; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc -global-isel=0 -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx900 < %s | FileCheck -check-prefixes=GFX9,GFX9-SDAG %s
; RUN: llc -global-isel=1 -new-reg-bank-select -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx900 < %s | FileCheck -check-prefixes=GFX9,GFX9-GISEL %s

; RUN: llc -global-isel=0 -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1010 < %s | FileCheck -check-prefixes=GFX10PLUS,GFX10-SDAG %s
; RUN: llc -global-isel=1 -new-reg-bank-select -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1010 < %s | FileCheck -check-prefixes=GFX10PLUS,GFX10-GISEL %s

; RUN: llc -global-isel=0 -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1100 -amdgpu-enable-delay-alu=0 < %s | FileCheck -check-prefixes=GFX10PLUS,GFX11-SDAG %s
; RUN: llc -global-isel=1 -new-reg-bank-select -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1100 -amdgpu-enable-delay-alu=0 < %s | FileCheck -check-prefixes=GFX10PLUS,GFX11-GISEL %s

; RUN: llc -global-isel=0 -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1200 < %s | FileCheck -check-prefixes=GFX12,GFX12-SDAG %s
; RUN: llc -global-isel=1 -new-reg-bank-select -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1200 < %s | FileCheck -check-prefixes=GFX12,GFX12-GISEL %s

define float @v_constained_fadd_f32_fpexcept_strict(float %x, float %y) #0 {
; GFX9-LABEL: v_constained_fadd_f32_fpexcept_strict:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    v_add_f32_e32 v0, v0, v1
; GFX9-NEXT:    s_setpc_b64 s[30:31]
;
; GFX10PLUS-LABEL: v_constained_fadd_f32_fpexcept_strict:
; GFX10PLUS:       ; %bb.0:
; GFX10PLUS-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10PLUS-NEXT:    v_add_f32_e32 v0, v0, v1
; GFX10PLUS-NEXT:    s_setpc_b64 s[30:31]
;
; GFX12-LABEL: v_constained_fadd_f32_fpexcept_strict:
; 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_add_f32_e32 v0, v0, v1
; GFX12-NEXT:    s_setpc_b64 s[30:31]
; GCN-SDAG-LABEL: v_constained_fadd_f32_fpexcept_strict:
; GCN-SDAG:       ; %bb.0:
; GCN-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GCN-SDAG-NEXT:    v_add_f32_e32 v0, v0, v1
; GCN-SDAG-NEXT:    s_setpc_b64 s[30:31]
; GCN-GISEL-LABEL: v_constained_fadd_f32_fpexcept_strict:
; GCN-GISEL:       ; %bb.0:
; GCN-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GCN-GISEL-NEXT:    v_add_f32_e32 v0, v0, v1
; GCN-GISEL-NEXT:    s_setpc_b64 s[30:31]
  %val = call float @llvm.experimental.constrained.fadd.f32(float %x, float %y, metadata !"round.tonearest", metadata !"fpexcept.strict")
  ret float %val
}

define float @v_constained_fadd_f32_fpexcept_ignore(float %x, float %y) #0 {
; GFX9-LABEL: v_constained_fadd_f32_fpexcept_ignore:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    v_add_f32_e32 v0, v0, v1
; GFX9-NEXT:    s_setpc_b64 s[30:31]
;
; GFX10PLUS-LABEL: v_constained_fadd_f32_fpexcept_ignore:
; GFX10PLUS:       ; %bb.0:
; GFX10PLUS-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10PLUS-NEXT:    v_add_f32_e32 v0, v0, v1
; GFX10PLUS-NEXT:    s_setpc_b64 s[30:31]
;
; GFX12-LABEL: v_constained_fadd_f32_fpexcept_ignore:
; 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_add_f32_e32 v0, v0, v1
; GFX12-NEXT:    s_setpc_b64 s[30:31]
; GCN-SDAG-LABEL: v_constained_fadd_f32_fpexcept_ignore:
; GCN-SDAG:       ; %bb.0:
; GCN-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GCN-SDAG-NEXT:    v_add_f32_e32 v0, v0, v1
; GCN-SDAG-NEXT:    s_setpc_b64 s[30:31]
; GCN-GISEL-LABEL: v_constained_fadd_f32_fpexcept_ignore:
; GCN-GISEL:       ; %bb.0:
; GCN-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GCN-GISEL-NEXT:    v_add_f32_e32 v0, v0, v1
; GCN-GISEL-NEXT:    s_setpc_b64 s[30:31]
  %val = call float @llvm.experimental.constrained.fadd.f32(float %x, float %y, metadata !"round.tonearest", metadata !"fpexcept.ignore")
  ret float %val
}

define float @v_constained_fadd_f32_fpexcept_maytrap(float %x, float %y) #0 {
; GFX9-LABEL: v_constained_fadd_f32_fpexcept_maytrap:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    v_add_f32_e32 v0, v0, v1
; GFX9-NEXT:    s_setpc_b64 s[30:31]
;
; GFX10PLUS-LABEL: v_constained_fadd_f32_fpexcept_maytrap:
; GFX10PLUS:       ; %bb.0:
; GFX10PLUS-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10PLUS-NEXT:    v_add_f32_e32 v0, v0, v1
; GFX10PLUS-NEXT:    s_setpc_b64 s[30:31]
;
; GFX12-LABEL: v_constained_fadd_f32_fpexcept_maytrap:
; 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_add_f32_e32 v0, v0, v1
; GFX12-NEXT:    s_setpc_b64 s[30:31]
; GCN-SDAG-LABEL: v_constained_fadd_f32_fpexcept_maytrap:
; GCN-SDAG:       ; %bb.0:
; GCN-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GCN-SDAG-NEXT:    v_add_f32_e32 v0, v0, v1
; GCN-SDAG-NEXT:    s_setpc_b64 s[30:31]
; GCN-GISEL-LABEL: v_constained_fadd_f32_fpexcept_maytrap:
; GCN-GISEL:       ; %bb.0:
; GCN-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GCN-GISEL-NEXT:    v_add_f32_e32 v0, v0, v1
; GCN-GISEL-NEXT:    s_setpc_b64 s[30:31]
  %val = call float @llvm.experimental.constrained.fadd.f32(float %x, float %y, metadata !"round.tonearest", metadata !"fpexcept.maytrap")
  ret float %val
}

define <2 x float> @v_constained_fadd_v2f32_fpexcept_strict(<2 x float> %x, <2 x float> %y) #0 {
; GFX9-LABEL: v_constained_fadd_v2f32_fpexcept_strict:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    v_add_f32_e32 v0, v0, v2
; GFX9-NEXT:    v_add_f32_e32 v1, v1, v3
; GFX9-NEXT:    s_setpc_b64 s[30:31]
;
; GFX10-SDAG-LABEL: v_constained_fadd_v2f32_fpexcept_strict:
; GFX10-SDAG:       ; %bb.0:
; GFX10-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-SDAG-NEXT:    v_add_f32_e32 v0, v0, v2
; GFX10-SDAG-NEXT:    v_add_f32_e32 v1, v1, v3
; GFX10-SDAG-NEXT:    s_setpc_b64 s[30:31]
;
; GFX10-GISEL-LABEL: v_constained_fadd_v2f32_fpexcept_strict:
; GFX10-GISEL:       ; %bb.0:
; GFX10-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-GISEL-NEXT:    v_add_f32_e32 v0, v0, v2
; GFX10-GISEL-NEXT:    v_add_f32_e32 v1, v1, v3
; GFX10-GISEL-NEXT:    s_setpc_b64 s[30:31]
;
; GFX11-SDAG-LABEL: v_constained_fadd_v2f32_fpexcept_strict:
; GFX11-SDAG:       ; %bb.0:
; GFX11-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-SDAG-NEXT:    v_dual_add_f32 v0, v0, v2 :: v_dual_add_f32 v1, v1, v3
; GFX11-SDAG-NEXT:    s_setpc_b64 s[30:31]
;
; GFX11-GISEL-LABEL: v_constained_fadd_v2f32_fpexcept_strict:
; GFX11-GISEL:       ; %bb.0:
; GFX11-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-GISEL-NEXT:    v_dual_add_f32 v0, v0, v2 :: v_dual_add_f32 v1, v1, v3
; GFX11-GISEL-NEXT:    s_setpc_b64 s[30:31]
;
; GFX12-LABEL: v_constained_fadd_v2f32_fpexcept_strict:
; 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_add_f32 v0, v0, v2 :: v_dual_add_f32 v1, v1, v3
; GFX12-NEXT:    s_setpc_b64 s[30:31]
; GFX10-LABEL: v_constained_fadd_v2f32_fpexcept_strict:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    v_add_f32_e32 v0, v0, v2
; GFX10-NEXT:    v_add_f32_e32 v1, v1, v3
; GFX10-NEXT:    s_setpc_b64 s[30:31]
; GFX11-LABEL: v_constained_fadd_v2f32_fpexcept_strict:
; GFX11:       ; %bb.0:
; GFX11-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-NEXT:    v_dual_add_f32 v0, v0, v2 :: v_dual_add_f32 v1, v1, v3
; GFX11-NEXT:    s_setpc_b64 s[30:31]
; GCN-SDAG-LABEL: v_constained_fadd_v2f32_fpexcept_strict:
; GCN-SDAG:       ; %bb.0:
; GCN-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GCN-SDAG-NEXT:    v_add_f32_e32 v0, v0, v2
; GCN-SDAG-NEXT:    v_add_f32_e32 v1, v1, v3
; GCN-SDAG-NEXT:    s_setpc_b64 s[30:31]
; GCN-GISEL-LABEL: v_constained_fadd_v2f32_fpexcept_strict:
; GCN-GISEL:       ; %bb.0:
; GCN-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GCN-GISEL-NEXT:    v_add_f32_e32 v0, v0, v2
; GCN-GISEL-NEXT:    v_add_f32_e32 v1, v1, v3
; GCN-GISEL-NEXT:    s_setpc_b64 s[30:31]
  %val = call <2 x float> @llvm.experimental.constrained.fadd.v2f32(<2 x float> %x, <2 x float> %y, metadata !"round.tonearest", metadata !"fpexcept.strict")
  ret <2 x float> %val
}

define <2 x float> @v_constained_fadd_v2f32_fpexcept_ignore(<2 x float> %x, <2 x float> %y) #0 {
; GFX9-LABEL: v_constained_fadd_v2f32_fpexcept_ignore:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    v_add_f32_e32 v0, v0, v2
; GFX9-NEXT:    v_add_f32_e32 v1, v1, v3
; GFX9-NEXT:    s_setpc_b64 s[30:31]
;
; GFX10-SDAG-LABEL: v_constained_fadd_v2f32_fpexcept_ignore:
; GFX10-SDAG:       ; %bb.0:
; GFX10-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-SDAG-NEXT:    v_add_f32_e32 v0, v0, v2
; GFX10-SDAG-NEXT:    v_add_f32_e32 v1, v1, v3
; GFX10-SDAG-NEXT:    s_setpc_b64 s[30:31]
;
; GFX10-GISEL-LABEL: v_constained_fadd_v2f32_fpexcept_ignore:
; GFX10-GISEL:       ; %bb.0:
; GFX10-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-GISEL-NEXT:    v_add_f32_e32 v0, v0, v2
; GFX10-GISEL-NEXT:    v_add_f32_e32 v1, v1, v3
; GFX10-GISEL-NEXT:    s_setpc_b64 s[30:31]
;
; GFX11-SDAG-LABEL: v_constained_fadd_v2f32_fpexcept_ignore:
; GFX11-SDAG:       ; %bb.0:
; GFX11-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-SDAG-NEXT:    v_dual_add_f32 v0, v0, v2 :: v_dual_add_f32 v1, v1, v3
; GFX11-SDAG-NEXT:    s_setpc_b64 s[30:31]
;
; GFX11-GISEL-LABEL: v_constained_fadd_v2f32_fpexcept_ignore:
; GFX11-GISEL:       ; %bb.0:
; GFX11-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-GISEL-NEXT:    v_dual_add_f32 v0, v0, v2 :: v_dual_add_f32 v1, v1, v3
; GFX11-GISEL-NEXT:    s_setpc_b64 s[30:31]
;
; GFX12-LABEL: v_constained_fadd_v2f32_fpexcept_ignore:
; 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_add_f32 v0, v0, v2 :: v_dual_add_f32 v1, v1, v3
; GFX12-NEXT:    s_setpc_b64 s[30:31]
; GFX10-LABEL: v_constained_fadd_v2f32_fpexcept_ignore:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    v_add_f32_e32 v0, v0, v2
; GFX10-NEXT:    v_add_f32_e32 v1, v1, v3
; GFX10-NEXT:    s_setpc_b64 s[30:31]
; GFX11-LABEL: v_constained_fadd_v2f32_fpexcept_ignore:
; GFX11:       ; %bb.0:
; GFX11-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-NEXT:    v_dual_add_f32 v0, v0, v2 :: v_dual_add_f32 v1, v1, v3
; GFX11-NEXT:    s_setpc_b64 s[30:31]
; GCN-SDAG-LABEL: v_constained_fadd_v2f32_fpexcept_ignore:
; GCN-SDAG:       ; %bb.0:
; GCN-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GCN-SDAG-NEXT:    v_add_f32_e32 v0, v0, v2
; GCN-SDAG-NEXT:    v_add_f32_e32 v1, v1, v3
; GCN-SDAG-NEXT:    s_setpc_b64 s[30:31]
; GCN-GISEL-LABEL: v_constained_fadd_v2f32_fpexcept_ignore:
; GCN-GISEL:       ; %bb.0:
; GCN-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GCN-GISEL-NEXT:    v_add_f32_e32 v0, v0, v2
; GCN-GISEL-NEXT:    v_add_f32_e32 v1, v1, v3
; GCN-GISEL-NEXT:    s_setpc_b64 s[30:31]
  %val = call <2 x float> @llvm.experimental.constrained.fadd.v2f32(<2 x float> %x, <2 x float> %y, metadata !"round.tonearest", metadata !"fpexcept.ignore")
  ret <2 x float> %val
}

define <2 x float> @v_constained_fadd_v2f32_fpexcept_maytrap(<2 x float> %x, <2 x float> %y) #0 {
; GFX9-LABEL: v_constained_fadd_v2f32_fpexcept_maytrap:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    v_add_f32_e32 v0, v0, v2
; GFX9-NEXT:    v_add_f32_e32 v1, v1, v3
; GFX9-NEXT:    s_setpc_b64 s[30:31]
;
; GFX10-SDAG-LABEL: v_constained_fadd_v2f32_fpexcept_maytrap:
; GFX10-SDAG:       ; %bb.0:
; GFX10-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-SDAG-NEXT:    v_add_f32_e32 v0, v0, v2
; GFX10-SDAG-NEXT:    v_add_f32_e32 v1, v1, v3
; GFX10-SDAG-NEXT:    s_setpc_b64 s[30:31]
;
; GFX10-GISEL-LABEL: v_constained_fadd_v2f32_fpexcept_maytrap:
; GFX10-GISEL:       ; %bb.0:
; GFX10-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-GISEL-NEXT:    v_add_f32_e32 v0, v0, v2
; GFX10-GISEL-NEXT:    v_add_f32_e32 v1, v1, v3
; GFX10-GISEL-NEXT:    s_setpc_b64 s[30:31]
;
; GFX11-SDAG-LABEL: v_constained_fadd_v2f32_fpexcept_maytrap:
; GFX11-SDAG:       ; %bb.0:
; GFX11-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-SDAG-NEXT:    v_dual_add_f32 v0, v0, v2 :: v_dual_add_f32 v1, v1, v3
; GFX11-SDAG-NEXT:    s_setpc_b64 s[30:31]
;
; GFX11-GISEL-LABEL: v_constained_fadd_v2f32_fpexcept_maytrap:
; GFX11-GISEL:       ; %bb.0:
; GFX11-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-GISEL-NEXT:    v_dual_add_f32 v0, v0, v2 :: v_dual_add_f32 v1, v1, v3
; GFX11-GISEL-NEXT:    s_setpc_b64 s[30:31]
;
; GFX12-LABEL: v_constained_fadd_v2f32_fpexcept_maytrap:
; 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_add_f32 v0, v0, v2 :: v_dual_add_f32 v1, v1, v3
; GFX12-NEXT:    s_setpc_b64 s[30:31]
; GFX10-LABEL: v_constained_fadd_v2f32_fpexcept_maytrap:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    v_add_f32_e32 v0, v0, v2
; GFX10-NEXT:    v_add_f32_e32 v1, v1, v3
; GFX10-NEXT:    s_setpc_b64 s[30:31]
; GFX11-LABEL: v_constained_fadd_v2f32_fpexcept_maytrap:
; GFX11:       ; %bb.0:
; GFX11-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-NEXT:    v_dual_add_f32 v0, v0, v2 :: v_dual_add_f32 v1, v1, v3
; GFX11-NEXT:    s_setpc_b64 s[30:31]
; GCN-SDAG-LABEL: v_constained_fadd_v2f32_fpexcept_maytrap:
; GCN-SDAG:       ; %bb.0:
; GCN-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GCN-SDAG-NEXT:    v_add_f32_e32 v0, v0, v2
; GCN-SDAG-NEXT:    v_add_f32_e32 v1, v1, v3
; GCN-SDAG-NEXT:    s_setpc_b64 s[30:31]
; GCN-GISEL-LABEL: v_constained_fadd_v2f32_fpexcept_maytrap:
; GCN-GISEL:       ; %bb.0:
; GCN-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GCN-GISEL-NEXT:    v_add_f32_e32 v0, v0, v2
; GCN-GISEL-NEXT:    v_add_f32_e32 v1, v1, v3
; GCN-GISEL-NEXT:    s_setpc_b64 s[30:31]
  %val = call <2 x float> @llvm.experimental.constrained.fadd.v2f32(<2 x float> %x, <2 x float> %y, metadata !"round.tonearest", metadata !"fpexcept.maytrap")
  ret <2 x float> %val
}

define <3 x float> @v_constained_fadd_v3f32_fpexcept_strict(<3 x float> %x, <3 x float> %y) #0 {
; GFX9-LABEL: v_constained_fadd_v3f32_fpexcept_strict:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    v_add_f32_e32 v0, v0, v3
; GFX9-NEXT:    v_add_f32_e32 v1, v1, v4
; GFX9-NEXT:    v_add_f32_e32 v2, v2, v5
; GFX9-NEXT:    s_setpc_b64 s[30:31]
;
; GFX10-SDAG-LABEL: v_constained_fadd_v3f32_fpexcept_strict:
; GFX10-SDAG:       ; %bb.0:
; GFX10-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-SDAG-NEXT:    v_add_f32_e32 v0, v0, v3
; GFX10-SDAG-NEXT:    v_add_f32_e32 v1, v1, v4
; GFX10-SDAG-NEXT:    v_add_f32_e32 v2, v2, v5
; GFX10-SDAG-NEXT:    s_setpc_b64 s[30:31]
;
; GFX10-GISEL-LABEL: v_constained_fadd_v3f32_fpexcept_strict:
; GFX10-GISEL:       ; %bb.0:
; GFX10-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-GISEL-NEXT:    v_add_f32_e32 v0, v0, v3
; GFX10-GISEL-NEXT:    v_add_f32_e32 v1, v1, v4
; GFX10-GISEL-NEXT:    v_add_f32_e32 v2, v2, v5
; GFX10-GISEL-NEXT:    s_setpc_b64 s[30:31]
;
; GFX11-SDAG-LABEL: v_constained_fadd_v3f32_fpexcept_strict:
; GFX11-SDAG:       ; %bb.0:
; GFX11-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-SDAG-NEXT:    v_dual_add_f32 v0, v0, v3 :: v_dual_add_f32 v1, v1, v4
; GFX11-SDAG-NEXT:    v_add_f32_e32 v2, v2, v5
; GFX11-SDAG-NEXT:    s_setpc_b64 s[30:31]
;
; GFX11-GISEL-LABEL: v_constained_fadd_v3f32_fpexcept_strict:
; GFX11-GISEL:       ; %bb.0:
; GFX11-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-GISEL-NEXT:    v_dual_add_f32 v0, v0, v3 :: v_dual_add_f32 v1, v1, v4
; GFX11-GISEL-NEXT:    v_add_f32_e32 v2, v2, v5
; GFX11-GISEL-NEXT:    s_setpc_b64 s[30:31]
;
; GFX12-LABEL: v_constained_fadd_v3f32_fpexcept_strict:
; 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_add_f32 v0, v0, v3 :: v_dual_add_f32 v1, v1, v4
; GFX12-NEXT:    v_add_f32_e32 v2, v2, v5
; GFX12-NEXT:    s_setpc_b64 s[30:31]
; GFX10-LABEL: v_constained_fadd_v3f32_fpexcept_strict:
; GFX10:       ; %bb.0:
; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-NEXT:    v_add_f32_e32 v0, v0, v3
; GFX10-NEXT:    v_add_f32_e32 v1, v1, v4
; GFX10-NEXT:    v_add_f32_e32 v2, v2, v5
; GFX10-NEXT:    s_setpc_b64 s[30:31]
; GFX11-LABEL: v_constained_fadd_v3f32_fpexcept_strict:
; GFX11:       ; %bb.0:
; GFX11-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-NEXT:    v_dual_add_f32 v0, v0, v3 :: v_dual_add_f32 v1, v1, v4
; GFX11-NEXT:    v_add_f32_e32 v2, v2, v5
; GFX11-NEXT:    s_setpc_b64 s[30:31]
; GCN-SDAG-LABEL: v_constained_fadd_v3f32_fpexcept_strict:
; GCN-SDAG:       ; %bb.0:
; GCN-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GCN-SDAG-NEXT:    v_add_f32_e32 v0, v0, v3
; GCN-SDAG-NEXT:    v_add_f32_e32 v1, v1, v4
; GCN-SDAG-NEXT:    v_add_f32_e32 v2, v2, v5
; GCN-SDAG-NEXT:    s_setpc_b64 s[30:31]
; GCN-GISEL-LABEL: v_constained_fadd_v3f32_fpexcept_strict:
; GCN-GISEL:       ; %bb.0:
; GCN-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GCN-GISEL-NEXT:    v_add_f32_e32 v0, v0, v3
; GCN-GISEL-NEXT:    v_add_f32_e32 v1, v1, v4
; GCN-GISEL-NEXT:    v_add_f32_e32 v2, v2, v5
; GCN-GISEL-NEXT:    s_setpc_b64 s[30:31]
  %val = call <3 x float> @llvm.experimental.constrained.fadd.v3f32(<3 x float> %x, <3 x float> %y, metadata !"round.tonearest", metadata !"fpexcept.strict")
  ret <3 x float> %val
}

define amdgpu_ps float @s_constained_fadd_f32_fpexcept_strict(float inreg %x, float inreg %y) #0 {
; GFX9-LABEL: s_constained_fadd_f32_fpexcept_strict:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    v_mov_b32_e32 v0, s3
; GFX9-NEXT:    v_add_f32_e32 v0, s2, v0
; GFX9-NEXT:    ; return to shader part epilog
;
; GFX10PLUS-LABEL: s_constained_fadd_f32_fpexcept_strict:
; GFX10PLUS:       ; %bb.0:
; GFX10PLUS-NEXT:    v_add_f32_e64 v0, s2, s3
; GFX10PLUS-NEXT:    ; return to shader part epilog
;
; GFX12-LABEL: s_constained_fadd_f32_fpexcept_strict:
; GFX12:       ; %bb.0:
; GFX12-NEXT:    s_add_f32 s0, s2, s3
; GFX12-NEXT:    s_delay_alu instid0(SALU_CYCLE_3)
; GFX12-NEXT:    v_mov_b32_e32 v0, s0
; GFX12-NEXT:    ; return to shader part epilog
; GCN-SDAG-LABEL: s_constained_fadd_f32_fpexcept_strict:
; GCN-SDAG:       ; %bb.0:
; GCN-SDAG-NEXT:    v_mov_b32_e32 v0, s3
; GCN-SDAG-NEXT:    v_add_f32_e32 v0, s2, v0
; GCN-SDAG-NEXT:    ; return to shader part epilog
; GCN-GISEL-LABEL: s_constained_fadd_f32_fpexcept_strict:
; GCN-GISEL:       ; %bb.0:
; GCN-GISEL-NEXT:    v_mov_b32_e32 v0, s3
; GCN-GISEL-NEXT:    v_add_f32_e32 v0, s2, v0
; GCN-GISEL-NEXT:    ; return to shader part epilog
  %val = call float @llvm.experimental.constrained.fadd.f32(float %x, float %y, metadata !"round.tonearest", metadata !"fpexcept.strict")
  ret float %val
}

define float @v_constained_fadd_f32_fpexcept_strict_fabs_lhs(float %x, float %y) #0 {
; GFX9-LABEL: v_constained_fadd_f32_fpexcept_strict_fabs_lhs:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    v_add_f32_e64 v0, |v0|, v1
; GFX9-NEXT:    s_setpc_b64 s[30:31]
;
; GFX10PLUS-LABEL: v_constained_fadd_f32_fpexcept_strict_fabs_lhs:
; GFX10PLUS:       ; %bb.0:
; GFX10PLUS-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10PLUS-NEXT:    v_add_f32_e64 v0, |v0|, v1
; GFX10PLUS-NEXT:    s_setpc_b64 s[30:31]
;
; GFX12-LABEL: v_constained_fadd_f32_fpexcept_strict_fabs_lhs:
; 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_add_f32_e64 v0, |v0|, v1
; GFX12-NEXT:    s_setpc_b64 s[30:31]
; GCN-SDAG-LABEL: v_constained_fadd_f32_fpexcept_strict_fabs_lhs:
; GCN-SDAG:       ; %bb.0:
; GCN-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GCN-SDAG-NEXT:    v_add_f32_e64 v0, |v0|, v1
; GCN-SDAG-NEXT:    s_setpc_b64 s[30:31]
; GCN-GISEL-LABEL: v_constained_fadd_f32_fpexcept_strict_fabs_lhs:
; GCN-GISEL:       ; %bb.0:
; GCN-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GCN-GISEL-NEXT:    v_add_f32_e64 v0, |v0|, v1
; GCN-GISEL-NEXT:    s_setpc_b64 s[30:31]
  %fabs.x = call float @llvm.fabs.f32(float %x) #0
  %val = call float @llvm.experimental.constrained.fadd.f32(float %fabs.x, float %y, metadata !"round.tonearest", metadata !"fpexcept.strict")
  ret float %val
}

define float @v_constained_fadd_f32_fpexcept_strict_fabs_rhs(float %x, float %y) #0 {
; GFX9-LABEL: v_constained_fadd_f32_fpexcept_strict_fabs_rhs:
; GFX9:       ; %bb.0:
; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-NEXT:    v_add_f32_e64 v0, v0, |v1|
; GFX9-NEXT:    s_setpc_b64 s[30:31]
;
; GFX10PLUS-LABEL: v_constained_fadd_f32_fpexcept_strict_fabs_rhs:
; GFX10PLUS:       ; %bb.0:
; GFX10PLUS-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10PLUS-NEXT:    v_add_f32_e64 v0, v0, |v1|
; GFX10PLUS-NEXT:    s_setpc_b64 s[30:31]
;
; GFX12-LABEL: v_constained_fadd_f32_fpexcept_strict_fabs_rhs:
; 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_add_f32_e64 v0, v0, |v1|
; GFX12-NEXT:    s_setpc_b64 s[30:31]
; GCN-SDAG-LABEL: v_constained_fadd_f32_fpexcept_strict_fabs_rhs:
; GCN-SDAG:       ; %bb.0:
; GCN-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GCN-SDAG-NEXT:    v_add_f32_e64 v0, v0, |v1|
; GCN-SDAG-NEXT:    s_setpc_b64 s[30:31]
; GCN-GISEL-LABEL: v_constained_fadd_f32_fpexcept_strict_fabs_rhs:
; GCN-GISEL:       ; %bb.0:
; GCN-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GCN-GISEL-NEXT:    v_add_f32_e64 v0, v0, |v1|
; GCN-GISEL-NEXT:    s_setpc_b64 s[30:31]
  %fabs.y = call float @llvm.fabs.f32(float %y) #0
  %val = call float @llvm.experimental.constrained.fadd.f32(float %x, float %fabs.y, metadata !"round.tonearest", metadata !"fpexcept.strict")
  ret float %val
}

define float @v_constained_fadd_f32_fpexcept_strict_fneg_fabs_lhs(float %x, float %y) #0 {
; GCN-SDAG-LABEL: v_constained_fadd_f32_fpexcept_strict_fneg_fabs_lhs:
; GCN-SDAG:       ; %bb.0:
; GCN-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GCN-SDAG-NEXT:    v_sub_f32_e64 v0, v1, |v0|
; GCN-SDAG-NEXT:    s_setpc_b64 s[30:31]
;
; GCN-GISEL-LABEL: v_constained_fadd_f32_fpexcept_strict_fneg_fabs_lhs:
; GCN-GISEL:       ; %bb.0:
; GCN-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GCN-GISEL-NEXT:    v_add_f32_e64 v0, -|v0|, v1
; GCN-GISEL-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-SDAG-LABEL: v_constained_fadd_f32_fpexcept_strict_fneg_fabs_lhs:
; GFX9-SDAG:       ; %bb.0:
; GFX9-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-SDAG-NEXT:    v_sub_f32_e64 v0, v1, |v0|
; GFX9-SDAG-NEXT:    s_setpc_b64 s[30:31]
;
; GFX9-GISEL-LABEL: v_constained_fadd_f32_fpexcept_strict_fneg_fabs_lhs:
; GFX9-GISEL:       ; %bb.0:
; GFX9-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX9-GISEL-NEXT:    v_add_f32_e64 v0, -|v0|, v1
; GFX9-GISEL-NEXT:    s_setpc_b64 s[30:31]
;
; GFX10-SDAG-LABEL: v_constained_fadd_f32_fpexcept_strict_fneg_fabs_lhs:
; GFX10-SDAG:       ; %bb.0:
; GFX10-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-SDAG-NEXT:    v_sub_f32_e64 v0, v1, |v0|
; GFX10-SDAG-NEXT:    s_setpc_b64 s[30:31]
;
; GFX10-GISEL-LABEL: v_constained_fadd_f32_fpexcept_strict_fneg_fabs_lhs:
; GFX10-GISEL:       ; %bb.0:
; GFX10-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10-GISEL-NEXT:    v_add_f32_e64 v0, -|v0|, v1
; GFX10-GISEL-NEXT:    s_setpc_b64 s[30:31]
;
; GFX11-SDAG-LABEL: v_constained_fadd_f32_fpexcept_strict_fneg_fabs_lhs:
; GFX11-SDAG:       ; %bb.0:
; GFX11-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-SDAG-NEXT:    v_sub_f32_e64 v0, v1, |v0|
; GFX11-SDAG-NEXT:    s_setpc_b64 s[30:31]
;
; GFX11-GISEL-LABEL: v_constained_fadd_f32_fpexcept_strict_fneg_fabs_lhs:
; GFX11-GISEL:       ; %bb.0:
; GFX11-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX11-GISEL-NEXT:    v_add_f32_e64 v0, -|v0|, v1
; GFX11-GISEL-NEXT:    s_setpc_b64 s[30:31]
;
; GFX12-SDAG-LABEL: v_constained_fadd_f32_fpexcept_strict_fneg_fabs_lhs:
; GFX12-SDAG:       ; %bb.0:
; GFX12-SDAG-NEXT:    s_wait_loadcnt_dscnt 0x0
; GFX12-SDAG-NEXT:    s_wait_expcnt 0x0
; GFX12-SDAG-NEXT:    s_wait_samplecnt 0x0
; GFX12-SDAG-NEXT:    s_wait_bvhcnt 0x0
; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
; GFX12-SDAG-NEXT:    v_sub_f32_e64 v0, v1, |v0|
; GFX12-SDAG-NEXT:    s_setpc_b64 s[30:31]
;
; GFX12-GISEL-LABEL: v_constained_fadd_f32_fpexcept_strict_fneg_fabs_lhs:
; GFX12-GISEL:       ; %bb.0:
; GFX12-GISEL-NEXT:    s_wait_loadcnt_dscnt 0x0
; GFX12-GISEL-NEXT:    s_wait_expcnt 0x0
; GFX12-GISEL-NEXT:    s_wait_samplecnt 0x0
; GFX12-GISEL-NEXT:    s_wait_bvhcnt 0x0
; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
; GFX12-GISEL-NEXT:    v_add_f32_e64 v0, -|v0|, v1
; GFX12-GISEL-NEXT:    s_setpc_b64 s[30:31]
  %fabs.x = call float @llvm.fabs.f32(float %x) #0
  %neg.fabs.x = fneg float %fabs.x
  %val = call float @llvm.experimental.constrained.fadd.f32(float %neg.fabs.x, float %y, metadata !"round.tonearest", metadata !"fpexcept.strict")
  ret float %val
}

declare float @llvm.fabs.f32(float)
declare float @llvm.experimental.constrained.fadd.f32(float, float, metadata, metadata)
declare <2 x float> @llvm.experimental.constrained.fadd.v2f32(<2 x float>, <2 x float>, metadata, metadata)
declare <3 x float> @llvm.experimental.constrained.fadd.v3f32(<3 x float>, <3 x float>, metadata, metadata)

attributes #0 = { strictfp }