ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-06-06 07:51:52 +00:00
Code Issues Packages Projects Releases Wiki Activity

Replace ENV variable with CMake option for toggling hipRTC in codegen

Browse Source 
tests.
This commit is contained in:
Mirza Halilcevic
2024-10-08 17:22:01 +00:00
parent 47486abb67
commit d3a96e5111
11 changed files with 65 additions and 68 deletions
Download Patch File Download Diff File Expand all files Collapse all files

38
codegen/test/gemm_multiple_d.cpp
View File

@@ -1,33 +1,29 @@
#include "common.hpp"
#include "ck/host/device_gemm_multiple_d/problem.hpp"
#include "ck/host/device_gemm_multiple_d/operation.hpp"
#include "ck/host/headers.hpp"
#include "ck/host/stringutils.hpp"
#include "ck/host/utils.hpp"
#include <algorithm>
#include <cmath>
#include <iterator>
#include <random>
#include <test.hpp>
#include "common.hpp"
#include <rtc/compile_kernel.hpp>
#include <rtc/hip.hpp>
#include <test.hpp>
#include <algorithm>
#include <cmath>
#include <fstream>
#include <iterator>
#include <random>
using half = _Float16;
// using half = __fp16;
const std::string gemm_compile_check = R"__ck__(
#include <${include}>
extern "C" __global__ void f(const ck::half_t* a, const ck::half_t* b, ck::half_t* c) {
using G = ${template};
constexpr auto desc =
G::make_descriptor(ck::make_naive_tensor_descriptor_packed(ck::make_tuple(${m},
${k})),
ck::make_naive_tensor_descriptor(ck::make_tuple(${n},
${k}), ck::make_tuple(1, ${n})), ck::make_tuple(),
ck::make_naive_tensor_descriptor_packed(ck::make_tuple(${m},
${n})));
constexpr auto desc = G::make_descriptor(ck::make_naive_tensor_descriptor_packed(ck::make_tuple(${m}, ${k})),
ck::make_naive_tensor_descriptor(ck::make_tuple(${n}, ${k}), ck::make_tuple(1, ${n})),
ck::make_tuple(),
ck::make_naive_tensor_descriptor_packed(ck::make_tuple(${m}, ${n})));
static_assert(desc.IsValid(), "Invalid ck gemm.");
@@ -69,15 +65,15 @@ TEST_CASE(test_problem_kernel)
{"m", std::to_string(prob.M)},
{"n", std::to_string(prob.N)},
{"k", std::to_string(prob.K)}});
auto srcs = get_headers_for_test();
auto srcs = get_headers_for_test();
srcs.push_back({"main.cpp", src});
rtc::compile_options options;
options.kernel_name = "f";
auto k = rtc::compile_kernel(srcs, options);
auto block_size = solution.GetTemplateParameter<std::size_t>("BlockSize");
auto m_per_block = solution.GetTemplateParameter<std::size_t>("MPerBlock");
auto n_per_block = solution.GetTemplateParameter<std::size_t>("NPerBlock");
auto grid_size = ck::host::integer_divide_ceil(prob.M, m_per_block) *
options.kernel_name = "f";
auto k = rtc::compile_kernel(srcs, options);
auto block_size = solution.GetTemplateParameter<std::size_t>("BlockSize");
auto m_per_block = solution.GetTemplateParameter<std::size_t>("MPerBlock");
auto n_per_block = solution.GetTemplateParameter<std::size_t>("NPerBlock");
auto grid_size = ck::host::integer_divide_ceil(prob.M, m_per_block) *
ck::host::integer_divide_ceil(prob.N, n_per_block);
k.launch(nullptr, grid_size * block_size, block_size)(a.data(), b.data(), c.data());