Skip to content

Commit bded3b3

Browse files
jroelofsjroelofs
authored
[llvm][AArch64] Improve the cost model for i128 div's (#107306)
1 parent 16dc65b commit bded3b3

File tree

2 files changed

+50
-1
lines changed

2 files changed

+50
-1
lines changed

‎llvm/lib/Target/AArch64/AArch64TargetTransformInfo.cpp

Lines changed: 7 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -3224,8 +3224,8 @@ InstructionCost AArch64TTIImpl::getArithmeticInstrCost(
32243224
}
32253225
[[fallthrough]];
32263226
case ISD::UDIV: {
3227+
auto VT = TLI->getValueType(DL, Ty);
32273228
if (Op2Info.isConstant() && Op2Info.isUniform()) {
3228-
auto VT = TLI->getValueType(DL, Ty);
32293229
if (TLI->isOperationLegalOrCustom(ISD::MULHU, VT)) {
32303230
// Vector signed division by constant are expanded to the
32313231
// sequence MULHS + ADD/SUB + SRA + SRL + ADD, and unsigned division
@@ -3240,6 +3240,12 @@ InstructionCost AArch64TTIImpl::getArithmeticInstrCost(
32403240
}
32413241
}
32423242

3243+
// div i128's are lowered as libcalls. Pass nullptr as (u)divti3 calls are
3244+
// emitted by the backend even when those functions are not declared in the
3245+
// module.
3246+
if (!VT.isVector() && VT.getSizeInBits() > 64)
3247+
return getCallInstrCost(/*Function*/ nullptr, Ty, {Ty, Ty}, CostKind);
3248+
32433249
InstructionCost Cost = BaseT::getArithmeticInstrCost(
32443250
Opcode, Ty, CostKind, Op1Info, Op2Info);
32453251
if (Ty->isVectorTy()) {

‎llvm/test/Analysis/CostModel/AArch64/div.ll

Lines changed: 43 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -5,6 +5,7 @@ target datalayout = "e-m:e-i8:8:32-i16:16:32-i64:64-i128:128-n32:64-S128"
55

66
define i32 @sdiv() {
77
; CHECK-LABEL: 'sdiv'
8+
; CHECK-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %I128 = sdiv i128 undef, undef
89
; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %I64 = sdiv i64 undef, undef
910
; CHECK-NEXT: Cost Model: Found an estimated cost of 28 for instruction: %V2i64 = sdiv <2 x i64> undef, undef
1011
; CHECK-NEXT: Cost Model: Found an estimated cost of 56 for instruction: %V4i64 = sdiv <4 x i64> undef, undef
@@ -23,6 +24,8 @@ define i32 @sdiv() {
2324
; CHECK-NEXT: Cost Model: Found an estimated cost of 672 for instruction: %V64i8 = sdiv <64 x i8> undef, undef
2425
; CHECK-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
2526
;
27+
%I128 = sdiv i128 undef, undef
28+
2629
%I64 = sdiv i64 undef, undef
2730
%V2i64 = sdiv <2 x i64> undef, undef
2831
%V4i64 = sdiv <4 x i64> undef, undef
@@ -48,6 +51,7 @@ define i32 @sdiv() {
4851

4952
define i32 @udiv() {
5053
; CHECK-LABEL: 'udiv'
54+
; CHECK-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %I128 = udiv i128 undef, undef
5155
; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %I64 = udiv i64 undef, undef
5256
; CHECK-NEXT: Cost Model: Found an estimated cost of 28 for instruction: %V2i64 = udiv <2 x i64> undef, undef
5357
; CHECK-NEXT: Cost Model: Found an estimated cost of 56 for instruction: %V4i64 = udiv <4 x i64> undef, undef
@@ -66,6 +70,8 @@ define i32 @udiv() {
6670
; CHECK-NEXT: Cost Model: Found an estimated cost of 672 for instruction: %V64i8 = udiv <64 x i8> undef, undef
6771
; CHECK-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
6872
;
73+
%I128 = udiv i128 undef, undef
74+
6975
%I64 = udiv i64 undef, undef
7076
%V2i64 = udiv <2 x i64> undef, undef
7177
%V4i64 = udiv <4 x i64> undef, undef
@@ -91,6 +97,7 @@ define i32 @udiv() {
9197

9298
define i32 @sdiv_const() {
9399
; CHECK-LABEL: 'sdiv_const'
100+
; CHECK-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %I128 = sdiv i128 undef, 7
94101
; CHECK-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %I64 = sdiv i64 undef, 7
95102
; CHECK-NEXT: Cost Model: Found an estimated cost of 28 for instruction: %V2i64 = sdiv <2 x i64> undef, <i64 6, i64 7>
96103
; CHECK-NEXT: Cost Model: Found an estimated cost of 56 for instruction: %V4i64 = sdiv <4 x i64> undef, <i64 4, i64 5, i64 6, i64 7>
@@ -109,6 +116,8 @@ define i32 @sdiv_const() {
109116
; CHECK-NEXT: Cost Model: Found an estimated cost of 672 for instruction: %V64i8 = sdiv <64 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
110117
; CHECK-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
111118
;
119+
%I128 = sdiv i128 undef, 7
120+
112121
%I64 = sdiv i64 undef, 7
113122
%V2i64 = sdiv <2 x i64> undef, <i64 6, i64 7>
114123
%V4i64 = sdiv <4 x i64> undef, <i64 4, i64 5, i64 6, i64 7>
@@ -134,6 +143,7 @@ define i32 @sdiv_const() {
134143

135144
define i32 @udiv_const() {
136145
; CHECK-LABEL: 'udiv_const'
146+
; CHECK-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %I128 = udiv i128 undef, 7
137147
; CHECK-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %I64 = udiv i64 undef, 7
138148
; CHECK-NEXT: Cost Model: Found an estimated cost of 28 for instruction: %V2i64 = udiv <2 x i64> undef, <i64 6, i64 7>
139149
; CHECK-NEXT: Cost Model: Found an estimated cost of 56 for instruction: %V4i64 = udiv <4 x i64> undef, <i64 4, i64 5, i64 6, i64 7>
@@ -152,6 +162,9 @@ define i32 @udiv_const() {
152162
; CHECK-NEXT: Cost Model: Found an estimated cost of 672 for instruction: %V64i8 = udiv <64 x i8> undef, <i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19>
153163
; CHECK-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
154164
;
165+
166+
%I128 = udiv i128 undef, 7
167+
155168
%I64 = udiv i64 undef, 7
156169
%V2i64 = udiv <2 x i64> undef, <i64 6, i64 7>
157170
%V4i64 = udiv <4 x i64> undef, <i64 4, i64 5, i64 6, i64 7>
@@ -177,6 +190,7 @@ define i32 @udiv_const() {
177190

178191
define i32 @sdiv_uniformconst() {
179192
; CHECK-LABEL: 'sdiv_uniformconst'
193+
; CHECK-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %I128 = sdiv i128 undef, 7
180194
; CHECK-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %I64 = sdiv i64 undef, 7
181195
; CHECK-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %V2i64 = sdiv <2 x i64> undef, <i64 7, i64 7>
182196
; CHECK-NEXT: Cost Model: Found an estimated cost of 20 for instruction: %V4i64 = sdiv <4 x i64> undef, <i64 7, i64 7, i64 7, i64 7>
@@ -195,6 +209,8 @@ define i32 @sdiv_uniformconst() {
195209
; CHECK-NEXT: Cost Model: Found an estimated cost of 320 for instruction: %V64i8 = sdiv <64 x i8> undef, <i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7>
196210
; CHECK-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
197211
;
212+
%I128 = sdiv i128 undef, 7
213+
198214
%I64 = sdiv i64 undef, 7
199215
%V2i64 = sdiv <2 x i64> undef, <i64 7, i64 7>
200216
%V4i64 = sdiv <4 x i64> undef, <i64 7, i64 7, i64 7, i64 7>
@@ -220,6 +236,7 @@ define i32 @sdiv_uniformconst() {
220236

221237
define i32 @udiv_uniformconst() {
222238
; CHECK-LABEL: 'udiv_uniformconst'
239+
; CHECK-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %I128 = udiv i128 undef, 7
223240
; CHECK-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %I64 = udiv i64 undef, 7
224241
; CHECK-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %V2i64 = udiv <2 x i64> undef, <i64 7, i64 7>
225242
; CHECK-NEXT: Cost Model: Found an estimated cost of 20 for instruction: %V4i64 = udiv <4 x i64> undef, <i64 7, i64 7, i64 7, i64 7>
@@ -238,6 +255,8 @@ define i32 @udiv_uniformconst() {
238255
; CHECK-NEXT: Cost Model: Found an estimated cost of 320 for instruction: %V64i8 = udiv <64 x i8> undef, <i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7>
239256
; CHECK-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
240257
;
258+
%I128 = udiv i128 undef, 7
259+
241260
%I64 = udiv i64 undef, 7
242261
%V2i64 = udiv <2 x i64> undef, <i64 7, i64 7>
243262
%V4i64 = udiv <4 x i64> undef, <i64 7, i64 7, i64 7, i64 7>
@@ -263,6 +282,7 @@ define i32 @udiv_uniformconst() {
263282

264283
define i32 @sdiv_constpow2() {
265284
; CHECK-LABEL: 'sdiv_constpow2'
285+
; CHECK-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %I128 = sdiv i128 undef, 16
266286
; CHECK-NEXT: Cost Model: Found an estimated cost of 5 for instruction: %I64 = sdiv i64 undef, 16
267287
; CHECK-NEXT: Cost Model: Found an estimated cost of 28 for instruction: %V2i64 = sdiv <2 x i64> undef, <i64 8, i64 16>
268288
; CHECK-NEXT: Cost Model: Found an estimated cost of 56 for instruction: %V4i64 = sdiv <4 x i64> undef, <i64 2, i64 4, i64 8, i64 16>
@@ -281,6 +301,8 @@ define i32 @sdiv_constpow2() {
281301
; CHECK-NEXT: Cost Model: Found an estimated cost of 672 for instruction: %V64i8 = sdiv <64 x i8> undef, <i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16>
282302
; CHECK-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
283303
;
304+
%I128 = sdiv i128 undef, 16
305+
284306
%I64 = sdiv i64 undef, 16
285307
%V2i64 = sdiv <2 x i64> undef, <i64 8, i64 16>
286308
%V4i64 = sdiv <4 x i64> undef, <i64 2, i64 4, i64 8, i64 16>
@@ -306,6 +328,7 @@ define i32 @sdiv_constpow2() {
306328

307329
define i32 @udiv_constpow2() {
308330
; CHECK-LABEL: 'udiv_constpow2'
331+
; CHECK-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %I128 = udiv i128 undef, 16
309332
; CHECK-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %I64 = udiv i64 undef, 16
310333
; CHECK-NEXT: Cost Model: Found an estimated cost of 28 for instruction: %V2i64 = udiv <2 x i64> undef, <i64 8, i64 16>
311334
; CHECK-NEXT: Cost Model: Found an estimated cost of 56 for instruction: %V4i64 = udiv <4 x i64> undef, <i64 2, i64 4, i64 8, i64 16>
@@ -324,6 +347,8 @@ define i32 @udiv_constpow2() {
324347
; CHECK-NEXT: Cost Model: Found an estimated cost of 672 for instruction: %V64i8 = udiv <64 x i8> undef, <i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16, i8 2, i8 4, i8 8, i8 16>
325348
; CHECK-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
326349
;
350+
%I128 = udiv i128 undef, 16
351+
327352
%I64 = udiv i64 undef, 16
328353
%V2i64 = udiv <2 x i64> undef, <i64 8, i64 16>
329354
%V4i64 = udiv <4 x i64> undef, <i64 2, i64 4, i64 8, i64 16>
@@ -349,6 +374,7 @@ define i32 @udiv_constpow2() {
349374

350375
define i32 @sdiv_uniformconstpow2() {
351376
; CHECK-LABEL: 'sdiv_uniformconstpow2'
377+
; CHECK-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %I128 = sdiv i128 undef, 16
352378
; CHECK-NEXT: Cost Model: Found an estimated cost of 5 for instruction: %I64 = sdiv i64 undef, 16
353379
; CHECK-NEXT: Cost Model: Found an estimated cost of 15 for instruction: %V2i64 = sdiv <2 x i64> undef, <i64 16, i64 16>
354380
; CHECK-NEXT: Cost Model: Found an estimated cost of 30 for instruction: %V4i64 = sdiv <4 x i64> undef, <i64 16, i64 16, i64 16, i64 16>
@@ -367,6 +393,8 @@ define i32 @sdiv_uniformconstpow2() {
367393
; CHECK-NEXT: Cost Model: Found an estimated cost of 396 for instruction: %V64i8 = sdiv <64 x i8> undef, <i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16>
368394
; CHECK-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
369395
;
396+
%I128 = sdiv i128 undef, 16
397+
370398
%I64 = sdiv i64 undef, 16
371399
%V2i64 = sdiv <2 x i64> undef, <i64 16, i64 16>
372400
%V4i64 = sdiv <4 x i64> undef, <i64 16, i64 16, i64 16, i64 16>
@@ -392,6 +420,7 @@ define i32 @sdiv_uniformconstpow2() {
392420

393421
define i32 @udiv_uniformconstpow2() {
394422
; CHECK-LABEL: 'udiv_uniformconstpow2'
423+
; CHECK-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %I128 = udiv i128 undef, 16
395424
; CHECK-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %I64 = udiv i64 undef, 16
396425
; CHECK-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %V2i64 = udiv <2 x i64> undef, <i64 16, i64 16>
397426
; CHECK-NEXT: Cost Model: Found an estimated cost of 20 for instruction: %V4i64 = udiv <4 x i64> undef, <i64 16, i64 16, i64 16, i64 16>
@@ -410,6 +439,8 @@ define i32 @udiv_uniformconstpow2() {
410439
; CHECK-NEXT: Cost Model: Found an estimated cost of 320 for instruction: %V64i8 = udiv <64 x i8> undef, <i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16, i8 16>
411440
; CHECK-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
412441
;
442+
%I128 = udiv i128 undef, 16
443+
413444
%I64 = udiv i64 undef, 16
414445
%V2i64 = udiv <2 x i64> undef, <i64 16, i64 16>
415446
%V4i64 = udiv <4 x i64> undef, <i64 16, i64 16, i64 16, i64 16>
@@ -435,6 +466,7 @@ define i32 @udiv_uniformconstpow2() {
435466

436467
define i32 @sdiv_constnegpow2() {
437468
; CHECK-LABEL: 'sdiv_constnegpow2'
469+
; CHECK-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %I128 = sdiv i128 undef, -16
438470
; CHECK-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %I64 = sdiv i64 undef, -16
439471
; CHECK-NEXT: Cost Model: Found an estimated cost of 28 for instruction: %V2i64 = sdiv <2 x i64> undef, <i64 -8, i64 -16>
440472
; CHECK-NEXT: Cost Model: Found an estimated cost of 56 for instruction: %V4i64 = sdiv <4 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16>
@@ -453,6 +485,8 @@ define i32 @sdiv_constnegpow2() {
453485
; CHECK-NEXT: Cost Model: Found an estimated cost of 672 for instruction: %V64i8 = sdiv <64 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
454486
; CHECK-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
455487
;
488+
%I128 = sdiv i128 undef, -16
489+
456490
%I64 = sdiv i64 undef, -16
457491
%V2i64 = sdiv <2 x i64> undef, <i64 -8, i64 -16>
458492
%V4i64 = sdiv <4 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16>
@@ -478,6 +512,7 @@ define i32 @sdiv_constnegpow2() {
478512

479513
define i32 @udiv_constnegpow2() {
480514
; CHECK-LABEL: 'udiv_constnegpow2'
515+
; CHECK-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %I128 = udiv i128 undef, -16
481516
; CHECK-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %I64 = udiv i64 undef, -16
482517
; CHECK-NEXT: Cost Model: Found an estimated cost of 28 for instruction: %V2i64 = udiv <2 x i64> undef, <i64 -8, i64 -16>
483518
; CHECK-NEXT: Cost Model: Found an estimated cost of 56 for instruction: %V4i64 = udiv <4 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16>
@@ -496,6 +531,8 @@ define i32 @udiv_constnegpow2() {
496531
; CHECK-NEXT: Cost Model: Found an estimated cost of 672 for instruction: %V64i8 = udiv <64 x i8> undef, <i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16, i8 -2, i8 -4, i8 -8, i8 -16>
497532
; CHECK-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
498533
;
534+
%I128 = udiv i128 undef, -16
535+
499536
%I64 = udiv i64 undef, -16
500537
%V2i64 = udiv <2 x i64> undef, <i64 -8, i64 -16>
501538
%V4i64 = udiv <4 x i64> undef, <i64 -2, i64 -4, i64 -8, i64 -16>
@@ -521,6 +558,7 @@ define i32 @udiv_constnegpow2() {
521558

522559
define i32 @sdiv_uniformconstnegpow2() {
523560
; CHECK-LABEL: 'sdiv_uniformconstnegpow2'
561+
; CHECK-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %I128 = sdiv i128 undef, -16
524562
; CHECK-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %I64 = sdiv i64 undef, -16
525563
; CHECK-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %V2i64 = sdiv <2 x i64> undef, <i64 -16, i64 -16>
526564
; CHECK-NEXT: Cost Model: Found an estimated cost of 20 for instruction: %V4i64 = sdiv <4 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16>
@@ -539,6 +577,8 @@ define i32 @sdiv_uniformconstnegpow2() {
539577
; CHECK-NEXT: Cost Model: Found an estimated cost of 320 for instruction: %V64i8 = sdiv <64 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
540578
; CHECK-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
541579
;
580+
%I128 = sdiv i128 undef, -16
581+
542582
%I64 = sdiv i64 undef, -16
543583
%V2i64 = sdiv <2 x i64> undef, <i64 -16, i64 -16>
544584
%V4i64 = sdiv <4 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16>
@@ -564,6 +604,7 @@ define i32 @sdiv_uniformconstnegpow2() {
564604

565605
define i32 @udiv_uniformconstnegpow2() {
566606
; CHECK-LABEL: 'udiv_uniformconstnegpow2'
607+
; CHECK-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %I128 = udiv i128 undef, -16
567608
; CHECK-NEXT: Cost Model: Found an estimated cost of 7 for instruction: %I64 = udiv i64 undef, -16
568609
; CHECK-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %V2i64 = udiv <2 x i64> undef, <i64 -16, i64 -16>
569610
; CHECK-NEXT: Cost Model: Found an estimated cost of 20 for instruction: %V4i64 = udiv <4 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16>
@@ -582,6 +623,8 @@ define i32 @udiv_uniformconstnegpow2() {
582623
; CHECK-NEXT: Cost Model: Found an estimated cost of 320 for instruction: %V64i8 = udiv <64 x i8> undef, <i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16, i8 -16>
583624
; CHECK-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
584625
;
626+
%I128 = udiv i128 undef, -16
627+
585628
%I64 = udiv i64 undef, -16
586629
%V2i64 = udiv <2 x i64> undef, <i64 -16, i64 -16>
587630
%V4i64 = udiv <4 x i64> undef, <i64 -16, i64 -16, i64 -16, i64 -16>

0 commit comments

Comments
 (0)