[Bench] Bump Compute Benchmarks version (intel#21497)

to include new UR CMake search and torch benchmark refactor.

Author: lukaszstolarczuk

devops/scripts/benchmarks/benches/compute/compute.py

@@ -40,8 +40,8 @@ def git_url(self) -> str:
         return "https://github.com/intel/compute-benchmarks.git"
 
     def git_hash(self) -> str:
-        # Mar 11, 2026
-        return "5daa36aef97c1a57ca94aed2b43ed5d6829bcc11"
+        # Mar 13, 2026
+        return "0237dade20798127c6432635b47c6bdb2f9d59b0"
 
     def setup(self) -> None:
         if options.sycl is None:
@@ -581,8 +581,8 @@ def createTorchGraphMultiQueueBench(variant_name: str, **kwargs):
                             variant_name,
                             profiler_type,
                             fixed_args={
-                                "workgroupCount": 512,
-                                "workgroupSize": 256,
+                                "KernelWGCount": 512,
+                                "KernelWGSize": 256,
                                 "Profiling": 0,
                                 "UseEvents": 0,
                             },
@@ -592,15 +592,15 @@ def createTorchGraphMultiQueueBench(variant_name: str, **kwargs):
                     benches += [
                         createTorchGraphMultiQueueBench(
                             "small",
-                            kernelsPerQueue=10,
+                            KernelsPerQueue=10,
                         ),
                         createTorchGraphMultiQueueBench(
                             "medium",
-                            kernelsPerQueue=32,
+                            KernelsPerQueue=32,
                         ),
                         createTorchGraphMultiQueueBench(
                             "large",
-                            kernelsPerQueue=64,
+                            KernelsPerQueue=64,
                         ),
                     ]
 

devops/scripts/benchmarks/tests/test_integration.py

@@ -285,7 +285,7 @@ def test_torch_l0(self):
                 {"pytorch", "L0"},
             )
             self._checkCase(
-                "torch_benchmark_l0 KernelSubmitGraphMultiQueue kernelsPerQueue 64 CPU count",
+                "torch_benchmark_l0 KernelSubmitGraphMultiQueue KernelsPerQueue 64 CPU count",
                 "KernelSubmitGraphMultiQueue large, CPU count",
                 {"pytorch", "L0"},
             )
@@ -334,7 +334,7 @@ def test_torch_sycl(self):
                 {"pytorch", "SYCL"},
             )
             self._checkCase(
-                "torch_benchmark_sycl KernelSubmitGraphMultiQueue kernelsPerQueue 32 CPU count",
+                "torch_benchmark_sycl KernelSubmitGraphMultiQueue KernelsPerQueue 32 CPU count",
                 "KernelSubmitGraphMultiQueue medium, CPU count",
                 {"pytorch", "SYCL"},
             )
@@ -388,7 +388,7 @@ def test_torch_syclpreview(self):
                 {"pytorch", "SYCL"},
             )
             self._checkCase(
-                "torch_benchmark_syclpreview KernelSubmitGraphMultiQueue kernelsPerQueue 10",
+                "torch_benchmark_syclpreview KernelSubmitGraphMultiQueue KernelsPerQueue 10",
                 "KernelSubmitGraphMultiQueue small",
                 {"pytorch", "SYCL"},
             )