ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-05-20 12:59:49 +00:00
Code Issues Packages Projects Releases Wiki Activity

Enable bf16 RNE on gfx950 (#2542)

Browse Source 
* Enable bf16 RNE for gfx950

* test bhalf

* fix

* fix

* Comments fixes

* fixes

* clean

* fix

[ROCm/composable_kernel commit: 685771b875]
This commit is contained in:
Bartłomiej Kocot
2025-07-28 00:47:17 +02:00
committed by GitHub
parent 457e3c8b08
commit 23ee2ddb5c
2 changed files with 63 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

46
test/data_type/test_bhalf.cpp
View File

@@ -2,8 +2,12 @@
// Copyright (c) 2025, Advanced Micro Devices, Inc. All rights reserved.
#include "gtest/gtest.h"
#include <hip/hip_runtime.h>
#include "ck/utility/data_type.hpp"
#include "ck/utility/type_convert.hpp"
#include "ck/host_utility/hip_check_error.hpp"
using ck::bhalf_t;
using ck::type_convert;
@@ -46,3 +50,45 @@ TEST(BHALF_T, MantisaExpOverflow)
ASSERT_TRUE(std::isnan(float_val));
ASSERT_TRUE(std::isnan(type_convert<float>(type_convert<bhalf_t>(float_val))));
}
__global__ void cast(const float input, float* output)
{
const bhalf_t bhalf_val = type_convert<bhalf_t>(input);
*output = type_convert<float>(bhalf_val);
}
TEST(BHALF_T, CastOnDevice)
{
constexpr int num_vals = 11;
const float abs_tol = std::pow(2, -7);
float float_vals[num_vals] = {0.5, 0.875, 1.5, 1, 2, 4, 8, 16, 32, 64, 128};
float* float_val_after_cast_dev;
float float_val_after_cast_host;
hip_check_error(hipMalloc(&float_val_after_cast_dev, sizeof(float)));
// Positive
for(int idx = 0; idx < num_vals; idx++)
{
cast<<<1, 1>>>(float_vals[idx], float_val_after_cast_dev);
hip_check_error(hipMemcpy(&float_val_after_cast_host,
float_val_after_cast_dev,
sizeof(float),
hipMemcpyDeviceToHost));
ASSERT_NEAR(float_val_after_cast_host, float_vals[idx], abs_tol);
}
// Negative
for(int idx = 0; idx < num_vals; idx++)
{
cast<<<1, 1>>>(-float_vals[idx], float_val_after_cast_dev);
hip_check_error(hipMemcpy(&float_val_after_cast_host,
float_val_after_cast_dev,
sizeof(float),
hipMemcpyDeviceToHost));
ASSERT_NEAR(float_val_after_cast_host, -float_vals[idx], abs_tol);
}
}