Skip to content

Commit f39e714

Browse files
paschalis-mpeispaschalis-mpeis
committed
[AArch64][CostModel] Improve scalar frem cost
In AArch64 the cost of scalar frem is the cost of a call to 'fmod'.
1 parent b8025d1 commit f39e714

File tree

3 files changed

+53
-45
lines changed

3 files changed

+53
-45
lines changed

‎llvm/lib/Target/AArch64/AArch64TargetTransformInfo.cpp

Lines changed: 8 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -2940,6 +2940,14 @@ InstructionCost AArch64TTIImpl::getArithmeticInstrCost(
29402940

29412941
return BaseT::getArithmeticInstrCost(Opcode, Ty, CostKind, Op1Info,
29422942
Op2Info);
2943+
case ISD::FREM:
2944+
if (!Ty->isVectorTy()) {
2945+
Function *F =
2946+
CxtI == nullptr ? nullptr : CxtI->getModule()->getFunction("fmod");
2947+
return getCallInstrCost(F, Ty, {Ty, Ty}, CostKind);
2948+
}
2949+
return BaseT::getArithmeticInstrCost(Opcode, Ty, CostKind, Op1Info,
2950+
Op2Info);
29432951
}
29442952
}
29452953

‎llvm/test/Analysis/CostModel/AArch64/arith-fp-frem.ll

Lines changed: 34 additions & 34 deletions
Original file line numberDiff line numberDiff line change
@@ -22,44 +22,44 @@ target triple = "aarch64-unknown-linux-gnu"
2222

2323
define void @frem_f64(ptr noalias %in.ptr, ptr noalias %out.ptr) {
2424
; NEON-NO-VECLIB-LABEL: 'frem_f64'
25-
; NEON-NO-VECLIB: LV: Found an estimated cost of 2 for VF 1 For instruction: %res = frem double %in, %in
26-
; NEON-NO-VECLIB: LV: Found an estimated cost of 8 for VF 2 For instruction: %res = frem double %in, %in
25+
; NEON-NO-VECLIB: LV: Found an estimated cost of 10 for VF 1 For instruction: %res = frem double %in, %in
26+
; NEON-NO-VECLIB: LV: Found an estimated cost of 24 for VF 2 For instruction: %res = frem double %in, %in
2727
;
2828
; SVE-NO-VECLIB-LABEL: 'frem_f64'
29-
; SVE-NO-VECLIB: LV: Found an estimated cost of 2 for VF 1 For instruction: %res = frem double %in, %in
30-
; SVE-NO-VECLIB: LV: Found an estimated cost of 8 for VF 2 For instruction: %res = frem double %in, %in
29+
; SVE-NO-VECLIB: LV: Found an estimated cost of 10 for VF 1 For instruction: %res = frem double %in, %in
30+
; SVE-NO-VECLIB: LV: Found an estimated cost of 24 for VF 2 For instruction: %res = frem double %in, %in
3131
; SVE-NO-VECLIB: LV: Found an estimated cost of Invalid for VF vscale x 1 For instruction: %res = frem double %in, %in
3232
; SVE-NO-VECLIB: LV: Found an estimated cost of Invalid for VF vscale x 2 For instruction: %res = frem double %in, %in
3333
;
3434
; NEON-ARMPL-LABEL: 'frem_f64'
35-
; NEON-ARMPL: LV: Found an estimated cost of 2 for VF 1 For instruction: %res = frem double %in, %in
36-
; NEON-ARMPL: LV: Found an estimated cost of 8 for VF 2 For instruction: %res = frem double %in, %in
35+
; NEON-ARMPL: LV: Found an estimated cost of 10 for VF 1 For instruction: %res = frem double %in, %in
36+
; NEON-ARMPL: LV: Found an estimated cost of 10 for VF 2 For instruction: %res = frem double %in, %in
3737
;
3838
; NEON-SLEEF-LABEL: 'frem_f64'
39-
; NEON-SLEEF: LV: Found an estimated cost of 2 for VF 1 For instruction: %res = frem double %in, %in
40-
; NEON-SLEEF: LV: Found an estimated cost of 8 for VF 2 For instruction: %res = frem double %in, %in
39+
; NEON-SLEEF: LV: Found an estimated cost of 10 for VF 1 For instruction: %res = frem double %in, %in
40+
; NEON-SLEEF: LV: Found an estimated cost of 10 for VF 2 For instruction: %res = frem double %in, %in
4141
;
4242
; SVE-ARMPL-LABEL: 'frem_f64'
43-
; SVE-ARMPL: LV: Found an estimated cost of 2 for VF 1 For instruction: %res = frem double %in, %in
44-
; SVE-ARMPL: LV: Found an estimated cost of 8 for VF 2 For instruction: %res = frem double %in, %in
43+
; SVE-ARMPL: LV: Found an estimated cost of 10 for VF 1 For instruction: %res = frem double %in, %in
44+
; SVE-ARMPL: LV: Found an estimated cost of 10 for VF 2 For instruction: %res = frem double %in, %in
4545
; SVE-ARMPL: LV: Found an estimated cost of Invalid for VF vscale x 1 For instruction: %res = frem double %in, %in
4646
; SVE-ARMPL: LV: Found an estimated cost of 10 for VF vscale x 2 For instruction: %res = frem double %in, %in
4747
;
4848
; SVE-SLEEF-LABEL: 'frem_f64'
49-
; SVE-SLEEF: LV: Found an estimated cost of 2 for VF 1 For instruction: %res = frem double %in, %in
50-
; SVE-SLEEF: LV: Found an estimated cost of 8 for VF 2 For instruction: %res = frem double %in, %in
49+
; SVE-SLEEF: LV: Found an estimated cost of 10 for VF 1 For instruction: %res = frem double %in, %in
50+
; SVE-SLEEF: LV: Found an estimated cost of 10 for VF 2 For instruction: %res = frem double %in, %in
5151
; SVE-SLEEF: LV: Found an estimated cost of Invalid for VF vscale x 1 For instruction: %res = frem double %in, %in
5252
; SVE-SLEEF: LV: Found an estimated cost of 10 for VF vscale x 2 For instruction: %res = frem double %in, %in
5353
;
5454
; SVE-ARMPL-TAILFOLD-LABEL: 'frem_f64'
55-
; SVE-ARMPL-TAILFOLD: LV: Found an estimated cost of 2 for VF 1 For instruction: %res = frem double %in, %in
56-
; SVE-ARMPL-TAILFOLD: LV: Found an estimated cost of 8 for VF 2 For instruction: %res = frem double %in, %in
55+
; SVE-ARMPL-TAILFOLD: LV: Found an estimated cost of 10 for VF 1 For instruction: %res = frem double %in, %in
56+
; SVE-ARMPL-TAILFOLD: LV: Found an estimated cost of 10 for VF 2 For instruction: %res = frem double %in, %in
5757
; SVE-ARMPL-TAILFOLD: LV: Found an estimated cost of Invalid for VF vscale x 1 For instruction: %res = frem double %in, %in
5858
; SVE-ARMPL-TAILFOLD: LV: Found an estimated cost of 10 for VF vscale x 2 For instruction: %res = frem double %in, %in
5959
;
6060
; SVE-SLEEF-TAILFOLD-LABEL: 'frem_f64'
61-
; SVE-SLEEF-TAILFOLD: LV: Found an estimated cost of 2 for VF 1 For instruction: %res = frem double %in, %in
62-
; SVE-SLEEF-TAILFOLD: LV: Found an estimated cost of 8 for VF 2 For instruction: %res = frem double %in, %in
61+
; SVE-SLEEF-TAILFOLD: LV: Found an estimated cost of 10 for VF 1 For instruction: %res = frem double %in, %in
62+
; SVE-SLEEF-TAILFOLD: LV: Found an estimated cost of 10 for VF 2 For instruction: %res = frem double %in, %in
6363
; SVE-SLEEF-TAILFOLD: LV: Found an estimated cost of Invalid for VF vscale x 1 For instruction: %res = frem double %in, %in
6464
; SVE-SLEEF-TAILFOLD: LV: Found an estimated cost of 10 for VF vscale x 2 For instruction: %res = frem double %in, %in
6565
;
@@ -83,55 +83,55 @@ define void @frem_f64(ptr noalias %in.ptr, ptr noalias %out.ptr) {
8383

8484
define void @frem_f32(ptr noalias %in.ptr, ptr noalias %out.ptr) {
8585
; NEON-NO-VECLIB-LABEL: 'frem_f32'
86-
; NEON-NO-VECLIB: LV: Found an estimated cost of 2 for VF 1 For instruction: %res = frem float %in, %in
87-
; NEON-NO-VECLIB: LV: Found an estimated cost of 8 for VF 2 For instruction: %res = frem float %in, %in
88-
; NEON-NO-VECLIB: LV: Found an estimated cost of 20 for VF 4 For instruction: %res = frem float %in, %in
86+
; NEON-NO-VECLIB: LV: Found an estimated cost of 10 for VF 1 For instruction: %res = frem float %in, %in
87+
; NEON-NO-VECLIB: LV: Found an estimated cost of 24 for VF 2 For instruction: %res = frem float %in, %in
88+
; NEON-NO-VECLIB: LV: Found an estimated cost of 52 for VF 4 For instruction: %res = frem float %in, %in
8989
;
9090
; SVE-NO-VECLIB-LABEL: 'frem_f32'
91-
; SVE-NO-VECLIB: LV: Found an estimated cost of 2 for VF 1 For instruction: %res = frem float %in, %in
92-
; SVE-NO-VECLIB: LV: Found an estimated cost of 8 for VF 2 For instruction: %res = frem float %in, %in
93-
; SVE-NO-VECLIB: LV: Found an estimated cost of 20 for VF 4 For instruction: %res = frem float %in, %in
91+
; SVE-NO-VECLIB: LV: Found an estimated cost of 10 for VF 1 For instruction: %res = frem float %in, %in
92+
; SVE-NO-VECLIB: LV: Found an estimated cost of 24 for VF 2 For instruction: %res = frem float %in, %in
93+
; SVE-NO-VECLIB: LV: Found an estimated cost of 52 for VF 4 For instruction: %res = frem float %in, %in
9494
; SVE-NO-VECLIB: LV: Found an estimated cost of Invalid for VF vscale x 1 For instruction: %res = frem float %in, %in
9595
; SVE-NO-VECLIB: LV: Found an estimated cost of Invalid for VF vscale x 2 For instruction: %res = frem float %in, %in
9696
; SVE-NO-VECLIB: LV: Found an estimated cost of Invalid for VF vscale x 4 For instruction: %res = frem float %in, %in
9797
;
9898
; NEON-ARMPL-LABEL: 'frem_f32'
99-
; NEON-ARMPL: LV: Found an estimated cost of 2 for VF 1 For instruction: %res = frem float %in, %in
100-
; NEON-ARMPL: LV: Found an estimated cost of 8 for VF 2 For instruction: %res = frem float %in, %in
99+
; NEON-ARMPL: LV: Found an estimated cost of 10 for VF 1 For instruction: %res = frem float %in, %in
100+
; NEON-ARMPL: LV: Found an estimated cost of 24 for VF 2 For instruction: %res = frem float %in, %in
101101
; NEON-ARMPL: LV: Found an estimated cost of 10 for VF 4 For instruction: %res = frem float %in, %in
102102
;
103103
; NEON-SLEEF-LABEL: 'frem_f32'
104-
; NEON-SLEEF: LV: Found an estimated cost of 2 for VF 1 For instruction: %res = frem float %in, %in
105-
; NEON-SLEEF: LV: Found an estimated cost of 8 for VF 2 For instruction: %res = frem float %in, %in
104+
; NEON-SLEEF: LV: Found an estimated cost of 10 for VF 1 For instruction: %res = frem float %in, %in
105+
; NEON-SLEEF: LV: Found an estimated cost of 24 for VF 2 For instruction: %res = frem float %in, %in
106106
; NEON-SLEEF: LV: Found an estimated cost of 10 for VF 4 For instruction: %res = frem float %in, %in
107107
;
108108
; SVE-ARMPL-LABEL: 'frem_f32'
109-
; SVE-ARMPL: LV: Found an estimated cost of 2 for VF 1 For instruction: %res = frem float %in, %in
110-
; SVE-ARMPL: LV: Found an estimated cost of 8 for VF 2 For instruction: %res = frem float %in, %in
109+
; SVE-ARMPL: LV: Found an estimated cost of 10 for VF 1 For instruction: %res = frem float %in, %in
110+
; SVE-ARMPL: LV: Found an estimated cost of 24 for VF 2 For instruction: %res = frem float %in, %in
111111
; SVE-ARMPL: LV: Found an estimated cost of 10 for VF 4 For instruction: %res = frem float %in, %in
112112
; SVE-ARMPL: LV: Found an estimated cost of Invalid for VF vscale x 1 For instruction: %res = frem float %in, %in
113113
; SVE-ARMPL: LV: Found an estimated cost of Invalid for VF vscale x 2 For instruction: %res = frem float %in, %in
114114
; SVE-ARMPL: LV: Found an estimated cost of 10 for VF vscale x 4 For instruction: %res = frem float %in, %in
115115
;
116116
; SVE-SLEEF-LABEL: 'frem_f32'
117-
; SVE-SLEEF: LV: Found an estimated cost of 2 for VF 1 For instruction: %res = frem float %in, %in
118-
; SVE-SLEEF: LV: Found an estimated cost of 8 for VF 2 For instruction: %res = frem float %in, %in
117+
; SVE-SLEEF: LV: Found an estimated cost of 10 for VF 1 For instruction: %res = frem float %in, %in
118+
; SVE-SLEEF: LV: Found an estimated cost of 24 for VF 2 For instruction: %res = frem float %in, %in
119119
; SVE-SLEEF: LV: Found an estimated cost of 10 for VF 4 For instruction: %res = frem float %in, %in
120120
; SVE-SLEEF: LV: Found an estimated cost of Invalid for VF vscale x 1 For instruction: %res = frem float %in, %in
121121
; SVE-SLEEF: LV: Found an estimated cost of Invalid for VF vscale x 2 For instruction: %res = frem float %in, %in
122122
; SVE-SLEEF: LV: Found an estimated cost of 10 for VF vscale x 4 For instruction: %res = frem float %in, %in
123123
;
124124
; SVE-ARMPL-TAILFOLD-LABEL: 'frem_f32'
125-
; SVE-ARMPL-TAILFOLD: LV: Found an estimated cost of 2 for VF 1 For instruction: %res = frem float %in, %in
126-
; SVE-ARMPL-TAILFOLD: LV: Found an estimated cost of 8 for VF 2 For instruction: %res = frem float %in, %in
125+
; SVE-ARMPL-TAILFOLD: LV: Found an estimated cost of 10 for VF 1 For instruction: %res = frem float %in, %in
126+
; SVE-ARMPL-TAILFOLD: LV: Found an estimated cost of 24 for VF 2 For instruction: %res = frem float %in, %in
127127
; SVE-ARMPL-TAILFOLD: LV: Found an estimated cost of 10 for VF 4 For instruction: %res = frem float %in, %in
128128
; SVE-ARMPL-TAILFOLD: LV: Found an estimated cost of Invalid for VF vscale x 1 For instruction: %res = frem float %in, %in
129129
; SVE-ARMPL-TAILFOLD: LV: Found an estimated cost of Invalid for VF vscale x 2 For instruction: %res = frem float %in, %in
130130
; SVE-ARMPL-TAILFOLD: LV: Found an estimated cost of 10 for VF vscale x 4 For instruction: %res = frem float %in, %in
131131
;
132132
; SVE-SLEEF-TAILFOLD-LABEL: 'frem_f32'
133-
; SVE-SLEEF-TAILFOLD: LV: Found an estimated cost of 2 for VF 1 For instruction: %res = frem float %in, %in
134-
; SVE-SLEEF-TAILFOLD: LV: Found an estimated cost of 8 for VF 2 For instruction: %res = frem float %in, %in
133+
; SVE-SLEEF-TAILFOLD: LV: Found an estimated cost of 10 for VF 1 For instruction: %res = frem float %in, %in
134+
; SVE-SLEEF-TAILFOLD: LV: Found an estimated cost of 24 for VF 2 For instruction: %res = frem float %in, %in
135135
; SVE-SLEEF-TAILFOLD: LV: Found an estimated cost of 10 for VF 4 For instruction: %res = frem float %in, %in
136136
; SVE-SLEEF-TAILFOLD: LV: Found an estimated cost of Invalid for VF vscale x 1 For instruction: %res = frem float %in, %in
137137
; SVE-SLEEF-TAILFOLD: LV: Found an estimated cost of Invalid for VF vscale x 2 For instruction: %res = frem float %in, %in

‎llvm/test/Analysis/CostModel/AArch64/arith-fp.ll

Lines changed: 11 additions & 11 deletions
Original file line numberDiff line numberDiff line change
@@ -197,17 +197,17 @@ define i32 @fdiv(i32 %arg) {
197197

198198
define i32 @frem(i32 %arg) {
199199
; CHECK-LABEL: 'frem'
200-
; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F16 = frem half undef, undef
201-
; CHECK-NEXT: Cost Model: Found an estimated cost of 20 for instruction: %V4F16 = frem <4 x half> undef, undef
202-
; CHECK-NEXT: Cost Model: Found an estimated cost of 44 for instruction: %V8F16 = frem <8 x half> undef, undef
203-
; CHECK-NEXT: Cost Model: Found an estimated cost of 88 for instruction: %V16F16 = frem <16 x half> undef, undef
204-
; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F32 = frem float undef, undef
205-
; CHECK-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V2F32 = frem <2 x float> undef, undef
206-
; CHECK-NEXT: Cost Model: Found an estimated cost of 20 for instruction: %V4F32 = frem <4 x float> undef, undef
207-
; CHECK-NEXT: Cost Model: Found an estimated cost of 40 for instruction: %V8F32 = frem <8 x float> undef, undef
208-
; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %F64 = frem double undef, undef
209-
; CHECK-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %V2F64 = frem <2 x double> undef, undef
210-
; CHECK-NEXT: Cost Model: Found an estimated cost of 16 for instruction: %V4F64 = frem <4 x double> undef, undef
200+
; CHECK-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %F16 = frem half undef, undef
201+
; CHECK-NEXT: Cost Model: Found an estimated cost of 52 for instruction: %V4F16 = frem <4 x half> undef, undef
202+
; CHECK-NEXT: Cost Model: Found an estimated cost of 108 for instruction: %V8F16 = frem <8 x half> undef, undef
203+
; CHECK-NEXT: Cost Model: Found an estimated cost of 216 for instruction: %V16F16 = frem <16 x half> undef, undef
204+
; CHECK-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %F32 = frem float undef, undef
205+
; CHECK-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %V2F32 = frem <2 x float> undef, undef
206+
; CHECK-NEXT: Cost Model: Found an estimated cost of 52 for instruction: %V4F32 = frem <4 x float> undef, undef
207+
; CHECK-NEXT: Cost Model: Found an estimated cost of 104 for instruction: %V8F32 = frem <8 x float> undef, undef
208+
; CHECK-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %F64 = frem double undef, undef
209+
; CHECK-NEXT: Cost Model: Found an estimated cost of 24 for instruction: %V2F64 = frem <2 x double> undef, undef
210+
; CHECK-NEXT: Cost Model: Found an estimated cost of 48 for instruction: %V4F64 = frem <4 x double> undef, undef
211211
; CHECK-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
212212
;
213213
%F16 = frem half undef, undef

0 commit comments

Comments
 (0)