Skip to content

AMDGPU: Handle f32 minimum3/maximum3 pattern for gfx950 #117737

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 1 commit into from
Nov 26, 2024
Merged

AMDGPU: Handle f32 minimum3/maximum3 pattern for gfx950 #117737

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
3 changes: 2 additions & 1 deletion llvm/lib/Target/AMDGPU/SIISelLowering.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -13346,7 +13346,8 @@ static bool supportsMin3Max3(const GCNSubtarget &Subtarget, unsigned Opc,
return (VT == MVT::f32) || (VT == MVT::f16 && Subtarget.hasMin3Max3_16());
case ISD::FMINIMUM:
case ISD::FMAXIMUM:
return (VT == MVT::f32 || VT == MVT::f16) && Subtarget.hasIEEEMinMax3();
return (VT == MVT::f32 && Subtarget.hasMinimum3Maximum3F32()) ||
(VT == MVT::f16 && Subtarget.hasMinimum3Maximum3F16());
case ISD::SMAX:
case ISD::SMIN:
case ISD::UMAX:
Expand Down
146 changes: 50 additions & 96 deletions llvm/test/CodeGen/AMDGPU/fmaximum3.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -31,8 +31,7 @@ define float @v_fmaximum3_f32(float %a, float %b, float %c) {
; GFX950-LABEL: v_fmaximum3_f32:
; GFX950: ; %bb.0:
; GFX950-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX950-NEXT: v_maximum3_f32 v0, v0, v1, v1
; GFX950-NEXT: v_maximum3_f32 v0, v0, v2, v2
; GFX950-NEXT: v_maximum3_f32 v0, v0, v1, v2
; GFX950-NEXT: s_setpc_b64 s[30:31]
%max0 = call float @llvm.maximum.f32(float %a, float %b)
%max1 = call float @llvm.maximum.f32(float %max0, float %c)
Expand Down Expand Up @@ -67,8 +66,7 @@ define float @v_fmaximum3_f32_commute(float %a, float %b, float %c) {
; GFX950-LABEL: v_fmaximum3_f32_commute:
; GFX950: ; %bb.0:
; GFX950-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX950-NEXT: v_maximum3_f32 v0, v0, v1, v1
; GFX950-NEXT: v_maximum3_f32 v0, v2, v0, v0
; GFX950-NEXT: v_maximum3_f32 v0, v2, v0, v1
; GFX950-NEXT: s_setpc_b64 s[30:31]
%max0 = call float @llvm.maximum.f32(float %a, float %b)
%max1 = call float @llvm.maximum.f32(float %c, float %max0)
Expand Down Expand Up @@ -102,9 +100,9 @@ define amdgpu_ps i32 @s_fmaximum3_f32(float inreg %a, float inreg %b, float inre
;
; GFX950-LABEL: s_fmaximum3_f32:
; GFX950: ; %bb.0:
; GFX950-NEXT: v_mov_b32_e32 v0, s0
; GFX950-NEXT: v_maximum3_f32 v0, v0, s1, s1
; GFX950-NEXT: v_maximum3_f32 v0, v0, s2, s2
; GFX950-NEXT: v_mov_b32_e32 v0, s1
; GFX950-NEXT: v_mov_b32_e32 v1, s2
; GFX950-NEXT: v_maximum3_f32 v0, s0, v0, v1
; GFX950-NEXT: s_nop 0
; GFX950-NEXT: v_readfirstlane_b32 s0, v0
; GFX950-NEXT: ; return to shader part epilog
Expand Down Expand Up @@ -143,8 +141,7 @@ define float @v_fmaximum3_f32_fabs0(float %a, float %b, float %c) {
; GFX950-LABEL: v_fmaximum3_f32_fabs0:
; GFX950: ; %bb.0:
; GFX950-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX950-NEXT: v_maximum3_f32 v0, |v0|, v1, v1
; GFX950-NEXT: v_maximum3_f32 v0, v0, v2, v2
; GFX950-NEXT: v_maximum3_f32 v0, |v0|, v1, v2
; GFX950-NEXT: s_setpc_b64 s[30:31]
%a.fabs = call float @llvm.fabs.f32(float %a)
%max0 = call float @llvm.maximum.f32(float %a.fabs, float %b)
Expand Down Expand Up @@ -180,8 +177,7 @@ define float @v_fmaximum3_f32_fabs1(float %a, float %b, float %c) {
; GFX950-LABEL: v_fmaximum3_f32_fabs1:
; GFX950: ; %bb.0:
; GFX950-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX950-NEXT: v_maximum3_f32 v0, v0, |v1|, |v1|
; GFX950-NEXT: v_maximum3_f32 v0, v0, v2, v2
; GFX950-NEXT: v_maximum3_f32 v0, v0, |v1|, v2
; GFX950-NEXT: s_setpc_b64 s[30:31]
%b.fabs = call float @llvm.fabs.f32(float %b)
%max0 = call float @llvm.maximum.f32(float %a, float %b.fabs)
Expand Down Expand Up @@ -217,8 +213,7 @@ define float @v_fmaximum3_f32_fabs2(float %a, float %b, float %c) {
; GFX950-LABEL: v_fmaximum3_f32_fabs2:
; GFX950: ; %bb.0:
; GFX950-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX950-NEXT: v_maximum3_f32 v0, v0, v1, v1
; GFX950-NEXT: v_maximum3_f32 v0, v0, |v2|, |v2|
; GFX950-NEXT: v_maximum3_f32 v0, v0, v1, |v2|
; GFX950-NEXT: s_setpc_b64 s[30:31]
%c.fabs = call float @llvm.fabs.f32(float %c)
%max0 = call float @llvm.maximum.f32(float %a, float %b)
Expand Down Expand Up @@ -254,8 +249,7 @@ define float @v_fmaximum3_f32_fabs_all(float %a, float %b, float %c) {
; GFX950-LABEL: v_fmaximum3_f32_fabs_all:
; GFX950: ; %bb.0:
; GFX950-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX950-NEXT: v_maximum3_f32 v0, |v0|, |v1|, |v1|
; GFX950-NEXT: v_maximum3_f32 v0, v0, |v2|, |v2|
; GFX950-NEXT: v_maximum3_f32 v0, |v0|, |v1|, |v2|
; GFX950-NEXT: s_setpc_b64 s[30:31]
%a.fabs = call float @llvm.fabs.f32(float %a)
%b.fabs = call float @llvm.fabs.f32(float %b)
Expand Down Expand Up @@ -293,8 +287,7 @@ define float @v_fmaximum3_f32_fneg_all(float %a, float %b, float %c) {
; GFX950-LABEL: v_fmaximum3_f32_fneg_all:
; GFX950: ; %bb.0:
; GFX950-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX950-NEXT: v_maximum3_f32 v0, -v0, -v1, -v1
; GFX950-NEXT: v_maximum3_f32 v0, v0, -v2, -v2
; GFX950-NEXT: v_maximum3_f32 v0, -v0, -v1, -v2
; GFX950-NEXT: s_setpc_b64 s[30:31]
%a.fneg = fneg float %a
%b.fneg = fneg float %b
Expand Down Expand Up @@ -332,8 +325,7 @@ define float @v_fmaximum3_f32_fneg_fabs_all(float %a, float %b, float %c) {
; GFX950-LABEL: v_fmaximum3_f32_fneg_fabs_all:
; GFX950: ; %bb.0:
; GFX950-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX950-NEXT: v_maximum3_f32 v0, -|v0|, -|v1|, -|v1|
; GFX950-NEXT: v_maximum3_f32 v0, v0, -|v2|, -|v2|
; GFX950-NEXT: v_maximum3_f32 v0, -|v0|, -|v1|, -|v2|
; GFX950-NEXT: s_setpc_b64 s[30:31]
%a.fabs = call float @llvm.fabs.f32(float %a)
%b.fabs = call float @llvm.fabs.f32(float %b)
Expand Down Expand Up @@ -374,8 +366,7 @@ define float @v_fmaximum3_f32_fneg0(float %a, float %b, float %c) {
; GFX950-LABEL: v_fmaximum3_f32_fneg0:
; GFX950: ; %bb.0:
; GFX950-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX950-NEXT: v_maximum3_f32 v0, -v0, v1, v1
; GFX950-NEXT: v_maximum3_f32 v0, v0, v2, v2
; GFX950-NEXT: v_maximum3_f32 v0, -v0, v1, v2
; GFX950-NEXT: s_setpc_b64 s[30:31]
%a.fneg = fneg float %a
%max0 = call float @llvm.maximum.f32(float %a.fneg, float %b)
Expand Down Expand Up @@ -411,8 +402,7 @@ define float @v_fmaximum3_f32_fneg1(float %a, float %b, float %c) {
; GFX950-LABEL: v_fmaximum3_f32_fneg1:
; GFX950: ; %bb.0:
; GFX950-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX950-NEXT: v_maximum3_f32 v0, v0, -v1, -v1
; GFX950-NEXT: v_maximum3_f32 v0, v0, v2, v2
; GFX950-NEXT: v_maximum3_f32 v0, v0, -v1, v2
; GFX950-NEXT: s_setpc_b64 s[30:31]
%b.fneg = fneg float %b
%max0 = call float @llvm.maximum.f32(float %a, float %b.fneg)
Expand Down Expand Up @@ -448,8 +438,7 @@ define float @v_fmaximum3_f32_fneg2(float %a, float %b, float %c) {
; GFX950-LABEL: v_fmaximum3_f32_fneg2:
; GFX950: ; %bb.0:
; GFX950-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX950-NEXT: v_maximum3_f32 v0, v0, v1, v1
; GFX950-NEXT: v_maximum3_f32 v0, v0, -v2, -v2
; GFX950-NEXT: v_maximum3_f32 v0, v0, v1, -v2
; GFX950-NEXT: s_setpc_b64 s[30:31]
%c.fneg = fneg float %c
%max0 = call float @llvm.maximum.f32(float %a, float %b)
Expand Down Expand Up @@ -486,8 +475,7 @@ define float @v_fmaximum3_f32_const0(float %b, float %c) {
; GFX950: ; %bb.0:
; GFX950-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX950-NEXT: s_mov_b32 s0, 0x41000000
; GFX950-NEXT: v_maximum3_f32 v0, v0, s0, s0
; GFX950-NEXT: v_maximum3_f32 v0, v0, v1, v1
; GFX950-NEXT: v_maximum3_f32 v0, v0, s0, v1
; GFX950-NEXT: s_setpc_b64 s[30:31]
%max0 = call float @llvm.maximum.f32(float 8.0, float %b)
%max1 = call float @llvm.maximum.f32(float %max0, float %c)
Expand Down Expand Up @@ -522,9 +510,8 @@ define float @v_fmaximum3_f32__const2(float %a, float %b) {
; GFX950-LABEL: v_fmaximum3_f32__const2:
; GFX950: ; %bb.0:
; GFX950-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX950-NEXT: v_maximum3_f32 v0, v0, v1, v1
; GFX950-NEXT: s_mov_b32 s0, 0x41000000
; GFX950-NEXT: v_maximum3_f32 v0, v0, s0, s0
; GFX950-NEXT: v_maximum3_f32 v0, v0, v1, s0
; GFX950-NEXT: s_setpc_b64 s[30:31]
%max0 = call float @llvm.maximum.f32(float %a, float %b)
%max1 = call float @llvm.maximum.f32(float %max0, float 8.0)
Expand Down Expand Up @@ -559,8 +546,7 @@ define float @v_fmaximum3_f32_inlineimm0(float %b, float %c) {
; GFX950-LABEL: v_fmaximum3_f32_inlineimm0:
; GFX950: ; %bb.0:
; GFX950-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX950-NEXT: v_maximum3_f32 v0, v0, 4.0, 4.0
; GFX950-NEXT: v_maximum3_f32 v0, v0, v1, v1
; GFX950-NEXT: v_maximum3_f32 v0, v0, 4.0, v1
; GFX950-NEXT: s_setpc_b64 s[30:31]
%max0 = call float @llvm.maximum.f32(float 4.0, float %b)
%max1 = call float @llvm.maximum.f32(float %max0, float %c)
Expand Down Expand Up @@ -595,8 +581,7 @@ define float @v_fmaximum3_f32__inlineimm(float %a, float %b) {
; GFX950-LABEL: v_fmaximum3_f32__inlineimm:
; GFX950: ; %bb.0:
; GFX950-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX950-NEXT: v_maximum3_f32 v0, v0, v1, v1
; GFX950-NEXT: v_maximum3_f32 v0, v0, 4.0, 4.0
; GFX950-NEXT: v_maximum3_f32 v0, v0, v1, 4.0
; GFX950-NEXT: s_setpc_b64 s[30:31]
%max0 = call float @llvm.maximum.f32(float %a, float %b)
%max1 = call float @llvm.maximum.f32(float %max0, float 4.0)
Expand Down Expand Up @@ -634,9 +619,8 @@ define float @v_fmaximum3_f32_const1_const2(float %a) {
; GFX950: ; %bb.0:
; GFX950-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX950-NEXT: s_mov_b32 s0, 0x41000000
; GFX950-NEXT: v_maximum3_f32 v0, v0, s0, s0
; GFX950-NEXT: s_mov_b32 s0, 0x41800000
; GFX950-NEXT: v_maximum3_f32 v0, v0, s0, s0
; GFX950-NEXT: v_mov_b32_e32 v1, 0x41800000
; GFX950-NEXT: v_maximum3_f32 v0, v0, s0, v1
; GFX950-NEXT: s_setpc_b64 s[30:31]
%max0 = call float @llvm.maximum.f32(float %a, float 8.0)
%max1 = call float @llvm.maximum.f32(float %max0, float 16.0)
Expand Down Expand Up @@ -680,10 +664,8 @@ define <2 x float> @v_fmaximum3_v2f32(<2 x float> %a, <2 x float> %b, <2 x float
; GFX950-LABEL: v_fmaximum3_v2f32:
; GFX950: ; %bb.0:
; GFX950-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX950-NEXT: v_maximum3_f32 v1, v1, v3, v3
; GFX950-NEXT: v_maximum3_f32 v0, v0, v2, v2
; GFX950-NEXT: v_maximum3_f32 v0, v4, v0, v0
; GFX950-NEXT: v_maximum3_f32 v1, v5, v1, v1
; GFX950-NEXT: v_maximum3_f32 v0, v4, v0, v2
; GFX950-NEXT: v_maximum3_f32 v1, v5, v1, v3
; GFX950-NEXT: s_setpc_b64 s[30:31]
%max0 = call <2 x float> @llvm.maximum.v2f32(<2 x float> %a, <2 x float> %b)
%max1 = call <2 x float> @llvm.maximum.v2f32(<2 x float> %c, <2 x float> %max0)
Expand Down Expand Up @@ -727,10 +709,8 @@ define <2 x float> @v_fmaximum3_v2f32_commute(<2 x float> %a, <2 x float> %b, <2
; GFX950-LABEL: v_fmaximum3_v2f32_commute:
; GFX950: ; %bb.0:
; GFX950-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX950-NEXT: v_maximum3_f32 v1, v1, v3, v3
; GFX950-NEXT: v_maximum3_f32 v0, v0, v2, v2
; GFX950-NEXT: v_maximum3_f32 v0, v0, v4, v4
; GFX950-NEXT: v_maximum3_f32 v1, v1, v5, v5
; GFX950-NEXT: v_maximum3_f32 v0, v0, v2, v4
; GFX950-NEXT: v_maximum3_f32 v1, v1, v3, v5
; GFX950-NEXT: s_setpc_b64 s[30:31]
%max0 = call <2 x float> @llvm.maximum.v2f32(<2 x float> %a, <2 x float> %b)
%max1 = call <2 x float> @llvm.maximum.v2f32(<2 x float> %max0, <2 x float> %c)
Expand Down Expand Up @@ -774,10 +754,8 @@ define <2 x float> @v_fmaximum3_v2f32__fabs_all(<2 x float> %a, <2 x float> %b,
; GFX950-LABEL: v_fmaximum3_v2f32__fabs_all:
; GFX950: ; %bb.0:
; GFX950-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX950-NEXT: v_maximum3_f32 v1, |v1|, |v3|, |v3|
; GFX950-NEXT: v_maximum3_f32 v0, |v0|, |v2|, |v2|
; GFX950-NEXT: v_maximum3_f32 v0, v0, |v4|, |v4|
; GFX950-NEXT: v_maximum3_f32 v1, v1, |v5|, |v5|
; GFX950-NEXT: v_maximum3_f32 v0, |v0|, |v2|, |v4|
; GFX950-NEXT: v_maximum3_f32 v1, |v1|, |v3|, |v5|
; GFX950-NEXT: s_setpc_b64 s[30:31]
%a.fabs = call <2 x float> @llvm.fabs.v2f32(<2 x float> %a)
%b.fabs = call <2 x float> @llvm.fabs.v2f32(<2 x float> %b)
Expand Down Expand Up @@ -824,10 +802,8 @@ define <2 x float> @v_fmaximum3_v2f32__fneg_all(<2 x float> %a, <2 x float> %b,
; GFX950-LABEL: v_fmaximum3_v2f32__fneg_all:
; GFX950: ; %bb.0:
; GFX950-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX950-NEXT: v_maximum3_f32 v1, -v1, -v3, -v3
; GFX950-NEXT: v_maximum3_f32 v0, -v0, -v2, -v2
; GFX950-NEXT: v_maximum3_f32 v0, v0, -v4, -v4
; GFX950-NEXT: v_maximum3_f32 v1, v1, -v5, -v5
; GFX950-NEXT: v_maximum3_f32 v0, -v0, -v2, -v4
; GFX950-NEXT: v_maximum3_f32 v1, -v1, -v3, -v5
; GFX950-NEXT: s_setpc_b64 s[30:31]
%a.fneg = fneg <2 x float> %a
%b.fneg = fneg <2 x float> %b
Expand Down Expand Up @@ -874,10 +850,8 @@ define <2 x float> @v_fmaximum3_v2f32__inlineimm1(<2 x float> %a, <2 x float> %c
; GFX950-LABEL: v_fmaximum3_v2f32__inlineimm1:
; GFX950: ; %bb.0:
; GFX950-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX950-NEXT: v_maximum3_f32 v1, v1, 2.0, 2.0
; GFX950-NEXT: v_maximum3_f32 v0, v0, 2.0, 2.0
; GFX950-NEXT: v_maximum3_f32 v0, v0, v2, v2
; GFX950-NEXT: v_maximum3_f32 v1, v1, v3, v3
; GFX950-NEXT: v_maximum3_f32 v0, v0, 2.0, v2
; GFX950-NEXT: v_maximum3_f32 v1, v1, 2.0, v3
; GFX950-NEXT: s_setpc_b64 s[30:31]
%max0 = call <2 x float> @llvm.maximum.v2f32(<2 x float> %a, <2 x float> <float 2.0, float 2.0>)
%max1 = call <2 x float> @llvm.maximum.v2f32(<2 x float> %max0, <2 x float> %c)
Expand Down Expand Up @@ -921,10 +895,8 @@ define <2 x float> @v_fmaximum3_v2f32__inlineimm2(<2 x float> %a, <2 x float> %b
; GFX950-LABEL: v_fmaximum3_v2f32__inlineimm2:
; GFX950: ; %bb.0:
; GFX950-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX950-NEXT: v_maximum3_f32 v1, v1, v3, v3
; GFX950-NEXT: v_maximum3_f32 v0, v0, v2, v2
; GFX950-NEXT: v_maximum3_f32 v0, v0, 4.0, 4.0
; GFX950-NEXT: v_maximum3_f32 v1, v1, 4.0, 4.0
; GFX950-NEXT: v_maximum3_f32 v0, v0, v2, 4.0
; GFX950-NEXT: v_maximum3_f32 v1, v1, v3, 4.0
; GFX950-NEXT: s_setpc_b64 s[30:31]
%max0 = call <2 x float> @llvm.maximum.v2f32(<2 x float> %a, <2 x float> %b)
%max1 = call <2 x float> @llvm.maximum.v2f32(<2 x float> %max0, <2 x float> <float 4.0, float 4.0>)
Expand Down Expand Up @@ -977,12 +949,9 @@ define <3 x float> @v_fmaximum3_v3f32(<3 x float> %a, <3 x float> %b, <3 x float
; GFX950-LABEL: v_fmaximum3_v3f32:
; GFX950: ; %bb.0:
; GFX950-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX950-NEXT: v_maximum3_f32 v2, v2, v5, v5
; GFX950-NEXT: v_maximum3_f32 v1, v1, v4, v4
; GFX950-NEXT: v_maximum3_f32 v0, v0, v3, v3
; GFX950-NEXT: v_maximum3_f32 v0, v6, v0, v0
; GFX950-NEXT: v_maximum3_f32 v1, v7, v1, v1
; GFX950-NEXT: v_maximum3_f32 v2, v8, v2, v2
; GFX950-NEXT: v_maximum3_f32 v0, v6, v0, v3
; GFX950-NEXT: v_maximum3_f32 v1, v7, v1, v4
; GFX950-NEXT: v_maximum3_f32 v2, v8, v2, v5
; GFX950-NEXT: s_setpc_b64 s[30:31]
%max0 = call <3 x float> @llvm.maximum.v3f32(<3 x float> %a, <3 x float> %b)
%max1 = call <3 x float> @llvm.maximum.v3f32(<3 x float> %c, <3 x float> %max0)
Expand Down Expand Up @@ -1035,12 +1004,9 @@ define <3 x float> @v_fmaximum3_v3f32_commute(<3 x float> %a, <3 x float> %b, <3
; GFX950-LABEL: v_fmaximum3_v3f32_commute:
; GFX950: ; %bb.0:
; GFX950-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX950-NEXT: v_maximum3_f32 v2, v2, v5, v5
; GFX950-NEXT: v_maximum3_f32 v1, v1, v4, v4
; GFX950-NEXT: v_maximum3_f32 v0, v0, v3, v3
; GFX950-NEXT: v_maximum3_f32 v0, v0, v6, v6
; GFX950-NEXT: v_maximum3_f32 v1, v1, v7, v7
; GFX950-NEXT: v_maximum3_f32 v2, v2, v8, v8
; GFX950-NEXT: v_maximum3_f32 v0, v0, v3, v6
; GFX950-NEXT: v_maximum3_f32 v1, v1, v4, v7
; GFX950-NEXT: v_maximum3_f32 v2, v2, v5, v8
; GFX950-NEXT: s_setpc_b64 s[30:31]
%max0 = call <3 x float> @llvm.maximum.v3f32(<3 x float> %a, <3 x float> %b)
%max1 = call <3 x float> @llvm.maximum.v3f32(<3 x float> %max0, <3 x float> %c)
Expand Down Expand Up @@ -1093,12 +1059,9 @@ define <3 x float> @v_fmaximum3_v3f32__fabs_all(<3 x float> %a, <3 x float> %b,
; GFX950-LABEL: v_fmaximum3_v3f32__fabs_all:
; GFX950: ; %bb.0:
; GFX950-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX950-NEXT: v_maximum3_f32 v2, |v2|, |v5|, |v5|
; GFX950-NEXT: v_maximum3_f32 v1, |v1|, |v4|, |v4|
; GFX950-NEXT: v_maximum3_f32 v0, |v0|, |v3|, |v3|
; GFX950-NEXT: v_maximum3_f32 v0, v0, |v6|, |v6|
; GFX950-NEXT: v_maximum3_f32 v1, v1, |v7|, |v7|
; GFX950-NEXT: v_maximum3_f32 v2, v2, |v8|, |v8|
; GFX950-NEXT: v_maximum3_f32 v0, |v0|, |v3|, |v6|
; GFX950-NEXT: v_maximum3_f32 v1, |v1|, |v4|, |v7|
; GFX950-NEXT: v_maximum3_f32 v2, |v2|, |v5|, |v8|
; GFX950-NEXT: s_setpc_b64 s[30:31]
%a.fabs = call <3 x float> @llvm.fabs.v3f32(<3 x float> %a)
%b.fabs = call <3 x float> @llvm.fabs.v3f32(<3 x float> %b)
Expand Down Expand Up @@ -1154,12 +1117,9 @@ define <3 x float> @v_fmaximum3_v3f32__fneg_all(<3 x float> %a, <3 x float> %b,
; GFX950-LABEL: v_fmaximum3_v3f32__fneg_all:
; GFX950: ; %bb.0:
; GFX950-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX950-NEXT: v_maximum3_f32 v2, -v2, -v5, -v5
; GFX950-NEXT: v_maximum3_f32 v1, -v1, -v4, -v4
; GFX950-NEXT: v_maximum3_f32 v0, -v0, -v3, -v3
; GFX950-NEXT: v_maximum3_f32 v0, v0, -v6, -v6
; GFX950-NEXT: v_maximum3_f32 v1, v1, -v7, -v7
; GFX950-NEXT: v_maximum3_f32 v2, v2, -v8, -v8
; GFX950-NEXT: v_maximum3_f32 v0, -v0, -v3, -v6
; GFX950-NEXT: v_maximum3_f32 v1, -v1, -v4, -v7
; GFX950-NEXT: v_maximum3_f32 v2, -v2, -v5, -v8
; GFX950-NEXT: s_setpc_b64 s[30:31]
%a.fneg = fneg <3 x float> %a
%b.fneg = fneg <3 x float> %b
Expand Down Expand Up @@ -1215,12 +1175,9 @@ define <3 x float> @v_fmaximum3_v3f32__inlineimm1(<3 x float> %a, <3 x float> %c
; GFX950-LABEL: v_fmaximum3_v3f32__inlineimm1:
; GFX950: ; %bb.0:
; GFX950-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX950-NEXT: v_maximum3_f32 v2, v2, 2.0, 2.0
; GFX950-NEXT: v_maximum3_f32 v1, v1, 2.0, 2.0
; GFX950-NEXT: v_maximum3_f32 v0, v0, 2.0, 2.0
; GFX950-NEXT: v_maximum3_f32 v0, v0, v3, v3
; GFX950-NEXT: v_maximum3_f32 v1, v1, v4, v4
; GFX950-NEXT: v_maximum3_f32 v2, v2, v5, v5
; GFX950-NEXT: v_maximum3_f32 v0, v0, 2.0, v3
; GFX950-NEXT: v_maximum3_f32 v1, v1, 2.0, v4
; GFX950-NEXT: v_maximum3_f32 v2, v2, 2.0, v5
; GFX950-NEXT: s_setpc_b64 s[30:31]
%max0 = call <3 x float> @llvm.maximum.v3f32(<3 x float> %a, <3 x float> <float 2.0, float 2.0, float 2.0>)
%max1 = call <3 x float> @llvm.maximum.v3f32(<3 x float> %max0, <3 x float> %c)
Expand Down Expand Up @@ -1273,12 +1230,9 @@ define <3 x float> @v_fmaximum3_v3f32__inlineimm2(<3 x float> %a, <3 x float> %b
; GFX950-LABEL: v_fmaximum3_v3f32__inlineimm2:
; GFX950: ; %bb.0:
; GFX950-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX950-NEXT: v_maximum3_f32 v2, v2, v5, v5
; GFX950-NEXT: v_maximum3_f32 v1, v1, v4, v4
; GFX950-NEXT: v_maximum3_f32 v0, v0, v3, v3
; GFX950-NEXT: v_maximum3_f32 v0, v0, 4.0, 4.0
; GFX950-NEXT: v_maximum3_f32 v1, v1, 4.0, 4.0
; GFX950-NEXT: v_maximum3_f32 v2, v2, 4.0, 4.0
; GFX950-NEXT: v_maximum3_f32 v0, v0, v3, 4.0
; GFX950-NEXT: v_maximum3_f32 v1, v1, v4, 4.0
; GFX950-NEXT: v_maximum3_f32 v2, v2, v5, 4.0
; GFX950-NEXT: s_setpc_b64 s[30:31]
%max0 = call <3 x float> @llvm.maximum.v3f32(<3 x float> %a, <3 x float> %b)
%max1 = call <3 x float> @llvm.maximum.v3f32(<3 x float> %max0, <3 x float> <float 4.0, float 4.0, float 4.0>)
Expand Down
Loading