[llvm] [DAG] Support saturated truncate (PR #99418)
via llvm-commits
llvm-commits at lists.llvm.org
Fri Aug 9 03:11:48 PDT 2024
https://github.com/ParkHanbum updated https://github.com/llvm/llvm-project/pull/99418
>From 8d81896a8b1f1137a2a182597551047c1092b1da Mon Sep 17 00:00:00 2001
From: hanbeom <kese111 at gmail.com>
Date: Fri, 9 Aug 2024 16:41:51 +0900
Subject: [PATCH 1/6] [DAG] Support saturated truncate
A truncate is considered saturated if no additional conversion is
required between the target and return values. If the target is
saturated when attempting to truncate from a vector, there is an
opportunity to optimize it.
Previously, each architecture had its own attempt at optimization,
leading to redundant code. This patch implements common logic by
introducing three new ISDs:
`ISD::TRUNCATE_SSAT_S`: When the operand is a signed value and
the range of values matches the range of signed values of the
destination type.
`ISD::TRUNCATE_SSAT_U`: When the operand is a signed value and
the range of values matches the range of unsigned values of the
destination type.
`ISD::TRUNCATE_USAT_U`: When the operand is an unsigned value and
the range of values matches the range of unsigned values of the
destination type.
These ISDs indicate a saturated truncate.
Fixes #85903
---
llvm/include/llvm/CodeGen/ISDOpcodes.h | 20 ++++
.../include/llvm/Target/TargetSelectionDAG.td | 3 +
llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp | 112 +++++++++++++++++-
.../SelectionDAG/SelectionDAGDumper.cpp | 3 +
llvm/lib/CodeGen/TargetLoweringBase.cpp | 5 +
5 files changed, 142 insertions(+), 1 deletion(-)
diff --git a/llvm/include/llvm/CodeGen/ISDOpcodes.h b/llvm/include/llvm/CodeGen/ISDOpcodes.h
index 5b657fb1712968..7305e3086fcd65 100644
--- a/llvm/include/llvm/CodeGen/ISDOpcodes.h
+++ b/llvm/include/llvm/CodeGen/ISDOpcodes.h
@@ -814,6 +814,26 @@ enum NodeType {
/// TRUNCATE - Completely drop the high bits.
TRUNCATE,
+ /// TRUNCATE_[SU]SAT_[SU] - Truncate for saturated operand
+ /// [SU] located in middle, prefix for `SAT` means indicates whether
+ /// existing truncate target was a signed operation. For examples,
+ /// If `truncate(smin(smax(x, C), C))` was saturated then become `S`.
+ /// If `truncate(umin(x, C))` was saturated then become `U`.
+ /// [SU] located in last indicates whether range of truncated values is
+ /// sign-saturated. For example, if `truncate(smin(smax(x, C), C))` is a
+ /// truncation to `i8`, then if value of C ranges from `-128 to 127`, it will
+ /// be saturated against signed values, resulting in `S`, which will combine
+ /// to `TRUNCATE_SSAT_S`. If the value of C ranges from `0 to 255`, it will
+ /// be saturated against unsigned values, resulting in `U`, which will
+ /// combine to `TRUNATE_SSAT_U`. Similarly, in `truncate(umin(x, C))`, if
+ /// value of C ranges from `0 to 255`, it becomes `U` because it is saturated
+ /// for unsigned values. As a result, it combines to `TRUNCATE_USAT_U`.
+ TRUNCATE_SSAT_S, // saturate signed input to signed result -
+ // truncate(smin(smax(x, C), C))
+ TRUNCATE_SSAT_U, // saturate signed input to unsigned result -
+ // truncate(smin(smax(x, 0), C))
+ TRUNCATE_USAT_U, // saturate unsigned input to unsigned result -
+ // truncate(umin(x, C))
/// [SU]INT_TO_FP - These operators convert integers (whose interpreted sign
/// depends on the first letter) to floating point.
diff --git a/llvm/include/llvm/Target/TargetSelectionDAG.td b/llvm/include/llvm/Target/TargetSelectionDAG.td
index 46044aab79a832..92d10a94bd81e5 100644
--- a/llvm/include/llvm/Target/TargetSelectionDAG.td
+++ b/llvm/include/llvm/Target/TargetSelectionDAG.td
@@ -477,6 +477,9 @@ def sext : SDNode<"ISD::SIGN_EXTEND", SDTIntExtendOp>;
def zext : SDNode<"ISD::ZERO_EXTEND", SDTIntExtendOp>;
def anyext : SDNode<"ISD::ANY_EXTEND" , SDTIntExtendOp>;
def trunc : SDNode<"ISD::TRUNCATE" , SDTIntTruncOp>;
+def truncssat_s : SDNode<"ISD::TRUNCATE_SSAT_S", SDTIntTruncOp>;
+def truncssat_u : SDNode<"ISD::TRUNCATE_SSAT_U", SDTIntTruncOp>;
+def truncusat_u : SDNode<"ISD::TRUNCATE_USAT_U", SDTIntTruncOp>;
def bitconvert : SDNode<"ISD::BITCAST" , SDTUnaryOp>;
def addrspacecast : SDNode<"ISD::ADDRSPACECAST", SDTUnaryOp>;
def freeze : SDNode<"ISD::FREEZE" , SDTFreeze>;
diff --git a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
index 060e66175d965c..7d2741f9a654ac 100644
--- a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
@@ -486,6 +486,7 @@ namespace {
SDValue visitSIGN_EXTEND_INREG(SDNode *N);
SDValue visitEXTEND_VECTOR_INREG(SDNode *N);
SDValue visitTRUNCATE(SDNode *N);
+ SDValue visitTRUNCATE_USAT(SDNode *N);
SDValue visitBITCAST(SDNode *N);
SDValue visitFREEZE(SDNode *N);
SDValue visitBUILD_PAIR(SDNode *N);
@@ -13203,7 +13204,9 @@ SDValue DAGCombiner::matchVSelectOpSizesWithSetCC(SDNode *Cast) {
unsigned CastOpcode = Cast->getOpcode();
assert((CastOpcode == ISD::SIGN_EXTEND || CastOpcode == ISD::ZERO_EXTEND ||
CastOpcode == ISD::TRUNCATE || CastOpcode == ISD::FP_EXTEND ||
- CastOpcode == ISD::FP_ROUND) &&
+ CastOpcode == ISD::TRUNCATE_SSAT_S ||
+ CastOpcode == ISD::TRUNCATE_SSAT_U ||
+ CastOpcode == ISD::TRUNCATE_USAT_U || CastOpcode == ISD::FP_ROUND) &&
"Unexpected opcode for vector select narrowing/widening");
// We only do this transform before legal ops because the pattern may be
@@ -14915,6 +14918,109 @@ SDValue DAGCombiner::visitEXTEND_VECTOR_INREG(SDNode *N) {
return SDValue();
}
+/// Detect patterns of truncation with unsigned saturation:
+///
+/// (truncate (umin (x, unsigned_max_of_dest_type)) to dest_type).
+/// Return the source value x to be truncated or SDValue() if the pattern was
+/// not matched.
+///
+static SDValue detectUSatUPattern(SDValue In, EVT VT) {
+ unsigned NumDstBits = VT.getScalarSizeInBits();
+ unsigned NumSrcBits = In.getScalarValueSizeInBits();
+ // Saturation with truncation. We truncate from InVT to VT.
+ assert(NumSrcBits > NumDstBits && "Unexpected types for truncate operation");
+
+ SDValue Min;
+ APInt UnsignedMax = APInt::getMaxValue(NumDstBits).zext(NumSrcBits);
+ if (sd_match(In, m_UMin(m_Value(Min), m_SpecificInt(UnsignedMax))))
+ return Min;
+
+ return SDValue();
+}
+
+/// Detect patterns of truncation with signed saturation:
+/// (truncate (smin (smax (x, signed_min_of_dest_type),
+/// signed_max_of_dest_type)) to dest_type)
+/// or:
+/// (truncate (smax (smin (x, signed_max_of_dest_type),
+/// signed_min_of_dest_type)) to dest_type).
+///
+/// Return the source value to be truncated or SDValue() if the pattern was not
+/// matched.
+static SDValue detectSSatSPattern(SDValue In, EVT VT) {
+ unsigned NumDstBits = VT.getScalarSizeInBits();
+ unsigned NumSrcBits = In.getScalarValueSizeInBits();
+ // Saturation with truncation. We truncate from InVT to VT.
+ assert(NumSrcBits > NumDstBits && "Unexpected types for truncate operation");
+
+ SDValue Val;
+ APInt SignedMax = APInt::getSignedMaxValue(NumDstBits).sext(NumSrcBits);
+ APInt SignedMin = APInt::getSignedMinValue(NumDstBits).sext(NumSrcBits);
+
+ if (sd_match(In, m_SMin(m_SMax(m_Value(Val), m_SpecificInt(SignedMin)),
+ m_SpecificInt(SignedMax))))
+ return Val;
+
+ if (sd_match(In, m_SMax(m_SMin(m_Value(Val), m_SpecificInt(SignedMax)),
+ m_SpecificInt(SignedMin))))
+ return Val;
+
+ return SDValue();
+}
+
+/// Detect patterns of truncation with unsigned saturation:
+static SDValue detectSSatUPattern(SDValue In, EVT VT, SelectionDAG &DAG,
+ const SDLoc &DL) {
+ unsigned NumDstBits = VT.getScalarSizeInBits();
+ unsigned NumSrcBits = In.getScalarValueSizeInBits();
+ // Saturation with truncation. We truncate from InVT to VT.
+ assert(NumSrcBits > NumDstBits && "Unexpected types for truncate operation");
+
+ SDValue Val;
+ APInt UnsignedMax = APInt::getMaxValue(NumDstBits).zext(NumSrcBits);
+ // Min == 0, Max is unsigned max of destination type.
+ if (sd_match(In, m_SMax(m_SMin(m_Value(Val), m_SpecificInt(UnsignedMax)),
+ m_Zero())))
+ return Val;
+
+ if (sd_match(In, m_SMin(m_SMax(m_Value(Val), m_Zero()),
+ m_SpecificInt(UnsignedMax))))
+ return Val;
+
+ if (sd_match(In, m_UMin(m_SMax(m_Value(Val), m_Zero()),
+ m_SpecificInt(UnsignedMax))))
+ return Val;
+
+ return SDValue();
+}
+
+static SDValue foldToSaturated(SDNode *N, EVT &VT, SDValue &Src, EVT &SrcVT,
+ SDLoc &DL, const TargetLowering &TLI,
+ SelectionDAG &DAG) {
+ auto AllowedTruncateSat = [&](unsigned Opc, EVT SrcVT, EVT VT) -> bool {
+ return (TLI.isOperationLegalOrCustom(Opc, SrcVT) &&
+ TLI.isTypeDesirableForOp(Opc, VT));
+ };
+
+ if (Src.getOpcode() == ISD::SMIN || Src.getOpcode() == ISD::SMAX) {
+ if (AllowedTruncateSat(ISD::TRUNCATE_SSAT_S, SrcVT, VT))
+ if (SDValue SSatVal = detectSSatSPattern(Src, VT))
+ return DAG.getNode(ISD::TRUNCATE_SSAT_S, DL, VT, SSatVal);
+ if (AllowedTruncateSat(ISD::TRUNCATE_SSAT_U, SrcVT, VT))
+ if (SDValue SSatVal = detectSSatUPattern(Src, VT, DAG, DL))
+ return DAG.getNode(ISD::TRUNCATE_SSAT_U, DL, VT, SSatVal);
+ } else if (Src.getOpcode() == ISD::UMIN) {
+ if (AllowedTruncateSat(ISD::TRUNCATE_SSAT_U, SrcVT, VT))
+ if (SDValue SSatVal = detectSSatUPattern(Src, VT, DAG, DL))
+ return DAG.getNode(ISD::TRUNCATE_SSAT_U, DL, VT, SSatVal);
+ if (AllowedTruncateSat(ISD::TRUNCATE_USAT_U, SrcVT, VT))
+ if (SDValue USatVal = detectUSatUPattern(Src, VT))
+ return DAG.getNode(ISD::TRUNCATE_USAT_U, DL, VT, USatVal);
+ }
+
+ return SDValue();
+}
+
SDValue DAGCombiner::visitTRUNCATE(SDNode *N) {
SDValue N0 = N->getOperand(0);
EVT VT = N->getValueType(0);
@@ -14930,6 +15036,10 @@ SDValue DAGCombiner::visitTRUNCATE(SDNode *N) {
if (N0.getOpcode() == ISD::TRUNCATE)
return DAG.getNode(ISD::TRUNCATE, DL, VT, N0.getOperand(0));
+ // fold saturated truncate
+ if (SDValue SaturatedTR = foldToSaturated(N, VT, N0, SrcVT, DL, TLI, DAG))
+ return SaturatedTR;
+
// fold (truncate c1) -> c1
if (SDValue C = DAG.FoldConstantArithmetic(ISD::TRUNCATE, DL, VT, {N0}))
return C;
diff --git a/llvm/lib/CodeGen/SelectionDAG/SelectionDAGDumper.cpp b/llvm/lib/CodeGen/SelectionDAG/SelectionDAGDumper.cpp
index 16fc52caebb757..46e8e54ee4ed7d 100644
--- a/llvm/lib/CodeGen/SelectionDAG/SelectionDAGDumper.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/SelectionDAGDumper.cpp
@@ -380,6 +380,9 @@ std::string SDNode::getOperationName(const SelectionDAG *G) const {
case ISD::SIGN_EXTEND_VECTOR_INREG: return "sign_extend_vector_inreg";
case ISD::ZERO_EXTEND_VECTOR_INREG: return "zero_extend_vector_inreg";
case ISD::TRUNCATE: return "truncate";
+ case ISD::TRUNCATE_SSAT_S: return "truncate_ssat_s";
+ case ISD::TRUNCATE_SSAT_U: return "truncate_ssat_u";
+ case ISD::TRUNCATE_USAT_U: return "truncate_usat_u";
case ISD::FP_ROUND: return "fp_round";
case ISD::STRICT_FP_ROUND: return "strict_fp_round";
case ISD::FP_EXTEND: return "fp_extend";
diff --git a/llvm/lib/CodeGen/TargetLoweringBase.cpp b/llvm/lib/CodeGen/TargetLoweringBase.cpp
index 6ca9955993d242..149b5dabee0565 100644
--- a/llvm/lib/CodeGen/TargetLoweringBase.cpp
+++ b/llvm/lib/CodeGen/TargetLoweringBase.cpp
@@ -753,6 +753,11 @@ void TargetLoweringBase::initActions() {
// Absolute difference
setOperationAction({ISD::ABDS, ISD::ABDU}, VT, Expand);
+ // Saturated trunc
+ setOperationAction(ISD::TRUNCATE_SSAT_S, VT, Expand);
+ setOperationAction(ISD::TRUNCATE_SSAT_U, VT, Expand);
+ setOperationAction(ISD::TRUNCATE_USAT_U, VT, Expand);
+
// These default to Expand so they will be expanded to CTLZ/CTTZ by default.
setOperationAction({ISD::CTLZ_ZERO_UNDEF, ISD::CTTZ_ZERO_UNDEF}, VT,
Expand);
>From 7e67747d6a4827eb63abda4fe48ed1714f3bb450 Mon Sep 17 00:00:00 2001
From: hanbeom <kese111 at gmail.com>
Date: Fri, 9 Aug 2024 16:58:07 +0900
Subject: [PATCH 2/6] [DAG] Combine `trunc[us]_u(fptoui(x))` to `fptoui_sat(x)`
---
llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp | 25 +++++++++++++++++++
1 file changed, 25 insertions(+)
diff --git a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
index 7d2741f9a654ac..788da40a6addd1 100644
--- a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
@@ -1909,6 +1909,8 @@ SDValue DAGCombiner::visit(SDNode *N) {
case ISD::ZERO_EXTEND_VECTOR_INREG:
case ISD::ANY_EXTEND_VECTOR_INREG: return visitEXTEND_VECTOR_INREG(N);
case ISD::TRUNCATE: return visitTRUNCATE(N);
+ case ISD::TRUNCATE_SSAT_U:
+ case ISD::TRUNCATE_USAT_U: return visitTRUNCATE_USAT(N);
case ISD::BITCAST: return visitBITCAST(N);
case ISD::BUILD_PAIR: return visitBUILD_PAIR(N);
case ISD::FADD: return visitFADD(N);
@@ -14918,6 +14920,29 @@ SDValue DAGCombiner::visitEXTEND_VECTOR_INREG(SDNode *N) {
return SDValue();
}
+SDValue DAGCombiner::visitTRUNCATE_USAT(SDNode *N) {
+ EVT VT = N->getValueType(0);
+ SDValue N0 = N->getOperand(0);
+
+ std::function<SDValue(SDValue)> MatchFPTOINT = [&](SDValue Val) -> SDValue {
+ if (Val.getOpcode() == ISD::FP_TO_UINT)
+ return Val;
+ return SDValue();
+ };
+
+ SDValue FPInstr = MatchFPTOINT(N0);
+ if (!FPInstr)
+ return SDValue();
+
+ EVT FPVT = FPInstr.getOperand(0).getValueType();
+ if (!DAG.getTargetLoweringInfo().shouldConvertFpToSat(ISD::FP_TO_UINT_SAT,
+ FPVT, VT))
+ return SDValue();
+ return DAG.getNode(ISD::FP_TO_UINT_SAT, SDLoc(FPInstr), VT,
+ FPInstr.getOperand(0),
+ DAG.getValueType(VT.getScalarType()));
+}
+
/// Detect patterns of truncation with unsigned saturation:
///
/// (truncate (umin (x, unsigned_max_of_dest_type)) to dest_type).
>From 7353da4914c033250f4d582186b67ddf4f54bd40 Mon Sep 17 00:00:00 2001
From: hanbeom <kese111 at gmail.com>
Date: Fri, 9 Aug 2024 17:18:42 +0900
Subject: [PATCH 3/6] [AArch64] Add tests for saturated truncate
---
llvm/test/CodeGen/AArch64/qmovn.ll | 169 +++++++++++++++++++++++++++++
1 file changed, 169 insertions(+)
diff --git a/llvm/test/CodeGen/AArch64/qmovn.ll b/llvm/test/CodeGen/AArch64/qmovn.ll
index 35c172adbad3d8..90dec7e3c35264 100644
--- a/llvm/test/CodeGen/AArch64/qmovn.ll
+++ b/llvm/test/CodeGen/AArch64/qmovn.ll
@@ -125,6 +125,117 @@ entry:
ret <2 x i32> %t
}
+define <2 x i32> @vqmovni64_smaxmin_u(<2 x i64> %s0) {
+; CHECK-LABEL: vqmovni64_smaxmin_u:
+; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: movi v1.2d, #0x000000ffffffff
+; CHECK-NEXT: cmgt v2.2d, v1.2d, v0.2d
+; CHECK-NEXT: bif v0.16b, v1.16b, v2.16b
+; CHECK-NEXT: cmgt v1.2d, v0.2d, #0
+; CHECK-NEXT: and v0.16b, v0.16b, v1.16b
+; CHECK-NEXT: xtn v0.2s, v0.2d
+; CHECK-NEXT: ret
+entry:
+ %c1 = icmp slt <2 x i64> %s0, <i64 4294967295, i64 4294967295>
+ %s1 = select <2 x i1> %c1, <2 x i64> %s0, <2 x i64> <i64 4294967295, i64 4294967295>
+ %c2 = icmp sgt <2 x i64> %s1, zeroinitializer
+ %s2 = select <2 x i1> %c2, <2 x i64> %s1, <2 x i64> zeroinitializer
+ %t = trunc <2 x i64> %s2 to <2 x i32>
+ ret <2 x i32> %t
+}
+
+define <2 x i32> @vqmovni64_sminmax_u(<2 x i64> %s0) {
+; CHECK-LABEL: vqmovni64_sminmax_u:
+; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: cmgt v1.2d, v0.2d, #0
+; CHECK-NEXT: movi v2.2d, #0x000000ffffffff
+; CHECK-NEXT: and v0.16b, v0.16b, v1.16b
+; CHECK-NEXT: cmgt v1.2d, v2.2d, v0.2d
+; CHECK-NEXT: and v0.16b, v0.16b, v1.16b
+; CHECK-NEXT: orn v0.16b, v0.16b, v1.16b
+; CHECK-NEXT: xtn v0.2s, v0.2d
+; CHECK-NEXT: ret
+entry:
+ %c1 = icmp sgt <2 x i64> %s0, zeroinitializer
+ %s1 = select <2 x i1> %c1, <2 x i64> %s0, <2 x i64> zeroinitializer
+ %c2 = icmp slt <2 x i64> %s1, <i64 4294967295, i64 4294967295>
+ %s2 = select <2 x i1> %c2, <2 x i64> %s1, <2 x i64> <i64 4294967295, i64 4294967295>
+ %t = trunc <2 x i64> %s2 to <2 x i32>
+ ret <2 x i32> %t
+}
+
+define <4 x i16> @vqmovni32_smaxmin_u(<4 x i32> %s0) {
+; CHECK-LABEL: vqmovni32_smaxmin_u:
+; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: movi v1.2d, #0x00ffff0000ffff
+; CHECK-NEXT: smin v0.4s, v0.4s, v1.4s
+; CHECK-NEXT: movi v1.2d, #0000000000000000
+; CHECK-NEXT: smax v0.4s, v0.4s, v1.4s
+; CHECK-NEXT: xtn v0.4h, v0.4s
+; CHECK-NEXT: ret
+entry:
+ %c1 = icmp slt <4 x i32> %s0, <i32 65535, i32 65535, i32 65535, i32 65535>
+ %s1 = select <4 x i1> %c1, <4 x i32> %s0, <4 x i32> <i32 65535, i32 65535, i32 65535, i32 65535>
+ %c2 = icmp sgt <4 x i32> %s1, zeroinitializer
+ %s2 = select <4 x i1> %c2, <4 x i32> %s1, <4 x i32> zeroinitializer
+ %t = trunc <4 x i32> %s2 to <4 x i16>
+ ret <4 x i16> %t
+}
+
+define <4 x i16> @vqmovni32_sminmax_u(<4 x i32> %s0) {
+; CHECK-LABEL: vqmovni32_sminmax_u:
+; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: movi v1.2d, #0000000000000000
+; CHECK-NEXT: smax v0.4s, v0.4s, v1.4s
+; CHECK-NEXT: movi v1.2d, #0x00ffff0000ffff
+; CHECK-NEXT: smin v0.4s, v0.4s, v1.4s
+; CHECK-NEXT: xtn v0.4h, v0.4s
+; CHECK-NEXT: ret
+entry:
+ %c1 = icmp sgt <4 x i32> %s0, zeroinitializer
+ %s1 = select <4 x i1> %c1, <4 x i32> %s0, <4 x i32> zeroinitializer
+ %c2 = icmp slt <4 x i32> %s1, <i32 65535, i32 65535, i32 65535, i32 65535>
+ %s2 = select <4 x i1> %c2, <4 x i32> %s1, <4 x i32> <i32 65535, i32 65535, i32 65535, i32 65535>
+ %t = trunc <4 x i32> %s2 to <4 x i16>
+ ret <4 x i16> %t
+}
+
+define <8 x i8> @vqmovni16_smaxmin_u(<8 x i16> %s0) {
+; CHECK-LABEL: vqmovni16_smaxmin_u:
+; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: movi v1.2d, #0xff00ff00ff00ff
+; CHECK-NEXT: movi v2.2d, #0000000000000000
+; CHECK-NEXT: smin v0.8h, v0.8h, v1.8h
+; CHECK-NEXT: smax v0.8h, v0.8h, v2.8h
+; CHECK-NEXT: xtn v0.8b, v0.8h
+; CHECK-NEXT: ret
+entry:
+ %c1 = icmp slt <8 x i16> %s0, <i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255>
+ %s1 = select <8 x i1> %c1, <8 x i16> %s0, <8 x i16> <i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255>
+ %c2 = icmp sgt <8 x i16> %s1, zeroinitializer
+ %s2 = select <8 x i1> %c2, <8 x i16> %s1, <8 x i16> zeroinitializer
+ %t = trunc <8 x i16> %s2 to <8 x i8>
+ ret <8 x i8> %t
+}
+
+define <8 x i8> @vqmovni16_sminmax_u(<8 x i16> %s0) {
+; CHECK-LABEL: vqmovni16_sminmax_u:
+; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: movi v1.2d, #0000000000000000
+; CHECK-NEXT: movi v2.2d, #0xff00ff00ff00ff
+; CHECK-NEXT: smax v0.8h, v0.8h, v1.8h
+; CHECK-NEXT: smin v0.8h, v0.8h, v2.8h
+; CHECK-NEXT: xtn v0.8b, v0.8h
+; CHECK-NEXT: ret
+entry:
+ %c1 = icmp sgt <8 x i16> %s0, zeroinitializer
+ %s1 = select <8 x i1> %c1, <8 x i16> %s0, <8 x i16> zeroinitializer
+ %c2 = icmp slt <8 x i16> %s1, <i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255>
+ %s2 = select <8 x i1> %c2, <8 x i16> %s1, <8 x i16> <i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255>
+ %t = trunc <8 x i16> %s2 to <8 x i8>
+ ret <8 x i8> %t
+}
+
define <2 x i32> @vqmovni64_umaxmin(<2 x i64> %s0) {
; CHECK-LABEL: vqmovni64_umaxmin:
; CHECK: // %bb.0: // %entry
@@ -347,3 +458,61 @@ entry:
%shuffle = shufflevector <2 x i32> %x, <2 x i32> %trunc, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
ret <4 x i32> %shuffle
}
+
+; Test the (concat_vectors (X), (trunc(smin(smax(Y, 0), 2^n))))) pattern.
+
+define <16 x i8> @sminsmax_range_unsigned_i16_to_i8(<8 x i8> %x, <8 x i16> %y) {
+; CHECK-LABEL: sminsmax_range_unsigned_i16_to_i8:
+; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: movi v2.2d, #0000000000000000
+; CHECK-NEXT: movi v3.2d, #0xff00ff00ff00ff
+; CHECK-NEXT: // kill: def $d0 killed $d0 def $q0
+; CHECK-NEXT: smax v1.8h, v1.8h, v2.8h
+; CHECK-NEXT: smin v1.8h, v1.8h, v3.8h
+; CHECK-NEXT: xtn2 v0.16b, v1.8h
+; CHECK-NEXT: ret
+entry:
+ %min = call <8 x i16> @llvm.smax.v8i16(<8 x i16> %y, <8 x i16> zeroinitializer)
+ %max = call <8 x i16> @llvm.smin.v8i16(<8 x i16> %min, <8 x i16> <i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255>)
+ %trunc = trunc <8 x i16> %max to <8 x i8>
+ %shuffle = shufflevector <8 x i8> %x, <8 x i8> %trunc, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+ ret <16 x i8> %shuffle
+}
+
+define <8 x i16> @sminsmax_range_unsigned_i32_to_i16(<4 x i16> %x, <4 x i32> %y) {
+; CHECK-LABEL: sminsmax_range_unsigned_i32_to_i16:
+; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: movi v2.2d, #0000000000000000
+; CHECK-NEXT: // kill: def $d0 killed $d0 def $q0
+; CHECK-NEXT: smax v1.4s, v1.4s, v2.4s
+; CHECK-NEXT: movi v2.2d, #0x00ffff0000ffff
+; CHECK-NEXT: smin v1.4s, v1.4s, v2.4s
+; CHECK-NEXT: xtn2 v0.8h, v1.4s
+; CHECK-NEXT: ret
+entry:
+ %smax = call <4 x i32> @llvm.smax.v4i32(<4 x i32> %y, <4 x i32> zeroinitializer)
+ %smin = call <4 x i32> @llvm.smin.v4i32(<4 x i32> %smax, <4 x i32> <i32 65535, i32 65535, i32 65535, i32 65535>)
+ %trunc = trunc <4 x i32> %smin to <4 x i16>
+ %shuffle = shufflevector <4 x i16> %x, <4 x i16> %trunc, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+ ret <8 x i16> %shuffle
+}
+
+define <4 x i32> @sminsmax_range_unsigned_i64_to_i32(<2 x i32> %x, <2 x i64> %y) {
+; CHECK-LABEL: sminsmax_range_unsigned_i64_to_i32:
+; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: cmgt v2.2d, v1.2d, #0
+; CHECK-NEXT: movi v3.2d, #0x000000ffffffff
+; CHECK-NEXT: // kill: def $d0 killed $d0 def $q0
+; CHECK-NEXT: and v1.16b, v1.16b, v2.16b
+; CHECK-NEXT: cmgt v2.2d, v3.2d, v1.2d
+; CHECK-NEXT: and v1.16b, v1.16b, v2.16b
+; CHECK-NEXT: orn v1.16b, v1.16b, v2.16b
+; CHECK-NEXT: xtn2 v0.4s, v1.2d
+; CHECK-NEXT: ret
+entry:
+ %smax = call <2 x i64> @llvm.smax.v2i64(<2 x i64> %y, <2 x i64> zeroinitializer)
+ %smin = call <2 x i64> @llvm.smin.v2i64(<2 x i64> %smax, <2 x i64> <i64 4294967295, i64 4294967295>)
+ %trunc = trunc <2 x i64> %smin to <2 x i32>
+ %shuffle = shufflevector <2 x i32> %x, <2 x i32> %trunc, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+ ret <4 x i32> %shuffle
+}
>From 7c1f22ea8d719d7c7b6802077ac7280b06aa3d64 Mon Sep 17 00:00:00 2001
From: hanbeom <kese111 at gmail.com>
Date: Thu, 1 Aug 2024 20:03:35 +0900
Subject: [PATCH 4/6] [AArch64] Add fail tests with unsupported types in
saturated truncate
---
llvm/test/CodeGen/AArch64/qmovn.ll | 111 +++++++++++++++++++++++++++++
1 file changed, 111 insertions(+)
diff --git a/llvm/test/CodeGen/AArch64/qmovn.ll b/llvm/test/CodeGen/AArch64/qmovn.ll
index 90dec7e3c35264..1eab8ab4862998 100644
--- a/llvm/test/CodeGen/AArch64/qmovn.ll
+++ b/llvm/test/CodeGen/AArch64/qmovn.ll
@@ -516,3 +516,114 @@ entry:
%shuffle = shufflevector <2 x i32> %x, <2 x i32> %trunc, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
ret <4 x i32> %shuffle
}
+
+; Type support varification - not supported with saturated value
+; i64 -> i16
+define <4 x i16> @sminsmax_range_unsigned_i64_to_i16(<2 x i16> %x, <2 x i64> %y) {
+; CHECK-LABEL: sminsmax_range_unsigned_i64_to_i16:
+; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: cmgt v2.2d, v1.2d, #0
+; CHECK-NEXT: movi v3.2d, #0x0000000000ffff
+; CHECK-NEXT: and v1.16b, v1.16b, v2.16b
+; CHECK-NEXT: cmgt v2.2d, v3.2d, v1.2d
+; CHECK-NEXT: bif v1.16b, v3.16b, v2.16b
+; CHECK-NEXT: xtn v1.2s, v1.2d
+; CHECK-NEXT: uzp1 v0.4h, v0.4h, v1.4h
+; CHECK-NEXT: ret
+entry:
+ %smax = call <2 x i64> @llvm.smax.v2i64(<2 x i64> %y, <2 x i64> zeroinitializer)
+ %smin = call <2 x i64> @llvm.smin.v2i64(<2 x i64> %smax, <2 x i64> <i64 65535, i64 65535>)
+ %trunc = trunc <2 x i64> %smin to <2 x i16>
+ %shuffle = shufflevector <2 x i16> %x, <2 x i16> %trunc, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+ ret <4 x i16> %shuffle
+}
+
+define <4 x i16> @sminsmax_range_signed_i64_to_i16(<2 x i16> %x, <2 x i64> %y) {
+; CHECK-LABEL: sminsmax_range_signed_i64_to_i16:
+; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: mov x8, #-32768 // =0xffffffffffff8000
+; CHECK-NEXT: dup v2.2d, x8
+; CHECK-NEXT: mov w8, #32767 // =0x7fff
+; CHECK-NEXT: cmgt v3.2d, v1.2d, v2.2d
+; CHECK-NEXT: bif v1.16b, v2.16b, v3.16b
+; CHECK-NEXT: dup v2.2d, x8
+; CHECK-NEXT: cmgt v3.2d, v2.2d, v1.2d
+; CHECK-NEXT: bif v1.16b, v2.16b, v3.16b
+; CHECK-NEXT: xtn v1.2s, v1.2d
+; CHECK-NEXT: uzp1 v0.4h, v0.4h, v1.4h
+; CHECK-NEXT: ret
+entry:
+ %smax = call <2 x i64> @llvm.smax.v2i64(<2 x i64> %y, <2 x i64> <i64 -32768, i64 -32768>)
+ %smin = call <2 x i64> @llvm.smin.v2i64(<2 x i64> %smax, <2 x i64> <i64 32767, i64 32767>)
+ %trunc = trunc <2 x i64> %smin to <2 x i16>
+ %shuffle = shufflevector <2 x i16> %x, <2 x i16> %trunc, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+ ret <4 x i16> %shuffle
+}
+
+define <4 x i16> @umin_range_unsigned_i64_to_i16(<2 x i16> %x, <2 x i64> %y) {
+; CHECK-LABEL: umin_range_unsigned_i64_to_i16:
+; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: movi v2.2d, #0x0000000000ffff
+; CHECK-NEXT: cmhi v3.2d, v2.2d, v1.2d
+; CHECK-NEXT: bif v1.16b, v2.16b, v3.16b
+; CHECK-NEXT: xtn v1.2s, v1.2d
+; CHECK-NEXT: uzp1 v0.4h, v0.4h, v1.4h
+; CHECK-NEXT: ret
+entry:
+ %umin = call <2 x i64> @llvm.umin.v2i64(<2 x i64> %y, <2 x i64> <i64 65535, i64 65535>)
+ %trunc = trunc <2 x i64> %umin to <2 x i16>
+ %shuffle = shufflevector <2 x i16> %x, <2 x i16> %trunc, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+ ret <4 x i16> %shuffle
+}
+
+; i32 -> i8
+define <8 x i8> @sminsmax_range_unsigned_i64_to_i8(<4 x i8> %x, <4 x i32> %y) {
+; CHECK-LABEL: sminsmax_range_unsigned_i64_to_i8:
+; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: movi v2.2d, #0000000000000000
+; CHECK-NEXT: smax v1.4s, v1.4s, v2.4s
+; CHECK-NEXT: movi v2.2d, #0x0000ff000000ff
+; CHECK-NEXT: smin v1.4s, v1.4s, v2.4s
+; CHECK-NEXT: xtn v1.4h, v1.4s
+; CHECK-NEXT: uzp1 v0.8b, v0.8b, v1.8b
+; CHECK-NEXT: ret
+entry:
+ %smax = call <4 x i32> @llvm.smax.v4i32(<4 x i32> %y, <4 x i32> zeroinitializer)
+ %smin = call <4 x i32> @llvm.smin.v4i32(<4 x i32> %smax, <4 x i32> <i32 255, i32 255, i32 255, i32 255>)
+ %trunc = trunc <4 x i32> %smin to <4 x i8>
+ %shuffle = shufflevector <4 x i8> %x, <4 x i8> %trunc, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+ ret <8 x i8> %shuffle
+}
+
+define <8 x i8> @sminsmax_range_signed_i32_to_i8(<4 x i8> %x, <4 x i32> %y) {
+; CHECK-LABEL: sminsmax_range_signed_i32_to_i8:
+; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: mvni v2.4s, #127
+; CHECK-NEXT: smax v1.4s, v1.4s, v2.4s
+; CHECK-NEXT: movi v2.4s, #127
+; CHECK-NEXT: smin v1.4s, v1.4s, v2.4s
+; CHECK-NEXT: xtn v1.4h, v1.4s
+; CHECK-NEXT: uzp1 v0.8b, v0.8b, v1.8b
+; CHECK-NEXT: ret
+entry:
+ %smax = call <4 x i32> @llvm.smax.v4i32(<4 x i32> %y, <4 x i32> <i32 -128, i32 -128, i32 -128, i32 -128>)
+ %smin = call <4 x i32> @llvm.smin.v4i32(<4 x i32> %smax, <4 x i32> <i32 127, i32 127, i32 127, i32 127>)
+ %trunc = trunc <4 x i32> %smin to <4 x i8>
+ %shuffle = shufflevector <4 x i8> %x, <4 x i8> %trunc, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+ ret <8 x i8> %shuffle
+}
+
+define <8 x i8> @umin_range_unsigned_i32_to_i8(<4 x i8> %x, <4 x i32> %y) {
+; CHECK-LABEL: umin_range_unsigned_i32_to_i8:
+; CHECK: // %bb.0: // %entry
+; CHECK-NEXT: movi v2.2d, #0x0000ff000000ff
+; CHECK-NEXT: umin v1.4s, v1.4s, v2.4s
+; CHECK-NEXT: xtn v1.4h, v1.4s
+; CHECK-NEXT: uzp1 v0.8b, v0.8b, v1.8b
+; CHECK-NEXT: ret
+entry:
+ %umin = call <4 x i32> @llvm.umin.v4i32(<4 x i32> %y, <4 x i32> <i32 255, i32 255, i32 255, i32 255>)
+ %trunc = trunc <4 x i32> %umin to <4 x i8>
+ %shuffle = shufflevector <4 x i8> %x, <4 x i8> %trunc, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+ ret <8 x i8> %shuffle
+}
>From 49c4481aebee59b536daa353d0fa890ff921194c Mon Sep 17 00:00:00 2001
From: hanbeom <kese111 at gmail.com>
Date: Tue, 16 Jul 2024 14:05:55 +0900
Subject: [PATCH 5/6] [AArch64] Support saturated truncate
Add support for saturated truncate with the following changes:
- Add action to Legal for types v8i16, v4i32, and v2i64
- Implement `isTypeDesirableForOp` to check for truncate conversions
- Add patterns for saturated truncate of supported types
---
.../Target/AArch64/AArch64ISelLowering.cpp | 18 +++
llvm/lib/Target/AArch64/AArch64ISelLowering.h | 5 +
llvm/lib/Target/AArch64/AArch64InstrInfo.td | 83 ++++++-----
llvm/test/CodeGen/AArch64/fpclamptosat_vec.ll | 56 +++-----
llvm/test/CodeGen/AArch64/qmovn.ll | 131 +++---------------
5 files changed, 108 insertions(+), 185 deletions(-)
diff --git a/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp b/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
index d86e52d49000ae..2ebe2ff2712d6d 100644
--- a/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
+++ b/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
@@ -1410,6 +1410,12 @@ AArch64TargetLowering::AArch64TargetLowering(const TargetMachine &TM,
}
}
+ for (MVT VT : {MVT::v8i16, MVT::v4i32, MVT::v2i64}) {
+ setOperationAction(ISD::TRUNCATE_SSAT_S, VT, Legal);
+ setOperationAction(ISD::TRUNCATE_SSAT_U, VT, Legal);
+ setOperationAction(ISD::TRUNCATE_USAT_U, VT, Legal);
+ }
+
if (Subtarget->hasSME()) {
setOperationAction(ISD::INTRINSIC_W_CHAIN, MVT::Other, Custom);
}
@@ -28730,6 +28736,18 @@ bool AArch64TargetLowering::hasInlineStackProbe(
MF.getInfo<AArch64FunctionInfo>()->hasStackProbing();
}
+bool AArch64TargetLowering::isTypeDesirableForOp(unsigned Opc, EVT VT) const {
+ switch (Opc) {
+ case ISD::TRUNCATE_SSAT_S:
+ case ISD::TRUNCATE_SSAT_U:
+ case ISD::TRUNCATE_USAT_U:
+ if (VT == MVT::v8i8 || VT == MVT::v4i16 || VT == MVT::v2i32)
+ return true;
+ }
+
+ return TargetLowering::isTypeDesirableForOp(Opc, VT);
+}
+
#ifndef NDEBUG
void AArch64TargetLowering::verifyTargetSDNode(const SDNode *N) const {
switch (N->getOpcode()) {
diff --git a/llvm/lib/Target/AArch64/AArch64ISelLowering.h b/llvm/lib/Target/AArch64/AArch64ISelLowering.h
index 81e15185f985d5..50e26612ac863e 100644
--- a/llvm/lib/Target/AArch64/AArch64ISelLowering.h
+++ b/llvm/lib/Target/AArch64/AArch64ISelLowering.h
@@ -743,6 +743,11 @@ class AArch64TargetLowering : public TargetLowering {
bool generateFMAsInMachineCombiner(EVT VT,
CodeGenOptLevel OptLevel) const override;
+ /// Return true if the target has native support for
+ /// the specified value type and it is 'desirable' to use the type for the
+ /// given node type.
+ bool isTypeDesirableForOp(unsigned Opc, EVT VT) const override;
+
const MCPhysReg *getScratchRegisters(CallingConv::ID CC) const override;
ArrayRef<MCPhysReg> getRoundingControlRegisters() const override;
diff --git a/llvm/lib/Target/AArch64/AArch64InstrInfo.td b/llvm/lib/Target/AArch64/AArch64InstrInfo.td
index 1053ba9242768a..9ed8d3ee3dcee0 100644
--- a/llvm/lib/Target/AArch64/AArch64InstrInfo.td
+++ b/llvm/lib/Target/AArch64/AArch64InstrInfo.td
@@ -5418,64 +5418,75 @@ def VImm7FFF: PatLeaf<(AArch64movi_msl (i32 127), (i32 264))>;
def VImm8000: PatLeaf<(AArch64mvni_msl (i32 127), (i32 264))>;
// trunc(umin(X, 255)) -> UQXTRN v8i8
-def : Pat<(v8i8 (trunc (umin (v8i16 V128:$Vn), (v8i16 VImmFF)))),
+def : Pat<(v8i8 (truncusat_u (v8i16 V128:$Vn))),
(UQXTNv8i8 V128:$Vn)>;
// trunc(umin(X, 65535)) -> UQXTRN v4i16
-def : Pat<(v4i16 (trunc (umin (v4i32 V128:$Vn), (v4i32 VImmFFFF)))),
+def : Pat<(v4i16 (truncusat_u (v4i32 V128:$Vn))),
(UQXTNv4i16 V128:$Vn)>;
+// trunc(umin(X, 4294967295)) -> UQXTRN v2i32
+def : Pat<(v2i32 (truncusat_u (v2i64 V128:$Vn))),
+ (UQXTNv2i32 V128:$Vn)>;
// trunc(smin(smax(X, -128), 128)) -> SQXTRN
-// with reversed min/max
-def : Pat<(v8i8 (trunc (smin (smax (v8i16 V128:$Vn), (v8i16 VImm80)),
- (v8i16 VImm7F)))),
- (SQXTNv8i8 V128:$Vn)>;
-def : Pat<(v8i8 (trunc (smax (smin (v8i16 V128:$Vn), (v8i16 VImm7F)),
- (v8i16 VImm80)))),
+def : Pat<(v8i8 (truncssat_s (v8i16 V128:$Vn))),
(SQXTNv8i8 V128:$Vn)>;
// trunc(smin(smax(X, -32768), 32767)) -> SQXTRN
-// with reversed min/max
-def : Pat<(v4i16 (trunc (smin (smax (v4i32 V128:$Vn), (v4i32 VImm8000)),
- (v4i32 VImm7FFF)))),
- (SQXTNv4i16 V128:$Vn)>;
-def : Pat<(v4i16 (trunc (smax (smin (v4i32 V128:$Vn), (v4i32 VImm7FFF)),
- (v4i32 VImm8000)))),
+def : Pat<(v4i16 (truncssat_s (v4i32 V128:$Vn))),
(SQXTNv4i16 V128:$Vn)>;
-
-// concat_vectors(Vd, trunc(umin(X, 255))) -> UQXTRN(Vd, Vn)
+// trunc(smin(smax(X, -2147483648), 2147483647)) -> SQXTRN
+def : Pat<(v2i32 (truncssat_s (v2i64 V128:$Vn))),
+ (SQXTNv2i32 V128:$Vn)>;
+// trunc(umin(smax(X, 0), 255)) -> SQXTUN
+def : Pat<(v8i8 (truncssat_u (v8i16 V128:$Vn))),
+ (SQXTUNv8i8 V128:$Vn)>;
+// trunc(umin(smax(X, 0), 65535)) -> SQXTUN
+def : Pat<(v4i16 (truncssat_u (v4i32 V128:$Vn))),
+ (SQXTUNv4i16 V128:$Vn)>;
+// trunc(umin(smax(X, 0), 4294967295)) -> SQXTUN
+def : Pat<(v2i32 (truncssat_u (v2i64 V128:$Vn))),
+ (SQXTUNv2i32 V128:$Vn)>;
+
+// truncusat_u
+// concat_vectors(Vd, truncusat_u(Vn)) ~> UQXTRN(Vd, Vn)
def : Pat<(v16i8 (concat_vectors
(v8i8 V64:$Vd),
- (v8i8 (trunc (umin (v8i16 V128:$Vn), (v8i16 VImmFF)))))),
+ (v8i8 (truncusat_u (v8i16 V128:$Vn))))),
(UQXTNv16i8 (INSERT_SUBREG (IMPLICIT_DEF), V64:$Vd, dsub), V128:$Vn)>;
-// concat_vectors(Vd, trunc(umin(X, 65535))) -> UQXTRN(Vd, Vn)
def : Pat<(v8i16 (concat_vectors
(v4i16 V64:$Vd),
- (v4i16 (trunc (umin (v4i32 V128:$Vn), (v4i32 VImmFFFF)))))),
+ (v4i16 (truncusat_u (v4i32 V128:$Vn))))),
(UQXTNv8i16 (INSERT_SUBREG (IMPLICIT_DEF), V64:$Vd, dsub), V128:$Vn)>;
+def : Pat<(v4i32 (concat_vectors
+ (v2i32 V64:$Vd),
+ (v2i32 (truncusat_u (v2i64 V128:$Vn))))),
+ (UQXTNv4i32 (INSERT_SUBREG (IMPLICIT_DEF), V64:$Vd, dsub), V128:$Vn)>;
-// concat_vectors(Vd, trunc(smin(smax Vm, -128), 127) ~> SQXTN2(Vd, Vn)
-// with reversed min/max
+// concat_vectors(Vd, truncssat_s(Vn)) ~> SQXTN2(Vd, Vn)
def : Pat<(v16i8 (concat_vectors
(v8i8 V64:$Vd),
- (v8i8 (trunc (smin (smax (v8i16 V128:$Vn), (v8i16 VImm80)),
- (v8i16 VImm7F)))))),
+ (v8i8 (truncssat_s (v8i16 V128:$Vn))))),
(SQXTNv16i8 (INSERT_SUBREG (IMPLICIT_DEF), V64:$Vd, dsub), V128:$Vn)>;
-def : Pat<(v16i8 (concat_vectors
- (v8i8 V64:$Vd),
- (v8i8 (trunc (smax (smin (v8i16 V128:$Vn), (v8i16 VImm7F)),
- (v8i16 VImm80)))))),
- (SQXTNv16i8 (INSERT_SUBREG (IMPLICIT_DEF), V64:$Vd, dsub), V128:$Vn)>;
-
-// concat_vectors(Vd, trunc(smin(smax Vm, -32768), 32767) ~> SQXTN2(Vd, Vn)
-// with reversed min/max
def : Pat<(v8i16 (concat_vectors
(v4i16 V64:$Vd),
- (v4i16 (trunc (smin (smax (v4i32 V128:$Vn), (v4i32 VImm8000)),
- (v4i32 VImm7FFF)))))),
+ (v4i16 (truncssat_s (v4i32 V128:$Vn))))),
(SQXTNv8i16 (INSERT_SUBREG (IMPLICIT_DEF), V64:$Vd, dsub), V128:$Vn)>;
+def : Pat<(v4i32 (concat_vectors
+ (v2i32 V64:$Vd),
+ (v2i32 (truncssat_s (v2i64 V128:$Vn))))),
+ (SQXTNv4i32 (INSERT_SUBREG (IMPLICIT_DEF), V64:$Vd, dsub), V128:$Vn)>;
+
+// concat_vectors(Vd, truncssat_u(Vn)) ~> SQXTUN2(Vd, Vn)
+def : Pat<(v16i8 (concat_vectors
+ (v8i8 V64:$Vd),
+ (v8i8 (truncssat_u (v8i16 V128:$Vn))))),
+ (SQXTUNv16i8 (INSERT_SUBREG (IMPLICIT_DEF), V64:$Vd, dsub), V128:$Vn)>;
def : Pat<(v8i16 (concat_vectors
(v4i16 V64:$Vd),
- (v4i16 (trunc (smax (smin (v4i32 V128:$Vn), (v4i32 VImm7FFF)),
- (v4i32 VImm8000)))))),
- (SQXTNv8i16 (INSERT_SUBREG (IMPLICIT_DEF), V64:$Vd, dsub), V128:$Vn)>;
+ (v4i16 (truncssat_u (v4i32 V128:$Vn))))),
+ (SQXTUNv8i16 (INSERT_SUBREG (IMPLICIT_DEF), V64:$Vd, dsub), V128:$Vn)>;
+def : Pat<(v4i32 (concat_vectors
+ (v2i32 V64:$Vd),
+ (v2i32 (truncssat_u (v2i64 V128:$Vn))))),
+ (SQXTUNv4i32 (INSERT_SUBREG (IMPLICIT_DEF), V64:$Vd, dsub), V128:$Vn)>;
// Select BSWAP vector instructions into REV instructions
def : Pat<(v4i16 (bswap (v4i16 V64:$Rn))),
diff --git a/llvm/test/CodeGen/AArch64/fpclamptosat_vec.ll b/llvm/test/CodeGen/AArch64/fpclamptosat_vec.ll
index 0138bef9c38454..9157bcba59e9bb 100644
--- a/llvm/test/CodeGen/AArch64/fpclamptosat_vec.ll
+++ b/llvm/test/CodeGen/AArch64/fpclamptosat_vec.ll
@@ -7,12 +7,8 @@
define <2 x i32> @stest_f64i32(<2 x double> %x) {
; CHECK-LABEL: stest_f64i32:
; CHECK: // %bb.0: // %entry
-; CHECK-NEXT: mov d1, v0.d[1]
-; CHECK-NEXT: fcvtzs w8, d0
-; CHECK-NEXT: fcvtzs w9, d1
-; CHECK-NEXT: fmov s0, w8
-; CHECK-NEXT: mov v0.s[1], w9
-; CHECK-NEXT: // kill: def $d0 killed $d0 killed $q0
+; CHECK-NEXT: fcvtzs v0.2d, v0.2d
+; CHECK-NEXT: sqxtn v0.2s, v0.2d
; CHECK-NEXT: ret
entry:
%conv = fptosi <2 x double> %x to <2 x i64>
@@ -45,12 +41,8 @@ entry:
define <2 x i32> @ustest_f64i32(<2 x double> %x) {
; CHECK-LABEL: ustest_f64i32:
; CHECK: // %bb.0: // %entry
-; CHECK-NEXT: mov d1, v0.d[1]
-; CHECK-NEXT: fcvtzu w8, d0
-; CHECK-NEXT: fcvtzu w9, d1
-; CHECK-NEXT: fmov s0, w8
-; CHECK-NEXT: mov v0.s[1], w9
-; CHECK-NEXT: // kill: def $d0 killed $d0 killed $q0
+; CHECK-NEXT: fcvtzs v0.2d, v0.2d
+; CHECK-NEXT: sqxtun v0.2s, v0.2d
; CHECK-NEXT: ret
entry:
%conv = fptosi <2 x double> %x to <2 x i64>
@@ -242,8 +234,8 @@ entry:
define <4 x i16> @ustest_f32i16(<4 x float> %x) {
; CHECK-LABEL: ustest_f32i16:
; CHECK: // %bb.0: // %entry
-; CHECK-NEXT: fcvtzu v0.4s, v0.4s
-; CHECK-NEXT: uqxtn v0.4h, v0.4s
+; CHECK-NEXT: fcvtzs v0.4s, v0.4s
+; CHECK-NEXT: sqxtun v0.4h, v0.4s
; CHECK-NEXT: ret
entry:
%conv = fptosi <4 x float> %x to <4 x i32>
@@ -308,10 +300,10 @@ define <8 x i16> @ustest_f16i16(<8 x half> %x) {
; CHECK-CVT: // %bb.0: // %entry
; CHECK-CVT-NEXT: fcvtl v1.4s, v0.4h
; CHECK-CVT-NEXT: fcvtl2 v0.4s, v0.8h
-; CHECK-CVT-NEXT: fcvtzu v1.4s, v1.4s
-; CHECK-CVT-NEXT: fcvtzu v2.4s, v0.4s
-; CHECK-CVT-NEXT: uqxtn v0.4h, v1.4s
-; CHECK-CVT-NEXT: uqxtn2 v0.8h, v2.4s
+; CHECK-CVT-NEXT: fcvtzs v1.4s, v1.4s
+; CHECK-CVT-NEXT: fcvtzs v2.4s, v0.4s
+; CHECK-CVT-NEXT: sqxtun v0.4h, v1.4s
+; CHECK-CVT-NEXT: sqxtun2 v0.8h, v2.4s
; CHECK-CVT-NEXT: ret
;
; CHECK-FP16-LABEL: ustest_f16i16:
@@ -656,12 +648,8 @@ entry:
define <2 x i32> @stest_f64i32_mm(<2 x double> %x) {
; CHECK-LABEL: stest_f64i32_mm:
; CHECK: // %bb.0: // %entry
-; CHECK-NEXT: mov d1, v0.d[1]
-; CHECK-NEXT: fcvtzs w8, d0
-; CHECK-NEXT: fcvtzs w9, d1
-; CHECK-NEXT: fmov s0, w8
-; CHECK-NEXT: mov v0.s[1], w9
-; CHECK-NEXT: // kill: def $d0 killed $d0 killed $q0
+; CHECK-NEXT: fcvtzs v0.2d, v0.2d
+; CHECK-NEXT: sqxtn v0.2s, v0.2d
; CHECK-NEXT: ret
entry:
%conv = fptosi <2 x double> %x to <2 x i64>
@@ -691,12 +679,8 @@ entry:
define <2 x i32> @ustest_f64i32_mm(<2 x double> %x) {
; CHECK-LABEL: ustest_f64i32_mm:
; CHECK: // %bb.0: // %entry
-; CHECK-NEXT: mov d1, v0.d[1]
-; CHECK-NEXT: fcvtzu w8, d0
-; CHECK-NEXT: fcvtzu w9, d1
-; CHECK-NEXT: fmov s0, w8
-; CHECK-NEXT: mov v0.s[1], w9
-; CHECK-NEXT: // kill: def $d0 killed $d0 killed $q0
+; CHECK-NEXT: fcvtzs v0.2d, v0.2d
+; CHECK-NEXT: sqxtun v0.2s, v0.2d
; CHECK-NEXT: ret
entry:
%conv = fptosi <2 x double> %x to <2 x i64>
@@ -868,8 +852,8 @@ entry:
define <4 x i16> @ustest_f32i16_mm(<4 x float> %x) {
; CHECK-LABEL: ustest_f32i16_mm:
; CHECK: // %bb.0: // %entry
-; CHECK-NEXT: fcvtzu v0.4s, v0.4s
-; CHECK-NEXT: uqxtn v0.4h, v0.4s
+; CHECK-NEXT: fcvtzs v0.4s, v0.4s
+; CHECK-NEXT: sqxtun v0.4h, v0.4s
; CHECK-NEXT: ret
entry:
%conv = fptosi <4 x float> %x to <4 x i32>
@@ -929,10 +913,10 @@ define <8 x i16> @ustest_f16i16_mm(<8 x half> %x) {
; CHECK-CVT: // %bb.0: // %entry
; CHECK-CVT-NEXT: fcvtl v1.4s, v0.4h
; CHECK-CVT-NEXT: fcvtl2 v0.4s, v0.8h
-; CHECK-CVT-NEXT: fcvtzu v1.4s, v1.4s
-; CHECK-CVT-NEXT: fcvtzu v2.4s, v0.4s
-; CHECK-CVT-NEXT: uqxtn v0.4h, v1.4s
-; CHECK-CVT-NEXT: uqxtn2 v0.8h, v2.4s
+; CHECK-CVT-NEXT: fcvtzs v1.4s, v1.4s
+; CHECK-CVT-NEXT: fcvtzs v2.4s, v0.4s
+; CHECK-CVT-NEXT: sqxtun v0.4h, v1.4s
+; CHECK-CVT-NEXT: sqxtun2 v0.8h, v2.4s
; CHECK-CVT-NEXT: ret
;
; CHECK-FP16-LABEL: ustest_f16i16_mm:
diff --git a/llvm/test/CodeGen/AArch64/qmovn.ll b/llvm/test/CodeGen/AArch64/qmovn.ll
index 1eab8ab4862998..2685ea9fb5d202 100644
--- a/llvm/test/CodeGen/AArch64/qmovn.ll
+++ b/llvm/test/CodeGen/AArch64/qmovn.ll
@@ -84,15 +84,7 @@ entry:
define <2 x i32> @vqmovni64_smaxmin(<2 x i64> %s0) {
; CHECK-LABEL: vqmovni64_smaxmin:
; CHECK: // %bb.0: // %entry
-; CHECK-NEXT: mov w8, #2147483647 // =0x7fffffff
-; CHECK-NEXT: dup v1.2d, x8
-; CHECK-NEXT: mov x8, #-2147483648 // =0xffffffff80000000
-; CHECK-NEXT: cmgt v2.2d, v1.2d, v0.2d
-; CHECK-NEXT: bif v0.16b, v1.16b, v2.16b
-; CHECK-NEXT: dup v1.2d, x8
-; CHECK-NEXT: cmgt v2.2d, v0.2d, v1.2d
-; CHECK-NEXT: bif v0.16b, v1.16b, v2.16b
-; CHECK-NEXT: xtn v0.2s, v0.2d
+; CHECK-NEXT: sqxtn v0.2s, v0.2d
; CHECK-NEXT: ret
entry:
%c1 = icmp slt <2 x i64> %s0, <i64 2147483647, i64 2147483647>
@@ -106,15 +98,7 @@ entry:
define <2 x i32> @vqmovni64_sminmax(<2 x i64> %s0) {
; CHECK-LABEL: vqmovni64_sminmax:
; CHECK: // %bb.0: // %entry
-; CHECK-NEXT: mov x8, #-2147483648 // =0xffffffff80000000
-; CHECK-NEXT: dup v1.2d, x8
-; CHECK-NEXT: mov w8, #2147483647 // =0x7fffffff
-; CHECK-NEXT: cmgt v2.2d, v0.2d, v1.2d
-; CHECK-NEXT: bif v0.16b, v1.16b, v2.16b
-; CHECK-NEXT: dup v1.2d, x8
-; CHECK-NEXT: cmgt v2.2d, v1.2d, v0.2d
-; CHECK-NEXT: bif v0.16b, v1.16b, v2.16b
-; CHECK-NEXT: xtn v0.2s, v0.2d
+; CHECK-NEXT: sqxtn v0.2s, v0.2d
; CHECK-NEXT: ret
entry:
%c1 = icmp sgt <2 x i64> %s0, <i64 -2147483648, i64 -2147483648>
@@ -128,12 +112,7 @@ entry:
define <2 x i32> @vqmovni64_smaxmin_u(<2 x i64> %s0) {
; CHECK-LABEL: vqmovni64_smaxmin_u:
; CHECK: // %bb.0: // %entry
-; CHECK-NEXT: movi v1.2d, #0x000000ffffffff
-; CHECK-NEXT: cmgt v2.2d, v1.2d, v0.2d
-; CHECK-NEXT: bif v0.16b, v1.16b, v2.16b
-; CHECK-NEXT: cmgt v1.2d, v0.2d, #0
-; CHECK-NEXT: and v0.16b, v0.16b, v1.16b
-; CHECK-NEXT: xtn v0.2s, v0.2d
+; CHECK-NEXT: sqxtun v0.2s, v0.2d
; CHECK-NEXT: ret
entry:
%c1 = icmp slt <2 x i64> %s0, <i64 4294967295, i64 4294967295>
@@ -147,13 +126,7 @@ entry:
define <2 x i32> @vqmovni64_sminmax_u(<2 x i64> %s0) {
; CHECK-LABEL: vqmovni64_sminmax_u:
; CHECK: // %bb.0: // %entry
-; CHECK-NEXT: cmgt v1.2d, v0.2d, #0
-; CHECK-NEXT: movi v2.2d, #0x000000ffffffff
-; CHECK-NEXT: and v0.16b, v0.16b, v1.16b
-; CHECK-NEXT: cmgt v1.2d, v2.2d, v0.2d
-; CHECK-NEXT: and v0.16b, v0.16b, v1.16b
-; CHECK-NEXT: orn v0.16b, v0.16b, v1.16b
-; CHECK-NEXT: xtn v0.2s, v0.2d
+; CHECK-NEXT: sqxtun v0.2s, v0.2d
; CHECK-NEXT: ret
entry:
%c1 = icmp sgt <2 x i64> %s0, zeroinitializer
@@ -167,11 +140,7 @@ entry:
define <4 x i16> @vqmovni32_smaxmin_u(<4 x i32> %s0) {
; CHECK-LABEL: vqmovni32_smaxmin_u:
; CHECK: // %bb.0: // %entry
-; CHECK-NEXT: movi v1.2d, #0x00ffff0000ffff
-; CHECK-NEXT: smin v0.4s, v0.4s, v1.4s
-; CHECK-NEXT: movi v1.2d, #0000000000000000
-; CHECK-NEXT: smax v0.4s, v0.4s, v1.4s
-; CHECK-NEXT: xtn v0.4h, v0.4s
+; CHECK-NEXT: sqxtun v0.4h, v0.4s
; CHECK-NEXT: ret
entry:
%c1 = icmp slt <4 x i32> %s0, <i32 65535, i32 65535, i32 65535, i32 65535>
@@ -185,11 +154,7 @@ entry:
define <4 x i16> @vqmovni32_sminmax_u(<4 x i32> %s0) {
; CHECK-LABEL: vqmovni32_sminmax_u:
; CHECK: // %bb.0: // %entry
-; CHECK-NEXT: movi v1.2d, #0000000000000000
-; CHECK-NEXT: smax v0.4s, v0.4s, v1.4s
-; CHECK-NEXT: movi v1.2d, #0x00ffff0000ffff
-; CHECK-NEXT: smin v0.4s, v0.4s, v1.4s
-; CHECK-NEXT: xtn v0.4h, v0.4s
+; CHECK-NEXT: sqxtun v0.4h, v0.4s
; CHECK-NEXT: ret
entry:
%c1 = icmp sgt <4 x i32> %s0, zeroinitializer
@@ -203,11 +168,7 @@ entry:
define <8 x i8> @vqmovni16_smaxmin_u(<8 x i16> %s0) {
; CHECK-LABEL: vqmovni16_smaxmin_u:
; CHECK: // %bb.0: // %entry
-; CHECK-NEXT: movi v1.2d, #0xff00ff00ff00ff
-; CHECK-NEXT: movi v2.2d, #0000000000000000
-; CHECK-NEXT: smin v0.8h, v0.8h, v1.8h
-; CHECK-NEXT: smax v0.8h, v0.8h, v2.8h
-; CHECK-NEXT: xtn v0.8b, v0.8h
+; CHECK-NEXT: sqxtun v0.8b, v0.8h
; CHECK-NEXT: ret
entry:
%c1 = icmp slt <8 x i16> %s0, <i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255, i16 255>
@@ -221,11 +182,7 @@ entry:
define <8 x i8> @vqmovni16_sminmax_u(<8 x i16> %s0) {
; CHECK-LABEL: vqmovni16_sminmax_u:
; CHECK: // %bb.0: // %entry
-; CHECK-NEXT: movi v1.2d, #0000000000000000
-; CHECK-NEXT: movi v2.2d, #0xff00ff00ff00ff
-; CHECK-NEXT: smax v0.8h, v0.8h, v1.8h
-; CHECK-NEXT: smin v0.8h, v0.8h, v2.8h
-; CHECK-NEXT: xtn v0.8b, v0.8h
+; CHECK-NEXT: sqxtun v0.8b, v0.8h
; CHECK-NEXT: ret
entry:
%c1 = icmp sgt <8 x i16> %s0, zeroinitializer
@@ -239,11 +196,7 @@ entry:
define <2 x i32> @vqmovni64_umaxmin(<2 x i64> %s0) {
; CHECK-LABEL: vqmovni64_umaxmin:
; CHECK: // %bb.0: // %entry
-; CHECK-NEXT: movi v1.2d, #0x000000ffffffff
-; CHECK-NEXT: cmhi v1.2d, v1.2d, v0.2d
-; CHECK-NEXT: and v0.16b, v0.16b, v1.16b
-; CHECK-NEXT: orn v0.16b, v0.16b, v1.16b
-; CHECK-NEXT: xtn v0.2s, v0.2d
+; CHECK-NEXT: uqxtn v0.2s, v0.2d
; CHECK-NEXT: ret
entry:
%c1 = icmp ult <2 x i64> %s0, <i64 4294967295, i64 4294967295>
@@ -285,16 +238,8 @@ entry:
define <4 x i32> @signed_minmax_v2i64_to_v4i32(<2 x i32> %x, <2 x i64> %y) {
; CHECK-LABEL: signed_minmax_v2i64_to_v4i32:
; CHECK: // %bb.0: // %entry
-; CHECK-NEXT: mov w8, #2147483647 // =0x7fffffff
; CHECK-NEXT: // kill: def $d0 killed $d0 def $q0
-; CHECK-NEXT: dup v2.2d, x8
-; CHECK-NEXT: mov x8, #-2147483648 // =0xffffffff80000000
-; CHECK-NEXT: cmgt v3.2d, v2.2d, v1.2d
-; CHECK-NEXT: bif v1.16b, v2.16b, v3.16b
-; CHECK-NEXT: dup v2.2d, x8
-; CHECK-NEXT: cmgt v3.2d, v1.2d, v2.2d
-; CHECK-NEXT: bif v1.16b, v2.16b, v3.16b
-; CHECK-NEXT: xtn2 v0.4s, v1.2d
+; CHECK-NEXT: sqxtn2 v0.4s, v1.2d
; CHECK-NEXT: ret
entry:
%min = call <2 x i64> @llvm.smin.v2i64(<2 x i64> %y, <2 x i64> <i64 2147483647, i64 2147483647>)
@@ -337,16 +282,8 @@ entry:
define <4 x i32> @signed_maxmin_v2i64_to_v4i32(<2 x i32> %x, <2 x i64> %y) {
; CHECK-LABEL: signed_maxmin_v2i64_to_v4i32:
; CHECK: // %bb.0: // %entry
-; CHECK-NEXT: mov x8, #-2147483648 // =0xffffffff80000000
; CHECK-NEXT: // kill: def $d0 killed $d0 def $q0
-; CHECK-NEXT: dup v2.2d, x8
-; CHECK-NEXT: mov w8, #2147483647 // =0x7fffffff
-; CHECK-NEXT: cmgt v3.2d, v1.2d, v2.2d
-; CHECK-NEXT: bif v1.16b, v2.16b, v3.16b
-; CHECK-NEXT: dup v2.2d, x8
-; CHECK-NEXT: cmgt v3.2d, v2.2d, v1.2d
-; CHECK-NEXT: bif v1.16b, v2.16b, v3.16b
-; CHECK-NEXT: xtn2 v0.4s, v1.2d
+; CHECK-NEXT: sqxtn2 v0.4s, v1.2d
; CHECK-NEXT: ret
entry:
%max = call <2 x i64> @llvm.smax.v2i64(<2 x i64> %y, <2 x i64> <i64 -2147483648, i64 -2147483648>)
@@ -387,12 +324,8 @@ entry:
define <4 x i32> @unsigned_v2i64_to_v4i32(<2 x i32> %x, <2 x i64> %y) {
; CHECK-LABEL: unsigned_v2i64_to_v4i32:
; CHECK: // %bb.0: // %entry
-; CHECK-NEXT: movi v2.2d, #0x000000ffffffff
; CHECK-NEXT: // kill: def $d0 killed $d0 def $q0
-; CHECK-NEXT: cmhi v2.2d, v2.2d, v1.2d
-; CHECK-NEXT: and v1.16b, v1.16b, v2.16b
-; CHECK-NEXT: orn v1.16b, v1.16b, v2.16b
-; CHECK-NEXT: xtn2 v0.4s, v1.2d
+; CHECK-NEXT: uqxtn2 v0.4s, v1.2d
; CHECK-NEXT: ret
entry:
%min = call <2 x i64> @llvm.umin.v2i64(<2 x i64> %y, <2 x i64> <i64 4294967295, i64 4294967295>)
@@ -406,12 +339,8 @@ entry:
define <16 x i8> @us_maxmin_v8i16_to_v16i8(<8 x i8> %x, <8 x i16> %y) {
; CHECK-LABEL: us_maxmin_v8i16_to_v16i8:
; CHECK: // %bb.0: // %entry
-; CHECK-NEXT: movi v2.2d, #0000000000000000
-; CHECK-NEXT: movi v3.2d, #0xff00ff00ff00ff
; CHECK-NEXT: // kill: def $d0 killed $d0 def $q0
-; CHECK-NEXT: smax v1.8h, v1.8h, v2.8h
-; CHECK-NEXT: smin v1.8h, v1.8h, v3.8h
-; CHECK-NEXT: xtn2 v0.16b, v1.8h
+; CHECK-NEXT: sqxtun2 v0.16b, v1.8h
; CHECK-NEXT: ret
entry:
%max = call <8 x i16> @llvm.smax.v8i16(<8 x i16> %y, <8 x i16> zeroinitializer)
@@ -424,12 +353,8 @@ entry:
define <8 x i16> @us_maxmin_v4i32_to_v8i16(<4 x i16> %x, <4 x i32> %y) {
; CHECK-LABEL: us_maxmin_v4i32_to_v8i16:
; CHECK: // %bb.0: // %entry
-; CHECK-NEXT: movi v2.2d, #0000000000000000
; CHECK-NEXT: // kill: def $d0 killed $d0 def $q0
-; CHECK-NEXT: smax v1.4s, v1.4s, v2.4s
-; CHECK-NEXT: movi v2.2d, #0x00ffff0000ffff
-; CHECK-NEXT: smin v1.4s, v1.4s, v2.4s
-; CHECK-NEXT: xtn2 v0.8h, v1.4s
+; CHECK-NEXT: sqxtun2 v0.8h, v1.4s
; CHECK-NEXT: ret
entry:
%max = call <4 x i32> @llvm.smax.v4i32(<4 x i32> %y, <4 x i32> zeroinitializer)
@@ -442,14 +367,8 @@ entry:
define <4 x i32> @us_maxmin_v2i64_to_v4i32(<2 x i32> %x, <2 x i64> %y) {
; CHECK-LABEL: us_maxmin_v2i64_to_v4i32:
; CHECK: // %bb.0: // %entry
-; CHECK-NEXT: cmgt v2.2d, v1.2d, #0
-; CHECK-NEXT: movi v3.2d, #0x000000ffffffff
; CHECK-NEXT: // kill: def $d0 killed $d0 def $q0
-; CHECK-NEXT: and v1.16b, v1.16b, v2.16b
-; CHECK-NEXT: cmgt v2.2d, v3.2d, v1.2d
-; CHECK-NEXT: and v1.16b, v1.16b, v2.16b
-; CHECK-NEXT: orn v1.16b, v1.16b, v2.16b
-; CHECK-NEXT: xtn2 v0.4s, v1.2d
+; CHECK-NEXT: sqxtun2 v0.4s, v1.2d
; CHECK-NEXT: ret
entry:
%max = call <2 x i64> @llvm.smax.v2i64(<2 x i64> %y, <2 x i64> zeroinitializer)
@@ -464,12 +383,8 @@ entry:
define <16 x i8> @sminsmax_range_unsigned_i16_to_i8(<8 x i8> %x, <8 x i16> %y) {
; CHECK-LABEL: sminsmax_range_unsigned_i16_to_i8:
; CHECK: // %bb.0: // %entry
-; CHECK-NEXT: movi v2.2d, #0000000000000000
-; CHECK-NEXT: movi v3.2d, #0xff00ff00ff00ff
; CHECK-NEXT: // kill: def $d0 killed $d0 def $q0
-; CHECK-NEXT: smax v1.8h, v1.8h, v2.8h
-; CHECK-NEXT: smin v1.8h, v1.8h, v3.8h
-; CHECK-NEXT: xtn2 v0.16b, v1.8h
+; CHECK-NEXT: sqxtun2 v0.16b, v1.8h
; CHECK-NEXT: ret
entry:
%min = call <8 x i16> @llvm.smax.v8i16(<8 x i16> %y, <8 x i16> zeroinitializer)
@@ -482,12 +397,8 @@ entry:
define <8 x i16> @sminsmax_range_unsigned_i32_to_i16(<4 x i16> %x, <4 x i32> %y) {
; CHECK-LABEL: sminsmax_range_unsigned_i32_to_i16:
; CHECK: // %bb.0: // %entry
-; CHECK-NEXT: movi v2.2d, #0000000000000000
; CHECK-NEXT: // kill: def $d0 killed $d0 def $q0
-; CHECK-NEXT: smax v1.4s, v1.4s, v2.4s
-; CHECK-NEXT: movi v2.2d, #0x00ffff0000ffff
-; CHECK-NEXT: smin v1.4s, v1.4s, v2.4s
-; CHECK-NEXT: xtn2 v0.8h, v1.4s
+; CHECK-NEXT: sqxtun2 v0.8h, v1.4s
; CHECK-NEXT: ret
entry:
%smax = call <4 x i32> @llvm.smax.v4i32(<4 x i32> %y, <4 x i32> zeroinitializer)
@@ -500,14 +411,8 @@ entry:
define <4 x i32> @sminsmax_range_unsigned_i64_to_i32(<2 x i32> %x, <2 x i64> %y) {
; CHECK-LABEL: sminsmax_range_unsigned_i64_to_i32:
; CHECK: // %bb.0: // %entry
-; CHECK-NEXT: cmgt v2.2d, v1.2d, #0
-; CHECK-NEXT: movi v3.2d, #0x000000ffffffff
; CHECK-NEXT: // kill: def $d0 killed $d0 def $q0
-; CHECK-NEXT: and v1.16b, v1.16b, v2.16b
-; CHECK-NEXT: cmgt v2.2d, v3.2d, v1.2d
-; CHECK-NEXT: and v1.16b, v1.16b, v2.16b
-; CHECK-NEXT: orn v1.16b, v1.16b, v2.16b
-; CHECK-NEXT: xtn2 v0.4s, v1.2d
+; CHECK-NEXT: sqxtun2 v0.4s, v1.2d
; CHECK-NEXT: ret
entry:
%smax = call <2 x i64> @llvm.smax.v2i64(<2 x i64> %y, <2 x i64> zeroinitializer)
>From 7db4887dbbdc3b05d2e7ef854a3990164af6a2c4 Mon Sep 17 00:00:00 2001
From: hanbeom <kese111 at gmail.com>
Date: Tue, 16 Jul 2024 14:14:40 +0900
Subject: [PATCH 6/6] [RISCV] Support saturated truncate
Add support for saturated truncate by implementing the following changes:
- Add `TRUNCATE_[SU]SAT_[SU]` to the Action target of `TRUNCATE`
- Add `TRUNCATE_[SU]SAT_[SU]` to the TargetLowering target of `TRUNCATE`
- Convert `TRUNCATE_SSAT_S` to `TRUNCATE_VECTOR_VL_SSAT`
- Convert `TRUNCATE_[SU]SAT_U` to `TRUNCATE_VECTOR_VL_USAT`
---
llvm/lib/Target/RISCV/RISCVISelLowering.cpp | 25 +++++--
.../RISCV/rvv/fixed-vectors-trunc-sat-clip.ll | 32 +++------
.../CodeGen/RISCV/rvv/fpclamptosat_vec.ll | 68 +++++++------------
.../RISCV/rvv/trunc-sat-clip-sdnode.ll | 32 +++------
.../RISCV/rvv/trunc-select-to-max-usat.ll | 57 +++++-----------
5 files changed, 80 insertions(+), 134 deletions(-)
diff --git a/llvm/lib/Target/RISCV/RISCVISelLowering.cpp b/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
index d40d4997d76149..704caeab90bb6e 100644
--- a/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
+++ b/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
@@ -853,7 +853,9 @@ RISCVTargetLowering::RISCVTargetLowering(const TargetMachine &TM,
// Integer VTs are lowered as a series of "RISCVISD::TRUNCATE_VECTOR_VL"
// nodes which truncate by one power of two at a time.
- setOperationAction(ISD::TRUNCATE, VT, Custom);
+ setOperationAction({ISD::TRUNCATE, ISD::TRUNCATE_SSAT_S,
+ ISD::TRUNCATE_SSAT_U, ISD::TRUNCATE_USAT_U},
+ VT, Custom);
// Custom-lower insert/extract operations to simplify patterns.
setOperationAction({ISD::INSERT_VECTOR_ELT, ISD::EXTRACT_VECTOR_ELT}, VT,
@@ -1168,7 +1170,9 @@ RISCVTargetLowering::RISCVTargetLowering(const TargetMachine &TM,
setOperationAction(ISD::SELECT, VT, Custom);
- setOperationAction(ISD::TRUNCATE, VT, Custom);
+ setOperationAction({ISD::TRUNCATE, ISD::TRUNCATE_SSAT_S,
+ ISD::TRUNCATE_SSAT_U, ISD::TRUNCATE_USAT_U},
+ VT, Custom);
setOperationAction(ISD::BITCAST, VT, Custom);
@@ -6395,6 +6399,9 @@ SDValue RISCVTargetLowering::LowerOperation(SDValue Op,
return DAG.getNode(RISCVISD::BREV8, DL, VT, BSwap);
}
case ISD::TRUNCATE:
+ case ISD::TRUNCATE_SSAT_S:
+ case ISD::TRUNCATE_SSAT_U:
+ case ISD::TRUNCATE_USAT_U:
// Only custom-lower vector truncates
if (!Op.getSimpleValueType().isVector())
return Op;
@@ -8234,7 +8241,8 @@ SDValue RISCVTargetLowering::lowerVectorMaskTruncLike(SDValue Op,
SDValue RISCVTargetLowering::lowerVectorTruncLike(SDValue Op,
SelectionDAG &DAG) const {
- bool IsVPTrunc = Op.getOpcode() == ISD::VP_TRUNCATE;
+ unsigned Opc = Op.getOpcode();
+ bool IsVPTrunc = Opc == ISD::VP_TRUNCATE;
SDLoc DL(Op);
MVT VT = Op.getSimpleValueType();
@@ -8279,11 +8287,18 @@ SDValue RISCVTargetLowering::lowerVectorTruncLike(SDValue Op,
getDefaultVLOps(SrcVT, ContainerVT, DL, DAG, Subtarget);
}
+ unsigned NewOpc;
+ if (Opc == ISD::TRUNCATE_SSAT_S)
+ NewOpc = RISCVISD::TRUNCATE_VECTOR_VL_SSAT;
+ else if (Opc == ISD::TRUNCATE_SSAT_U || Opc == ISD::TRUNCATE_USAT_U)
+ NewOpc = RISCVISD::TRUNCATE_VECTOR_VL_USAT;
+ else
+ NewOpc = RISCVISD::TRUNCATE_VECTOR_VL;
+
do {
SrcEltVT = MVT::getIntegerVT(SrcEltVT.getSizeInBits() / 2);
MVT ResultVT = ContainerVT.changeVectorElementType(SrcEltVT);
- Result = DAG.getNode(RISCVISD::TRUNCATE_VECTOR_VL, DL, ResultVT, Result,
- Mask, VL);
+ Result = DAG.getNode(NewOpc, DL, ResultVT, Result, Mask, VL);
} while (SrcEltVT != DstEltVT);
if (SrcVT.isFixedLengthVector())
diff --git a/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-trunc-sat-clip.ll b/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-trunc-sat-clip.ll
index 4e367bb0d70cd1..e2f540e991fd0b 100644
--- a/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-trunc-sat-clip.ll
+++ b/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-trunc-sat-clip.ll
@@ -101,10 +101,8 @@ define void @trunc_sat_u8u16_notopt(ptr %x, ptr %y) {
define void @trunc_sat_u8u16_maxmin(ptr %x, ptr %y) {
; CHECK-LABEL: trunc_sat_u8u16_maxmin:
; CHECK: # %bb.0:
-; CHECK-NEXT: vsetivli zero, 4, e16, mf2, ta, ma
+; CHECK-NEXT: vsetivli zero, 4, e8, mf4, ta, ma
; CHECK-NEXT: vle16.v v8, (a0)
-; CHECK-NEXT: vmax.vx v8, v8, zero
-; CHECK-NEXT: vsetvli zero, zero, e8, mf4, ta, ma
; CHECK-NEXT: vnclipu.wi v8, v8, 0
; CHECK-NEXT: vse8.v v8, (a1)
; CHECK-NEXT: ret
@@ -119,10 +117,8 @@ define void @trunc_sat_u8u16_maxmin(ptr %x, ptr %y) {
define void @trunc_sat_u8u16_minmax(ptr %x, ptr %y) {
; CHECK-LABEL: trunc_sat_u8u16_minmax:
; CHECK: # %bb.0:
-; CHECK-NEXT: vsetivli zero, 4, e16, mf2, ta, ma
+; CHECK-NEXT: vsetivli zero, 4, e8, mf4, ta, ma
; CHECK-NEXT: vle16.v v8, (a0)
-; CHECK-NEXT: vmax.vx v8, v8, zero
-; CHECK-NEXT: vsetvli zero, zero, e8, mf4, ta, ma
; CHECK-NEXT: vnclipu.wi v8, v8, 0
; CHECK-NEXT: vse8.v v8, (a1)
; CHECK-NEXT: ret
@@ -356,10 +352,8 @@ define void @trunc_sat_u32u64_min(ptr %x, ptr %y) {
define void @trunc_sat_u32u64_maxmin(ptr %x, ptr %y) {
; CHECK-LABEL: trunc_sat_u32u64_maxmin:
; CHECK: # %bb.0:
-; CHECK-NEXT: vsetivli zero, 4, e64, m2, ta, ma
+; CHECK-NEXT: vsetivli zero, 4, e32, m1, ta, ma
; CHECK-NEXT: vle64.v v8, (a0)
-; CHECK-NEXT: vmax.vx v8, v8, zero
-; CHECK-NEXT: vsetvli zero, zero, e32, m1, ta, ma
; CHECK-NEXT: vnclipu.wi v10, v8, 0
; CHECK-NEXT: vse32.v v10, (a1)
; CHECK-NEXT: ret
@@ -374,10 +368,8 @@ define void @trunc_sat_u32u64_maxmin(ptr %x, ptr %y) {
define void @trunc_sat_u32u64_minmax(ptr %x, ptr %y) {
; CHECK-LABEL: trunc_sat_u32u64_minmax:
; CHECK: # %bb.0:
-; CHECK-NEXT: vsetivli zero, 4, e64, m2, ta, ma
+; CHECK-NEXT: vsetivli zero, 4, e32, m1, ta, ma
; CHECK-NEXT: vle64.v v8, (a0)
-; CHECK-NEXT: vmax.vx v8, v8, zero
-; CHECK-NEXT: vsetvli zero, zero, e32, m1, ta, ma
; CHECK-NEXT: vnclipu.wi v10, v8, 0
; CHECK-NEXT: vse32.v v10, (a1)
; CHECK-NEXT: ret
@@ -445,10 +437,8 @@ define void @trunc_sat_u8u32_min(ptr %x, ptr %y) {
define void @trunc_sat_u8u32_maxmin(ptr %x, ptr %y) {
; CHECK-LABEL: trunc_sat_u8u32_maxmin:
; CHECK: # %bb.0:
-; CHECK-NEXT: vsetivli zero, 4, e32, m1, ta, ma
+; CHECK-NEXT: vsetivli zero, 4, e16, mf2, ta, ma
; CHECK-NEXT: vle32.v v8, (a0)
-; CHECK-NEXT: vmax.vx v8, v8, zero
-; CHECK-NEXT: vsetvli zero, zero, e16, mf2, ta, ma
; CHECK-NEXT: vnclipu.wi v8, v8, 0
; CHECK-NEXT: vsetvli zero, zero, e8, mf4, ta, ma
; CHECK-NEXT: vnclipu.wi v8, v8, 0
@@ -465,10 +455,8 @@ define void @trunc_sat_u8u32_maxmin(ptr %x, ptr %y) {
define void @trunc_sat_u8u32_minmax(ptr %x, ptr %y) {
; CHECK-LABEL: trunc_sat_u8u32_minmax:
; CHECK: # %bb.0:
-; CHECK-NEXT: vsetivli zero, 4, e32, m1, ta, ma
+; CHECK-NEXT: vsetivli zero, 4, e16, mf2, ta, ma
; CHECK-NEXT: vle32.v v8, (a0)
-; CHECK-NEXT: vmax.vx v8, v8, zero
-; CHECK-NEXT: vsetvli zero, zero, e16, mf2, ta, ma
; CHECK-NEXT: vnclipu.wi v8, v8, 0
; CHECK-NEXT: vsetvli zero, zero, e8, mf4, ta, ma
; CHECK-NEXT: vnclipu.wi v8, v8, 0
@@ -544,10 +532,8 @@ define void @trunc_sat_u8u64_min(ptr %x, ptr %y) {
define void @trunc_sat_u8u64_maxmin(ptr %x, ptr %y) {
; CHECK-LABEL: trunc_sat_u8u64_maxmin:
; CHECK: # %bb.0:
-; CHECK-NEXT: vsetivli zero, 4, e64, m2, ta, ma
+; CHECK-NEXT: vsetivli zero, 4, e32, m1, ta, ma
; CHECK-NEXT: vle64.v v8, (a0)
-; CHECK-NEXT: vmax.vx v8, v8, zero
-; CHECK-NEXT: vsetvli zero, zero, e32, m1, ta, ma
; CHECK-NEXT: vnclipu.wi v10, v8, 0
; CHECK-NEXT: vsetvli zero, zero, e16, mf2, ta, ma
; CHECK-NEXT: vnclipu.wi v8, v10, 0
@@ -566,10 +552,8 @@ define void @trunc_sat_u8u64_maxmin(ptr %x, ptr %y) {
define void @trunc_sat_u8u64_minmax(ptr %x, ptr %y) {
; CHECK-LABEL: trunc_sat_u8u64_minmax:
; CHECK: # %bb.0:
-; CHECK-NEXT: vsetivli zero, 4, e64, m2, ta, ma
+; CHECK-NEXT: vsetivli zero, 4, e32, m1, ta, ma
; CHECK-NEXT: vle64.v v8, (a0)
-; CHECK-NEXT: vmax.vx v8, v8, zero
-; CHECK-NEXT: vsetvli zero, zero, e32, m1, ta, ma
; CHECK-NEXT: vnclipu.wi v10, v8, 0
; CHECK-NEXT: vsetvli zero, zero, e16, mf2, ta, ma
; CHECK-NEXT: vnclipu.wi v8, v10, 0
diff --git a/llvm/test/CodeGen/RISCV/rvv/fpclamptosat_vec.ll b/llvm/test/CodeGen/RISCV/rvv/fpclamptosat_vec.ll
index 3e2db3fa4685dd..ffbcebf621fd7c 100644
--- a/llvm/test/CodeGen/RISCV/rvv/fpclamptosat_vec.ll
+++ b/llvm/test/CodeGen/RISCV/rvv/fpclamptosat_vec.ll
@@ -113,7 +113,6 @@ define <2 x i32> @ustest_f64i32(<2 x double> %x) {
; CHECK-V: # %bb.0: # %entry
; CHECK-V-NEXT: vsetivli zero, 2, e64, m1, ta, ma
; CHECK-V-NEXT: vfcvt.rtz.x.f.v v8, v8
-; CHECK-V-NEXT: vmax.vx v8, v8, zero
; CHECK-V-NEXT: vsetvli zero, zero, e32, mf2, ta, ma
; CHECK-V-NEXT: vnclipu.wi v8, v8, 0
; CHECK-V-NEXT: ret
@@ -304,9 +303,6 @@ define <4 x i32> @ustest_f32i32(<4 x float> %x) {
; CHECK-V: # %bb.0: # %entry
; CHECK-V-NEXT: vsetivli zero, 4, e32, m1, ta, ma
; CHECK-V-NEXT: vfwcvt.rtz.x.f.v v10, v8
-; CHECK-V-NEXT: vsetvli zero, zero, e64, m2, ta, ma
-; CHECK-V-NEXT: vmax.vx v10, v10, zero
-; CHECK-V-NEXT: vsetvli zero, zero, e32, m1, ta, ma
; CHECK-V-NEXT: vnclipu.wi v8, v10, 0
; CHECK-V-NEXT: ret
entry:
@@ -801,17 +797,16 @@ define <4 x i32> @ustest_f16i32(<4 x half> %x) {
; CHECK-V-NEXT: call __extendhfsf2
; CHECK-V-NEXT: fcvt.l.s a0, fa0, rtz
; CHECK-V-NEXT: vsetivli zero, 2, e64, m1, ta, ma
-; CHECK-V-NEXT: vmv.s.x v8, a0
+; CHECK-V-NEXT: vmv.s.x v10, a0
; CHECK-V-NEXT: addi a0, sp, 16
-; CHECK-V-NEXT: vl1r.v v9, (a0) # Unknown-size Folded Reload
-; CHECK-V-NEXT: vslideup.vi v8, v9, 1
+; CHECK-V-NEXT: vl1r.v v8, (a0) # Unknown-size Folded Reload
+; CHECK-V-NEXT: vslideup.vi v10, v8, 1
; CHECK-V-NEXT: csrr a0, vlenb
; CHECK-V-NEXT: add a0, sp, a0
; CHECK-V-NEXT: addi a0, a0, 16
-; CHECK-V-NEXT: vl2r.v v10, (a0) # Unknown-size Folded Reload
+; CHECK-V-NEXT: vl2r.v v8, (a0) # Unknown-size Folded Reload
; CHECK-V-NEXT: vsetivli zero, 4, e64, m2, ta, ma
-; CHECK-V-NEXT: vslideup.vi v8, v10, 2
-; CHECK-V-NEXT: vmax.vx v10, v8, zero
+; CHECK-V-NEXT: vslideup.vi v10, v8, 2
; CHECK-V-NEXT: vsetvli zero, zero, e32, m1, ta, ma
; CHECK-V-NEXT: vnclipu.wi v8, v10, 0
; CHECK-V-NEXT: csrr a0, vlenb
@@ -944,9 +939,8 @@ define <2 x i16> @ustest_f64i16(<2 x double> %x) {
; CHECK-V: # %bb.0: # %entry
; CHECK-V-NEXT: vsetivli zero, 2, e32, mf2, ta, ma
; CHECK-V-NEXT: vfncvt.rtz.x.f.w v9, v8
-; CHECK-V-NEXT: vmax.vx v8, v9, zero
; CHECK-V-NEXT: vsetvli zero, zero, e16, mf4, ta, ma
-; CHECK-V-NEXT: vnclipu.wi v8, v8, 0
+; CHECK-V-NEXT: vnclipu.wi v8, v9, 0
; CHECK-V-NEXT: ret
entry:
%conv = fptosi <2 x double> %x to <2 x i32>
@@ -1139,7 +1133,6 @@ define <4 x i16> @ustest_f32i16(<4 x float> %x) {
; CHECK-V: # %bb.0: # %entry
; CHECK-V-NEXT: vsetivli zero, 4, e32, m1, ta, ma
; CHECK-V-NEXT: vfcvt.rtz.x.f.v v8, v8
-; CHECK-V-NEXT: vmax.vx v8, v8, zero
; CHECK-V-NEXT: vsetvli zero, zero, e16, mf2, ta, ma
; CHECK-V-NEXT: vnclipu.wi v8, v8, 0
; CHECK-V-NEXT: ret
@@ -2114,24 +2107,23 @@ define <8 x i16> @ustest_f16i16(<8 x half> %x) {
; CHECK-V-NEXT: call __extendhfsf2
; CHECK-V-NEXT: fcvt.l.s a0, fa0, rtz
; CHECK-V-NEXT: vsetivli zero, 2, e32, mf2, ta, ma
-; CHECK-V-NEXT: vmv.s.x v8, a0
+; CHECK-V-NEXT: vmv.s.x v10, a0
; CHECK-V-NEXT: addi a0, sp, 16
-; CHECK-V-NEXT: vl1r.v v9, (a0) # Unknown-size Folded Reload
-; CHECK-V-NEXT: vslideup.vi v8, v9, 1
+; CHECK-V-NEXT: vl1r.v v8, (a0) # Unknown-size Folded Reload
+; CHECK-V-NEXT: vslideup.vi v10, v8, 1
; CHECK-V-NEXT: csrr a0, vlenb
; CHECK-V-NEXT: add a0, sp, a0
; CHECK-V-NEXT: addi a0, a0, 16
-; CHECK-V-NEXT: vl1r.v v9, (a0) # Unknown-size Folded Reload
+; CHECK-V-NEXT: vl1r.v v8, (a0) # Unknown-size Folded Reload
; CHECK-V-NEXT: vsetivli zero, 4, e32, m1, ta, ma
-; CHECK-V-NEXT: vslideup.vi v8, v9, 2
+; CHECK-V-NEXT: vslideup.vi v10, v8, 2
; CHECK-V-NEXT: csrr a0, vlenb
; CHECK-V-NEXT: slli a0, a0, 1
; CHECK-V-NEXT: add a0, sp, a0
; CHECK-V-NEXT: addi a0, a0, 16
-; CHECK-V-NEXT: vl2r.v v10, (a0) # Unknown-size Folded Reload
+; CHECK-V-NEXT: vl2r.v v8, (a0) # Unknown-size Folded Reload
; CHECK-V-NEXT: vsetivli zero, 8, e32, m2, ta, ma
-; CHECK-V-NEXT: vslideup.vi v8, v10, 4
-; CHECK-V-NEXT: vmax.vx v10, v8, zero
+; CHECK-V-NEXT: vslideup.vi v10, v8, 4
; CHECK-V-NEXT: vsetvli zero, zero, e16, m1, ta, ma
; CHECK-V-NEXT: vnclipu.wi v8, v10, 0
; CHECK-V-NEXT: csrr a0, vlenb
@@ -3473,7 +3465,6 @@ define <2 x i32> @ustest_f64i32_mm(<2 x double> %x) {
; CHECK-V: # %bb.0: # %entry
; CHECK-V-NEXT: vsetivli zero, 2, e64, m1, ta, ma
; CHECK-V-NEXT: vfcvt.rtz.x.f.v v8, v8
-; CHECK-V-NEXT: vmax.vx v8, v8, zero
; CHECK-V-NEXT: vsetvli zero, zero, e32, mf2, ta, ma
; CHECK-V-NEXT: vnclipu.wi v8, v8, 0
; CHECK-V-NEXT: ret
@@ -3659,9 +3650,6 @@ define <4 x i32> @ustest_f32i32_mm(<4 x float> %x) {
; CHECK-V: # %bb.0: # %entry
; CHECK-V-NEXT: vsetivli zero, 4, e32, m1, ta, ma
; CHECK-V-NEXT: vfwcvt.rtz.x.f.v v10, v8
-; CHECK-V-NEXT: vsetvli zero, zero, e64, m2, ta, ma
-; CHECK-V-NEXT: vmax.vx v10, v10, zero
-; CHECK-V-NEXT: vsetvli zero, zero, e32, m1, ta, ma
; CHECK-V-NEXT: vnclipu.wi v8, v10, 0
; CHECK-V-NEXT: ret
entry:
@@ -4151,17 +4139,16 @@ define <4 x i32> @ustest_f16i32_mm(<4 x half> %x) {
; CHECK-V-NEXT: call __extendhfsf2
; CHECK-V-NEXT: fcvt.l.s a0, fa0, rtz
; CHECK-V-NEXT: vsetivli zero, 2, e64, m1, ta, ma
-; CHECK-V-NEXT: vmv.s.x v8, a0
+; CHECK-V-NEXT: vmv.s.x v10, a0
; CHECK-V-NEXT: addi a0, sp, 16
-; CHECK-V-NEXT: vl1r.v v9, (a0) # Unknown-size Folded Reload
-; CHECK-V-NEXT: vslideup.vi v8, v9, 1
+; CHECK-V-NEXT: vl1r.v v8, (a0) # Unknown-size Folded Reload
+; CHECK-V-NEXT: vslideup.vi v10, v8, 1
; CHECK-V-NEXT: csrr a0, vlenb
; CHECK-V-NEXT: add a0, sp, a0
; CHECK-V-NEXT: addi a0, a0, 16
-; CHECK-V-NEXT: vl2r.v v10, (a0) # Unknown-size Folded Reload
+; CHECK-V-NEXT: vl2r.v v8, (a0) # Unknown-size Folded Reload
; CHECK-V-NEXT: vsetivli zero, 4, e64, m2, ta, ma
-; CHECK-V-NEXT: vslideup.vi v8, v10, 2
-; CHECK-V-NEXT: vmax.vx v10, v8, zero
+; CHECK-V-NEXT: vslideup.vi v10, v8, 2
; CHECK-V-NEXT: vsetvli zero, zero, e32, m1, ta, ma
; CHECK-V-NEXT: vnclipu.wi v8, v10, 0
; CHECK-V-NEXT: csrr a0, vlenb
@@ -4289,9 +4276,8 @@ define <2 x i16> @ustest_f64i16_mm(<2 x double> %x) {
; CHECK-V: # %bb.0: # %entry
; CHECK-V-NEXT: vsetivli zero, 2, e32, mf2, ta, ma
; CHECK-V-NEXT: vfncvt.rtz.x.f.w v9, v8
-; CHECK-V-NEXT: vmax.vx v8, v9, zero
; CHECK-V-NEXT: vsetvli zero, zero, e16, mf4, ta, ma
-; CHECK-V-NEXT: vnclipu.wi v8, v8, 0
+; CHECK-V-NEXT: vnclipu.wi v8, v9, 0
; CHECK-V-NEXT: ret
entry:
%conv = fptosi <2 x double> %x to <2 x i32>
@@ -4479,7 +4465,6 @@ define <4 x i16> @ustest_f32i16_mm(<4 x float> %x) {
; CHECK-V: # %bb.0: # %entry
; CHECK-V-NEXT: vsetivli zero, 4, e32, m1, ta, ma
; CHECK-V-NEXT: vfcvt.rtz.x.f.v v8, v8
-; CHECK-V-NEXT: vmax.vx v8, v8, zero
; CHECK-V-NEXT: vsetvli zero, zero, e16, mf2, ta, ma
; CHECK-V-NEXT: vnclipu.wi v8, v8, 0
; CHECK-V-NEXT: ret
@@ -5449,24 +5434,23 @@ define <8 x i16> @ustest_f16i16_mm(<8 x half> %x) {
; CHECK-V-NEXT: call __extendhfsf2
; CHECK-V-NEXT: fcvt.l.s a0, fa0, rtz
; CHECK-V-NEXT: vsetivli zero, 2, e32, mf2, ta, ma
-; CHECK-V-NEXT: vmv.s.x v8, a0
+; CHECK-V-NEXT: vmv.s.x v10, a0
; CHECK-V-NEXT: addi a0, sp, 16
-; CHECK-V-NEXT: vl1r.v v9, (a0) # Unknown-size Folded Reload
-; CHECK-V-NEXT: vslideup.vi v8, v9, 1
+; CHECK-V-NEXT: vl1r.v v8, (a0) # Unknown-size Folded Reload
+; CHECK-V-NEXT: vslideup.vi v10, v8, 1
; CHECK-V-NEXT: csrr a0, vlenb
; CHECK-V-NEXT: add a0, sp, a0
; CHECK-V-NEXT: addi a0, a0, 16
-; CHECK-V-NEXT: vl1r.v v9, (a0) # Unknown-size Folded Reload
+; CHECK-V-NEXT: vl1r.v v8, (a0) # Unknown-size Folded Reload
; CHECK-V-NEXT: vsetivli zero, 4, e32, m1, ta, ma
-; CHECK-V-NEXT: vslideup.vi v8, v9, 2
+; CHECK-V-NEXT: vslideup.vi v10, v8, 2
; CHECK-V-NEXT: csrr a0, vlenb
; CHECK-V-NEXT: slli a0, a0, 1
; CHECK-V-NEXT: add a0, sp, a0
; CHECK-V-NEXT: addi a0, a0, 16
-; CHECK-V-NEXT: vl2r.v v10, (a0) # Unknown-size Folded Reload
+; CHECK-V-NEXT: vl2r.v v8, (a0) # Unknown-size Folded Reload
; CHECK-V-NEXT: vsetivli zero, 8, e32, m2, ta, ma
-; CHECK-V-NEXT: vslideup.vi v8, v10, 4
-; CHECK-V-NEXT: vmax.vx v10, v8, zero
+; CHECK-V-NEXT: vslideup.vi v10, v8, 4
; CHECK-V-NEXT: vsetvli zero, zero, e16, m1, ta, ma
; CHECK-V-NEXT: vnclipu.wi v8, v10, 0
; CHECK-V-NEXT: csrr a0, vlenb
diff --git a/llvm/test/CodeGen/RISCV/rvv/trunc-sat-clip-sdnode.ll b/llvm/test/CodeGen/RISCV/rvv/trunc-sat-clip-sdnode.ll
index 01a90d8a33b6ec..f43faadc532f26 100644
--- a/llvm/test/CodeGen/RISCV/rvv/trunc-sat-clip-sdnode.ll
+++ b/llvm/test/CodeGen/RISCV/rvv/trunc-sat-clip-sdnode.ll
@@ -102,9 +102,7 @@ define void @trunc_sat_u8u16_maxmin(ptr %x, ptr %y) {
; CHECK-LABEL: trunc_sat_u8u16_maxmin:
; CHECK: # %bb.0:
; CHECK-NEXT: vl1re16.v v8, (a0)
-; CHECK-NEXT: vsetvli a0, zero, e16, m1, ta, ma
-; CHECK-NEXT: vmax.vx v8, v8, zero
-; CHECK-NEXT: vsetvli zero, zero, e8, mf2, ta, ma
+; CHECK-NEXT: vsetvli a0, zero, e8, mf2, ta, ma
; CHECK-NEXT: vnclipu.wi v8, v8, 0
; CHECK-NEXT: vse8.v v8, (a1)
; CHECK-NEXT: ret
@@ -120,9 +118,7 @@ define void @trunc_sat_u8u16_minmax(ptr %x, ptr %y) {
; CHECK-LABEL: trunc_sat_u8u16_minmax:
; CHECK: # %bb.0:
; CHECK-NEXT: vl1re16.v v8, (a0)
-; CHECK-NEXT: vsetvli a0, zero, e16, m1, ta, ma
-; CHECK-NEXT: vmax.vx v8, v8, zero
-; CHECK-NEXT: vsetvli zero, zero, e8, mf2, ta, ma
+; CHECK-NEXT: vsetvli a0, zero, e8, mf2, ta, ma
; CHECK-NEXT: vnclipu.wi v8, v8, 0
; CHECK-NEXT: vse8.v v8, (a1)
; CHECK-NEXT: ret
@@ -357,9 +353,7 @@ define void @trunc_sat_u32u64_maxmin(ptr %x, ptr %y) {
; CHECK-LABEL: trunc_sat_u32u64_maxmin:
; CHECK: # %bb.0:
; CHECK-NEXT: vl4re64.v v8, (a0)
-; CHECK-NEXT: vsetvli a0, zero, e64, m4, ta, ma
-; CHECK-NEXT: vmax.vx v8, v8, zero
-; CHECK-NEXT: vsetvli zero, zero, e32, m2, ta, ma
+; CHECK-NEXT: vsetvli a0, zero, e32, m2, ta, ma
; CHECK-NEXT: vnclipu.wi v12, v8, 0
; CHECK-NEXT: vs2r.v v12, (a1)
; CHECK-NEXT: ret
@@ -375,9 +369,7 @@ define void @trunc_sat_u32u64_minmax(ptr %x, ptr %y) {
; CHECK-LABEL: trunc_sat_u32u64_minmax:
; CHECK: # %bb.0:
; CHECK-NEXT: vl4re64.v v8, (a0)
-; CHECK-NEXT: vsetvli a0, zero, e64, m4, ta, ma
-; CHECK-NEXT: vmax.vx v8, v8, zero
-; CHECK-NEXT: vsetvli zero, zero, e32, m2, ta, ma
+; CHECK-NEXT: vsetvli a0, zero, e32, m2, ta, ma
; CHECK-NEXT: vnclipu.wi v12, v8, 0
; CHECK-NEXT: vs2r.v v12, (a1)
; CHECK-NEXT: ret
@@ -446,9 +438,7 @@ define void @trunc_sat_u8u32_maxmin(ptr %x, ptr %y) {
; CHECK-LABEL: trunc_sat_u8u32_maxmin:
; CHECK: # %bb.0:
; CHECK-NEXT: vl2re32.v v8, (a0)
-; CHECK-NEXT: vsetvli a0, zero, e32, m2, ta, ma
-; CHECK-NEXT: vmax.vx v8, v8, zero
-; CHECK-NEXT: vsetvli zero, zero, e16, m1, ta, ma
+; CHECK-NEXT: vsetvli a0, zero, e16, m1, ta, ma
; CHECK-NEXT: vnclipu.wi v10, v8, 0
; CHECK-NEXT: vsetvli zero, zero, e8, mf2, ta, ma
; CHECK-NEXT: vnclipu.wi v8, v10, 0
@@ -466,9 +456,7 @@ define void @trunc_sat_u8u32_minmax(ptr %x, ptr %y) {
; CHECK-LABEL: trunc_sat_u8u32_minmax:
; CHECK: # %bb.0:
; CHECK-NEXT: vl2re32.v v8, (a0)
-; CHECK-NEXT: vsetvli a0, zero, e32, m2, ta, ma
-; CHECK-NEXT: vmax.vx v8, v8, zero
-; CHECK-NEXT: vsetvli zero, zero, e16, m1, ta, ma
+; CHECK-NEXT: vsetvli a0, zero, e16, m1, ta, ma
; CHECK-NEXT: vnclipu.wi v10, v8, 0
; CHECK-NEXT: vsetvli zero, zero, e8, mf2, ta, ma
; CHECK-NEXT: vnclipu.wi v8, v10, 0
@@ -545,9 +533,7 @@ define void @trunc_sat_u8u64_maxmin(ptr %x, ptr %y) {
; CHECK-LABEL: trunc_sat_u8u64_maxmin:
; CHECK: # %bb.0:
; CHECK-NEXT: vl4re64.v v8, (a0)
-; CHECK-NEXT: vsetvli a0, zero, e64, m4, ta, ma
-; CHECK-NEXT: vmax.vx v8, v8, zero
-; CHECK-NEXT: vsetvli zero, zero, e32, m2, ta, ma
+; CHECK-NEXT: vsetvli a0, zero, e32, m2, ta, ma
; CHECK-NEXT: vnclipu.wi v12, v8, 0
; CHECK-NEXT: vsetvli zero, zero, e16, m1, ta, ma
; CHECK-NEXT: vnclipu.wi v8, v12, 0
@@ -567,9 +553,7 @@ define void @trunc_sat_u8u64_minmax(ptr %x, ptr %y) {
; CHECK-LABEL: trunc_sat_u8u64_minmax:
; CHECK: # %bb.0:
; CHECK-NEXT: vl4re64.v v8, (a0)
-; CHECK-NEXT: vsetvli a0, zero, e64, m4, ta, ma
-; CHECK-NEXT: vmax.vx v8, v8, zero
-; CHECK-NEXT: vsetvli zero, zero, e32, m2, ta, ma
+; CHECK-NEXT: vsetvli a0, zero, e32, m2, ta, ma
; CHECK-NEXT: vnclipu.wi v12, v8, 0
; CHECK-NEXT: vsetvli zero, zero, e16, m1, ta, ma
; CHECK-NEXT: vnclipu.wi v8, v12, 0
diff --git a/llvm/test/CodeGen/RISCV/rvv/trunc-select-to-max-usat.ll b/llvm/test/CodeGen/RISCV/rvv/trunc-select-to-max-usat.ll
index 28d7588b9347a7..992ea8f8c18a5e 100644
--- a/llvm/test/CodeGen/RISCV/rvv/trunc-select-to-max-usat.ll
+++ b/llvm/test/CodeGen/RISCV/rvv/trunc-select-to-max-usat.ll
@@ -4,9 +4,7 @@
define <4 x i8> @test_v4i16_v4i8(<4 x i16> %x) {
; CHECK-LABEL: test_v4i16_v4i8:
; CHECK: # %bb.0:
-; CHECK-NEXT: vsetivli zero, 4, e16, mf2, ta, ma
-; CHECK-NEXT: vmax.vx v8, v8, zero
-; CHECK-NEXT: vsetvli zero, zero, e8, mf4, ta, ma
+; CHECK-NEXT: vsetivli zero, 4, e8, mf4, ta, ma
; CHECK-NEXT: vnclipu.wi v8, v8, 0
; CHECK-NEXT: ret
%a = icmp sgt <4 x i16> %x, zeroinitializer
@@ -20,9 +18,7 @@ define <4 x i8> @test_v4i16_v4i8(<4 x i16> %x) {
define <4 x i8> @test_v4i32_v4i8(<4 x i32> %x) {
; CHECK-LABEL: test_v4i32_v4i8:
; CHECK: # %bb.0:
-; CHECK-NEXT: vsetivli zero, 4, e32, m1, ta, ma
-; CHECK-NEXT: vmax.vx v8, v8, zero
-; CHECK-NEXT: vsetvli zero, zero, e16, mf2, ta, ma
+; CHECK-NEXT: vsetivli zero, 4, e16, mf2, ta, ma
; CHECK-NEXT: vnclipu.wi v8, v8, 0
; CHECK-NEXT: vsetvli zero, zero, e8, mf4, ta, ma
; CHECK-NEXT: vnclipu.wi v8, v8, 0
@@ -38,9 +34,7 @@ define <4 x i8> @test_v4i32_v4i8(<4 x i32> %x) {
define <4 x i8> @test_v4i64_v4i8(<4 x i64> %x) {
; CHECK-LABEL: test_v4i64_v4i8:
; CHECK: # %bb.0:
-; CHECK-NEXT: vsetivli zero, 4, e64, m2, ta, ma
-; CHECK-NEXT: vmax.vx v8, v8, zero
-; CHECK-NEXT: vsetvli zero, zero, e32, m1, ta, ma
+; CHECK-NEXT: vsetivli zero, 4, e32, m1, ta, ma
; CHECK-NEXT: vnclipu.wi v10, v8, 0
; CHECK-NEXT: vsetvli zero, zero, e16, mf2, ta, ma
; CHECK-NEXT: vnclipu.wi v8, v10, 0
@@ -58,9 +52,7 @@ define <4 x i8> @test_v4i64_v4i8(<4 x i64> %x) {
define <4 x i16> @test_v4i32_v4i16(<4 x i32> %x) {
; CHECK-LABEL: test_v4i32_v4i16:
; CHECK: # %bb.0:
-; CHECK-NEXT: vsetivli zero, 4, e32, m1, ta, ma
-; CHECK-NEXT: vmax.vx v8, v8, zero
-; CHECK-NEXT: vsetvli zero, zero, e16, mf2, ta, ma
+; CHECK-NEXT: vsetivli zero, 4, e16, mf2, ta, ma
; CHECK-NEXT: vnclipu.wi v8, v8, 0
; CHECK-NEXT: ret
%a = icmp sgt <4 x i32> %x, zeroinitializer
@@ -74,9 +66,7 @@ define <4 x i16> @test_v4i32_v4i16(<4 x i32> %x) {
define <4 x i16> @test_v4i64_v4i16(<4 x i64> %x) {
; CHECK-LABEL: test_v4i64_v4i16:
; CHECK: # %bb.0:
-; CHECK-NEXT: vsetivli zero, 4, e64, m2, ta, ma
-; CHECK-NEXT: vmax.vx v8, v8, zero
-; CHECK-NEXT: vsetvli zero, zero, e32, m1, ta, ma
+; CHECK-NEXT: vsetivli zero, 4, e32, m1, ta, ma
; CHECK-NEXT: vnclipu.wi v10, v8, 0
; CHECK-NEXT: vsetvli zero, zero, e16, mf2, ta, ma
; CHECK-NEXT: vnclipu.wi v8, v10, 0
@@ -92,10 +82,9 @@ define <4 x i16> @test_v4i64_v4i16(<4 x i64> %x) {
define <4 x i32> @test_v4i64_v4i32(<4 x i64> %x) {
; CHECK-LABEL: test_v4i64_v4i32:
; CHECK: # %bb.0:
-; CHECK-NEXT: vsetivli zero, 4, e64, m2, ta, ma
-; CHECK-NEXT: vmax.vx v10, v8, zero
-; CHECK-NEXT: vsetvli zero, zero, e32, m1, ta, ma
-; CHECK-NEXT: vnclipu.wi v8, v10, 0
+; CHECK-NEXT: vsetivli zero, 4, e32, m1, ta, ma
+; CHECK-NEXT: vnclipu.wi v10, v8, 0
+; CHECK-NEXT: vmv.v.v v8, v10
; CHECK-NEXT: ret
%a = icmp sgt <4 x i64> %x, zeroinitializer
%b = sext <4 x i1> %a to <4 x i64>
@@ -108,9 +97,7 @@ define <4 x i32> @test_v4i64_v4i32(<4 x i64> %x) {
define <vscale x 4 x i8> @test_nxv4i16_nxv4i8(<vscale x 4 x i16> %x) {
; CHECK-LABEL: test_nxv4i16_nxv4i8:
; CHECK: # %bb.0:
-; CHECK-NEXT: vsetvli a0, zero, e16, m1, ta, ma
-; CHECK-NEXT: vmax.vx v8, v8, zero
-; CHECK-NEXT: vsetvli zero, zero, e8, mf2, ta, ma
+; CHECK-NEXT: vsetvli a0, zero, e8, mf2, ta, ma
; CHECK-NEXT: vnclipu.wi v8, v8, 0
; CHECK-NEXT: ret
%a = icmp sgt <vscale x 4 x i16> %x, zeroinitializer
@@ -124,9 +111,7 @@ define <vscale x 4 x i8> @test_nxv4i16_nxv4i8(<vscale x 4 x i16> %x) {
define <vscale x 4 x i8> @test_nxv4i32_nxv4i8(<vscale x 4 x i32> %x) {
; CHECK-LABEL: test_nxv4i32_nxv4i8:
; CHECK: # %bb.0:
-; CHECK-NEXT: vsetvli a0, zero, e32, m2, ta, ma
-; CHECK-NEXT: vmax.vx v8, v8, zero
-; CHECK-NEXT: vsetvli zero, zero, e16, m1, ta, ma
+; CHECK-NEXT: vsetvli a0, zero, e16, m1, ta, ma
; CHECK-NEXT: vnclipu.wi v10, v8, 0
; CHECK-NEXT: vsetvli zero, zero, e8, mf2, ta, ma
; CHECK-NEXT: vnclipu.wi v8, v10, 0
@@ -142,9 +127,7 @@ define <vscale x 4 x i8> @test_nxv4i32_nxv4i8(<vscale x 4 x i32> %x) {
define <vscale x 4 x i8> @test_nxv4i64_nxv4i8(<vscale x 4 x i64> %x) {
; CHECK-LABEL: test_nxv4i64_nxv4i8:
; CHECK: # %bb.0:
-; CHECK-NEXT: vsetvli a0, zero, e64, m4, ta, ma
-; CHECK-NEXT: vmax.vx v8, v8, zero
-; CHECK-NEXT: vsetvli zero, zero, e32, m2, ta, ma
+; CHECK-NEXT: vsetvli a0, zero, e32, m2, ta, ma
; CHECK-NEXT: vnclipu.wi v12, v8, 0
; CHECK-NEXT: vsetvli zero, zero, e16, m1, ta, ma
; CHECK-NEXT: vnclipu.wi v8, v12, 0
@@ -162,10 +145,9 @@ define <vscale x 4 x i8> @test_nxv4i64_nxv4i8(<vscale x 4 x i64> %x) {
define <vscale x 4 x i16> @test_nxv4i32_nxv4i16(<vscale x 4 x i32> %x) {
; CHECK-LABEL: test_nxv4i32_nxv4i16:
; CHECK: # %bb.0:
-; CHECK-NEXT: vsetvli a0, zero, e32, m2, ta, ma
-; CHECK-NEXT: vmax.vx v10, v8, zero
-; CHECK-NEXT: vsetvli zero, zero, e16, m1, ta, ma
-; CHECK-NEXT: vnclipu.wi v8, v10, 0
+; CHECK-NEXT: vsetvli a0, zero, e16, m1, ta, ma
+; CHECK-NEXT: vnclipu.wi v10, v8, 0
+; CHECK-NEXT: vmv.v.v v8, v10
; CHECK-NEXT: ret
%a = icmp sgt <vscale x 4 x i32> %x, zeroinitializer
%b = sext <vscale x 4 x i1> %a to <vscale x 4 x i32>
@@ -178,9 +160,7 @@ define <vscale x 4 x i16> @test_nxv4i32_nxv4i16(<vscale x 4 x i32> %x) {
define <vscale x 4 x i16> @test_nxv4i64_nxv4i16(<vscale x 4 x i64> %x) {
; CHECK-LABEL: test_nxv4i64_nxv4i16:
; CHECK: # %bb.0:
-; CHECK-NEXT: vsetvli a0, zero, e64, m4, ta, ma
-; CHECK-NEXT: vmax.vx v8, v8, zero
-; CHECK-NEXT: vsetvli zero, zero, e32, m2, ta, ma
+; CHECK-NEXT: vsetvli a0, zero, e32, m2, ta, ma
; CHECK-NEXT: vnclipu.wi v12, v8, 0
; CHECK-NEXT: vsetvli zero, zero, e16, m1, ta, ma
; CHECK-NEXT: vnclipu.wi v8, v12, 0
@@ -196,10 +176,9 @@ define <vscale x 4 x i16> @test_nxv4i64_nxv4i16(<vscale x 4 x i64> %x) {
define <vscale x 4 x i32> @test_nxv4i64_nxv4i32(<vscale x 4 x i64> %x) {
; CHECK-LABEL: test_nxv4i64_nxv4i32:
; CHECK: # %bb.0:
-; CHECK-NEXT: vsetvli a0, zero, e64, m4, ta, ma
-; CHECK-NEXT: vmax.vx v12, v8, zero
-; CHECK-NEXT: vsetvli zero, zero, e32, m2, ta, ma
-; CHECK-NEXT: vnclipu.wi v8, v12, 0
+; CHECK-NEXT: vsetvli a0, zero, e32, m2, ta, ma
+; CHECK-NEXT: vnclipu.wi v12, v8, 0
+; CHECK-NEXT: vmv.v.v v8, v12
; CHECK-NEXT: ret
%a = icmp sgt <vscale x 4 x i64> %x, zeroinitializer
%b = sext <vscale x 4 x i1> %a to <vscale x 4 x i64>
More information about the llvm-commits
mailing list