@@ -262,7 +262,10 @@ class CudaTargetInfo<string targetName, list<Aspect> aspectList, int subGroupSiz
262262 assert !eq(subGroupSize, 32), "sub-group size for Cuda must be equal to 32 and not " # subGroupSize # ".";
263263}
264264
265- defvar CudaMinAspects = !listconcat(AllUSMAspects, [AspectGpu, AspectFp64, AspectOnline_compiler, AspectOnline_linker,
265+ defvar CudaMinUSMAspects = [AspectUsm_device_allocations, AspectUsm_host_allocations, AspectUsm_shared_allocations];
266+ defvar CudaSM90USMAspects = [AspectUsm_system_allocations, AspectUsm_atomic_host_allocations, AspectUsm_atomic_shared_allocations];
267+
268+ defvar CudaMinAspects = !listconcat(CudaMinUSMAspects, [AspectGpu, AspectFp64, AspectOnline_compiler, AspectOnline_linker,
266269 AspectQueue_profiling, AspectExt_intel_pci_address, AspectExt_intel_max_mem_bandwidth, AspectExt_intel_memory_bus_width,
267270 AspectExt_intel_device_info_uuid, AspectExt_oneapi_native_assert, AspectExt_intel_free_memory, AspectExt_intel_device_id,
268271 AspectExt_intel_memory_clock_rate, AspectExt_oneapi_ballot_group, AspectExt_oneapi_fixed_size_group,
@@ -292,9 +295,9 @@ def : CudaTargetInfo<"nvidia_gpu_sm_87", !listconcat(CudaMinAspects, CudaBindles
292295 [AspectFp16, AspectAtomic64, AspectExt_oneapi_cuda_async_barrier])>;
293296def : CudaTargetInfo<"nvidia_gpu_sm_89", !listconcat(CudaMinAspects, CudaBindlessImagesAspects,
294297 [AspectFp16, AspectAtomic64, AspectExt_oneapi_cuda_async_barrier])>;
295- def : CudaTargetInfo<"nvidia_gpu_sm_90", !listconcat(CudaMinAspects, CudaBindlessImagesAspects,
298+ def : CudaTargetInfo<"nvidia_gpu_sm_90", !listconcat(CudaMinAspects, CudaSM90USMAspects, CudaBindlessImagesAspects,
296299 [AspectFp16, AspectAtomic64, AspectExt_oneapi_cuda_async_barrier, AspectExt_oneapi_cuda_cluster_group])>;
297- def : CudaTargetInfo<"nvidia_gpu_sm_90a", !listconcat(CudaMinAspects, CudaBindlessImagesAspects,
300+ def : CudaTargetInfo<"nvidia_gpu_sm_90a", !listconcat(CudaMinAspects, CudaSM90USMAspects, CudaBindlessImagesAspects,
298301 [AspectFp16, AspectAtomic64, AspectExt_oneapi_cuda_async_barrier, AspectExt_oneapi_cuda_cluster_group])>;
299302
300303//
0 commit comments