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

Update bandwidth and latency calculations

Browse Source 
- Refined bandwidth and latency calculations for improved accuracy
This commit is contained in:
avinashkethineedi
2025-02-17 06:18:46 +00:00
parent 40bd8a38a0
commit c155636da4
26 changed files with 283 additions and 270 deletions
Download Patch File Download Diff File Expand all files Collapse all files
tests/functional_tests/team_reduction_tester.cpp
+9 -8
View File
@@ -73,11 +73,13 @@ rocshmem_team_t team_reduce_world_dup;
* DEVICE TEST KERNEL
*****************************************************************************/
template <typename T1, ROCSHMEM_OP T2>
__global__ void TeamReductionTest(int loop, int skip, uint64_t *timer,
T1 *s_buf, T1 *r_buf, int size, TestType type,
__global__ void TeamReductionTest(int loop, int skip, long long int *start_time,
long long int *end_time, T1 *s_buf, T1 *r_buf,
int size, TestType type,
ShmemContextType ctx_type,
rocshmem_team_t team) {
__shared__ rocshmem_ctx_t ctx;
int wg_id = get_flat_grid_id();
rocshmem_wg_init();
rocshmem_wg_ctx_create(ctx_type, &ctx);
@@ -86,10 +88,9 @@ __global__ void TeamReductionTest(int loop, int skip, uint64_t *timer,
__syncthreads();
uint64_t start;
for (int i = 0; i < loop + skip; i++) {
if (i == skip && hipThreadIdx_x == 0) {
start = rocshmem_timer();
start_time[wg_id] = wall_clock64();
}
wg_team_reduce<T1, T2>(ctx, team, r_buf, s_buf, size);
rocshmem_ctx_wg_barrier_all(ctx);
@@ -98,7 +99,7 @@ __global__ void TeamReductionTest(int loop, int skip, uint64_t *timer,
__syncthreads();
if (hipThreadIdx_x == 0) {
timer[hipBlockIdx_x] = rocshmem_timer() - start;
end_time[wg_id] = wall_clock64();
}
rocshmem_wg_ctx_destroy(&ctx);
@@ -138,9 +139,9 @@ void TeamReductionTester<T1, T2>::launchKernel(dim3 gridSize, dim3 blockSize,
size_t shared_bytes = 0;
hipLaunchKernelGGL(HIP_KERNEL_NAME(TeamReductionTest<T1, T2>), gridSize,
blockSize, shared_bytes, stream, loop, args.skip, timer,
s_buf, r_buf, size, _type, _shmem_context,
team_reduce_world_dup);
blockSize, shared_bytes, stream, loop, args.skip,
start_time, end_time, s_buf, r_buf, size, _type,
_shmem_context, team_reduce_world_dup);
num_msgs = loop + args.skip;
num_timed_msgs = loop;