Skip to content
Open
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
16 changes: 16 additions & 0 deletions devops/scripts/benchmarks/benches/compute/compute.py
Original file line number Diff line number Diff line change
Expand Up @@ -558,6 +558,22 @@ def createTorchLinearKernelSizeBench(variant_name: str, **kwargs):
)
)

# Add TorchEventRecordQuery benchmarks
for runtime in filter(lambda x: x != RUNTIMES.UR, RUNTIMES):
for profiler_type in list(PROFILERS):
benches.append(
TorchEventRecordQuery(
self,
runtime,
"medium",
profiler_type,
Profiling=0,
KernelWGCount=256,
KernelWGSize=512,
EventQueryIterations=1000,
)
)

#
# Note: Graph benchmarks segfault on pvc on L0
#
Expand Down
19 changes: 19 additions & 0 deletions devops/scripts/benchmarks/benches/compute/compute_torch.py
Original file line number Diff line number Diff line change
Expand Up @@ -251,6 +251,25 @@ def __init__(
)


class TorchEventRecordQuery(TorchBenchmark):
def __init__(
self,
suite,
runtime: RUNTIMES,
variant_name: str,
profiler_type: PROFILERS,
**kwargs,
):
super().__init__(
suite,
runtime,
"KernelSubmitEventRecordQuery",
variant_name,
profiler_type,
**kwargs,
)


class TorchGraphVllmMock(TorchBenchmark):
def __init__(
self,
Expand Down
10 changes: 10 additions & 0 deletions devops/scripts/benchmarks/tests/test_integration.py
Original file line number Diff line number Diff line change
Expand Up @@ -231,6 +231,11 @@ def test_torch_l0(self):
# "KernelSubmitEventRecordWait medium",
# {"pytorch", "L0"},
# )
self._checkCase(
"torch_benchmark_l0 KernelSubmitEventRecordQuery EventQueryIterations 1000, KernelWGCount 256, KernelWGSize 512, Profiling 0 CPU count",
"KernelSubmitEventRecordQuery medium, CPU count",
{"pytorch", "L0"},
)
self._checkCase(
"torch_benchmark_l0 KernelSubmitSingleQueue KernelDataType Int32, KernelWGCount 4096, KernelWGSize 512",
"KernelSubmitSingleQueue Int32Large",
Expand Down Expand Up @@ -292,6 +297,11 @@ def test_torch_sycl(self):
"KernelSubmitEventRecordWait medium, CPU count",
{"pytorch", "SYCL"},
)
self._checkCase(
"torch_benchmark_sycl KernelSubmitEventRecordQuery EventQueryIterations 1000, KernelWGCount 256, KernelWGSize 512, Profiling 0",
"KernelSubmitEventRecordQuery medium",
{"pytorch", "SYCL"},
)
self._checkCase(
"torch_benchmark_sycl KernelSubmitSingleQueue KernelDataType Mixed, KernelWGCount 512, KernelWGSize 256",
"KernelSubmitSingleQueue MixedMedium",
Expand Down
Loading