leavelet/cutlass
1
0
Fork 0
Code Issues Pull Requests Actions 24 Packages Projects Releases Wiki Activity

CUTLASS 3.5.1 (#1623)

Browse Source 
* CUTLASS 3.5.1

* updates, optimizations, fixes
This commit is contained in:
Vijay Thakkar
2024-07-29 08:46:24 -04:00
committed by GitHub
parent 56b46e2d13
commit be60a0b272
312 changed files with 19793 additions and 6775 deletions
Download Patch File Download Diff File Expand all files Collapse all files
tools/profiler/src/operation_profiler.cu
+24 -6
View File
@@ -345,6 +345,7 @@ int OperationProfiler::profile_all(
// For each operation in manifest
int matched_operation_count = 0;
int profiled_operation_count = 0;
for (auto const& operation_ptr : manifest) {
library::Operation const *operation = operation_ptr.get();
@@ -434,7 +435,7 @@ int OperationProfiler::profile_all(
// If there was an internal error, consume the CUDA error and move to the next operation.
(void)cudaGetLastError();
report.append_results(results_);
report.append_result(model_result_);
continue;
}
else if (status != Status::kSuccess) {
@@ -522,25 +523,42 @@ int OperationProfiler::profile_all(
operation,
problem_space,
problem);
// Count op as profiled, even it failed to profile
profiled_operation_count++;
}
report.append_results(results_);
results_.clear();
}
} // if op satisfied compute capacity
if (!continue_profiling) {
// break out of `for op in manifest` loop and move to next problem
// `for each problem in problem space` conditional check on not continue profiling
break;
}
}
} // for op in manifest
// If we did not find any kernels that match our filters and error_on_no_match was set, report an error
if (options.profiling.error_on_no_match && matched_operation_count <= 0) {
#if !NDEBUG
std::cout << "Error: No matching kernels found with kernel selection filters [--error_on_no_match]" << std::endl;
std::cerr << "Error: No matching kernels found with kernel selection filters [--error_on_no_match]" << std::endl;
#endif
retval = 1;
retval |= 1;
// Stop profiling on error no match
continue_profiling = false;
}
}
if (options.profiling.error_if_nothing_is_profiled && options.profiling.enabled && profiled_operation_count <= 0) {
#if !NDEBUG
std::cerr << "Error: No kernels profiled found with kernel selection filters [--error_if_nothing_is_profiled]" << std::endl;
#endif
retval |= 1;
// Stop profiling on error no match
continue_profiling = false;
}
} // for each problem in problem space
return retval;
}