[llvm] 05b16af - [DAGCombiner] Add combine for vector interleave of splats (#151110)
via llvm-commits
llvm-commits at lists.llvm.org
Fri Aug 1 01:58:08 PDT 2025
Author: David Sherwood
Date: 2025-08-01T09:58:05+01:00
New Revision: 05b16aff0f4d5132799d7d4470ca13db3afe13ae
URL: https://github.com/llvm/llvm-project/commit/05b16aff0f4d5132799d7d4470ca13db3afe13ae
DIFF: https://github.com/llvm/llvm-project/commit/05b16aff0f4d5132799d7d4470ca13db3afe13ae.diff
LOG: [DAGCombiner] Add combine for vector interleave of splats (#151110)
This patch adds two DAG combines:
1. vector_interleave(splat, splat, ...) -> {splat,splat,...}
2. concat_vectors(splat, splat, ...) -> wide_splat
where all the input splats are identical. Both of these
together enable us to fold
concat_vectors(vector_interleave(splat, splat, ...))
into a wide splat. Post-legalisation we must only do the
concat_vector combine if the wider type and splat operation
is legal.
For fixed-width vectors the DAG combine only occurs for
interleave factors of 3 or more, however it's not currently
safe to test this for AArch64 since there isn't any lowering
support for fixed-width interleaves. I've only added
fixed-width tests for RISCV.
Added:
Modified:
llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
llvm/test/CodeGen/AArch64/complex-deinterleaving-reductions-predicated-scalable.ll
llvm/test/CodeGen/AArch64/complex-deinterleaving-reductions-scalable.ll
llvm/test/CodeGen/AArch64/fixed-vector-interleave.ll
llvm/test/CodeGen/AArch64/sve-vector-interleave.ll
llvm/test/CodeGen/RISCV/rvv/vector-interleave-fixed.ll
llvm/test/CodeGen/RISCV/rvv/vector-interleave.ll
Removed:
################################################################################
diff --git a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
index a43020ee62281..68927518ffa28 100644
--- a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
@@ -331,6 +331,11 @@ namespace {
return CombineTo(N, To, 2, AddTo);
}
+ SDValue CombineTo(SDNode *N, SmallVectorImpl<SDValue> *To,
+ bool AddTo = true) {
+ return CombineTo(N, To->data(), To->size(), AddTo);
+ }
+
void CommitTargetLoweringOpt(const TargetLowering::TargetLoweringOpt &TLO);
private:
@@ -541,6 +546,7 @@ namespace {
SDValue visitEXTRACT_VECTOR_ELT(SDNode *N);
SDValue visitBUILD_VECTOR(SDNode *N);
SDValue visitCONCAT_VECTORS(SDNode *N);
+ SDValue visitVECTOR_INTERLEAVE(SDNode *N);
SDValue visitEXTRACT_SUBVECTOR(SDNode *N);
SDValue visitVECTOR_SHUFFLE(SDNode *N);
SDValue visitSCALAR_TO_VECTOR(SDNode *N);
@@ -2021,6 +2027,7 @@ SDValue DAGCombiner::visit(SDNode *N) {
case ISD::EXTRACT_VECTOR_ELT: return visitEXTRACT_VECTOR_ELT(N);
case ISD::BUILD_VECTOR: return visitBUILD_VECTOR(N);
case ISD::CONCAT_VECTORS: return visitCONCAT_VECTORS(N);
+ case ISD::VECTOR_INTERLEAVE: return visitVECTOR_INTERLEAVE(N);
case ISD::EXTRACT_SUBVECTOR: return visitEXTRACT_SUBVECTOR(N);
case ISD::VECTOR_SHUFFLE: return visitVECTOR_SHUFFLE(N);
case ISD::SCALAR_TO_VECTOR: return visitSCALAR_TO_VECTOR(N);
@@ -25282,6 +25289,28 @@ static SDValue combineConcatVectorOfShuffleAndItsOperands(
return DAG.getVectorShuffle(VT, dl, ShufOps[0], ShufOps[1], Mask);
}
+static SDValue combineConcatVectorOfSplats(SDNode *N, SelectionDAG &DAG,
+ const TargetLowering &TLI,
+ bool LegalTypes,
+ bool LegalOperations) {
+ EVT VT = N->getValueType(0);
+
+ // Post-legalization we can only create wider SPLAT_VECTOR operations if both
+ // the type and operation is legal. The Hexagon target has custom
+ // legalization for SPLAT_VECTOR that splits the operation into two parts and
+ // concatenates them. Therefore, custom lowering must also be rejected in
+ // order to avoid an infinite loop.
+ if ((LegalTypes && !TLI.isTypeLegal(VT)) ||
+ (LegalOperations && !TLI.isOperationLegal(ISD::SPLAT_VECTOR, VT)))
+ return SDValue();
+
+ SDValue Op0 = N->getOperand(0);
+ if (!llvm::all_equal(N->op_values()) || Op0.getOpcode() != ISD::SPLAT_VECTOR)
+ return SDValue();
+
+ return DAG.getNode(ISD::SPLAT_VECTOR, SDLoc(N), VT, Op0.getOperand(0));
+}
+
SDValue DAGCombiner::visitCONCAT_VECTORS(SDNode *N) {
// If we only have one input vector, we don't need to do any concatenation.
if (N->getNumOperands() == 1)
@@ -25405,6 +25434,10 @@ SDValue DAGCombiner::visitCONCAT_VECTORS(SDNode *N) {
return DAG.getBuildVector(VT, SDLoc(N), Opnds);
}
+ if (SDValue V =
+ combineConcatVectorOfSplats(N, DAG, TLI, LegalTypes, LegalOperations))
+ return V;
+
// Fold CONCAT_VECTORS of only bitcast scalars (or undef) to BUILD_VECTOR.
// FIXME: Add support for concat_vectors(bitcast(vec0),bitcast(vec1),...).
if (SDValue V = combineConcatVectorOfScalars(N, DAG))
@@ -25473,6 +25506,21 @@ SDValue DAGCombiner::visitCONCAT_VECTORS(SDNode *N) {
return SDValue();
}
+SDValue DAGCombiner::visitVECTOR_INTERLEAVE(SDNode *N) {
+ // Check to see if all operands are identical.
+ if (!llvm::all_equal(N->op_values()))
+ return SDValue();
+
+ // Check to see if the identical operand is a splat.
+ if (!DAG.isSplatValue(N->getOperand(0)))
+ return SDValue();
+
+ // interleave splat(X), splat(X).... --> splat(X), splat(X)....
+ SmallVector<SDValue, 4> Ops;
+ Ops.append(N->op_values().begin(), N->op_values().end());
+ return CombineTo(N, &Ops);
+}
+
// Helper that peeks through INSERT_SUBVECTOR/CONCAT_VECTORS to find
// if the subvector can be sourced for free.
static SDValue getSubVectorSrc(SDValue V, unsigned Index, EVT SubVT) {
diff --git a/llvm/test/CodeGen/AArch64/complex-deinterleaving-reductions-predicated-scalable.ll b/llvm/test/CodeGen/AArch64/complex-deinterleaving-reductions-predicated-scalable.ll
index 880bd2904154c..d67aa08125f74 100644
--- a/llvm/test/CodeGen/AArch64/complex-deinterleaving-reductions-predicated-scalable.ll
+++ b/llvm/test/CodeGen/AArch64/complex-deinterleaving-reductions-predicated-scalable.ll
@@ -14,20 +14,19 @@ target triple = "aarch64"
define %"class.std::complex" @complex_mul_v2f64(ptr %a, ptr %b) {
; CHECK-LABEL: complex_mul_v2f64:
; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: movi v0.2d, #0000000000000000
; CHECK-NEXT: movi v1.2d, #0000000000000000
; CHECK-NEXT: mov w8, #100 // =0x64
-; CHECK-NEXT: cntd x9
; CHECK-NEXT: whilelo p1.d, xzr, x8
+; CHECK-NEXT: cntd x9
; CHECK-NEXT: rdvl x10, #2
-; CHECK-NEXT: mov x11, x9
; CHECK-NEXT: ptrue p0.d
-; CHECK-NEXT: zip2 z0.d, z1.d, z1.d
-; CHECK-NEXT: zip1 z1.d, z1.d, z1.d
+; CHECK-NEXT: mov x11, x9
; CHECK-NEXT: .LBB0_1: // %vector.body
; CHECK-NEXT: // =>This Inner Loop Header: Depth=1
; CHECK-NEXT: zip2 p2.d, p1.d, p1.d
-; CHECK-NEXT: mov z6.d, z1.d
-; CHECK-NEXT: mov z7.d, z0.d
+; CHECK-NEXT: mov z6.d, z0.d
+; CHECK-NEXT: mov z7.d, z1.d
; CHECK-NEXT: zip1 p1.d, p1.d, p1.d
; CHECK-NEXT: ld1d { z2.d }, p2/z, [x0, #1, mul vl]
; CHECK-NEXT: ld1d { z4.d }, p2/z, [x1, #1, mul vl]
@@ -39,14 +38,14 @@ define %"class.std::complex" @complex_mul_v2f64(ptr %a, ptr %b) {
; CHECK-NEXT: fcmla z6.d, p0/m, z5.d, z3.d, #0
; CHECK-NEXT: fcmla z7.d, p0/m, z4.d, z2.d, #90
; CHECK-NEXT: fcmla z6.d, p0/m, z5.d, z3.d, #90
-; CHECK-NEXT: mov z0.d, p2/m, z7.d
-; CHECK-NEXT: mov z1.d, p1/m, z6.d
+; CHECK-NEXT: mov z1.d, p2/m, z7.d
+; CHECK-NEXT: mov z0.d, p1/m, z6.d
; CHECK-NEXT: whilelo p1.d, x11, x8
; CHECK-NEXT: add x11, x11, x9
; CHECK-NEXT: b.mi .LBB0_1
; CHECK-NEXT: // %bb.2: // %exit.block
-; CHECK-NEXT: uzp1 z2.d, z1.d, z0.d
-; CHECK-NEXT: uzp2 z1.d, z1.d, z0.d
+; CHECK-NEXT: uzp1 z2.d, z0.d, z1.d
+; CHECK-NEXT: uzp2 z1.d, z0.d, z1.d
; CHECK-NEXT: faddv d0, p0, z2.d
; CHECK-NEXT: faddv d1, p0, z1.d
; CHECK-NEXT: // kill: def $d0 killed $d0 killed $z0
@@ -111,21 +110,20 @@ exit.block: ; preds = %vector.body
define %"class.std::complex" @complex_mul_predicated_v2f64(ptr %a, ptr %b, ptr %cond) {
; CHECK-LABEL: complex_mul_predicated_v2f64:
; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: movi v0.2d, #0000000000000000
; CHECK-NEXT: movi v1.2d, #0000000000000000
; CHECK-NEXT: cntd x9
-; CHECK-NEXT: mov w11, #100 // =0x64
; CHECK-NEXT: neg x10, x9
+; CHECK-NEXT: mov w11, #100 // =0x64
; CHECK-NEXT: ptrue p0.d
; CHECK-NEXT: mov x8, xzr
; CHECK-NEXT: and x10, x10, x11
; CHECK-NEXT: rdvl x11, #2
-; CHECK-NEXT: zip2 z0.d, z1.d, z1.d
-; CHECK-NEXT: zip1 z1.d, z1.d, z1.d
; CHECK-NEXT: .LBB1_1: // %vector.body
; CHECK-NEXT: // =>This Inner Loop Header: Depth=1
; CHECK-NEXT: ld1w { z2.d }, p0/z, [x2, x8, lsl #2]
-; CHECK-NEXT: mov z6.d, z1.d
-; CHECK-NEXT: mov z7.d, z0.d
+; CHECK-NEXT: mov z6.d, z0.d
+; CHECK-NEXT: mov z7.d, z1.d
; CHECK-NEXT: add x8, x8, x9
; CHECK-NEXT: cmpne p1.d, p0/z, z2.d, #0
; CHECK-NEXT: cmp x10, x8
@@ -141,12 +139,12 @@ define %"class.std::complex" @complex_mul_predicated_v2f64(ptr %a, ptr %b, ptr %
; CHECK-NEXT: fcmla z6.d, p0/m, z5.d, z3.d, #0
; CHECK-NEXT: fcmla z7.d, p0/m, z4.d, z2.d, #90
; CHECK-NEXT: fcmla z6.d, p0/m, z5.d, z3.d, #90
-; CHECK-NEXT: mov z0.d, p2/m, z7.d
-; CHECK-NEXT: mov z1.d, p1/m, z6.d
+; CHECK-NEXT: mov z1.d, p2/m, z7.d
+; CHECK-NEXT: mov z0.d, p1/m, z6.d
; CHECK-NEXT: b.ne .LBB1_1
; CHECK-NEXT: // %bb.2: // %exit.block
-; CHECK-NEXT: uzp1 z2.d, z1.d, z0.d
-; CHECK-NEXT: uzp2 z1.d, z1.d, z0.d
+; CHECK-NEXT: uzp1 z2.d, z0.d, z1.d
+; CHECK-NEXT: uzp2 z1.d, z0.d, z1.d
; CHECK-NEXT: faddv d0, p0, z2.d
; CHECK-NEXT: faddv d1, p0, z1.d
; CHECK-NEXT: // kill: def $d0 killed $d0 killed $z0
@@ -213,21 +211,20 @@ exit.block: ; preds = %vector.body
define %"class.std::complex" @complex_mul_predicated_x2_v2f64(ptr %a, ptr %b, ptr %cond) {
; CHECK-LABEL: complex_mul_predicated_x2_v2f64:
; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: movi v0.2d, #0000000000000000
; CHECK-NEXT: movi v1.2d, #0000000000000000
; CHECK-NEXT: mov w8, #100 // =0x64
-; CHECK-NEXT: cntd x9
; CHECK-NEXT: whilelo p1.d, xzr, x8
+; CHECK-NEXT: cntd x9
; CHECK-NEXT: rdvl x10, #2
-; CHECK-NEXT: cnth x11
; CHECK-NEXT: ptrue p0.d
+; CHECK-NEXT: cnth x11
; CHECK-NEXT: mov x12, x9
-; CHECK-NEXT: zip2 z0.d, z1.d, z1.d
-; CHECK-NEXT: zip1 z1.d, z1.d, z1.d
; CHECK-NEXT: .LBB2_1: // %vector.body
; CHECK-NEXT: // =>This Inner Loop Header: Depth=1
; CHECK-NEXT: ld1w { z2.d }, p1/z, [x2]
-; CHECK-NEXT: mov z6.d, z1.d
-; CHECK-NEXT: mov z7.d, z0.d
+; CHECK-NEXT: mov z6.d, z0.d
+; CHECK-NEXT: mov z7.d, z1.d
; CHECK-NEXT: add x2, x2, x11
; CHECK-NEXT: and z2.d, z2.d, #0xffffffff
; CHECK-NEXT: cmpne p1.d, p1/z, z2.d, #0
@@ -243,14 +240,14 @@ define %"class.std::complex" @complex_mul_predicated_x2_v2f64(ptr %a, ptr %b, pt
; CHECK-NEXT: fcmla z6.d, p0/m, z5.d, z3.d, #0
; CHECK-NEXT: fcmla z7.d, p0/m, z4.d, z2.d, #90
; CHECK-NEXT: fcmla z6.d, p0/m, z5.d, z3.d, #90
-; CHECK-NEXT: mov z0.d, p2/m, z7.d
-; CHECK-NEXT: mov z1.d, p1/m, z6.d
+; CHECK-NEXT: mov z1.d, p2/m, z7.d
+; CHECK-NEXT: mov z0.d, p1/m, z6.d
; CHECK-NEXT: whilelo p1.d, x12, x8
; CHECK-NEXT: add x12, x12, x9
; CHECK-NEXT: b.mi .LBB2_1
; CHECK-NEXT: // %bb.2: // %exit.block
-; CHECK-NEXT: uzp1 z2.d, z1.d, z0.d
-; CHECK-NEXT: uzp2 z1.d, z1.d, z0.d
+; CHECK-NEXT: uzp1 z2.d, z0.d, z1.d
+; CHECK-NEXT: uzp2 z1.d, z0.d, z1.d
; CHECK-NEXT: faddv d0, p0, z2.d
; CHECK-NEXT: faddv d1, p0, z1.d
; CHECK-NEXT: // kill: def $d0 killed $d0 killed $z0
diff --git a/llvm/test/CodeGen/AArch64/complex-deinterleaving-reductions-scalable.ll b/llvm/test/CodeGen/AArch64/complex-deinterleaving-reductions-scalable.ll
index 29be231920305..0646ca4948e1d 100644
--- a/llvm/test/CodeGen/AArch64/complex-deinterleaving-reductions-scalable.ll
+++ b/llvm/test/CodeGen/AArch64/complex-deinterleaving-reductions-scalable.ll
@@ -14,15 +14,14 @@ target triple = "aarch64"
define %"class.std::complex" @complex_mul_v2f64(ptr %a, ptr %b) {
; CHECK-LABEL: complex_mul_v2f64:
; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: movi v0.2d, #0000000000000000
; CHECK-NEXT: movi v1.2d, #0000000000000000
; CHECK-NEXT: cntd x8
-; CHECK-NEXT: mov w10, #100 // =0x64
; CHECK-NEXT: neg x9, x8
+; CHECK-NEXT: mov w10, #100 // =0x64
; CHECK-NEXT: ptrue p0.d
; CHECK-NEXT: and x9, x9, x10
; CHECK-NEXT: rdvl x10, #2
-; CHECK-NEXT: zip2 z0.d, z1.d, z1.d
-; CHECK-NEXT: zip1 z1.d, z1.d, z1.d
; CHECK-NEXT: .LBB0_1: // %vector.body
; CHECK-NEXT: // =>This Inner Loop Header: Depth=1
; CHECK-NEXT: ldr z2, [x0, #1, mul vl]
@@ -32,14 +31,14 @@ define %"class.std::complex" @complex_mul_v2f64(ptr %a, ptr %b) {
; CHECK-NEXT: ldr z5, [x1]
; CHECK-NEXT: add x1, x1, x10
; CHECK-NEXT: add x0, x0, x10
-; CHECK-NEXT: fcmla z1.d, p0/m, z5.d, z3.d, #0
-; CHECK-NEXT: fcmla z0.d, p0/m, z4.d, z2.d, #0
-; CHECK-NEXT: fcmla z1.d, p0/m, z5.d, z3.d, #90
-; CHECK-NEXT: fcmla z0.d, p0/m, z4.d, z2.d, #90
+; CHECK-NEXT: fcmla z0.d, p0/m, z5.d, z3.d, #0
+; CHECK-NEXT: fcmla z1.d, p0/m, z4.d, z2.d, #0
+; CHECK-NEXT: fcmla z0.d, p0/m, z5.d, z3.d, #90
+; CHECK-NEXT: fcmla z1.d, p0/m, z4.d, z2.d, #90
; CHECK-NEXT: b.ne .LBB0_1
; CHECK-NEXT: // %bb.2: // %exit.block
-; CHECK-NEXT: uzp1 z2.d, z1.d, z0.d
-; CHECK-NEXT: uzp2 z1.d, z1.d, z0.d
+; CHECK-NEXT: uzp1 z2.d, z0.d, z1.d
+; CHECK-NEXT: uzp2 z1.d, z0.d, z1.d
; CHECK-NEXT: faddv d0, p0, z2.d
; CHECK-NEXT: faddv d1, p0, z1.d
; CHECK-NEXT: // kill: def $d0 killed $d0 killed $z0
@@ -183,17 +182,16 @@ exit.block: ; preds = %vector.body
define %"class.std::complex" @complex_mul_v2f64_unrolled(ptr %a, ptr %b) {
; CHECK-LABEL: complex_mul_v2f64_unrolled:
; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: movi v0.2d, #0000000000000000
; CHECK-NEXT: movi v1.2d, #0000000000000000
; CHECK-NEXT: cntw x8
-; CHECK-NEXT: mov w10, #1000 // =0x3e8
+; CHECK-NEXT: movi v2.2d, #0000000000000000
+; CHECK-NEXT: movi v3.2d, #0000000000000000
; CHECK-NEXT: neg x9, x8
+; CHECK-NEXT: mov w10, #1000 // =0x3e8
; CHECK-NEXT: ptrue p0.d
; CHECK-NEXT: and x9, x9, x10
; CHECK-NEXT: rdvl x10, #4
-; CHECK-NEXT: zip2 z0.d, z1.d, z1.d
-; CHECK-NEXT: zip1 z1.d, z1.d, z1.d
-; CHECK-NEXT: mov z2.d, z1.d
-; CHECK-NEXT: mov z3.d, z0.d
; CHECK-NEXT: .LBB2_1: // %vector.body
; CHECK-NEXT: // =>This Inner Loop Header: Depth=1
; CHECK-NEXT: ldr z4, [x0, #1, mul vl]
@@ -207,20 +205,20 @@ define %"class.std::complex" @complex_mul_v2f64_unrolled(ptr %a, ptr %b) {
; CHECK-NEXT: ldr z18, [x1, #3, mul vl]
; CHECK-NEXT: ldr z19, [x1, #2, mul vl]
; CHECK-NEXT: add x1, x1, x10
-; CHECK-NEXT: fcmla z1.d, p0/m, z16.d, z5.d, #0
-; CHECK-NEXT: fcmla z0.d, p0/m, z7.d, z4.d, #0
+; CHECK-NEXT: fcmla z0.d, p0/m, z16.d, z5.d, #0
+; CHECK-NEXT: fcmla z1.d, p0/m, z7.d, z4.d, #0
; CHECK-NEXT: fcmla z3.d, p0/m, z18.d, z6.d, #0
; CHECK-NEXT: fcmla z2.d, p0/m, z19.d, z17.d, #0
-; CHECK-NEXT: fcmla z1.d, p0/m, z16.d, z5.d, #90
-; CHECK-NEXT: fcmla z0.d, p0/m, z7.d, z4.d, #90
+; CHECK-NEXT: fcmla z0.d, p0/m, z16.d, z5.d, #90
+; CHECK-NEXT: fcmla z1.d, p0/m, z7.d, z4.d, #90
; CHECK-NEXT: fcmla z3.d, p0/m, z18.d, z6.d, #90
; CHECK-NEXT: fcmla z2.d, p0/m, z19.d, z17.d, #90
; CHECK-NEXT: b.ne .LBB2_1
; CHECK-NEXT: // %bb.2: // %exit.block
; CHECK-NEXT: uzp1 z4.d, z2.d, z3.d
-; CHECK-NEXT: uzp1 z5.d, z1.d, z0.d
+; CHECK-NEXT: uzp1 z5.d, z0.d, z1.d
; CHECK-NEXT: uzp2 z2.d, z2.d, z3.d
-; CHECK-NEXT: uzp2 z0.d, z1.d, z0.d
+; CHECK-NEXT: uzp2 z0.d, z0.d, z1.d
; CHECK-NEXT: fadd z1.d, z4.d, z5.d
; CHECK-NEXT: fadd z2.d, z2.d, z0.d
; CHECK-NEXT: faddv d0, p0, z1.d
@@ -310,15 +308,15 @@ define dso_local %"class.std::complex" @reduction_mix(ptr %a, ptr %b, ptr noalia
; CHECK-LABEL: reduction_mix:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: movi v2.2d, #0000000000000000
+; CHECK-NEXT: movi v0.2d, #0000000000000000
; CHECK-NEXT: cntd x9
-; CHECK-NEXT: mov w11, #100 // =0x64
+; CHECK-NEXT: movi v1.2d, #0000000000000000
; CHECK-NEXT: neg x10, x9
+; CHECK-NEXT: mov w11, #100 // =0x64
; CHECK-NEXT: ptrue p0.d
; CHECK-NEXT: mov x8, xzr
; CHECK-NEXT: and x10, x10, x11
; CHECK-NEXT: rdvl x11, #2
-; CHECK-NEXT: zip2 z0.d, z2.d, z2.d
-; CHECK-NEXT: zip1 z1.d, z2.d, z2.d
; CHECK-NEXT: .LBB3_1: // %vector.body
; CHECK-NEXT: // =>This Inner Loop Header: Depth=1
; CHECK-NEXT: ldr z3, [x0]
@@ -327,13 +325,13 @@ define dso_local %"class.std::complex" @reduction_mix(ptr %a, ptr %b, ptr noalia
; CHECK-NEXT: ld1w { z5.d }, p0/z, [x3, x8, lsl #2]
; CHECK-NEXT: add x8, x8, x9
; CHECK-NEXT: cmp x10, x8
-; CHECK-NEXT: fadd z0.d, z4.d, z0.d
-; CHECK-NEXT: fadd z1.d, z3.d, z1.d
+; CHECK-NEXT: fadd z1.d, z4.d, z1.d
+; CHECK-NEXT: fadd z0.d, z3.d, z0.d
; CHECK-NEXT: add z2.d, z5.d, z2.d
; CHECK-NEXT: b.ne .LBB3_1
; CHECK-NEXT: // %bb.2: // %middle.block
-; CHECK-NEXT: uzp2 z3.d, z1.d, z0.d
-; CHECK-NEXT: uzp1 z1.d, z1.d, z0.d
+; CHECK-NEXT: uzp2 z3.d, z0.d, z1.d
+; CHECK-NEXT: uzp1 z1.d, z0.d, z1.d
; CHECK-NEXT: uaddv d2, p0, z2.d
; CHECK-NEXT: faddv d0, p0, z3.d
; CHECK-NEXT: faddv d1, p0, z1.d
diff --git a/llvm/test/CodeGen/AArch64/fixed-vector-interleave.ll b/llvm/test/CodeGen/AArch64/fixed-vector-interleave.ll
index a9618fdc2dec3..05ecc9e7b49d4 100644
--- a/llvm/test/CodeGen/AArch64/fixed-vector-interleave.ll
+++ b/llvm/test/CodeGen/AArch64/fixed-vector-interleave.ll
@@ -131,18 +131,83 @@ define <4 x i64> @interleave2_v4i64(<2 x i64> %vec0, <2 x i64> %vec1) {
ret <4 x i64> %retval
}
+define <4 x i16> @interleave2_same_const_splat_v4i16() {
+; CHECK-SD-LABEL: interleave2_same_const_splat_v4i16:
+; CHECK-SD: // %bb.0:
+; CHECK-SD-NEXT: movi v0.4h, #3
+; CHECK-SD-NEXT: ret
+;
+; CHECK-GI-LABEL: interleave2_same_const_splat_v4i16:
+; CHECK-GI: // %bb.0:
+; CHECK-GI-NEXT: mov w8, #3 // =0x3
+; CHECK-GI-NEXT: fmov s0, w8
+; CHECK-GI-NEXT: mov v0.h[1], w8
+; CHECK-GI-NEXT: zip1 v0.4h, v0.4h, v0.4h
+; CHECK-GI-NEXT: ret
+ %retval = call <4 x i16> @llvm.vector.interleave2.v4i16(<2 x i16> splat(i16 3), <2 x i16> splat(i16 3))
+ ret <4 x i16> %retval
+}
+
+define <4 x i16> @interleave2_
diff _const_splat_v4i16() {
+; CHECK-SD-LABEL: interleave2_
diff _const_splat_v4i16:
+; CHECK-SD: // %bb.0:
+; CHECK-SD-NEXT: adrp x8, .LCPI11_0
+; CHECK-SD-NEXT: ldr d0, [x8, :lo12:.LCPI11_0]
+; CHECK-SD-NEXT: ret
+;
+; CHECK-GI-LABEL: interleave2_
diff _const_splat_v4i16:
+; CHECK-GI: // %bb.0:
+; CHECK-GI-NEXT: mov w8, #3 // =0x3
+; CHECK-GI-NEXT: mov w9, #4 // =0x4
+; CHECK-GI-NEXT: fmov s0, w8
+; CHECK-GI-NEXT: fmov s1, w9
+; CHECK-GI-NEXT: mov v0.h[1], w8
+; CHECK-GI-NEXT: mov v1.h[1], w9
+; CHECK-GI-NEXT: zip1 v0.4h, v0.4h, v1.4h
+; CHECK-GI-NEXT: ret
+ %retval = call <4 x i16> @llvm.vector.interleave2.v4i16(<2 x i16> splat(i16 3), <2 x i16> splat(i16 4))
+ ret <4 x i16> %retval
+}
-; Float declarations
-declare <4 x half> @llvm.vector.interleave2.v4f16(<2 x half>, <2 x half>)
-declare <8 x half> @llvm.vector.interleave2.v8f16(<4 x half>, <4 x half>)
-declare <16 x half> @llvm.vector.interleave2.v16f16(<8 x half>, <8 x half>)
-declare <4 x float> @llvm.vector.interleave2.v4f32(<2 x float>, <2 x float>)
-declare <8 x float> @llvm.vector.interleave2.v8f32(<4 x float>, <4 x float>)
-declare <4 x double> @llvm.vector.interleave2.v4f64(<2 x double>, <2 x double>)
-
-; Integer declarations
-declare <32 x i8> @llvm.vector.interleave2.v32i8(<16 x i8>, <16 x i8>)
-declare <16 x i16> @llvm.vector.interleave2.v16i16(<8 x i16>, <8 x i16>)
-declare <8 x i32> @llvm.vector.interleave2.v8i32(<4 x i32>, <4 x i32>)
-declare <4 x i64> @llvm.vector.interleave2.v4i64(<2 x i64>, <2 x i64>)
+define <4 x i16> @interleave2_same_nonconst_splat_v4i16(i16 %a) {
+; CHECK-SD-LABEL: interleave2_same_nonconst_splat_v4i16:
+; CHECK-SD: // %bb.0:
+; CHECK-SD-NEXT: dup v0.4h, w0
+; CHECK-SD-NEXT: ret
+;
+; CHECK-GI-LABEL: interleave2_same_nonconst_splat_v4i16:
+; CHECK-GI: // %bb.0:
+; CHECK-GI-NEXT: dup v0.4h, w0
+; CHECK-GI-NEXT: zip1 v0.4h, v0.4h, v0.4h
+; CHECK-GI-NEXT: ret
+ %ins = insertelement <2 x i16> poison, i16 %a, i32 0
+ %splat = shufflevector <2 x i16> %ins, <2 x i16> poison, <2 x i32> <i32 0, i32 0>
+ %retval = call <4 x i16> @llvm.vector.interleave2.v4i16(<2 x i16> %splat, <2 x i16> %splat)
+ ret <4 x i16> %retval
+}
+
+define <4 x i16> @interleave2_
diff _nonconst_splat_v4i16(i16 %a, i16 %b) {
+; CHECK-SD-LABEL: interleave2_
diff _nonconst_splat_v4i16:
+; CHECK-SD: // %bb.0:
+; CHECK-SD-NEXT: fmov s0, w0
+; CHECK-SD-NEXT: mov v0.h[1], w0
+; CHECK-SD-NEXT: mov v0.h[2], w1
+; CHECK-SD-NEXT: mov v0.h[3], w1
+; CHECK-SD-NEXT: rev32 v1.4h, v0.4h
+; CHECK-SD-NEXT: uzp1 v0.4h, v0.4h, v1.4h
+; CHECK-SD-NEXT: ret
+;
+; CHECK-GI-LABEL: interleave2_
diff _nonconst_splat_v4i16:
+; CHECK-GI: // %bb.0:
+; CHECK-GI-NEXT: dup v0.4h, w0
+; CHECK-GI-NEXT: dup v1.4h, w1
+; CHECK-GI-NEXT: zip1 v0.4h, v0.4h, v1.4h
+; CHECK-GI-NEXT: ret
+ %ins1 = insertelement <2 x i16> poison, i16 %a, i32 0
+ %splat1 = shufflevector <2 x i16> %ins1, <2 x i16> poison, <2 x i32> <i32 0, i32 0>
+ %ins2 = insertelement <2 x i16> poison, i16 %b, i32 0
+ %splat2 = shufflevector <2 x i16> %ins2, <2 x i16> poison, <2 x i32> <i32 0, i32 0>
+ %retval = call <4 x i16> @llvm.vector.interleave2.v4i16(<2 x i16> %splat1, <2 x i16> %splat2)
+ ret <4 x i16> %retval
+}
diff --git a/llvm/test/CodeGen/AArch64/sve-vector-interleave.ll b/llvm/test/CodeGen/AArch64/sve-vector-interleave.ll
index 52cb2d9ebe343..c7fb2db53d2a3 100644
--- a/llvm/test/CodeGen/AArch64/sve-vector-interleave.ll
+++ b/llvm/test/CodeGen/AArch64/sve-vector-interleave.ll
@@ -267,7 +267,7 @@ define <vscale x 32 x i16> @interleave4_nxv8i16(<vscale x 8 x i16> %vec0, <vscal
; SME2-NEXT: // kill: def $z0 killed $z0 killed $z0_z1_z2_z3 def $z0_z1_z2_z3
; SME2-NEXT: zip { z0.h - z3.h }, { z0.h - z3.h }
; SME2-NEXT: ret
- %retval = call <vscale x 32 x i16> @llvm.vector.interleave4.nxv8i16(<vscale x 8 x i16> %vec0, <vscale x 8 x i16> %vec1, <vscale x 8 x i16> %vec2, <vscale x 8 x i16> %vec3)
+ %retval = call <vscale x 32 x i16> @llvm.vector.interleave4.nxv32i16(<vscale x 8 x i16> %vec0, <vscale x 8 x i16> %vec1, <vscale x 8 x i16> %vec2, <vscale x 8 x i16> %vec3)
ret <vscale x 32 x i16> %retval
}
@@ -540,30 +540,81 @@ define <vscale x 4 x i32> @interleave2_nxv2i32(<vscale x 2 x i32> %vec0, <vscale
ret <vscale x 4 x i32> %retval
}
-; Float declarations
-declare <vscale x 4 x half> @llvm.vector.interleave2.nxv4f16(<vscale x 2 x half>, <vscale x 2 x half>)
-declare <vscale x 8 x half> @llvm.vector.interleave2.nxv8f16(<vscale x 4 x half>, <vscale x 4 x half>)
-declare <vscale x 16 x half> @llvm.vector.interleave2.nxv16f16(<vscale x 8 x half>, <vscale x 8 x half>)
-declare <vscale x 4 x float> @llvm.vector.interleave2.nxv4f32(<vscale x 2 x float>, <vscale x 2 x float>)
-declare <vscale x 8 x float> @llvm.vector.interleave2.nxv8f32(<vscale x 4 x float>, <vscale x 4 x float>)
-declare <vscale x 4 x double> @llvm.vector.interleave2.nxv4f64(<vscale x 2 x double>, <vscale x 2 x double>)
+define <vscale x 4 x i16> @interleave2_same_const_splat_nxv4i16() {
+; CHECK-LABEL: interleave2_same_const_splat_nxv4i16:
+; CHECK: // %bb.0:
+; CHECK-NEXT: mov z0.s, #3 // =0x3
+; CHECK-NEXT: ret
+ %retval = call <vscale x 4 x i16> @llvm.vector.interleave2.nxv4i16(<vscale x 2 x i16> splat(i16 3), <vscale x 2 x i16> splat(i16 3))
+ ret <vscale x 4 x i16> %retval
+}
+
+define <vscale x 4 x i16> @interleave2_
diff _const_splat_nxv4i16() {
+; SVE-LABEL: interleave2_
diff _const_splat_nxv4i16:
+; SVE: // %bb.0:
+; SVE-NEXT: mov z0.d, #4 // =0x4
+; SVE-NEXT: mov z1.d, #3 // =0x3
+; SVE-NEXT: zip2 z2.d, z1.d, z0.d
+; SVE-NEXT: zip1 z0.d, z1.d, z0.d
+; SVE-NEXT: uzp1 z0.s, z0.s, z2.s
+; SVE-NEXT: ret
+;
+; SME2-LABEL: interleave2_
diff _const_splat_nxv4i16:
+; SME2: // %bb.0:
+; SME2-NEXT: mov z0.d, #4 // =0x4
+; SME2-NEXT: mov z1.d, #3 // =0x3
+; SME2-NEXT: zip { z0.d, z1.d }, z1.d, z0.d
+; SME2-NEXT: uzp1 z0.s, z0.s, z1.s
+; SME2-NEXT: ret
+ %retval = call <vscale x 4 x i16> @llvm.vector.interleave2.v4i16(<vscale x 2 x i16> splat(i16 3), <vscale x 2 x i16> splat(i16 4))
+ ret <vscale x 4 x i16> %retval
+}
-; Integer declarations
-declare <vscale x 32 x i8> @llvm.vector.interleave2.nxv32i8(<vscale x 16 x i8>, <vscale x 16 x i8>)
-declare <vscale x 16 x i16> @llvm.vector.interleave2.nxv16i16(<vscale x 8 x i16>, <vscale x 8 x i16>)
-declare <vscale x 8 x i32> @llvm.vector.interleave2.nxv8i32(<vscale x 4 x i32>, <vscale x 4 x i32>)
-declare <vscale x 4 x i64> @llvm.vector.interleave2.nxv4i64(<vscale x 2 x i64>, <vscale x 2 x i64>)
+define <vscale x 4 x i16> @interleave2_same_nonconst_splat_nxv4i16(i16 %a) {
+; CHECK-LABEL: interleave2_same_nonconst_splat_nxv4i16:
+; CHECK: // %bb.0:
+; CHECK-NEXT: mov z0.s, w0
+; CHECK-NEXT: ret
+ %ins = insertelement <vscale x 2 x i16> poison, i16 %a, i32 0
+ %splat = shufflevector <vscale x 2 x i16> %ins, <vscale x 2 x i16> poison, <vscale x 2 x i32> zeroinitializer
+ %retval = call <vscale x 4 x i16> @llvm.vector.interleave2.nxv4i16(<vscale x 2 x i16> %splat, <vscale x 2 x i16> %splat)
+ ret <vscale x 4 x i16> %retval
+}
-; Predicated
-declare <vscale x 32 x i1> @llvm.vector.interleave2.nxv32i1(<vscale x 16 x i1>, <vscale x 16 x i1>)
-declare <vscale x 16 x i1> @llvm.vector.interleave2.nxv16i1(<vscale x 8 x i1>, <vscale x 8 x i1>)
-declare <vscale x 8 x i1> @llvm.vector.interleave2.nxv8i1(<vscale x 4 x i1>, <vscale x 4 x i1>)
-declare <vscale x 4 x i1> @llvm.vector.interleave2.nxv4i1(<vscale x 2 x i1>, <vscale x 2 x i1>)
-
-; Illegal type size
-declare <vscale x 16 x i32> @llvm.vector.interleave2.nxv16i32(<vscale x 8 x i32>, <vscale x 8 x i32>)
-declare <vscale x 8 x i64> @llvm.vector.interleave2.nxv8i64(<vscale x 4 x i64>, <vscale x 4 x i64>)
-
-declare <vscale x 16 x i8> @llvm.vector.interleave2.nxv16i8(<vscale x 8 x i8>, <vscale x 8 x i8>)
-declare <vscale x 8 x i16> @llvm.vector.interleave2.nxv8i16(<vscale x 4 x i16>, <vscale x 4 x i16>)
-declare <vscale x 4 x i32> @llvm.vector.interleave2.nxv4i32(<vscale x 2 x i32>, <vscale x 2 x i32>)
+define <vscale x 4 x i16> @interleave2_
diff _nonconst_splat_nxv4i16(i16 %a, i16 %b) {
+; SVE-LABEL: interleave2_
diff _nonconst_splat_nxv4i16:
+; SVE: // %bb.0:
+; SVE-NEXT: // kill: def $w1 killed $w1 def $x1
+; SVE-NEXT: // kill: def $w0 killed $w0 def $x0
+; SVE-NEXT: mov z0.d, x0
+; SVE-NEXT: mov z1.d, x1
+; SVE-NEXT: zip2 z2.d, z0.d, z1.d
+; SVE-NEXT: zip1 z0.d, z0.d, z1.d
+; SVE-NEXT: uzp1 z0.s, z0.s, z2.s
+; SVE-NEXT: ret
+;
+; SME2-LABEL: interleave2_
diff _nonconst_splat_nxv4i16:
+; SME2: // %bb.0:
+; SME2-NEXT: // kill: def $w1 killed $w1 def $x1
+; SME2-NEXT: // kill: def $w0 killed $w0 def $x0
+; SME2-NEXT: mov z0.d, x0
+; SME2-NEXT: mov z1.d, x1
+; SME2-NEXT: zip { z0.d, z1.d }, z0.d, z1.d
+; SME2-NEXT: uzp1 z0.s, z0.s, z1.s
+; SME2-NEXT: ret
+ %ins1 = insertelement <vscale x 2 x i16> poison, i16 %a, i32 0
+ %splat1 = shufflevector <vscale x 2 x i16> %ins1, <vscale x 2 x i16> poison, <vscale x 2 x i32> zeroinitializer
+ %ins2 = insertelement <vscale x 2 x i16> poison, i16 %b, i32 0
+ %splat2 = shufflevector <vscale x 2 x i16> %ins2, <vscale x 2 x i16> poison, <vscale x 2 x i32> zeroinitializer
+ %retval = call <vscale x 4 x i16> @llvm.vector.interleave2.nxv4i16(<vscale x 2 x i16> %splat1, <vscale x 2 x i16> %splat2)
+ ret <vscale x 4 x i16> %retval
+}
+
+define <vscale x 8 x i16> @interleave4_same_const_splat_nxv8i16() {
+; CHECK-LABEL: interleave4_same_const_splat_nxv8i16:
+; CHECK: // %bb.0:
+; CHECK-NEXT: mov z0.h, #3 // =0x3
+; CHECK-NEXT: ret
+ %retval = call <vscale x 8 x i16> @llvm.vector.interleave4.nxv8i16(<vscale x 2 x i16> splat(i16 3), <vscale x 2 x i16> splat(i16 3), <vscale x 2 x i16> splat(i16 3), <vscale x 2 x i16> splat(i16 3))
+ ret <vscale x 8 x i16> %retval
+}
diff --git a/llvm/test/CodeGen/RISCV/rvv/vector-interleave-fixed.ll b/llvm/test/CodeGen/RISCV/rvv/vector-interleave-fixed.ll
index 3dc83d50ee3f3..38d38f78c6054 100644
--- a/llvm/test/CodeGen/RISCV/rvv/vector-interleave-fixed.ll
+++ b/llvm/test/CodeGen/RISCV/rvv/vector-interleave-fixed.ll
@@ -1636,3 +1636,49 @@ define <8 x half> @vector_interleave8_v8f16_v1f16(<1 x half> %a, <1 x half> %b,
%res = call <8 x half> @llvm.vector.interleave8.v8f16(<1 x half> %a, <1 x half> %b, <1 x half> %c, <1 x half> %d, <1 x half> %e, <1 x half> %f, <1 x half> %g, <1 x half> %h)
ret <8 x half> %res
}
+
+define <8 x i16> @interleave4_const_splat_v8i16(<2 x i16> %a) {
+; CHECK-LABEL: interleave4_const_splat_v8i16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetivli zero, 8, e16, m1, ta, ma
+; CHECK-NEXT: vmv.v.i v8, 3
+; CHECK-NEXT: ret
+;
+; ZVBB-LABEL: interleave4_const_splat_v8i16:
+; ZVBB: # %bb.0:
+; ZVBB-NEXT: vsetivli zero, 8, e16, m1, ta, ma
+; ZVBB-NEXT: vmv.v.i v8, 3
+; ZVBB-NEXT: ret
+;
+; ZIP-LABEL: interleave4_const_splat_v8i16:
+; ZIP: # %bb.0:
+; ZIP-NEXT: vsetivli zero, 8, e16, m1, ta, ma
+; ZIP-NEXT: vmv.v.i v8, 3
+; ZIP-NEXT: ret
+ %retval = call <8 x i16> @llvm.vector.interleave4.v8i16(<2 x i16> splat(i16 3), <2 x i16> splat(i16 3), <2 x i16> splat(i16 3), <2 x i16> splat(i16 3))
+ ret <8 x i16> %retval
+}
+
+define <8 x i16> @interleave4_same_nonconst_splat_v8i16(i16 %a) {
+; CHECK-LABEL: interleave4_same_nonconst_splat_v8i16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetivli zero, 8, e16, m1, ta, ma
+; CHECK-NEXT: vmv.v.x v8, a0
+; CHECK-NEXT: ret
+;
+; ZVBB-LABEL: interleave4_same_nonconst_splat_v8i16:
+; ZVBB: # %bb.0:
+; ZVBB-NEXT: vsetivli zero, 8, e16, m1, ta, ma
+; ZVBB-NEXT: vmv.v.x v8, a0
+; ZVBB-NEXT: ret
+;
+; ZIP-LABEL: interleave4_same_nonconst_splat_v8i16:
+; ZIP: # %bb.0:
+; ZIP-NEXT: vsetivli zero, 8, e16, m1, ta, ma
+; ZIP-NEXT: vmv.v.x v8, a0
+; ZIP-NEXT: ret
+ %ins = insertelement <2 x i16> poison, i16 %a, i32 0
+ %splat = shufflevector <2 x i16> %ins, <2 x i16> poison, <2 x i32> zeroinitializer
+ %retval = call <8 x i16> @llvm.vector.interleave4.v8i16(<2 x i16> %splat, <2 x i16> %splat, <2 x i16> %splat, <2 x i16> %splat)
+ ret <8 x i16> %retval
+}
diff --git a/llvm/test/CodeGen/RISCV/rvv/vector-interleave.ll b/llvm/test/CodeGen/RISCV/rvv/vector-interleave.ll
index 01cc5c58b24ce..ee38257f09cd5 100644
--- a/llvm/test/CodeGen/RISCV/rvv/vector-interleave.ll
+++ b/llvm/test/CodeGen/RISCV/rvv/vector-interleave.ll
@@ -14947,3 +14947,147 @@ define <vscale x 16 x double> @vector_interleave_nxv16f64_nxv2f64(<vscale x 2 x
%res = call <vscale x 16 x double> @llvm.vector.interleave8.nxv16f64(<vscale x 2 x double> %v0, <vscale x 2 x double> %v1, <vscale x 2 x double> %v2, <vscale x 2 x double> %v3, <vscale x 2 x double> %v4, <vscale x 2 x double> %v5, <vscale x 2 x double> %v6, <vscale x 2 x double> %v7)
ret <vscale x 16 x double> %res
}
+
+define <vscale x 4 x i16> @interleave2_same_const_splat_nxv4i16() {
+; CHECK-LABEL: interleave2_same_const_splat_nxv4i16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli a0, zero, e16, m1, ta, ma
+; CHECK-NEXT: vmv.v.i v8, 3
+; CHECK-NEXT: ret
+;
+; ZVBB-LABEL: interleave2_same_const_splat_nxv4i16:
+; ZVBB: # %bb.0:
+; ZVBB-NEXT: vsetvli a0, zero, e16, m1, ta, ma
+; ZVBB-NEXT: vmv.v.i v8, 3
+; ZVBB-NEXT: ret
+ %retval = call <vscale x 4 x i16> @llvm.vector.interleave2.nxv4i16(<vscale x 2 x i16> splat(i16 3), <vscale x 2 x i16> splat(i16 3))
+ ret <vscale x 4 x i16> %retval
+}
+
+define <vscale x 4 x i16> @interleave2_
diff _const_splat_nxv4i16() {
+; V-LABEL: interleave2_
diff _const_splat_nxv4i16:
+; V: # %bb.0:
+; V-NEXT: vsetvli a0, zero, e16, mf2, ta, ma
+; V-NEXT: vmv.v.i v9, 3
+; V-NEXT: li a0, 4
+; V-NEXT: vmv.v.i v10, -1
+; V-NEXT: vwaddu.vx v8, v9, a0
+; V-NEXT: vwmaccu.vx v8, a0, v10
+; V-NEXT: csrr a0, vlenb
+; V-NEXT: srli a0, a0, 2
+; V-NEXT: vsetvli a1, zero, e16, m1, ta, ma
+; V-NEXT: vslidedown.vx v9, v8, a0
+; V-NEXT: vslideup.vx v8, v9, a0
+; V-NEXT: ret
+;
+; ZVBB-LABEL: interleave2_
diff _const_splat_nxv4i16:
+; ZVBB: # %bb.0:
+; ZVBB-NEXT: vsetvli a0, zero, e16, mf2, ta, ma
+; ZVBB-NEXT: vmv.v.i v8, 4
+; ZVBB-NEXT: li a0, 3
+; ZVBB-NEXT: vwsll.vi v9, v8, 16
+; ZVBB-NEXT: vwaddu.wx v8, v9, a0
+; ZVBB-NEXT: csrr a0, vlenb
+; ZVBB-NEXT: srli a0, a0, 2
+; ZVBB-NEXT: vsetvli a1, zero, e16, m1, ta, ma
+; ZVBB-NEXT: vslidedown.vx v9, v8, a0
+; ZVBB-NEXT: vslideup.vx v8, v9, a0
+; ZVBB-NEXT: ret
+;
+; ZIP-LABEL: interleave2_
diff _const_splat_nxv4i16:
+; ZIP: # %bb.0:
+; ZIP-NEXT: vsetvli a0, zero, e16, mf2, ta, ma
+; ZIP-NEXT: vmv.v.i v9, 4
+; ZIP-NEXT: vmv.v.i v10, 3
+; ZIP-NEXT: csrr a0, vlenb
+; ZIP-NEXT: ri.vzip2b.vv v11, v10, v9
+; ZIP-NEXT: ri.vzip2a.vv v8, v10, v9
+; ZIP-NEXT: srli a0, a0, 2
+; ZIP-NEXT: vsetvli a1, zero, e16, m1, ta, ma
+; ZIP-NEXT: vslideup.vx v8, v11, a0
+; ZIP-NEXT: ret
+ %retval = call <vscale x 4 x i16> @llvm.vector.interleave2.v4i16(<vscale x 2 x i16> splat(i16 3), <vscale x 2 x i16> splat(i16 4))
+ ret <vscale x 4 x i16> %retval
+}
+
+define <vscale x 4 x i16> @interleave2_same_nonconst_splat_nxv4i16(i16 %a) {
+; CHECK-LABEL: interleave2_same_nonconst_splat_nxv4i16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli a1, zero, e16, m1, ta, ma
+; CHECK-NEXT: vmv.v.x v8, a0
+; CHECK-NEXT: ret
+;
+; ZVBB-LABEL: interleave2_same_nonconst_splat_nxv4i16:
+; ZVBB: # %bb.0:
+; ZVBB-NEXT: vsetvli a1, zero, e16, m1, ta, ma
+; ZVBB-NEXT: vmv.v.x v8, a0
+; ZVBB-NEXT: ret
+ %ins = insertelement <vscale x 2 x i16> poison, i16 %a, i32 0
+ %splat = shufflevector <vscale x 2 x i16> %ins, <vscale x 2 x i16> poison, <vscale x 2 x i32> zeroinitializer
+ %retval = call <vscale x 4 x i16> @llvm.vector.interleave2.nxv4i16(<vscale x 2 x i16> %splat, <vscale x 2 x i16> %splat)
+ ret <vscale x 4 x i16> %retval
+}
+
+define <vscale x 4 x i16> @interleave2_
diff _nonconst_splat_nxv4i16(i16 %a, i16 %b) {
+; V-LABEL: interleave2_
diff _nonconst_splat_nxv4i16:
+; V: # %bb.0:
+; V-NEXT: vsetvli a2, zero, e16, mf2, ta, ma
+; V-NEXT: vmv.v.x v9, a0
+; V-NEXT: vmv.v.i v10, -1
+; V-NEXT: csrr a0, vlenb
+; V-NEXT: vwaddu.vx v8, v9, a1
+; V-NEXT: vwmaccu.vx v8, a1, v10
+; V-NEXT: srli a0, a0, 2
+; V-NEXT: vsetvli a1, zero, e16, m1, ta, ma
+; V-NEXT: vslidedown.vx v9, v8, a0
+; V-NEXT: vslideup.vx v8, v9, a0
+; V-NEXT: ret
+;
+; ZVBB-LABEL: interleave2_
diff _nonconst_splat_nxv4i16:
+; ZVBB: # %bb.0:
+; ZVBB-NEXT: vsetvli a2, zero, e16, mf2, ta, ma
+; ZVBB-NEXT: vmv.v.x v8, a1
+; ZVBB-NEXT: csrr a1, vlenb
+; ZVBB-NEXT: vwsll.vi v9, v8, 16
+; ZVBB-NEXT: vwaddu.wx v8, v9, a0
+; ZVBB-NEXT: srli a1, a1, 2
+; ZVBB-NEXT: vsetvli a0, zero, e16, m1, ta, ma
+; ZVBB-NEXT: vslidedown.vx v9, v8, a1
+; ZVBB-NEXT: vslideup.vx v8, v9, a1
+; ZVBB-NEXT: ret
+;
+; ZIP-LABEL: interleave2_
diff _nonconst_splat_nxv4i16:
+; ZIP: # %bb.0:
+; ZIP-NEXT: vsetvli a2, zero, e16, mf2, ta, ma
+; ZIP-NEXT: vmv.v.x v9, a0
+; ZIP-NEXT: vmv.v.x v10, a1
+; ZIP-NEXT: csrr a0, vlenb
+; ZIP-NEXT: ri.vzip2b.vv v11, v9, v10
+; ZIP-NEXT: ri.vzip2a.vv v8, v9, v10
+; ZIP-NEXT: srli a0, a0, 2
+; ZIP-NEXT: vsetvli a1, zero, e16, m1, ta, ma
+; ZIP-NEXT: vslideup.vx v8, v11, a0
+; ZIP-NEXT: ret
+ %ins1 = insertelement <vscale x 2 x i16> poison, i16 %a, i32 0
+ %splat1 = shufflevector <vscale x 2 x i16> %ins1, <vscale x 2 x i16> poison, <vscale x 2 x i32> zeroinitializer
+ %ins2 = insertelement <vscale x 2 x i16> poison, i16 %b, i32 0
+ %splat2 = shufflevector <vscale x 2 x i16> %ins2, <vscale x 2 x i16> poison, <vscale x 2 x i32> zeroinitializer
+ %retval = call <vscale x 4 x i16> @llvm.vector.interleave2.nxv4i16(<vscale x 2 x i16> %splat1, <vscale x 2 x i16> %splat2)
+ ret <vscale x 4 x i16> %retval
+}
+
+define <vscale x 8 x i16> @interleave4_same_const_splat_nxv8i16() {
+; CHECK-LABEL: interleave4_same_const_splat_nxv8i16:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli a0, zero, e16, m2, ta, ma
+; CHECK-NEXT: vmv.v.i v8, 3
+; CHECK-NEXT: ret
+;
+; ZVBB-LABEL: interleave4_same_const_splat_nxv8i16:
+; ZVBB: # %bb.0:
+; ZVBB-NEXT: vsetvli a0, zero, e16, m2, ta, ma
+; ZVBB-NEXT: vmv.v.i v8, 3
+; ZVBB-NEXT: ret
+ %retval = call <vscale x 8 x i16> @llvm.vector.interleave4.nxv8i16(<vscale x 2 x i16> splat(i16 3), <vscale x 2 x i16> splat(i16 3), <vscale x 2 x i16> splat(i16 3), <vscale x 2 x i16> splat(i16 3))
+ ret <vscale x 8 x i16> %retval
+}
More information about the llvm-commits
mailing list