[SYCL] Align Non-Uniform Groups with PR#14604 (#19238)

This PR implements the non-uniform groups API redesign proposed in [PR
14604](#14604)

1. Removed redundant `_group` suffix from class names
2. Used more descriptive function names that clearly indicate the
partitioning behaviour
3. United similar functionality to reduce cognitive overhead

  * fixed_size_group → chunk
  * tangle_group → tangle
  * ballot_group → fragment
  * opportunistic_group → merged into fragment
  * get_ballot_group() → binary_partition()
  * get_fixed_size_group<N>() → chunked_partition<N>()
  * get_tangle_group() → entangle()

---------

Signed-off-by: Larsen, Steffen <[email protected]>
Co-authored-by: Andrei Zibrov <[email protected]>
Co-authored-by: John Pennycook <[email protected]>
Co-authored-by: Udit Kumar Agarwal <[email protected]>

Author: 4 people

llvm/include/llvm/SYCLLowerIR/DeviceConfigFile.td

@@ -70,10 +70,9 @@ def AspectExt_oneapi_unique_addressing_per_dim : Aspect<"ext_oneapi_unique_addre
 def AspectExt_oneapi_bindless_images_sample_1d_usm : Aspect<"ext_oneapi_bindless_images_sample_1d_usm">;
 def AspectExt_oneapi_bindless_images_sample_2d_usm : Aspect<"ext_oneapi_bindless_images_sample_2d_usm">;
 def AspectExt_intel_esimd : Aspect<"ext_intel_esimd">;
-def AspectExt_oneapi_ballot_group : Aspect<"ext_oneapi_ballot_group">;
-def AspectExt_oneapi_fixed_size_group : Aspect<"ext_oneapi_fixed_size_group">;
-def AspectExt_oneapi_opportunistic_group : Aspect<"ext_oneapi_opportunistic_group">;
-def AspectExt_oneapi_tangle_group : Aspect<"ext_oneapi_tangle_group">;
+def AspectExt_oneapi_fragment : Aspect<"ext_oneapi_fragment">;
+def AspectExt_oneapi_chunk : Aspect<"ext_oneapi_chunk">;
+def AspectExt_oneapi_tangle : Aspect<"ext_oneapi_tangle">;
 def AspectExt_intel_matrix : Aspect<"ext_intel_matrix">;
 def AspectExt_oneapi_is_composite : Aspect<"ext_oneapi_is_composite">;
 def AspectExt_oneapi_is_component : Aspect<"ext_oneapi_is_component">;
@@ -155,8 +154,8 @@ def : TargetInfo<"__TestAspectList",
     AspectExt_oneapi_bindless_sampled_image_fetch_3d,
     AspectExt_oneapi_bindless_images_gather,
     AspectExt_intel_esimd,
-    AspectExt_oneapi_ballot_group, AspectExt_oneapi_fixed_size_group, AspectExt_oneapi_opportunistic_group,
-    AspectExt_oneapi_tangle_group, AspectExt_intel_matrix, AspectExt_oneapi_is_composite, AspectExt_oneapi_is_component,
+    AspectExt_oneapi_fragment, AspectExt_oneapi_chunk,
+    AspectExt_oneapi_tangle, AspectExt_intel_matrix, AspectExt_oneapi_is_composite, AspectExt_oneapi_is_component,
     AspectExt_oneapi_graph, AspectExt_oneapi_limited_graph, AspectExt_oneapi_private_alloca, 
     AspectExt_oneapi_queue_profiling_tag, AspectExt_oneapi_virtual_mem, AspectExt_oneapi_cuda_cluster_group, 
     AspectExt_intel_fpga_task_sequence,
@@ -181,9 +180,9 @@ def : TargetInfo<"__TestDeprecatedAspectList",
 defvar IntelCpuAspects = [
     AspectCpu, AspectFp16, AspectFp64, AspectQueue_profiling, AspectAtomic64,
     AspectExt_oneapi_srgb, AspectExt_oneapi_native_assert,
-    AspectExt_intel_legacy_image, AspectExt_oneapi_ballot_group,
-    AspectExt_oneapi_fixed_size_group, AspectExt_oneapi_opportunistic_group,
-    AspectExt_oneapi_tangle_group, AspectExt_oneapi_private_alloca, 
+    AspectExt_intel_legacy_image, AspectExt_oneapi_fragment,
+    AspectExt_oneapi_chunk,
+    AspectExt_oneapi_tangle, AspectExt_oneapi_private_alloca, 
     AspectOnline_compiler, AspectOnline_linker, AspectExt_intel_gpu_slices,
     AspectExt_intel_gpu_subslices_per_slice, AspectExt_intel_gpu_eu_count_per_subslice,
     AspectExt_intel_gpu_hw_threads_per_eu, AspectExt_intel_device_id, 
@@ -207,9 +206,9 @@ defvar Fp16Fp64Atomic64 = [AspectFp16, AspectFp64, AspectAtomic64];
 defvar Fp16Atomic64 = [AspectFp16, AspectAtomic64];
 defvar Sg8_16_32 = [8, 16, 32];
 defvar Sg16_32 = [16, 32];
-defvar IntelGPUBaseAspects = [AspectExt_intel_esimd, AspectExt_oneapi_ballot_group,
-    AspectExt_oneapi_fixed_size_group, AspectExt_oneapi_opportunistic_group,
-    AspectExt_oneapi_tangle_group];
+defvar IntelGPUBaseAspects = [AspectExt_intel_esimd, AspectExt_oneapi_fragment,
+    AspectExt_oneapi_chunk,
+    AspectExt_oneapi_tangle];
 class IntelTargetInfo<string Name, list<Aspect> Aspects, list<int> subGroupSizesList>
 : TargetInfo<Name, IntelGPUBaseAspects # Aspects, subGroupSizesList>;
 // Note: only the "canonical" target names are listed here - see
@@ -283,8 +282,8 @@ defvar CudaSM90USMAspects = [AspectUsm_system_allocations, AspectUsm_atomic_host
 defvar CudaMinAspects = !listconcat(CudaMinUSMAspects, [AspectGpu, AspectFp64, AspectOnline_compiler, AspectOnline_linker,
     AspectQueue_profiling, AspectExt_intel_pci_address, AspectExt_intel_memory_bus_width,
     AspectExt_intel_device_info_uuid, AspectExt_oneapi_native_assert, AspectExt_intel_free_memory, AspectExt_intel_device_id,
-    AspectExt_intel_memory_clock_rate, AspectExt_oneapi_ballot_group, AspectExt_oneapi_fixed_size_group,
-    AspectExt_oneapi_opportunistic_group, AspectExt_oneapi_graph, AspectExt_oneapi_limited_graph]);
+    AspectExt_intel_memory_clock_rate, AspectExt_oneapi_fragment, AspectExt_oneapi_chunk,
+    AspectExt_oneapi_graph, AspectExt_oneapi_limited_graph]);
 // Bindless images aspects are partially supported on CUDA and disabled by default at the moment.
 defvar CudaBindlessImagesAspects = [AspectExt_oneapi_bindless_images, AspectExt_oneapi_bindless_images_shared_usm,
     AspectExt_oneapi_bindless_images_1d_usm, AspectExt_oneapi_bindless_images_2d_usm, AspectExt_oneapi_external_memory_import,