74 lines · plain
1; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx942 -filetype=obj < %s | llvm-objdump -s -j .rodata - | FileCheck --check-prefix=OBJDUMP %s2; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx942 < %s | FileCheck --check-prefix=ASM %s3 4; OBJDUMP: Contents of section .rodata:5; OBJDUMP-NEXT: 0000 00000000 00000000 10010000 00000000 ................6; OBJDUMP-NEXT: 0010 00000000 00000000 00000000 00000000 ................7; OBJDUMP-NEXT: 0020 00000000 00000000 00000000 00000000 ................8; OBJDUMP-NOT: 0030 0000af00 94130000 1a000400 00000000 ................9; OBJDUMP-NEXT: 0030 8000af00 98130000 1e000400 00000000 ................10 11; ASM-LABEL: amdhsa_kernarg_preload_4_implicit_6:12; ASM: .amdhsa_user_sgpr_count 1213; ASM: .amdhsa_next_free_sgpr 1214; ASM: ; TotalNumSgprs: 1815; ASM: ; NumSGPRsForWavesPerEU: 1816 17; Test that we include preloaded SGPRs in the GRANULATED_WAVEFRONT_SGPR_COUNT18; feild that are not explicitly referenced in the kernel. This test has 6 implicit19; user SPGRs enabled, 4 preloaded kernarg SGPRs, plus 6 extra SGPRs allocated20; for flat scratch, ect. The total number of allocated SGPRs encoded in the21; kernel descriptor should be 16. That's a 1 in the KD field since the granule22; size is 8 and it's NumGranules - 1. The encoding for that looks like '40'.23 24define amdgpu_kernel void @amdhsa_kernarg_preload_4_implicit_6(i128 inreg) { ret void }25 26; OBJDUMP-NEXT: 0040 00000000 00000000 20010000 00000000 ........ .......27; OBJDUMP-NEXT: 0050 00000000 00000000 00000000 00000000 ................28; OBJDUMP-NEXT: 0060 00000000 00000000 00000000 00000000 ................29; OBJDUMP-NEXT: 0070 4000af00 94000000 08000800 00000000 @...............30 31; ASM-LABEL: amdhsa_kernarg_preload_8_implicit_2:32; ASM: .amdhsa_user_sgpr_count 1033; ASM: .amdhsa_next_free_sgpr 1034; ASM: ; TotalNumSgprs: 1635; ASM: ; NumSGPRsForWavesPerEU: 1636 37; Only the kernarg_ptr is enabled so we should have 8 preload kernarg SGPRs, 238; implicit, and 6 extra.39 40define amdgpu_kernel void @amdhsa_kernarg_preload_8_implicit_2(i256 inreg) #0 { ret void }41 42; OBJDUMP-NEXT: 0080 00000000 00000000 08010000 00000000 ................43; OBJDUMP-NEXT: 0090 00000000 00000000 00000000 00000000 ................44; OBJDUMP-NEXT: 00a0 00000000 00000000 00000000 00000000 ................45; OBJDUMP-NEXT: 00b0 4000af00 86000000 08000100 00000000 @...............46 47; ASM-LABEL: amdhsa_kernarg_preload_1_implicit_2:48; ASM: .amdhsa_user_sgpr_count 349; ASM: .amdhsa_next_free_sgpr 350; ASM: ; TotalNumSgprs: 951; ASM: ; NumSGPRsForWavesPerEU: 952 53; 1 preload, 2 implicit, 6 extra. Rounds up to 16 SGPRs in the KD.54 55define amdgpu_kernel void @amdhsa_kernarg_preload_1_implicit_2(i32 inreg) #0 { ret void }56 57; OBJDUMP-NEXT: 00c0 00000000 00000000 08010000 00000000 ................58; OBJDUMP-NEXT: 00d0 00000000 00000000 00000000 00000000 ................59; OBJDUMP-NEXT: 00e0 00000000 00000000 00000000 00000000 ................60; OBJDUMP-NEXT: 00f0 0000af00 84000000 08000000 00000000 ................61 62; ASM-LABEL: amdhsa_kernarg_preload_0_implicit_2:63; ASM: .amdhsa_user_sgpr_count 264; ASM: .amdhsa_next_free_sgpr 065; ASM: ; TotalNumSgprs: 666; ASM: ; NumSGPRsForWavesPerEU: 667 68; 0 preload kernarg SGPRs, 2 implicit, 6 extra. Rounds up to 8 SGPRs in the KD.69; Encoded like '00'.70 71define amdgpu_kernel void @amdhsa_kernarg_preload_0_implicit_2(i32) #0 { ret void }72 73attributes #0 = { "amdgpu-agpr-alloc"="0" "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-lds-kernel-id" "amdgpu-no-multigrid-sync-arg" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-cluster-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-cluster-id-y" "amdgpu-no-workgroup-id-z" "amdgpu-no-cluster-id-z" "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z" "uniform-work-group-size"="false" }74