ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-04-19 22:39:03 +00:00
Code Issues Packages Projects Releases Wiki Activity

[CK_BUILDER] Remove cmath include (#3508)

Browse Source 
Remove the dependency from device_tensor_generator.hpp and fix a typo from a previous force push. The changes replace standard library math functions with their ck::math equivalents and define PI as a local constant instead of computing it using std::acos.

Key changes:

* Removed #include header dependency
* Replaced std::acos(-1.0) with hardcoded PI constant 3.141592653f
* Replaced std::sqrt, std::cos, and std::sin with ck::math equivalents
This commit is contained in:
John Shumway
2026-01-02 16:58:35 -08:00
committed by GitHub
parent 355ce9230d
commit 4670df5ca6
1 changed files with 6 additions and 8 deletions
Download Patch File Download Diff File Expand all files Collapse all files

14
include/ck/library/utility/device_tensor_generator.hpp
View File

@@ -7,7 +7,6 @@
#include "ck/utility/common_header.hpp"
#include "ck/library/utility/device_tensor_generator.hpp"
#include "ck/utility/data_type.hpp"
#include <cmath>
// use xorshift for now since it is simple. Should be suitable enough, but feel free to switch in
// the future
@@ -107,7 +106,7 @@ template <typename T>
__global__ void
fill_tensor_norm_rand_fp_values(T* p, float sigma, float mean, uint64_t buffer_element_size)
{
static constexpr PI = std::acos(-1.0);
static constexpr float PI = 3.141592653f;
// initial values
ran_state_u32 s = ran_init();
float norm[2];
@@ -116,12 +115,11 @@ fill_tensor_norm_rand_fp_values(T* p, float sigma, float mean, uint64_t buffer_e
{
if(j % (2 / ck::packed_size_v<T>) == 0)
{
float u1 = ran_gen_round_u32(s) * (1.0f / 4294967296.0f);
float u2 = ran_gen_round_u32(s) * (1.0f / 4294967296.0f);
norm[0] =
sigma * std::sqrt(-2.0f * ck::math::log(u1)) * std::cos(2.0f * PI * u2) + mean;
norm[1] =
sigma * std::sqrt(-2.0f * ck::math::log(u1)) * std::sin(2.0f * PI * u2) + mean;
float u1 = ran_gen_round_u32(s) * (1.0f / 4294967296.0f);
float u2 = ran_gen_round_u32(s) * (1.0f / 4294967296.0f);
float scale = sigma * ck::math::sqrt(-2.0f * ck::math::log(u1));
norm[0] = scale * ck::math::cos(2.0f * PI * u2) + mean;
norm[1] = scale * ck::math::sin(2.0f * PI * u2) + mean;
}
if constexpr(ck::is_same_v<T, ck::f4x2_pk_t>)