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
1267 lines
58 KiB
LLVM
1267 lines
58 KiB
LLVM
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 6
|
|
; RUN: llc -global-isel -new-reg-bank-select -mtriple=amdgcn-mesa-mesa3d -mcpu=fiji < %s | FileCheck -check-prefixes=GFX8 %s
|
|
; RUN: llc -global-isel -new-reg-bank-select -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx900 < %s | FileCheck -check-prefixes=GFX9,GFX900 %s
|
|
; RUN: llc -global-isel -new-reg-bank-select -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx942 < %s | FileCheck -check-prefixes=GFX9,GFX942 %s
|
|
; RUN: llc -global-isel -new-reg-bank-select -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1100 < %s | FileCheck -check-prefixes=GFX11 %s
|
|
; RUN: llc -global-isel -new-reg-bank-select -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1200 < %s | FileCheck -check-prefixes=GFX12 %s
|
|
|
|
define void @v_constained_fma_f64_fpexcept_strict_uni(double inreg %x, double inreg %y, double inreg %z, ptr addrspace(1) %out) #0 {
|
|
; GFX8-LABEL: v_constained_fma_f64_fpexcept_strict_uni:
|
|
; GFX8: ; %bb.0:
|
|
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX8-NEXT: v_mov_b32_e32 v2, s18
|
|
; GFX8-NEXT: v_mov_b32_e32 v4, s20
|
|
; GFX8-NEXT: v_mov_b32_e32 v3, s19
|
|
; GFX8-NEXT: v_mov_b32_e32 v5, s21
|
|
; GFX8-NEXT: v_fma_f64 v[2:3], s[16:17], v[2:3], v[4:5]
|
|
; GFX8-NEXT: flat_store_dwordx2 v[0:1], v[2:3]
|
|
; GFX8-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX8-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX900-LABEL: v_constained_fma_f64_fpexcept_strict_uni:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, s18
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, s20
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, s19
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, s21
|
|
; GFX900-NEXT: v_fma_f64 v[2:3], s[16:17], v[2:3], v[4:5]
|
|
; GFX900-NEXT: global_store_dwordx2 v[0:1], v[2:3], off
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_constained_fma_f64_fpexcept_strict_uni:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b64_e32 v[2:3], s[2:3]
|
|
; GFX942-NEXT: v_mov_b64_e32 v[4:5], s[16:17]
|
|
; GFX942-NEXT: v_fma_f64 v[2:3], s[0:1], v[2:3], v[4:5]
|
|
; GFX942-NEXT: global_store_dwordx2 v[0:1], v[2:3], off
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX11-LABEL: v_constained_fma_f64_fpexcept_strict_uni:
|
|
; GFX11: ; %bb.0:
|
|
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX11-NEXT: v_dual_mov_b32 v2, s16 :: v_dual_mov_b32 v3, s17
|
|
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
|
; GFX11-NEXT: v_fma_f64 v[2:3], s[0:1], s[2:3], v[2:3]
|
|
; GFX11-NEXT: global_store_b64 v[0:1], v[2:3], off
|
|
; GFX11-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX12-LABEL: v_constained_fma_f64_fpexcept_strict_uni:
|
|
; GFX12: ; %bb.0:
|
|
; GFX12-NEXT: s_wait_loadcnt_dscnt 0x0
|
|
; GFX12-NEXT: s_wait_expcnt 0x0
|
|
; GFX12-NEXT: s_wait_samplecnt 0x0
|
|
; GFX12-NEXT: s_wait_bvhcnt 0x0
|
|
; GFX12-NEXT: s_wait_kmcnt 0x0
|
|
; GFX12-NEXT: v_dual_mov_b32 v2, s16 :: v_dual_mov_b32 v3, s17
|
|
; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
|
; GFX12-NEXT: v_fma_f64 v[2:3], s[0:1], s[2:3], v[2:3]
|
|
; GFX12-NEXT: global_store_b64 v[0:1], v[2:3], off
|
|
; GFX12-NEXT: s_setpc_b64 s[30:31]
|
|
%val = call double @llvm.experimental.constrained.fma.f64(double %x, double %y, double %z, metadata !"round.tonearest", metadata !"fpexcept.strict")
|
|
store double %val, ptr addrspace(1) %out
|
|
ret void
|
|
}
|
|
|
|
define void @v_constained_fma_f64_fpexcept_strict_div(double %x, double %y, double %z, ptr addrspace(1) %out) #0 {
|
|
; GFX8-LABEL: v_constained_fma_f64_fpexcept_strict_div:
|
|
; GFX8: ; %bb.0:
|
|
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX8-NEXT: v_fma_f64 v[0:1], v[0:1], v[2:3], v[4:5]
|
|
; GFX8-NEXT: flat_store_dwordx2 v[6:7], v[0:1]
|
|
; GFX8-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX8-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX9-LABEL: v_constained_fma_f64_fpexcept_strict_div:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: v_fma_f64 v[0:1], v[0:1], v[2:3], v[4:5]
|
|
; GFX9-NEXT: global_store_dwordx2 v[6:7], v[0:1], off
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX11-LABEL: v_constained_fma_f64_fpexcept_strict_div:
|
|
; GFX11: ; %bb.0:
|
|
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX11-NEXT: v_fma_f64 v[0:1], v[0:1], v[2:3], v[4:5]
|
|
; GFX11-NEXT: global_store_b64 v[6:7], v[0:1], off
|
|
; GFX11-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX12-LABEL: v_constained_fma_f64_fpexcept_strict_div:
|
|
; GFX12: ; %bb.0:
|
|
; GFX12-NEXT: s_wait_loadcnt_dscnt 0x0
|
|
; GFX12-NEXT: s_wait_expcnt 0x0
|
|
; GFX12-NEXT: s_wait_samplecnt 0x0
|
|
; GFX12-NEXT: s_wait_bvhcnt 0x0
|
|
; GFX12-NEXT: s_wait_kmcnt 0x0
|
|
; GFX12-NEXT: v_fma_f64 v[0:1], v[0:1], v[2:3], v[4:5]
|
|
; GFX12-NEXT: global_store_b64 v[6:7], v[0:1], off
|
|
; GFX12-NEXT: s_setpc_b64 s[30:31]
|
|
%val = call double @llvm.experimental.constrained.fma.f64(double %x, double %y, double %z, metadata !"round.tonearest", metadata !"fpexcept.strict")
|
|
store double %val, ptr addrspace(1) %out
|
|
ret void
|
|
}
|
|
|
|
define void @v_constained_fma_v2f64_fpexcept_strict_uni(<2 x double> inreg %x, <2 x double> inreg %y, <2 x double> inreg %z, ptr addrspace(1) %out) #0 {
|
|
; GFX8-LABEL: v_constained_fma_v2f64_fpexcept_strict_uni:
|
|
; GFX8: ; %bb.0:
|
|
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX8-NEXT: v_mov_b32_e32 v2, s20
|
|
; GFX8-NEXT: v_mov_b32_e32 v4, s24
|
|
; GFX8-NEXT: v_mov_b32_e32 v3, s21
|
|
; GFX8-NEXT: v_mov_b32_e32 v5, s25
|
|
; GFX8-NEXT: v_fma_f64 v[2:3], s[16:17], v[2:3], v[4:5]
|
|
; GFX8-NEXT: v_mov_b32_e32 v4, s22
|
|
; GFX8-NEXT: v_mov_b32_e32 v6, s26
|
|
; GFX8-NEXT: v_mov_b32_e32 v5, s23
|
|
; GFX8-NEXT: v_mov_b32_e32 v7, s27
|
|
; GFX8-NEXT: v_fma_f64 v[4:5], s[18:19], v[4:5], v[6:7]
|
|
; GFX8-NEXT: v_readfirstlane_b32 s4, v2
|
|
; GFX8-NEXT: v_readfirstlane_b32 s5, v3
|
|
; GFX8-NEXT: v_readfirstlane_b32 s6, v4
|
|
; GFX8-NEXT: v_readfirstlane_b32 s7, v5
|
|
; GFX8-NEXT: v_mov_b32_e32 v2, s4
|
|
; GFX8-NEXT: v_mov_b32_e32 v3, s5
|
|
; GFX8-NEXT: v_mov_b32_e32 v4, s6
|
|
; GFX8-NEXT: v_mov_b32_e32 v5, s7
|
|
; GFX8-NEXT: flat_store_dwordx4 v[0:1], v[2:5]
|
|
; GFX8-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX8-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX900-LABEL: v_constained_fma_v2f64_fpexcept_strict_uni:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, s20
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, s24
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, s21
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, s25
|
|
; GFX900-NEXT: v_fma_f64 v[2:3], s[16:17], v[2:3], v[4:5]
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, s22
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, s26
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, s23
|
|
; GFX900-NEXT: v_mov_b32_e32 v7, s27
|
|
; GFX900-NEXT: v_fma_f64 v[4:5], s[18:19], v[4:5], v[6:7]
|
|
; GFX900-NEXT: v_readfirstlane_b32 s4, v2
|
|
; GFX900-NEXT: v_readfirstlane_b32 s5, v3
|
|
; GFX900-NEXT: v_readfirstlane_b32 s6, v4
|
|
; GFX900-NEXT: v_readfirstlane_b32 s7, v5
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, s4
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, s5
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, s6
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, s7
|
|
; GFX900-NEXT: global_store_dwordx4 v[0:1], v[2:5], off
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_constained_fma_v2f64_fpexcept_strict_uni:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b64_e32 v[2:3], s[16:17]
|
|
; GFX942-NEXT: v_mov_b64_e32 v[4:5], s[20:21]
|
|
; GFX942-NEXT: v_fma_f64 v[2:3], s[0:1], v[2:3], v[4:5]
|
|
; GFX942-NEXT: v_mov_b64_e32 v[4:5], s[22:23]
|
|
; GFX942-NEXT: v_readfirstlane_b32 s0, v2
|
|
; GFX942-NEXT: v_readfirstlane_b32 s1, v3
|
|
; GFX942-NEXT: v_mov_b64_e32 v[2:3], s[18:19]
|
|
; GFX942-NEXT: v_fma_f64 v[2:3], s[2:3], v[2:3], v[4:5]
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_readfirstlane_b32 s2, v2
|
|
; GFX942-NEXT: v_readfirstlane_b32 s3, v3
|
|
; GFX942-NEXT: s_nop 1
|
|
; GFX942-NEXT: v_mov_b64_e32 v[4:5], s[2:3]
|
|
; GFX942-NEXT: v_mov_b64_e32 v[2:3], s[0:1]
|
|
; GFX942-NEXT: global_store_dwordx4 v[0:1], v[2:5], off
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX11-LABEL: v_constained_fma_v2f64_fpexcept_strict_uni:
|
|
; GFX11: ; %bb.0:
|
|
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX11-NEXT: v_dual_mov_b32 v4, s22 :: v_dual_mov_b32 v5, s23
|
|
; GFX11-NEXT: v_dual_mov_b32 v2, s20 :: v_dual_mov_b32 v3, s21
|
|
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
|
|
; GFX11-NEXT: v_fma_f64 v[4:5], s[2:3], s[18:19], v[4:5]
|
|
; GFX11-NEXT: v_fma_f64 v[2:3], s[0:1], s[16:17], v[2:3]
|
|
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3)
|
|
; GFX11-NEXT: v_readfirstlane_b32 s3, v5
|
|
; GFX11-NEXT: v_readfirstlane_b32 s2, v4
|
|
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_4)
|
|
; GFX11-NEXT: v_readfirstlane_b32 s0, v2
|
|
; GFX11-NEXT: v_readfirstlane_b32 s1, v3
|
|
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_2)
|
|
; GFX11-NEXT: v_dual_mov_b32 v5, s3 :: v_dual_mov_b32 v4, s2
|
|
; GFX11-NEXT: v_dual_mov_b32 v2, s0 :: v_dual_mov_b32 v3, s1
|
|
; GFX11-NEXT: global_store_b128 v[0:1], v[2:5], off
|
|
; GFX11-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX12-LABEL: v_constained_fma_v2f64_fpexcept_strict_uni:
|
|
; GFX12: ; %bb.0:
|
|
; GFX12-NEXT: s_wait_loadcnt_dscnt 0x0
|
|
; GFX12-NEXT: s_wait_expcnt 0x0
|
|
; GFX12-NEXT: s_wait_samplecnt 0x0
|
|
; GFX12-NEXT: s_wait_bvhcnt 0x0
|
|
; GFX12-NEXT: s_wait_kmcnt 0x0
|
|
; GFX12-NEXT: v_dual_mov_b32 v4, s22 :: v_dual_mov_b32 v5, s23
|
|
; GFX12-NEXT: v_dual_mov_b32 v2, s20 :: v_dual_mov_b32 v3, s21
|
|
; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
|
|
; GFX12-NEXT: v_fma_f64 v[4:5], s[2:3], s[18:19], v[4:5]
|
|
; GFX12-NEXT: v_fma_f64 v[2:3], s[0:1], s[16:17], v[2:3]
|
|
; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3)
|
|
; GFX12-NEXT: v_readfirstlane_b32 s3, v5
|
|
; GFX12-NEXT: v_readfirstlane_b32 s2, v4
|
|
; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_4)
|
|
; GFX12-NEXT: v_readfirstlane_b32 s0, v2
|
|
; GFX12-NEXT: v_readfirstlane_b32 s1, v3
|
|
; GFX12-NEXT: s_wait_alu depctr_va_sdst(0)
|
|
; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_2)
|
|
; GFX12-NEXT: v_dual_mov_b32 v5, s3 :: v_dual_mov_b32 v4, s2
|
|
; GFX12-NEXT: v_dual_mov_b32 v2, s0 :: v_dual_mov_b32 v3, s1
|
|
; GFX12-NEXT: global_store_b128 v[0:1], v[2:5], off
|
|
; GFX12-NEXT: s_setpc_b64 s[30:31]
|
|
|
|
; FIXME: Optimize readfirstlane
|
|
%val = call <2 x double> @llvm.experimental.constrained.fma.v2f64(<2 x double> %x, <2 x double> %y, <2 x double> %z, metadata !"round.tonearest", metadata !"fpexcept.strict")
|
|
store <2 x double> %val, ptr addrspace(1) %out
|
|
ret void
|
|
}
|
|
|
|
define void @v_constained_fma_v2f64_fpexcept_strict_div(<2 x double> %x, <2 x double> %y, <2 x double> %z, ptr addrspace(1) %out) #0 {
|
|
; GFX8-LABEL: v_constained_fma_v2f64_fpexcept_strict_div:
|
|
; GFX8: ; %bb.0:
|
|
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX8-NEXT: v_fma_f64 v[0:1], v[0:1], v[4:5], v[8:9]
|
|
; GFX8-NEXT: v_fma_f64 v[2:3], v[2:3], v[6:7], v[10:11]
|
|
; GFX8-NEXT: flat_store_dwordx4 v[12:13], v[0:3]
|
|
; GFX8-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX8-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX9-LABEL: v_constained_fma_v2f64_fpexcept_strict_div:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: v_fma_f64 v[0:1], v[0:1], v[4:5], v[8:9]
|
|
; GFX9-NEXT: v_fma_f64 v[2:3], v[2:3], v[6:7], v[10:11]
|
|
; GFX9-NEXT: global_store_dwordx4 v[12:13], v[0:3], off
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX11-LABEL: v_constained_fma_v2f64_fpexcept_strict_div:
|
|
; GFX11: ; %bb.0:
|
|
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX11-NEXT: v_fma_f64 v[0:1], v[0:1], v[4:5], v[8:9]
|
|
; GFX11-NEXT: v_fma_f64 v[2:3], v[2:3], v[6:7], v[10:11]
|
|
; GFX11-NEXT: global_store_b128 v[12:13], v[0:3], off
|
|
; GFX11-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX12-LABEL: v_constained_fma_v2f64_fpexcept_strict_div:
|
|
; GFX12: ; %bb.0:
|
|
; GFX12-NEXT: s_wait_loadcnt_dscnt 0x0
|
|
; GFX12-NEXT: s_wait_expcnt 0x0
|
|
; GFX12-NEXT: s_wait_samplecnt 0x0
|
|
; GFX12-NEXT: s_wait_bvhcnt 0x0
|
|
; GFX12-NEXT: s_wait_kmcnt 0x0
|
|
; GFX12-NEXT: v_fma_f64 v[0:1], v[0:1], v[4:5], v[8:9]
|
|
; GFX12-NEXT: v_fma_f64 v[2:3], v[2:3], v[6:7], v[10:11]
|
|
; GFX12-NEXT: global_store_b128 v[12:13], v[0:3], off
|
|
; GFX12-NEXT: s_setpc_b64 s[30:31]
|
|
%val = call <2 x double> @llvm.experimental.constrained.fma.v2f64(<2 x double> %x, <2 x double> %y, <2 x double> %z, metadata !"round.tonearest", metadata !"fpexcept.strict")
|
|
store <2 x double> %val, ptr addrspace(1) %out
|
|
ret void
|
|
}
|
|
|
|
define void @v_constained_fma_v3f64_fpexcept_strict_uni(<3 x double> inreg %x, <3 x double> inreg %y, <3 x double> inreg %z, ptr addrspace(1) %out) #0 {
|
|
; GFX8-LABEL: v_constained_fma_v3f64_fpexcept_strict_uni:
|
|
; GFX8: ; %bb.0:
|
|
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX8-NEXT: v_readfirstlane_b32 s4, v0
|
|
; GFX8-NEXT: v_readfirstlane_b32 s5, v1
|
|
; GFX8-NEXT: v_mov_b32_e32 v0, s22
|
|
; GFX8-NEXT: v_mov_b32_e32 v6, s28
|
|
; GFX8-NEXT: v_mov_b32_e32 v1, s23
|
|
; GFX8-NEXT: v_mov_b32_e32 v7, s29
|
|
; GFX8-NEXT: v_fma_f64 v[0:1], s[16:17], v[0:1], v[6:7]
|
|
; GFX8-NEXT: v_mov_b32_e32 v6, s24
|
|
; GFX8-NEXT: v_mov_b32_e32 v9, s5
|
|
; GFX8-NEXT: v_mov_b32_e32 v7, s25
|
|
; GFX8-NEXT: v_mov_b32_e32 v8, s4
|
|
; GFX8-NEXT: v_fma_f64 v[6:7], s[18:19], v[6:7], v[8:9]
|
|
; GFX8-NEXT: v_readfirstlane_b32 s8, v2
|
|
; GFX8-NEXT: v_readfirstlane_b32 s9, v3
|
|
; GFX8-NEXT: v_readfirstlane_b32 s4, v0
|
|
; GFX8-NEXT: v_readfirstlane_b32 s5, v1
|
|
; GFX8-NEXT: v_mov_b32_e32 v0, s26
|
|
; GFX8-NEXT: v_mov_b32_e32 v2, s8
|
|
; GFX8-NEXT: v_mov_b32_e32 v1, s27
|
|
; GFX8-NEXT: v_mov_b32_e32 v3, s9
|
|
; GFX8-NEXT: v_fma_f64 v[8:9], s[20:21], v[0:1], v[2:3]
|
|
; GFX8-NEXT: v_readfirstlane_b32 s6, v6
|
|
; GFX8-NEXT: v_readfirstlane_b32 s7, v7
|
|
; GFX8-NEXT: v_mov_b32_e32 v0, s4
|
|
; GFX8-NEXT: v_mov_b32_e32 v1, s5
|
|
; GFX8-NEXT: v_mov_b32_e32 v2, s6
|
|
; GFX8-NEXT: v_mov_b32_e32 v3, s7
|
|
; GFX8-NEXT: flat_store_dwordx4 v[4:5], v[0:3]
|
|
; GFX8-NEXT: s_nop 0
|
|
; GFX8-NEXT: v_add_u32_e32 v0, vcc, 16, v4
|
|
; GFX8-NEXT: v_addc_u32_e32 v1, vcc, 0, v5, vcc
|
|
; GFX8-NEXT: flat_store_dwordx2 v[0:1], v[8:9]
|
|
; GFX8-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX8-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX900-LABEL: v_constained_fma_v3f64_fpexcept_strict_uni:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_readfirstlane_b32 s4, v0
|
|
; GFX900-NEXT: v_readfirstlane_b32 s5, v1
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, s22
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, s28
|
|
; GFX900-NEXT: v_mov_b32_e32 v1, s23
|
|
; GFX900-NEXT: v_mov_b32_e32 v7, s29
|
|
; GFX900-NEXT: v_fma_f64 v[0:1], s[16:17], v[0:1], v[6:7]
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, s24
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, s5
|
|
; GFX900-NEXT: v_mov_b32_e32 v7, s25
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, s4
|
|
; GFX900-NEXT: v_fma_f64 v[6:7], s[18:19], v[6:7], v[8:9]
|
|
; GFX900-NEXT: v_readfirstlane_b32 s7, v3
|
|
; GFX900-NEXT: v_readfirstlane_b32 s6, v2
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, s26
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, s7
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, s27
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, s6
|
|
; GFX900-NEXT: v_fma_f64 v[8:9], s[20:21], v[2:3], v[8:9]
|
|
; GFX900-NEXT: v_readfirstlane_b32 s4, v0
|
|
; GFX900-NEXT: v_readfirstlane_b32 s5, v1
|
|
; GFX900-NEXT: v_readfirstlane_b32 s6, v6
|
|
; GFX900-NEXT: v_readfirstlane_b32 s7, v7
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, s4
|
|
; GFX900-NEXT: v_mov_b32_e32 v1, s5
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, s6
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, s7
|
|
; GFX900-NEXT: global_store_dwordx4 v[4:5], v[0:3], off
|
|
; GFX900-NEXT: global_store_dwordx2 v[4:5], v[8:9], off offset:16
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_constained_fma_v3f64_fpexcept_strict_uni:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b64_e32 v[2:3], s[18:19]
|
|
; GFX942-NEXT: v_mov_b64_e32 v[4:5], s[24:25]
|
|
; GFX942-NEXT: v_fma_f64 v[2:3], s[0:1], v[2:3], v[4:5]
|
|
; GFX942-NEXT: v_mov_b64_e32 v[4:5], s[26:27]
|
|
; GFX942-NEXT: v_readfirstlane_b32 s0, v2
|
|
; GFX942-NEXT: v_readfirstlane_b32 s1, v3
|
|
; GFX942-NEXT: v_mov_b64_e32 v[2:3], s[20:21]
|
|
; GFX942-NEXT: v_fma_f64 v[2:3], s[2:3], v[2:3], v[4:5]
|
|
; GFX942-NEXT: v_mov_b64_e32 v[4:5], s[28:29]
|
|
; GFX942-NEXT: v_readfirstlane_b32 s2, v2
|
|
; GFX942-NEXT: v_readfirstlane_b32 s3, v3
|
|
; GFX942-NEXT: v_mov_b64_e32 v[2:3], s[22:23]
|
|
; GFX942-NEXT: v_fma_f64 v[6:7], s[16:17], v[2:3], v[4:5]
|
|
; GFX942-NEXT: v_mov_b64_e32 v[4:5], s[2:3]
|
|
; GFX942-NEXT: v_mov_b64_e32 v[2:3], s[0:1]
|
|
; GFX942-NEXT: global_store_dwordx4 v[0:1], v[2:5], off
|
|
; GFX942-NEXT: global_store_dwordx2 v[0:1], v[6:7], off offset:16
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX11-LABEL: v_constained_fma_v3f64_fpexcept_strict_uni:
|
|
; GFX11: ; %bb.0:
|
|
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX11-NEXT: v_dual_mov_b32 v4, s26 :: v_dual_mov_b32 v5, s27
|
|
; GFX11-NEXT: v_dual_mov_b32 v2, s24 :: v_dual_mov_b32 v3, s25
|
|
; GFX11-NEXT: v_dual_mov_b32 v6, s28 :: v_dual_mov_b32 v7, s29
|
|
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
|
|
; GFX11-NEXT: v_fma_f64 v[4:5], s[2:3], s[20:21], v[4:5]
|
|
; GFX11-NEXT: v_fma_f64 v[2:3], s[0:1], s[18:19], v[2:3]
|
|
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
|
|
; GFX11-NEXT: v_fma_f64 v[6:7], s[16:17], s[22:23], v[6:7]
|
|
; GFX11-NEXT: v_readfirstlane_b32 s3, v5
|
|
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
|
|
; GFX11-NEXT: v_readfirstlane_b32 s2, v4
|
|
; GFX11-NEXT: v_readfirstlane_b32 s0, v2
|
|
; GFX11-NEXT: v_readfirstlane_b32 s1, v3
|
|
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_2)
|
|
; GFX11-NEXT: v_dual_mov_b32 v5, s3 :: v_dual_mov_b32 v4, s2
|
|
; GFX11-NEXT: v_dual_mov_b32 v2, s0 :: v_dual_mov_b32 v3, s1
|
|
; GFX11-NEXT: s_clause 0x1
|
|
; GFX11-NEXT: global_store_b128 v[0:1], v[2:5], off
|
|
; GFX11-NEXT: global_store_b64 v[0:1], v[6:7], off offset:16
|
|
; GFX11-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX12-LABEL: v_constained_fma_v3f64_fpexcept_strict_uni:
|
|
; GFX12: ; %bb.0:
|
|
; GFX12-NEXT: s_wait_loadcnt_dscnt 0x0
|
|
; GFX12-NEXT: s_wait_expcnt 0x0
|
|
; GFX12-NEXT: s_wait_samplecnt 0x0
|
|
; GFX12-NEXT: s_wait_bvhcnt 0x0
|
|
; GFX12-NEXT: s_wait_kmcnt 0x0
|
|
; GFX12-NEXT: v_dual_mov_b32 v4, s26 :: v_dual_mov_b32 v5, s27
|
|
; GFX12-NEXT: v_dual_mov_b32 v2, s24 :: v_dual_mov_b32 v3, s25
|
|
; GFX12-NEXT: v_dual_mov_b32 v6, s28 :: v_dual_mov_b32 v7, s29
|
|
; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
|
|
; GFX12-NEXT: v_fma_f64 v[4:5], s[2:3], s[20:21], v[4:5]
|
|
; GFX12-NEXT: v_fma_f64 v[2:3], s[0:1], s[18:19], v[2:3]
|
|
; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
|
|
; GFX12-NEXT: v_fma_f64 v[6:7], s[16:17], s[22:23], v[6:7]
|
|
; GFX12-NEXT: v_readfirstlane_b32 s3, v5
|
|
; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
|
|
; GFX12-NEXT: v_readfirstlane_b32 s2, v4
|
|
; GFX12-NEXT: v_readfirstlane_b32 s0, v2
|
|
; GFX12-NEXT: v_readfirstlane_b32 s1, v3
|
|
; GFX12-NEXT: s_wait_alu depctr_va_sdst(0)
|
|
; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_2)
|
|
; GFX12-NEXT: v_dual_mov_b32 v5, s3 :: v_dual_mov_b32 v4, s2
|
|
; GFX12-NEXT: v_dual_mov_b32 v2, s0 :: v_dual_mov_b32 v3, s1
|
|
; GFX12-NEXT: s_clause 0x1
|
|
; GFX12-NEXT: global_store_b128 v[0:1], v[2:5], off
|
|
; GFX12-NEXT: global_store_b64 v[0:1], v[6:7], off offset:16
|
|
; GFX12-NEXT: s_setpc_b64 s[30:31]
|
|
%val = call <3 x double> @llvm.experimental.constrained.fma.v3f64(<3 x double> %x, <3 x double> %y, <3 x double> %z, metadata !"round.tonearest", metadata !"fpexcept.strict")
|
|
store <3 x double> %val, ptr addrspace(1) %out
|
|
ret void
|
|
}
|
|
|
|
define void @v_constained_fma_v3f64_fpexcept_strict_div(<3 x double> %x, <3 x double> %y, <3 x double> %z, ptr addrspace(1) %out) #0 {
|
|
; GFX8-LABEL: v_constained_fma_v3f64_fpexcept_strict_div:
|
|
; GFX8: ; %bb.0:
|
|
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX8-NEXT: v_fma_f64 v[0:1], v[0:1], v[6:7], v[12:13]
|
|
; GFX8-NEXT: v_fma_f64 v[2:3], v[2:3], v[8:9], v[14:15]
|
|
; GFX8-NEXT: v_fma_f64 v[4:5], v[4:5], v[10:11], v[16:17]
|
|
; GFX8-NEXT: flat_store_dwordx4 v[18:19], v[0:3]
|
|
; GFX8-NEXT: s_nop 0
|
|
; GFX8-NEXT: v_add_u32_e32 v0, vcc, 16, v18
|
|
; GFX8-NEXT: v_addc_u32_e32 v1, vcc, 0, v19, vcc
|
|
; GFX8-NEXT: flat_store_dwordx2 v[0:1], v[4:5]
|
|
; GFX8-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX8-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX9-LABEL: v_constained_fma_v3f64_fpexcept_strict_div:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: v_fma_f64 v[0:1], v[0:1], v[6:7], v[12:13]
|
|
; GFX9-NEXT: v_fma_f64 v[2:3], v[2:3], v[8:9], v[14:15]
|
|
; GFX9-NEXT: v_fma_f64 v[4:5], v[4:5], v[10:11], v[16:17]
|
|
; GFX9-NEXT: global_store_dwordx4 v[18:19], v[0:3], off
|
|
; GFX9-NEXT: global_store_dwordx2 v[18:19], v[4:5], off offset:16
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX11-LABEL: v_constained_fma_v3f64_fpexcept_strict_div:
|
|
; GFX11: ; %bb.0:
|
|
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX11-NEXT: v_fma_f64 v[0:1], v[0:1], v[6:7], v[12:13]
|
|
; GFX11-NEXT: v_fma_f64 v[2:3], v[2:3], v[8:9], v[14:15]
|
|
; GFX11-NEXT: v_fma_f64 v[4:5], v[4:5], v[10:11], v[16:17]
|
|
; GFX11-NEXT: s_clause 0x1
|
|
; GFX11-NEXT: global_store_b128 v[18:19], v[0:3], off
|
|
; GFX11-NEXT: global_store_b64 v[18:19], v[4:5], off offset:16
|
|
; GFX11-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX12-LABEL: v_constained_fma_v3f64_fpexcept_strict_div:
|
|
; GFX12: ; %bb.0:
|
|
; GFX12-NEXT: s_wait_loadcnt_dscnt 0x0
|
|
; GFX12-NEXT: s_wait_expcnt 0x0
|
|
; GFX12-NEXT: s_wait_samplecnt 0x0
|
|
; GFX12-NEXT: s_wait_bvhcnt 0x0
|
|
; GFX12-NEXT: s_wait_kmcnt 0x0
|
|
; GFX12-NEXT: v_fma_f64 v[0:1], v[0:1], v[6:7], v[12:13]
|
|
; GFX12-NEXT: v_fma_f64 v[2:3], v[2:3], v[8:9], v[14:15]
|
|
; GFX12-NEXT: v_fma_f64 v[4:5], v[4:5], v[10:11], v[16:17]
|
|
; GFX12-NEXT: s_clause 0x1
|
|
; GFX12-NEXT: global_store_b128 v[18:19], v[0:3], off
|
|
; GFX12-NEXT: global_store_b64 v[18:19], v[4:5], off offset:16
|
|
; GFX12-NEXT: s_setpc_b64 s[30:31]
|
|
%val = call <3 x double> @llvm.experimental.constrained.fma.v3f64(<3 x double> %x, <3 x double> %y, <3 x double> %z, metadata !"round.tonearest", metadata !"fpexcept.strict")
|
|
store <3 x double> %val, ptr addrspace(1) %out
|
|
ret void
|
|
}
|
|
|
|
define void @v_constained_fma_v4f64_fpexcept_strict_uni(<4 x double> inreg %x, <4 x double> inreg %y, <4 x double> inreg %z, ptr addrspace(1) %out) #0 {
|
|
; GFX8-LABEL: v_constained_fma_v4f64_fpexcept_strict_uni:
|
|
; GFX8: ; %bb.0:
|
|
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX8-NEXT: v_readfirstlane_b32 s6, v2
|
|
; GFX8-NEXT: v_readfirstlane_b32 s4, v0
|
|
; GFX8-NEXT: v_readfirstlane_b32 s5, v1
|
|
; GFX8-NEXT: v_readfirstlane_b32 s7, v3
|
|
; GFX8-NEXT: v_mov_b32_e32 v0, s24
|
|
; GFX8-NEXT: v_mov_b32_e32 v2, s6
|
|
; GFX8-NEXT: v_readfirstlane_b32 s8, v4
|
|
; GFX8-NEXT: v_mov_b32_e32 v1, s25
|
|
; GFX8-NEXT: v_mov_b32_e32 v3, s7
|
|
; GFX8-NEXT: v_readfirstlane_b32 s9, v5
|
|
; GFX8-NEXT: v_fma_f64 v[0:1], s[16:17], v[0:1], v[2:3]
|
|
; GFX8-NEXT: v_mov_b32_e32 v2, s26
|
|
; GFX8-NEXT: v_mov_b32_e32 v4, s8
|
|
; GFX8-NEXT: v_readfirstlane_b32 s10, v6
|
|
; GFX8-NEXT: v_mov_b32_e32 v3, s27
|
|
; GFX8-NEXT: v_mov_b32_e32 v5, s9
|
|
; GFX8-NEXT: v_readfirstlane_b32 s11, v7
|
|
; GFX8-NEXT: v_fma_f64 v[2:3], s[18:19], v[2:3], v[4:5]
|
|
; GFX8-NEXT: v_mov_b32_e32 v4, s28
|
|
; GFX8-NEXT: v_mov_b32_e32 v6, s10
|
|
; GFX8-NEXT: v_readfirstlane_b32 s12, v8
|
|
; GFX8-NEXT: v_mov_b32_e32 v5, s29
|
|
; GFX8-NEXT: v_mov_b32_e32 v7, s11
|
|
; GFX8-NEXT: v_readfirstlane_b32 s13, v9
|
|
; GFX8-NEXT: v_fma_f64 v[4:5], s[20:21], v[4:5], v[6:7]
|
|
; GFX8-NEXT: v_mov_b32_e32 v7, s5
|
|
; GFX8-NEXT: v_mov_b32_e32 v8, s12
|
|
; GFX8-NEXT: v_mov_b32_e32 v6, s4
|
|
; GFX8-NEXT: v_mov_b32_e32 v9, s13
|
|
; GFX8-NEXT: v_fma_f64 v[6:7], s[22:23], v[6:7], v[8:9]
|
|
; GFX8-NEXT: v_readfirstlane_b32 s4, v0
|
|
; GFX8-NEXT: v_readfirstlane_b32 s5, v1
|
|
; GFX8-NEXT: v_readfirstlane_b32 s6, v2
|
|
; GFX8-NEXT: v_readfirstlane_b32 s7, v3
|
|
; GFX8-NEXT: v_mov_b32_e32 v0, s4
|
|
; GFX8-NEXT: v_readfirstlane_b32 s8, v4
|
|
; GFX8-NEXT: v_mov_b32_e32 v1, s5
|
|
; GFX8-NEXT: v_mov_b32_e32 v2, s6
|
|
; GFX8-NEXT: v_mov_b32_e32 v3, s7
|
|
; GFX8-NEXT: v_readfirstlane_b32 s9, v5
|
|
; GFX8-NEXT: v_readfirstlane_b32 s10, v6
|
|
; GFX8-NEXT: v_readfirstlane_b32 s11, v7
|
|
; GFX8-NEXT: flat_store_dwordx4 v[10:11], v[0:3]
|
|
; GFX8-NEXT: v_add_u32_e32 v4, vcc, 16, v10
|
|
; GFX8-NEXT: v_mov_b32_e32 v0, s8
|
|
; GFX8-NEXT: v_addc_u32_e32 v5, vcc, 0, v11, vcc
|
|
; GFX8-NEXT: v_mov_b32_e32 v1, s9
|
|
; GFX8-NEXT: v_mov_b32_e32 v2, s10
|
|
; GFX8-NEXT: v_mov_b32_e32 v3, s11
|
|
; GFX8-NEXT: flat_store_dwordx4 v[4:5], v[0:3]
|
|
; GFX8-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX8-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX900-LABEL: v_constained_fma_v4f64_fpexcept_strict_uni:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_readfirstlane_b32 s6, v2
|
|
; GFX900-NEXT: v_readfirstlane_b32 s4, v0
|
|
; GFX900-NEXT: v_readfirstlane_b32 s5, v1
|
|
; GFX900-NEXT: v_readfirstlane_b32 s7, v3
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, s24
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, s6
|
|
; GFX900-NEXT: v_readfirstlane_b32 s8, v4
|
|
; GFX900-NEXT: v_mov_b32_e32 v1, s25
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, s7
|
|
; GFX900-NEXT: v_readfirstlane_b32 s9, v5
|
|
; GFX900-NEXT: v_fma_f64 v[0:1], s[16:17], v[0:1], v[2:3]
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, s26
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, s8
|
|
; GFX900-NEXT: v_readfirstlane_b32 s10, v6
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, s27
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, s9
|
|
; GFX900-NEXT: v_readfirstlane_b32 s11, v7
|
|
; GFX900-NEXT: v_fma_f64 v[2:3], s[18:19], v[2:3], v[4:5]
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, s28
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, s10
|
|
; GFX900-NEXT: v_readfirstlane_b32 s12, v8
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, s29
|
|
; GFX900-NEXT: v_mov_b32_e32 v7, s11
|
|
; GFX900-NEXT: v_readfirstlane_b32 s13, v9
|
|
; GFX900-NEXT: v_fma_f64 v[4:5], s[20:21], v[4:5], v[6:7]
|
|
; GFX900-NEXT: v_mov_b32_e32 v7, s5
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, s12
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, s4
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, s13
|
|
; GFX900-NEXT: v_fma_f64 v[6:7], s[22:23], v[6:7], v[8:9]
|
|
; GFX900-NEXT: v_readfirstlane_b32 s4, v0
|
|
; GFX900-NEXT: v_readfirstlane_b32 s5, v1
|
|
; GFX900-NEXT: v_readfirstlane_b32 s6, v2
|
|
; GFX900-NEXT: v_readfirstlane_b32 s7, v3
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, s4
|
|
; GFX900-NEXT: v_readfirstlane_b32 s8, v4
|
|
; GFX900-NEXT: v_mov_b32_e32 v1, s5
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, s6
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, s7
|
|
; GFX900-NEXT: v_readfirstlane_b32 s9, v5
|
|
; GFX900-NEXT: v_readfirstlane_b32 s10, v6
|
|
; GFX900-NEXT: v_readfirstlane_b32 s11, v7
|
|
; GFX900-NEXT: global_store_dwordx4 v[10:11], v[0:3], off
|
|
; GFX900-NEXT: s_nop 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, s8
|
|
; GFX900-NEXT: v_mov_b32_e32 v1, s9
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, s10
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, s11
|
|
; GFX900-NEXT: global_store_dwordx4 v[10:11], v[0:3], off offset:16
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_constained_fma_v4f64_fpexcept_strict_uni:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_readfirstlane_b32 s4, v0
|
|
; GFX942-NEXT: v_readfirstlane_b32 s5, v1
|
|
; GFX942-NEXT: v_readfirstlane_b32 s6, v2
|
|
; GFX942-NEXT: v_readfirstlane_b32 s7, v3
|
|
; GFX942-NEXT: v_mov_b64_e32 v[0:1], s[20:21]
|
|
; GFX942-NEXT: v_mov_b64_e32 v[2:3], s[28:29]
|
|
; GFX942-NEXT: v_fma_f64 v[0:1], s[0:1], v[0:1], v[2:3]
|
|
; GFX942-NEXT: v_mov_b64_e32 v[2:3], s[4:5]
|
|
; GFX942-NEXT: v_readfirstlane_b32 s0, v0
|
|
; GFX942-NEXT: v_readfirstlane_b32 s1, v1
|
|
; GFX942-NEXT: v_mov_b64_e32 v[0:1], s[22:23]
|
|
; GFX942-NEXT: v_fma_f64 v[0:1], s[2:3], v[0:1], v[2:3]
|
|
; GFX942-NEXT: v_mov_b64_e32 v[2:3], s[6:7]
|
|
; GFX942-NEXT: v_readfirstlane_b32 s2, v0
|
|
; GFX942-NEXT: v_readfirstlane_b32 s3, v1
|
|
; GFX942-NEXT: v_mov_b64_e32 v[0:1], s[24:25]
|
|
; GFX942-NEXT: v_readfirstlane_b32 s8, v4
|
|
; GFX942-NEXT: v_readfirstlane_b32 s9, v5
|
|
; GFX942-NEXT: v_fma_f64 v[0:1], s[16:17], v[0:1], v[2:3]
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_readfirstlane_b32 s4, v0
|
|
; GFX942-NEXT: v_readfirstlane_b32 s5, v1
|
|
; GFX942-NEXT: v_mov_b64_e32 v[0:1], s[26:27]
|
|
; GFX942-NEXT: v_mov_b64_e32 v[2:3], s[8:9]
|
|
; GFX942-NEXT: v_fma_f64 v[0:1], s[18:19], v[0:1], v[2:3]
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_readfirstlane_b32 s6, v0
|
|
; GFX942-NEXT: v_readfirstlane_b32 s7, v1
|
|
; GFX942-NEXT: v_mov_b64_e32 v[0:1], s[0:1]
|
|
; GFX942-NEXT: v_mov_b64_e32 v[2:3], s[2:3]
|
|
; GFX942-NEXT: global_store_dwordx4 v[6:7], v[0:3], off
|
|
; GFX942-NEXT: s_nop 1
|
|
; GFX942-NEXT: v_mov_b64_e32 v[0:1], s[4:5]
|
|
; GFX942-NEXT: v_mov_b64_e32 v[2:3], s[6:7]
|
|
; GFX942-NEXT: global_store_dwordx4 v[6:7], v[0:3], off offset:16
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX11-LABEL: v_constained_fma_v4f64_fpexcept_strict_uni:
|
|
; GFX11: ; %bb.0:
|
|
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX11-NEXT: v_readfirstlane_b32 s4, v0
|
|
; GFX11-NEXT: v_readfirstlane_b32 s5, v1
|
|
; GFX11-NEXT: v_readfirstlane_b32 s8, v4
|
|
; GFX11-NEXT: v_readfirstlane_b32 s9, v5
|
|
; GFX11-NEXT: v_readfirstlane_b32 s6, v2
|
|
; GFX11-NEXT: v_readfirstlane_b32 s7, v3
|
|
; GFX11-NEXT: v_dual_mov_b32 v0, s28 :: v_dual_mov_b32 v1, s29
|
|
; GFX11-NEXT: v_dual_mov_b32 v2, s4 :: v_dual_mov_b32 v3, s5
|
|
; GFX11-NEXT: v_dual_mov_b32 v8, s8 :: v_dual_mov_b32 v9, s9
|
|
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
|
|
; GFX11-NEXT: v_dual_mov_b32 v4, s6 :: v_dual_mov_b32 v5, s7
|
|
; GFX11-NEXT: v_fma_f64 v[0:1], s[0:1], s[20:21], v[0:1]
|
|
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
|
|
; GFX11-NEXT: v_fma_f64 v[2:3], s[2:3], s[22:23], v[2:3]
|
|
; GFX11-NEXT: v_fma_f64 v[8:9], s[18:19], s[26:27], v[8:9]
|
|
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
|
|
; GFX11-NEXT: v_fma_f64 v[4:5], s[16:17], s[24:25], v[4:5]
|
|
; GFX11-NEXT: v_readfirstlane_b32 s0, v0
|
|
; GFX11-NEXT: v_readfirstlane_b32 s1, v1
|
|
; GFX11-NEXT: v_readfirstlane_b32 s2, v2
|
|
; GFX11-NEXT: v_readfirstlane_b32 s3, v3
|
|
; GFX11-NEXT: v_readfirstlane_b32 s7, v9
|
|
; GFX11-NEXT: v_readfirstlane_b32 s4, v4
|
|
; GFX11-NEXT: v_readfirstlane_b32 s5, v5
|
|
; GFX11-NEXT: v_readfirstlane_b32 s6, v8
|
|
; GFX11-NEXT: v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
|
|
; GFX11-NEXT: v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s3
|
|
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_3)
|
|
; GFX11-NEXT: v_dual_mov_b32 v11, s7 :: v_dual_mov_b32 v10, s6
|
|
; GFX11-NEXT: v_dual_mov_b32 v9, s5 :: v_dual_mov_b32 v8, s4
|
|
; GFX11-NEXT: s_clause 0x1
|
|
; GFX11-NEXT: global_store_b128 v[6:7], v[0:3], off
|
|
; GFX11-NEXT: global_store_b128 v[6:7], v[8:11], off offset:16
|
|
; GFX11-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX12-LABEL: v_constained_fma_v4f64_fpexcept_strict_uni:
|
|
; GFX12: ; %bb.0:
|
|
; GFX12-NEXT: s_wait_loadcnt_dscnt 0x0
|
|
; GFX12-NEXT: s_wait_expcnt 0x0
|
|
; GFX12-NEXT: s_wait_samplecnt 0x0
|
|
; GFX12-NEXT: s_wait_bvhcnt 0x0
|
|
; GFX12-NEXT: s_wait_kmcnt 0x0
|
|
; GFX12-NEXT: v_readfirstlane_b32 s4, v0
|
|
; GFX12-NEXT: v_readfirstlane_b32 s5, v1
|
|
; GFX12-NEXT: v_readfirstlane_b32 s8, v4
|
|
; GFX12-NEXT: v_readfirstlane_b32 s9, v5
|
|
; GFX12-NEXT: v_readfirstlane_b32 s6, v2
|
|
; GFX12-NEXT: v_readfirstlane_b32 s7, v3
|
|
; GFX12-NEXT: v_dual_mov_b32 v0, s28 :: v_dual_mov_b32 v1, s29
|
|
; GFX12-NEXT: s_wait_alu depctr_va_sdst(0)
|
|
; GFX12-NEXT: v_dual_mov_b32 v2, s4 :: v_dual_mov_b32 v3, s5
|
|
; GFX12-NEXT: v_dual_mov_b32 v8, s8 :: v_dual_mov_b32 v9, s9
|
|
; GFX12-NEXT: v_dual_mov_b32 v4, s6 :: v_dual_mov_b32 v5, s7
|
|
; GFX12-NEXT: v_fma_f64 v[0:1], s[0:1], s[20:21], v[0:1]
|
|
; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
|
|
; GFX12-NEXT: v_fma_f64 v[2:3], s[2:3], s[22:23], v[2:3]
|
|
; GFX12-NEXT: v_fma_f64 v[8:9], s[18:19], s[26:27], v[8:9]
|
|
; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
|
|
; GFX12-NEXT: v_fma_f64 v[4:5], s[16:17], s[24:25], v[4:5]
|
|
; GFX12-NEXT: v_readfirstlane_b32 s0, v0
|
|
; GFX12-NEXT: v_readfirstlane_b32 s1, v1
|
|
; GFX12-NEXT: v_readfirstlane_b32 s2, v2
|
|
; GFX12-NEXT: v_readfirstlane_b32 s3, v3
|
|
; GFX12-NEXT: v_readfirstlane_b32 s7, v9
|
|
; GFX12-NEXT: v_readfirstlane_b32 s4, v4
|
|
; GFX12-NEXT: v_readfirstlane_b32 s5, v5
|
|
; GFX12-NEXT: v_readfirstlane_b32 s6, v8
|
|
; GFX12-NEXT: s_wait_alu depctr_va_sdst(0)
|
|
; GFX12-NEXT: v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
|
|
; GFX12-NEXT: v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s3
|
|
; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_3)
|
|
; GFX12-NEXT: v_dual_mov_b32 v11, s7 :: v_dual_mov_b32 v10, s6
|
|
; GFX12-NEXT: v_dual_mov_b32 v9, s5 :: v_dual_mov_b32 v8, s4
|
|
; GFX12-NEXT: s_clause 0x1
|
|
; GFX12-NEXT: global_store_b128 v[6:7], v[0:3], off
|
|
; GFX12-NEXT: global_store_b128 v[6:7], v[8:11], off offset:16
|
|
; GFX12-NEXT: s_setpc_b64 s[30:31]
|
|
%val = call <4 x double> @llvm.experimental.constrained.fma.v4f64(<4 x double> %x, <4 x double> %y, <4 x double> %z, metadata !"round.tonearest", metadata !"fpexcept.strict")
|
|
store <4 x double> %val, ptr addrspace(1) %out
|
|
ret void
|
|
}
|
|
|
|
define void @v_constained_fma_v4f64_fpexcept_strict_div(<4 x double> %x, <4 x double> %y, <4 x double> %z, ptr addrspace(1) %out) #0 {
|
|
; GFX8-LABEL: v_constained_fma_v4f64_fpexcept_strict_div:
|
|
; GFX8: ; %bb.0:
|
|
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX8-NEXT: v_fma_f64 v[0:1], v[0:1], v[8:9], v[16:17]
|
|
; GFX8-NEXT: v_fma_f64 v[2:3], v[2:3], v[10:11], v[18:19]
|
|
; GFX8-NEXT: v_fma_f64 v[4:5], v[4:5], v[12:13], v[20:21]
|
|
; GFX8-NEXT: v_fma_f64 v[6:7], v[6:7], v[14:15], v[22:23]
|
|
; GFX8-NEXT: flat_store_dwordx4 v[24:25], v[0:3]
|
|
; GFX8-NEXT: s_nop 0
|
|
; GFX8-NEXT: v_add_u32_e32 v0, vcc, 16, v24
|
|
; GFX8-NEXT: v_addc_u32_e32 v1, vcc, 0, v25, vcc
|
|
; GFX8-NEXT: flat_store_dwordx4 v[0:1], v[4:7]
|
|
; GFX8-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX8-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX9-LABEL: v_constained_fma_v4f64_fpexcept_strict_div:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: v_fma_f64 v[0:1], v[0:1], v[8:9], v[16:17]
|
|
; GFX9-NEXT: v_fma_f64 v[2:3], v[2:3], v[10:11], v[18:19]
|
|
; GFX9-NEXT: v_fma_f64 v[4:5], v[4:5], v[12:13], v[20:21]
|
|
; GFX9-NEXT: v_fma_f64 v[6:7], v[6:7], v[14:15], v[22:23]
|
|
; GFX9-NEXT: global_store_dwordx4 v[24:25], v[0:3], off
|
|
; GFX9-NEXT: global_store_dwordx4 v[24:25], v[4:7], off offset:16
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX11-LABEL: v_constained_fma_v4f64_fpexcept_strict_div:
|
|
; GFX11: ; %bb.0:
|
|
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX11-NEXT: v_fma_f64 v[0:1], v[0:1], v[8:9], v[16:17]
|
|
; GFX11-NEXT: v_fma_f64 v[2:3], v[2:3], v[10:11], v[18:19]
|
|
; GFX11-NEXT: v_fma_f64 v[4:5], v[4:5], v[12:13], v[20:21]
|
|
; GFX11-NEXT: v_fma_f64 v[6:7], v[6:7], v[14:15], v[22:23]
|
|
; GFX11-NEXT: s_clause 0x1
|
|
; GFX11-NEXT: global_store_b128 v[24:25], v[0:3], off
|
|
; GFX11-NEXT: global_store_b128 v[24:25], v[4:7], off offset:16
|
|
; GFX11-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX12-LABEL: v_constained_fma_v4f64_fpexcept_strict_div:
|
|
; GFX12: ; %bb.0:
|
|
; GFX12-NEXT: s_wait_loadcnt_dscnt 0x0
|
|
; GFX12-NEXT: s_wait_expcnt 0x0
|
|
; GFX12-NEXT: s_wait_samplecnt 0x0
|
|
; GFX12-NEXT: s_wait_bvhcnt 0x0
|
|
; GFX12-NEXT: s_wait_kmcnt 0x0
|
|
; GFX12-NEXT: v_fma_f64 v[0:1], v[0:1], v[8:9], v[16:17]
|
|
; GFX12-NEXT: v_fma_f64 v[2:3], v[2:3], v[10:11], v[18:19]
|
|
; GFX12-NEXT: v_fma_f64 v[4:5], v[4:5], v[12:13], v[20:21]
|
|
; GFX12-NEXT: v_fma_f64 v[6:7], v[6:7], v[14:15], v[22:23]
|
|
; GFX12-NEXT: s_clause 0x1
|
|
; GFX12-NEXT: global_store_b128 v[24:25], v[0:3], off
|
|
; GFX12-NEXT: global_store_b128 v[24:25], v[4:7], off offset:16
|
|
; GFX12-NEXT: s_setpc_b64 s[30:31]
|
|
%val = call <4 x double> @llvm.experimental.constrained.fma.v4f64(<4 x double> %x, <4 x double> %y, <4 x double> %z, metadata !"round.tonearest", metadata !"fpexcept.strict")
|
|
store <4 x double> %val, ptr addrspace(1) %out
|
|
ret void
|
|
}
|
|
|
|
define void @v_constained_fma_f64_fpexcept_strict_fneg_uni(double inreg %x, double inreg %y, double inreg %z, ptr addrspace(1) %out) #0 {
|
|
; GFX8-LABEL: v_constained_fma_f64_fpexcept_strict_fneg_uni:
|
|
; GFX8: ; %bb.0:
|
|
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX8-NEXT: v_mov_b32_e32 v2, s16
|
|
; GFX8-NEXT: v_mov_b32_e32 v4, s18
|
|
; GFX8-NEXT: v_mov_b32_e32 v3, s17
|
|
; GFX8-NEXT: v_mov_b32_e32 v5, s19
|
|
; GFX8-NEXT: v_fma_f64 v[2:3], v[2:3], v[4:5], -s[20:21]
|
|
; GFX8-NEXT: flat_store_dwordx2 v[0:1], v[2:3]
|
|
; GFX8-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX8-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX900-LABEL: v_constained_fma_f64_fpexcept_strict_fneg_uni:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, s16
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, s18
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, s17
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, s19
|
|
; GFX900-NEXT: v_fma_f64 v[2:3], v[2:3], v[4:5], -s[20:21]
|
|
; GFX900-NEXT: global_store_dwordx2 v[0:1], v[2:3], off
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_constained_fma_f64_fpexcept_strict_fneg_uni:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b64_e32 v[2:3], s[0:1]
|
|
; GFX942-NEXT: v_mov_b64_e32 v[4:5], s[2:3]
|
|
; GFX942-NEXT: v_fma_f64 v[2:3], v[2:3], v[4:5], -s[16:17]
|
|
; GFX942-NEXT: global_store_dwordx2 v[0:1], v[2:3], off
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX11-LABEL: v_constained_fma_f64_fpexcept_strict_fneg_uni:
|
|
; GFX11: ; %bb.0:
|
|
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX11-NEXT: v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s3
|
|
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
|
; GFX11-NEXT: v_fma_f64 v[2:3], s[0:1], v[2:3], -s[16:17]
|
|
; GFX11-NEXT: global_store_b64 v[0:1], v[2:3], off
|
|
; GFX11-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX12-LABEL: v_constained_fma_f64_fpexcept_strict_fneg_uni:
|
|
; GFX12: ; %bb.0:
|
|
; GFX12-NEXT: s_wait_loadcnt_dscnt 0x0
|
|
; GFX12-NEXT: s_wait_expcnt 0x0
|
|
; GFX12-NEXT: s_wait_samplecnt 0x0
|
|
; GFX12-NEXT: s_wait_bvhcnt 0x0
|
|
; GFX12-NEXT: s_wait_kmcnt 0x0
|
|
; GFX12-NEXT: v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s3
|
|
; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
|
; GFX12-NEXT: v_fma_f64 v[2:3], s[0:1], v[2:3], -s[16:17]
|
|
; GFX12-NEXT: global_store_b64 v[0:1], v[2:3], off
|
|
; GFX12-NEXT: s_setpc_b64 s[30:31]
|
|
%neg.z = fneg double %z
|
|
%val = call double @llvm.experimental.constrained.fma.f64(double %x, double %y, double %neg.z, metadata !"round.tonearest", metadata !"fpexcept.strict")
|
|
store double %val, ptr addrspace(1) %out
|
|
ret void
|
|
}
|
|
|
|
define void @v_constained_fma_f64_fpexcept_strict_fneg_div(double %x, double %y, double %z, ptr addrspace(1) %out) #0 {
|
|
; GFX8-LABEL: v_constained_fma_f64_fpexcept_strict_fneg_div:
|
|
; GFX8: ; %bb.0:
|
|
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX8-NEXT: v_fma_f64 v[0:1], v[0:1], v[2:3], -v[4:5]
|
|
; GFX8-NEXT: flat_store_dwordx2 v[6:7], v[0:1]
|
|
; GFX8-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX8-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX9-LABEL: v_constained_fma_f64_fpexcept_strict_fneg_div:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: v_fma_f64 v[0:1], v[0:1], v[2:3], -v[4:5]
|
|
; GFX9-NEXT: global_store_dwordx2 v[6:7], v[0:1], off
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX11-LABEL: v_constained_fma_f64_fpexcept_strict_fneg_div:
|
|
; GFX11: ; %bb.0:
|
|
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX11-NEXT: v_fma_f64 v[0:1], v[0:1], v[2:3], -v[4:5]
|
|
; GFX11-NEXT: global_store_b64 v[6:7], v[0:1], off
|
|
; GFX11-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX12-LABEL: v_constained_fma_f64_fpexcept_strict_fneg_div:
|
|
; GFX12: ; %bb.0:
|
|
; GFX12-NEXT: s_wait_loadcnt_dscnt 0x0
|
|
; GFX12-NEXT: s_wait_expcnt 0x0
|
|
; GFX12-NEXT: s_wait_samplecnt 0x0
|
|
; GFX12-NEXT: s_wait_bvhcnt 0x0
|
|
; GFX12-NEXT: s_wait_kmcnt 0x0
|
|
; GFX12-NEXT: v_fma_f64 v[0:1], v[0:1], v[2:3], -v[4:5]
|
|
; GFX12-NEXT: global_store_b64 v[6:7], v[0:1], off
|
|
; GFX12-NEXT: s_setpc_b64 s[30:31]
|
|
%neg.z = fneg double %z
|
|
%val = call double @llvm.experimental.constrained.fma.f64(double %x, double %y, double %neg.z, metadata !"round.tonearest", metadata !"fpexcept.strict")
|
|
store double %val, ptr addrspace(1) %out
|
|
ret void
|
|
}
|
|
|
|
define void @v_constained_fma_f64_fpexcept_strict_fneg_fneg_uni(double inreg %x, double inreg %y, double inreg %z, ptr addrspace(1) %out) #0 {
|
|
; GFX8-LABEL: v_constained_fma_f64_fpexcept_strict_fneg_fneg_uni:
|
|
; GFX8: ; %bb.0:
|
|
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX8-NEXT: v_mov_b32_e32 v2, s18
|
|
; GFX8-NEXT: v_mov_b32_e32 v4, s20
|
|
; GFX8-NEXT: v_mov_b32_e32 v3, s19
|
|
; GFX8-NEXT: v_mov_b32_e32 v5, s21
|
|
; GFX8-NEXT: v_fma_f64 v[2:3], -s[16:17], -v[2:3], v[4:5]
|
|
; GFX8-NEXT: flat_store_dwordx2 v[0:1], v[2:3]
|
|
; GFX8-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX8-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX900-LABEL: v_constained_fma_f64_fpexcept_strict_fneg_fneg_uni:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, s18
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, s20
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, s19
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, s21
|
|
; GFX900-NEXT: v_fma_f64 v[2:3], -s[16:17], -v[2:3], v[4:5]
|
|
; GFX900-NEXT: global_store_dwordx2 v[0:1], v[2:3], off
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_constained_fma_f64_fpexcept_strict_fneg_fneg_uni:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b64_e32 v[2:3], s[2:3]
|
|
; GFX942-NEXT: v_mov_b64_e32 v[4:5], s[16:17]
|
|
; GFX942-NEXT: v_fma_f64 v[2:3], -s[0:1], -v[2:3], v[4:5]
|
|
; GFX942-NEXT: global_store_dwordx2 v[0:1], v[2:3], off
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX11-LABEL: v_constained_fma_f64_fpexcept_strict_fneg_fneg_uni:
|
|
; GFX11: ; %bb.0:
|
|
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX11-NEXT: v_dual_mov_b32 v2, s16 :: v_dual_mov_b32 v3, s17
|
|
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
|
; GFX11-NEXT: v_fma_f64 v[2:3], -s[0:1], -s[2:3], v[2:3]
|
|
; GFX11-NEXT: global_store_b64 v[0:1], v[2:3], off
|
|
; GFX11-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX12-LABEL: v_constained_fma_f64_fpexcept_strict_fneg_fneg_uni:
|
|
; GFX12: ; %bb.0:
|
|
; GFX12-NEXT: s_wait_loadcnt_dscnt 0x0
|
|
; GFX12-NEXT: s_wait_expcnt 0x0
|
|
; GFX12-NEXT: s_wait_samplecnt 0x0
|
|
; GFX12-NEXT: s_wait_bvhcnt 0x0
|
|
; GFX12-NEXT: s_wait_kmcnt 0x0
|
|
; GFX12-NEXT: v_dual_mov_b32 v2, s16 :: v_dual_mov_b32 v3, s17
|
|
; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
|
; GFX12-NEXT: v_fma_f64 v[2:3], -s[0:1], -s[2:3], v[2:3]
|
|
; GFX12-NEXT: global_store_b64 v[0:1], v[2:3], off
|
|
; GFX12-NEXT: s_setpc_b64 s[30:31]
|
|
%neg.x = fneg double %x
|
|
%neg.y = fneg double %y
|
|
%val = call double @llvm.experimental.constrained.fma.f64(double %neg.x, double %neg.y, double %z, metadata !"round.tonearest", metadata !"fpexcept.strict")
|
|
store double %val, ptr addrspace(1) %out
|
|
ret void
|
|
}
|
|
|
|
define void @v_constained_fma_f64_fpexcept_strict_fneg_fneg_div(double %x, double %y, double %z, ptr addrspace(1) %out) #0 {
|
|
; GFX8-LABEL: v_constained_fma_f64_fpexcept_strict_fneg_fneg_div:
|
|
; GFX8: ; %bb.0:
|
|
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX8-NEXT: v_fma_f64 v[0:1], -v[0:1], -v[2:3], v[4:5]
|
|
; GFX8-NEXT: flat_store_dwordx2 v[6:7], v[0:1]
|
|
; GFX8-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX8-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX9-LABEL: v_constained_fma_f64_fpexcept_strict_fneg_fneg_div:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: v_fma_f64 v[0:1], -v[0:1], -v[2:3], v[4:5]
|
|
; GFX9-NEXT: global_store_dwordx2 v[6:7], v[0:1], off
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX11-LABEL: v_constained_fma_f64_fpexcept_strict_fneg_fneg_div:
|
|
; GFX11: ; %bb.0:
|
|
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX11-NEXT: v_fma_f64 v[0:1], -v[0:1], -v[2:3], v[4:5]
|
|
; GFX11-NEXT: global_store_b64 v[6:7], v[0:1], off
|
|
; GFX11-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX12-LABEL: v_constained_fma_f64_fpexcept_strict_fneg_fneg_div:
|
|
; GFX12: ; %bb.0:
|
|
; GFX12-NEXT: s_wait_loadcnt_dscnt 0x0
|
|
; GFX12-NEXT: s_wait_expcnt 0x0
|
|
; GFX12-NEXT: s_wait_samplecnt 0x0
|
|
; GFX12-NEXT: s_wait_bvhcnt 0x0
|
|
; GFX12-NEXT: s_wait_kmcnt 0x0
|
|
; GFX12-NEXT: v_fma_f64 v[0:1], -v[0:1], -v[2:3], v[4:5]
|
|
; GFX12-NEXT: global_store_b64 v[6:7], v[0:1], off
|
|
; GFX12-NEXT: s_setpc_b64 s[30:31]
|
|
%neg.x = fneg double %x
|
|
%neg.y = fneg double %y
|
|
%val = call double @llvm.experimental.constrained.fma.f64(double %neg.x, double %neg.y, double %z, metadata !"round.tonearest", metadata !"fpexcept.strict")
|
|
store double %val, ptr addrspace(1) %out
|
|
ret void
|
|
}
|
|
|
|
define void @v_constained_fma_f64_fpexcept_strict_fabs_fabs_uni(double inreg %x, double inreg %y, double inreg %z, ptr addrspace(1) %out) #0 {
|
|
; GFX8-LABEL: v_constained_fma_f64_fpexcept_strict_fabs_fabs_uni:
|
|
; GFX8: ; %bb.0:
|
|
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX8-NEXT: v_mov_b32_e32 v2, s18
|
|
; GFX8-NEXT: v_mov_b32_e32 v4, s20
|
|
; GFX8-NEXT: v_mov_b32_e32 v3, s19
|
|
; GFX8-NEXT: v_mov_b32_e32 v5, s21
|
|
; GFX8-NEXT: v_fma_f64 v[2:3], |s[16:17]|, |v[2:3]|, v[4:5]
|
|
; GFX8-NEXT: flat_store_dwordx2 v[0:1], v[2:3]
|
|
; GFX8-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX8-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX900-LABEL: v_constained_fma_f64_fpexcept_strict_fabs_fabs_uni:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, s18
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, s20
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, s19
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, s21
|
|
; GFX900-NEXT: v_fma_f64 v[2:3], |s[16:17]|, |v[2:3]|, v[4:5]
|
|
; GFX900-NEXT: global_store_dwordx2 v[0:1], v[2:3], off
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_constained_fma_f64_fpexcept_strict_fabs_fabs_uni:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b64_e32 v[2:3], s[2:3]
|
|
; GFX942-NEXT: v_mov_b64_e32 v[4:5], s[16:17]
|
|
; GFX942-NEXT: v_fma_f64 v[2:3], |s[0:1]|, |v[2:3]|, v[4:5]
|
|
; GFX942-NEXT: global_store_dwordx2 v[0:1], v[2:3], off
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX11-LABEL: v_constained_fma_f64_fpexcept_strict_fabs_fabs_uni:
|
|
; GFX11: ; %bb.0:
|
|
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX11-NEXT: v_dual_mov_b32 v2, s16 :: v_dual_mov_b32 v3, s17
|
|
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
|
; GFX11-NEXT: v_fma_f64 v[2:3], |s[0:1]|, |s[2:3]|, v[2:3]
|
|
; GFX11-NEXT: global_store_b64 v[0:1], v[2:3], off
|
|
; GFX11-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX12-LABEL: v_constained_fma_f64_fpexcept_strict_fabs_fabs_uni:
|
|
; GFX12: ; %bb.0:
|
|
; GFX12-NEXT: s_wait_loadcnt_dscnt 0x0
|
|
; GFX12-NEXT: s_wait_expcnt 0x0
|
|
; GFX12-NEXT: s_wait_samplecnt 0x0
|
|
; GFX12-NEXT: s_wait_bvhcnt 0x0
|
|
; GFX12-NEXT: s_wait_kmcnt 0x0
|
|
; GFX12-NEXT: v_dual_mov_b32 v2, s16 :: v_dual_mov_b32 v3, s17
|
|
; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_1)
|
|
; GFX12-NEXT: v_fma_f64 v[2:3], |s[0:1]|, |s[2:3]|, v[2:3]
|
|
; GFX12-NEXT: global_store_b64 v[0:1], v[2:3], off
|
|
; GFX12-NEXT: s_setpc_b64 s[30:31]
|
|
%neg.x = call double @llvm.fabs.f64(double %x) #0
|
|
%neg.y = call double @llvm.fabs.f64(double %y) #0
|
|
%val = call double @llvm.experimental.constrained.fma.f64(double %neg.x, double %neg.y, double %z, metadata !"round.tonearest", metadata !"fpexcept.strict")
|
|
store double %val, ptr addrspace(1) %out
|
|
ret void
|
|
}
|
|
|
|
define void @v_constained_fma_f64_fpexcept_strict_fabs_fabs_div(double %x, double %y, double %z, ptr addrspace(1) %out) #0 {
|
|
; GFX8-LABEL: v_constained_fma_f64_fpexcept_strict_fabs_fabs_div:
|
|
; GFX8: ; %bb.0:
|
|
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX8-NEXT: v_fma_f64 v[0:1], |v[0:1]|, |v[2:3]|, v[4:5]
|
|
; GFX8-NEXT: flat_store_dwordx2 v[6:7], v[0:1]
|
|
; GFX8-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX8-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX9-LABEL: v_constained_fma_f64_fpexcept_strict_fabs_fabs_div:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: v_fma_f64 v[0:1], |v[0:1]|, |v[2:3]|, v[4:5]
|
|
; GFX9-NEXT: global_store_dwordx2 v[6:7], v[0:1], off
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX11-LABEL: v_constained_fma_f64_fpexcept_strict_fabs_fabs_div:
|
|
; GFX11: ; %bb.0:
|
|
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX11-NEXT: v_fma_f64 v[0:1], |v[0:1]|, |v[2:3]|, v[4:5]
|
|
; GFX11-NEXT: global_store_b64 v[6:7], v[0:1], off
|
|
; GFX11-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX12-LABEL: v_constained_fma_f64_fpexcept_strict_fabs_fabs_div:
|
|
; GFX12: ; %bb.0:
|
|
; GFX12-NEXT: s_wait_loadcnt_dscnt 0x0
|
|
; GFX12-NEXT: s_wait_expcnt 0x0
|
|
; GFX12-NEXT: s_wait_samplecnt 0x0
|
|
; GFX12-NEXT: s_wait_bvhcnt 0x0
|
|
; GFX12-NEXT: s_wait_kmcnt 0x0
|
|
; GFX12-NEXT: v_fma_f64 v[0:1], |v[0:1]|, |v[2:3]|, v[4:5]
|
|
; GFX12-NEXT: global_store_b64 v[6:7], v[0:1], off
|
|
; GFX12-NEXT: s_setpc_b64 s[30:31]
|
|
%neg.x = call double @llvm.fabs.f64(double %x) #0
|
|
%neg.y = call double @llvm.fabs.f64(double %y) #0
|
|
%val = call double @llvm.experimental.constrained.fma.f64(double %neg.x, double %neg.y, double %z, metadata !"round.tonearest", metadata !"fpexcept.strict")
|
|
store double %val, ptr addrspace(1) %out
|
|
ret void
|
|
}
|
|
|
|
define void @v_constained_fma_v2f64_fpexcept_strict_fneg_fneg_uni(<2 x double> inreg %x, <2 x double> inreg %y, <2 x double> inreg %z, ptr addrspace(1) %out) #0 {
|
|
; GFX8-LABEL: v_constained_fma_v2f64_fpexcept_strict_fneg_fneg_uni:
|
|
; GFX8: ; %bb.0:
|
|
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX8-NEXT: v_mov_b32_e32 v2, s20
|
|
; GFX8-NEXT: v_mov_b32_e32 v6, s24
|
|
; GFX8-NEXT: v_mov_b32_e32 v3, s21
|
|
; GFX8-NEXT: v_mov_b32_e32 v7, s25
|
|
; GFX8-NEXT: v_mov_b32_e32 v4, s22
|
|
; GFX8-NEXT: v_fma_f64 v[2:3], -s[16:17], -v[2:3], v[6:7]
|
|
; GFX8-NEXT: v_mov_b32_e32 v6, s26
|
|
; GFX8-NEXT: v_mov_b32_e32 v5, s23
|
|
; GFX8-NEXT: v_mov_b32_e32 v7, s27
|
|
; GFX8-NEXT: v_fma_f64 v[4:5], -s[18:19], -v[4:5], v[6:7]
|
|
; GFX8-NEXT: v_readfirstlane_b32 s4, v2
|
|
; GFX8-NEXT: v_readfirstlane_b32 s5, v3
|
|
; GFX8-NEXT: v_readfirstlane_b32 s6, v4
|
|
; GFX8-NEXT: v_readfirstlane_b32 s7, v5
|
|
; GFX8-NEXT: v_mov_b32_e32 v2, s4
|
|
; GFX8-NEXT: v_mov_b32_e32 v3, s5
|
|
; GFX8-NEXT: v_mov_b32_e32 v4, s6
|
|
; GFX8-NEXT: v_mov_b32_e32 v5, s7
|
|
; GFX8-NEXT: flat_store_dwordx4 v[0:1], v[2:5]
|
|
; GFX8-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX8-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX900-LABEL: v_constained_fma_v2f64_fpexcept_strict_fneg_fneg_uni:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, s20
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, s24
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, s21
|
|
; GFX900-NEXT: v_mov_b32_e32 v7, s25
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, s22
|
|
; GFX900-NEXT: v_fma_f64 v[2:3], -s[16:17], -v[2:3], v[6:7]
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, s26
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, s23
|
|
; GFX900-NEXT: v_mov_b32_e32 v7, s27
|
|
; GFX900-NEXT: v_fma_f64 v[4:5], -s[18:19], -v[4:5], v[6:7]
|
|
; GFX900-NEXT: v_readfirstlane_b32 s4, v2
|
|
; GFX900-NEXT: v_readfirstlane_b32 s5, v3
|
|
; GFX900-NEXT: v_readfirstlane_b32 s6, v4
|
|
; GFX900-NEXT: v_readfirstlane_b32 s7, v5
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, s4
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, s5
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, s6
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, s7
|
|
; GFX900-NEXT: global_store_dwordx4 v[0:1], v[2:5], off
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_constained_fma_v2f64_fpexcept_strict_fneg_fneg_uni:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b64_e32 v[2:3], s[16:17]
|
|
; GFX942-NEXT: v_mov_b64_e32 v[6:7], s[20:21]
|
|
; GFX942-NEXT: v_fma_f64 v[2:3], -s[0:1], -v[2:3], v[6:7]
|
|
; GFX942-NEXT: v_mov_b64_e32 v[4:5], s[18:19]
|
|
; GFX942-NEXT: v_readfirstlane_b32 s0, v2
|
|
; GFX942-NEXT: v_readfirstlane_b32 s1, v3
|
|
; GFX942-NEXT: v_mov_b64_e32 v[2:3], s[22:23]
|
|
; GFX942-NEXT: v_fma_f64 v[2:3], -s[2:3], -v[4:5], v[2:3]
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_readfirstlane_b32 s2, v2
|
|
; GFX942-NEXT: v_readfirstlane_b32 s3, v3
|
|
; GFX942-NEXT: s_nop 1
|
|
; GFX942-NEXT: v_mov_b64_e32 v[4:5], s[2:3]
|
|
; GFX942-NEXT: v_mov_b64_e32 v[2:3], s[0:1]
|
|
; GFX942-NEXT: global_store_dwordx4 v[0:1], v[2:5], off
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX11-LABEL: v_constained_fma_v2f64_fpexcept_strict_fneg_fneg_uni:
|
|
; GFX11: ; %bb.0:
|
|
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX11-NEXT: v_dual_mov_b32 v4, s22 :: v_dual_mov_b32 v5, s23
|
|
; GFX11-NEXT: v_dual_mov_b32 v2, s20 :: v_dual_mov_b32 v3, s21
|
|
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
|
|
; GFX11-NEXT: v_fma_f64 v[4:5], -s[2:3], -s[18:19], v[4:5]
|
|
; GFX11-NEXT: v_fma_f64 v[2:3], -s[0:1], -s[16:17], v[2:3]
|
|
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3)
|
|
; GFX11-NEXT: v_readfirstlane_b32 s3, v5
|
|
; GFX11-NEXT: v_readfirstlane_b32 s2, v4
|
|
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_4)
|
|
; GFX11-NEXT: v_readfirstlane_b32 s0, v2
|
|
; GFX11-NEXT: v_readfirstlane_b32 s1, v3
|
|
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_2)
|
|
; GFX11-NEXT: v_dual_mov_b32 v5, s3 :: v_dual_mov_b32 v4, s2
|
|
; GFX11-NEXT: v_dual_mov_b32 v2, s0 :: v_dual_mov_b32 v3, s1
|
|
; GFX11-NEXT: global_store_b128 v[0:1], v[2:5], off
|
|
; GFX11-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX12-LABEL: v_constained_fma_v2f64_fpexcept_strict_fneg_fneg_uni:
|
|
; GFX12: ; %bb.0:
|
|
; GFX12-NEXT: s_wait_loadcnt_dscnt 0x0
|
|
; GFX12-NEXT: s_wait_expcnt 0x0
|
|
; GFX12-NEXT: s_wait_samplecnt 0x0
|
|
; GFX12-NEXT: s_wait_bvhcnt 0x0
|
|
; GFX12-NEXT: s_wait_kmcnt 0x0
|
|
; GFX12-NEXT: v_dual_mov_b32 v4, s22 :: v_dual_mov_b32 v5, s23
|
|
; GFX12-NEXT: v_dual_mov_b32 v2, s20 :: v_dual_mov_b32 v3, s21
|
|
; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
|
|
; GFX12-NEXT: v_fma_f64 v[4:5], -s[2:3], -s[18:19], v[4:5]
|
|
; GFX12-NEXT: v_fma_f64 v[2:3], -s[0:1], -s[16:17], v[2:3]
|
|
; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3)
|
|
; GFX12-NEXT: v_readfirstlane_b32 s3, v5
|
|
; GFX12-NEXT: v_readfirstlane_b32 s2, v4
|
|
; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_4)
|
|
; GFX12-NEXT: v_readfirstlane_b32 s0, v2
|
|
; GFX12-NEXT: v_readfirstlane_b32 s1, v3
|
|
; GFX12-NEXT: s_wait_alu depctr_va_sdst(0)
|
|
; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_2)
|
|
; GFX12-NEXT: v_dual_mov_b32 v5, s3 :: v_dual_mov_b32 v4, s2
|
|
; GFX12-NEXT: v_dual_mov_b32 v2, s0 :: v_dual_mov_b32 v3, s1
|
|
; GFX12-NEXT: global_store_b128 v[0:1], v[2:5], off
|
|
; GFX12-NEXT: s_setpc_b64 s[30:31]
|
|
%neg.x = fneg <2 x double> %x
|
|
%neg.y = fneg <2 x double> %y
|
|
%val = call <2 x double> @llvm.experimental.constrained.fma.v2f64(<2 x double> %neg.x, <2 x double> %neg.y, <2 x double> %z, metadata !"round.tonearest", metadata !"fpexcept.strict")
|
|
store <2 x double> %val, ptr addrspace(1) %out
|
|
ret void
|
|
}
|
|
|
|
define void @v_constained_fma_v2f64_fpexcept_strict_fneg_fneg_div(<2 x double> %x, <2 x double> %y, <2 x double> %z, ptr addrspace(1) %out) #0 {
|
|
; GFX8-LABEL: v_constained_fma_v2f64_fpexcept_strict_fneg_fneg_div:
|
|
; GFX8: ; %bb.0:
|
|
; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX8-NEXT: v_fma_f64 v[0:1], -v[0:1], -v[4:5], v[8:9]
|
|
; GFX8-NEXT: v_fma_f64 v[2:3], -v[2:3], -v[6:7], v[10:11]
|
|
; GFX8-NEXT: flat_store_dwordx4 v[12:13], v[0:3]
|
|
; GFX8-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX8-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX9-LABEL: v_constained_fma_v2f64_fpexcept_strict_fneg_fneg_div:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: v_fma_f64 v[0:1], -v[0:1], -v[4:5], v[8:9]
|
|
; GFX9-NEXT: v_fma_f64 v[2:3], -v[2:3], -v[6:7], v[10:11]
|
|
; GFX9-NEXT: global_store_dwordx4 v[12:13], v[0:3], off
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX11-LABEL: v_constained_fma_v2f64_fpexcept_strict_fneg_fneg_div:
|
|
; GFX11: ; %bb.0:
|
|
; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX11-NEXT: v_fma_f64 v[0:1], -v[0:1], -v[4:5], v[8:9]
|
|
; GFX11-NEXT: v_fma_f64 v[2:3], -v[2:3], -v[6:7], v[10:11]
|
|
; GFX11-NEXT: global_store_b128 v[12:13], v[0:3], off
|
|
; GFX11-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX12-LABEL: v_constained_fma_v2f64_fpexcept_strict_fneg_fneg_div:
|
|
; GFX12: ; %bb.0:
|
|
; GFX12-NEXT: s_wait_loadcnt_dscnt 0x0
|
|
; GFX12-NEXT: s_wait_expcnt 0x0
|
|
; GFX12-NEXT: s_wait_samplecnt 0x0
|
|
; GFX12-NEXT: s_wait_bvhcnt 0x0
|
|
; GFX12-NEXT: s_wait_kmcnt 0x0
|
|
; GFX12-NEXT: v_fma_f64 v[0:1], -v[0:1], -v[4:5], v[8:9]
|
|
; GFX12-NEXT: v_fma_f64 v[2:3], -v[2:3], -v[6:7], v[10:11]
|
|
; GFX12-NEXT: global_store_b128 v[12:13], v[0:3], off
|
|
; GFX12-NEXT: s_setpc_b64 s[30:31]
|
|
%neg.x = fneg <2 x double> %x
|
|
%neg.y = fneg <2 x double> %y
|
|
%val = call <2 x double> @llvm.experimental.constrained.fma.v2f64(<2 x double> %neg.x, <2 x double> %neg.y, <2 x double> %z, metadata !"round.tonearest", metadata !"fpexcept.strict")
|
|
store <2 x double> %val, ptr addrspace(1) %out
|
|
ret void
|
|
}
|
|
|
|
declare double @llvm.fabs.f64(double)
|
|
declare double @llvm.experimental.constrained.fma.f64(double, double, double, metadata, metadata)
|
|
declare <2 x double> @llvm.experimental.constrained.fma.v2f64(<2 x double>, <2 x double>, <2 x double>, metadata, metadata)
|
|
declare <3 x double> @llvm.experimental.constrained.fma.v3f64(<3 x double>, <3 x double>, <3 x double>, metadata, metadata)
|
|
declare <4 x double> @llvm.experimental.constrained.fma.v4f64(<4 x double>, <4 x double>, <4 x double>, metadata, metadata)
|
|
|
|
attributes #0 = { strictfp }
|