mirror of
https://github.com/ROCm/composable_kernel.git
synced 2026-04-20 14:59:17 +00:00
Add json dump support to output details from CK/CKTile Examples. (#2551)
* Adding RapidJson Library * Adding Json Dumps in all CK_Tile Examples Not verified yet * Adding json to cktile Batched Transpose * adding json dumps to layernorm2d_fwd * Adding json dump to flatmm_basic * Adding RapidJson Library * Adding Json Dumps in all CK_Tile Examples Not verified yet * Adding json to cktile Batched Transpose * adding json dumps to layernorm2d_fwd * Adding json dump to flatmm_basic * Adding json in 03_gemm * Add json dump to 16_batched_gemm * Add json dump to gemm_multi_d_fp16 * Add json dump to grouped_gemm * fix fmha_bwd/fwd * Fix clang-format errors exclude include/rapidjson in jenkins as its a third-party library * Saparating function and defination. * Update Documentation of 03_gemm * Refactoring as per code review * Disable fp8 instances on unsupported targets (#2592) * Restrict building of gemm_universal_preshuffle_f8 instances to specific targets in CMakeLists.txt * Add condition to skip gemm_xdl_universal_preshuffle_f8 instances for unsupported targets in CMakeLists.txt * Add conditions to skip unsupported targets for gemm_universal_preshuffle_f8 and gemm_xdl_universal_preshuffle_f8 instances in CMakeLists.txt * Refine conditions to exclude gemm_universal_preshuffle_f8 instances for unsupported targets in CMakeLists.txt --------- Co-authored-by: AviralGoelAMD <aviralgoel@amd.com> * fix clang format * remove duplicate lines of code from library/src/tensor_operation_instance/gpu/CMakeLists.txt * Fixing Readme and unifying jsondumps * adding moe_smoothquant * adding fused_moe * Fixing Readme for batched_gemm * Fixing Readme for grouped_gemm * adding flatmm * adding gemm_multi_d_fp16 * adding elementwise * adding File name when json is dumped * Fixing Reduce after merge * adding batched_transpose * Adding Warptile in Gemm * Fixing Clang Format --------- Co-authored-by: Aviral Goel <aviral.goel@amd.com> Co-authored-by: AviralGoelAMD <aviralgoel@amd.com> Co-authored-by: illsilin_amdeng <Illia.Silin@amd.com>
This commit is contained in:
rahjain-amd
@@ -17,19 +17,21 @@ This will result in an executable `build/bin/tile_example_gemm_multi_d_fp16`
|
||||
## example
|
||||
```
|
||||
args:
|
||||
-m M dimensions - (Default: 3840)
|
||||
-n N dimensions - (Default: 4096)
|
||||
-k K dimensions - (Default: 4096)
|
||||
-a_layout Tensor A layout (default:R)
|
||||
-b_layout Tensor B layout (default:C)
|
||||
-ds_layout Tensor D layout (default:R)
|
||||
-e_layout Tensor E layout (default:R)
|
||||
-stride_a Tensor A strides - (Default: 0)
|
||||
-stride_b Tensor B strides - (Default: 0)
|
||||
-stride_e Tensor C strides - (Default: 0)
|
||||
-stride_ds Tensor D strides - (Default: 0)
|
||||
-validate 0. No validation, 1. Validation on GPU. (Default: 1)
|
||||
-warmup Number of iterations before benchmark the kernel. (Default: 10)
|
||||
-repeat Number of iterations to benchmark the kernel. (Default: 100)
|
||||
-kbatch kbatch for SplitK. (Default 1)
|
||||
-m m dimension (default:3840)
|
||||
-n n dimension (default:4096)
|
||||
-k k dimension (default:4096)
|
||||
-a_layout A tensor data layout - Row by default (default:R)
|
||||
-b_layout B tensor data layout - Col by default (default:C)
|
||||
-ds_layout Ds tensor data layout - Row by default (default:R)
|
||||
-e_layout E tensor data layout - Row by default (default:R)
|
||||
-stride_a Tensor A stride (default:0)
|
||||
-stride_b Tensor B stride (default:0)
|
||||
-stride_ds Tensor Ds stride (default:0)
|
||||
-stride_e Tensor E stride (default:0)
|
||||
-v 0. No validation, 1. Validation on GPU (default:1)
|
||||
-warmup number of iterations before benchmark the kernel (default:50)
|
||||
-repeat number of iterations to benchmark the kernel (default:100)
|
||||
-kbatch kbatch for SplitK (default:1)
|
||||
-json 0: No Json, 1: Dump Results in Json format (default:0)
|
||||
-jsonfile json file name to dump results (default:cktile_gemm_multi_d_fp16.json)
|
||||
```
|
||||
|
||||
@@ -58,7 +58,9 @@ auto create_args(int argc, char* argv[])
|
||||
.insert("v", "1", "0. No validation, 1. Validation on GPU")
|
||||
.insert("warmup", "50", "number of iterations before benchmark the kernel")
|
||||
.insert("repeat", "100", "number of iterations to benchmark the kernel")
|
||||
.insert("kbatch", "1", "kbatch for SplitK");
|
||||
.insert("kbatch", "1", "kbatch for SplitK")
|
||||
.insert("json", "0", "0: No Json, 1: Dump Results in Json format")
|
||||
.insert("jsonfile", "cktile_gemm_multi_d_fp16.json", "json file name to dump results");
|
||||
|
||||
bool result = arg_parser.parse(argc, argv);
|
||||
return std::make_tuple(result, arg_parser);
|
||||
|
||||
@@ -3,6 +3,7 @@
|
||||
|
||||
#pragma once
|
||||
#include <cstddef>
|
||||
#include "json_dump.hpp"
|
||||
|
||||
template <typename ADataType,
|
||||
typename BDataType,
|
||||
@@ -54,30 +55,6 @@ float invoke_gemm_multi_d(const void* a_m_k_dev_buf,
|
||||
CDEElementWise>(
|
||||
gemm_descs, ck_tile::stream_config{nullptr, true, 1, n_warmup, n_repeat});
|
||||
|
||||
std::string op_name{"Gemm Multiple-D"};
|
||||
static constexpr ck_tile::index_t NumDTensor = DsDataType::size();
|
||||
|
||||
std::size_t flop = 0, num_btype = 0;
|
||||
|
||||
flop += std::size_t(2) * M * N * K;
|
||||
|
||||
ck_tile::static_for<0, NumDTensor, 1>{}([&](auto i) {
|
||||
num_btype += sizeof(ck_tile::remove_cvref_t<std::tuple_element_t<i, DsDataType>>) * M * N;
|
||||
flop += sizeof(ck_tile::remove_cvref_t<std::tuple_element_t<i, DsDataType>>) * M * N;
|
||||
});
|
||||
|
||||
num_btype += sizeof(ADataType) * M * K + sizeof(BDataType) * K * N + sizeof(EDataType) * M * N;
|
||||
|
||||
float tflops = static_cast<float>(flop) / 1.E9 / ave_time;
|
||||
float gb_per_sec = num_btype / 1.E6 / ave_time;
|
||||
|
||||
std::cout << "Run Gemm Multiple-D kernel with:\n";
|
||||
std::cout << "M =" << M << " N =" << N << " K =" << K << "\n";
|
||||
std::cout << "StrideA = " << StrideA << " StrideB = " << StrideB << " StrideE = " << StrideE
|
||||
<< "\n";
|
||||
std::cout << "Perf: " << ave_time << " ms, " << tflops << " TFlops, " << gb_per_sec << " GB/s, "
|
||||
<< "\n";
|
||||
|
||||
return ave_time;
|
||||
}
|
||||
|
||||
@@ -159,29 +136,53 @@ int run_multiple_d_gemm_example_with_layouts(int argc,
|
||||
|
||||
std::array<ck_tile::index_t, DsDataType::size()> stridesDs = {StrideD0, StrideD1};
|
||||
|
||||
invoke_gemm_multi_d<ADataType,
|
||||
BDataType,
|
||||
DsDataType,
|
||||
AccDataType,
|
||||
EDataType,
|
||||
ALayout,
|
||||
BLayout,
|
||||
DsLayout,
|
||||
ELayout,
|
||||
CDElementWiseFn>(a_m_k_dev_buf.GetDeviceBuffer(),
|
||||
b_k_n_dev_buf.GetDeviceBuffer(),
|
||||
ds_ptr_buf,
|
||||
e_m_n_dev_buf.GetDeviceBuffer(),
|
||||
M,
|
||||
N,
|
||||
K,
|
||||
StrideA,
|
||||
StrideB,
|
||||
stridesDs,
|
||||
StrideE,
|
||||
n_warmup,
|
||||
n_repeat,
|
||||
k_batch);
|
||||
float ave_time = invoke_gemm_multi_d<ADataType,
|
||||
BDataType,
|
||||
DsDataType,
|
||||
AccDataType,
|
||||
EDataType,
|
||||
ALayout,
|
||||
BLayout,
|
||||
DsLayout,
|
||||
ELayout,
|
||||
CDElementWiseFn>(a_m_k_dev_buf.GetDeviceBuffer(),
|
||||
b_k_n_dev_buf.GetDeviceBuffer(),
|
||||
ds_ptr_buf,
|
||||
e_m_n_dev_buf.GetDeviceBuffer(),
|
||||
M,
|
||||
N,
|
||||
K,
|
||||
StrideA,
|
||||
StrideB,
|
||||
stridesDs,
|
||||
StrideE,
|
||||
n_warmup,
|
||||
n_repeat,
|
||||
k_batch);
|
||||
|
||||
std::string op_name{"Gemm Multiple-D"};
|
||||
static constexpr ck_tile::index_t NumDTensor = DsDataType::size();
|
||||
|
||||
std::size_t flop = 0, num_btype = 0;
|
||||
|
||||
flop += std::size_t(2) * M * N * K;
|
||||
|
||||
ck_tile::static_for<0, NumDTensor, 1>{}([&](auto i) {
|
||||
num_btype += sizeof(ck_tile::remove_cvref_t<std::tuple_element_t<i, DsDataType>>) * M * N;
|
||||
flop += sizeof(ck_tile::remove_cvref_t<std::tuple_element_t<i, DsDataType>>) * M * N;
|
||||
});
|
||||
|
||||
num_btype += sizeof(ADataType) * M * K + sizeof(BDataType) * K * N + sizeof(EDataType) * M * N;
|
||||
|
||||
float tflops = static_cast<float>(flop) / 1.E9 / ave_time;
|
||||
float gb_per_sec = num_btype / 1.E6 / ave_time;
|
||||
|
||||
std::cout << "Run Gemm Multiple-D kernel with:\n";
|
||||
std::cout << "M =" << M << " N =" << N << " K =" << K << "\n";
|
||||
std::cout << "StrideA = " << StrideA << " StrideB = " << StrideB << " StrideE = " << StrideE
|
||||
<< "\n";
|
||||
std::cout << "Perf: " << ave_time << " ms, " << tflops << " TFlops, " << gb_per_sec << " GB/s, "
|
||||
<< "\n";
|
||||
|
||||
e_m_n_dev_buf.FromDevice(e_m_n_device_result.data());
|
||||
|
||||
@@ -217,6 +218,24 @@ int run_multiple_d_gemm_example_with_layouts(int argc,
|
||||
<< std::endl;
|
||||
std::cout << "The CPU veification result is: " << (pass ? "correct" : "fail") << std::endl;
|
||||
}
|
||||
|
||||
if(arg_parser.get_int("json") == 1)
|
||||
{
|
||||
dump_gemm_multi_d_fp16_json_results(arg_parser.get_str("jsonfile"),
|
||||
op_name,
|
||||
M,
|
||||
N,
|
||||
K,
|
||||
StrideA,
|
||||
StrideB,
|
||||
StrideD0,
|
||||
StrideD1,
|
||||
StrideE,
|
||||
pass,
|
||||
ave_time,
|
||||
tflops,
|
||||
gb_per_sec);
|
||||
}
|
||||
return pass;
|
||||
}
|
||||
|
||||
|
||||
Reference in New Issue
Block a user