derek.holloway
7502bf64a4
CodeQL / Analyze (actions) (push) Canceled after 0s
CodeQL / Analyze (python) (push) Canceled after 0s
pre-commit / pre-commit (push) Canceled after 0s
Multi-Arch CI / setup (push) Canceled after 0s
Publish build_manylinux_x86_64 images / publish_build_manylinux_x86_64 (push) Canceled after 0s
Setup (Multi-Arch) / setup (push) Canceled after 0s
Multi-Arch CI / Linux::${{ fromJSON(needs.setup.outputs.linux_build_config || '{}').build_variant_label || 'skip' }} (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Publish no_rocm_image_ubi10 image / publish_no_rocm_image_ubi10 (push) Canceled after 0s
Multi-Arch CI / Windows::${{ fromJSON(needs.setup.outputs.windows_build_config || '{}').build_variant_label || 'skip' }} (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Multi-Arch CI / Manifest Diff (push) Canceled after 0s
TheRock Manifest Diff Report / Generate Manifest Diff Report (push) Canceled after 0s
Multi-Arch CI / CI Summary (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04 (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04_ocl_rt images / publish_no_rocm_image_ubuntu24_04_ocl_rt (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_rocgdb (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04_rocgdb images / publish_no_rocm_image_ubuntu24_04_rocgdb (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_ocl_rt (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Unit Tests / Unit Tests :: ubuntu-24.04 (push) Canceled after 0s
Unit Tests / Unit Tests :: windows-2022 (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_media (push) Canceled after 0s
Unit Tests / Unit Tests Summary (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_openmpi (push) Canceled after 0s
940 lines
43 KiB
LLVM
940 lines
43 KiB
LLVM
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
|
|
; RUN: llc -mtriple=amdgcn -mattr=+fast-fmaf,+mad-mac-f32-insts -denormal-fp-math-f32=preserve-sign < %s | FileCheck -check-prefix=GCN -check-prefix=GCN-FLUSH %s
|
|
; RUN: llc -mtriple=amdgcn -mattr=-fast-fmaf,+mad-mac-f32-insts -denormal-fp-math-f32=preserve-sign < %s | FileCheck -check-prefix=GCN -check-prefix=GCN-FLUSH %s
|
|
|
|
; RUN: llc -mtriple=amdgcn -mattr=+fast-fmaf,+mad-mac-f32-insts -denormal-fp-math-f32=ieee < %s | FileCheck -check-prefix=GCN -check-prefix=GCN-FASTFMA %s
|
|
; RUN: llc -mtriple=amdgcn -mattr=-fast-fmaf,+mad-mac-f32-insts -denormal-fp-math-f32=ieee < %s | FileCheck -check-prefix=GCN -check-prefix=GCN-SLOWFMA %s
|
|
|
|
; FIXME: This should also fold when fma is actually fast if an FMA
|
|
; exists in the original program.
|
|
|
|
; (fadd (fma x, y, (fmul u, v), z) -> (fma x, y (fma u, v, z))
|
|
define amdgpu_kernel void @fast_add_fmuladd_fmul() #0 {
|
|
; GCN-FLUSH-LABEL: fast_add_fmuladd_fmul:
|
|
; GCN-FLUSH: ; %bb.0:
|
|
; GCN-FLUSH-NEXT: s_mov_b32 s3, 0xf000
|
|
; GCN-FLUSH-NEXT: s_mov_b32 s2, -1
|
|
; GCN-FLUSH-NEXT: buffer_load_dword v0, off, s[0:3], 0 glc
|
|
; GCN-FLUSH-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FLUSH-NEXT: buffer_load_dword v1, off, s[0:3], 0 glc
|
|
; GCN-FLUSH-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FLUSH-NEXT: buffer_load_dword v2, off, s[0:3], 0 glc
|
|
; GCN-FLUSH-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FLUSH-NEXT: buffer_load_dword v3, off, s[0:3], 0 glc
|
|
; GCN-FLUSH-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FLUSH-NEXT: buffer_load_dword v4, off, s[0:3], 0 glc
|
|
; GCN-FLUSH-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FLUSH-NEXT: v_mac_f32_e32 v2, v3, v4
|
|
; GCN-FLUSH-NEXT: v_mac_f32_e32 v2, v0, v1
|
|
; GCN-FLUSH-NEXT: buffer_store_dword v2, off, s[0:3], 0
|
|
; GCN-FLUSH-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FLUSH-NEXT: s_endpgm
|
|
;
|
|
; GCN-FASTFMA-LABEL: fast_add_fmuladd_fmul:
|
|
; GCN-FASTFMA: ; %bb.0:
|
|
; GCN-FASTFMA-NEXT: s_mov_b32 s3, 0xf000
|
|
; GCN-FASTFMA-NEXT: s_mov_b32 s2, -1
|
|
; GCN-FASTFMA-NEXT: buffer_load_dword v0, off, s[0:3], 0 glc
|
|
; GCN-FASTFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FASTFMA-NEXT: buffer_load_dword v1, off, s[0:3], 0 glc
|
|
; GCN-FASTFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FASTFMA-NEXT: buffer_load_dword v2, off, s[0:3], 0 glc
|
|
; GCN-FASTFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FASTFMA-NEXT: buffer_load_dword v3, off, s[0:3], 0 glc
|
|
; GCN-FASTFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FASTFMA-NEXT: buffer_load_dword v4, off, s[0:3], 0 glc
|
|
; GCN-FASTFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FASTFMA-NEXT: v_fma_f32 v2, v3, v4, v2
|
|
; GCN-FASTFMA-NEXT: v_fma_f32 v0, v0, v1, v2
|
|
; GCN-FASTFMA-NEXT: buffer_store_dword v0, off, s[0:3], 0
|
|
; GCN-FASTFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FASTFMA-NEXT: s_endpgm
|
|
;
|
|
; GCN-SLOWFMA-LABEL: fast_add_fmuladd_fmul:
|
|
; GCN-SLOWFMA: ; %bb.0:
|
|
; GCN-SLOWFMA-NEXT: s_mov_b32 s3, 0xf000
|
|
; GCN-SLOWFMA-NEXT: s_mov_b32 s2, -1
|
|
; GCN-SLOWFMA-NEXT: buffer_load_dword v0, off, s[0:3], 0 glc
|
|
; GCN-SLOWFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-SLOWFMA-NEXT: buffer_load_dword v1, off, s[0:3], 0 glc
|
|
; GCN-SLOWFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-SLOWFMA-NEXT: buffer_load_dword v2, off, s[0:3], 0 glc
|
|
; GCN-SLOWFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-SLOWFMA-NEXT: buffer_load_dword v3, off, s[0:3], 0 glc
|
|
; GCN-SLOWFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-SLOWFMA-NEXT: buffer_load_dword v4, off, s[0:3], 0 glc
|
|
; GCN-SLOWFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-SLOWFMA-NEXT: v_mul_f32_e32 v3, v3, v4
|
|
; GCN-SLOWFMA-NEXT: v_mul_f32_e32 v0, v0, v1
|
|
; GCN-SLOWFMA-NEXT: v_add_f32_e32 v0, v0, v3
|
|
; GCN-SLOWFMA-NEXT: v_add_f32_e32 v0, v0, v2
|
|
; GCN-SLOWFMA-NEXT: buffer_store_dword v0, off, s[0:3], 0
|
|
; GCN-SLOWFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-SLOWFMA-NEXT: s_endpgm
|
|
%x = load volatile float, ptr addrspace(1) poison
|
|
%y = load volatile float, ptr addrspace(1) poison
|
|
%z = load volatile float, ptr addrspace(1) poison
|
|
%u = load volatile float, ptr addrspace(1) poison
|
|
%v = load volatile float, ptr addrspace(1) poison
|
|
%mul.u.v = fmul fast float %u, %v
|
|
%fma = call fast float @llvm.fmuladd.f32(float %x, float %y, float %mul.u.v)
|
|
%add = fadd fast float %fma, %z
|
|
store volatile float %add, ptr addrspace(1) poison
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_kernel void @fast_sub_fmuladd_fmul() #0 {
|
|
; GCN-FLUSH-LABEL: fast_sub_fmuladd_fmul:
|
|
; GCN-FLUSH: ; %bb.0:
|
|
; GCN-FLUSH-NEXT: s_mov_b32 s3, 0xf000
|
|
; GCN-FLUSH-NEXT: s_mov_b32 s2, -1
|
|
; GCN-FLUSH-NEXT: buffer_load_dword v0, off, s[0:3], 0 glc
|
|
; GCN-FLUSH-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FLUSH-NEXT: buffer_load_dword v1, off, s[0:3], 0 glc
|
|
; GCN-FLUSH-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FLUSH-NEXT: buffer_load_dword v2, off, s[0:3], 0 glc
|
|
; GCN-FLUSH-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FLUSH-NEXT: buffer_load_dword v3, off, s[0:3], 0 glc
|
|
; GCN-FLUSH-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FLUSH-NEXT: buffer_load_dword v4, off, s[0:3], 0 glc
|
|
; GCN-FLUSH-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FLUSH-NEXT: v_mad_f32 v2, v3, v4, -v2
|
|
; GCN-FLUSH-NEXT: v_mac_f32_e32 v2, v0, v1
|
|
; GCN-FLUSH-NEXT: buffer_store_dword v2, off, s[0:3], 0
|
|
; GCN-FLUSH-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FLUSH-NEXT: s_endpgm
|
|
;
|
|
; GCN-FASTFMA-LABEL: fast_sub_fmuladd_fmul:
|
|
; GCN-FASTFMA: ; %bb.0:
|
|
; GCN-FASTFMA-NEXT: s_mov_b32 s3, 0xf000
|
|
; GCN-FASTFMA-NEXT: s_mov_b32 s2, -1
|
|
; GCN-FASTFMA-NEXT: buffer_load_dword v0, off, s[0:3], 0 glc
|
|
; GCN-FASTFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FASTFMA-NEXT: buffer_load_dword v1, off, s[0:3], 0 glc
|
|
; GCN-FASTFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FASTFMA-NEXT: buffer_load_dword v2, off, s[0:3], 0 glc
|
|
; GCN-FASTFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FASTFMA-NEXT: buffer_load_dword v3, off, s[0:3], 0 glc
|
|
; GCN-FASTFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FASTFMA-NEXT: buffer_load_dword v4, off, s[0:3], 0 glc
|
|
; GCN-FASTFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FASTFMA-NEXT: v_fma_f32 v2, v3, v4, -v2
|
|
; GCN-FASTFMA-NEXT: v_fma_f32 v0, v0, v1, v2
|
|
; GCN-FASTFMA-NEXT: buffer_store_dword v0, off, s[0:3], 0
|
|
; GCN-FASTFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FASTFMA-NEXT: s_endpgm
|
|
;
|
|
; GCN-SLOWFMA-LABEL: fast_sub_fmuladd_fmul:
|
|
; GCN-SLOWFMA: ; %bb.0:
|
|
; GCN-SLOWFMA-NEXT: s_mov_b32 s3, 0xf000
|
|
; GCN-SLOWFMA-NEXT: s_mov_b32 s2, -1
|
|
; GCN-SLOWFMA-NEXT: buffer_load_dword v0, off, s[0:3], 0 glc
|
|
; GCN-SLOWFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-SLOWFMA-NEXT: buffer_load_dword v1, off, s[0:3], 0 glc
|
|
; GCN-SLOWFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-SLOWFMA-NEXT: buffer_load_dword v2, off, s[0:3], 0 glc
|
|
; GCN-SLOWFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-SLOWFMA-NEXT: buffer_load_dword v3, off, s[0:3], 0 glc
|
|
; GCN-SLOWFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-SLOWFMA-NEXT: buffer_load_dword v4, off, s[0:3], 0 glc
|
|
; GCN-SLOWFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-SLOWFMA-NEXT: v_mul_f32_e32 v3, v3, v4
|
|
; GCN-SLOWFMA-NEXT: v_mul_f32_e32 v0, v0, v1
|
|
; GCN-SLOWFMA-NEXT: v_add_f32_e32 v0, v0, v3
|
|
; GCN-SLOWFMA-NEXT: v_sub_f32_e32 v0, v0, v2
|
|
; GCN-SLOWFMA-NEXT: buffer_store_dword v0, off, s[0:3], 0
|
|
; GCN-SLOWFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-SLOWFMA-NEXT: s_endpgm
|
|
%x = load volatile float, ptr addrspace(1) poison
|
|
%y = load volatile float, ptr addrspace(1) poison
|
|
%z = load volatile float, ptr addrspace(1) poison
|
|
%u = load volatile float, ptr addrspace(1) poison
|
|
%v = load volatile float, ptr addrspace(1) poison
|
|
%mul.u.v = fmul fast float %u, %v
|
|
%fma = call fast float @llvm.fmuladd.f32(float %x, float %y, float %mul.u.v)
|
|
%add = fsub fast float %fma, %z
|
|
store volatile float %add, ptr addrspace(1) poison
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_kernel void @fast_add_fmuladd_fmul_multi_use_mul() #0 {
|
|
; GCN-FLUSH-LABEL: fast_add_fmuladd_fmul_multi_use_mul:
|
|
; GCN-FLUSH: ; %bb.0:
|
|
; GCN-FLUSH-NEXT: s_mov_b32 s3, 0xf000
|
|
; GCN-FLUSH-NEXT: s_mov_b32 s2, -1
|
|
; GCN-FLUSH-NEXT: buffer_load_dword v0, off, s[0:3], 0 glc
|
|
; GCN-FLUSH-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FLUSH-NEXT: buffer_load_dword v1, off, s[0:3], 0 glc
|
|
; GCN-FLUSH-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FLUSH-NEXT: buffer_load_dword v2, off, s[0:3], 0 glc
|
|
; GCN-FLUSH-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FLUSH-NEXT: buffer_load_dword v3, off, s[0:3], 0 glc
|
|
; GCN-FLUSH-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FLUSH-NEXT: buffer_load_dword v4, off, s[0:3], 0 glc
|
|
; GCN-FLUSH-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FLUSH-NEXT: v_mul_f32_e32 v3, v3, v4
|
|
; GCN-FLUSH-NEXT: buffer_store_dword v3, off, s[0:3], 0
|
|
; GCN-FLUSH-NEXT: s_waitcnt vmcnt(0) expcnt(0)
|
|
; GCN-FLUSH-NEXT: v_mac_f32_e32 v3, v0, v1
|
|
; GCN-FLUSH-NEXT: v_add_f32_e32 v0, v3, v2
|
|
; GCN-FLUSH-NEXT: buffer_store_dword v0, off, s[0:3], 0
|
|
; GCN-FLUSH-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FLUSH-NEXT: s_endpgm
|
|
;
|
|
; GCN-FASTFMA-LABEL: fast_add_fmuladd_fmul_multi_use_mul:
|
|
; GCN-FASTFMA: ; %bb.0:
|
|
; GCN-FASTFMA-NEXT: s_mov_b32 s3, 0xf000
|
|
; GCN-FASTFMA-NEXT: s_mov_b32 s2, -1
|
|
; GCN-FASTFMA-NEXT: buffer_load_dword v0, off, s[0:3], 0 glc
|
|
; GCN-FASTFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FASTFMA-NEXT: buffer_load_dword v1, off, s[0:3], 0 glc
|
|
; GCN-FASTFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FASTFMA-NEXT: buffer_load_dword v2, off, s[0:3], 0 glc
|
|
; GCN-FASTFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FASTFMA-NEXT: buffer_load_dword v3, off, s[0:3], 0 glc
|
|
; GCN-FASTFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FASTFMA-NEXT: buffer_load_dword v4, off, s[0:3], 0 glc
|
|
; GCN-FASTFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FASTFMA-NEXT: v_mul_f32_e32 v3, v3, v4
|
|
; GCN-FASTFMA-NEXT: buffer_store_dword v3, off, s[0:3], 0
|
|
; GCN-FASTFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FASTFMA-NEXT: v_fma_f32 v0, v0, v1, v3
|
|
; GCN-FASTFMA-NEXT: v_add_f32_e32 v0, v0, v2
|
|
; GCN-FASTFMA-NEXT: buffer_store_dword v0, off, s[0:3], 0
|
|
; GCN-FASTFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FASTFMA-NEXT: s_endpgm
|
|
;
|
|
; GCN-SLOWFMA-LABEL: fast_add_fmuladd_fmul_multi_use_mul:
|
|
; GCN-SLOWFMA: ; %bb.0:
|
|
; GCN-SLOWFMA-NEXT: s_mov_b32 s3, 0xf000
|
|
; GCN-SLOWFMA-NEXT: s_mov_b32 s2, -1
|
|
; GCN-SLOWFMA-NEXT: buffer_load_dword v0, off, s[0:3], 0 glc
|
|
; GCN-SLOWFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-SLOWFMA-NEXT: buffer_load_dword v1, off, s[0:3], 0 glc
|
|
; GCN-SLOWFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-SLOWFMA-NEXT: buffer_load_dword v2, off, s[0:3], 0 glc
|
|
; GCN-SLOWFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-SLOWFMA-NEXT: buffer_load_dword v3, off, s[0:3], 0 glc
|
|
; GCN-SLOWFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-SLOWFMA-NEXT: buffer_load_dword v4, off, s[0:3], 0 glc
|
|
; GCN-SLOWFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-SLOWFMA-NEXT: v_mul_f32_e32 v3, v3, v4
|
|
; GCN-SLOWFMA-NEXT: v_mul_f32_e32 v0, v0, v1
|
|
; GCN-SLOWFMA-NEXT: buffer_store_dword v3, off, s[0:3], 0
|
|
; GCN-SLOWFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-SLOWFMA-NEXT: v_add_f32_e32 v0, v0, v3
|
|
; GCN-SLOWFMA-NEXT: v_add_f32_e32 v0, v0, v2
|
|
; GCN-SLOWFMA-NEXT: buffer_store_dword v0, off, s[0:3], 0
|
|
; GCN-SLOWFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-SLOWFMA-NEXT: s_endpgm
|
|
%x = load volatile float, ptr addrspace(1) poison
|
|
%y = load volatile float, ptr addrspace(1) poison
|
|
%z = load volatile float, ptr addrspace(1) poison
|
|
%u = load volatile float, ptr addrspace(1) poison
|
|
%v = load volatile float, ptr addrspace(1) poison
|
|
%mul.u.v = fmul fast float %u, %v
|
|
store volatile float %mul.u.v, ptr addrspace(1) poison
|
|
%fma = call fast float @llvm.fmuladd.f32(float %x, float %y, float %mul.u.v)
|
|
%add = fadd fast float %fma, %z
|
|
store volatile float %add, ptr addrspace(1) poison
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_kernel void @fast_add_fmuladd_fmul_multi_use_mul_commute() #0 {
|
|
; GCN-FLUSH-LABEL: fast_add_fmuladd_fmul_multi_use_mul_commute:
|
|
; GCN-FLUSH: ; %bb.0:
|
|
; GCN-FLUSH-NEXT: s_mov_b32 s3, 0xf000
|
|
; GCN-FLUSH-NEXT: s_mov_b32 s2, -1
|
|
; GCN-FLUSH-NEXT: buffer_load_dword v0, off, s[0:3], 0 glc
|
|
; GCN-FLUSH-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FLUSH-NEXT: buffer_load_dword v1, off, s[0:3], 0 glc
|
|
; GCN-FLUSH-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FLUSH-NEXT: buffer_load_dword v2, off, s[0:3], 0 glc
|
|
; GCN-FLUSH-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FLUSH-NEXT: buffer_load_dword v3, off, s[0:3], 0 glc
|
|
; GCN-FLUSH-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FLUSH-NEXT: buffer_load_dword v4, off, s[0:3], 0 glc
|
|
; GCN-FLUSH-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FLUSH-NEXT: v_mul_f32_e32 v3, v3, v4
|
|
; GCN-FLUSH-NEXT: buffer_store_dword v3, off, s[0:3], 0
|
|
; GCN-FLUSH-NEXT: s_waitcnt vmcnt(0) expcnt(0)
|
|
; GCN-FLUSH-NEXT: v_mac_f32_e32 v3, v0, v1
|
|
; GCN-FLUSH-NEXT: v_add_f32_e32 v0, v2, v3
|
|
; GCN-FLUSH-NEXT: buffer_store_dword v0, off, s[0:3], 0
|
|
; GCN-FLUSH-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FLUSH-NEXT: s_endpgm
|
|
;
|
|
; GCN-FASTFMA-LABEL: fast_add_fmuladd_fmul_multi_use_mul_commute:
|
|
; GCN-FASTFMA: ; %bb.0:
|
|
; GCN-FASTFMA-NEXT: s_mov_b32 s3, 0xf000
|
|
; GCN-FASTFMA-NEXT: s_mov_b32 s2, -1
|
|
; GCN-FASTFMA-NEXT: buffer_load_dword v0, off, s[0:3], 0 glc
|
|
; GCN-FASTFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FASTFMA-NEXT: buffer_load_dword v1, off, s[0:3], 0 glc
|
|
; GCN-FASTFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FASTFMA-NEXT: buffer_load_dword v2, off, s[0:3], 0 glc
|
|
; GCN-FASTFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FASTFMA-NEXT: buffer_load_dword v3, off, s[0:3], 0 glc
|
|
; GCN-FASTFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FASTFMA-NEXT: buffer_load_dword v4, off, s[0:3], 0 glc
|
|
; GCN-FASTFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FASTFMA-NEXT: v_mul_f32_e32 v3, v3, v4
|
|
; GCN-FASTFMA-NEXT: buffer_store_dword v3, off, s[0:3], 0
|
|
; GCN-FASTFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FASTFMA-NEXT: v_fma_f32 v0, v0, v1, v3
|
|
; GCN-FASTFMA-NEXT: v_add_f32_e32 v0, v2, v0
|
|
; GCN-FASTFMA-NEXT: buffer_store_dword v0, off, s[0:3], 0
|
|
; GCN-FASTFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FASTFMA-NEXT: s_endpgm
|
|
;
|
|
; GCN-SLOWFMA-LABEL: fast_add_fmuladd_fmul_multi_use_mul_commute:
|
|
; GCN-SLOWFMA: ; %bb.0:
|
|
; GCN-SLOWFMA-NEXT: s_mov_b32 s3, 0xf000
|
|
; GCN-SLOWFMA-NEXT: s_mov_b32 s2, -1
|
|
; GCN-SLOWFMA-NEXT: buffer_load_dword v0, off, s[0:3], 0 glc
|
|
; GCN-SLOWFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-SLOWFMA-NEXT: buffer_load_dword v1, off, s[0:3], 0 glc
|
|
; GCN-SLOWFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-SLOWFMA-NEXT: buffer_load_dword v2, off, s[0:3], 0 glc
|
|
; GCN-SLOWFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-SLOWFMA-NEXT: buffer_load_dword v3, off, s[0:3], 0 glc
|
|
; GCN-SLOWFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-SLOWFMA-NEXT: buffer_load_dword v4, off, s[0:3], 0 glc
|
|
; GCN-SLOWFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-SLOWFMA-NEXT: v_mul_f32_e32 v3, v3, v4
|
|
; GCN-SLOWFMA-NEXT: v_mul_f32_e32 v0, v0, v1
|
|
; GCN-SLOWFMA-NEXT: buffer_store_dword v3, off, s[0:3], 0
|
|
; GCN-SLOWFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-SLOWFMA-NEXT: v_add_f32_e32 v0, v0, v3
|
|
; GCN-SLOWFMA-NEXT: v_add_f32_e32 v0, v2, v0
|
|
; GCN-SLOWFMA-NEXT: buffer_store_dword v0, off, s[0:3], 0
|
|
; GCN-SLOWFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-SLOWFMA-NEXT: s_endpgm
|
|
%x = load volatile float, ptr addrspace(1) poison
|
|
%y = load volatile float, ptr addrspace(1) poison
|
|
%z = load volatile float, ptr addrspace(1) poison
|
|
%u = load volatile float, ptr addrspace(1) poison
|
|
%v = load volatile float, ptr addrspace(1) poison
|
|
%mul.u.v = fmul fast float %u, %v
|
|
store volatile float %mul.u.v, ptr addrspace(1) poison
|
|
%fma = call fast float @llvm.fmuladd.f32(float %x, float %y, float %mul.u.v)
|
|
%add = fadd fast float %z, %fma
|
|
store volatile float %add, ptr addrspace(1) poison
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_kernel void @fast_add_fmuladd_fmul_multi_use_fmuladd() #0 {
|
|
; GCN-FLUSH-LABEL: fast_add_fmuladd_fmul_multi_use_fmuladd:
|
|
; GCN-FLUSH: ; %bb.0:
|
|
; GCN-FLUSH-NEXT: s_mov_b32 s3, 0xf000
|
|
; GCN-FLUSH-NEXT: s_mov_b32 s2, -1
|
|
; GCN-FLUSH-NEXT: buffer_load_dword v0, off, s[0:3], 0 glc
|
|
; GCN-FLUSH-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FLUSH-NEXT: buffer_load_dword v1, off, s[0:3], 0 glc
|
|
; GCN-FLUSH-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FLUSH-NEXT: buffer_load_dword v2, off, s[0:3], 0 glc
|
|
; GCN-FLUSH-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FLUSH-NEXT: buffer_load_dword v3, off, s[0:3], 0 glc
|
|
; GCN-FLUSH-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FLUSH-NEXT: buffer_load_dword v4, off, s[0:3], 0 glc
|
|
; GCN-FLUSH-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FLUSH-NEXT: v_mul_f32_e32 v3, v3, v4
|
|
; GCN-FLUSH-NEXT: v_mac_f32_e32 v3, v0, v1
|
|
; GCN-FLUSH-NEXT: buffer_store_dword v3, off, s[0:3], 0
|
|
; GCN-FLUSH-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FLUSH-NEXT: v_add_f32_e32 v0, v3, v2
|
|
; GCN-FLUSH-NEXT: buffer_store_dword v0, off, s[0:3], 0
|
|
; GCN-FLUSH-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FLUSH-NEXT: s_endpgm
|
|
;
|
|
; GCN-FASTFMA-LABEL: fast_add_fmuladd_fmul_multi_use_fmuladd:
|
|
; GCN-FASTFMA: ; %bb.0:
|
|
; GCN-FASTFMA-NEXT: s_mov_b32 s3, 0xf000
|
|
; GCN-FASTFMA-NEXT: s_mov_b32 s2, -1
|
|
; GCN-FASTFMA-NEXT: buffer_load_dword v0, off, s[0:3], 0 glc
|
|
; GCN-FASTFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FASTFMA-NEXT: buffer_load_dword v1, off, s[0:3], 0 glc
|
|
; GCN-FASTFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FASTFMA-NEXT: buffer_load_dword v2, off, s[0:3], 0 glc
|
|
; GCN-FASTFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FASTFMA-NEXT: buffer_load_dword v3, off, s[0:3], 0 glc
|
|
; GCN-FASTFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FASTFMA-NEXT: buffer_load_dword v4, off, s[0:3], 0 glc
|
|
; GCN-FASTFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FASTFMA-NEXT: v_mul_f32_e32 v3, v3, v4
|
|
; GCN-FASTFMA-NEXT: v_fma_f32 v0, v0, v1, v3
|
|
; GCN-FASTFMA-NEXT: buffer_store_dword v0, off, s[0:3], 0
|
|
; GCN-FASTFMA-NEXT: s_waitcnt vmcnt(0) expcnt(0)
|
|
; GCN-FASTFMA-NEXT: v_add_f32_e32 v0, v0, v2
|
|
; GCN-FASTFMA-NEXT: buffer_store_dword v0, off, s[0:3], 0
|
|
; GCN-FASTFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FASTFMA-NEXT: s_endpgm
|
|
;
|
|
; GCN-SLOWFMA-LABEL: fast_add_fmuladd_fmul_multi_use_fmuladd:
|
|
; GCN-SLOWFMA: ; %bb.0:
|
|
; GCN-SLOWFMA-NEXT: s_mov_b32 s3, 0xf000
|
|
; GCN-SLOWFMA-NEXT: s_mov_b32 s2, -1
|
|
; GCN-SLOWFMA-NEXT: buffer_load_dword v0, off, s[0:3], 0 glc
|
|
; GCN-SLOWFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-SLOWFMA-NEXT: buffer_load_dword v1, off, s[0:3], 0 glc
|
|
; GCN-SLOWFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-SLOWFMA-NEXT: buffer_load_dword v2, off, s[0:3], 0 glc
|
|
; GCN-SLOWFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-SLOWFMA-NEXT: buffer_load_dword v3, off, s[0:3], 0 glc
|
|
; GCN-SLOWFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-SLOWFMA-NEXT: buffer_load_dword v4, off, s[0:3], 0 glc
|
|
; GCN-SLOWFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-SLOWFMA-NEXT: v_mul_f32_e32 v3, v3, v4
|
|
; GCN-SLOWFMA-NEXT: v_mul_f32_e32 v0, v0, v1
|
|
; GCN-SLOWFMA-NEXT: v_add_f32_e32 v0, v0, v3
|
|
; GCN-SLOWFMA-NEXT: buffer_store_dword v0, off, s[0:3], 0
|
|
; GCN-SLOWFMA-NEXT: s_waitcnt vmcnt(0) expcnt(0)
|
|
; GCN-SLOWFMA-NEXT: v_add_f32_e32 v0, v0, v2
|
|
; GCN-SLOWFMA-NEXT: buffer_store_dword v0, off, s[0:3], 0
|
|
; GCN-SLOWFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-SLOWFMA-NEXT: s_endpgm
|
|
%x = load volatile float, ptr addrspace(1) poison
|
|
%y = load volatile float, ptr addrspace(1) poison
|
|
%z = load volatile float, ptr addrspace(1) poison
|
|
%u = load volatile float, ptr addrspace(1) poison
|
|
%v = load volatile float, ptr addrspace(1) poison
|
|
%mul.u.v = fmul fast float %u, %v
|
|
%fma = call fast float @llvm.fmuladd.f32(float %x, float %y, float %mul.u.v)
|
|
store volatile float %fma, ptr addrspace(1) poison
|
|
%add = fadd fast float %fma, %z
|
|
store volatile float %add, ptr addrspace(1) poison
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_kernel void @fast_add_fmuladd_fmul_multi_use_fmuladd_commute() #0 {
|
|
; GCN-FLUSH-LABEL: fast_add_fmuladd_fmul_multi_use_fmuladd_commute:
|
|
; GCN-FLUSH: ; %bb.0:
|
|
; GCN-FLUSH-NEXT: s_mov_b32 s3, 0xf000
|
|
; GCN-FLUSH-NEXT: s_mov_b32 s2, -1
|
|
; GCN-FLUSH-NEXT: buffer_load_dword v0, off, s[0:3], 0 glc
|
|
; GCN-FLUSH-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FLUSH-NEXT: buffer_load_dword v1, off, s[0:3], 0 glc
|
|
; GCN-FLUSH-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FLUSH-NEXT: buffer_load_dword v2, off, s[0:3], 0 glc
|
|
; GCN-FLUSH-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FLUSH-NEXT: buffer_load_dword v3, off, s[0:3], 0 glc
|
|
; GCN-FLUSH-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FLUSH-NEXT: buffer_load_dword v4, off, s[0:3], 0 glc
|
|
; GCN-FLUSH-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FLUSH-NEXT: v_mul_f32_e32 v3, v3, v4
|
|
; GCN-FLUSH-NEXT: v_mac_f32_e32 v3, v0, v1
|
|
; GCN-FLUSH-NEXT: buffer_store_dword v3, off, s[0:3], 0
|
|
; GCN-FLUSH-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FLUSH-NEXT: v_add_f32_e32 v0, v2, v3
|
|
; GCN-FLUSH-NEXT: buffer_store_dword v0, off, s[0:3], 0
|
|
; GCN-FLUSH-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FLUSH-NEXT: s_endpgm
|
|
;
|
|
; GCN-FASTFMA-LABEL: fast_add_fmuladd_fmul_multi_use_fmuladd_commute:
|
|
; GCN-FASTFMA: ; %bb.0:
|
|
; GCN-FASTFMA-NEXT: s_mov_b32 s3, 0xf000
|
|
; GCN-FASTFMA-NEXT: s_mov_b32 s2, -1
|
|
; GCN-FASTFMA-NEXT: buffer_load_dword v0, off, s[0:3], 0 glc
|
|
; GCN-FASTFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FASTFMA-NEXT: buffer_load_dword v1, off, s[0:3], 0 glc
|
|
; GCN-FASTFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FASTFMA-NEXT: buffer_load_dword v2, off, s[0:3], 0 glc
|
|
; GCN-FASTFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FASTFMA-NEXT: buffer_load_dword v3, off, s[0:3], 0 glc
|
|
; GCN-FASTFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FASTFMA-NEXT: buffer_load_dword v4, off, s[0:3], 0 glc
|
|
; GCN-FASTFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FASTFMA-NEXT: v_mul_f32_e32 v3, v3, v4
|
|
; GCN-FASTFMA-NEXT: v_fma_f32 v0, v0, v1, v3
|
|
; GCN-FASTFMA-NEXT: buffer_store_dword v0, off, s[0:3], 0
|
|
; GCN-FASTFMA-NEXT: s_waitcnt vmcnt(0) expcnt(0)
|
|
; GCN-FASTFMA-NEXT: v_add_f32_e32 v0, v2, v0
|
|
; GCN-FASTFMA-NEXT: buffer_store_dword v0, off, s[0:3], 0
|
|
; GCN-FASTFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FASTFMA-NEXT: s_endpgm
|
|
;
|
|
; GCN-SLOWFMA-LABEL: fast_add_fmuladd_fmul_multi_use_fmuladd_commute:
|
|
; GCN-SLOWFMA: ; %bb.0:
|
|
; GCN-SLOWFMA-NEXT: s_mov_b32 s3, 0xf000
|
|
; GCN-SLOWFMA-NEXT: s_mov_b32 s2, -1
|
|
; GCN-SLOWFMA-NEXT: buffer_load_dword v0, off, s[0:3], 0 glc
|
|
; GCN-SLOWFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-SLOWFMA-NEXT: buffer_load_dword v1, off, s[0:3], 0 glc
|
|
; GCN-SLOWFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-SLOWFMA-NEXT: buffer_load_dword v2, off, s[0:3], 0 glc
|
|
; GCN-SLOWFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-SLOWFMA-NEXT: buffer_load_dword v3, off, s[0:3], 0 glc
|
|
; GCN-SLOWFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-SLOWFMA-NEXT: buffer_load_dword v4, off, s[0:3], 0 glc
|
|
; GCN-SLOWFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-SLOWFMA-NEXT: v_mul_f32_e32 v3, v3, v4
|
|
; GCN-SLOWFMA-NEXT: v_mul_f32_e32 v0, v0, v1
|
|
; GCN-SLOWFMA-NEXT: v_add_f32_e32 v0, v0, v3
|
|
; GCN-SLOWFMA-NEXT: buffer_store_dword v0, off, s[0:3], 0
|
|
; GCN-SLOWFMA-NEXT: s_waitcnt vmcnt(0) expcnt(0)
|
|
; GCN-SLOWFMA-NEXT: v_add_f32_e32 v0, v2, v0
|
|
; GCN-SLOWFMA-NEXT: buffer_store_dword v0, off, s[0:3], 0
|
|
; GCN-SLOWFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-SLOWFMA-NEXT: s_endpgm
|
|
%x = load volatile float, ptr addrspace(1) poison
|
|
%y = load volatile float, ptr addrspace(1) poison
|
|
%z = load volatile float, ptr addrspace(1) poison
|
|
%u = load volatile float, ptr addrspace(1) poison
|
|
%v = load volatile float, ptr addrspace(1) poison
|
|
%mul.u.v = fmul fast float %u, %v
|
|
%fma = call fast float @llvm.fmuladd.f32(float %x, float %y, float %mul.u.v)
|
|
store volatile float %fma, ptr addrspace(1) poison
|
|
%add = fadd fast float %z, %fma
|
|
store volatile float %add, ptr addrspace(1) poison
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_kernel void @fast_sub_fmuladd_fmul_multi_use_mul() #0 {
|
|
; GCN-FLUSH-LABEL: fast_sub_fmuladd_fmul_multi_use_mul:
|
|
; GCN-FLUSH: ; %bb.0:
|
|
; GCN-FLUSH-NEXT: s_mov_b32 s3, 0xf000
|
|
; GCN-FLUSH-NEXT: s_mov_b32 s2, -1
|
|
; GCN-FLUSH-NEXT: buffer_load_dword v0, off, s[0:3], 0 glc
|
|
; GCN-FLUSH-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FLUSH-NEXT: buffer_load_dword v1, off, s[0:3], 0 glc
|
|
; GCN-FLUSH-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FLUSH-NEXT: buffer_load_dword v2, off, s[0:3], 0 glc
|
|
; GCN-FLUSH-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FLUSH-NEXT: buffer_load_dword v3, off, s[0:3], 0 glc
|
|
; GCN-FLUSH-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FLUSH-NEXT: buffer_load_dword v4, off, s[0:3], 0 glc
|
|
; GCN-FLUSH-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FLUSH-NEXT: v_mul_f32_e32 v3, v3, v4
|
|
; GCN-FLUSH-NEXT: v_mad_f32 v0, v0, v1, v3
|
|
; GCN-FLUSH-NEXT: v_sub_f32_e32 v0, v0, v2
|
|
; GCN-FLUSH-NEXT: buffer_store_dword v3, off, s[0:3], 0
|
|
; GCN-FLUSH-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FLUSH-NEXT: buffer_store_dword v0, off, s[0:3], 0
|
|
; GCN-FLUSH-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FLUSH-NEXT: s_endpgm
|
|
;
|
|
; GCN-FASTFMA-LABEL: fast_sub_fmuladd_fmul_multi_use_mul:
|
|
; GCN-FASTFMA: ; %bb.0:
|
|
; GCN-FASTFMA-NEXT: s_mov_b32 s3, 0xf000
|
|
; GCN-FASTFMA-NEXT: s_mov_b32 s2, -1
|
|
; GCN-FASTFMA-NEXT: buffer_load_dword v0, off, s[0:3], 0 glc
|
|
; GCN-FASTFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FASTFMA-NEXT: buffer_load_dword v1, off, s[0:3], 0 glc
|
|
; GCN-FASTFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FASTFMA-NEXT: buffer_load_dword v2, off, s[0:3], 0 glc
|
|
; GCN-FASTFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FASTFMA-NEXT: buffer_load_dword v3, off, s[0:3], 0 glc
|
|
; GCN-FASTFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FASTFMA-NEXT: buffer_load_dword v4, off, s[0:3], 0 glc
|
|
; GCN-FASTFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FASTFMA-NEXT: v_mul_f32_e32 v3, v3, v4
|
|
; GCN-FASTFMA-NEXT: v_fma_f32 v0, v0, v1, v3
|
|
; GCN-FASTFMA-NEXT: v_sub_f32_e32 v0, v0, v2
|
|
; GCN-FASTFMA-NEXT: buffer_store_dword v3, off, s[0:3], 0
|
|
; GCN-FASTFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FASTFMA-NEXT: buffer_store_dword v0, off, s[0:3], 0
|
|
; GCN-FASTFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FASTFMA-NEXT: s_endpgm
|
|
;
|
|
; GCN-SLOWFMA-LABEL: fast_sub_fmuladd_fmul_multi_use_mul:
|
|
; GCN-SLOWFMA: ; %bb.0:
|
|
; GCN-SLOWFMA-NEXT: s_mov_b32 s3, 0xf000
|
|
; GCN-SLOWFMA-NEXT: s_mov_b32 s2, -1
|
|
; GCN-SLOWFMA-NEXT: buffer_load_dword v0, off, s[0:3], 0 glc
|
|
; GCN-SLOWFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-SLOWFMA-NEXT: buffer_load_dword v1, off, s[0:3], 0 glc
|
|
; GCN-SLOWFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-SLOWFMA-NEXT: buffer_load_dword v2, off, s[0:3], 0 glc
|
|
; GCN-SLOWFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-SLOWFMA-NEXT: buffer_load_dword v3, off, s[0:3], 0 glc
|
|
; GCN-SLOWFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-SLOWFMA-NEXT: buffer_load_dword v4, off, s[0:3], 0 glc
|
|
; GCN-SLOWFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-SLOWFMA-NEXT: v_mul_f32_e32 v3, v3, v4
|
|
; GCN-SLOWFMA-NEXT: v_mul_f32_e32 v0, v0, v1
|
|
; GCN-SLOWFMA-NEXT: v_add_f32_e32 v0, v0, v3
|
|
; GCN-SLOWFMA-NEXT: v_sub_f32_e32 v0, v0, v2
|
|
; GCN-SLOWFMA-NEXT: buffer_store_dword v3, off, s[0:3], 0
|
|
; GCN-SLOWFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-SLOWFMA-NEXT: buffer_store_dword v0, off, s[0:3], 0
|
|
; GCN-SLOWFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-SLOWFMA-NEXT: s_endpgm
|
|
%x = load volatile float, ptr addrspace(1) poison
|
|
%y = load volatile float, ptr addrspace(1) poison
|
|
%z = load volatile float, ptr addrspace(1) poison
|
|
%u = load volatile float, ptr addrspace(1) poison
|
|
%v = load volatile float, ptr addrspace(1) poison
|
|
%mul.u.v = fmul fast float %u, %v
|
|
%fma = call fast float @llvm.fmuladd.f32(float %x, float %y, float %mul.u.v)
|
|
%sub = fsub fast float %fma, %z
|
|
store volatile float %mul.u.v, ptr addrspace(1) poison
|
|
store volatile float %sub, ptr addrspace(1) poison
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_kernel void @fast_sub_fmuladd_fmul_multi_use_fmuladd_lhs() #0 {
|
|
; GCN-FLUSH-LABEL: fast_sub_fmuladd_fmul_multi_use_fmuladd_lhs:
|
|
; GCN-FLUSH: ; %bb.0:
|
|
; GCN-FLUSH-NEXT: s_mov_b32 s3, 0xf000
|
|
; GCN-FLUSH-NEXT: s_mov_b32 s2, -1
|
|
; GCN-FLUSH-NEXT: buffer_load_dword v0, off, s[0:3], 0 glc
|
|
; GCN-FLUSH-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FLUSH-NEXT: buffer_load_dword v1, off, s[0:3], 0 glc
|
|
; GCN-FLUSH-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FLUSH-NEXT: buffer_load_dword v2, off, s[0:3], 0 glc
|
|
; GCN-FLUSH-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FLUSH-NEXT: buffer_load_dword v3, off, s[0:3], 0 glc
|
|
; GCN-FLUSH-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FLUSH-NEXT: buffer_load_dword v4, off, s[0:3], 0 glc
|
|
; GCN-FLUSH-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FLUSH-NEXT: v_mul_f32_e32 v3, v3, v4
|
|
; GCN-FLUSH-NEXT: v_mac_f32_e32 v3, v0, v1
|
|
; GCN-FLUSH-NEXT: v_sub_f32_e32 v0, v3, v2
|
|
; GCN-FLUSH-NEXT: buffer_store_dword v3, off, s[0:3], 0
|
|
; GCN-FLUSH-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FLUSH-NEXT: buffer_store_dword v0, off, s[0:3], 0
|
|
; GCN-FLUSH-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FLUSH-NEXT: s_endpgm
|
|
;
|
|
; GCN-FASTFMA-LABEL: fast_sub_fmuladd_fmul_multi_use_fmuladd_lhs:
|
|
; GCN-FASTFMA: ; %bb.0:
|
|
; GCN-FASTFMA-NEXT: s_mov_b32 s3, 0xf000
|
|
; GCN-FASTFMA-NEXT: s_mov_b32 s2, -1
|
|
; GCN-FASTFMA-NEXT: buffer_load_dword v0, off, s[0:3], 0 glc
|
|
; GCN-FASTFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FASTFMA-NEXT: buffer_load_dword v1, off, s[0:3], 0 glc
|
|
; GCN-FASTFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FASTFMA-NEXT: buffer_load_dword v2, off, s[0:3], 0 glc
|
|
; GCN-FASTFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FASTFMA-NEXT: buffer_load_dword v3, off, s[0:3], 0 glc
|
|
; GCN-FASTFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FASTFMA-NEXT: buffer_load_dword v4, off, s[0:3], 0 glc
|
|
; GCN-FASTFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FASTFMA-NEXT: v_mul_f32_e32 v3, v3, v4
|
|
; GCN-FASTFMA-NEXT: v_fma_f32 v0, v0, v1, v3
|
|
; GCN-FASTFMA-NEXT: v_sub_f32_e32 v1, v0, v2
|
|
; GCN-FASTFMA-NEXT: buffer_store_dword v0, off, s[0:3], 0
|
|
; GCN-FASTFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FASTFMA-NEXT: buffer_store_dword v1, off, s[0:3], 0
|
|
; GCN-FASTFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FASTFMA-NEXT: s_endpgm
|
|
;
|
|
; GCN-SLOWFMA-LABEL: fast_sub_fmuladd_fmul_multi_use_fmuladd_lhs:
|
|
; GCN-SLOWFMA: ; %bb.0:
|
|
; GCN-SLOWFMA-NEXT: s_mov_b32 s3, 0xf000
|
|
; GCN-SLOWFMA-NEXT: s_mov_b32 s2, -1
|
|
; GCN-SLOWFMA-NEXT: buffer_load_dword v0, off, s[0:3], 0 glc
|
|
; GCN-SLOWFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-SLOWFMA-NEXT: buffer_load_dword v1, off, s[0:3], 0 glc
|
|
; GCN-SLOWFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-SLOWFMA-NEXT: buffer_load_dword v2, off, s[0:3], 0 glc
|
|
; GCN-SLOWFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-SLOWFMA-NEXT: buffer_load_dword v3, off, s[0:3], 0 glc
|
|
; GCN-SLOWFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-SLOWFMA-NEXT: buffer_load_dword v4, off, s[0:3], 0 glc
|
|
; GCN-SLOWFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-SLOWFMA-NEXT: v_mul_f32_e32 v3, v3, v4
|
|
; GCN-SLOWFMA-NEXT: v_mul_f32_e32 v0, v0, v1
|
|
; GCN-SLOWFMA-NEXT: v_add_f32_e32 v0, v0, v3
|
|
; GCN-SLOWFMA-NEXT: v_sub_f32_e32 v1, v0, v2
|
|
; GCN-SLOWFMA-NEXT: buffer_store_dword v0, off, s[0:3], 0
|
|
; GCN-SLOWFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-SLOWFMA-NEXT: buffer_store_dword v1, off, s[0:3], 0
|
|
; GCN-SLOWFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-SLOWFMA-NEXT: s_endpgm
|
|
%x = load volatile float, ptr addrspace(1) poison
|
|
%y = load volatile float, ptr addrspace(1) poison
|
|
%z = load volatile float, ptr addrspace(1) poison
|
|
%u = load volatile float, ptr addrspace(1) poison
|
|
%v = load volatile float, ptr addrspace(1) poison
|
|
%mul.u.v = fmul fast float %u, %v
|
|
%fma = call fast float @llvm.fmuladd.f32(float %x, float %y, float %mul.u.v)
|
|
%add = fsub fast float %fma, %z
|
|
store volatile float %fma, ptr addrspace(1) poison
|
|
store volatile float %add, ptr addrspace(1) poison
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_kernel void @fast_sub_fmuladd_fmul_multi_use_fmuladd_rhs() #0 {
|
|
; GCN-FLUSH-LABEL: fast_sub_fmuladd_fmul_multi_use_fmuladd_rhs:
|
|
; GCN-FLUSH: ; %bb.0:
|
|
; GCN-FLUSH-NEXT: s_mov_b32 s3, 0xf000
|
|
; GCN-FLUSH-NEXT: s_mov_b32 s2, -1
|
|
; GCN-FLUSH-NEXT: buffer_load_dword v0, off, s[0:3], 0 glc
|
|
; GCN-FLUSH-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FLUSH-NEXT: buffer_load_dword v1, off, s[0:3], 0 glc
|
|
; GCN-FLUSH-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FLUSH-NEXT: buffer_load_dword v2, off, s[0:3], 0 glc
|
|
; GCN-FLUSH-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FLUSH-NEXT: buffer_load_dword v3, off, s[0:3], 0 glc
|
|
; GCN-FLUSH-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FLUSH-NEXT: buffer_load_dword v4, off, s[0:3], 0 glc
|
|
; GCN-FLUSH-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FLUSH-NEXT: v_mul_f32_e32 v3, v3, v4
|
|
; GCN-FLUSH-NEXT: v_mac_f32_e32 v3, v0, v1
|
|
; GCN-FLUSH-NEXT: v_sub_f32_e32 v0, v2, v3
|
|
; GCN-FLUSH-NEXT: buffer_store_dword v3, off, s[0:3], 0
|
|
; GCN-FLUSH-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FLUSH-NEXT: buffer_store_dword v0, off, s[0:3], 0
|
|
; GCN-FLUSH-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FLUSH-NEXT: s_endpgm
|
|
;
|
|
; GCN-FASTFMA-LABEL: fast_sub_fmuladd_fmul_multi_use_fmuladd_rhs:
|
|
; GCN-FASTFMA: ; %bb.0:
|
|
; GCN-FASTFMA-NEXT: s_mov_b32 s3, 0xf000
|
|
; GCN-FASTFMA-NEXT: s_mov_b32 s2, -1
|
|
; GCN-FASTFMA-NEXT: buffer_load_dword v0, off, s[0:3], 0 glc
|
|
; GCN-FASTFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FASTFMA-NEXT: buffer_load_dword v1, off, s[0:3], 0 glc
|
|
; GCN-FASTFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FASTFMA-NEXT: buffer_load_dword v2, off, s[0:3], 0 glc
|
|
; GCN-FASTFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FASTFMA-NEXT: buffer_load_dword v3, off, s[0:3], 0 glc
|
|
; GCN-FASTFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FASTFMA-NEXT: buffer_load_dword v4, off, s[0:3], 0 glc
|
|
; GCN-FASTFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FASTFMA-NEXT: v_mul_f32_e32 v3, v3, v4
|
|
; GCN-FASTFMA-NEXT: v_fma_f32 v0, v0, v1, v3
|
|
; GCN-FASTFMA-NEXT: v_sub_f32_e32 v1, v2, v0
|
|
; GCN-FASTFMA-NEXT: buffer_store_dword v0, off, s[0:3], 0
|
|
; GCN-FASTFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FASTFMA-NEXT: buffer_store_dword v1, off, s[0:3], 0
|
|
; GCN-FASTFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FASTFMA-NEXT: s_endpgm
|
|
;
|
|
; GCN-SLOWFMA-LABEL: fast_sub_fmuladd_fmul_multi_use_fmuladd_rhs:
|
|
; GCN-SLOWFMA: ; %bb.0:
|
|
; GCN-SLOWFMA-NEXT: s_mov_b32 s3, 0xf000
|
|
; GCN-SLOWFMA-NEXT: s_mov_b32 s2, -1
|
|
; GCN-SLOWFMA-NEXT: buffer_load_dword v0, off, s[0:3], 0 glc
|
|
; GCN-SLOWFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-SLOWFMA-NEXT: buffer_load_dword v1, off, s[0:3], 0 glc
|
|
; GCN-SLOWFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-SLOWFMA-NEXT: buffer_load_dword v2, off, s[0:3], 0 glc
|
|
; GCN-SLOWFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-SLOWFMA-NEXT: buffer_load_dword v3, off, s[0:3], 0 glc
|
|
; GCN-SLOWFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-SLOWFMA-NEXT: buffer_load_dword v4, off, s[0:3], 0 glc
|
|
; GCN-SLOWFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-SLOWFMA-NEXT: v_mul_f32_e32 v3, v3, v4
|
|
; GCN-SLOWFMA-NEXT: v_mul_f32_e32 v0, v0, v1
|
|
; GCN-SLOWFMA-NEXT: v_add_f32_e32 v0, v0, v3
|
|
; GCN-SLOWFMA-NEXT: v_sub_f32_e32 v1, v2, v0
|
|
; GCN-SLOWFMA-NEXT: buffer_store_dword v0, off, s[0:3], 0
|
|
; GCN-SLOWFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-SLOWFMA-NEXT: buffer_store_dword v1, off, s[0:3], 0
|
|
; GCN-SLOWFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-SLOWFMA-NEXT: s_endpgm
|
|
%x = load volatile float, ptr addrspace(1) poison
|
|
%y = load volatile float, ptr addrspace(1) poison
|
|
%z = load volatile float, ptr addrspace(1) poison
|
|
%u = load volatile float, ptr addrspace(1) poison
|
|
%v = load volatile float, ptr addrspace(1) poison
|
|
%mul.u.v = fmul fast float %u, %v
|
|
%fma = call fast float @llvm.fmuladd.f32(float %x, float %y, float %mul.u.v)
|
|
%add = fsub fast float %z, %fma
|
|
store volatile float %fma, ptr addrspace(1) poison
|
|
store volatile float %add, ptr addrspace(1) poison
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_kernel void @fast_sub_fmuladd_fpext_fmul_multi_use_fmuladd_lhs() #0 {
|
|
; GCN-FLUSH-LABEL: fast_sub_fmuladd_fpext_fmul_multi_use_fmuladd_lhs:
|
|
; GCN-FLUSH: ; %bb.0:
|
|
; GCN-FLUSH-NEXT: s_mov_b32 s3, 0xf000
|
|
; GCN-FLUSH-NEXT: s_mov_b32 s2, -1
|
|
; GCN-FLUSH-NEXT: buffer_load_dword v0, off, s[0:3], 0 glc
|
|
; GCN-FLUSH-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FLUSH-NEXT: buffer_load_dword v1, off, s[0:3], 0 glc
|
|
; GCN-FLUSH-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FLUSH-NEXT: buffer_load_dword v2, off, s[0:3], 0 glc
|
|
; GCN-FLUSH-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FLUSH-NEXT: buffer_load_ushort v3, off, s[0:3], 0 glc
|
|
; GCN-FLUSH-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FLUSH-NEXT: buffer_load_ushort v4, off, s[0:3], 0 glc
|
|
; GCN-FLUSH-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FLUSH-NEXT: v_cvt_f32_f16_e32 v3, v3
|
|
; GCN-FLUSH-NEXT: v_cvt_f32_f16_e32 v4, v4
|
|
; GCN-FLUSH-NEXT: v_mul_f32_e32 v3, v3, v4
|
|
; GCN-FLUSH-NEXT: v_cvt_f16_f32_e32 v3, v3
|
|
; GCN-FLUSH-NEXT: v_cvt_f32_f16_e32 v3, v3
|
|
; GCN-FLUSH-NEXT: v_mac_f32_e32 v3, v0, v1
|
|
; GCN-FLUSH-NEXT: v_sub_f32_e32 v0, v3, v2
|
|
; GCN-FLUSH-NEXT: buffer_store_dword v3, off, s[0:3], 0
|
|
; GCN-FLUSH-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FLUSH-NEXT: buffer_store_dword v0, off, s[0:3], 0
|
|
; GCN-FLUSH-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FLUSH-NEXT: s_endpgm
|
|
;
|
|
; GCN-FASTFMA-LABEL: fast_sub_fmuladd_fpext_fmul_multi_use_fmuladd_lhs:
|
|
; GCN-FASTFMA: ; %bb.0:
|
|
; GCN-FASTFMA-NEXT: s_mov_b32 s3, 0xf000
|
|
; GCN-FASTFMA-NEXT: s_mov_b32 s2, -1
|
|
; GCN-FASTFMA-NEXT: buffer_load_dword v0, off, s[0:3], 0 glc
|
|
; GCN-FASTFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FASTFMA-NEXT: buffer_load_dword v1, off, s[0:3], 0 glc
|
|
; GCN-FASTFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FASTFMA-NEXT: buffer_load_dword v2, off, s[0:3], 0 glc
|
|
; GCN-FASTFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FASTFMA-NEXT: buffer_load_ushort v3, off, s[0:3], 0 glc
|
|
; GCN-FASTFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FASTFMA-NEXT: buffer_load_ushort v4, off, s[0:3], 0 glc
|
|
; GCN-FASTFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FASTFMA-NEXT: v_cvt_f32_f16_e32 v3, v3
|
|
; GCN-FASTFMA-NEXT: v_cvt_f32_f16_e32 v4, v4
|
|
; GCN-FASTFMA-NEXT: v_mul_f32_e32 v3, v3, v4
|
|
; GCN-FASTFMA-NEXT: v_cvt_f16_f32_e32 v3, v3
|
|
; GCN-FASTFMA-NEXT: v_cvt_f32_f16_e32 v3, v3
|
|
; GCN-FASTFMA-NEXT: v_fma_f32 v0, v0, v1, v3
|
|
; GCN-FASTFMA-NEXT: v_sub_f32_e32 v1, v0, v2
|
|
; GCN-FASTFMA-NEXT: buffer_store_dword v0, off, s[0:3], 0
|
|
; GCN-FASTFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FASTFMA-NEXT: buffer_store_dword v1, off, s[0:3], 0
|
|
; GCN-FASTFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FASTFMA-NEXT: s_endpgm
|
|
;
|
|
; GCN-SLOWFMA-LABEL: fast_sub_fmuladd_fpext_fmul_multi_use_fmuladd_lhs:
|
|
; GCN-SLOWFMA: ; %bb.0:
|
|
; GCN-SLOWFMA-NEXT: s_mov_b32 s3, 0xf000
|
|
; GCN-SLOWFMA-NEXT: s_mov_b32 s2, -1
|
|
; GCN-SLOWFMA-NEXT: buffer_load_dword v0, off, s[0:3], 0 glc
|
|
; GCN-SLOWFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-SLOWFMA-NEXT: buffer_load_dword v1, off, s[0:3], 0 glc
|
|
; GCN-SLOWFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-SLOWFMA-NEXT: buffer_load_dword v2, off, s[0:3], 0 glc
|
|
; GCN-SLOWFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-SLOWFMA-NEXT: buffer_load_ushort v3, off, s[0:3], 0 glc
|
|
; GCN-SLOWFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-SLOWFMA-NEXT: buffer_load_ushort v4, off, s[0:3], 0 glc
|
|
; GCN-SLOWFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-SLOWFMA-NEXT: v_cvt_f32_f16_e32 v3, v3
|
|
; GCN-SLOWFMA-NEXT: v_cvt_f32_f16_e32 v4, v4
|
|
; GCN-SLOWFMA-NEXT: v_mul_f32_e32 v0, v0, v1
|
|
; GCN-SLOWFMA-NEXT: v_mul_f32_e32 v1, v3, v4
|
|
; GCN-SLOWFMA-NEXT: v_cvt_f16_f32_e32 v1, v1
|
|
; GCN-SLOWFMA-NEXT: v_cvt_f32_f16_e32 v1, v1
|
|
; GCN-SLOWFMA-NEXT: v_add_f32_e32 v0, v0, v1
|
|
; GCN-SLOWFMA-NEXT: v_sub_f32_e32 v1, v0, v2
|
|
; GCN-SLOWFMA-NEXT: buffer_store_dword v0, off, s[0:3], 0
|
|
; GCN-SLOWFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-SLOWFMA-NEXT: buffer_store_dword v1, off, s[0:3], 0
|
|
; GCN-SLOWFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-SLOWFMA-NEXT: s_endpgm
|
|
%x = load volatile float, ptr addrspace(1) poison
|
|
%y = load volatile float, ptr addrspace(1) poison
|
|
%z = load volatile float, ptr addrspace(1) poison
|
|
%u = load volatile half, ptr addrspace(1) poison
|
|
%v = load volatile half, ptr addrspace(1) poison
|
|
%mul.u.v.half = fmul fast half %u, %v
|
|
%mul.u.v = fpext half %mul.u.v.half to float
|
|
%fma = call fast float @llvm.fmuladd.f32(float %x, float %y, float %mul.u.v)
|
|
%add = fsub fast float %fma, %z
|
|
store volatile float %fma, ptr addrspace(1) poison
|
|
store volatile float %add, ptr addrspace(1) poison
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_kernel void @fast_sub_fmuladd_fpext_fmul_multi_use_fmuladd_rhs() #0 {
|
|
; GCN-FLUSH-LABEL: fast_sub_fmuladd_fpext_fmul_multi_use_fmuladd_rhs:
|
|
; GCN-FLUSH: ; %bb.0:
|
|
; GCN-FLUSH-NEXT: s_mov_b32 s3, 0xf000
|
|
; GCN-FLUSH-NEXT: s_mov_b32 s2, -1
|
|
; GCN-FLUSH-NEXT: buffer_load_dword v0, off, s[0:3], 0 glc
|
|
; GCN-FLUSH-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FLUSH-NEXT: buffer_load_dword v1, off, s[0:3], 0 glc
|
|
; GCN-FLUSH-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FLUSH-NEXT: buffer_load_dword v2, off, s[0:3], 0 glc
|
|
; GCN-FLUSH-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FLUSH-NEXT: buffer_load_ushort v3, off, s[0:3], 0 glc
|
|
; GCN-FLUSH-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FLUSH-NEXT: buffer_load_ushort v4, off, s[0:3], 0 glc
|
|
; GCN-FLUSH-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FLUSH-NEXT: v_cvt_f32_f16_e32 v3, v3
|
|
; GCN-FLUSH-NEXT: v_cvt_f32_f16_e32 v4, v4
|
|
; GCN-FLUSH-NEXT: v_mul_f32_e32 v3, v3, v4
|
|
; GCN-FLUSH-NEXT: v_cvt_f16_f32_e32 v3, v3
|
|
; GCN-FLUSH-NEXT: v_cvt_f32_f16_e32 v3, v3
|
|
; GCN-FLUSH-NEXT: v_mac_f32_e32 v3, v0, v1
|
|
; GCN-FLUSH-NEXT: v_sub_f32_e32 v0, v2, v3
|
|
; GCN-FLUSH-NEXT: buffer_store_dword v3, off, s[0:3], 0
|
|
; GCN-FLUSH-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FLUSH-NEXT: buffer_store_dword v0, off, s[0:3], 0
|
|
; GCN-FLUSH-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FLUSH-NEXT: s_endpgm
|
|
;
|
|
; GCN-FASTFMA-LABEL: fast_sub_fmuladd_fpext_fmul_multi_use_fmuladd_rhs:
|
|
; GCN-FASTFMA: ; %bb.0:
|
|
; GCN-FASTFMA-NEXT: s_mov_b32 s3, 0xf000
|
|
; GCN-FASTFMA-NEXT: s_mov_b32 s2, -1
|
|
; GCN-FASTFMA-NEXT: buffer_load_dword v0, off, s[0:3], 0 glc
|
|
; GCN-FASTFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FASTFMA-NEXT: buffer_load_dword v1, off, s[0:3], 0 glc
|
|
; GCN-FASTFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FASTFMA-NEXT: buffer_load_dword v2, off, s[0:3], 0 glc
|
|
; GCN-FASTFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FASTFMA-NEXT: buffer_load_ushort v3, off, s[0:3], 0 glc
|
|
; GCN-FASTFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FASTFMA-NEXT: buffer_load_ushort v4, off, s[0:3], 0 glc
|
|
; GCN-FASTFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FASTFMA-NEXT: v_cvt_f32_f16_e32 v3, v3
|
|
; GCN-FASTFMA-NEXT: v_cvt_f32_f16_e32 v4, v4
|
|
; GCN-FASTFMA-NEXT: v_mul_f32_e32 v3, v3, v4
|
|
; GCN-FASTFMA-NEXT: v_cvt_f16_f32_e32 v3, v3
|
|
; GCN-FASTFMA-NEXT: v_cvt_f32_f16_e32 v3, v3
|
|
; GCN-FASTFMA-NEXT: v_fma_f32 v0, v0, v1, v3
|
|
; GCN-FASTFMA-NEXT: v_sub_f32_e32 v1, v2, v0
|
|
; GCN-FASTFMA-NEXT: buffer_store_dword v0, off, s[0:3], 0
|
|
; GCN-FASTFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FASTFMA-NEXT: buffer_store_dword v1, off, s[0:3], 0
|
|
; GCN-FASTFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-FASTFMA-NEXT: s_endpgm
|
|
;
|
|
; GCN-SLOWFMA-LABEL: fast_sub_fmuladd_fpext_fmul_multi_use_fmuladd_rhs:
|
|
; GCN-SLOWFMA: ; %bb.0:
|
|
; GCN-SLOWFMA-NEXT: s_mov_b32 s3, 0xf000
|
|
; GCN-SLOWFMA-NEXT: s_mov_b32 s2, -1
|
|
; GCN-SLOWFMA-NEXT: buffer_load_dword v0, off, s[0:3], 0 glc
|
|
; GCN-SLOWFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-SLOWFMA-NEXT: buffer_load_dword v1, off, s[0:3], 0 glc
|
|
; GCN-SLOWFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-SLOWFMA-NEXT: buffer_load_dword v2, off, s[0:3], 0 glc
|
|
; GCN-SLOWFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-SLOWFMA-NEXT: buffer_load_ushort v3, off, s[0:3], 0 glc
|
|
; GCN-SLOWFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-SLOWFMA-NEXT: buffer_load_ushort v4, off, s[0:3], 0 glc
|
|
; GCN-SLOWFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-SLOWFMA-NEXT: v_cvt_f32_f16_e32 v3, v3
|
|
; GCN-SLOWFMA-NEXT: v_cvt_f32_f16_e32 v4, v4
|
|
; GCN-SLOWFMA-NEXT: v_mul_f32_e32 v0, v0, v1
|
|
; GCN-SLOWFMA-NEXT: v_mul_f32_e32 v1, v3, v4
|
|
; GCN-SLOWFMA-NEXT: v_cvt_f16_f32_e32 v1, v1
|
|
; GCN-SLOWFMA-NEXT: v_cvt_f32_f16_e32 v1, v1
|
|
; GCN-SLOWFMA-NEXT: v_add_f32_e32 v0, v0, v1
|
|
; GCN-SLOWFMA-NEXT: v_sub_f32_e32 v1, v2, v0
|
|
; GCN-SLOWFMA-NEXT: buffer_store_dword v0, off, s[0:3], 0
|
|
; GCN-SLOWFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-SLOWFMA-NEXT: buffer_store_dword v1, off, s[0:3], 0
|
|
; GCN-SLOWFMA-NEXT: s_waitcnt vmcnt(0)
|
|
; GCN-SLOWFMA-NEXT: s_endpgm
|
|
%x = load volatile float, ptr addrspace(1) poison
|
|
%y = load volatile float, ptr addrspace(1) poison
|
|
%z = load volatile float, ptr addrspace(1) poison
|
|
%u = load volatile half, ptr addrspace(1) poison
|
|
%v = load volatile half, ptr addrspace(1) poison
|
|
%mul.u.v.half = fmul fast half %u, %v
|
|
%mul.u.v = fpext half %mul.u.v.half to float
|
|
%fma = call fast float @llvm.fmuladd.f32(float %x, float %y, float %mul.u.v)
|
|
%add = fsub fast float %z, %fma
|
|
store volatile float %fma, ptr addrspace(1) poison
|
|
store volatile float %add, ptr addrspace(1) poison
|
|
ret void
|
|
}
|
|
|
|
declare float @llvm.fma.f32(float, float, float) #1
|
|
declare float @llvm.fmuladd.f32(float, float, float) #1
|
|
|
|
attributes #0 = { nounwind }
|
|
attributes #1 = { nounwind readnone }
|
|
;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
|
|
; GCN: {{.*}}
|