chengjunlu
commented
1 year ago Disassemble of the SPIRV IR from the DPCPP example:
; SPIR-V
; Version: 1.4
; Generator: Khronos LLVM/SPIR-V Translator; 14
; Bound: 229
; Schema: 0
OpCapability Addresses ; 0x00000014
OpCapability Linkage ; 0x0000001c
OpCapability Kernel ; 0x00000024
OpCapability Vector16 ; 0x0000002c
OpCapability Int64 ; 0x00000034
OpCapability GenericPointer ; 0x0000003c
OpCapability Int8 ; 0x00000044
OpCapability SubgroupDispatch ; 0x0000004c
OpCapability IndirectReferencesINTEL ; 0x00000054
OpCapability VectorComputeINTEL ; 0x0000005c
OpCapability ExpectAssumeKHR ; 0x00000064
OpCapability MemoryAccessAliasingINTEL ; 0x0000006c
OpCapability OptNoneINTEL ; 0x00000074
OpExtension "SPV_INTEL_function_pointers" ; 0x0000007c
OpExtension "SPV_INTEL_memory_access_aliasing" ; 0x0000009c
OpExtension "SPV_INTEL_optnone" ; 0x000000c4
OpExtension "SPV_INTEL_vector_compute" ; 0x000000dc
OpExtension "SPV_KHR_expect_assume" ; 0x000000fc
%1 = OpExtInstImport "OpenCL.std" ; 0x00000118
OpMemoryModel Physical64 OpenCL ; 0x0000012c
OpEntryPoint Kernel %43 "_ZTSZZ4testILb1EEbvENKUlRN4sycl3_V17handlerEE_clES3_EUlNS1_7nd_itemILi1EEEE_" ; 0x00000138
OpExecutionMode %43 ContractionOff ; 0x00000194
OpExecutionMode %43 SubgroupSize 16 ; 0x000001a0
OpSource Unknown 100000 ; 0x000001b0
OpName %__spirv_BuiltInSubgroupId "__spirv_BuiltInSubgroupId" ; 0x000001bc
OpName %__spirv_BuiltInSubgroupLocalInvocationId "__spirv_BuiltInSubgroupLocalInvocationId" ; 0x000001e0
OpName %__spirv_BuiltInWorkgroupId "__spirv_BuiltInWorkgroupId" ; 0x00000214
OpName %__spirv_BuiltInGlobalLinearId "__spirv_BuiltInGlobalLinearId" ; 0x00000238
OpName %__spirv_BuiltInWorkgroupSize "__spirv_BuiltInWorkgroupSize" ; 0x00000260
OpName %_Z33__regcall3____builtin_invoke_simdILb1EfPFNSt12experimental4simdIfNS0_10__simd_abiILNS0_12_StorageKindE2ELi16EEEEERFS5_PfS5_iES6_S5_jEJPS7_S6_fjEvET0_T1_DpT2__2 "_Z33__regcall3____builtin_invoke_simdILb1EfPFNSt12experimental4simdIfNS0_10__simd_abiILNS0_12_StorageKindE2ELi16EEEEERFS5_PfS5_iES6_S5_jEJPS7_S6_fjEvET0_T1_DpT2__2" ; 0x00000288
OpName %_Z33__regcall3____builtin_invoke_simdILb1EfPFNSt12experimental4simdIfNS0_10__simd_abiILNS0_12_StorageKindE2ELi16EEEEERFS5_PfS5_iES6_S5_jEJPS7_S6_fjEvET0_T1_DpT2__4 "_Z33__regcall3____builtin_invoke_simdILb1EfPFNSt12experimental4simdIfNS0_10__simd_abiILNS0_12_StorageKindE2ELi16EEEEERFS5_PfS5_iES6_S5_jEJPS7_S6_fjEvET0_T1_DpT2__4" ; 0x00000334
OpName %_Z33__regcall3____builtin_invoke_simdILb1EvPFvRFvPfNSt12experimental4simdIfNS1_10__simd_abiILNS1_12_StorageKindE2ELi16EEEEEiES0_S6_jEJPS7_S0_fjEvET0_T1_DpT2__6 "_Z33__regcall3____builtin_invoke_simdILb1EvPFvRFvPfNSt12experimental4simdIfNS1_10__simd_abiILNS1_12_StorageKindE2ELi16EEEEEiES0_S6_jEJPS7_S0_fjEvET0_T1_DpT2__6" ; 0x000003e0
OpName %llvm_genx_svm_block_ld_unaligned_v16f32_i64 "llvm.genx.svm.block.ld.unaligned.v16f32.i64" ; 0x00000488
OpName %__itt_offload_wi_start_wrapper "__itt_offload_wi_start_wrapper" ; 0x000004bc
OpName %_ZN4sycl3_V13ext6oneapi12experimental6detail33__regcall3__simd_func_call_helperILi16ERFNSt12experimental4simdIfNS6_10__simd_abiILNS6_12_StorageKindE2ELi16EEEEEPfSB_iEJNS3_7uniformISC_EEfNSF_IjEEEEENSt13invoke_resultIT0_JDpNS4_9spmd2simdIT1_XT_EvE4typeEEE4typeESJ_SO__3 "_ZN4sycl3_V13ext6oneapi12experimental6detail33__regcall3__simd_func_call_helperILi16ERFNSt12experimental4simdIfNS6_10__simd_abiILNS6_12_StorageKindE2ELi16EEEEEPfSB_iEJNS3_7uniformISC_EEfNSF_IjEEEEENSt13invoke_resultIT0_JDpNS4_9spmd2simdIT1_XT_EvE4typeEEE4typeESJ_SO__3" ; 0x000004e4
OpName %_ZN4sycl3_V13ext6oneapi12experimental6detail33__regcall3__simd_func_call_helperILi16ERFNSt12experimental4simdIfNS6_10__simd_abiILNS6_12_StorageKindE2ELi16EEEEEPfSB_iEJNS3_7uniformISC_EEfNSF_IjEEEEENSt13invoke_resultIT0_JDpNS4_9spmd2simdIT1_XT_EvE4typeEEE4typeESJ_SO__1 "_ZN4sycl3_V13ext6oneapi12experimental6detail33__regcall3__simd_func_call_helperILi16ERFNSt12experimental4simdIfNS6_10__simd_abiILNS6_12_StorageKindE2ELi16EEEEEPfSB_iEJNS3_7uniformISC_EEfNSF_IjEEEEENSt13invoke_resultIT0_JDpNS4_9spmd2simdIT1_XT_EvE4typeEEE4typeESJ_SO__1" ; 0x000005fc
OpName %_ZN4sycl3_V13ext6oneapi12experimental6detail33__regcall3__simd_func_call_helperILi16ERFvPfNSt12experimental4simdIfNS7_10__simd_abiILNS7_12_StorageKindE2ELi16EEEEEiEJNS3_7uniformIS6_EEfNSF_IjEEEEENSt13invoke_resultIT0_JDpNS4_9spmd2simdIT1_XT_EvE4typeEEE4typeESJ_SO__5 "_ZN4sycl3_V13ext6oneapi12experimental6detail33__regcall3__simd_func_call_helperILi16ERFvPfNSt12experimental4simdIfNS7_10__simd_abiILNS7_12_StorageKindE2ELi16EEEEEiEJNS3_7uniformIS6_EEfNSF_IjEEEEENSt13invoke_resultIT0_JDpNS4_9spmd2simdIT1_XT_EvE4typeEEE4typeESJ_SO__5" ; 0x00000714
OpName %__itt_offload_wi_finish_wrapper "__itt_offload_wi_finish_wrapper" ; 0x00000828
OpName %__itt_offload_wi_start_stub "__itt_offload_wi_start_stub" ; 0x00000850
OpName %__itt_offload_wi_finish_stub "__itt_offload_wi_finish_stub" ; 0x00000874
OpName %_Z24__regcall3__SIMD_CALLEE1PfNSt12experimental4simdIfNS0_10__simd_abiILNS0_12_StorageKindE2ELi16EEEEEi "_Z24__regcall3__SIMD_CALLEE1PfNSt12experimental4simdIfNS0_10__simd_abiILNS0_12_StorageKindE2ELi16EEEEEi" ; 0x0000089c
OpName %_esimd ".esimd" ; 0x0000090c
OpName %_Z24__regcall3__SIMD_CALLEE2PfNSt12experimental4simdIfNS0_10__simd_abiILNS0_12_StorageKindE2ELi16EEEEEi "_Z24__regcall3__SIMD_CALLEE2PfNSt12experimental4simdIfNS0_10__simd_abiILNS0_12_StorageKindE2ELi16EEEEEi" ; 0x0000091c
OpName %_esimd_0 ".esimd" ; 0x0000098c
OpName %_Z28__regcall3__SIMD_CALLEE_VOIDPfNSt12experimental4simdIfNS0_10__simd_abiILNS0_12_StorageKindE2ELi16EEEEEi "_Z28__regcall3__SIMD_CALLEE_VOIDPfNSt12experimental4simdIfNS0_10__simd_abiILNS0_12_StorageKindE2ELi16EEEEEi" ; 0x0000099c
OpName %_esimd_1 ".esimd" ; 0x00000a10
OpName %_esimd_i ".esimd.i" ; 0x00000a20
OpName %_esimd_i_0 ".esimd.i" ; 0x00000a34
OpName %_esimd_i_1 ".esimd.i" ; 0x00000a48
%52 = OpAliasDomainDeclINTEL ; 0x00000a5c
%53 = OpAliasScopeDeclINTEL %52 ; 0x00000a64
%54 = OpAliasDomainDeclINTEL ; 0x00000a70
%55 = OpAliasScopeDeclINTEL %54 ; 0x00000a78
%56 = OpAliasDomainDeclINTEL ; 0x00000a84
%57 = OpAliasScopeDeclINTEL %56 ; 0x00000a8c
%58 = OpAliasScopeListDeclINTEL %53 %55 %57 ; 0x00000a98
%61 = OpAliasDomainDeclINTEL ; 0x00000aac
%62 = OpAliasScopeDeclINTEL %61 ; 0x00000ab4
%63 = OpAliasScopeListDeclINTEL %62 ; 0x00000ac0
%71 = OpAliasDomainDeclINTEL ; 0x00000acc
%72 = OpAliasScopeDeclINTEL %71 ; 0x00000ad4
%73 = OpAliasScopeListDeclINTEL %72 ; 0x00000ae0
OpDecorate %__spirv_BuiltInSubgroupId LinkageAttributes "__spirv_BuiltInSubgroupId" Import ; 0x00000aec
OpDecorate %__spirv_BuiltInSubgroupId Constant ; 0x00000b18
OpDecorate %__spirv_BuiltInSubgroupId BuiltIn SubgroupId ; 0x00000b24
OpDecorate %__spirv_BuiltInSubgroupId Alignment 4 ; 0x00000b34
OpDecorate %__spirv_BuiltInSubgroupLocalInvocationId LinkageAttributes "__spirv_BuiltInSubgroupLocalInvocationId" Import ; 0x00000b44
OpDecorate %__spirv_BuiltInSubgroupLocalInvocationId Constant ; 0x00000b80
OpDecorate %__spirv_BuiltInSubgroupLocalInvocationId BuiltIn SubgroupLocalInvocationId ; 0x00000b8c
OpDecorate %__spirv_BuiltInSubgroupLocalInvocationId Alignment 4 ; 0x00000b9c
OpDecorate %__spirv_BuiltInWorkgroupId LinkageAttributes "__spirv_BuiltInWorkgroupId" Import ; 0x00000bac
OpDecorate %__spirv_BuiltInWorkgroupId Constant ; 0x00000bd8
OpDecorate %__spirv_BuiltInWorkgroupId BuiltIn WorkgroupId ; 0x00000be4
OpDecorate %__spirv_BuiltInWorkgroupId Alignment 32 ; 0x00000bf4
OpDecorate %__spirv_BuiltInGlobalLinearId LinkageAttributes "__spirv_BuiltInGlobalLinearId" Import ; 0x00000c04
OpDecorate %__spirv_BuiltInGlobalLinearId Constant ; 0x00000c34
OpDecorate %__spirv_BuiltInGlobalLinearId BuiltIn GlobalLinearId ; 0x00000c40
OpDecorate %__spirv_BuiltInGlobalLinearId Alignment 8 ; 0x00000c50
OpDecorate %__spirv_BuiltInWorkgroupSize LinkageAttributes "__spirv_BuiltInWorkgroupSize" Import ; 0x00000c60
OpDecorate %__spirv_BuiltInWorkgroupSize Constant ; 0x00000c90
OpDecorate %__spirv_BuiltInWorkgroupSize BuiltIn WorkgroupSize ; 0x00000c9c
OpDecorate %__spirv_BuiltInWorkgroupSize Alignment 32 ; 0x00000cac
OpDecorate %_Z33__regcall3____builtin_invoke_simdILb1EfPFNSt12experimental4simdIfNS0_10__simd_abiILNS0_12_StorageKindE2ELi16EEEEERFS5_PfS5_iES6_S5_jEJPS7_S6_fjEvET0_T1_DpT2__2 LinkageAttributes "_Z33__regcall3____builtin_invoke_simdILb1EfPFNSt12experimental4simdIfNS0_10__simd_abiILNS0_12_StorageKindE2ELi16EEEEERFS5_PfS5_iES6_S5_jEJPS7_S6_fjEvET0_T1_DpT2__2" Import ; 0x00000cbc
OpDecorate %_Z33__regcall3____builtin_invoke_simdILb1EfPFNSt12experimental4simdIfNS0_10__simd_abiILNS0_12_StorageKindE2ELi16EEEEERFS5_PfS5_iES6_S5_jEJPS7_S6_fjEvET0_T1_DpT2__4 LinkageAttributes "_Z33__regcall3____builtin_invoke_simdILb1EfPFNSt12experimental4simdIfNS0_10__simd_abiILNS0_12_StorageKindE2ELi16EEEEERFS5_PfS5_iES6_S5_jEJPS7_S6_fjEvET0_T1_DpT2__4" Import ; 0x00000d70
OpDecorate %_Z33__regcall3____builtin_invoke_simdILb1EvPFvRFvPfNSt12experimental4simdIfNS1_10__simd_abiILNS1_12_StorageKindE2ELi16EEEEEiES0_S6_jEJPS7_S0_fjEvET0_T1_DpT2__6 LinkageAttributes "_Z33__regcall3____builtin_invoke_simdILb1EvPFvRFvPfNSt12experimental4simdIfNS1_10__simd_abiILNS1_12_StorageKindE2ELi16EEEEEiES0_S6_jEJPS7_S0_fjEvET0_T1_DpT2__6" Import ; 0x00000e24
OpDecorate %llvm_genx_svm_block_ld_unaligned_v16f32_i64 LinkageAttributes "llvm.genx.svm.block.ld.unaligned.v16f32.i64" Import ; 0x00000ed4
OpDecorate %llvm_genx_svm_block_ld_unaligned_v16f32_i64 VectorComputeFunctionINTEL ; 0x00000f10
OpDecorate %44 Alignment 4 ; 0x00000f1c
OpDecorate %45 Alignment 4 ; 0x00000f2c
OpDecorate %46 Alignment 4 ; 0x00000f3c
OpDecorate %__itt_offload_wi_start_wrapper LinkageAttributes "__itt_offload_wi_start_wrapper" Export ; 0x00000f4c
OpDecorate %77 NoSignedWrap ; 0x00000f7c
OpDecorate %77 NoUnsignedWrap ; 0x00000f88
OpDecorate %_ZN4sycl3_V13ext6oneapi12experimental6detail33__regcall3__simd_func_call_helperILi16ERFNSt12experimental4simdIfNS6_10__simd_abiILNS6_12_StorageKindE2ELi16EEEEEPfSB_iEJNS3_7uniformISC_EEfNSF_IjEEEEENSt13invoke_resultIT0_JDpNS4_9spmd2simdIT1_XT_EvE4typeEEE4typeESJ_SO__3 LinkageAttributes "_ZN4sycl3_V13ext6oneapi12experimental6detail33__regcall3__simd_func_call_helperILi16ERFNSt12experimental4simdIfNS6_10__simd_abiILNS6_12_StorageKindE2ELi16EEEEEPfSB_iEJNS3_7uniformISC_EEfNSF_IjEEEEENSt13invoke_resultIT0_JDpNS4_9spmd2simdIT1_XT_EvE4typeEEE4typeESJ_SO__3" Export ; 0x00000f94
OpDecorate %_ZN4sycl3_V13ext6oneapi12experimental6detail33__regcall3__simd_func_call_helperILi16ERFNSt12experimental4simdIfNS6_10__simd_abiILNS6_12_StorageKindE2ELi16EEEEEPfSB_iEJNS3_7uniformISC_EEfNSF_IjEEEEENSt13invoke_resultIT0_JDpNS4_9spmd2simdIT1_XT_EvE4typeEEE4typeESJ_SO__3 ReferencedIndirectlyINTEL ; 0x000010b4
OpDecorate %_ZN4sycl3_V13ext6oneapi12experimental6detail33__regcall3__simd_func_call_helperILi16ERFNSt12experimental4simdIfNS6_10__simd_abiILNS6_12_StorageKindE2ELi16EEEEEPfSB_iEJNS3_7uniformISC_EEfNSF_IjEEEEENSt13invoke_resultIT0_JDpNS4_9spmd2simdIT1_XT_EvE4typeEEE4typeESJ_SO__3 StackCallINTEL ; 0x000010c0
OpDecorate %_ZN4sycl3_V13ext6oneapi12experimental6detail33__regcall3__simd_func_call_helperILi16ERFNSt12experimental4simdIfNS6_10__simd_abiILNS6_12_StorageKindE2ELi16EEEEEPfSB_iEJNS3_7uniformISC_EEfNSF_IjEEEEENSt13invoke_resultIT0_JDpNS4_9spmd2simdIT1_XT_EvE4typeEEE4typeESJ_SO__3 VectorComputeFunctionINTEL ; 0x000010cc
OpDecorate %_ZN4sycl3_V13ext6oneapi12experimental6detail33__regcall3__simd_func_call_helperILi16ERFNSt12experimental4simdIfNS6_10__simd_abiILNS6_12_StorageKindE2ELi16EEEEEPfSB_iEJNS3_7uniformISC_EEfNSF_IjEEEEENSt13invoke_resultIT0_JDpNS4_9spmd2simdIT1_XT_EvE4typeEEE4typeESJ_SO__1 LinkageAttributes "_ZN4sycl3_V13ext6oneapi12experimental6detail33__regcall3__simd_func_call_helperILi16ERFNSt12experimental4simdIfNS6_10__simd_abiILNS6_12_StorageKindE2ELi16EEEEEPfSB_iEJNS3_7uniformISC_EEfNSF_IjEEEEENSt13invoke_resultIT0_JDpNS4_9spmd2simdIT1_XT_EvE4typeEEE4typeESJ_SO__1" Export ; 0x000010d8
OpDecorate %_ZN4sycl3_V13ext6oneapi12experimental6detail33__regcall3__simd_func_call_helperILi16ERFNSt12experimental4simdIfNS6_10__simd_abiILNS6_12_StorageKindE2ELi16EEEEEPfSB_iEJNS3_7uniformISC_EEfNSF_IjEEEEENSt13invoke_resultIT0_JDpNS4_9spmd2simdIT1_XT_EvE4typeEEE4typeESJ_SO__1 ReferencedIndirectlyINTEL ; 0x000011f8
OpDecorate %_ZN4sycl3_V13ext6oneapi12experimental6detail33__regcall3__simd_func_call_helperILi16ERFNSt12experimental4simdIfNS6_10__simd_abiILNS6_12_StorageKindE2ELi16EEEEEPfSB_iEJNS3_7uniformISC_EEfNSF_IjEEEEENSt13invoke_resultIT0_JDpNS4_9spmd2simdIT1_XT_EvE4typeEEE4typeESJ_SO__1 StackCallINTEL ; 0x00001204
OpDecorate %_ZN4sycl3_V13ext6oneapi12experimental6detail33__regcall3__simd_func_call_helperILi16ERFNSt12experimental4simdIfNS6_10__simd_abiILNS6_12_StorageKindE2ELi16EEEEEPfSB_iEJNS3_7uniformISC_EEfNSF_IjEEEEENSt13invoke_resultIT0_JDpNS4_9spmd2simdIT1_XT_EvE4typeEEE4typeESJ_SO__1 VectorComputeFunctionINTEL ; 0x00001210
OpDecorate %94 FPFastMathMode Fast ; 0x0000121c
OpDecorate %_ZN4sycl3_V13ext6oneapi12experimental6detail33__regcall3__simd_func_call_helperILi16ERFvPfNSt12experimental4simdIfNS7_10__simd_abiILNS7_12_StorageKindE2ELi16EEEEEiEJNS3_7uniformIS6_EEfNSF_IjEEEEENSt13invoke_resultIT0_JDpNS4_9spmd2simdIT1_XT_EvE4typeEEE4typeESJ_SO__5 LinkageAttributes "_ZN4sycl3_V13ext6oneapi12experimental6detail33__regcall3__simd_func_call_helperILi16ERFvPfNSt12experimental4simdIfNS7_10__simd_abiILNS7_12_StorageKindE2ELi16EEEEEiEJNS3_7uniformIS6_EEfNSF_IjEEEEENSt13invoke_resultIT0_JDpNS4_9spmd2simdIT1_XT_EvE4typeEEE4typeESJ_SO__5" Export ; 0x0000122c
OpDecorate %_ZN4sycl3_V13ext6oneapi12experimental6detail33__regcall3__simd_func_call_helperILi16ERFvPfNSt12experimental4simdIfNS7_10__simd_abiILNS7_12_StorageKindE2ELi16EEEEEiEJNS3_7uniformIS6_EEfNSF_IjEEEEENSt13invoke_resultIT0_JDpNS4_9spmd2simdIT1_XT_EvE4typeEEE4typeESJ_SO__5 ReferencedIndirectlyINTEL ; 0x00001348
OpDecorate %_ZN4sycl3_V13ext6oneapi12experimental6detail33__regcall3__simd_func_call_helperILi16ERFvPfNSt12experimental4simdIfNS7_10__simd_abiILNS7_12_StorageKindE2ELi16EEEEEiEJNS3_7uniformIS6_EEfNSF_IjEEEEENSt13invoke_resultIT0_JDpNS4_9spmd2simdIT1_XT_EvE4typeEEE4typeESJ_SO__5 StackCallINTEL ; 0x00001354
OpDecorate %_ZN4sycl3_V13ext6oneapi12experimental6detail33__regcall3__simd_func_call_helperILi16ERFvPfNSt12experimental4simdIfNS7_10__simd_abiILNS7_12_StorageKindE2ELi16EEEEEiEJNS3_7uniformIS6_EEfNSF_IjEEEEENSt13invoke_resultIT0_JDpNS4_9spmd2simdIT1_XT_EvE4typeEEE4typeESJ_SO__5 VectorComputeFunctionINTEL ; 0x00001360
OpDecorate %__itt_offload_wi_finish_wrapper LinkageAttributes "__itt_offload_wi_finish_wrapper" Export ; 0x0000136c
OpDecorate %110 Alignment 8 ; 0x0000139c
OpDecorate %112 SpecId 4285822057 ; 0x000013ac
OpDecorate %__itt_offload_wi_start_stub LinkageAttributes "__itt_offload_wi_start_stub" Export ; 0x000013bc
OpDecorate %147 Alignment 8 ; 0x000013e8
OpDecorate %148 SpecId 4285822057 ; 0x000013f8
OpDecorate %__itt_offload_wi_finish_stub LinkageAttributes "__itt_offload_wi_finish_stub" Export ; 0x00001408
OpDecorate %167 Alignment 8 ; 0x00001438
OpDecorate %168 Alignment 8 ; 0x00001448
OpDecorate %170 Alignment 4 ; 0x00001458
OpDecorate %177 Alignment 8 ; 0x00001468
OpDecorate %178 Alignment 8 ; 0x00001478
OpDecorate %_Z24__regcall3__SIMD_CALLEE1PfNSt12experimental4simdIfNS0_10__simd_abiILNS0_12_StorageKindE2ELi16EEEEEi LinkageAttributes "_Z24__regcall3__SIMD_CALLEE1PfNSt12experimental4simdIfNS0_10__simd_abiILNS0_12_StorageKindE2ELi16EEEEEi" Export ; 0x00001488
OpDecorate %_Z24__regcall3__SIMD_CALLEE1PfNSt12experimental4simdIfNS0_10__simd_abiILNS0_12_StorageKindE2ELi16EEEEEi ReferencedIndirectlyINTEL ; 0x00001500
OpDecorate %_Z24__regcall3__SIMD_CALLEE1PfNSt12experimental4simdIfNS0_10__simd_abiILNS0_12_StorageKindE2ELi16EEEEEi VectorComputeFunctionINTEL ; 0x0000150c
OpDecorate %190 FPFastMathMode Fast ; 0x00001518
OpDecorate %_Z24__regcall3__SIMD_CALLEE2PfNSt12experimental4simdIfNS0_10__simd_abiILNS0_12_StorageKindE2ELi16EEEEEi LinkageAttributes "_Z24__regcall3__SIMD_CALLEE2PfNSt12experimental4simdIfNS0_10__simd_abiILNS0_12_StorageKindE2ELi16EEEEEi" Export ; 0x00001528
OpDecorate %_Z24__regcall3__SIMD_CALLEE2PfNSt12experimental4simdIfNS0_10__simd_abiILNS0_12_StorageKindE2ELi16EEEEEi ReferencedIndirectlyINTEL ; 0x000015a0
OpDecorate %_Z24__regcall3__SIMD_CALLEE2PfNSt12experimental4simdIfNS0_10__simd_abiILNS0_12_StorageKindE2ELi16EEEEEi VectorComputeFunctionINTEL ; 0x000015ac
OpDecorate %200 FPFastMathMode Fast ; 0x000015b8
OpDecorate %201 FPFastMathMode Fast ; 0x000015c8
OpDecorate %_Z28__regcall3__SIMD_CALLEE_VOIDPfNSt12experimental4simdIfNS0_10__simd_abiILNS0_12_StorageKindE2ELi16EEEEEi LinkageAttributes "_Z28__regcall3__SIMD_CALLEE_VOIDPfNSt12experimental4simdIfNS0_10__simd_abiILNS0_12_StorageKindE2ELi16EEEEEi" Export ; 0x000015d8
OpDecorate %_Z28__regcall3__SIMD_CALLEE_VOIDPfNSt12experimental4simdIfNS0_10__simd_abiILNS0_12_StorageKindE2ELi16EEEEEi ReferencedIndirectlyINTEL ; 0x00001654
OpDecorate %_Z28__regcall3__SIMD_CALLEE_VOIDPfNSt12experimental4simdIfNS0_10__simd_abiILNS0_12_StorageKindE2ELi16EEEEEi VectorComputeFunctionINTEL ; 0x00001660
OpDecorate %216 FPFastMathMode Fast ; 0x0000166c
OpDecorate %217 FPFastMathMode Fast ; 0x0000167c
OpDecorate %223 FPFastMathMode Fast ; 0x0000168c
%uint = OpTypeInt 32 0 ; 0x0000169c
%ulong = OpTypeInt 64 0 ; 0x000016ac
%uchar = OpTypeInt 8 0 ; 0x000016bc
%uint_4 = OpConstant %uint 4 ; 0x000016cc
%uint_6 = OpConstant %uint 6 ; 0x000016dc
%ulong_2147483648 = OpConstant %ulong 2147483648 ; 0x000016ec
%ulong_3 = OpConstant %ulong 3 ; 0x00001700
%112 = OpSpecConstant %uchar 0 ; 0x00001714
%uchar_0 = OpConstant %uchar 0 ; 0x00001724
%ulong_0 = OpConstant %ulong 0 ; 0x00001734
%ulong_1 = OpConstant %ulong 1 ; 0x00001748
%ulong_2 = OpConstant %ulong 2 ; 0x0000175c
%148 = OpSpecConstant %uchar 0 ; 0x00001770
%_ptr_CrossWorkgroup_uint = OpTypePointer CrossWorkgroup %uint ; 0x00001780
%v3ulong = OpTypeVector %ulong 3 ; 0x00001790
%_ptr_CrossWorkgroup_v3ulong = OpTypePointer CrossWorkgroup %v3ulong ; 0x000017a0
%_ptr_CrossWorkgroup_ulong = OpTypePointer CrossWorkgroup %ulong ; 0x000017b0
%float = OpTypeFloat 32 ; 0x000017c0
%v16float = OpTypeVector %float 16 ; 0x000017cc
%_ptr_Generic_float = OpTypePointer Generic %float ; 0x000017dc
%16 = OpTypeFunction %v16float %_ptr_Generic_float %v16float %uint ; 0x000017ec
%_ptr_Function_16 = OpTypePointer Function %16 ; 0x00001804
%18 = OpTypeFunction %float %_ptr_Function_16 %_ptr_Generic_float %float %uint ; 0x00001814
%void = OpTypeVoid ; 0x00001830
%30 = OpTypeFunction %void %_ptr_Generic_float %v16float %uint ; 0x00001838
%_ptr_Function_30 = OpTypePointer Function %30 ; 0x00001850
%32 = OpTypeFunction %void %_ptr_Function_30 %_ptr_Generic_float %float %uint ; 0x00001860
%38 = OpTypeFunction %v16float %ulong ; 0x0000187c
%_ptr_CrossWorkgroup_float = OpTypePointer CrossWorkgroup %float ; 0x0000188c
%42 = OpTypeFunction %void %_ptr_CrossWorkgroup_float %_ptr_CrossWorkgroup_float %_ptr_CrossWorkgroup_float ; 0x0000189c
%48 = OpTypeFunction %void ; 0x000018b4
%bool = OpTypeBool ; 0x000018c0
%_arr_ulong_ulong_3 = OpTypeArray %ulong %ulong_3 ; 0x000018c8
%_ptr_Function__arr_ulong_ulong_3 = OpTypePointer Function %_arr_ulong_ulong_3 ; 0x000018d8
%_ptr_Function_uchar = OpTypePointer Function %uchar ; 0x000018e8
%_ptr_Function_ulong = OpTypePointer Function %ulong ; 0x000018f8
%_ptr_Generic_ulong = OpTypePointer Generic %ulong ; 0x00001908
%138 = OpTypeFunction %void %_ptr_Generic_ulong %ulong %uint ; 0x00001918
%160 = OpTypeFunction %void %_ptr_Generic_ulong %ulong ; 0x00001930
%_ptr_Function__ptr_Generic_ulong = OpTypePointer Function %_ptr_Generic_ulong ; 0x00001944
%_ptr_Function_uint = OpTypePointer Function %uint ; 0x00001954
%_ptr_Generic__ptr_Generic_ulong = OpTypePointer Generic %_ptr_Generic_ulong ; 0x00001964
%_ptr_Generic_uint = OpTypePointer Generic %uint ; 0x00001974
%__spirv_BuiltInSubgroupId = OpVariable %_ptr_CrossWorkgroup_uint CrossWorkgroup ; 0x00001984
%__spirv_BuiltInSubgroupLocalInvocationId = OpVariable %_ptr_CrossWorkgroup_uint CrossWorkgroup ; 0x00001994
%__spirv_BuiltInWorkgroupId = OpVariable %_ptr_CrossWorkgroup_v3ulong CrossWorkgroup ; 0x000019a4
%__spirv_BuiltInGlobalLinearId = OpVariable %_ptr_CrossWorkgroup_ulong CrossWorkgroup ; 0x000019b4
%__spirv_BuiltInWorkgroupSize = OpVariable %_ptr_CrossWorkgroup_v3ulong CrossWorkgroup ; 0x000019c4
%_Z33__regcall3____builtin_invoke_simdILb1EfPFNSt12experimental4simdIfNS0_10__simd_abiILNS0_12_StorageKindE2ELi16EEEEERFS5_PfS5_iES6_S5_jEJPS7_S6_fjEvET0_T1_DpT2__2 = OpFunction %float None %18 ; 0x000019d4
%20 = OpFunctionParameter %_ptr_Function_16 ; 0x000019e8
%21 = OpFunctionParameter %_ptr_Generic_float ; 0x000019f4
%22 = OpFunctionParameter %float ; 0x00001a00
%23 = OpFunctionParameter %uint ; 0x00001a0c
OpFunctionEnd ; 0x00001a18
%_Z33__regcall3____builtin_invoke_simdILb1EfPFNSt12experimental4simdIfNS0_10__simd_abiILNS0_12_StorageKindE2ELi16EEEEERFS5_PfS5_iES6_S5_jEJPS7_S6_fjEvET0_T1_DpT2__4 = OpFunction %float None %18 ; 0x00001a1c
%25 = OpFunctionParameter %_ptr_Function_16 ; 0x00001a30
%26 = OpFunctionParameter %_ptr_Generic_float ; 0x00001a3c
%27 = OpFunctionParameter %float ; 0x00001a48
%28 = OpFunctionParameter %uint ; 0x00001a54
OpFunctionEnd ; 0x00001a60
%_Z33__regcall3____builtin_invoke_simdILb1EvPFvRFvPfNSt12experimental4simdIfNS1_10__simd_abiILNS1_12_StorageKindE2ELi16EEEEEiES0_S6_jEJPS7_S0_fjEvET0_T1_DpT2__6 = OpFunction %void None %32 ; 0x00001a64
%34 = OpFunctionParameter %_ptr_Function_30 ; 0x00001a78
%35 = OpFunctionParameter %_ptr_Generic_float ; 0x00001a84
%36 = OpFunctionParameter %float ; 0x00001a90
%37 = OpFunctionParameter %uint ; 0x00001a9c
OpFunctionEnd ; 0x00001aa8
%llvm_genx_svm_block_ld_unaligned_v16f32_i64 = OpFunction %v16float Inline %38 ; 0x00001aac
%40 = OpFunctionParameter %ulong ; 0x00001ac0
OpFunctionEnd ; 0x00001acc
%43 = OpFunction %void None %42 ; 0x00001ad0
%44 = OpFunctionParameter %_ptr_CrossWorkgroup_float ; 0x00001ae4
%45 = OpFunctionParameter %_ptr_CrossWorkgroup_float ; 0x00001af0
%46 = OpFunctionParameter %_ptr_CrossWorkgroup_float ; 0x00001afc
%47 = OpLabel ; 0x00001b08
%50 = OpFunctionCall %void %__itt_offload_wi_start_wrapper ; 0x00001b10
%51 = OpPtrCastToGeneric %_ptr_Generic_float %44 ; 0x00001b20
%59 = OpLoad %v3ulong %__spirv_BuiltInWorkgroupId Aligned|NoAliasINTELMask 32 %58 ; 0x00001b30
%60 = OpCompositeExtract %ulong %59 0 ; 0x00001b4c
%64 = OpLoad %uint %__spirv_BuiltInSubgroupId Aligned|NoAliasINTELMask 4 %63 ; 0x00001b60
%66 = OpShiftLeftLogical %uint %64 %uint_4 ; 0x00001b7c
%67 = OpUConvert %uint %60 ; 0x00001b90
%69 = OpShiftLeftLogical %uint %67 %uint_6 ; 0x00001ba0
%70 = OpIAdd %uint %69 %66 ; 0x00001bb4
%74 = OpLoad %uint %__spirv_BuiltInSubgroupLocalInvocationId Aligned|NoAliasINTELMask 4 %73 ; 0x00001bc8
%75 = OpUConvert %ulong %74 ; 0x00001be4
%76 = OpUConvert %ulong %70 ; 0x00001bf4
%77 = OpIAdd %ulong %75 %76 ; 0x00001c04
%80 = OpULessThan %bool %77 %ulong_2147483648 ; 0x00001c18
OpAssumeTrueKHR %80 ; 0x00001c2c
%81 = OpInBoundsPtrAccessChain %_ptr_CrossWorkgroup_float %45 %77 ; 0x00001c34
%82 = OpLoad %float %81 Aligned 4 ; 0x00001c48
%87 = OpFunctionCall %float %_Z33__regcall3____builtin_invoke_simdILb1EfPFNSt12experimental4simdIfNS0_10__simd_abiILNS0_12_StorageKindE2ELi16EEEEERFS5_PfS5_iES6_S5_jEJPS7_S6_fjEvET0_T1_DpT2__4 %_ZN4sycl3_V13ext6oneapi12experimental6detail33__regcall3__simd_func_call_helperILi16ERFNSt12experimental4simdIfNS6_10__simd_abiILNS6_12_StorageKindE2ELi16EEEEEPfSB_iEJNS3_7uniformISC_EEfNSF_IjEEEEENSt13invoke_resultIT0_JDpNS4_9spmd2simdIT1_XT_EvE4typeEEE4typeESJ_SO__3 %51 %82 %70 ; 0x00001c60
%88 = OpLoad %float %81 Aligned 4 ; 0x00001c80
%93 = OpFunctionCall %float %_Z33__regcall3____builtin_invoke_simdILb1EfPFNSt12experimental4simdIfNS0_10__simd_abiILNS0_12_StorageKindE2ELi16EEEEERFS5_PfS5_iES6_S5_jEJPS7_S6_fjEvET0_T1_DpT2__2 %_ZN4sycl3_V13ext6oneapi12experimental6detail33__regcall3__simd_func_call_helperILi16ERFNSt12experimental4simdIfNS6_10__simd_abiILNS6_12_StorageKindE2ELi16EEEEEPfSB_iEJNS3_7uniformISC_EEfNSF_IjEEEEENSt13invoke_resultIT0_JDpNS4_9spmd2simdIT1_XT_EvE4typeEEE4typeESJ_SO__1 %51 %88 %70 ; 0x00001c98
%94 = OpFAdd %float %87 %93 ; 0x00001cb8
%95 = OpLoad %float %81 Aligned 4 ; 0x00001ccc
%100 = OpFunctionCall %void %_Z33__regcall3____builtin_invoke_simdILb1EvPFvRFvPfNSt12experimental4simdIfNS1_10__simd_abiILNS1_12_StorageKindE2ELi16EEEEEiES0_S6_jEJPS7_S0_fjEvET0_T1_DpT2__6 %_ZN4sycl3_V13ext6oneapi12experimental6detail33__regcall3__simd_func_call_helperILi16ERFvPfNSt12experimental4simdIfNS7_10__simd_abiILNS7_12_StorageKindE2ELi16EEEEEiEJNS3_7uniformIS6_EEfNSF_IjEEEEENSt13invoke_resultIT0_JDpNS4_9spmd2simdIT1_XT_EvE4typeEEE4typeESJ_SO__5 %51 %95 %70 ; 0x00001ce4
%101 = OpInBoundsPtrAccessChain %_ptr_CrossWorkgroup_float %46 %77 ; 0x00001d04
OpStore %101 %94 Aligned 4 ; 0x00001d18
%103 = OpFunctionCall %void %__itt_offload_wi_finish_wrapper ; 0x00001d2c
OpReturn ; 0x00001d3c
OpFunctionEnd ; 0x00001d40
%__itt_offload_wi_start_wrapper = OpFunction %void Inline %48 ; 0x00001d44
%104 = OpLabel ; 0x00001d58
%110 = OpVariable %_ptr_Function__arr_ulong_ulong_3 Function ; 0x00001d60
%114 = OpIEqual %bool %112 %uchar_0 ; 0x00001d70
OpBranchConditional %114 %106 %105 ; 0x00001d84
%105 = OpLabel ; 0x00001d94
%116 = OpBitcast %_ptr_Function_uchar %110 ; 0x00001d9c
OpLifetimeStart %116 24 ; 0x00001dac
%119 = OpInBoundsPtrAccessChain %_ptr_Function_ulong %110 %ulong_0 %ulong_0 ; 0x00001db8
%121 = OpPtrCastToGeneric %_ptr_Generic_ulong %119 ; 0x00001dd0
%122 = OpLoad %v3ulong %__spirv_BuiltInWorkgroupId Aligned 32 ; 0x00001de0
%123 = OpCompositeExtract %ulong %122 0 ; 0x00001df8
OpStore %119 %123 Aligned 8 ; 0x00001e0c
%125 = OpInBoundsPtrAccessChain %_ptr_Function_ulong %110 %ulong_0 %ulong_1 ; 0x00001e20
%126 = OpCompositeExtract %ulong %122 1 ; 0x00001e38
OpStore %125 %126 Aligned 8 ; 0x00001e4c
%128 = OpInBoundsPtrAccessChain %_ptr_Function_ulong %110 %ulong_0 %ulong_2 ; 0x00001e60
%129 = OpCompositeExtract %ulong %122 2 ; 0x00001e78
OpStore %128 %129 Aligned 8 ; 0x00001e8c
%130 = OpLoad %ulong %__spirv_BuiltInGlobalLinearId Aligned 8 ; 0x00001ea0
%131 = OpLoad %v3ulong %__spirv_BuiltInWorkgroupSize Aligned 32 ; 0x00001eb8
%132 = OpCompositeExtract %ulong %131 0 ; 0x00001ed0
%133 = OpCompositeExtract %ulong %131 1 ; 0x00001ee4
%134 = OpIMul %ulong %132 %133 ; 0x00001ef8
%135 = OpCompositeExtract %ulong %131 2 ; 0x00001f0c
%136 = OpIMul %ulong %134 %135 ; 0x00001f20
%137 = OpUConvert %uint %136 ; 0x00001f34
%143 = OpFunctionCall %void %__itt_offload_wi_start_stub %121 %130 %137 ; 0x00001f44
OpLifetimeStop %116 24 ; 0x00001f60
OpBranch %106 ; 0x00001f6c
%106 = OpLabel ; 0x00001f74
OpReturn ; 0x00001f7c
OpFunctionEnd ; 0x00001f80
%_ZN4sycl3_V13ext6oneapi12experimental6detail33__regcall3__simd_func_call_helperILi16ERFNSt12experimental4simdIfNS6_10__simd_abiILNS6_12_StorageKindE2ELi16EEEEEPfSB_iEJNS3_7uniformISC_EEfNSF_IjEEEEENSt13invoke_resultIT0_JDpNS4_9spmd2simdIT1_XT_EvE4typeEEE4typeESJ_SO__3 = OpFunction %v16float None %16 ; 0x00001f84
%84 = OpFunctionParameter %_ptr_Generic_float ; 0x00001f98
%85 = OpFunctionParameter %v16float ; 0x00001fa4
%86 = OpFunctionParameter %uint ; 0x00001fb0
%218 = OpLabel ; 0x00001fbc
%219 = OpSConvert %ulong %86 ; 0x00001fc4
%220 = OpInBoundsPtrAccessChain %_ptr_Generic_float %84 %219 ; 0x00001fd4
%221 = OpConvertPtrToU %ulong %220 ; 0x00001fe8
%_esimd_i_0 = OpFunctionCall %v16float %llvm_genx_svm_block_ld_unaligned_v16f32_i64 %221 ; 0x00001ff8
%223 = OpFAdd %v16float %_esimd_i_0 %85 ; 0x0000200c
OpReturnValue %223 ; 0x00002020
OpFunctionEnd ; 0x00002028
%_ZN4sycl3_V13ext6oneapi12experimental6detail33__regcall3__simd_func_call_helperILi16ERFNSt12experimental4simdIfNS6_10__simd_abiILNS6_12_StorageKindE2ELi16EEEEEPfSB_iEJNS3_7uniformISC_EEfNSF_IjEEEEENSt13invoke_resultIT0_JDpNS4_9spmd2simdIT1_XT_EvE4typeEEE4typeESJ_SO__1 = OpFunction %v16float None %16 ; 0x0000202c
%90 = OpFunctionParameter %_ptr_Generic_float ; 0x00002040
%91 = OpFunctionParameter %v16float ; 0x0000204c
%92 = OpFunctionParameter %uint ; 0x00002058
%211 = OpLabel ; 0x00002064
%212 = OpSConvert %ulong %92 ; 0x0000206c
%213 = OpInBoundsPtrAccessChain %_ptr_Generic_float %90 %212 ; 0x0000207c
%214 = OpConvertPtrToU %ulong %213 ; 0x00002090
%_esimd_i = OpFunctionCall %v16float %llvm_genx_svm_block_ld_unaligned_v16f32_i64 %214 ; 0x000020a0
%216 = OpFAdd %v16float %_esimd_i %91 ; 0x000020b4
%217 = OpFAdd %v16float %216 %216 ; 0x000020c8
OpReturnValue %217 ; 0x000020dc
OpFunctionEnd ; 0x000020e4
%_ZN4sycl3_V13ext6oneapi12experimental6detail33__regcall3__simd_func_call_helperILi16ERFvPfNSt12experimental4simdIfNS7_10__simd_abiILNS7_12_StorageKindE2ELi16EEEEEiEJNS3_7uniformIS6_EEfNSF_IjEEEEENSt13invoke_resultIT0_JDpNS4_9spmd2simdIT1_XT_EvE4typeEEE4typeESJ_SO__5 = OpFunction %void None %30 ; 0x000020e8
%97 = OpFunctionParameter %_ptr_Generic_float ; 0x000020fc
%98 = OpFunctionParameter %v16float ; 0x00002108
%99 = OpFunctionParameter %uint ; 0x00002114
%224 = OpLabel ; 0x00002120
%225 = OpSConvert %ulong %99 ; 0x00002128
%226 = OpInBoundsPtrAccessChain %_ptr_Generic_float %97 %225 ; 0x00002138
%227 = OpConvertPtrToU %ulong %226 ; 0x0000214c
%_esimd_i_1 = OpFunctionCall %v16float %llvm_genx_svm_block_ld_unaligned_v16f32_i64 %227 ; 0x0000215c
OpReturn ; 0x00002170
OpFunctionEnd ; 0x00002174
%__itt_offload_wi_finish_wrapper = OpFunction %void Inline %48 ; 0x00002178
%144 = OpLabel ; 0x0000218c
%147 = OpVariable %_ptr_Function__arr_ulong_ulong_3 Function ; 0x00002194
%149 = OpIEqual %bool %148 %uchar_0 ; 0x000021a4
OpBranchConditional %149 %146 %145 ; 0x000021b8
%145 = OpLabel ; 0x000021c8
%150 = OpBitcast %_ptr_Function_uchar %147 ; 0x000021d0
OpLifetimeStart %150 24 ; 0x000021e0
%151 = OpInBoundsPtrAccessChain %_ptr_Function_ulong %147 %ulong_0 %ulong_0 ; 0x000021ec
%152 = OpPtrCastToGeneric %_ptr_Generic_ulong %151 ; 0x00002204
%153 = OpLoad %v3ulong %__spirv_BuiltInWorkgroupId Aligned 32 ; 0x00002214
%154 = OpCompositeExtract %ulong %153 0 ; 0x0000222c
OpStore %151 %154 Aligned 8 ; 0x00002240
%155 = OpInBoundsPtrAccessChain %_ptr_Function_ulong %147 %ulong_0 %ulong_1 ; 0x00002254
%156 = OpCompositeExtract %ulong %153 1 ; 0x0000226c
OpStore %155 %156 Aligned 8 ; 0x00002280
%157 = OpInBoundsPtrAccessChain %_ptr_Function_ulong %147 %ulong_0 %ulong_2 ; 0x00002294
%158 = OpCompositeExtract %ulong %153 2 ; 0x000022ac
OpStore %157 %158 Aligned 8 ; 0x000022c0
%159 = OpLoad %ulong %__spirv_BuiltInGlobalLinearId Aligned 8 ; 0x000022d4
%164 = OpFunctionCall %void %__itt_offload_wi_finish_stub %152 %159 ; 0x000022ec
OpLifetimeStop %150 24 ; 0x00002304
OpBranch %146 ; 0x00002310
%146 = OpLabel ; 0x00002318
OpReturn ; 0x00002320
OpFunctionEnd ; 0x00002324
%__itt_offload_wi_start_stub = OpFunction %void DontInline|OptNoneINTEL %138 ; 0x00002328
%140 = OpFunctionParameter %_ptr_Generic_ulong ; 0x0000233c
%141 = OpFunctionParameter %ulong ; 0x00002348
%142 = OpFunctionParameter %uint ; 0x00002354
%165 = OpLabel ; 0x00002360
%167 = OpVariable %_ptr_Function__ptr_Generic_ulong Function ; 0x00002368
%168 = OpVariable %_ptr_Function_ulong Function ; 0x00002378
%170 = OpVariable %_ptr_Function_uint Function ; 0x00002388
%172 = OpPtrCastToGeneric %_ptr_Generic__ptr_Generic_ulong %167 ; 0x00002398
%173 = OpPtrCastToGeneric %_ptr_Generic_ulong %168 ; 0x000023a8
%175 = OpPtrCastToGeneric %_ptr_Generic_uint %170 ; 0x000023b8
OpStore %172 %140 Aligned 8 ; 0x000023c8
OpStore %173 %141 Aligned 8 ; 0x000023dc
OpStore %175 %142 Aligned 4 ; 0x000023f0
OpReturn ; 0x00002404
OpFunctionEnd ; 0x00002408
%__itt_offload_wi_finish_stub = OpFunction %void DontInline|OptNoneINTEL %160 ; 0x0000240c
%162 = OpFunctionParameter %_ptr_Generic_ulong ; 0x00002420
%163 = OpFunctionParameter %ulong ; 0x0000242c
%176 = OpLabel ; 0x00002438
%177 = OpVariable %_ptr_Function__ptr_Generic_ulong Function ; 0x00002440
%178 = OpVariable %_ptr_Function_ulong Function ; 0x00002450
%179 = OpPtrCastToGeneric %_ptr_Generic__ptr_Generic_ulong %177 ; 0x00002460
%180 = OpPtrCastToGeneric %_ptr_Generic_ulong %178 ; 0x00002470
OpStore %179 %162 Aligned 8 ; 0x00002480
OpStore %180 %163 Aligned 8 ; 0x00002494
OpReturn ; 0x000024a8
OpFunctionEnd ; 0x000024ac
%_Z24__regcall3__SIMD_CALLEE1PfNSt12experimental4simdIfNS0_10__simd_abiILNS0_12_StorageKindE2ELi16EEEEEi = OpFunction %v16float Inline %16 ; 0x000024b0
%182 = OpFunctionParameter %_ptr_Generic_float ; 0x000024c4
%183 = OpFunctionParameter %v16float ; 0x000024d0
%184 = OpFunctionParameter %uint ; 0x000024dc
%185 = OpLabel ; 0x000024e8
%186 = OpSConvert %ulong %184 ; 0x000024f0
%187 = OpInBoundsPtrAccessChain %_ptr_Generic_float %182 %186 ; 0x00002500
%188 = OpConvertPtrToU %ulong %187 ; 0x00002514
%_esimd = OpFunctionCall %v16float %llvm_genx_svm_block_ld_unaligned_v16f32_i64 %188 ; 0x00002524
%190 = OpFAdd %v16float %_esimd %183 ; 0x00002538
OpReturnValue %190 ; 0x0000254c
OpFunctionEnd ; 0x00002554
%_Z24__regcall3__SIMD_CALLEE2PfNSt12experimental4simdIfNS0_10__simd_abiILNS0_12_StorageKindE2ELi16EEEEEi = OpFunction %v16float Inline %16 ; 0x00002558
%192 = OpFunctionParameter %_ptr_Generic_float ; 0x0000256c
%193 = OpFunctionParameter %v16float ; 0x00002578
%194 = OpFunctionParameter %uint ; 0x00002584
%195 = OpLabel ; 0x00002590
%196 = OpSConvert %ulong %194 ; 0x00002598
%197 = OpInBoundsPtrAccessChain %_ptr_Generic_float %192 %196 ; 0x000025a8
%198 = OpConvertPtrToU %ulong %197 ; 0x000025bc
%_esimd_0 = OpFunctionCall %v16float %llvm_genx_svm_block_ld_unaligned_v16f32_i64 %198 ; 0x000025cc
%200 = OpFAdd %v16float %_esimd_0 %193 ; 0x000025e0
%201 = OpFAdd %v16float %200 %200 ; 0x000025f4
OpReturnValue %201 ; 0x00002608
OpFunctionEnd ; 0x00002610
%_Z28__regcall3__SIMD_CALLEE_VOIDPfNSt12experimental4simdIfNS0_10__simd_abiILNS0_12_StorageKindE2ELi16EEEEEi = OpFunction %void Inline %30 ; 0x00002614
%203 = OpFunctionParameter %_ptr_Generic_float ; 0x00002628
%204 = OpFunctionParameter %v16float ; 0x00002634
%205 = OpFunctionParameter %uint ; 0x00002640
%206 = OpLabel ; 0x0000264c
%207 = OpSConvert %ulong %205 ; 0x00002654
%208 = OpInBoundsPtrAccessChain %_ptr_Generic_float %203 %207 ; 0x00002664
%209 = OpConvertPtrToU %ulong %208 ; 0x00002678
%_esimd_1 = OpFunctionCall %v16float %llvm_genx_svm_block_ld_unaligned_v16f32_i64 %209 ; 0x00002688
OpReturn ; 0x0000269c
OpFunctionEnd ; 0x000026a0
Jianhui-Li
Background
The Triton kernel is generated as SIMT major SPIRV kernel. It is because some component has to be used with SIMT paradigm. Like: Intel math library is only SIMT version. But for some minor portion of the kernel, we might have to use the SIMD paradigm for performance or functionality. Like: we may want to use the VC intrinsic in the Triton kernel.
We are working on enabling SIMT->SIMD calling convention on Triton kernel. https://github.com/intel/llvm/blob/sycl/sycl/doc/extensions/experimental/sycl_ext_oneapi_invoke_simd.asciidoc
By doing so, we can codegen SIMD paradigm code for parts of the kernel.
The requirements
We referred the SPIRV generated by the DPCPP which is SIMT+SIMD. We need to refer the SPIRV kernel function directly thru a function pointer. https://github.com/intel/llvm/blob/sycl/sycl/doc/design/spirv-extensions/SPV_INTEL_function_pointers.asciidoc
We need the SPIRV dialect to support this.