AI/rocm-systems
2
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

GFX12 PC Sampling support (#186)

Browse Source 
The GFX12 host-trap PC sampling support in SDK and V3.
Introducing parser tests specific to GFX12.

Co-authored-by: vlaindic_amdeng <vladimir.indic@amd.com>
This commit is contained in:
systems-assistant[bot]
2025-09-22 13:17:00 +02:00
committed by GitHub
parent 997b36f5bc
commit 63a723a287
21 changed files with 1325 additions and 350 deletions
Download Patch File Download Diff File Expand all files Collapse all files
projects/rocprofiler-sdk/tests/bin/pc-sampling/exec-mask-manipulation/exec_mask_manipulation.cpp
+10 -4
View File
@@ -497,19 +497,25 @@ kernel3(const float c)
void
run_kernel()
{
for(int i = 1; i <= 64; i++)
int wave_size = 0;
HIP_API_CALL(hipDeviceGetAttribute(&wave_size, hipDeviceAttributeWarpSize, 0));
// Get device properties to retrieve GFXIP version
uint32_t num_blocks = BLOCK_SIZE;
for(int i = 1; i <= wave_size; i++)
{
if(i % 2 == 1)
kernel1<<<BLOCK_SIZE, i>>>(i);
kernel1<<<num_blocks, i>>>(i);
else
kernel2<<<BLOCK_SIZE, i>>>(i);
kernel2<<<num_blocks, i>>>(i);
check_hip_error();
HIP_API_CALL(hipDeviceSynchronize());
}
float arg = 0;
kernel3<<<BLOCK_SIZE, 4 * 64>>>(arg);
kernel3<<<num_blocks, 4 * wave_size>>>(arg);
check_hip_error();
HIP_API_CALL(hipDeviceSynchronize());
}