Skip to content

[Bench] Add SYCL MemcpyExecute MT benchmark #18604

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 1 commit into from
May 23, 2025
Merged

[Bench] Add SYCL MemcpyExecute MT benchmark #18604

Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
47 changes: 37 additions & 10 deletions devops/scripts/benchmarks/benches/compute.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -49,7 +49,7 @@ def git_url(self) -> str:
return "https://github.com/intel/compute-benchmarks.git"

def git_hash(self) -> str:
return "49a8c6314875c57fee9b59aea16e721572e3021d"
return "3283b5edb8bf771c519625af741b5db7a37b0111"

def setup(self):
if options.sycl is None:
Expand Down Expand Up @@ -196,11 +196,12 @@ def benchmarks(self) -> list[Benchmark]:
# Add UR-specific benchmarks
if options.ur is not None:
benches += [
MemcpyExecute(self, 400, 1, 102400, 10, 1, 1, 1, 1),
MemcpyExecute(self, 400, 1, 102400, 10, 0, 1, 1, 1),
MemcpyExecute(self, 100, 4, 102400, 10, 1, 1, 0, 1),
MemcpyExecute(self, 100, 4, 102400, 10, 1, 1, 0, 0),
MemcpyExecute(self, 4096, 4, 1024, 10, 0, 1, 0, 1),
MemcpyExecute(self, RUNTIMES.UR, 400, 1, 102400, 10, 1, 1, 1, 1, 0),
MemcpyExecute(self, RUNTIMES.UR, 400, 1, 102400, 10, 0, 1, 1, 1, 0),
MemcpyExecute(self, RUNTIMES.UR, 100, 4, 102400, 10, 1, 1, 0, 1, 0),
MemcpyExecute(self, RUNTIMES.UR, 100, 4, 102400, 10, 1, 1, 0, 0, 0),
MemcpyExecute(self, RUNTIMES.UR, 4096, 4, 1024, 10, 0, 1, 0, 1, 0),
MemcpyExecute(self, RUNTIMES.UR, 4096, 4, 1024, 10, 0, 1, 0, 1, 1),
UsmMemoryAllocation(self, RUNTIMES.UR, "Device", 256, "Both"),
UsmMemoryAllocation(self, RUNTIMES.UR, "Device", 256 * 1024, "Both"),
UsmBatchMemoryAllocation(self, RUNTIMES.UR, "Device", 128, 256, "Both"),
Expand All @@ -211,6 +212,20 @@ def benchmarks(self) -> list[Benchmark]:
self, RUNTIMES.UR, "Device", 128, 128 * 1024, "Both"
),
]
benches += [
MemcpyExecute(
self, RUNTIMES.SYCL_PREVIEW, 4096, 1, 1024, 40, 1, 1, 0, 1, 0
),
MemcpyExecute(
self, RUNTIMES.SYCL_PREVIEW, 4096, 1, 1024, 40, 1, 1, 0, 1, 1
),
MemcpyExecute(
self, RUNTIMES.SYCL_PREVIEW, 4096, 4, 1024, 10, 1, 1, 0, 1, 0
),
MemcpyExecute(
self, RUNTIMES.SYCL_PREVIEW, 4096, 4, 1024, 10, 1, 1, 0, 1, 1
),
]

return benches

Expand Down Expand Up @@ -533,6 +548,7 @@ class MemcpyExecute(ComputeBenchmark):
def __init__(
self,
bench,
runtime: RUNTIMES,
numOpsPerThread,
numThreads,
allocSize,
Expand All @@ -541,7 +557,9 @@ def __init__(
dstUSM,
useEvent,
useCopyOffload,
useBarrier,
):
self.runtime = runtime
self.numOpsPerThread = numOpsPerThread
self.numThreads = numThreads
self.allocSize = allocSize
Expand All @@ -550,7 +568,10 @@ def __init__(
self.dstUSM = dstUSM
self.useEvents = useEvent
self.useCopyOffload = useCopyOffload
super().__init__(bench, "multithread_benchmark_ur", "MemcpyExecute")
self.useBarrier = useBarrier
super().__init__(
bench, f"multithread_benchmark_{self.runtime.value}", "MemcpyExecute"
)

def extra_env_vars(self) -> dict:
if not self.useCopyOffload:
Expand All @@ -560,9 +581,10 @@ def extra_env_vars(self) -> dict:

def name(self):
return (
f"multithread_benchmark_ur MemcpyExecute opsPerThread:{self.numOpsPerThread}, numThreads:{self.numThreads}, allocSize:{self.allocSize} srcUSM:{self.srcUSM} dstUSM:{self.dstUSM}"
f"multithread_benchmark_{self.runtime.value} MemcpyExecute opsPerThread:{self.numOpsPerThread}, numThreads:{self.numThreads}, allocSize:{self.allocSize} srcUSM:{self.srcUSM} dstUSM:{self.dstUSM}"
+ (" without events" if not self.useEvents else "")
+ (" without copy offload" if not self.useCopyOffload else "")
+ (" with barrier" if self.useBarrier else "")
)

def explicit_group(self):
Expand All @@ -571,21 +593,25 @@ def explicit_group(self):
+ str(self.numOpsPerThread)
+ " numThreads: "
+ str(self.numThreads)
+ " allocSize: "
+ str(self.allocSize)
)

def description(self) -> str:
src_type = "device" if self.srcUSM == 1 else "host"
dst_type = "device" if self.dstUSM == 1 else "host"
events = "with" if self.useEvents else "without"
copy_offload = "with" if self.useCopyOffload else "without"
with_barrier = "with" if self.useBarrier else "without"
return (
f"Measures multithreaded memory copy performance with {self.numThreads} threads "
f"each performing {self.numOpsPerThread} operations on {self.allocSize} bytes "
f"from {src_type} to {dst_type} memory {events} events {copy_offload} driver copy offload."
f"from {src_type} to {dst_type} memory {events} events {copy_offload} driver copy offload "
f"{with_barrier} barrier. "
)

def get_tags(self):
return ["memory", "latency", "UR", "micro"]
return ["memory", "latency", runtime_to_tag_name(self.runtime), "micro"]

def bin_args(self) -> list[str]:
return [
Expand All @@ -599,6 +625,7 @@ def bin_args(self) -> list[str]:
f"--iterations={self.iterations}",
f"--SrcUSM={self.srcUSM}",
f"--DstUSM={self.dstUSM}",
f"--UseBarrier={self.useBarrier}",
]


Expand Down
Loading