Skip to content

GitLab

Commit 930b2872 authored by Harisankar Sadasivan's avatar Harisankar Sadasivan
Browse files

best performing kernel for GEMV codex problem with M=1 with inverted B matrix

parents a1e17d18 a4f72a31
Hide whitespace changes
Inline Side-by-side
Showing with 2895 additions and 161 deletions
example/53_gemv_splitk/gemv_splitk_fp16.cpp
View file @ 930b2872
......@@ -10,7 +10,7 @@ using CDataType = ck::half_t;
using AccDataType = float;
using ALayout = Row;
using BLayout = Col;
using BLayout = Row; // Col;
using CLayout = Row;
using AElementOp = PassThrough;
......@@ -19,9 +19,9 @@ using CElementOp = PassThrough;
static constexpr auto GemmMNPadding = ck::tensor_operation::device::GemmSpecialization::MNPadding;
#define K1 8 // K1PerThread:2,4,8
#define K0 4 // K0PerBlock:1,2,3,4...32
#define N1 2 // Nperthread:2,4,8
#define K1 4
#define K0 3
#define N1 2
#define B 64 // block-size:64
// clang-format off
......@@ -31,7 +31,7 @@ using DeviceGemvInstance = ck::tensor_operation::device::deviceGemvDl/*
// ######| | | | | | | | Operation| Operation| Operation| | | | | | | | | | KBatch_K0_M0_M1_K1| KBatch_K0_M0_M1_K1| ArrangeOrder| Order| KBatch_K0_M0_M1_K1 | ContiguousDimOrder| KBatch_K0_M0_M1_K1 | Order| | | Order| | |
// ######| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | |
//< ADataType, BDataType, CDataType, AccDataType, ALayout, BLayout, CLayout, AElementOp, BElementOp, CElementOp, GemmMNPadding, 64, 1, 64, 32, 2, 1, 1, 1, S<1, 1, 1, 2>, S<32, 1, 1, 1>, S<1, 2, 0, 3>, S<1, 2, 0, 3>, S<1, 1, 1, 2>, S<1, 2, 0, 3>, S<1, 1, 1, 2>, S<1, 2, 0, 3>, 3, 2, S<0, 1, 2, 3, 4, 5>, 5, 1>;*/
< ADataType, BDataType, CDataType, AccDataType, ALayout, BLayout, CLayout, AElementOp, BElementOp, CElementOp, GemmMNPadding, B, 1, B*N1, K0, K1, 1, N1, 1, S<1,1, 1, 1, K1>, S<1,K0, 1, 1, 1>,S<0,1,2,3,4>, S<0,1,2,3,4>, S<1,1, 1, 1, K1>, S<0,1,2,3,4>, S<1,1, 1, 1, 2>, S<0,1,2,3,4>, 4, K1, S<0, 1, 2, 3, 4, 5>, 5, N1>;
< ADataType, BDataType, CDataType, AccDataType, ALayout, BLayout, CLayout, AElementOp, BElementOp, CElementOp, GemmMNPadding, B, 1, B*N1, K0, K1, 1, N1, 1, S<1,1, 1, 1, K1>, S<1,K0, 1, 1, 1>,S<0,1,2,3,4>, S<0,1,2,3,4>, S<1,1, 1, 1, K1>, S<0,1,2,3,4>, S<1,1, 1, 1, 2>, S<0,1,2,3,4>, 3, N1, S<0, 1, 2, 3, 4, 5>, 5, N1>;
// clang-format on
using ReferenceGemmInstance = ck::tensor_operation::host::
......
example/CMakeLists.txt
View file @ 930b2872
......@@ -7,20 +7,114 @@ add_custom_target(examples)
function(add_example_executable EXAMPLE_NAME FILE_NAME)
message("adding example ${EXAMPLE_NAME}")
add_executable(${EXAMPLE_NAME} ${FILE_NAME})
target_link_libraries(${EXAMPLE_NAME} PRIVATE utility)
add_test(NAME ${EXAMPLE_NAME} COMMAND $<TARGET_FILE:${EXAMPLE_NAME}> ${ARGN})
add_dependencies(examples ${EXAMPLE_NAME})
add_dependencies(check ${EXAMPLE_NAME})
rocm_install(TARGETS ${EXAMPLE_NAME} COMPONENT examples)
set(result 1)
if(DEFINED DTYPES)
foreach(source IN LISTS FILE_NAME)
set(test 0)
foreach(type IN LISTS DTYPES)
if(type MATCHES "fp16")
set(type1 "_f16")
elseif(type MATCHES "fp32")
set(type1 "_f32")
elseif(type MATCHES "fp8")
set(type1 "_f8")
elseif(type MATCHES "bf16")
set(type1 "_b16")
elseif(type MATCHES "fp64")
set(type1 "_f64")
elseif(type MATCHES "int8")
set(type1 "_i8")
endif()
if("${source}" MATCHES "${type}" OR "${source}" MATCHES "${type1}")
#if filename matches any selected type, exit type loop and do no exclude the file from the list
set(test 0)
break()
elseif((source MATCHES "fp8" OR source MATCHES "fp32" OR source MATCHES "fp64" OR source MATCHES "bf16" OR source MATCHES "int8" OR source MATCHES "fp16" OR
source MATCHES "_f8" OR source MATCHES "_f32" OR source MATCHES "_f64" OR source MATCHES "_i8" OR source MATCHES "_f16" OR source MATCHES "_b16") AND
NOT(source MATCHES type OR source MATCHES type1))
#if filename contains a type which doesn't match any selected type, mark it for removal
set(test 1)
endif()
endforeach()
if(test EQUAL 1)
message("removing example source file ${source} ")
list(REMOVE_ITEM FILE_NAME "${source}")
endif()
endforeach()
endif()
foreach(source IN LISTS FILE_NAME)
if(NOT DEFINED DL_KERNELS AND source MATCHES "_dl")
message("removing dl example ${source} ")
list(REMOVE_ITEM FILE_NAME "${source}")
endif()
endforeach()
#only continue if there are some source files left on the list
if(FILE_NAME)
add_executable(${EXAMPLE_NAME} ${FILE_NAME})
target_link_libraries(${EXAMPLE_NAME} PRIVATE utility)
add_test(NAME ${EXAMPLE_NAME} COMMAND $<TARGET_FILE:${EXAMPLE_NAME}> ${ARGN})
add_dependencies(examples ${EXAMPLE_NAME})
add_dependencies(check ${EXAMPLE_NAME})
rocm_install(TARGETS ${EXAMPLE_NAME} COMPONENT examples)
set(result 0)
endif()
#message("add_example returns ${result}")
set(result ${result} PARENT_SCOPE)
endfunction(add_example_executable EXAMPLE_NAME)
function(add_example_executable_no_testing EXAMPLE_NAME FILE_NAME)
message("adding example ${EXAMPLE_NAME}")
add_executable(${EXAMPLE_NAME} ${FILE_NAME})
target_link_libraries(${EXAMPLE_NAME} PRIVATE utility)
add_dependencies(examples ${EXAMPLE_NAME})
rocm_install(TARGETS ${EXAMPLE_NAME} COMPONENT examples)
set(result 1)
if(DEFINED DTYPES)
foreach(source IN LISTS FILE_NAME)
set(test 0)
foreach(type IN LISTS DTYPES)
if(type MATCHES "fp16")
set(type1 "_f16")
elseif(type MATCHES "fp32")
set(type1 "_f32")
elseif(type MATCHES "fp8")
set(type1 "_f8")
elseif(type MATCHES "bf16")
set(type1 "_b16")
elseif(type MATCHES "fp64")
set(type1 "_f64")
elseif(type MATCHES "int8")
set(type1 "_i8")
endif()
if("${source}" MATCHES "${type}" OR "${source}" MATCHES "${type1}")
#if filename matches any selected type, exit type loop and do no exclude the file from the list
set(test 0)
break()
elseif((source MATCHES "fp8" OR source MATCHES "fp32" OR source MATCHES "fp64" OR source MATCHES "bf16" OR source MATCHES "int8" OR source MATCHES "fp16" OR
source MATCHES "_f8" OR source MATCHES "_f32" OR source MATCHES "_f64" OR source MATCHES "_i8" OR source MATCHES "_f16" OR source MATCHES "_b16") AND
NOT(source MATCHES type OR source MATCHES type1))
#if filename contains a type which doesn't match any selected type, mark it for removal
set(test 1)
endif()
endforeach()
if(test EQUAL 1)
message("removing example ${source} ")
list(REMOVE_ITEM FILE_NAME "${source}")
endif()
endforeach()
endif()
foreach(source IN LISTS FILE_NAME)
if(NOT DEFINED DL_KERNELS AND source MATCHES "_dl")
message("removing dl example ${source} ")
list(REMOVE_ITEM FILE_NAME "${source}")
endif()
endforeach()
#only continue if there are some source files left on the list
if(FILE_NAME)
add_executable(${EXAMPLE_NAME} ${FILE_NAME})
target_link_libraries(${EXAMPLE_NAME} PRIVATE utility)
add_dependencies(examples ${EXAMPLE_NAME})
rocm_install(TARGETS ${EXAMPLE_NAME} COMPONENT examples)
set(result 0)
endif()
#message("add_example returns ${result}")
set(result ${result} PARENT_SCOPE)
endfunction(add_example_executable_no_testing EXAMPLE_NAME)
# add all example subdir
......
include/ck/host_utility/kernel_launch.hpp
View file @ 930b2872
......@@ -34,6 +34,7 @@ float launch_and_time_kernel(const StreamConfig& stream_config,
#endif
// warm up
kernel<<<grid_dim, block_dim, lds_byte, stream_config.stream_id_>>>(args...);
hip_check_error(hipGetLastError());
const int nrepeat = 10;
#if DEBUG_LOG
......@@ -50,6 +51,7 @@ float launch_and_time_kernel(const StreamConfig& stream_config,
for(int i = 0; i < nrepeat; ++i)
{
kernel<<<grid_dim, block_dim, lds_byte, stream_config.stream_id_>>>(args...);
hip_check_error(hipGetLastError());
}
hip_check_error(hipEventRecord(stop, stream_config.stream_id_));
......@@ -64,11 +66,13 @@ float launch_and_time_kernel(const StreamConfig& stream_config,
else
{
kernel<<<grid_dim, block_dim, lds_byte, stream_config.stream_id_>>>(args...);
hip_check_error(hipGetLastError());
return 0;
}
#else
kernel<<<grid_dim, block_dim, lds_byte, stream_config.stream_id_>>>(args...);
hip_check_error(hipGetLastError());
return 0;
#endif
......@@ -101,6 +105,7 @@ float launch_and_time_kernel_with_preprocess(const StreamConfig& stream_config,
// warm up
preprocess();
kernel<<<grid_dim, block_dim, lds_byte, stream_config.stream_id_>>>(args...);
hip_check_error(hipGetLastError());
const int nrepeat = 10;
#if DEBUG_LOG
......@@ -118,6 +123,7 @@ float launch_and_time_kernel_with_preprocess(const StreamConfig& stream_config,
{
preprocess();
kernel<<<grid_dim, block_dim, lds_byte, stream_config.stream_id_>>>(args...);
hip_check_error(hipGetLastError());
}
hip_check_error(hipEventRecord(stop, stream_config.stream_id_));
......@@ -133,11 +139,13 @@ float launch_and_time_kernel_with_preprocess(const StreamConfig& stream_config,
{
preprocess();
kernel<<<grid_dim, block_dim, lds_byte, stream_config.stream_id_>>>(args...);
hip_check_error(hipGetLastError());
return 0;
}
#else
kernel<<<grid_dim, block_dim, lds_byte, stream_config.stream_id_>>>(args...);
hip_check_error(hipGetLastError());
return 0;
#endif
......
include/ck/tensor_operation/gpu/block/blockwise_gemm_xdlops.hpp
View file @ 930b2872
......@@ -28,7 +28,8 @@ MakeGemmMmaTileDescriptor_MN0_MN1_MN2_K(const TileDesc_K0_MN_K1&)
}
template <index_t BlockSize,
typename FloatAB,
typename FloatA,
typename FloatB,
typename FloatAcc,
typename AK0MK1BlockDesc,
typename BK0NK1BlockDesc,
......@@ -58,7 +59,7 @@ struct BlockwiseGemmXdlops_k0mk1_k0nk1_m0n0m1n1m2m3m4n2_v1
static constexpr index_t A_K1 = AK0MK1BlockDesc{}.GetLength(I2);
static constexpr index_t B_K1 = BK0NK1BlockDesc{}.GetLength(I2);
static constexpr auto xdlops_gemm = XdlopsGemm<FloatAB, MPerXDL, NPerXDL, KPack>{};
static constexpr auto xdlops_gemm = XdlopsGemm<FloatA, MPerXDL, NPerXDL, KPack, FloatB>{};
static constexpr index_t KPerThread = KPerBlock / xdlops_gemm.K0PerXdlops;
......@@ -294,9 +295,9 @@ struct BlockwiseGemmXdlops_k0mk1_k0nk1_m0n0m1n1m2m3m4n2_v1
const BBlockBuffer& b_block_buf,
CThreadBuffer& c_thread_buf) const
{
auto a_thread_buf = make_static_buffer<AddressSpaceEnum::Vgpr, FloatAB>(
auto a_thread_buf = make_static_buffer<AddressSpaceEnum::Vgpr, FloatA>(
a_thread_desc_.GetElementSpaceSize());
auto b_thread_buf = make_static_buffer<AddressSpaceEnum::Vgpr, FloatAB>(
auto b_thread_buf = make_static_buffer<AddressSpaceEnum::Vgpr, FloatB>(
b_thread_desc_.GetElementSpaceSize());
static_for<0, MRepeat, 1>{}([&](auto m0) {
......@@ -318,25 +319,27 @@ struct BlockwiseGemmXdlops_k0mk1_k0nk1_m0n0m1n1m2m3m4n2_v1
b_thread_buf);
static_for<0, KPerThread, KPack>{}([&](auto k) {
vector_type<FloatAB, KPack> a_thread_vec;
vector_type<FloatAB, KPack> b_thread_vec;
vector_type<FloatA, KPack> a_thread_vec;
vector_type<FloatB, KPack> b_thread_vec;
static_for<0, KPack, 1>{}([&](auto i) {
a_thread_vec.template AsType<FloatAB>()(i) = a_thread_buf
a_thread_vec.template AsType<FloatA>()(i) = a_thread_buf
[Number<a_thread_desc_.CalculateOffset(make_tuple(0, 0, 0, k + i))>{}];
b_thread_vec.template AsType<FloatAB>()(i) = b_thread_buf
b_thread_vec.template AsType<FloatB>()(i) = b_thread_buf
[Number<b_thread_desc_.CalculateOffset(make_tuple(0, 0, 0, k + i))>{}];
});
using mfma_input_type =
typename vector_type<FloatAB, xdlops_gemm.K1PerXdlops>::type;
using mfma_input_type_a =
typename vector_type<FloatA, xdlops_gemm.K1PerXdlops>::type;
using mfma_input_type_b =
typename vector_type<FloatB, xdlops_gemm.K1PerXdlops>::type;
constexpr index_t c_offset =
c_thread_desc_.CalculateOffset(make_tuple(m0, n0, 0));
xdlops_gemm.template Run(
a_thread_vec.template AsType<mfma_input_type>(),
b_thread_vec.template AsType<mfma_input_type>(),
a_thread_vec.template AsType<mfma_input_type_a>(),
b_thread_vec.template AsType<mfma_input_type_b>(),
c_thread_buf.GetVectorTypeReference(Number<c_offset>{}));
});
});
......@@ -356,8 +359,8 @@ struct BlockwiseGemmXdlops_k0mk1_k0nk1_m0n0m1n1m2m3m4n2_v1
static constexpr auto c_thread_desc_ = make_naive_tensor_descriptor_packed(
make_tuple(Number<MRepeat>{}, Number<NRepeat>{}, xdlops_gemm.GetRegSizePerXdlops()));
using AThreadCopy = ThreadwiseTensorSliceTransfer_v4<FloatAB,
FloatAB,
using AThreadCopy = ThreadwiseTensorSliceTransfer_v4<FloatA,
FloatA,
decltype(a_block_desc_m0_m1_m2_k),
decltype(a_thread_desc_),
Sequence<1, 1, 1, KPerThread>,
......@@ -366,8 +369,8 @@ struct BlockwiseGemmXdlops_k0mk1_k0nk1_m0n0m1n1m2m3m4n2_v1
A_K1,
A_K1>;
using BThreadCopy = ThreadwiseTensorSliceTransfer_v4<FloatAB,
FloatAB,
using BThreadCopy = ThreadwiseTensorSliceTransfer_v4<FloatB,
FloatB,
decltype(b_block_desc_n0_n1_n2_k),
decltype(b_thread_desc_),
Sequence<1, 1, 1, KPerThread>,
......@@ -385,7 +388,8 @@ struct BlockwiseGemmXdlops_k0mk1_k0nk1_m0n0m1n1m2m3m4n2_v1
// the latest ROCm release. For unsupported compilers, inter-wave loop scheduler falls back to the
// default loop scheduler which is given by the macro CK_EXPERIMENTAL_INTER_WAVE_SCHEDULING=0
template <index_t BlockSize,
typename FloatAB,
typename FloatA,
typename FloatB,
typename FloatAcc,
typename AK0MK1BlockDesc,
typename BK0NK1BlockDesc,
......@@ -397,7 +401,8 @@ template <index_t BlockSize,
index_t NumMacClusters = CK_EXPERIMENTAL_INTER_WAVE_SCHEDULING_MAC_CLUSTERS>
struct BlockwiseGemmXdlopsInterwave_k0mk1_k0nk1_m0n0m1n1m2m3m4n2_v1
: public BlockwiseGemmXdlops_k0mk1_k0nk1_m0n0m1n1m2m3m4n2_v1<BlockSize,
FloatAB,
FloatA,
FloatB,
FloatAcc,
AK0MK1BlockDesc,
BK0NK1BlockDesc,
......@@ -408,7 +413,8 @@ struct BlockwiseGemmXdlopsInterwave_k0mk1_k0nk1_m0n0m1n1m2m3m4n2_v1
KPack>
{
using Base = BlockwiseGemmXdlops_k0mk1_k0nk1_m0n0m1n1m2m3m4n2_v1<BlockSize,
FloatAB,
FloatA,
FloatB,
FloatAcc,
AK0MK1BlockDesc,
BK0NK1BlockDesc,
......@@ -440,9 +446,9 @@ struct BlockwiseGemmXdlopsInterwave_k0mk1_k0nk1_m0n0m1n1m2m3m4n2_v1
const BBlockBuffer& b_block_buf,
CThreadBuffer& c_thread_buf) const
{
auto a_thread_buf = make_static_buffer<AddressSpaceEnum::Vgpr, FloatAB>(
auto a_thread_buf = make_static_buffer<AddressSpaceEnum::Vgpr, FloatA>(
a_thread_desc_.GetElementSpaceSize());
auto b_thread_buf = make_static_buffer<AddressSpaceEnum::Vgpr, FloatAB>(
auto b_thread_buf = make_static_buffer<AddressSpaceEnum::Vgpr, FloatB>(
b_thread_desc_.GetElementSpaceSize());
static_for<0, KPerThread, KPerInnerLoop>{}([&](auto k) {
......@@ -479,20 +485,22 @@ struct BlockwiseGemmXdlopsInterwave_k0mk1_k0nk1_m0n0m1n1m2m3m4n2_v1
static_for<0, KPerInnerLoop, KPack>{}([&](auto k_) {
static_for<0, MRepeat, 1>{}([&](auto m0) {
static_for<0, NRepeat, 1>{}([&](auto n0) {
vector_type<FloatAB, KPack> a_thread_vec;
vector_type<FloatAB, KPack> b_thread_vec;
vector_type<FloatA, KPack> a_thread_vec;
vector_type<FloatB, KPack> b_thread_vec;
static_for<0, KPack, 1>{}([&](auto i) {
a_thread_vec.template AsType<FloatAB>()(i) =
a_thread_vec.template AsType<FloatA>()(i) =
a_thread_buf[Number<a_thread_desc_.CalculateOffset(
make_tuple(m0, 0, 0, k_ + i))>{}];
b_thread_vec.template AsType<FloatAB>()(i) =
b_thread_vec.template AsType<FloatB>()(i) =
b_thread_buf[Number<b_thread_desc_.CalculateOffset(
make_tuple(n0, 0, 0, k_ + i))>{}];
});
using mfma_input_type =
typename vector_type<FloatAB, xdlops_gemm.K1PerXdlops>::type;
using mfma_input_type_a =
typename vector_type<FloatA, xdlops_gemm.K1PerXdlops>::type;
using mfma_input_type_b =
typename vector_type<FloatB, xdlops_gemm.K1PerXdlops>::type;
constexpr index_t c_offset =
c_thread_desc_.CalculateOffset(make_tuple(m0, n0, 0));
......@@ -514,8 +522,8 @@ struct BlockwiseGemmXdlopsInterwave_k0mk1_k0nk1_m0n0m1n1m2m3m4n2_v1
// TODO: insert setprio in more precise manner since we
// could have more than >1 MFMA instructions in single call
xdlops_gemm.template Run(
a_thread_vec.template AsType<mfma_input_type>(),
b_thread_vec.template AsType<mfma_input_type>(),
a_thread_vec.template AsType<mfma_input_type_a>(),
b_thread_vec.template AsType<mfma_input_type_b>(),
c_thread_buf.GetVectorTypeReference(Number<c_offset>{}));
if constexpr(k_.value == 0 && m0.value == 0 && n0.value == 0)
{
......@@ -541,8 +549,8 @@ struct BlockwiseGemmXdlopsInterwave_k0mk1_k0nk1_m0n0m1n1m2m3m4n2_v1
static constexpr auto b_thread_desc_ = make_naive_tensor_descriptor_packed(
make_tuple(Number<NRepeat>{}, I1, I1, Number<KPerInnerLoop>{}));
using AThreadCopy = ThreadwiseTensorSliceTransfer_v4<FloatAB,
FloatAB,
using AThreadCopy = ThreadwiseTensorSliceTransfer_v4<FloatA,
FloatA,
decltype(a_block_desc_m0_m1_m2_k),
decltype(a_thread_desc_),
Sequence<1, 1, 1, KPerInnerLoop>,
......@@ -551,8 +559,8 @@ struct BlockwiseGemmXdlopsInterwave_k0mk1_k0nk1_m0n0m1n1m2m3m4n2_v1
A_K1,
A_K1>;
using BThreadCopy = ThreadwiseTensorSliceTransfer_v4<FloatAB,
FloatAB,
using BThreadCopy = ThreadwiseTensorSliceTransfer_v4<FloatB,
FloatB,
decltype(b_block_desc_n0_n1_n2_k),
decltype(b_thread_desc_),
Sequence<1, 1, 1, KPerInnerLoop>,
......@@ -568,7 +576,8 @@ struct BlockwiseGemmXdlopsInterwave_k0mk1_k0nk1_m0n0m1n1m2m3m4n2_v1
};
template <index_t BlockSize,
typename FloatAB,
typename FloatA,
typename FloatB,
typename FloatAcc,
typename AK0MK1BlockDesc,
typename BK0NK1BlockDesc,
......@@ -583,7 +592,8 @@ constexpr auto BlockwiseGemmXdlops_k0mk1_k0nk1_m0n0m1n1m2m3m4n2_Selector()
if constexpr(LoopSched == LoopScheduler::Default)
{
return BlockwiseGemmXdlops_k0mk1_k0nk1_m0n0m1n1m2m3m4n2_v1<BlockSize,
FloatAB,
FloatA,
FloatB,
FloatAcc,
AK0MK1BlockDesc,
BK0NK1BlockDesc,
......@@ -596,7 +606,8 @@ constexpr auto BlockwiseGemmXdlops_k0mk1_k0nk1_m0n0m1n1m2m3m4n2_Selector()
else if constexpr(LoopSched == LoopScheduler::Interwave)
{
return BlockwiseGemmXdlopsInterwave_k0mk1_k0nk1_m0n0m1n1m2m3m4n2_v1<BlockSize,
FloatAB,
FloatA,
FloatB,
FloatAcc,
AK0MK1BlockDesc,
BK0NK1BlockDesc,
......@@ -618,26 +629,27 @@ constexpr auto BlockwiseGemmXdlops_k0mk1_k0nk1_m0n0m1n1m2m3m4n2_Selector()
* 3. configurable k index starting position and step size after each FMA/XDL instruction
*/
template <index_t BlockSize,
typename FloatAB,
typename FloatAcc,
typename ATileDesc,
typename BTileDesc,
typename AMmaTileDesc,
typename BMmaTileDesc,
index_t MPerBlock,
index_t NPerBlock,
index_t KPerBlock,
index_t MPerXDL,
index_t NPerXDL,
index_t MRepeat,
index_t NRepeat,
index_t KPack,
bool TransposeC = false,
index_t AMmaKStride =
KPack* XdlopsGemm<FloatAB, MPerXDL, NPerXDL, KPack, TransposeC>{}.K0PerXdlops,
index_t BMmaKStride =
KPack* XdlopsGemm<FloatAB, MPerXDL, NPerXDL, KPack, TransposeC>{}.K0PerXdlops>
template <
index_t BlockSize,
typename FloatAB,
typename FloatAcc,
typename ATileDesc,
typename BTileDesc,
typename AMmaTileDesc,
typename BMmaTileDesc,
index_t MPerBlock,
index_t NPerBlock,
index_t KPerBlock,
index_t MPerXDL,
index_t NPerXDL,
index_t MRepeat,
index_t NRepeat,
index_t KPack,
bool TransposeC = false,
index_t AMmaKStride =
KPack* XdlopsGemm<FloatAB, MPerXDL, NPerXDL, KPack, FloatAB, TransposeC>{}.K0PerXdlops,
index_t BMmaKStride =
KPack* XdlopsGemm<FloatAB, MPerXDL, NPerXDL, KPack, FloatAB, TransposeC>{}.K0PerXdlops>
struct BlockwiseGemmXdlops_v2
{
static constexpr auto I0 = Number<0>{};
......@@ -654,7 +666,8 @@ struct BlockwiseGemmXdlops_v2
static constexpr index_t A_K1 = ATileDesc{}.GetLength(I2);
static constexpr index_t B_K1 = BTileDesc{}.GetLength(I2);
static constexpr auto xdlops_gemm = XdlopsGemm<FloatAB, MPerXDL, NPerXDL, KPack, TransposeC>{};
static constexpr auto xdlops_gemm =
XdlopsGemm<FloatAB, MPerXDL, NPerXDL, KPack, FloatAB, TransposeC>{};
static constexpr index_t KPerThread = KPerBlock / xdlops_gemm.K0PerXdlops;
......
include/ck/tensor_operation/gpu/block/thread_group_tensor_slice_transfer_v7r2.hpp 0 → 100755
View file @ 930b2872
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
#include "ck/utility/common_header.hpp"
#include "ck/tensor_description/tensor_descriptor.hpp"
#include "ck/tensor_description/tensor_descriptor_helper.hpp"
#include "ck/tensor_description/cluster_descriptor.hpp"
#include "ck/tensor_operation/gpu/thread/threadwise_tensor_slice_transfer_v7r2.hpp"
#include "ck/utility/is_detected.hpp"
namespace ck {
// Thread-group level multi-source, multi-destination tensor slice data movement
// Assume:
// 1. All sources and destinations are DynamicBuffer
// 2. Same VectorDim and ScalerPerVector for all sources and destinations
// 3. DstInMemOps are per destination tensor
// 4. ThreadTransferSrcResetCoordinateAfterRunFlags are per source tensor
// 5. ThreadTransferDstResetCoordinateAfterRunFlags are per destination tensor
//
// Does following things to avoid scratch memory issue
// 1. Pass tensor descritpors by reference (or tuple of references)
// 2. Does not keep reference to tensor descriptor
// 3. Does not construct new tensor coordinate when call Run()
template <typename ThreadGroup,
typename SrcDatas,
typename DstDatas,
typename SrcDescs,
typename DstDescs,
typename ElementwiseOperation,
typename DstInMemOps, // Sequence<InMemoryDataOperationEnum ...>
typename SliceLengths,
typename ThreadClusterLengths,
typename ThreadClusterArrangeOrder,
typename SrcDimAccessOrder,
typename DstDimAccessOrder,
index_t SrcVectorDim,
index_t DstVectorDim,
index_t SrcScalarPerVector,
index_t DstScalarPerVector,
typename ThreadTransferSrcResetCoordinateAfterRunFlags,
typename ThreadTransferDstResetCoordinateAfterRunFlags>
struct ThreadGroupTensorSliceTransfer_v7r2
{
static constexpr index_t nDim =
remove_cvref_t<tuple_element_t<0, SrcDescs>>::GetNumOfDimension();
static constexpr index_t nSrc = remove_cvref_t<SrcDescs>::Size();
static constexpr index_t nDst = remove_cvref_t<DstDescs>::Size();
using Index = MultiIndex<nDim>;
static constexpr auto thread_slice_lengths = SliceLengths{} / ThreadClusterLengths{};
__device__ constexpr ThreadGroupTensorSliceTransfer_v7r2(
const SrcDescs& src_descs,
const StaticallyIndexedArray<Index, nSrc>& src_block_slice_origins,
const DstDescs& dst_descs,
const StaticallyIndexedArray<Index, nDst>& dst_block_slice_origins,
const ElementwiseOperation& element_op)
: threadwise_transfer_(src_descs,
StaticallyIndexedArray<Index, nSrc>{},
dst_descs,
StaticallyIndexedArray<Index, nDst>{},
element_op)
{
static_assert(nSrc == SrcDatas::Size() && nSrc == SrcDescs::Size() &&
nSrc == ThreadTransferSrcResetCoordinateAfterRunFlags::Size() &&
nDst == DstDatas::Size() && nDst == DstDescs::Size() &&
nDst == ThreadTransferDstResetCoordinateAfterRunFlags::Size(),
"wrong!");
static_for<0, nSrc, 1>{}([&](auto i) {
static_assert(
nDim == remove_cvref_t<tuple_element_t<i.value, SrcDescs>>::GetNumOfDimension(),
"wrong!");
});
static_for<0, nDst, 1>{}([&](auto i) {
static_assert(
nDim == remove_cvref_t<tuple_element_t<i.value, DstDescs>>::GetNumOfDimension(),
"wrong!");
});
static_assert(nDim == ThreadClusterLengths::Size() &&
nDim == ThreadClusterArrangeOrder::Size() &&
nDim == SrcDimAccessOrder::Size() && nDim == DstDimAccessOrder::Size(),
"wrong! nDim not consistent");
static_assert(
is_same<SliceLengths, decltype(thread_slice_lengths * ThreadClusterLengths{})>{},
"wrong! threads should be mapped to cover entire slicing window");
static_assert(ThreadGroup::GetNumOfThread() >= thread_cluster_desc_.GetElementSize(),
"wrong! ThreadGroup::GetNumOfThread() too small");
if(ThreadGroup::GetNumOfThread() == thread_cluster_desc_.GetElementSize() or
ThreadGroup::GetThreadId() < thread_cluster_desc_.GetElementSize())
{
const auto thread_cluster_idx = thread_cluster_desc_.CalculateBottomIndex(
make_multi_index(get_thread_local_1d_id()));
const auto thread_data_idx_begin = thread_cluster_idx * thread_slice_lengths;
const auto src_thread_slice_origins = generate_tuple(
[&](auto i) { return src_block_slice_origins[i] + thread_data_idx_begin; },
Number<nSrc>{});
const auto dst_thread_slice_origins = generate_tuple(
[&](auto i) { return dst_block_slice_origins[i] + thread_data_idx_begin; },
Number<nDst>{});
threadwise_transfer_.SetSrcSliceOrigins(src_descs, src_thread_slice_origins);
threadwise_transfer_.SetDstSliceOrigins(dst_descs, dst_thread_slice_origins);
}
}
template <typename SrcBuffers>
__device__ void RunRead(const SrcDescs& src_descs, const SrcBuffers& src_bufs)
{
if(ThreadGroup::GetNumOfThread() == thread_cluster_desc_.GetElementSize() or
ThreadGroup::GetThreadId() < thread_cluster_desc_.GetElementSize())
{
threadwise_transfer_.RunRead(src_descs, src_bufs);
}
}
template <typename T>
using is_tuple = decltype(std::declval<T&>().IsTuple());
template <typename DstBuffers>
__device__ void RunWrite(const DstDescs& dst_descs, DstBuffers dst_bufs)
{
if(ThreadGroup::GetNumOfThread() == thread_cluster_desc_.GetElementSize() or
ThreadGroup::GetThreadId() < thread_cluster_desc_.GetElementSize())
{
if constexpr(is_detected<is_tuple, decltype(dst_bufs)>::value)
threadwise_transfer_.RunWrite(dst_descs, dst_bufs);
else
threadwise_transfer_.RunWrite(dst_descs, tie(dst_bufs));
}
}
template <typename SrcBuffers, typename DstBuffers>
__device__ void Run(const SrcDescs& src_descs,
const SrcBuffers& src_bufs,
const DstDescs& dst_descs,
DstBuffers dst_bufs)
{
RunRead(src_descs, src_bufs);
RunWrite(dst_descs, dst_bufs);
}
template <index_t ISrc>
__device__ void
MoveSrcSliceWindow(const SrcDescs& src_descs, Number<ISrc> iSrc, const Index& step)
{
if(ThreadGroup::GetNumOfThread() == thread_cluster_desc_.GetElementSize() or
ThreadGroup::GetThreadId() < thread_cluster_desc_.GetElementSize())
{
threadwise_transfer_.MoveSrcSliceWindow(src_descs, iSrc, step);
}
}
__device__ void MoveSrcSliceWindow(const SrcDescs& src_descs, const Index& step)
{
static_for<0, SrcDescs::Size(), 1>{}(
[&](auto i) { MoveSrcSliceWindow(src_descs, i, step); });
}
template <index_t IDst>
__device__ void
MoveDstSliceWindow(const DstDescs& dst_descs, Number<IDst> iDst, const Index& step)
{
if(ThreadGroup::GetNumOfThread() == thread_cluster_desc_.GetElementSize() or
ThreadGroup::GetThreadId() < thread_cluster_desc_.GetElementSize())
{
threadwise_transfer_.MoveDstSliceWindow(dst_descs, iDst, step);
}
}
__device__ void MoveDstSliceWindow(const DstDescs& dst_descs, const Index& step)
{
static_for<0, DstDescs::Size(), 1>{}(
[&](auto i) { MoveDstSliceWindow(dst_descs, i, step); });
}
private:
static constexpr auto thread_cluster_desc_ =
make_cluster_descriptor(ThreadClusterLengths{}, ThreadClusterArrangeOrder{});
using ThreadwiseTransfer =
ThreadwiseTensorSliceTransfer_v7r2<SrcDatas,
DstDatas,
SrcDescs,
DstDescs,
ElementwiseOperation,
DstInMemOps,
decltype(thread_slice_lengths),
SrcDimAccessOrder,
DstDimAccessOrder,
SrcVectorDim,
DstVectorDim,
SrcScalarPerVector,
DstScalarPerVector,
ThreadTransferSrcResetCoordinateAfterRunFlags,
ThreadTransferDstResetCoordinateAfterRunFlags>;
ThreadwiseTransfer threadwise_transfer_;
};
} // namespace ck
include/ck/tensor_operation/gpu/device/conv_tensor_rearrange_op.hpp 0 → 100755
View file @ 930b2872
// SPDX-License-Identifier: MIT
// Copyright (c) 2023, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
namespace ck {
namespace conv_tensor_rearrange_op {
struct BaseConvTensorRearrangeOp
{
};
struct ImageToColumn : public BaseConvTensorRearrangeOp
{
static constexpr const char* name = "Image to Column";
};
struct ColumnToImage : public BaseConvTensorRearrangeOp
{
static constexpr const char* name = "Column to Image";
};
template <typename Op,
typename std::enable_if<std::is_base_of<BaseConvTensorRearrangeOp, Op>::value,
bool>::type = false>
std::ostream& operator<<(std::ostream& os, const BaseConvTensorRearrangeOp&)
{
os << Op::name;
return os;
}
} // namespace conv_tensor_rearrange_op
} // namespace ck
include/ck/tensor_operation/gpu/device/device_image_to_column.hpp → include/ck/tensor_operation/gpu/device/device_conv_tensor_rearrange.hpp
View file @ 930b2872
......@@ -12,21 +12,26 @@ namespace tensor_operation {
namespace device {
/**
* \brief Image to column.
* \brief Convolution Tensor Rearrange.
*
* This Device operator converts image ([G, N, Di, Hi, Wi, C]) to the gemm
* problem([N * Do * Ho * Wo, Z * Y * X * C]). G must be equal to 1.
* This Device operator supports conversion image ([G, N, Di, Hi, Wi, C]) to
* the gemm problem([N * Do * Ho * Wo, Z * Y * X * C]) (Image to Column) and
* conversion gemm form to the image (Column to Image).
*
* Note that G must be equal to 1.
*
* \tparam NDimSpatial Number of spatial dimensions.
* \tparam InputLayout Input Layout.
* \tparam ImageLayout Input Layout.
* \tparam InputDataType Input Data Type.
* \tparam OutputDataType Output Data Type.
* \tparam ConvTensorRearrangeOp Operation type: ImageToColumn, ColumnToImage.
*/
template <index_t NDimSpatial,
typename InputLayout,
typename ImageLayout,
typename InputDataType,
typename OutputDataType>
struct DeviceImageToColumn : public BaseOperator
typename OutputDataType,
typename ConvTensorRearrangeOp>
struct DeviceConvTensorRearrange : public BaseOperator
{
/**
......@@ -39,8 +44,8 @@ struct DeviceImageToColumn : public BaseOperator
* \param input_spatial_lengths Input spatial lengths.
* \param filter_spatial_lengths Filter spatial lengths.
* \param output_spatial_lengths Output spatial lengths.
* \param input_g_n_c_wis_strides Input strides in order [G, N, C, D, H, W].
* \param output_m_k_strides Output strides.
* \param image_g_n_c_wis_strides Image strides in order [G, N, C, D, H, W].
* \param gemm_m_k_strides Gemm form strides.
* \param conv_filter_strides Convolution filter strides.
* \param conv_filter_dilations Convolution filter dilations.
* \param input_left_pads Convolution left pads.
......@@ -55,8 +60,8 @@ struct DeviceImageToColumn : public BaseOperator
const std::array<index_t, NDimSpatial>& input_spatial_lengths,
const std::array<index_t, NDimSpatial>& filter_spatial_lengths,
const std::array<index_t, NDimSpatial>& output_spatial_lengths,
const std::array<index_t, NDimSpatial + 3>& input_g_n_c_wis_strides,
const std::array<index_t, 2>& output_m_k_strides,
const std::array<index_t, NDimSpatial + 3>& image_g_n_c_wis_strides,
const std::array<index_t, 2>& gemm_m_k_strides,
const std::array<index_t, NDimSpatial>& conv_filter_strides,
const std::array<index_t, NDimSpatial>& conv_filter_dilations,
const std::array<index_t, NDimSpatial>& input_left_pads,
......
include/ck/tensor_operation/gpu/device/device_gemm_multiple_abd.hpp 0 → 100755
View file @ 930b2872
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
#include <array>
#include "ck/tensor_operation/gpu/device/device_base.hpp"
namespace ck {
namespace tensor_operation {
namespace device {
// GEMM:
// input : A0[M, K], B0[K, N],
// input : D0[M, N], D1[M, N], ...
// output : E[M, N]
// C = a_op(A) * b_op(B)
// E = cde_op(C, D0, D1, ...)
// Assume:
// D0, D1, ... and E have the same layout
template <typename AsLayout,
typename BsLayout,
typename DsLayout,
typename ELayout,
typename AsDataType,
typename BsDataType,
typename DsDataType,
typename EDataType,
typename AElementwiseOperation,
typename BElementwiseOperation,
typename CDEElementwiseOperation>
struct DeviceGemmMultipleABD : public BaseOperator
{
static constexpr index_t NumATensor = AsDataType::Size();
static constexpr index_t NumBTensor = BsDataType::Size();
static constexpr index_t NumDTensor = DsDataType::Size();
virtual std::unique_ptr<BaseArgument>
MakeArgumentPointer(std::array<const void*, NumATensor> p_as,
std::array<const void*, NumBTensor> p_bs,
std::array<const void*, NumDTensor> p_ds,
void* p_e,
ck::index_t M,
ck::index_t N,
ck::index_t K,
std::array<ck::index_t, NumATensor> StrideAs,
std::array<ck::index_t, NumBTensor> StrideBs,
std::array<ck::index_t, NumDTensor> StrideDs,
ck::index_t StrideE,
AElementwiseOperation a_element_op,
BElementwiseOperation b_element_op,
CDEElementwiseOperation cde_element_op) = 0;
virtual std::unique_ptr<BaseInvoker> MakeInvokerPointer() = 0;
};
} // namespace device
} // namespace tensor_operation
} // namespace ck
include/ck/tensor_operation/gpu/device/device_grouped_conv_bwd_data_multiple_d.hpp
View file @ 930b2872
......@@ -29,7 +29,9 @@ template <ck::index_t NDimSpatial,
typename EDataType,
typename AElementwiseOperation,
typename BElementwiseOperation,
typename CDEElementwiseOperation>
typename CDEElementwiseOperation,
typename AComputeType = ADataType,
typename BComputeType = AComputeType>
struct DeviceGroupedConvBwdDataMultipleD : public BaseOperator
{
static constexpr index_t NumDTensor = DsDataType::Size();
......
include/ck/tensor_operation/gpu/device/device_grouped_conv_bwd_weight.hpp
View file @ 930b2872
......@@ -20,7 +20,9 @@ template <ck::index_t NDimSpatial,
typename OutDataType,
typename InElementwiseOperation,
typename WeiElementwiseOperation,
typename OutElementwiseOperation>
typename OutElementwiseOperation,
typename ComputeTypeA = InDataType,
typename ComputeTypeB = ComputeTypeA>
struct DeviceGroupedConvBwdWeight : public BaseOperator
{
virtual std::unique_ptr<BaseArgument>
......
include/ck/tensor_operation/gpu/device/device_grouped_conv_fwd_multiple_d.hpp
View file @ 930b2872
......@@ -29,7 +29,8 @@ template <index_t NDimSpatial,
typename EDataType,
typename AElementwiseOperation,
typename BElementwiseOperation,
typename CDEElementwiseOperation>
typename CDEElementwiseOperation,
typename ComputeType = ADataType>
struct DeviceGroupedConvFwdMultipleD : public BaseOperator
{
static constexpr index_t NumDTensor = DsDataType::Size();
......
include/ck/tensor_operation/gpu/device/device_grouped_gemm_splitk.hpp
View file @ 930b2872
......@@ -8,6 +8,57 @@ namespace ck {
namespace tensor_operation {
namespace device {
///
/// @brief Structure representing single GEMM problem arguments.
///
/// The pointer to the vector of those structures is passed
/// to the GroupedGEMM entry point kernel.
///
struct GroupedGemmKernelArguments
{
__host__ __device__ GroupedGemmKernelArguments(const void* p_a_grid_,
const void* p_b_grid_,
void* p_c_grid_,
index_t M_,
index_t N_,
index_t K_,
index_t StrideA_,
index_t StrideB_,
index_t StrideC_)
: p_a_grid{p_a_grid_},
p_b_grid{p_b_grid_},
p_c_grid{p_c_grid_},
M{M_},
N{N_},
K{K_},
StrideA{StrideA_},
StrideB{StrideB_},
StrideC{StrideC_}
{
}
const void* p_a_grid;
const void* p_b_grid;
void* p_c_grid;
index_t M;
index_t N;
index_t K;
index_t StrideA;
index_t StrideB;
index_t StrideC;
void Print() const
{
std::cout << "arg {"
<< "M:" << M << ", "
<< "N:" << N << ", "
<< "K:" << K << ", "
<< "SA:" << StrideA << ", "
<< "SB:" << StrideB << ", "
<< "SC:" << StrideC << "}" << std::endl;
}
};
template <typename ALayout,
typename BLayout,
typename DsLayout,
......@@ -31,7 +82,28 @@ struct DeviceGroupedGemmSplitK : public DeviceGroupedGemm<ALayout,
BElementwiseOperation,
CElementwiseOperation>
{
virtual void SetKBatchSize(BaseArgument* p_arg, index_t kbatch) const = 0;
//----------------------------------------------------------------------------------------------
/// @brief Sets the k batch size.
///
/// @param p_arg Pointer to the Argument we're going to change.
/// @param[in] kbatch The kbatch value.
///
virtual void SetKBatchSize([[maybe_unused]] BaseArgument* p_arg,
[[maybe_unused]] index_t kbatch) const
{
}
//----------------------------------------------------------------------------------------------
/// @brief Sets the device kernel arguments pointer.
///
/// @param p_arg The pointer to the Argument we're going to update.
/// @param[in] p_dev_kernel_args The pointer to the device memory which contains kernel
/// arguments.
///
virtual void SetDeviceKernelArgs([[maybe_unused]] BaseArgument* p_arg,
[[maybe_unused]] const void* p_dev_kernel_args) const
{
}
};
} // namespace device
......
include/ck/tensor_operation/gpu/device/device_softmax.hpp
View file @ 930b2872
......@@ -22,22 +22,22 @@ template <typename InDataType,
index_t NumReduceDim>
struct DeviceSoftmax : public BaseOperator
{
//
// @brief Makes a pointer to Argument class.
//
// @param[in] inLengths Input tensor extent(s) from high to low dimension
// @param[in] inStrides Input tensor stride(s) from high to low dimension
// @param[in] reduceDims The dimension(s) the normalization operation is applied
// @param[in] alpha double type value
// @param[in] beta double type value
// @param[in] in_dev Typeless const pointer in device memory storing the input
// tensor
// @param out_dev Typeless pointer in device memory storing the output tensor
// @param[in] in_elementwise_op The input elementwise operation.
// @param[in] acc_elementwise_op The accumulation elementwise operation.
//
// @return Unique pointer to the Argument class.
//
///
/// @brief Makes a pointer to Argument class.
///
/// @param[in] inLengths Input tensor extent(s) from high to low dimension
/// @param[in] inStrides Input tensor stride(s) from high to low dimension
/// @param[in] reduceDims The dimension(s) the normalization operation is applied
/// @param[in] alpha double type value
/// @param[in] beta double type value
/// @param[in] in_dev Typeless const pointer in device memory storing the input
/// tensor
/// @param out_dev Typeless pointer in device memory storing the output tensor
/// @param[in] in_elementwise_op The input elementwise operation.
/// @param[in] acc_elementwise_op The accumulation elementwise operation.
///
/// @return Unique pointer to the Argument class.
///
virtual std::unique_ptr<BaseArgument>
MakeArgumentPointer(const std::vector<index_t> inLengths,
const std::vector<index_t> inStrides,
......
include/ck/tensor_operation/gpu/device/impl/device_column_to_image_impl.hpp 0 → 100755
View file @ 930b2872
// SPDX-License-Identifier: MIT
// Copyright (c) 2023, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
#include "ck/tensor_description/tensor_descriptor.hpp"
#include "ck/tensor_description/tensor_descriptor_helper.hpp"
#include "ck/tensor_operation/gpu/device/device_conv_tensor_rearrange.hpp"
#include "ck/tensor_operation/gpu/grid/gridwise_tensor_rearrange.hpp"
#include "ck/host_utility/kernel_launch.hpp"
#include "ck/tensor_operation/gpu/device/convolution_backward_data_specialization.hpp"
#include "ck/tensor_operation/gpu/grid/block_to_ctile_map.hpp"
#include "ck/tensor_operation/operator_transform/transform_conv_fwd_to_gemm.hpp"
#include "ck/tensor_operation/gpu/device/matrix_padder.hpp"
#include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
#include "ck/tensor_operation/gpu/device/conv_tensor_rearrange_op.hpp"
#include "ck/host_utility/io.hpp"
namespace ck {
namespace tensor_operation {
namespace device {
// Image to column for input layout NDHWC:
// input : image converted to the gemm problem [N * Do * Ho * Wo, Z * Y * X * C]
// output : image [N, Di, Hi, Wi, C]
template <index_t NDimSpatial,
typename ImageLayout,
typename InputDataType,
typename OutputDataType,
index_t BlockSize,
index_t MPerBlock,
index_t KPerBlock,
typename ThreadClusterLengths,
index_t ScalarPerVector,
typename std::enable_if<NDimSpatial >= 1 && NDimSpatial <= 3, bool>::type = false>
struct DeviceColumnToImageImpl
: public DeviceConvTensorRearrange<NDimSpatial,
ImageLayout,
InputDataType,
OutputDataType,
conv_tensor_rearrange_op::ColumnToImage>
{
static constexpr auto I0 = Number<0>{};
static constexpr auto I1 = Number<1>{};
static constexpr auto I2 = Number<2>{};
static constexpr auto ZIdx = Number<I0>{};
static constexpr auto YIdx = NDimSpatial == 1 ? I0 : Number<NDimSpatial - I2>{};
static constexpr auto XIdx = Number<NDimSpatial - I1>{};
static constexpr auto spatial_offset = Number<3>{};
static constexpr auto conv_to_gemm_transformer =
TransformConvFwdToGemm<NDimSpatial, ConvolutionForwardSpecialization::Default>{};
static constexpr auto matrix_padder =
MatrixPadder<GemmSpecialization::MKPadding, index_t, index_t, index_t>{
MPerBlock, 0 /* NPerBlock*/, KPerBlock};
// Calculate number of independent filters for given conv params
static index_t GetNumberOfIndependentFilters(const index_t input_spatial_len,
const index_t left_pad,
const index_t right_pad,
const index_t filter_len,
const index_t filter_stride,
const index_t filter_dilation,
const index_t image_offset)
{
const index_t x_eff = (filter_len - 1) * filter_dilation + 1;
const index_t next_filter_padded =
math::integer_divide_ceil(x_eff, filter_stride) * filter_stride;
// If filter_stride >= x_eff then each filter is independent
const index_t independent_filter_stride =
filter_stride >= x_eff ? filter_stride : next_filter_padded;
const index_t w_eff = input_spatial_len - image_offset + left_pad + right_pad - x_eff;
// There are no independent filters
if(w_eff < 0)
return 0;
const index_t independent_kernels_num = w_eff / independent_filter_stride + 1;
return independent_kernels_num;
}
// Make column form descriptor
static auto
MakeInputDescriptor_M_K(const ck::index_t N,
const ck::index_t C,
const std::array<index_t, NDimSpatial>& filter_spatial_lengths,
const std::array<index_t, NDimSpatial>& output_spatial_lengths,
const std::array<index_t, NDimSpatial>& conv_filter_strides,
const std::array<index_t, 2>& gemm_m_k_strides,
const std::array<index_t, NDimSpatial>& independent_filters,
const std::array<index_t, NDimSpatial>& effs)
{
const index_t DoHoWo = ck::accumulate_n<index_t>(
output_spatial_lengths.begin(), NDimSpatial, 1, std::multiplies<>());
const index_t CZYX =
C * ck::accumulate_n<index_t>(
filter_spatial_lengths.begin(), NDimSpatial, 1, std::multiplies<>());
const index_t NStride = DoHoWo * gemm_m_k_strides[I0] * gemm_m_k_strides[I1];
// Calculate the appropriate stride for each set of independent filters
// in each dimension
const index_t WStride =
math::integer_divide_ceil(effs[XIdx], conv_filter_strides[XIdx]) * gemm_m_k_strides[I0];
const index_t HStride = math::integer_divide_ceil(effs[YIdx], conv_filter_strides[YIdx]) *
output_spatial_lengths[XIdx] * gemm_m_k_strides[I0];
const index_t DStride = math::integer_divide_ceil(effs[ZIdx], conv_filter_strides[ZIdx]) *
output_spatial_lengths[YIdx] * output_spatial_lengths[XIdx] *
gemm_m_k_strides[I0];
// Create descriptor for independent filters in each dimension and
// then merge them into column form
if constexpr(NDimSpatial == 1)
{
const auto desc_gemm_form =
make_naive_tensor_descriptor(make_tuple(N, independent_filters[XIdx], CZYX),
make_tuple(NStride, WStride, gemm_m_k_strides[I1]));
const auto desc_gemm_form_merged_filters = transform_tensor_descriptor(
desc_gemm_form,
make_tuple(make_merge_transform(make_tuple(N, independent_filters[XIdx])),
make_pass_through_transform(CZYX)),
make_tuple(Sequence<0, 1>{}, Sequence<2>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
const auto desc_m_k = matrix_padder.PadADescriptor_M_K(desc_gemm_form_merged_filters);
return desc_m_k;
}
else if constexpr(NDimSpatial == 2)
{
const auto desc_gemm_form = make_naive_tensor_descriptor(
make_tuple(N, independent_filters[YIdx], independent_filters[XIdx], CZYX),
make_tuple(NStride, HStride, WStride, gemm_m_k_strides[I1]));
const auto desc_gemm_form_merged_filters = transform_tensor_descriptor(
desc_gemm_form,
make_tuple(make_merge_transform(
make_tuple(N, independent_filters[YIdx], independent_filters[XIdx])),
make_pass_through_transform(CZYX)),
make_tuple(Sequence<0, 1, 2>{}, Sequence<3>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
const auto desc_m_k = matrix_padder.PadADescriptor_M_K(desc_gemm_form_merged_filters);
return desc_m_k;
}
else if constexpr(NDimSpatial == 3)
{
const auto desc_gemm_form = make_naive_tensor_descriptor(
make_tuple(N,
independent_filters[ZIdx],
independent_filters[YIdx],
independent_filters[XIdx],
CZYX),
make_tuple(NStride, DStride, HStride, WStride, gemm_m_k_strides[I1]));
const auto desc_gemm_form_merged_filters = transform_tensor_descriptor(
desc_gemm_form,
make_tuple(make_merge_transform(make_tuple(N,
independent_filters[ZIdx],
independent_filters[YIdx],
independent_filters[XIdx])),
make_pass_through_transform(CZYX)),
make_tuple(Sequence<0, 1, 2, 3>{}, Sequence<4>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}));
const auto desc_m_k = matrix_padder.PadADescriptor_M_K(desc_gemm_form_merged_filters);
return desc_m_k;
}
}
// Use MakeADescriptor_M_K from grouped convolution forward
static auto
MakeOutDescriptor_M_K(const ck::index_t N,
const ck::index_t C,
const std::array<index_t, NDimSpatial>& input_spatial_lengths,
const std::array<index_t, NDimSpatial>& filter_spatial_lengths,
const std::array<index_t, NDimSpatial + 3>& image_g_n_c_wis_strides,
const std::array<index_t, NDimSpatial>& conv_filter_strides,
const std::array<index_t, NDimSpatial>& conv_filter_dilations,
const std::array<index_t, NDimSpatial>& input_left_pads,
const std::array<index_t, NDimSpatial>& input_right_pads,
const std::array<index_t, NDimSpatial>& image_offsets,
const std::array<index_t, NDimSpatial>& independent_filters,
const std::array<index_t, NDimSpatial>& effs)
{
std::array<index_t, NDimSpatial + 3> a_g_n_c_wis_lengths{1};
std::array<index_t, NDimSpatial + 3> b_g_k_c_xs_lengths{1};
std::array<index_t, NDimSpatial + 3> c_g_n_k_wos_lengths{1};
auto copy = [](const auto& x, auto& y, index_t dst_offset) {
std::copy(x.begin(), x.end(), y.begin() + dst_offset);
};
copy(input_spatial_lengths, a_g_n_c_wis_lengths, spatial_offset);
copy(filter_spatial_lengths, b_g_k_c_xs_lengths, spatial_offset);
// Calculate descriptor only for independent filters
copy(independent_filters, c_g_n_k_wos_lengths, spatial_offset);
// fill only significant values (C and N)
a_g_n_c_wis_lengths[I1] = N;
a_g_n_c_wis_lengths[I2] = C;
b_g_k_c_xs_lengths[I2] = C;
c_g_n_k_wos_lengths[I1] = N;
// Modify pads to apply offsets
std::array<index_t, NDimSpatial> input_left_pads_with_offset;
for(index_t i = 0; i < NDimSpatial; i++)
{
input_left_pads_with_offset[i] = math::max(0, input_left_pads[i] - image_offsets[i]);
}
// Modify input spatial lengths to apply offsets
for(index_t i = 0; i < NDimSpatial; i++)
{
a_g_n_c_wis_lengths[i + spatial_offset] -=
math::max(0, image_offsets[i] - input_left_pads[i]);
}
// Strides to next independent filters
std::array<index_t, NDimSpatial> independent_filter_strides;
for(index_t i = 0; i < NDimSpatial; i++)
{
index_t independent_filter_stride =
math::integer_divide_ceil(effs[i], conv_filter_strides[i]) * conv_filter_strides[i];
// If conv stride is greater than whole filter size, use conv stride
independent_filter_strides[i] = conv_filter_strides[i] >= effs[i]
? conv_filter_strides[i]
: independent_filter_stride;
}
// Calculate image form descriptor for the modified convolution problem
const auto in_gemmmraw_gemmkraw_desc =
conv_to_gemm_transformer.template MakeADescriptor_M_K<ImageLayout>(
a_g_n_c_wis_lengths,
image_g_n_c_wis_strides,
b_g_k_c_xs_lengths,
{}, // not needed for A Descriptor
c_g_n_k_wos_lengths,
{}, // not needed for A Descriptor
// conv_filter_strides,
independent_filter_strides,
conv_filter_dilations,
input_left_pads_with_offset,
input_right_pads);
const auto in_gemmm_gemmk_desc =
matrix_padder.PadADescriptor_M_K(in_gemmmraw_gemmkraw_desc);
return in_gemmm_gemmk_desc;
}
using InputGridDesc =
remove_cvref_t<decltype(MakeInputDescriptor_M_K(1, 1, {}, {}, {}, {}, {}, {}))>;
using OutputGridDesc = remove_cvref_t<decltype(MakeOutDescriptor_M_K(
1, 1, {}, {}, {}, {}, {}, {}, {}, {}, {}, {}))>;
using Block2ETileMap = remove_cvref_t<
decltype(BlockToCTileMap_M00_N0_M01Adapt<MPerBlock, KPerBlock, InputGridDesc>(
InputGridDesc{}))>;
using GridwiseTensorRearrangeKernel = GridwiseTensorRearrange<InputGridDesc,
InputDataType,
OutputGridDesc,
OutputDataType,
BlockSize,
MPerBlock,
KPerBlock,
ThreadClusterLengths,
ScalarPerVector,
InMemoryDataOperationEnum::Add,
Block2ETileMap>;
struct Argument : public BaseArgument
{
Argument(const void* p_in, // input image
void* p_out, // output image
const ck::index_t N,
const ck::index_t C,
const std::array<index_t, NDimSpatial>& input_spatial_lengths,
const std::array<index_t, NDimSpatial>& filter_spatial_lengths,
const std::array<index_t, NDimSpatial>& output_spatial_lengths,
const std::array<index_t, NDimSpatial + 3>& image_g_n_c_wis_strides,
const std::array<index_t, 2>& gemm_m_k_strides,
const std::array<index_t, NDimSpatial>& conv_filter_strides,
const std::array<index_t, NDimSpatial>& conv_filter_dilations,
const std::array<index_t, NDimSpatial>& input_left_pads,
const std::array<index_t, NDimSpatial>& input_right_pads)
: C_(C),
X_(filter_spatial_lengths[NDimSpatial - I1]),
p_in_{static_cast<const InputDataType*>(p_in)},
p_out_{static_cast<OutputDataType*>(p_out)},
image_g_n_c_wis_strides_{image_g_n_c_wis_strides},
conv_filter_strides_{conv_filter_strides},
conv_filter_dilations_{conv_filter_dilations},
input_left_pads_{input_left_pads},
input_right_pads_{input_right_pads}
{
const index_t x_eff =
(filter_spatial_lengths[XIdx] - 1) * conv_filter_dilations[XIdx] + 1;
const index_t y_eff =
NDimSpatial < 2
? I1
: (filter_spatial_lengths[YIdx] - 1) * conv_filter_dilations[YIdx] + 1;
const index_t z_eff =
NDimSpatial < 3
? I1
: (filter_spatial_lengths[ZIdx] - 1) * conv_filter_dilations[ZIdx] + 1;
// Iterate over sets of independent filters
for(int z_img_offset = 0; z_img_offset < z_eff;
z_img_offset += conv_filter_strides[ZIdx])
{
for(int y_img_offset = 0; y_img_offset < y_eff;
y_img_offset += conv_filter_strides[YIdx])
{
for(int x_img_offset = 0; x_img_offset < x_eff;
x_img_offset += conv_filter_strides[XIdx])
{
std::array<index_t, NDimSpatial> image_offsets;
std::array<index_t, NDimSpatial> effs;
// Calculate the starting offset for a given set of
// independent filters
if constexpr(NDimSpatial == 1)
{
image_offsets = {x_img_offset};
effs = {x_eff};
}
if constexpr(NDimSpatial == 2)
{
image_offsets = {y_img_offset, x_img_offset};
effs = {y_eff, x_eff};
}
else if constexpr(NDimSpatial == 3)
{
image_offsets = {z_img_offset, y_img_offset, x_img_offset};
effs = {z_eff, y_eff, x_eff};
}
std::array<index_t, NDimSpatial> independent_filters;
for(index_t i = 0; i < NDimSpatial; i++)
{
independent_filters[i] =
GetNumberOfIndependentFilters(input_spatial_lengths[i],
input_left_pads[i],
input_right_pads[i],
filter_spatial_lengths[i],
conv_filter_strides[i],
conv_filter_dilations[i],
image_offsets[i]);
}
const index_t independent_filters_acum = ck::accumulate_n<index_t>(
independent_filters.begin(), NDimSpatial, 1, std::multiplies<>());
if(independent_filters_acum <= 0)
continue;
const auto in_grid_desc_m_k =
MakeInputDescriptor_M_K(N,
C,
filter_spatial_lengths,
output_spatial_lengths,
conv_filter_strides,
gemm_m_k_strides,
independent_filters,
effs);
const auto out_grid_desc_m_k =
MakeOutDescriptor_M_K(N,
C,
input_spatial_lengths,
filter_spatial_lengths,
image_g_n_c_wis_strides,
conv_filter_strides,
conv_filter_dilations,
input_left_pads,
input_right_pads,
image_offsets,
independent_filters,
effs);
in_grid_desc_m_k_container_.push_back(in_grid_desc_m_k);
out_grid_desc_m_k_container_.push_back(out_grid_desc_m_k);
const index_t x_idx = x_img_offset / conv_filter_strides[XIdx];
const index_t y_idx = y_img_offset / conv_filter_strides[YIdx];
const index_t z_idx = z_img_offset / conv_filter_strides[ZIdx];
const index_t x_offset_with_pad =
math::max(0, x_img_offset - input_left_pads[XIdx]);
const index_t y_offset_with_pad =
math::max(0, y_img_offset - input_left_pads[YIdx]);
const index_t z_offset_with_pad =
math::max(0, z_img_offset - input_left_pads[ZIdx]);
// Memory offsets to next set of independent filters,
// move to independent filters in each dimension
const index_t in_offset =
x_idx * gemm_m_k_strides[0] +
y_idx * gemm_m_k_strides[0] * output_spatial_lengths[XIdx] +
z_idx * gemm_m_k_strides[0] * output_spatial_lengths[YIdx] *
output_spatial_lengths[XIdx];
// Move to independent filters in appropriate dimensions
const index_t out_offset =
x_offset_with_pad * image_g_n_c_wis_strides[spatial_offset + XIdx] +
y_offset_with_pad * image_g_n_c_wis_strides[spatial_offset + YIdx] +
z_offset_with_pad * image_g_n_c_wis_strides[spatial_offset + ZIdx];
const InputDataType* p_in_with_offset =
static_cast<const InputDataType*>(p_in) + in_offset;
OutputDataType* p_out_with_offset =
static_cast<OutputDataType*>(p_out) + out_offset;
p_in_container_.push_back(p_in_with_offset);
p_out_container_.push_back(p_out_with_offset);
}
}
}
}
void Print() const
{
for(std::size_t i = 0; i < in_grid_desc_m_k_container_.size(); i++)
{
std::cout << in_grid_desc_m_k_container_[i] << std::endl;
std::cout << out_grid_desc_m_k_container_[i] << std::endl;
}
}
const ck::index_t C_;
const ck::index_t X_;
const InputDataType* p_in_;
OutputDataType* p_out_;
const std::array<index_t, NDimSpatial + 3>& image_g_n_c_wis_strides_;
const std::array<index_t, NDimSpatial>& conv_filter_strides_;
const std::array<index_t, NDimSpatial>& conv_filter_dilations_;
const std::array<index_t, NDimSpatial>& input_left_pads_;
const std::array<index_t, NDimSpatial>& input_right_pads_;
std::vector<InputGridDesc> in_grid_desc_m_k_container_;
std::vector<OutputGridDesc> out_grid_desc_m_k_container_;
std::vector<const InputDataType*> p_in_container_;
std::vector<OutputDataType*> p_out_container_;
};
struct Invoker : public BaseInvoker
{
float Run(const Argument& arg, const StreamConfig& stream_config = StreamConfig{})
{
if(stream_config.log_level_ > 0)
{
arg.Print();
}
float elapsed_time = 0.f;
const auto kernel = kernel_tensor_rearrange<InputGridDesc,
InputDataType,
OutputGridDesc,
OutputDataType,
Block2ETileMap,
GridwiseTensorRearrangeKernel>;
// Execute each set of independent filters
for(std::size_t i = 0; i < arg.in_grid_desc_m_k_container_.size(); i++)
{
const auto block_2_tile_map =
BlockToCTileMap_M00_N0_M01Adapt<MPerBlock, KPerBlock, InputGridDesc>(
arg.out_grid_desc_m_k_container_[i]);
const index_t grid_size =
block_2_tile_map.CalculateGridSize(arg.in_grid_desc_m_k_container_[i]);
elapsed_time += launch_and_time_kernel(stream_config,
kernel,
dim3(grid_size),
dim3(BlockSize),
0,
arg.in_grid_desc_m_k_container_[i],
arg.p_in_container_[i],
arg.out_grid_desc_m_k_container_[i],
arg.p_out_container_[i],
block_2_tile_map);
}
return elapsed_time;
}
float Run(const BaseArgument* p_arg,
const StreamConfig& stream_config = StreamConfig{}) override
{
return Run(*dynamic_cast<const Argument*>(p_arg), stream_config);
}
};
bool IsSupportedArgument(const Argument& arg)
{
using namespace tensor_layout::convolution;
if constexpr(!(std::is_same_v<ImageLayout, GNWC> || std::is_same_v<ImageLayout, GNHWC> ||
std::is_same_v<ImageLayout, GNDHWC>))
{
return false;
}
const auto w_pad_left = arg.input_left_pads_[NDimSpatial - I1];
const auto w_pad_right = arg.input_right_pads_[NDimSpatial - I1];
const auto dilation_x = arg.conv_filter_dilations_[NDimSpatial - I1];
const auto stride_x = arg.conv_filter_strides_[NDimSpatial - I1];
bool is_w_packed = arg.image_g_n_c_wis_strides_[NDimSpatial + I2] == arg.C_;
bool is_c_packed = arg.image_g_n_c_wis_strides_[I2] == 1;
// check vector acces with c not packed
if(!is_c_packed && ScalarPerVector != 1)
return false;
// check vector access of filter window row (only C if C is not packed)
if(!is_w_packed && arg.C_ % ScalarPerVector != 0)
return false;
// check vector access of filter window row (X * C)
if(arg.X_ * arg.C_ % ScalarPerVector != 0)
return false;
// check vector access of pads (w_pad_left/w_pad_right * C)
if(w_pad_left * arg.C_ % ScalarPerVector != 0 ||
w_pad_right * arg.C_ % ScalarPerVector != 0)
return false;
// check vector access of with stride and pad
if((w_pad_left != 0 || w_pad_right != 0) && stride_x > 1 && arg.C_ % ScalarPerVector != 0)
return false;
// check vector access of with dilation
if(dilation_x > 1 && arg.C_ % ScalarPerVector != 0)
return false;
bool valid = true;
for(std::size_t i = 0; i < arg.in_grid_desc_m_k_container_.size(); i++)
{
valid &= GridwiseTensorRearrangeKernel::CheckValidity(
arg.in_grid_desc_m_k_container_[i], arg.out_grid_desc_m_k_container_[i]);
}
return valid;
}
bool IsSupportedArgument(const BaseArgument* p_arg) override
{
return IsSupportedArgument(*dynamic_cast<const Argument*>(p_arg));
}
static auto MakeArgument(const void* p_in, // input image
void* p_out, // output image
const ck::index_t N,
const ck::index_t C,
const std::array<index_t, NDimSpatial>& input_spatial_lengths,
const std::array<index_t, NDimSpatial>& filter_spatial_lengths,
const std::array<index_t, NDimSpatial>& output_spatial_lengths,
const std::array<index_t, NDimSpatial + 3>& image_g_n_c_wis_strides,
const std::array<index_t, 2>& gemm_m_k_strides,
const std::array<index_t, NDimSpatial>& conv_filter_strides,
const std::array<index_t, NDimSpatial>& conv_filter_dilations,
const std::array<index_t, NDimSpatial>& input_left_pads,
const std::array<index_t, NDimSpatial>& input_right_pads)
{
return Argument{static_cast<const InputDataType*>(p_in),
static_cast<OutputDataType*>(p_out),
N,
C,
input_spatial_lengths,
filter_spatial_lengths,
output_spatial_lengths,
image_g_n_c_wis_strides,
gemm_m_k_strides,
conv_filter_strides,
conv_filter_dilations,
input_left_pads,
input_right_pads};
}
static auto MakeInvoker() { return Invoker{}; }
std::unique_ptr<BaseArgument>
MakeArgumentPointer(const void* p_in, // input image
void* p_out, // output image
const ck::index_t N,
const ck::index_t C,
const std::array<index_t, NDimSpatial>& input_spatial_lengths,
const std::array<index_t, NDimSpatial>& filter_spatial_lengths,
const std::array<index_t, NDimSpatial>& output_spatial_lengths,
const std::array<index_t, NDimSpatial + 3>& image_g_n_c_wis_strides,
const std::array<index_t, 2>& gemm_m_k_strides,
const std::array<index_t, NDimSpatial>& conv_filter_strides,
const std::array<index_t, NDimSpatial>& conv_filter_dilations,
const std::array<index_t, NDimSpatial>& input_left_pads,
const std::array<index_t, NDimSpatial>& input_right_pads) override
{
return std::make_unique<Argument>(static_cast<const InputDataType*>(p_in),
static_cast<OutputDataType*>(p_out),
N,
C,
input_spatial_lengths,
filter_spatial_lengths,
output_spatial_lengths,
image_g_n_c_wis_strides,
gemm_m_k_strides,
conv_filter_strides,
conv_filter_dilations,
input_left_pads,
input_right_pads);
}
std::unique_ptr<BaseInvoker> MakeInvokerPointer() override
{
return std::make_unique<Invoker>(Invoker{});
}
std::string GetTypeString() const override
{
auto str = std::stringstream();
// clang-format off
str << "DeviceColumnToImage"
<< "<"
<< BlockSize << ", "
<< MPerBlock << ", "
<< KPerBlock << ", "
<< ScalarPerVector
<< ">";
// clang-format on
return str.str();
}
};
} // namespace device
} // namespace tensor_operation
} // namespace ck
include/ck/tensor_operation/gpu/device/impl/device_gemm_multiple_abd_xdl_cshuffle.hpp 0 → 100755
View file @ 930b2872
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
#include <iostream>
#include <sstream>
#include "ck/utility/common_header.hpp"
#include "ck/tensor_description/tensor_descriptor.hpp"
#include "ck/tensor_description/tensor_descriptor_helper.hpp"
#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
#include "ck/tensor_operation/gpu/device/device_gemm_multiple_abd.hpp"
#include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
#include "ck/tensor_operation/gpu/device/matrix_padder.hpp"
#include "ck/tensor_operation/gpu/grid/gridwise_gemm_multiple_abd_xdl_cshuffle.hpp"
#include "ck/host_utility/device_prop.hpp"
#include "ck/host_utility/kernel_launch.hpp"
namespace ck {
template <typename GridwiseGemm,
typename AsPointer,
typename BsPointer,
typename DsPointer,
typename EDataType,
typename AElementwiseOperation,
typename BElementwiseOperation,
typename CDEElementwiseOperation,
typename AsGridDesc_AK0_M_AK1,
typename BsGridDesc_BK0_N_BK1,
typename DsGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock,
typename EGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock,
typename Block2ETileMap,
bool HasMainKBlockLoop>
__global__ void
#if CK_USE_LAUNCH_BOUNDS
__launch_bounds__(CK_MAX_THREAD_PER_BLOCK, CK_MIN_BLOCK_PER_CU)
#endif
kernel_gemm_multiple_abd_xdl_cshuffle(
AsPointer p_as_grid,
BsPointer p_bs_grid,
DsPointer p_ds_grid,
EDataType* __restrict__ p_e_grid,
const AElementwiseOperation a_element_op,
const BElementwiseOperation b_element_op,
const CDEElementwiseOperation cde_element_op,
const AsGridDesc_AK0_M_AK1 as_grid_desc_ak0_m_ak1,
const BsGridDesc_BK0_N_BK1 bs_grid_desc_bk0_n_bk1,
const DsGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock
ds_grid_desc_mblock_mperblock_nblock_nperblock,
const EGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock
e_grid_desc_mblock_mperblock_nblock_nperblock,
const Block2ETileMap block_2_etile_map)
{
#if(!defined(__HIP_DEVICE_COMPILE__) || defined(__gfx908__) || defined(__gfx90a__) || \
defined(__gfx940__) || defined(__gfx941__) || defined(__gfx942__))
__shared__ char p_shared[GridwiseGemm::GetSharedMemoryNumberOfByte()];
GridwiseGemm::template Run<HasMainKBlockLoop>(p_as_grid,
p_bs_grid,
p_ds_grid,
p_e_grid,
p_shared,
a_element_op,
b_element_op,
cde_element_op,
as_grid_desc_ak0_m_ak1,
bs_grid_desc_bk0_n_bk1,
ds_grid_desc_mblock_mperblock_nblock_nperblock,
e_grid_desc_mblock_mperblock_nblock_nperblock,
block_2_etile_map);
#else
ignore = p_as_grid;
ignore = p_bs_grid;
ignore = p_ds_grid;
ignore = p_e_grid;
ignore = a_element_op;
ignore = b_element_op;
ignore = cde_element_op;
ignore = as_grid_desc_ak0_m_ak1;
ignore = bs_grid_desc_bk0_n_bk1;
ignore = ds_grid_desc_mblock_mperblock_nblock_nperblock;
ignore = e_grid_desc_mblock_mperblock_nblock_nperblock;
ignore = block_2_etile_map;
#endif
}
} // namespace ck
namespace ck {
namespace tensor_operation {
namespace device {
// GEMM:
// input : A[M, K]
// input : B[N, K]
// input : D0[M, N], D1[M, N], ...
// output : E[M, N]
// C = a_op(A) * b_op(B)
// E = cde_op(C, D0, D1, ...)
// Assume:
// D0, D1, ... and E have the same layout
template <typename AsLayout,
typename BsLayout,
typename DsLayout,
typename ELayout,
typename AsDataType,
typename BsDataType,
typename AccDataType,
typename CShuffleDataType,
typename DsDataType,
typename EDataType,
typename AElementwiseOperation,
typename BElementwiseOperation,
typename CDEElementwiseOperation,
GemmSpecialization GemmSpec,
index_t NumGemmKPrefetchStage,
index_t BlockSize,
index_t MPerBlock,
index_t NPerBlock,
index_t KPerBlock,
index_t AK1,
index_t BK1,
index_t MPerXDL,
index_t NPerXDL,
index_t MXdlPerWave,
index_t NXdlPerWave,
typename ABlockTransferThreadClusterLengths_AK0_M_AK1,
typename ABlockTransferThreadClusterArrangeOrder,
typename ABlockTransferSrcAccessOrder,
index_t ABlockTransferSrcVectorDim,
index_t ABlockTransferSrcScalarPerVector,
index_t ABlockTransferDstScalarPerVector_AK1,
index_t ABlockLdsExtraM,
typename BBlockTransferThreadClusterLengths_BK0_N_BK1,
typename BBlockTransferThreadClusterArrangeOrder,
typename BBlockTransferSrcAccessOrder,
index_t BBlockTransferSrcVectorDim,
index_t BBlockTransferSrcScalarPerVector,
index_t BBlockTransferDstScalarPerVector_BK1,
index_t BBlockLdsExtraN,
index_t CShuffleMXdlPerWavePerShuffle,
index_t CShuffleNXdlPerWavePerShuffle,
typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock,
index_t CDEBlockTransferScalarPerVector_NPerBlock,
LoopScheduler LoopSched = make_default_loop_scheduler(),
PipelineVersion PipelineVer = PipelineVersion::v1>
struct DeviceGemmMultipleABD_Xdl_CShuffle : public DeviceGemmMultipleABD<AsLayout,
BsLayout,
DsLayout,
ELayout,
AsDataType,
BsDataType,
DsDataType,
EDataType,
AElementwiseOperation,
BElementwiseOperation,
CDEElementwiseOperation>
{
using DeviceOp = DeviceGemmMultipleABD_Xdl_CShuffle;
static constexpr index_t NumATensor = AsDataType::Size();
static constexpr index_t NumBTensor = BsDataType::Size();
static constexpr index_t NumDTensor = DsDataType::Size();
static constexpr auto I0 = Number<0>{};
static constexpr auto I1 = Number<1>{};
static constexpr auto I2 = Number<2>{};
static constexpr auto I3 = Number<3>{};
#if 0
static constexpr auto matrix_padder =
MatrixPadder<GemmSpec, index_t, index_t, index_t>{MPerBlock, NPerBlock, KPerBlock};
static auto MakeAGridDescriptor_M_K(index_t MRaw, index_t KRaw, index_t StrideAs)
{
const auto a_grid_desc_mraw_kraw = [&]() {
if constexpr(is_same_v<tensor_layout::gemm::RowMajor, AsLayout>)
{
return make_naive_tensor_descriptor(make_tuple(MRaw, KRaw),
make_tuple(StrideAs, I1));
}
else if constexpr(is_same_v<tensor_layout::gemm::ColumnMajor, AsLayout>)
{
return make_naive_tensor_descriptor(make_tuple(MRaw, KRaw),
make_tuple(I1, StrideAs));
}
}();
return matrix_padder.PadADescriptor_M_K(a_grid_desc_mraw_kraw);
}
static auto MakeBGridDescriptor_N_K(index_t KRaw, index_t NRaw, index_t StrideBs)
{
const auto b_grid_desc_nraw_kraw = [&]() {
if constexpr(is_same<tensor_layout::gemm::RowMajor, BsLayout>::value)
{
return make_naive_tensor_descriptor(make_tuple(NRaw, KRaw),
make_tuple(I1, StrideBs));
}
else if constexpr(is_same<tensor_layout::gemm::ColumnMajor, BsLayout>::value)
{
return make_naive_tensor_descriptor(make_tuple(NRaw, KRaw),
make_tuple(StrideBs, I1));
}
}();
return matrix_padder.PadBDescriptor_N_K(b_grid_desc_nraw_kraw);
}
template <typename ELay>
static auto MakeEGridDescriptor_M_N(index_t MRaw, index_t NRaw, index_t StrideE)
{
const auto e_grid_desc_mraw_nraw = [&]() {
if constexpr(is_same<tensor_layout::gemm::RowMajor, ELay>::value)
{
return make_naive_tensor_descriptor(make_tuple(MRaw, NRaw),
make_tuple(StrideE, I1));
}
else if constexpr(is_same<tensor_layout::gemm::ColumnMajor, ELay>::value)
{
return make_naive_tensor_descriptor(make_tuple(MRaw, NRaw),
make_tuple(I1, StrideE));
}
}();
return matrix_padder.PadCDescriptor_M_N(e_grid_desc_mraw_nraw);
}
static auto MakeDsGridDescriptor_M_N(const std::array<index_t, NumDTensor>& MRaws,
const std::array<index_t, NumDTensor>& NRaws,
const std::array<index_t, NumDTensor>& DsStride)
{
return generate_tuple(
[&](auto i) {
using DLayout = remove_cvref_t<tuple_element_t<i.value, DsLayout>>;
return DeviceOp::MakeEGridDescriptor_M_N<DLayout>(MRaws[i], NRaws[i], DsStride[i]);
},
Number<NumDTensor>{});
}
#endif
using ComputeDataType = EDataType;
// GridwiseGemm
using GridwiseGemm = GridwiseGemmMultipleABD_xdl_cshuffle<
AsDataType,
BsDataType,
ComputeDataType,
AccDataType,
CShuffleDataType,
DsDataType,
EDataType,
AElementwiseOperation,
BElementwiseOperation,
CDEElementwiseOperation,
InMemoryDataOperationEnum::Set,
NumGemmKPrefetchStage,
BlockSize,
MPerBlock,
NPerBlock,
KPerBlock,
AK1,
BK1,
MPerXDL,
NPerXDL,
MXdlPerWave,
NXdlPerWave,
ABlockTransferThreadClusterLengths_AK0_M_AK1,
ABlockTransferThreadClusterArrangeOrder,
ABlockTransferSrcAccessOrder,
ABlockTransferSrcVectorDim,
ABlockTransferSrcScalarPerVector,
ABlockTransferDstScalarPerVector_AK1,
false,
ABlockLdsExtraM,
BBlockTransferThreadClusterLengths_BK0_N_BK1,
BBlockTransferThreadClusterArrangeOrder,
BBlockTransferSrcAccessOrder,
BBlockTransferSrcVectorDim,
BBlockTransferSrcScalarPerVector,
BBlockTransferDstScalarPerVector_BK1,
false,
BBlockLdsExtraN,
CShuffleMXdlPerWavePerShuffle,
CShuffleNXdlPerWavePerShuffle,
CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock,
CDEBlockTransferScalarPerVector_NPerBlock,
LoopSched,
PipelineVer>;
// desc for problem definition
using AsGridDesc_M_K =
remove_cvref_t<decltype(GridwiseGemm::template MakeAsGridDescriptor_M_K<AsLayout, GemmSpec>(
{}, {}, {}))>;
using BsGridDesc_N_K =
remove_cvref_t<decltype(GridwiseGemm::template MakeBsGridDescriptor_N_K<BsLayout, GemmSpec>(
{}, {}, {}))>;
using DsGridDesc_M_N =
remove_cvref_t<decltype(GridwiseGemm::template MakeDsGridDescriptor_M_N<DsLayout, GemmSpec>(
{}, {}, {}))>;
using EGridDesc_M_N =
decltype(GridwiseGemm::template MakeEGridDescriptor_M_N<ELayout, GemmSpec>(1, 1, 1));
// desc for blockwise copy
using AsGridDesc_AK0_M_AK1 =
remove_cvref_t<decltype(GridwiseGemm::MakeAsGridDescriptor_AK0_M_AK1(AsGridDesc_M_K{}))>;
using BsGridDesc_BK0_N_BK1 =
remove_cvref_t<decltype(GridwiseGemm::MakeBsGridDescriptor_BK0_N_BK1(BsGridDesc_N_K{}))>;
using DsGridDesc_MBlock_MPerBlock_NBlock_NPerBlock = remove_cvref_t<
decltype(GridwiseGemm::MakeDsGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock(
DsGridDesc_M_N{}))>;
using EGridDesc_MBlock_MPerBlock_NBlock_NPerBlock =
remove_cvref_t<decltype(GridwiseGemm::MakeEGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock(
EGridDesc_M_N{}))>;
// block-to-e-tile map
using Block2ETileMap =
remove_cvref_t<decltype(GridwiseGemm::MakeBlock2ETileMap(EGridDesc_M_N{}))>;
// Argument
struct Argument : public BaseArgument
{
Argument(std::array<const void*, NumATensor> p_as_grid,
std::array<const void*, NumBTensor> p_bs_grid,
std::array<const void*, NumDTensor> p_ds_grid,
void* p_e_grid,
index_t MRaw,
index_t NRaw,
index_t KRaw,
std::array<index_t, NumATensor> StrideAs,
std::array<index_t, NumBTensor> StrideBs,
std::array<index_t, NumDTensor> StrideDs,
index_t StrideE,
AElementwiseOperation a_element_op,
BElementwiseOperation b_element_op,
CDEElementwiseOperation cde_element_op)
: p_as_grid_{},
p_bs_grid_{},
p_ds_grid_{},
p_e_grid_{static_cast<EDataType*>(p_e_grid)},
as_grid_desc_m_k_{},
bs_grid_desc_n_k_{},
ds_grid_desc_m_n_{},
e_grid_desc_m_n_{GridwiseGemm::template MakeEGridDescriptor_M_N<ELayout, GemmSpec>(
MRaw, NRaw, StrideE)},
as_grid_desc_ak0_m_ak1_{},
bs_grid_desc_bk0_n_bk1_{},
ds_grid_desc_mblock_mperblock_nblock_nperblock_{},
e_grid_desc_mblock_mperblock_nblock_nperblock_{},
block_2_etile_map_{GridwiseGemm::MakeBlock2ETileMap(e_grid_desc_m_n_)},
a_element_op_{a_element_op},
b_element_op_{b_element_op},
cde_element_op_{cde_element_op},
MRaw_{MRaw},
NRaw_{NRaw},
KRaw_{KRaw}
{
// populate pointer, desc for As
static_for<0, NumATensor, 1>{}([&](auto i) {
using ALayout = remove_cvref_t<tuple_element_t<i.value, AsLayout>>;
using ADataType = remove_cvref_t<tuple_element_t<i.value, AsDataType>>;
// A pointer
p_as_grid_(i) = static_cast<const ADataType*>(p_as_grid[i]);
// A desc
as_grid_desc_m_k_(i) =
GridwiseGemm::template MakeAGridDescriptor_M_K<ALayout, GemmSpec>(
MRaw, KRaw, StrideAs[i]);
});
// populate pointer, desc for Bs
static_for<0, NumBTensor, 1>{}([&](auto i) {
using BLayout = remove_cvref_t<tuple_element_t<i.value, BsLayout>>;
using BDataType = remove_cvref_t<tuple_element_t<i.value, BsDataType>>;
// B pointer
p_bs_grid_(i) = static_cast<const BDataType*>(p_bs_grid[i]);
// B desc
bs_grid_desc_n_k_(i) =
GridwiseGemm::template MakeBGridDescriptor_N_K<BLayout, GemmSpec>(
KRaw, NRaw, StrideBs[i]);
});
// populate pointer, desc for Ds
static_for<0, NumDTensor, 1>{}([&](auto i) {
using DLayout = remove_cvref_t<tuple_element_t<i.value, DsLayout>>;
using DDataType = remove_cvref_t<tuple_element_t<i.value, DsDataType>>;
// D pointer
p_ds_grid_(i) = static_cast<const DDataType*>(p_ds_grid[i]);
// D desc
ds_grid_desc_m_n_(i) =
GridwiseGemm::template MakeEGridDescriptor_M_N<DLayout, GemmSpec>(
MRaw, NRaw, StrideDs[i]);
});
// populate desc for Ds/E
if(GridwiseGemm::CheckValidity(as_grid_desc_m_k_,
bs_grid_desc_n_k_,
ds_grid_desc_m_n_,
e_grid_desc_m_n_,
block_2_etile_map_))
{
as_grid_desc_ak0_m_ak1_ =
GridwiseGemm::MakeAsGridDescriptor_AK0_M_AK1(as_grid_desc_m_k_);
bs_grid_desc_bk0_n_bk1_ =
GridwiseGemm::MakeBsGridDescriptor_BK0_N_BK1(bs_grid_desc_n_k_);
ds_grid_desc_mblock_mperblock_nblock_nperblock_ =
GridwiseGemm::MakeDsGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock(
ds_grid_desc_m_n_);
e_grid_desc_mblock_mperblock_nblock_nperblock_ =
GridwiseGemm::MakeEGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock(
e_grid_desc_m_n_);
}
}
void Print() const
{
// std::cout << "A[M, K]: " << as_grid_desc_m_k_ << std::endl;
// std::cout << "B[N, K]: " << bs_grid_desc_n_k_ << std::endl;
// static_for<0, NumDTensor, 1>{}(
//[&](auto i) { std::cout << "Ds[M, N]: " << ds_grid_desc_m_n_[i] << std::endl; });
// std::cout << "E[M, N]: " << e_grid_desc_m_n_ << std::endl;
}
// private:
// pointers
typename GridwiseGemm::AsGridPointer p_as_grid_;
typename GridwiseGemm::BsGridPointer p_bs_grid_;
typename GridwiseGemm::DsGridPointer p_ds_grid_;
EDataType* p_e_grid_;
// tensor descriptors for problem definiton
AsGridDesc_M_K as_grid_desc_m_k_;
BsGridDesc_N_K bs_grid_desc_n_k_;
DsGridDesc_M_N ds_grid_desc_m_n_;
EGridDesc_M_N e_grid_desc_m_n_;
// tensor descriptors for block/thread-wise copy
AsGridDesc_AK0_M_AK1 as_grid_desc_ak0_m_ak1_;
BsGridDesc_BK0_N_BK1 bs_grid_desc_bk0_n_bk1_;
DsGridDesc_MBlock_MPerBlock_NBlock_NPerBlock
ds_grid_desc_mblock_mperblock_nblock_nperblock_;
EGridDesc_MBlock_MPerBlock_NBlock_NPerBlock e_grid_desc_mblock_mperblock_nblock_nperblock_;
// block-to-e-tile map
Block2ETileMap block_2_etile_map_;
// element-wise op
AElementwiseOperation a_element_op_;
BElementwiseOperation b_element_op_;
CDEElementwiseOperation cde_element_op_;
// for checking vector load/store
index_t MRaw_;
index_t NRaw_;
index_t KRaw_;
};
// Invoker
struct Invoker : public BaseInvoker
{
using Argument = DeviceOp::Argument;
float Run(const Argument& arg, const StreamConfig& stream_config = StreamConfig{})
{
if(!GridwiseGemm::CheckValidity(arg.as_grid_desc_m_k_,
arg.bs_grid_desc_n_k_,
arg.ds_grid_desc_m_n_,
arg.e_grid_desc_m_n_,
arg.block_2_etile_map_))
{
throw std::runtime_error("wrong! GridwiseGemm has invalid setting");
}
const index_t grid_size =
arg.block_2_etile_map_.CalculateGridSize(arg.e_grid_desc_m_n_);
auto launch_kernel = [&](auto has_main_k_block_loop) {
constexpr bool has_main_loop = has_main_k_block_loop.value;
const auto kernel = kernel_gemm_multiple_abd_xdl_cshuffle<
GridwiseGemm,
typename GridwiseGemm::AsGridPointer,
typename GridwiseGemm::BsGridPointer,
typename GridwiseGemm::DsGridPointer,
EDataType,
AElementwiseOperation,
BElementwiseOperation,
CDEElementwiseOperation,
DeviceOp::AsGridDesc_AK0_M_AK1,
DeviceOp::BsGridDesc_BK0_N_BK1,
DeviceOp::DsGridDesc_MBlock_MPerBlock_NBlock_NPerBlock,
DeviceOp::EGridDesc_MBlock_MPerBlock_NBlock_NPerBlock,
DeviceOp::Block2ETileMap,
has_main_loop>;
return launch_and_time_kernel(stream_config,
kernel,
dim3(grid_size),
dim3(BlockSize),
0,
arg.p_as_grid_,
arg.p_bs_grid_,
arg.p_ds_grid_,
arg.p_e_grid_,
arg.a_element_op_,
arg.b_element_op_,
arg.cde_element_op_,
arg.as_grid_desc_ak0_m_ak1_,
arg.bs_grid_desc_bk0_n_bk1_,
arg.ds_grid_desc_mblock_mperblock_nblock_nperblock_,
arg.e_grid_desc_mblock_mperblock_nblock_nperblock_,
arg.block_2_etile_map_);
};
const auto K = arg.as_grid_desc_m_k_[I0].GetLength(I1);
if(GridwiseGemm::CalculateHasMainKBlockLoop(K))
{
return launch_kernel(integral_constant<bool, true>{});
}
else
{
return launch_kernel(integral_constant<bool, false>{});
}
}
// polymorphic
float Run(const BaseArgument* p_arg,
const StreamConfig& stream_config = StreamConfig{}) override
{
return Run(*dynamic_cast<const Argument*>(p_arg), stream_config);
}
};
static bool IsSupportedArgument(const Argument& arg)
{
if(!ck::is_xdl_supported())
{
return false;
}
// check vector load/store
{
using Row = ck::tensor_layout::gemm::RowMajor;
using Col = ck::tensor_layout::gemm::ColumnMajor;
bool all_valid = true;
static_for<0, NumATensor, 1>{}([&](auto i) {
using ALayout = remove_cvref_t<tuple_element_t<i.value, AsLayout>>;
// check vector load of A
if constexpr(is_same_v<ALayout, Row> && ABlockTransferSrcVectorDim == 2)
{
if(arg.KRaw_ % ABlockTransferSrcScalarPerVector != 0)
{
all_valid = false;
}
}
else if constexpr(is_same_v<ALayout, Col> && ABlockTransferSrcVectorDim == 1)
{
// FIXME: not rigorous
if(arg.MRaw_ % ABlockTransferSrcScalarPerVector != 0)
{
all_valid = false;
}
}
else
{
all_valid = false;
}
});
static_for<0, NumBTensor, 1>{}([&](auto i) {
using BLayout = remove_cvref_t<tuple_element_t<i.value, BsLayout>>;
// check vector laod of B
if constexpr(is_same_v<BLayout, Col> && BBlockTransferSrcVectorDim == 2)
{
if(arg.KRaw_ % BBlockTransferSrcScalarPerVector != 0)
{
all_valid = false;
}
}
else if constexpr(is_same_v<BLayout, Row> && BBlockTransferSrcVectorDim == 1)
{
// FIXME: not rigorous
if(arg.NRaw_ % BBlockTransferSrcScalarPerVector != 0)
{
all_valid = false;
}
}
else
{
all_valid = false;
}
});
// check vector load of Ds
// only support RowMajor for now
static_for<0, NumDTensor, 1>{}([&](auto i) {
using DLayout = remove_cvref_t<tuple_element_t<i.value, DsLayout>>;
if constexpr(!is_same_v<DLayout, Row>)
{
all_valid = false;
}
});
if(!all_valid)
{
return false;
}
// check vector store of E
// only support RowMajor for now
if constexpr(is_same_v<ELayout, Row>)
{
if(arg.NRaw_ % CDEBlockTransferScalarPerVector_NPerBlock != 0)
{
return false;
}
}
else
{
return false;
}
}
return GridwiseGemm::CheckValidity(arg.as_grid_desc_m_k_,
arg.bs_grid_desc_n_k_,
arg.ds_grid_desc_m_n_,
arg.e_grid_desc_m_n_,
arg.block_2_etile_map_);
}
// polymorphic
bool IsSupportedArgument(const BaseArgument* p_arg) override
{
return IsSupportedArgument(*dynamic_cast<const Argument*>(p_arg));
}
static auto MakeArgument(std::array<const void*, NumATensor> p_as,
std::array<const void*, NumBTensor> p_bs,
std::array<const void*, NumDTensor> p_ds,
void* p_e,
index_t MRaw,
index_t NRaw,
index_t KRaw,
std::array<index_t, NumATensor> StrideAs,
std::array<index_t, NumBTensor> StrideBs,
std::array<index_t, NumDTensor> StrideDs,
index_t StrideE,
AElementwiseOperation a_element_op,
BElementwiseOperation b_element_op,
CDEElementwiseOperation cde_element_op)
{
return Argument{p_as,
p_bs,
p_ds,
p_e,
MRaw,
NRaw,
KRaw,
StrideAs,
StrideBs,
StrideDs,
StrideE,
a_element_op,
b_element_op,
cde_element_op};
}
static auto MakeInvoker() { return Invoker{}; }
// polymorphic
std::unique_ptr<BaseArgument>
MakeArgumentPointer(std::array<const void*, NumATensor> p_as,
std::array<const void*, NumBTensor> p_bs,
std::array<const void*, NumDTensor> p_ds,
void* p_e,
index_t MRaw,
index_t NRaw,
index_t KRaw,
std::array<ck::index_t, NumATensor> StrideAs,
std::array<ck::index_t, NumBTensor> StrideBs,
std::array<ck::index_t, NumDTensor> StrideDs,
index_t StrideE,
AElementwiseOperation a_element_op,
BElementwiseOperation b_element_op,
CDEElementwiseOperation cde_element_op) override
{
return std::make_unique<Argument>(p_as,
p_bs,
p_ds,
p_e,
MRaw,
NRaw,
KRaw,
StrideAs,
StrideBs,
StrideDs,
StrideE,
a_element_op,
b_element_op,
cde_element_op);
}
// polymorphic
std::unique_ptr<BaseInvoker> MakeInvokerPointer() override
{
return std::make_unique<Invoker>(Invoker{});
}
// polymorphic
std::string GetTypeString() const override
{
auto str = std::stringstream();
std::map<LoopScheduler, std::string> LoopSchedToString{
{LoopScheduler::Default, "Default"}, {LoopScheduler::Interwave, "Interwave"}};
std::map<PipelineVersion, std::string> PipelineVersionToString{{PipelineVersion::v1, "v1"},
{PipelineVersion::v2, "v2"}};
// clang-format off
str << "DeviceGemmMultipleABD_Xdl_CShuffle"
<< "<"
<< BlockSize << ", "
<< MPerBlock << ", "
<< NPerBlock << ", "
<< KPerBlock << ", "
<< AK1 << ", "
<< BK1 << ", "
<< MPerXDL << ", "
<< NPerXDL << ", "
<< MXdlPerWave << ", "
<< NXdlPerWave << ", "
<< ABlockTransferSrcScalarPerVector << ", "
<< BBlockTransferSrcScalarPerVector << ", "
<< CShuffleMXdlPerWavePerShuffle << ", "
<< CShuffleNXdlPerWavePerShuffle << ", "
<< getGemmSpecializationString(GemmSpec)
<< ">"
<< " LoopScheduler: "
<< LoopSchedToString[LoopSched] << ", "
<< "PipelineVersion: "
<< PipelineVersionToString[PipelineVer];
// clang-format on
return str.str();
}
};
} // namespace device
} // namespace tensor_operation
} // namespace ck
include/ck/tensor_operation/gpu/device/impl/device_gemm_xdl_cshuffle.hpp
View file @ 930b2872
......@@ -66,7 +66,8 @@ template <typename ALayout,
index_t CShuffleBlockTransferScalarPerVector_NPerBlock,
LoopScheduler LoopSched = make_default_loop_scheduler(),
PipelineVersion PipelineVer = PipelineVersion::v1,
typename ComputeType = CDataType>
typename ComputeTypeA = CDataType,
typename ComputeTypeB = ComputeTypeA>
struct DeviceGemm_Xdl_CShuffle : public DeviceGemm<ALayout,
BLayout,
CLayout,
......@@ -131,7 +132,8 @@ struct DeviceGemm_Xdl_CShuffle : public DeviceGemm<ALayout,
CShuffleBlockTransferScalarPerVector_NPerBlock,
LoopSched,
PipelineVer,
ComputeType>;
ComputeTypeA,
ComputeTypeB>;
using Argument = typename GridwiseGemm::Argument;
......
include/ck/tensor_operation/gpu/device/impl/device_gemm_xdl_splitk_c_shuffle.hpp
View file @ 930b2872
......@@ -168,7 +168,7 @@ struct DeviceGemmXdlSplitKCShuffle : public DeviceGemmSplitK<ALayout,
stream_config.stream_id_));
ave_time = launch_and_time_kernel(
stream_config, kernel, dim3(gdx, gdy, gdz), dim3(BlockSize), 0, karg, b2c_map);
stream_config, kernel, dim3(gdx, gdy, gdz), dim3(BlockSize), 0, karg);
};
if(has_main_k0_block_loop)
......
include/ck/tensor_operation/gpu/device/impl/device_gemv_splitk.hpp
View file @ 930b2872
......@@ -303,7 +303,7 @@ struct deviceGemvDl : public DeviceGemv<ALayout,
StrideC,
GridwiseGemv::CalculateMPadded(M),
GridwiseGemv::CalculateNPadded(N),
GridwiseGemv::CalculateKPadded(K, KBatch),
K,
GridwiseGemv::CalculateK0(K, KBatch),
KBatch}; // //
}
......@@ -336,7 +336,7 @@ struct deviceGemvDl : public DeviceGemv<ALayout,
StrideC,
GridwiseGemv::CalculateMPadded(M),
GridwiseGemv::CalculateNPadded(N),
GridwiseGemv::CalculateKPadded(K, KBatch),
K,
GridwiseGemv::CalculateK0(K, KBatch),
KBatch); // //
}
......
include/ck/tensor_operation/gpu/device/impl/device_grouped_conv_bwd_data_multiple_d_wmma_cshuffle.hpp 0 → 100755
View file @ 930b2872
// SPDX-License-Identifier: MIT
// Copyright (c) 2023, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
#include <iostream>
#include <sstream>
#include "ck/utility/common_header.hpp"
#include "ck/tensor_description/tensor_descriptor.hpp"
#include "ck/tensor_description/tensor_descriptor_helper.hpp"
#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
#include "ck/tensor_operation/gpu/device/device_grouped_conv_bwd_data_multiple_d.hpp"
#include "ck/tensor_operation/gpu/device/convolution_backward_data_specialization.hpp"
#include "ck/tensor_operation/operator_transform/transform_conv_bwd_data_to_gemm_v1.hpp"
#include "ck/tensor_operation/gpu/grid/gridwise_gemm_multiple_d_wmma_cshuffle.hpp"
#include "ck/tensor_operation/gpu/device/impl/device_grouped_conv_utils.hpp"
#include "ck/host_utility/device_prop.hpp"
#include "ck/host_utility/kernel_launch.hpp"
#include "ck/host_utility/io.hpp"
namespace ck {
namespace tensor_operation {
namespace device {
// Conv backward data multiple D:
// input : output image A: [G, N, K, Ho, Wo]
// input : weight B: [G, K, C, Y, X],
// input : D0, D1, ... : [G, N, K, Ho, Wo]
// output : input image E: [G, N, C, Hi, Wi]
// C = a_op(A) * b_op(B)
// E = cde_op(C, D0, D1, ...)
template <index_t NDimSpatial,
typename ALayout, // output image
typename BLayout, // weight
typename DsLayout, // bias
typename ELayout, // input image
typename ADataType, // output image
typename BDataType, // weight
typename AccDataType,
typename CShuffleDataType,
typename DsDataType, // bias
typename EDataType, // input image
typename AElementwiseOp, // output image
typename BElementwiseOp, // weight
typename CDEElementwiseOp, // C, bias, and input image
ConvolutionBackwardDataSpecialization ConvBackwardDataSpecialization,
ck::index_t BlockSize,
ck::index_t MPerBlock,
ck::index_t NPerBlock,
ck::index_t K0PerBlock,
ck::index_t K1,
ck::index_t MPerWMMA,
ck::index_t NPerWMMA,
ck::index_t MRepeat,
ck::index_t NRepeat,
typename ABlockTransferThreadClusterLengths_AK0_M_AK1,
typename ABlockTransferThreadClusterArrangeOrder,
typename ABlockTransferSrcAccessOrder,
index_t ABlockTransferSrcVectorDim,
index_t ABlockTransferSrcScalarPerVector,
index_t ABlockTransferDstScalarPerVector_AK1,
bool ABlockLdsExtraM,
typename BBlockTransferThreadClusterLengths_BK0_N_BK1,
typename BBlockTransferThreadClusterArrangeOrder,
typename BBlockTransferSrcAccessOrder,
index_t BBlockTransferSrcVectorDim,
index_t BBlockTransferSrcScalarPerVector,
index_t BBlockTransferDstScalarPerVector_BK1,
bool BBlockLdsExtraN,
index_t CShuffleMRepeatPerShuffle,
index_t CShuffleNRepeatPerShuffle,
typename CDEShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock,
index_t CDEShuffleBlockTransferScalarPerVector_NPerBlock,
index_t NumGemmKPrefetchStage = 1,
LoopScheduler LoopSched = make_default_loop_scheduler(),
ck::PipelineVersion PipelineVer = ck::PipelineVersion::v1>
struct DeviceGroupedConvBwdDataMultipleD_Wmma_CShuffle
: public DeviceGroupedConvBwdDataMultipleD<NDimSpatial,
ALayout, // output image
BLayout, // weight
DsLayout, // bias
ELayout, // input image
ADataType, // output image
BDataType, // weight
DsDataType, // bias
EDataType, // input image
AElementwiseOp,
BElementwiseOp,
CDEElementwiseOp>
{
// TODO: Extend support for more spatial dimensions.
static_assert(NDimSpatial == 2 || NDimSpatial == 3,
"wrong! only implemented for 2D and 3D now");
using DeviceOp = DeviceGroupedConvBwdDataMultipleD_Wmma_CShuffle;
static constexpr index_t NumDTensor = DsDataType::Size();
// TODO: Add support for different A and B data types.
using ABDataType = ADataType;
static constexpr auto I0 = Number<0>{};
static constexpr auto I1 = Number<1>{};
static constexpr auto I2 = Number<2>{};
static constexpr auto I3 = Number<3>{};
static constexpr index_t KPerBlock = K0PerBlock * K1;
static constexpr auto transform_conv_to_gemm =
TransformConvBwdDataToGemm_v1<NDimSpatial,
ConvBackwardDataSpecialization,
K1,
K1,
MPerBlock,
NPerBlock,
KPerBlock,
true /* DoPadGemmM */,
true /* DoPadGemmN */>{};
static auto GetDummyABDsEGridDescriptor()
{
const std::array<index_t, NDimSpatial + 3> dummy_tensor_lengths = {1};
const std::array<index_t, NDimSpatial + 3> dummy_tensor_strides = {1};
const std::array<index_t, NDimSpatial> dummy_spatial_lengths = {1};
const auto a_grid_desc_ak0_m_ak1 =
transform_conv_to_gemm.template MakeADescriptor_AK0_M_AK1<ALayout>(
dummy_tensor_lengths,
dummy_tensor_strides,
dummy_tensor_lengths,
dummy_tensor_strides,
dummy_tensor_lengths,
dummy_tensor_strides,
dummy_spatial_lengths,
dummy_spatial_lengths,
dummy_spatial_lengths,
dummy_spatial_lengths,
dummy_spatial_lengths);
const auto b_grid_desc_bk0_n_bk1 =
transform_conv_to_gemm.template MakeBDescriptor_BK0_N_BK1<BLayout>(
dummy_tensor_lengths,
dummy_tensor_strides,
dummy_tensor_lengths,
dummy_tensor_strides,
dummy_tensor_lengths,
dummy_tensor_strides,
dummy_spatial_lengths,
dummy_spatial_lengths,
dummy_spatial_lengths,
dummy_spatial_lengths,
dummy_spatial_lengths);
const auto ds_grid_desc_m_n = generate_tuple(
[&](auto i) {
using DLayout = remove_cvref_t<tuple_element_t<i.value, DsLayout>>;
return transform_conv_to_gemm.template MakeCDescriptor_M_N<DLayout>(
dummy_tensor_lengths,
dummy_tensor_strides,
dummy_tensor_lengths,
dummy_tensor_strides,
dummy_tensor_lengths,
dummy_tensor_strides,
dummy_spatial_lengths,
dummy_spatial_lengths,
dummy_spatial_lengths,
dummy_spatial_lengths,
dummy_spatial_lengths);
},
Number<NumDTensor>{});
const auto e_grid_desc_m_n =
transform_conv_to_gemm.template MakeCDescriptor_M_N<ELayout>(dummy_tensor_lengths,
dummy_tensor_strides,
dummy_tensor_lengths,
dummy_tensor_strides,
dummy_tensor_lengths,
dummy_tensor_strides,
dummy_spatial_lengths,
dummy_spatial_lengths,
dummy_spatial_lengths,
dummy_spatial_lengths,
dummy_spatial_lengths);
return make_tuple(
a_grid_desc_ak0_m_ak1, b_grid_desc_bk0_n_bk1, ds_grid_desc_m_n, e_grid_desc_m_n);
}
// desc
using ABDsEGridDesc = decltype(GetDummyABDsEGridDescriptor());
using AGridDesc_AK0_M_AK1 = remove_cvref_t<tuple_element_t<0, ABDsEGridDesc>>;
using BGridDesc_BK0_N_BK1 = remove_cvref_t<tuple_element_t<1, ABDsEGridDesc>>;
using DsGridDesc_M_N = remove_cvref_t<tuple_element_t<2, ABDsEGridDesc>>;
using EGridDesc_M_N = remove_cvref_t<tuple_element_t<3, ABDsEGridDesc>>;
// GridwiseGemm
using GridwiseGemm = GridwiseGemmMultipleD_k0mk1_k0nk1_mn_wmma_cshuffle<
// DataType Family
ADataType,
BDataType,
AccDataType,
CShuffleDataType,
DsDataType,
EDataType,
// InMemory Data Descriptor
AGridDesc_AK0_M_AK1,
BGridDesc_BK0_N_BK1,
DsGridDesc_M_N,
EGridDesc_M_N,
// ElementwiseOp Family
AElementwiseOp,
BElementwiseOp,
CDEElementwiseOp,
InMemoryDataOperationEnum::Set,
// Tiling Family
MPerBlock,
NPerBlock,
K0PerBlock,
MPerWMMA,
NPerWMMA,
K1,
MRepeat,
NRepeat,
// ThreadCluster Family
BlockSize,
ABlockTransferThreadClusterLengths_AK0_M_AK1,
ABlockTransferThreadClusterArrangeOrder,
ABlockTransferSrcAccessOrder,
ABlockTransferSrcVectorDim,
ABlockTransferSrcScalarPerVector,
ABlockTransferDstScalarPerVector_AK1,
false,
ABlockLdsExtraM,
BBlockTransferThreadClusterLengths_BK0_N_BK1,
BBlockTransferThreadClusterArrangeOrder,
BBlockTransferSrcAccessOrder,
BBlockTransferSrcVectorDim,
BBlockTransferSrcScalarPerVector,
BBlockTransferDstScalarPerVector_BK1,
false,
BBlockLdsExtraN,
CShuffleMRepeatPerShuffle,
CShuffleNRepeatPerShuffle,
CDEShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock,
CDEShuffleBlockTransferScalarPerVector_NPerBlock,
NumGemmKPrefetchStage,
LoopSched,
PipelineVer>;
using DsGridDesc_MBlock_MPerBlock_NBlock_NPerBlock =
decltype(GridwiseGemm::MakeDsGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock(
DsGridDesc_M_N{}));
using EGridDesc_MBlock_MPerBlock_NBlock_NPerBlock =
decltype(GridwiseGemm::MakeEGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock(
EGridDesc_M_N{}));
// Argument
struct Argument : public BaseArgument
{
Argument(const void* p_a, // output image
const void* p_b, // weight
const std::array<const void*, NumDTensor>& p_ds, // bias
void* p_e, // input image
const std::array<index_t, NDimSpatial + 3>& a_g_n_k_wos_lengths,
const std::array<index_t, NDimSpatial + 3>& a_g_n_k_wos_strides,
const std::array<index_t, NDimSpatial + 3>& b_g_k_c_xs_lengths,
const std::array<index_t, NDimSpatial + 3>& b_g_k_c_xs_strides,
const std::array<std::array<index_t, NDimSpatial + 3>, NumDTensor>&
ds_g_n_c_wis_lengths,
const std::array<std::array<index_t, NDimSpatial + 3>, NumDTensor>&
ds_g_n_c_wis_strides,
const std::array<index_t, NDimSpatial + 3>& e_g_n_c_wis_lengths,
const std::array<index_t, NDimSpatial + 3>& e_g_n_c_wis_strides,
const std::array<index_t, NDimSpatial>& conv_filter_strides,
const std::array<index_t, NDimSpatial>& conv_filter_dilations,
const std::array<index_t, NDimSpatial>& input_left_pads,
const std::array<index_t, NDimSpatial>& input_right_pads,
const AElementwiseOp& a_element_op,
const BElementwiseOp& b_element_op,
const CDEElementwiseOp& cde_element_op)
: p_a_grid_{static_cast<const ADataType*>(p_a)},
p_b_grid_{static_cast<const BDataType*>(p_b)},
p_ds_grid_{},
p_e_grid_{static_cast<EDataType*>(p_e)},
num_group_{a_g_n_k_wos_lengths[0]},
a_element_op_{a_element_op},
b_element_op_{b_element_op},
cde_element_op_{cde_element_op},
a_g_n_k_wos_lengths_{a_g_n_k_wos_lengths},
a_g_n_k_wos_strides_{a_g_n_k_wos_strides},
b_g_k_c_xs_lengths_{b_g_k_c_xs_lengths},
b_g_k_c_xs_strides_{b_g_k_c_xs_strides},
ds_g_n_c_wis_lengths_{ds_g_n_c_wis_lengths},
ds_g_n_c_wis_strides_{ds_g_n_c_wis_strides},
e_g_n_c_wis_lengths_{e_g_n_c_wis_lengths},
e_g_n_c_wis_strides_{e_g_n_c_wis_strides},
conv_filter_strides_{conv_filter_strides},
conv_filter_dilations_{conv_filter_dilations},
input_left_pads_{input_left_pads},
input_right_pads_{input_right_pads}
{
// populate Ds pointer
static_for<0, NumDTensor, 1>{}([&](auto i) {
using DDataType = remove_cvref_t<tuple_element_t<i.value, DsDataType>>;
p_ds_grid_(i) = static_cast<const DDataType*>(p_ds[i]);
});
// A/B/Ds/E Batch Stride
compute_ptr_offset_of_batch_.BatchStrideA_ = a_g_n_k_wos_strides[0];
compute_ptr_offset_of_batch_.BatchStrideB_ = b_g_k_c_xs_strides[0];
compute_ptr_offset_of_batch_.BatchStrideE_ = e_g_n_c_wis_strides[0];
static_for<0, NumDTensor, 1>{}([&](auto i) {
compute_ptr_offset_of_batch_.BatchStrideDs_(i) = ds_g_n_c_wis_strides[i][0];
});
static constexpr auto NonSpatialDimsNum = Number<3>{};
static constexpr auto DIdx = Number<NonSpatialDimsNum>{};
static constexpr auto HIdx =
NDimSpatial == 2 ? Number<NonSpatialDimsNum>{} : Number<NonSpatialDimsNum + 1>{};
static constexpr auto WIdx = NDimSpatial == 2 ? Number<NonSpatialDimsNum + 1>{}
: Number<NonSpatialDimsNum + 2>{};
static constexpr auto ZIdx = Number<NonSpatialDimsNum>{};
static constexpr auto YIdx =
NDimSpatial == 2 ? Number<NonSpatialDimsNum>{} : Number<NonSpatialDimsNum + 1>{};
static constexpr auto XIdx = NDimSpatial == 2 ? Number<NonSpatialDimsNum + 1>{}
: Number<NonSpatialDimsNum + 2>{};
// problem definition
const index_t Z = b_g_k_c_xs_lengths[ZIdx];
const index_t Y = b_g_k_c_xs_lengths[YIdx];
const index_t X = b_g_k_c_xs_lengths[XIdx];
const index_t ConvStrideD = conv_filter_strides[DIdx - NonSpatialDimsNum];
const index_t ConvStrideH = conv_filter_strides[HIdx - NonSpatialDimsNum];
const index_t ConvStrideW = conv_filter_strides[WIdx - NonSpatialDimsNum];
const index_t ConvDilationD = conv_filter_dilations[DIdx - NonSpatialDimsNum];
const index_t ConvDilationH = conv_filter_dilations[HIdx - NonSpatialDimsNum];
const index_t ConvDilationW = conv_filter_dilations[WIdx - NonSpatialDimsNum];
const auto GcdStrideDilationD = math::gcd(ConvStrideD, ConvDilationD);
const auto GcdStrideDilationH = math::gcd(ConvStrideH, ConvDilationH);
const auto GcdStrideDilationW = math::gcd(ConvStrideW, ConvDilationW);
const auto ZTilde = NDimSpatial == 3 ? ConvStrideD / GcdStrideDilationD : 1;
const auto YTilde = ConvStrideH / GcdStrideDilationH;
const auto XTilde = ConvStrideW / GcdStrideDilationW;
for(index_t i_ztilde = 0; i_ztilde < ZTilde; ++i_ztilde)
{
for(index_t i_ytilde = 0; i_ytilde < YTilde; ++i_ytilde)
{
for(index_t i_xtilde = 0; i_xtilde < XTilde; ++i_xtilde)
{
// check slice is valid
const auto ZDotSlice =
NDimSpatial == 3 ? math::integer_divide_ceil(Z - i_ztilde, ZTilde) : 1;
const auto YDotSlice = math::integer_divide_ceil(Y - i_ytilde, YTilde);
const auto XDotSlice = math::integer_divide_ceil(X - i_xtilde, XTilde);
if(YDotSlice * XDotSlice * ZDotSlice <= 0)
{
continue;
}
std::array<index_t, NDimSpatial> tildes;
if constexpr(NDimSpatial == 2)
{
tildes = {i_ytilde, i_xtilde};
}
else if constexpr(NDimSpatial == 3)
{
tildes = {i_ztilde, i_ytilde, i_xtilde};
}
else
{
throw std::runtime_error("wrong! only implemented for 2D and 3D now");
}
const auto a_grid_desc_ak0_m_ak1 =
transform_conv_to_gemm.template MakeADescriptor_AK0_M_AK1<ALayout>(
a_g_n_k_wos_lengths,
a_g_n_k_wos_strides,
b_g_k_c_xs_lengths,
b_g_k_c_xs_strides,
e_g_n_c_wis_lengths,
e_g_n_c_wis_strides,
conv_filter_strides,
conv_filter_dilations,
input_left_pads,
input_right_pads,
tildes);
const auto b_grid_desc_bk0_n_bk1 =
transform_conv_to_gemm.template MakeBDescriptor_BK0_N_BK1<BLayout>(
a_g_n_k_wos_lengths,
a_g_n_k_wos_strides,
b_g_k_c_xs_lengths,
b_g_k_c_xs_strides,
e_g_n_c_wis_lengths,
e_g_n_c_wis_strides,
conv_filter_strides,
conv_filter_dilations,
input_left_pads,
input_right_pads,
tildes);
DsGridDesc_M_N ds_grid_desc_m_n;
// populate Ds desc
static_for<0, NumDTensor, 1>{}([&](auto i) {
using DLayout = remove_cvref_t<tuple_element_t<i.value, DsLayout>>;
ds_grid_desc_m_n(i) =
transform_conv_to_gemm.template MakeCDescriptor_M_N<DLayout>(
a_g_n_k_wos_lengths,
a_g_n_k_wos_strides,
b_g_k_c_xs_lengths,
b_g_k_c_xs_strides,
ds_g_n_c_wis_lengths[i],
ds_g_n_c_wis_strides[i],
conv_filter_strides,
conv_filter_dilations,
input_left_pads,
input_right_pads,
tildes);
});
const auto e_grid_desc_m_n =
transform_conv_to_gemm.template MakeCDescriptor_M_N<ELayout>(
a_g_n_k_wos_lengths,
a_g_n_k_wos_strides,
b_g_k_c_xs_lengths,
b_g_k_c_xs_strides,
e_g_n_c_wis_lengths,
e_g_n_c_wis_strides,
conv_filter_strides,
conv_filter_dilations,
input_left_pads,
input_right_pads,
tildes);
// for check validity
ds_grid_desc_m_n_container_.push_back(ds_grid_desc_m_n);
e_grid_desc_m_n_container_.push_back(e_grid_desc_m_n);
// desc for blockwise copy
a_grid_desc_ak0_m_ak1_container_.push_back(a_grid_desc_ak0_m_ak1);
b_grid_desc_bk0_n_bk1_container_.push_back(b_grid_desc_bk0_n_bk1);
// block-to-e-tile-map
auto block_2_ctile_map = GridwiseGemm::MakeDefaultBlock2CTileMap(
e_grid_desc_m_n, 1 /* M01 */, 1 /* N01 */);
block_2_ctile_map_container_.push_back(block_2_ctile_map);
ds_grid_desc_mblock_mperblock_nblock_nperblock_container_.push_back(
GridwiseGemm::MakeDsGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock(
ds_grid_desc_m_n));
e_grid_desc_mblock_mperblock_nblock_nperblock_container_.push_back(
GridwiseGemm::MakeEGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock(
e_grid_desc_m_n));
}
}
}
}
void Print() const
{
for(std::size_t i = 0; i < a_grid_desc_ak0_m_ak1_container_.size(); i++)
{
std::cout << "a_grid_desc_ak0_m_ak1_container_"
<< a_grid_desc_ak0_m_ak1_container_[i] << std::endl;
std::cout << "b_grid_desc_bk0_n_bk1_container_"
<< b_grid_desc_bk0_n_bk1_container_[i] << std::endl;
static_for<0, NumDTensor, 1>{}([&](auto j) {
std::cout << "ds_grid_desc_mblock_mperblock_nblock_nperblock_container_"
<< ds_grid_desc_mblock_mperblock_nblock_nperblock_container_[i][j]
<< std::endl;
});
std::cout << "e_grid_desc_mblock_mperblock_nblock_nperblock_container_"
<< e_grid_desc_mblock_mperblock_nblock_nperblock_container_[i]
<< std::endl;
}
}
// pointers
const ADataType* p_a_grid_;
const BDataType* p_b_grid_;
typename GridwiseGemm::DsGridPointer p_ds_grid_;
EDataType* p_e_grid_;
// tensor descriptor for problem definition
index_t num_group_;
std::vector<DsGridDesc_M_N> ds_grid_desc_m_n_container_;
std::vector<EGridDesc_M_N> e_grid_desc_m_n_container_;
// tensor descriptor for block-wise copy
std::vector<AGridDesc_AK0_M_AK1> a_grid_desc_ak0_m_ak1_container_;
std::vector<BGridDesc_BK0_N_BK1> b_grid_desc_bk0_n_bk1_container_;
std::vector<DsGridDesc_MBlock_MPerBlock_NBlock_NPerBlock>
ds_grid_desc_mblock_mperblock_nblock_nperblock_container_;
std::vector<EGridDesc_MBlock_MPerBlock_NBlock_NPerBlock>
e_grid_desc_mblock_mperblock_nblock_nperblock_container_;
// block-to-e-tile map
std::vector<typename GridwiseGemm::DefaultBlock2CTileMap> block_2_ctile_map_container_;
// for computing batch offset
ComputePtrOffsetOfStridedBatch<NumDTensor> compute_ptr_offset_of_batch_;
// element-wise op
AElementwiseOp a_element_op_;
BElementwiseOp b_element_op_;
CDEElementwiseOp cde_element_op_;
// for checking IsSupportedArgument()
std::array<index_t, NDimSpatial + 3> a_g_n_k_wos_lengths_;
std::array<index_t, NDimSpatial + 3> a_g_n_k_wos_strides_;
std::array<index_t, NDimSpatial + 3> b_g_k_c_xs_lengths_;
std::array<index_t, NDimSpatial + 3> b_g_k_c_xs_strides_;
std::array<std::array<index_t, NDimSpatial + 3>, NumDTensor> ds_g_n_c_wis_lengths_;
std::array<std::array<index_t, NDimSpatial + 3>, NumDTensor> ds_g_n_c_wis_strides_;
std::array<index_t, NDimSpatial + 3> e_g_n_c_wis_lengths_;
std::array<index_t, NDimSpatial + 3> e_g_n_c_wis_strides_;
std::array<index_t, NDimSpatial> conv_filter_strides_;
std::array<index_t, NDimSpatial> conv_filter_dilations_;
std::array<index_t, NDimSpatial> input_left_pads_;
std::array<index_t, NDimSpatial> input_right_pads_;
};
// Invoker
struct Invoker : public BaseInvoker
{
using Argument = DeviceOp::Argument;
float Run(const Argument& arg, const StreamConfig& stream_config = StreamConfig{})
{
if(stream_config.log_level_ > 0)
{
arg.Print();
}
float ave_time = 0;
for(std::size_t i = 0; i < arg.a_grid_desc_ak0_m_ak1_container_.size(); i++)
{
const index_t grid_size = arg.block_2_ctile_map_container_[i].CalculateGridSize(
arg.e_grid_desc_m_n_container_[i]) *
arg.num_group_;
const auto GemmK = arg.a_grid_desc_ak0_m_ak1_container_[i].GetLength(I0) *
arg.a_grid_desc_ak0_m_ak1_container_[i].GetLength(I2);
auto launch_kernel = [&](auto has_main_k_block_loop) {
constexpr bool has_main_loop = has_main_k_block_loop.value;
const auto kernel = kernel_grouped_conv_fwd_multiple_d_wmma_cshuffle<
GridwiseGemm,
ADataType,
BDataType,
typename GridwiseGemm::DsGridPointer,
EDataType,
AElementwiseOp,
BElementwiseOp,
CDEElementwiseOp,
DeviceOp::AGridDesc_AK0_M_AK1,
DeviceOp::BGridDesc_BK0_N_BK1,
typename GridwiseGemm::DsGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock,
typename GridwiseGemm::EGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock,
remove_reference_t<typename GridwiseGemm::DefaultBlock2CTileMap>,
ComputePtrOffsetOfStridedBatch<NumDTensor>,
has_main_loop>;
return launch_and_time_kernel(
stream_config,
kernel,
dim3(grid_size),
dim3(BlockSize),
0,
arg.p_a_grid_,
arg.p_b_grid_,
arg.p_ds_grid_,
arg.p_e_grid_,
arg.a_element_op_,
arg.b_element_op_,
arg.cde_element_op_,
arg.a_g_n_k_wos_lengths_[0], // Group count
arg.a_grid_desc_ak0_m_ak1_container_[i],
arg.b_grid_desc_bk0_n_bk1_container_[i],
arg.ds_grid_desc_mblock_mperblock_nblock_nperblock_container_[i],
arg.e_grid_desc_mblock_mperblock_nblock_nperblock_container_[i],
arg.block_2_ctile_map_container_[i],
arg.compute_ptr_offset_of_batch_);
};
if(GridwiseGemm::CalculateHasMainKBlockLoop(GemmK))
{
ave_time += launch_kernel(integral_constant<bool, true>{});
}
else
{
ave_time += launch_kernel(integral_constant<bool, false>{});
}
}
return ave_time;
}
float Run(const BaseArgument* p_arg,
const StreamConfig& stream_config = StreamConfig{}) override
{
return Run(*dynamic_cast<const Argument*>(p_arg), stream_config);
}
};
static bool IsSupportedArgument(const Argument& arg)
{
// check device
if(get_device_name() == "gfx1100" || get_device_name() == "gfx1101" ||
ck::get_device_name() == "gfx1102")
{
if constexpr(!(is_same_v<AccDataType, float> || is_same_v<AccDataType, int32_t>))
{
return false;
}
}
else
{
return false;
}
const index_t ConvK = arg.b_g_k_c_xs_lengths_[1];
const index_t ConvC = arg.b_g_k_c_xs_lengths_[2];
// Specialization
if constexpr(ConvBackwardDataSpecialization ==
ConvolutionBackwardDataSpecialization::Filter1x1Stride1Pad0)
{
// check if it's a 1x1 convolution with stride=1 and no padding
for(int i = 0; i < NDimSpatial; i++)
{
if(!(arg.b_g_k_c_xs_lengths_[3 + i] == 1 && arg.conv_filter_strides_[i] == 1 &&
arg.input_left_pads_[i] == 0 && arg.input_right_pads_[i] == 0))
{
return false;
}
}
}
// vector load for A matrix from global memory to LDS
if constexpr(is_same_v<ALayout, tensor_layout::convolution::GNHWK> ||
is_same_v<ALayout, tensor_layout::convolution::GNDHWK> ||
is_same_v<ALayout, tensor_layout::convolution::NHWGK> ||
is_same_v<ALayout, tensor_layout::convolution::NDHWGK>)
{
if(!(ABlockTransferSrcVectorDim == 2 && ConvK % ABlockTransferSrcScalarPerVector == 0))
{
return false;
}
}
else
{
return false;
}
// vector load for B matrix from global memory to LDS
if constexpr(is_same_v<BLayout, tensor_layout::convolution::GKYXC> ||
is_same_v<BLayout, tensor_layout::convolution::GKZYXC>)
{
if(!(BBlockTransferSrcVectorDim == 1 && ConvC % BBlockTransferSrcScalarPerVector == 0))
{
return false;
}
}
else
{
return false;
}
// vector store for Ds
bool ds_valid = true;
static_for<0, NumDTensor, 1>{}([&](auto i) {
using DLayout = remove_cvref_t<tuple_element_t<i.value, DsLayout>>;
if constexpr(is_same_v<DLayout, tensor_layout::convolution::GNHWC> ||
is_same_v<DLayout, tensor_layout::convolution::GNDHWC> ||
is_same_v<DLayout, tensor_layout::convolution::NHWGC> ||
is_same_v<DLayout, tensor_layout::convolution::NDHWGC> ||
is_same_v<DLayout, tensor_layout::convolution::G_NHW_C> ||
is_same_v<DLayout, tensor_layout::convolution::GC> ||
is_same_v<DLayout, tensor_layout::convolution::G_C>)
{
// vector load D matrix from global memory
if(!(ConvC % CDEShuffleBlockTransferScalarPerVector_NPerBlock == 0))
{
ds_valid = false;
}
}
else
{
ds_valid = false;
}
});
if(!ds_valid)
{
return false;
}
// vector store for E
if constexpr(is_same_v<ELayout, tensor_layout::convolution::GNHWC> ||
is_same_v<ELayout, tensor_layout::convolution::GNDHWC> ||
is_same_v<ELayout, tensor_layout::convolution::NHWGC> ||
is_same_v<ELayout, tensor_layout::convolution::NDHWGC>)
{
// vector store C matrix into global memory
if(!(ConvC % CDEShuffleBlockTransferScalarPerVector_NPerBlock == 0))
{
return false;
}
}
else
{
return false;
}
// Gridwise GEMM size
for(std::size_t i = 0; i < arg.a_grid_desc_ak0_m_ak1_container_.size(); i++)
{
if(!GridwiseGemm::CheckValidity(arg.a_grid_desc_ak0_m_ak1_container_[i],
arg.b_grid_desc_bk0_n_bk1_container_[i],
arg.ds_grid_desc_m_n_container_[i],
arg.e_grid_desc_m_n_container_[i],
arg.block_2_ctile_map_container_[i]))
{
return false;
}
}
return true;
}
bool IsSupportedArgument(const BaseArgument* p_arg) override
{
return IsSupportedArgument(*dynamic_cast<const Argument*>(p_arg));
}
static auto
MakeArgument(const void* p_a, // output image
const void* p_b, // weight
const std::array<const void*, NumDTensor>& p_ds, // bias
void* p_e, // input image
const std::array<index_t, NDimSpatial + 3>& a_g_n_k_wos_lengths, // output image
const std::array<index_t, NDimSpatial + 3>& a_g_n_k_wos_strides, // output image
const std::array<index_t, NDimSpatial + 3>& b_g_k_c_xs_lengths, // weight
const std::array<index_t, NDimSpatial + 3>& b_g_k_c_xs_strides, // weight
const std::array<std::array<index_t, NDimSpatial + 3>, NumDTensor>&
ds_g_n_c_wis_lengths, // bias
const std::array<std::array<index_t, NDimSpatial + 3>, NumDTensor>&
ds_g_n_c_wis_strides, // bias
const std::array<index_t, NDimSpatial + 3>& e_g_n_c_wis_lengths, // input image
const std::array<index_t, NDimSpatial + 3>& e_g_n_c_wis_strides, // input image
const std::array<index_t, NDimSpatial>& conv_filter_strides,
const std::array<index_t, NDimSpatial>& conv_filter_dilations,
const std::array<index_t, NDimSpatial>& input_left_pads,
const std::array<index_t, NDimSpatial>& input_right_pads,
const AElementwiseOp& a_element_op,
const BElementwiseOp& b_element_op,
const CDEElementwiseOp& cde_element_op)
{
return Argument{p_a,
p_b,
p_ds,
p_e,
a_g_n_k_wos_lengths,
a_g_n_k_wos_strides,
b_g_k_c_xs_lengths,
b_g_k_c_xs_strides,
ds_g_n_c_wis_lengths,
ds_g_n_c_wis_strides,
e_g_n_c_wis_lengths,
e_g_n_c_wis_strides,
conv_filter_strides,
conv_filter_dilations,
input_left_pads,
input_right_pads,
a_element_op,
b_element_op,
cde_element_op};
}
static auto MakeInvoker() { return Invoker{}; }
std::unique_ptr<BaseArgument> MakeArgumentPointer(
const void* p_a, // output image
const void* p_b, // weight
const std::array<const void*, NumDTensor>& p_ds, // bias
void* p_e, // input image
const std::array<index_t, NDimSpatial + 3>& a_g_n_k_wos_lengths, // output image
const std::array<index_t, NDimSpatial + 3>& a_g_n_k_wos_strides, // output image
const std::array<index_t, NDimSpatial + 3>& b_g_k_c_xs_lengths, // weight
const std::array<index_t, NDimSpatial + 3>& b_g_k_c_xs_strides, // weight
const std::array<std::array<index_t, NDimSpatial + 3>, NumDTensor>&
ds_g_n_c_wis_lengths, // bias
const std::array<std::array<index_t, NDimSpatial + 3>, NumDTensor>&
ds_g_n_c_wis_strides, // bias
const std::array<index_t, NDimSpatial + 3>& e_g_n_c_wis_lengths, // input image
const std::array<index_t, NDimSpatial + 3>& e_g_n_c_wis_strides, // input image
const std::array<index_t, NDimSpatial>& conv_filter_strides,
const std::array<index_t, NDimSpatial>& conv_filter_dilations,
const std::array<index_t, NDimSpatial>& input_left_pads,
const std::array<index_t, NDimSpatial>& input_right_pads,
const AElementwiseOp& a_element_op,
const BElementwiseOp& b_element_op,
const CDEElementwiseOp& cde_element_op) override
{
return std::make_unique<Argument>(p_a,
p_b,
p_ds,
p_e,
a_g_n_k_wos_lengths,
a_g_n_k_wos_strides,
b_g_k_c_xs_lengths,
b_g_k_c_xs_strides,
ds_g_n_c_wis_lengths,
ds_g_n_c_wis_strides,
e_g_n_c_wis_lengths,
e_g_n_c_wis_strides,
conv_filter_strides,
conv_filter_dilations,
input_left_pads,
input_right_pads,
a_element_op,
b_element_op,
cde_element_op);
}
std::unique_ptr<BaseInvoker> MakeInvokerPointer() override
{
return std::make_unique<Invoker>(Invoker{});
}
std::string GetTypeString() const override
{
auto str = std::stringstream();
// clang-format off
str << "DeviceGroupedConvBwdDataMultipleD_Wmma_CShuffle"
<< "<"
<< BlockSize << ", "
<< MPerBlock << ", "
<< NPerBlock << ", "
<< KPerBlock << ", "
<< getConvBackwardDataSpecializationString(ConvBackwardDataSpecialization) << ", "
<< K1 << ", "
<< ABlockTransferSrcScalarPerVector << ", "
<< BBlockTransferSrcScalarPerVector
<< ">";
// clang-format on
return str.str();
}
};
} // namespace device
} // namespace tensor_operation
} // namespace ck
include/ck/tensor_operation/gpu/device/impl/device_grouped_conv_bwd_data_multiple_d_xdl_cshuffle_v1.hpp
View file @ 930b2872
......@@ -14,6 +14,7 @@
#include "ck/tensor_operation/gpu/device/convolution_backward_data_specialization.hpp"
#include "ck/tensor_operation/operator_transform/transform_conv_bwd_data_to_gemm_v1.hpp"
#include "ck/tensor_operation/gpu/grid/gridwise_gemm_multiple_d_xdl_cshuffle.hpp"
#include "ck/tensor_operation/gpu/device/impl/device_grouped_conv_utils.hpp"
#include "ck/host_utility/device_prop.hpp"
#include "ck/host_utility/kernel_launch.hpp"
#include "ck/host_utility/io.hpp"
......@@ -24,51 +25,6 @@ namespace device {
namespace {
template <index_t NumDTensor>
struct ComputePtrOffsetOfStridedBatch
{
ComputePtrOffsetOfStridedBatch() = default;
ComputePtrOffsetOfStridedBatch(index_t BatchStrideA,
index_t BatchStrideB,
Array<ck::index_t, NumDTensor> BatchStrideDs,
index_t BatchStrideE)
: BatchStrideA_(BatchStrideA),
BatchStrideB_(BatchStrideB),
BatchStrideDs_(BatchStrideDs),
BatchStrideE_(BatchStrideE)
{
}
__host__ __device__ constexpr long_index_t GetAPtrOffset(index_t g_idx) const
{
return g_idx * static_cast<long_index_t>(BatchStrideA_);
}
__host__ __device__ constexpr long_index_t GetBPtrOffset(index_t g_idx) const
{
return g_idx * static_cast<long_index_t>(BatchStrideB_);
}
__host__ __device__ constexpr auto GetDsPtrOffset(index_t g_idx) const
{
Array<long_index_t, NumDTensor> ds_offset;
static_for<0, NumDTensor, 1>{}(
[&](auto i) { ds_offset(i) = g_idx * static_cast<long_index_t>(BatchStrideDs_[i]); });
return ds_offset;
}
__host__ __device__ constexpr long_index_t GetEPtrOffset(index_t g_idx) const
{
return g_idx * static_cast<long_index_t>(BatchStrideE_);
}
index_t BatchStrideA_;
index_t BatchStrideB_;
Array<ck::index_t, NumDTensor> BatchStrideDs_;
index_t BatchStrideE_;
};
/*
* \brief Wrapper function of GridwiseGemm::Run to realize BatchedGEMM.
*
......@@ -242,7 +198,9 @@ template <index_t NDimSpatial,
index_t CShuffleNXdlPerWavePerShuffle,
typename CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock,
index_t CDEBlockTransferScalarPerVector_NPerBlock,
LoopScheduler LoopSched = make_default_loop_scheduler()>
LoopScheduler LoopSched = make_default_loop_scheduler(),
typename AComputeType = ADataType,
typename BComputeType = AComputeType>
struct DeviceGroupedConvBwdDataMultipleD_Xdl_CShuffle_v1
: public DeviceGroupedConvBwdDataMultipleD<NDimSpatial,
ALayout, // output image
......@@ -255,9 +213,11 @@ struct DeviceGroupedConvBwdDataMultipleD_Xdl_CShuffle_v1
EDataType, // input image
AElementwiseOp,
BElementwiseOp,
CDEElementwiseOp>
CDEElementwiseOp,
AComputeType,
BComputeType>
{
// FIXME
// TODO: Extend support for more spatial dimensions.
static_assert(NDimSpatial == 2 || NDimSpatial == 3,
"wrong! only implemented for 2D and 3D now");
......@@ -265,7 +225,7 @@ struct DeviceGroupedConvBwdDataMultipleD_Xdl_CShuffle_v1
static constexpr index_t NumDTensor = DsDataType::Size();
// TODO make A/B datatype different
// TODO: Add support for different A and B data types.
using ABDataType = ADataType;
static constexpr auto I0 = Number<0>{};
......@@ -356,9 +316,9 @@ struct DeviceGroupedConvBwdDataMultipleD_Xdl_CShuffle_v1
// GridwiseGemm
using GridwiseGemm = GridwiseGemmMultipleD_xdl_cshuffle<
ABDataType, // TODO: distinguish A/B datatype
ABDataType, // TODO: distinguish A/B datatype
ABDataType, // TODO: distinguish A/B datatype
ABDataType,
ABDataType,
AComputeType,
AccDataType,
CShuffleDataType,
DsDataType,
......@@ -398,7 +358,9 @@ struct DeviceGroupedConvBwdDataMultipleD_Xdl_CShuffle_v1
CShuffleNXdlPerWavePerShuffle,
CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock,
CDEBlockTransferScalarPerVector_NPerBlock,
LoopSched>;
LoopSched,
PipelineVersion::v1,
BComputeType>;
template <typename Desc_K0_M_K1>
static auto transform_k0_m_k1_to_m_k(const Desc_K0_M_K1& desc_k0_m_k1)
......
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