;
...@@ -1701,28 +2111,33 @@ void marlin_mm_f16i4(const void* A, const void* B, void* C, void* s, ...@@ -1701,28 +2111,33 @@ void marlin_mm_f16i4(const void* A, const void* B, void* C, void* s,
thread_m_blocks = exec_cfg.max_m_blocks; thread_m_blocks = exec_cfg.max_m_blocks;
} }
// Define kernel configurations
if (false) { if (false) {
} }
CALL_IF(4, 32, 2, 256) GPTQ_CALL_IF(4, 16, 4, 256)
CALL_IF(4, 16, 4, 256) GPTQ_CALL_IF(4, 8, 8, 256)
CALL_IF(4, 8, 8, 256) GPTQ_CALL_IF(4, 8, 4, 128)
CALL_IF(4, 8, 4, 128) GPTQ_CALL_IF(4, 4, 8, 128)
CALL_IF(4, 4, 8, 128) GPTQ_CALL_IF(8, 16, 4, 256)
CALL_IF(8, 32, 2, 256) GPTQ_CALL_IF(8, 8, 8, 256)
CALL_IF(8, 16, 4, 256) GPTQ_CALL_IF(8, 8, 4, 128)
CALL_IF(8, 8, 8, 256) GPTQ_CALL_IF(8, 4, 8, 128)
CALL_IF(8, 8, 4, 128)
CALL_IF(8, 4, 8, 128) AWQ_CALL_IF(4, 16, 4, 256)
AWQ_CALL_IF(4, 8, 8, 256)
AWQ_CALL_IF(4, 8, 4, 128)
AWQ_CALL_IF(4, 4, 8, 128)
AWQ_CALL_IF(8, 16, 4, 256)
AWQ_CALL_IF(8, 8, 8, 256)
AWQ_CALL_IF(8, 8, 4, 128)
AWQ_CALL_IF(8, 4, 8, 128)
else { else {
TORCH_CHECK(false, "Unsupported shapes: MNK = [" + str(prob_m) + ", " + TORCH_CHECK(false, "Unsupported shapes: MNK = [", prob_m, ", ", prob_n,
str(prob_n) + ", " + str(prob_k) + "]" + ", ", prob_k, "]", ", has_act_order = ", has_act_order,
", has_act_order = " + str(has_act_order) + ", num_groups = ", num_groups, ", group_size = ", group_size,
", num_groups = " + str(num_groups) + ", thread_m_blocks = ", thread_m_blocks,
", group_size = " + str(group_size) + ", thread_n_blocks = ", thread_n_blocks,
", thread_m_blocks = " + str(thread_m_blocks) + ", thread_k_blocks = ", thread_k_blocks,
", thread_n_blocks = " + str(thread_n_blocks) + ", num_bits = ", num_bits);
", thread_k_blocks = " + str(thread_k_blocks));
} }
A_ptr += 16 * thread_m_blocks * (prob_k / 8) * par; A_ptr += 16 * thread_m_blocks * (prob_k / 8) * par;
...@@ -1730,17 +2145,28 @@ void marlin_mm_f16i4(const void* A, const void* B, void* C, void* s, ...@@ -1730,17 +2145,28 @@ void marlin_mm_f16i4(const void* A, const void* B, void* C, void* s,
} }
} }
} // namespace gptq_marlin } // namespace marlin
torch::Tensor gptq_marlin_gemm(torch::Tensor& a, torch::Tensor& b_q_weight, torch::Tensor gptq_marlin_gemm(torch::Tensor& a, torch::Tensor& b_q_weight,
torch::Tensor& b_scales, torch::Tensor& g_idx, torch::Tensor& b_scales, torch::Tensor& b_zeros,
torch::Tensor& perm, torch::Tensor& workspace, torch::Tensor& g_idx, torch::Tensor& perm,
int64_t num_bits, int64_t size_m, int64_t size_n, torch::Tensor& workspace,
int64_t size_k, bool is_k_full) { vllm::ScalarTypeTorchPtr const& b_q_type,
// Verify num_bits int64_t size_m, int64_t size_n, int64_t size_k,
TORCH_CHECK(num_bits == 4 || num_bits == 8, bool is_k_full, bool has_zp,
"num_bits must be 4 or 8. Got = ", num_bits); bool use_fp32_reduce) {
int pack_factor = 32 / num_bits; if (has_zp) {
TORCH_CHECK(*b_q_type == vllm::kU4 || *b_q_type == vllm::kU8,
"b_q_type must be u4 or u8 when has_zp = True. Got = ",
b_q_type->str());
} else {
TORCH_CHECK(
*b_q_type == vllm::kU4B8 || *b_q_type == vllm::kU8B128,
"b_q_type must be uint4b8 or uint8b128 when has_zp = False. Got = ",
b_q_type->str());
}
int pack_factor = 32 / b_q_type->size_bits();
// Verify A // Verify A
TORCH_CHECK(a.size(0) == size_m, "Shape mismatch: a.size(0) = ", a.size(0), TORCH_CHECK(a.size(0) == size_m, "Shape mismatch: a.size(0) = ", a.size(0),
...@@ -1749,16 +2175,15 @@ torch::Tensor gptq_marlin_gemm(torch::Tensor& a, torch::Tensor& b_q_weight, ...@@ -1749,16 +2175,15 @@ torch::Tensor gptq_marlin_gemm(torch::Tensor& a, torch::Tensor& b_q_weight,
", size_k = ", size_k); ", size_k = ", size_k);
// Verify B // Verify B
TORCH_CHECK(size_k % gptq_marlin::tile_size == 0, "size_k = ", size_k, TORCH_CHECK(size_k % marlin::tile_size == 0, "size_k = ", size_k,
" is not divisible by tile_size = ", gptq_marlin::tile_size); " is not divisible by tile_size = ", marlin::tile_size);
TORCH_CHECK((size_k / gptq_marlin::tile_size) == b_q_weight.size(0), TORCH_CHECK((size_k / marlin::tile_size) == b_q_weight.size(0),
"Shape mismatch: b_q_weight.size(0) = ", b_q_weight.size(0), "Shape mismatch: b_q_weight.size(0) = ", b_q_weight.size(0),
", size_k = ", size_k, ", tile_size = ", gptq_marlin::tile_size); ", size_k = ", size_k, ", tile_size = ", marlin::tile_size);
TORCH_CHECK(b_q_weight.size(1) % gptq_marlin::tile_size == 0, TORCH_CHECK(b_q_weight.size(1) % marlin::tile_size == 0,
"b_q_weight.size(1) = ", b_q_weight.size(1), "b_q_weight.size(1) = ", b_q_weight.size(1),
" is not divisible by tile_size = ", gptq_marlin::tile_size); " is not divisible by tile_size = ", marlin::tile_size);
int actual_size_n = int actual_size_n = (b_q_weight.size(1) / marlin::tile_size) * pack_factor;
(b_q_weight.size(1) / gptq_marlin::tile_size) * pack_factor;
TORCH_CHECK(size_n == actual_size_n, "size_n = ", size_n, TORCH_CHECK(size_n == actual_size_n, "size_n = ", size_n,
", actual_size_n = ", actual_size_n); ", actual_size_n = ", actual_size_n);
...@@ -1772,6 +2197,9 @@ torch::Tensor gptq_marlin_gemm(torch::Tensor& a, torch::Tensor& b_q_weight, ...@@ -1772,6 +2197,9 @@ torch::Tensor gptq_marlin_gemm(torch::Tensor& a, torch::Tensor& b_q_weight,
TORCH_CHECK(b_scales.device().is_cuda(), "b_scales is not on GPU"); TORCH_CHECK(b_scales.device().is_cuda(), "b_scales is not on GPU");
TORCH_CHECK(b_scales.is_contiguous(), "b_scales is not contiguous"); TORCH_CHECK(b_scales.is_contiguous(), "b_scales is not contiguous");
TORCH_CHECK(b_zeros.device().is_cuda(), "b_zeros is not on GPU");
TORCH_CHECK(b_zeros.is_contiguous(), "b_zeros is not contiguous");
TORCH_CHECK(g_idx.device().is_cuda(), "g_idx is not on GPU"); TORCH_CHECK(g_idx.device().is_cuda(), "g_idx is not on GPU");
TORCH_CHECK(g_idx.is_contiguous(), "g_idx is not contiguous"); TORCH_CHECK(g_idx.is_contiguous(), "g_idx is not contiguous");
...@@ -1784,6 +2212,17 @@ torch::Tensor gptq_marlin_gemm(torch::Tensor& a, torch::Tensor& b_q_weight, ...@@ -1784,6 +2212,17 @@ torch::Tensor gptq_marlin_gemm(torch::Tensor& a, torch::Tensor& b_q_weight,
torch::Tensor c = torch::empty({size_m, size_n}, options); torch::Tensor c = torch::empty({size_m, size_n}, options);
torch::Tensor a_tmp = torch::empty({size_m, size_k}, options); torch::Tensor a_tmp = torch::empty({size_m, size_k}, options);
// Alloc C tmp buffer that is going to be used for the global reduce
int reduce_max_m = marlin::determine_reduce_max_m(size_m, marlin::max_par);
int reduce_n = size_n;
auto options_fp32 =
torch::TensorOptions().dtype(at::kFloat).device(a.device());
if (!use_fp32_reduce) {
reduce_max_m = 0;
reduce_n = 0;
}
torch::Tensor c_tmp = torch::empty({reduce_max_m, reduce_n}, options_fp32);
// thread_k: `k` size of a thread_tile in `weights` (can usually be left as // thread_k: `k` size of a thread_tile in `weights` (can usually be left as
// auto -1) // auto -1)
int thread_k = -1; int thread_k = -1;
...@@ -1805,8 +2244,8 @@ torch::Tensor gptq_marlin_gemm(torch::Tensor& a, torch::Tensor& b_q_weight, ...@@ -1805,8 +2244,8 @@ torch::Tensor gptq_marlin_gemm(torch::Tensor& a, torch::Tensor& b_q_weight,
int group_size = -1; int group_size = -1;
bool has_act_order = g_idx.size(0) != 0; bool has_act_order = g_idx.size(0) != 0;
int b_rank = b_scales.sizes().size(); int rank = b_scales.sizes().size();
TORCH_CHECK(b_rank == 2, "b_scales rank = ", b_rank, " is not 2"); TORCH_CHECK(rank == 2, "b_scales rank = ", rank, " is not 2");
TORCH_CHECK(b_scales.size(1) == size_n, "b_scales dim 1 = ", b_scales.size(1), TORCH_CHECK(b_scales.size(1) == size_n, "b_scales dim 1 = ", b_scales.size(1),
" is not size_n = ", size_n); " is not size_n = ", size_n);
num_groups = b_scales.size(0); num_groups = b_scales.size(0);
...@@ -1832,34 +2271,45 @@ torch::Tensor gptq_marlin_gemm(torch::Tensor& a, torch::Tensor& b_q_weight, ...@@ -1832,34 +2271,45 @@ torch::Tensor gptq_marlin_gemm(torch::Tensor& a, torch::Tensor& b_q_weight,
} }
} }
// Verify b_zeros
if (has_zp) {
int rank = b_zeros.sizes().size();
TORCH_CHECK(rank == 2, "b_zeros rank = ", rank, " is not 2");
TORCH_CHECK(b_zeros.size(0) == num_groups,
"b_zeros dim 0 = ", b_zeros.size(0),
" is not num_groups = ", num_groups);
TORCH_CHECK(b_zeros.size(1) == size_n / pack_factor,
"b_zeros dim 1 = ", b_scales.size(1),
" is not size_n / pack_factor = ", size_n / pack_factor);
}
// Verify workspace size // Verify workspace size
TORCH_CHECK( TORCH_CHECK(size_n % marlin::min_thread_n == 0, "size_n = ", size_n,
size_n % gptq_marlin::min_thread_n == 0, "size_n = ", size_n, ", is not divisible by min_thread_n = ", marlin::min_thread_n);
", is not divisible by min_thread_n = ", gptq_marlin::min_thread_n); int min_workspace_size = (size_n / marlin::min_thread_n) * marlin::max_par;
int min_workspace_size =
(size_n / gptq_marlin::min_thread_n) * gptq_marlin::max_par;
TORCH_CHECK(workspace.numel() >= min_workspace_size, TORCH_CHECK(workspace.numel() >= min_workspace_size,
"workspace.numel = ", workspace.numel(), "workspace.numel = ", workspace.numel(),
" is below min_workspace_size = ", min_workspace_size); " is below min_workspace_size = ", min_workspace_size);
int dev = a.get_device(); int dev = a.get_device();
if (a.scalar_type() == at::ScalarType::Half) { if (a.scalar_type() == at::ScalarType::Half) {
gptq_marlin::marlin_mm_f16i4<half>( marlin::marlin_mm<half>(
a.data_ptr<at::Half>(), b_q_weight.data_ptr(), c.data_ptr<at::Half>(), a.data_ptr<at::Half>(), b_q_weight.data_ptr(), c.data_ptr<at::Half>(),
b_scales.data_ptr<at::Half>(), g_idx.data_ptr(), perm.data_ptr(), c_tmp.data_ptr<float>(), b_scales.data_ptr<at::Half>(),
b_zeros.data_ptr(), g_idx.data_ptr(), perm.data_ptr(),
a_tmp.data_ptr<at::Half>(), size_m, size_n, size_k, a_tmp.data_ptr<at::Half>(), size_m, size_n, size_k,
workspace.data_ptr(), num_bits, has_act_order, is_k_full, num_groups, workspace.data_ptr(), *b_q_type, has_act_order, is_k_full, has_zp,
group_size, dev, at::cuda::getCurrentCUDAStream(dev), thread_k, num_groups, group_size, dev, at::cuda::getCurrentCUDAStream(dev),
thread_n, sms, gptq_marlin::max_par); thread_k, thread_n, sms, marlin::max_par, use_fp32_reduce);
} else if (a.scalar_type() == at::ScalarType::BFloat16) { } else if (a.scalar_type() == at::ScalarType::BFloat16) {
gptq_marlin::marlin_mm_f16i4<nv_bfloat16>( marlin::marlin_mm<nv_bfloat16>(
a.data_ptr<at::BFloat16>(), b_q_weight.data_ptr(), a.data_ptr<at::BFloat16>(), b_q_weight.data_ptr(),
c.data_ptr<at::BFloat16>(), b_scales.data_ptr<at::BFloat16>(), c.data_ptr<at::BFloat16>(), c_tmp.data_ptr<float>(),
g_idx.data_ptr(), perm.data_ptr(), a_tmp.data_ptr<at::BFloat16>(), b_scales.data_ptr<at::BFloat16>(), b_zeros.data_ptr(), g_idx.data_ptr(),
size_m, size_n, size_k, workspace.data_ptr(), num_bits, has_act_order, perm.data_ptr(), a_tmp.data_ptr<at::BFloat16>(), size_m, size_n, size_k,
is_k_full, num_groups, group_size, dev, workspace.data_ptr(), *b_q_type, has_act_order, is_k_full, has_zp,
at::cuda::getCurrentCUDAStream(dev), thread_k, thread_n, sms, num_groups, group_size, dev, at::cuda::getCurrentCUDAStream(dev),
gptq_marlin::max_par); thread_k, thread_n, sms, marlin::max_par, use_fp32_reduce);
} else { } else {
TORCH_CHECK(false, "gpt_marlin_gemm only supports bfloat16 and float16"); TORCH_CHECK(false, "gpt_marlin_gemm only supports bfloat16 and float16");
} }
......
#include "gptq_marlin.cuh" #include "marlin.cuh"
namespace gptq_marlin {
static constexpr int repack_stages = 8;
static constexpr int repack_threads = 256;
static constexpr int tile_k_size = tile_size;
static constexpr int tile_n_size = tile_k_size * 4;
#if defined(__CUDA_ARCH__) && __CUDA_ARCH__ < 800 #if defined(__CUDA_ARCH__) && __CUDA_ARCH__ < 800
namespace marlin {
template <int const num_threads, int const num_bits, bool const has_perm> template <int const num_threads, int const num_bits, bool const has_perm>
__global__ void marlin_repack_kernel( __global__ void gptq_marlin_repack_kernel(
uint32_t const* __restrict__ b_q_weight_ptr, uint32_t const* __restrict__ b_q_weight_ptr,
uint32_t const* __restrict__ perm_ptr, uint32_t* __restrict__ out_ptr, uint32_t const* __restrict__ perm_ptr, uint32_t* __restrict__ out_ptr,
int size_k, int size_n) {} int size_k, int size_n) {}
} // namespace gptq_marlin } // namespace marlin
torch::Tensor gptq_marlin_repack(torch::Tensor& b_q_weight, torch::Tensor& perm, torch::Tensor gptq_marlin_repack(torch::Tensor& b_q_weight, torch::Tensor& perm,
int64_t size_k, int64_t size_n, int64_t size_k, int64_t size_n,
...@@ -29,8 +22,10 @@ torch::Tensor gptq_marlin_repack(torch::Tensor& b_q_weight, torch::Tensor& perm, ...@@ -29,8 +22,10 @@ torch::Tensor gptq_marlin_repack(torch::Tensor& b_q_weight, torch::Tensor& perm,
#else #else
namespace marlin {
template <int const num_threads, int const num_bits, bool const has_perm> template <int const num_threads, int const num_bits, bool const has_perm>
__global__ void marlin_repack_kernel( __global__ void gptq_marlin_repack_kernel(
uint32_t const* __restrict__ b_q_weight_ptr, uint32_t const* __restrict__ b_q_weight_ptr,
uint32_t const* __restrict__ perm_ptr, uint32_t* __restrict__ out_ptr, uint32_t const* __restrict__ perm_ptr, uint32_t* __restrict__ out_ptr,
int size_k, int size_n) { int size_k, int size_n) {
...@@ -259,28 +254,28 @@ __global__ void marlin_repack_kernel( ...@@ -259,28 +254,28 @@ __global__ void marlin_repack_kernel(
} }
} }
} // namespace gptq_marlin } // namespace marlin
#define CALL_IF(NUM_BITS, HAS_PERM) \ #define CALL_IF(NUM_BITS, HAS_PERM) \
else if (num_bits == NUM_BITS && has_perm == HAS_PERM) { \ else if (num_bits == NUM_BITS && has_perm == HAS_PERM) { \
cudaFuncSetAttribute( \ cudaFuncSetAttribute( \
gptq_marlin::marlin_repack_kernel<gptq_marlin::repack_threads, \ marlin::gptq_marlin_repack_kernel<marlin::repack_threads, NUM_BITS, \
NUM_BITS, HAS_PERM>, \ HAS_PERM>, \
cudaFuncAttributeMaxDynamicSharedMemorySize, max_shared_mem); \ cudaFuncAttributeMaxDynamicSharedMemorySize, max_shared_mem); \
gptq_marlin::marlin_repack_kernel<gptq_marlin::repack_threads, NUM_BITS, \ marlin::gptq_marlin_repack_kernel<marlin::repack_threads, NUM_BITS, \
HAS_PERM> \ HAS_PERM> \
<<<blocks, gptq_marlin::repack_threads, max_shared_mem, stream>>>( \ <<<blocks, marlin::repack_threads, max_shared_mem, stream>>>( \
b_q_weight_ptr, perm_ptr, out_ptr, size_k, size_n); \ b_q_weight_ptr, perm_ptr, out_ptr, size_k, size_n); \
} }
torch::Tensor gptq_marlin_repack(torch::Tensor& b_q_weight, torch::Tensor& perm, torch::Tensor gptq_marlin_repack(torch::Tensor& b_q_weight, torch::Tensor& perm,
int64_t size_k, int64_t size_n, int64_t size_k, int64_t size_n,
int64_t num_bits) { int64_t num_bits) {
// Verify compatibility with marlin tile of 16x64 // Verify compatibility with marlin tile of 16x64
TORCH_CHECK(size_k % gptq_marlin::tile_k_size == 0, "size_k = ", size_k, TORCH_CHECK(size_k % marlin::tile_k_size == 0, "size_k = ", size_k,
" is not divisible by tile_k_size = ", gptq_marlin::tile_k_size); " is not divisible by tile_k_size = ", marlin::tile_k_size);
TORCH_CHECK(size_n % gptq_marlin::tile_n_size == 0, "size_n = ", size_n, TORCH_CHECK(size_n % marlin::tile_n_size == 0, "size_n = ", size_n,
" is not divisible by tile_n_size = ", gptq_marlin::tile_n_size); " is not divisible by tile_n_size = ", marlin::tile_n_size);
TORCH_CHECK(num_bits == 4 || num_bits == 8, TORCH_CHECK(num_bits == 4 || num_bits == 8,
"num_bits must be 4 or 8. Got = ", num_bits); "num_bits must be 4 or 8. Got = ", num_bits);
...@@ -308,10 +303,9 @@ torch::Tensor gptq_marlin_repack(torch::Tensor& b_q_weight, torch::Tensor& perm, ...@@ -308,10 +303,9 @@ torch::Tensor gptq_marlin_repack(torch::Tensor& b_q_weight, torch::Tensor& perm,
auto options = torch::TensorOptions() auto options = torch::TensorOptions()
.dtype(b_q_weight.dtype()) .dtype(b_q_weight.dtype())
.device(b_q_weight.device()); .device(b_q_weight.device());
torch::Tensor out = torch::Tensor out = torch::empty(
torch::empty({size_k / gptq_marlin::tile_size, {size_k / marlin::tile_size, size_n * marlin::tile_size / pack_factor},
size_n * gptq_marlin::tile_size / pack_factor}, options);
options);
// Detect if there is act_order // Detect if there is act_order
bool has_perm = perm.size(0) != 0; bool has_perm = perm.size(0) != 0;
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment