Skip to content

Commit 530676f

Browse files
davemgreendavemgreen
committed
[AArch64] Fold away zext of extract of uzp.
Similar to #107201, this comes up from the lowering of zext of deinterleaving shuffles. Patterns such as ext(extract_subvector(uzp(a, b))) can be converted to a simple and to perform the extract/zext from a uzp1. Uzp2 can be handled with an extra shift, and due to the existing legalization we could have and / shift between which can be combined in. Mostly this reduces instruction count or increases the amount of parallelism in the sequence.
1 parent 3413f95 commit 530676f

File tree

2 files changed

+135
-93
lines changed

2 files changed

+135
-93
lines changed

‎llvm/lib/Target/AArch64/AArch64ISelLowering.cpp

Lines changed: 66 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -22240,6 +22240,70 @@ static SDValue performZExtDeinterleaveShuffleCombine(SDNode *N,
2224022240
DAG.getConstant((1 << InVT.getScalarSizeInBits()) - 1, DL, VT));
2224122241
}
2224222242

22243+
// This comes up similar to the above when lowering deinterleaving shuffles from
22244+
// zexts. We have legalized the operations in the generally case to
22245+
// zext(extract_subvector(uzp(a, b))), which can be converted to and(a, mask) if
22246+
// the extract is to the low half and the uzp is uzp1. There would be an extra
22247+
// shift if the uzp was uzp2 to grab the upper half. Due to the combine above
22248+
// there could also be an existing and / shift that can be combined in, either
22249+
// before of after the extract.
22250+
static SDValue performZExtUZPCombine(SDNode *N, SelectionDAG &DAG) {
22251+
EVT VT = N->getValueType(0);
22252+
if (N->getOpcode() != ISD::ZERO_EXTEND ||
22253+
(VT != MVT::v2i64 && VT != MVT::v4i32 && VT != MVT::v8i16))
22254+
return SDValue();
22255+
22256+
SDValue Op = N->getOperand(0);
22257+
unsigned ExtOffset = (unsigned)-1;
22258+
if (Op.getOpcode() == ISD::EXTRACT_SUBVECTOR) {
22259+
ExtOffset = Op.getConstantOperandVal(1);
22260+
Op = Op.getOperand(0);
22261+
}
22262+
22263+
unsigned Shift = 0;
22264+
APInt Mask = APInt::getLowBitsSet(VT.getScalarSizeInBits(),
22265+
Op.getValueType().getScalarSizeInBits());
22266+
22267+
if (Op.getOpcode() == AArch64ISD::VLSHR) {
22268+
Shift = Op.getConstantOperandVal(1);
22269+
Op = Op.getOperand(0);
22270+
Mask = Mask.lshr(Shift);
22271+
}
22272+
if (Op.getOpcode() == ISD::AND &&
22273+
ISD::isConstantSplatVector(Op.getOperand(1).getNode(), Mask)) {
22274+
Op = Op.getOperand(0);
22275+
Mask = Mask.zext(VT.getScalarSizeInBits());
22276+
} else if (Op.getOpcode() == AArch64ISD::BICi) {
22277+
Mask = ~APInt(Op.getValueType().getScalarSizeInBits(),
22278+
Op.getConstantOperandVal(1) << Op.getConstantOperandVal(2));
22279+
Mask = Mask.zext(VT.getScalarSizeInBits());
22280+
Op = Op.getOperand(0);
22281+
}
22282+
22283+
if (ExtOffset == (unsigned)-1) {
22284+
if (Op.getOpcode() == ISD::EXTRACT_SUBVECTOR) {
22285+
ExtOffset = Op.getConstantOperandVal(1);
22286+
Op = Op.getOperand(0);
22287+
} else
22288+
return SDValue();
22289+
}
22290+
if (ExtOffset != 0 && ExtOffset != VT.getVectorNumElements())
22291+
return SDValue();
22292+
22293+
if (Op.getOpcode() != AArch64ISD::UZP1 && Op.getOpcode() != AArch64ISD::UZP2)
22294+
return SDValue();
22295+
if (Op.getOpcode() == AArch64ISD::UZP2)
22296+
Shift += VT.getScalarSizeInBits() / 2;
22297+
22298+
SDLoc DL(N);
22299+
SDValue BC = DAG.getNode(AArch64ISD::NVCAST, DL, VT,
22300+
Op.getOperand(ExtOffset == 0 ? 0 : 1));
22301+
if (Shift != 0)
22302+
BC = DAG.getNode(AArch64ISD::VLSHR, DL, VT, BC,
22303+
DAG.getConstant(Shift, DL, MVT::i32));
22304+
return DAG.getNode(ISD::AND, DL, VT, BC, DAG.getConstant(Mask, DL, VT));
22305+
}
22306+
2224322307
static SDValue performExtendCombine(SDNode *N,
2224422308
TargetLowering::DAGCombinerInfo &DCI,
2224522309
SelectionDAG &DAG) {
@@ -22262,6 +22326,8 @@ static SDValue performExtendCombine(SDNode *N,
2226222326

2226322327
if (SDValue R = performZExtDeinterleaveShuffleCombine(N, DAG))
2226422328
return R;
22329+
if (SDValue R = performZExtUZPCombine(N, DAG))
22330+
return R;
2226522331

2226622332
if (N->getValueType(0).isFixedLengthVector() &&
2226722333
N->getOpcode() == ISD::SIGN_EXTEND &&

‎llvm/test/CodeGen/AArch64/zext-shuffle.ll

Lines changed: 69 additions & 93 deletions
Original file line numberDiff line numberDiff line change
@@ -90,10 +90,11 @@ define <4 x i64> @v2i64_i16_04812(<16 x i16> %a) {
9090
define <4 x i64> @v2i64_i16_15913(<16 x i16> %a) {
9191
; CHECK-LABEL: v2i64_i16_15913:
9292
; CHECK: // %bb.0:
93-
; CHECK-NEXT: uzp1 v0.4s, v0.4s, v1.4s
94-
; CHECK-NEXT: ushr v0.4s, v0.4s, #16
95-
; CHECK-NEXT: ushll2 v1.2d, v0.4s, #0
96-
; CHECK-NEXT: ushll v0.2d, v0.2s, #0
93+
; CHECK-NEXT: movi v2.2d, #0x0000000000ffff
94+
; CHECK-NEXT: ushr v0.2d, v0.2d, #16
95+
; CHECK-NEXT: ushr v1.2d, v1.2d, #16
96+
; CHECK-NEXT: and v0.16b, v0.16b, v2.16b
97+
; CHECK-NEXT: and v1.16b, v1.16b, v2.16b
9798
; CHECK-NEXT: ret
9899
%s1 = shufflevector <16 x i16> %a, <16 x i16> undef, <4 x i32> <i32 1, i32 5, i32 9, i32 13>
99100
%z1 = zext <4 x i16> %s1 to <4 x i64>
@@ -117,10 +118,8 @@ define <4 x i64> @v2i64_i16_261014(<16 x i16> %a) {
117118
define <4 x i64> @v2i64_i16_371115(<16 x i16> %a) {
118119
; CHECK-LABEL: v2i64_i16_371115:
119120
; CHECK: // %bb.0:
120-
; CHECK-NEXT: uzp2 v0.4s, v0.4s, v1.4s
121-
; CHECK-NEXT: ushr v0.4s, v0.4s, #16
122-
; CHECK-NEXT: ushll2 v1.2d, v0.4s, #0
123-
; CHECK-NEXT: ushll v0.2d, v0.2s, #0
121+
; CHECK-NEXT: ushr v0.2d, v0.2d, #48
122+
; CHECK-NEXT: ushr v1.2d, v1.2d, #48
124123
; CHECK-NEXT: ret
125124
%s1 = shufflevector <16 x i16> %a, <16 x i16> undef, <4 x i32> <i32 3, i32 7, i32 11, i32 15>
126125
%z1 = zext <4 x i16> %s1 to <4 x i64>
@@ -142,8 +141,7 @@ define <4 x i32> @v4i32_0246(<8 x i16> %a, <8 x i16> %b) {
142141
define <4 x i32> @v4i32_1357(<8 x i16> %a, <8 x i16> %b) {
143142
; CHECK-LABEL: v4i32_1357:
144143
; CHECK: // %bb.0:
145-
; CHECK-NEXT: uzp2 v0.8h, v0.8h, v0.8h
146-
; CHECK-NEXT: ushll v0.4s, v0.4h, #0
144+
; CHECK-NEXT: ushr v0.4s, v0.4s, #16
147145
; CHECK-NEXT: ret
148146
%c = shufflevector <8 x i16> %a, <8 x i16> %b, <4 x i32> <i32 1, i32 3, i32 5, i32 7>
149147
%d = zext <4 x i16> %c to <4 x i32>
@@ -210,8 +208,7 @@ define <8 x i16> @v8i16_0246(<16 x i8> %a, <16 x i8> %b) {
210208
define <8 x i16> @v8i16_1357(<16 x i8> %a, <16 x i8> %b) {
211209
; CHECK-LABEL: v8i16_1357:
212210
; CHECK: // %bb.0:
213-
; CHECK-NEXT: uzp2 v0.16b, v0.16b, v0.16b
214-
; CHECK-NEXT: ushll v0.8h, v0.8b, #0
211+
; CHECK-NEXT: ushr v0.8h, v0.8h, #8
215212
; CHECK-NEXT: ret
216213
%c = shufflevector <16 x i8> %a, <16 x i8> %b, <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15>
217214
%d = zext <8 x i8> %c to <8 x i16>
@@ -278,8 +275,7 @@ define <8 x i32> @v8i32_0246(<16 x i8> %a, <16 x i8> %b) {
278275
define <8 x i32> @v8i32_1357(<16 x i8> %a, <16 x i8> %b) {
279276
; CHECK-LABEL: v8i32_1357:
280277
; CHECK: // %bb.0:
281-
; CHECK-NEXT: uzp2 v0.16b, v0.16b, v0.16b
282-
; CHECK-NEXT: ushll v0.8h, v0.8b, #0
278+
; CHECK-NEXT: ushr v0.8h, v0.8h, #8
283279
; CHECK-NEXT: ushll2 v1.4s, v0.8h, #0
284280
; CHECK-NEXT: ushll v0.4s, v0.4h, #0
285281
; CHECK-NEXT: ret
@@ -291,10 +287,9 @@ define <8 x i32> @v8i32_1357(<16 x i8> %a, <16 x i8> %b) {
291287
define <8 x i32> @v8i32_04812(<16 x i8> %a, <16 x i8> %b) {
292288
; CHECK-LABEL: v8i32_04812:
293289
; CHECK: // %bb.0:
294-
; CHECK-NEXT: uzp1 v0.8h, v0.8h, v1.8h
295-
; CHECK-NEXT: bic v0.8h, #255, lsl #8
296-
; CHECK-NEXT: ushll2 v1.4s, v0.8h, #0
297-
; CHECK-NEXT: ushll v0.4s, v0.4h, #0
290+
; CHECK-NEXT: movi v2.2d, #0x0000ff000000ff
291+
; CHECK-NEXT: and v0.16b, v0.16b, v2.16b
292+
; CHECK-NEXT: and v1.16b, v1.16b, v2.16b
298293
; CHECK-NEXT: ret
299294
%c = shufflevector <16 x i8> %a, <16 x i8> %b, <8 x i32> <i32 0, i32 4, i32 8, i32 12, i32 16, i32 20, i32 24, i32 28>
300295
%d = zext <8 x i8> %c to <8 x i32>
@@ -304,10 +299,11 @@ define <8 x i32> @v8i32_04812(<16 x i8> %a, <16 x i8> %b) {
304299
define <8 x i32> @v8i32_15913(<16 x i8> %a, <16 x i8> %b) {
305300
; CHECK-LABEL: v8i32_15913:
306301
; CHECK: // %bb.0:
307-
; CHECK-NEXT: uzp1 v0.8h, v0.8h, v1.8h
308-
; CHECK-NEXT: ushr v0.8h, v0.8h, #8
309-
; CHECK-NEXT: ushll2 v1.4s, v0.8h, #0
310-
; CHECK-NEXT: ushll v0.4s, v0.4h, #0
302+
; CHECK-NEXT: movi v2.2d, #0x0000ff000000ff
303+
; CHECK-NEXT: ushr v0.4s, v0.4s, #8
304+
; CHECK-NEXT: ushr v1.4s, v1.4s, #8
305+
; CHECK-NEXT: and v0.16b, v0.16b, v2.16b
306+
; CHECK-NEXT: and v1.16b, v1.16b, v2.16b
311307
; CHECK-NEXT: ret
312308
%c = shufflevector <16 x i8> %a, <16 x i8> %b, <8 x i32> <i32 1, i32 5, i32 9, i32 13, i32 17, i32 21, i32 25, i32 29>
313309
%d = zext <8 x i8> %c to <8 x i32>
@@ -317,10 +313,10 @@ define <8 x i32> @v8i32_15913(<16 x i8> %a, <16 x i8> %b) {
317313
define <8 x i32> @v8i32_261014(<16 x i8> %a, <16 x i8> %b) {
318314
; CHECK-LABEL: v8i32_261014:
319315
; CHECK: // %bb.0:
320-
; CHECK-NEXT: uzp2 v0.8h, v0.8h, v1.8h
321-
; CHECK-NEXT: bic v0.8h, #255, lsl #8
322-
; CHECK-NEXT: ushll2 v1.4s, v0.8h, #0
323-
; CHECK-NEXT: ushll v0.4s, v0.4h, #0
316+
; CHECK-NEXT: ushr v0.4s, v0.4s, #16
317+
; CHECK-NEXT: ushr v1.4s, v1.4s, #16
318+
; CHECK-NEXT: bic v0.4s, #255, lsl #8
319+
; CHECK-NEXT: bic v1.4s, #255, lsl #8
324320
; CHECK-NEXT: ret
325321
%c = shufflevector <16 x i8> %a, <16 x i8> %b, <8 x i32> <i32 2, i32 6, i32 10, i32 14, i32 18, i32 22, i32 26, i32 30>
326322
%d = zext <8 x i8> %c to <8 x i32>
@@ -330,10 +326,8 @@ define <8 x i32> @v8i32_261014(<16 x i8> %a, <16 x i8> %b) {
330326
define <8 x i32> @v8i32_371115(<16 x i8> %a, <16 x i8> %b) {
331327
; CHECK-LABEL: v8i32_371115:
332328
; CHECK: // %bb.0:
333-
; CHECK-NEXT: uzp2 v0.8h, v0.8h, v1.8h
334-
; CHECK-NEXT: ushr v0.8h, v0.8h, #8
335-
; CHECK-NEXT: ushll2 v1.4s, v0.8h, #0
336-
; CHECK-NEXT: ushll v0.4s, v0.4h, #0
329+
; CHECK-NEXT: ushr v0.4s, v0.4s, #24
330+
; CHECK-NEXT: ushr v1.4s, v1.4s, #24
337331
; CHECK-NEXT: ret
338332
%c = shufflevector <16 x i8> %a, <16 x i8> %b, <8 x i32> <i32 3, i32 7, i32 11, i32 15, i32 19, i32 23, i32 27, i32 31>
339333
%d = zext <8 x i8> %c to <8 x i32>
@@ -407,77 +401,59 @@ define <8 x i64> @zext_load_add(ptr %p) {
407401
define <8 x double> @uitofp_fadd(<32 x i16> %l) {
408402
; CHECK-LABEL: uitofp_fadd:
409403
; CHECK: // %bb.0:
410-
; CHECK-NEXT: uzp1 v5.4s, v0.4s, v3.4s
411-
; CHECK-NEXT: uzp1 v6.4s, v0.4s, v1.4s
412-
; CHECK-NEXT: uzp2 v0.4s, v0.4s, v1.4s
413-
; CHECK-NEXT: movi d4, #0x00ffff0000ffff
414-
; CHECK-NEXT: uzp1 v7.4s, v2.4s, v3.4s
415-
; CHECK-NEXT: uzp2 v2.4s, v2.4s, v3.4s
416-
; CHECK-NEXT: ext v16.16b, v6.16b, v6.16b, #8
417-
; CHECK-NEXT: ext v5.16b, v5.16b, v5.16b, #8
418-
; CHECK-NEXT: uzp2 v1.4s, v0.4s, v3.4s
419-
; CHECK-NEXT: and v17.8b, v6.8b, v4.8b
420-
; CHECK-NEXT: and v18.8b, v7.8b, v4.8b
421-
; CHECK-NEXT: ushr v6.2s, v6.2s, #16
422-
; CHECK-NEXT: ushr v7.2s, v7.2s, #16
423-
; CHECK-NEXT: and v21.8b, v0.8b, v4.8b
424-
; CHECK-NEXT: and v22.8b, v2.8b, v4.8b
425-
; CHECK-NEXT: ushr v2.2s, v2.2s, #16
426-
; CHECK-NEXT: and v19.8b, v16.8b, v4.8b
427-
; CHECK-NEXT: and v20.8b, v5.8b, v4.8b
428-
; CHECK-NEXT: ushll v3.2d, v17.2s, #0
429-
; CHECK-NEXT: ushll v17.2d, v18.2s, #0
430-
; CHECK-NEXT: ext v1.16b, v1.16b, v1.16b, #8
431-
; CHECK-NEXT: ushr v16.2s, v16.2s, #16
432-
; CHECK-NEXT: ushr v5.2s, v5.2s, #16
433-
; CHECK-NEXT: ushll v6.2d, v6.2s, #0
434-
; CHECK-NEXT: ushll v7.2d, v7.2s, #0
435-
; CHECK-NEXT: ushll v18.2d, v19.2s, #0
436-
; CHECK-NEXT: ushll v19.2d, v20.2s, #0
437-
; CHECK-NEXT: ext v20.16b, v0.16b, v0.16b, #8
438-
; CHECK-NEXT: ushr v0.2s, v0.2s, #16
439-
; CHECK-NEXT: ushll v16.2d, v16.2s, #0
440-
; CHECK-NEXT: ushll v21.2d, v21.2s, #0
441-
; CHECK-NEXT: ushll v5.2d, v5.2s, #0
442-
; CHECK-NEXT: ushll v22.2d, v22.2s, #0
443-
; CHECK-NEXT: ushll v2.2d, v2.2s, #0
444-
; CHECK-NEXT: ucvtf v3.2d, v3.2d
445-
; CHECK-NEXT: ucvtf v17.2d, v17.2d
446-
; CHECK-NEXT: ucvtf v6.2d, v6.2d
447-
; CHECK-NEXT: and v23.8b, v20.8b, v4.8b
448-
; CHECK-NEXT: and v4.8b, v1.8b, v4.8b
449-
; CHECK-NEXT: ushr v20.2s, v20.2s, #16
450-
; CHECK-NEXT: ushr v1.2s, v1.2s, #16
451-
; CHECK-NEXT: ushll v0.2d, v0.2s, #0
452-
; CHECK-NEXT: ucvtf v7.2d, v7.2d
404+
; CHECK-NEXT: movi v4.2d, #0x0000000000ffff
405+
; CHECK-NEXT: ushr v5.2d, v0.2d, #16
406+
; CHECK-NEXT: ushr v6.2d, v1.2d, #16
407+
; CHECK-NEXT: ushr v7.2d, v2.2d, #16
408+
; CHECK-NEXT: ushr v17.2d, v3.2d, #16
409+
; CHECK-NEXT: ushr v20.2d, v0.2d, #32
410+
; CHECK-NEXT: ushr v22.2d, v1.2d, #32
411+
; CHECK-NEXT: ushr v23.2d, v2.2d, #32
412+
; CHECK-NEXT: ushr v24.2d, v3.2d, #32
413+
; CHECK-NEXT: and v16.16b, v0.16b, v4.16b
414+
; CHECK-NEXT: and v18.16b, v1.16b, v4.16b
415+
; CHECK-NEXT: and v19.16b, v2.16b, v4.16b
416+
; CHECK-NEXT: and v21.16b, v3.16b, v4.16b
417+
; CHECK-NEXT: and v5.16b, v5.16b, v4.16b
418+
; CHECK-NEXT: and v6.16b, v6.16b, v4.16b
419+
; CHECK-NEXT: and v7.16b, v7.16b, v4.16b
420+
; CHECK-NEXT: and v17.16b, v17.16b, v4.16b
421+
; CHECK-NEXT: and v20.16b, v20.16b, v4.16b
422+
; CHECK-NEXT: and v22.16b, v22.16b, v4.16b
423+
; CHECK-NEXT: and v23.16b, v23.16b, v4.16b
424+
; CHECK-NEXT: and v4.16b, v24.16b, v4.16b
425+
; CHECK-NEXT: ushr v0.2d, v0.2d, #48
426+
; CHECK-NEXT: ushr v1.2d, v1.2d, #48
427+
; CHECK-NEXT: ushr v2.2d, v2.2d, #48
428+
; CHECK-NEXT: ushr v3.2d, v3.2d, #48
429+
; CHECK-NEXT: ucvtf v16.2d, v16.2d
453430
; CHECK-NEXT: ucvtf v18.2d, v18.2d
454431
; CHECK-NEXT: ucvtf v19.2d, v19.2d
455-
; CHECK-NEXT: ucvtf v16.2d, v16.2d
456-
; CHECK-NEXT: ushll v23.2d, v23.2s, #0
457-
; CHECK-NEXT: ushll v4.2d, v4.2s, #0
458-
; CHECK-NEXT: ushll v20.2d, v20.2s, #0
459-
; CHECK-NEXT: ushll v1.2d, v1.2s, #0
460-
; CHECK-NEXT: ucvtf v5.2d, v5.2d
461432
; CHECK-NEXT: ucvtf v21.2d, v21.2d
433+
; CHECK-NEXT: ucvtf v5.2d, v5.2d
434+
; CHECK-NEXT: ucvtf v6.2d, v6.2d
435+
; CHECK-NEXT: ucvtf v7.2d, v7.2d
436+
; CHECK-NEXT: ucvtf v17.2d, v17.2d
437+
; CHECK-NEXT: ucvtf v20.2d, v20.2d
462438
; CHECK-NEXT: ucvtf v22.2d, v22.2d
463-
; CHECK-NEXT: ucvtf v0.2d, v0.2d
464-
; CHECK-NEXT: ucvtf v2.2d, v2.2d
465439
; CHECK-NEXT: ucvtf v23.2d, v23.2d
466440
; CHECK-NEXT: ucvtf v4.2d, v4.2d
467-
; CHECK-NEXT: ucvtf v20.2d, v20.2d
441+
; CHECK-NEXT: ucvtf v0.2d, v0.2d
468442
; CHECK-NEXT: ucvtf v1.2d, v1.2d
469-
; CHECK-NEXT: fadd v16.2d, v18.2d, v16.2d
470-
; CHECK-NEXT: fadd v7.2d, v17.2d, v7.2d
471-
; CHECK-NEXT: fadd v5.2d, v19.2d, v5.2d
472-
; CHECK-NEXT: fadd v3.2d, v3.2d, v6.2d
473-
; CHECK-NEXT: fadd v0.2d, v21.2d, v0.2d
474-
; CHECK-NEXT: fadd v2.2d, v22.2d, v2.2d
475-
; CHECK-NEXT: fadd v4.2d, v4.2d, v1.2d
476-
; CHECK-NEXT: fadd v1.2d, v23.2d, v20.2d
477-
; CHECK-NEXT: fadd v0.2d, v3.2d, v0.2d
443+
; CHECK-NEXT: ucvtf v2.2d, v2.2d
444+
; CHECK-NEXT: ucvtf v3.2d, v3.2d
445+
; CHECK-NEXT: fadd v5.2d, v16.2d, v5.2d
446+
; CHECK-NEXT: fadd v17.2d, v21.2d, v17.2d
447+
; CHECK-NEXT: fadd v7.2d, v19.2d, v7.2d
448+
; CHECK-NEXT: fadd v6.2d, v18.2d, v6.2d
449+
; CHECK-NEXT: fadd v0.2d, v20.2d, v0.2d
450+
; CHECK-NEXT: fadd v1.2d, v22.2d, v1.2d
451+
; CHECK-NEXT: fadd v3.2d, v4.2d, v3.2d
452+
; CHECK-NEXT: fadd v2.2d, v23.2d, v2.2d
453+
; CHECK-NEXT: fadd v0.2d, v5.2d, v0.2d
454+
; CHECK-NEXT: fadd v1.2d, v6.2d, v1.2d
478455
; CHECK-NEXT: fadd v2.2d, v7.2d, v2.2d
479-
; CHECK-NEXT: fadd v1.2d, v16.2d, v1.2d
480-
; CHECK-NEXT: fadd v3.2d, v5.2d, v4.2d
456+
; CHECK-NEXT: fadd v3.2d, v17.2d, v3.2d
481457
; CHECK-NEXT: ret
482458
%s1 = shufflevector <32 x i16> %l, <32 x i16> undef, <8 x i32> <i32 0, i32 4, i32 8, i32 12, i32 16, i32 20, i32 24, i32 28>
483459
%z1 = uitofp <8 x i16> %s1 to <8 x double>

0 commit comments

Comments
 (0)