ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-05-14 18:17:44 +00:00
Code Issues Packages Projects Releases Wiki Activity

fix a bug with non-dword-aligned offset when OOB, in case crash (#616)

Browse Source 
Co-authored-by: zjing14 <zhangjing14@gmail.com>

[ROCm/composable_kernel commit: 76fcdc60e9]
This commit is contained in:
carlushuang
2023-03-09 22:07:24 +08:00
committed by GitHub
parent 9ce65cae0e
commit ca7b3a4f58
1 changed files with 4 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
include/ck/utility/amd_buffer_addressing.hpp
View File

@@ -1030,7 +1030,7 @@ amd_buffer_load_invalid_element_return_zero(const T* p_src_wave,
constexpr index_t vector_size = scalar_type<vector_t>::vector_size;
#if CK_EXPERIMENTAL_USE_BUFFER_LOAD_OOB_CHECK_OFFSET_TRICK
uint32_t src_addr_shift = src_thread_element_valid ? 0 : 0x7fffffff;
uint32_t src_addr_shift = src_thread_element_valid ? 0 : 0x80000000;
return amd_buffer_load_impl<scalar_t, vector_size>(
src_wave_buffer_resource, src_addr_shift + src_thread_addr_offset, 0);
@@ -1091,7 +1091,7 @@ __device__ void amd_buffer_store(const typename vector_type_maker<T, N>::type::t
constexpr index_t vector_size = scalar_type<vector_t>::vector_size;
#if CK_EXPERIMENTAL_USE_BUFFER_STORE_OOB_CHECK_OFFSET_TRICK
uint32_t dst_addr_shift = dst_thread_element_valid ? 0 : 0x7fffffff;
uint32_t dst_addr_shift = dst_thread_element_valid ? 0 : 0x80000000;
amd_buffer_store_impl<scalar_t, vector_size>(
src_thread_data, dst_wave_buffer_resource, dst_addr_shift + dst_thread_addr_offset, 0);
@@ -1126,7 +1126,7 @@ amd_buffer_atomic_add(const typename vector_type_maker<T, N>::type::type src_thr
constexpr index_t vector_size = scalar_type<vector_t>::vector_size;
#if CK_EXPERIMENTAL_USE_BUFFER_ATOMIC_ADD_OOB_CHECK_OFFSET_TRICK
uint32_t dst_addr_shift = dst_thread_element_valid ? 0 : 0x7fffffff;
uint32_t dst_addr_shift = dst_thread_element_valid ? 0 : 0x80000000;
amd_buffer_atomic_add_impl<scalar_t, vector_size>(
src_thread_data, dst_wave_buffer_resource, dst_addr_shift + dst_thread_addr_offset, 0);
@@ -1161,7 +1161,7 @@ amd_buffer_atomic_max(const typename vector_type_maker<T, N>::type::type src_thr
constexpr index_t vector_size = scalar_type<vector_t>::vector_size;
#if CK_EXPERIMENTAL_USE_BUFFER_ATOMIC_MAX_OOB_CHECK_OFFSET_TRICK
uint32_t dst_addr_shift = dst_thread_element_valid ? 0 : 0x7fffffff;
uint32_t dst_addr_shift = dst_thread_element_valid ? 0 : 0x80000000;
amd_buffer_atomic_max_impl<scalar_t, vector_size>(
src_thread_data, dst_wave_buffer_resource, dst_addr_shift + dst_thread_addr_offset, 0);