[llvm] [RISCV][GlobalISel] Vector Extension vadd Legalizer (PR #71400)
Jiahan Xie via llvm-commits
llvm-commits at lists.llvm.org
Mon Nov 6 10:14:32 PST 2023
https://github.com/jiahanxie353 updated https://github.com/llvm/llvm-project/pull/71400
>From 77b8c70fc25d1b4f27ff5e424eca09463bcd218f Mon Sep 17 00:00:00 2001
From: jiahanxie353 <jx353 at cornell.edu>
Date: Sat, 4 Nov 2023 10:04:22 -0400
Subject: [PATCH 1/4] [CodeGen][MachineVerifier] Use TypeSize instead of
unsigned for getRegSizeInBits
---
.../include/llvm/CodeGen/TargetRegisterInfo.h | 6 ++---
llvm/lib/CodeGen/MachineVerifier.cpp | 18 ++++++++-------
llvm/lib/CodeGen/TargetRegisterInfo.cpp | 19 ++++++++-------
.../RISCV/GlobalISel/irtranslator/fallback.ll | 4 ++--
llvm/test/MachineVerifier/copy-scalable.mir | 23 +++++++++++++++++++
5 files changed, 47 insertions(+), 23 deletions(-)
create mode 100644 llvm/test/MachineVerifier/copy-scalable.mir
diff --git a/llvm/include/llvm/CodeGen/TargetRegisterInfo.h b/llvm/include/llvm/CodeGen/TargetRegisterInfo.h
index 337fab735a09522..4fb6ba7c26930af 100644
--- a/llvm/include/llvm/CodeGen/TargetRegisterInfo.h
+++ b/llvm/include/llvm/CodeGen/TargetRegisterInfo.h
@@ -283,8 +283,8 @@ class TargetRegisterInfo : public MCRegisterInfo {
// DenseMapInfo<unsigned> uses -1u and -2u.
/// Return the size in bits of a register from class RC.
- unsigned getRegSizeInBits(const TargetRegisterClass &RC) const {
- return getRegClassInfo(RC).RegSize;
+ TypeSize getRegSizeInBits(const TargetRegisterClass &RC) const {
+ return TypeSize::Fixed(getRegClassInfo(RC).RegSize);
}
/// Return the size in bytes of the stack slot allocated to hold a spilled
@@ -858,7 +858,7 @@ class TargetRegisterInfo : public MCRegisterInfo {
const TargetRegisterClass *RC) const = 0;
/// Returns size in bits of a phys/virtual/generic register.
- unsigned getRegSizeInBits(Register Reg, const MachineRegisterInfo &MRI) const;
+ TypeSize getRegSizeInBits(Register Reg, const MachineRegisterInfo &MRI) const;
/// Get the weight in units of pressure for this register unit.
virtual unsigned getRegUnitWeight(unsigned RegUnit) const = 0;
diff --git a/llvm/lib/CodeGen/MachineVerifier.cpp b/llvm/lib/CodeGen/MachineVerifier.cpp
index dadaf60fa09da04..dc15f0d3b842304 100644
--- a/llvm/lib/CodeGen/MachineVerifier.cpp
+++ b/llvm/lib/CodeGen/MachineVerifier.cpp
@@ -1937,8 +1937,8 @@ void MachineVerifier::visitMachineInstrBefore(const MachineInstr *MI) {
// If we have only one valid type, this is likely a copy between a virtual
// and physical register.
- unsigned SrcSize = 0;
- unsigned DstSize = 0;
+ TypeSize SrcSize = TRI->getRegSizeInBits(SrcReg, *MRI);
+ TypeSize DstSize = TRI->getRegSizeInBits(DstReg, *MRI);
if (SrcReg.isPhysical() && DstTy.isValid()) {
const TargetRegisterClass *SrcRC =
TRI->getMinimalPhysRegClassLLT(SrcReg, DstTy);
@@ -1946,9 +1946,6 @@ void MachineVerifier::visitMachineInstrBefore(const MachineInstr *MI) {
SrcSize = TRI->getRegSizeInBits(*SrcRC);
}
- if (SrcSize == 0)
- SrcSize = TRI->getRegSizeInBits(SrcReg, *MRI);
-
if (DstReg.isPhysical() && SrcTy.isValid()) {
const TargetRegisterClass *DstRC =
TRI->getMinimalPhysRegClassLLT(DstReg, SrcTy);
@@ -1956,10 +1953,15 @@ void MachineVerifier::visitMachineInstrBefore(const MachineInstr *MI) {
DstSize = TRI->getRegSizeInBits(*DstRC);
}
- if (DstSize == 0)
- DstSize = TRI->getRegSizeInBits(DstReg, *MRI);
+ // If this is a copy from physical register to virtual register, and if the
+ // Dst is scalable and the Src is fixed, then the Dst can only hold the Src
+ // if the minimum size Dst can hold is at least as big as Src.
+ if (SrcReg.isPhysical() && DstReg.isVirtual() && DstSize.isScalable() &&
+ !SrcSize.isScalable() &&
+ DstSize.getKnownMinValue() <= SrcSize.getFixedValue())
+ break;
- if (SrcSize != 0 && DstSize != 0 && SrcSize != DstSize) {
+ if (SrcSize.isNonZero() && DstSize.isNonZero() && SrcSize != DstSize) {
if (!DstOp.getSubReg() && !SrcOp.getSubReg()) {
report("Copy Instruction is illegal with mismatching sizes", MI);
errs() << "Def Size = " << DstSize << ", Src Size = " << SrcSize
diff --git a/llvm/lib/CodeGen/TargetRegisterInfo.cpp b/llvm/lib/CodeGen/TargetRegisterInfo.cpp
index 1bb35f40facfd0f..c50b1cf9422717a 100644
--- a/llvm/lib/CodeGen/TargetRegisterInfo.cpp
+++ b/llvm/lib/CodeGen/TargetRegisterInfo.cpp
@@ -499,7 +499,7 @@ bool TargetRegisterInfo::regmaskSubsetEqual(const uint32_t *mask0,
return true;
}
-unsigned
+TypeSize
TargetRegisterInfo::getRegSizeInBits(Register Reg,
const MachineRegisterInfo &MRI) const {
const TargetRegisterClass *RC{};
@@ -508,16 +508,15 @@ TargetRegisterInfo::getRegSizeInBits(Register Reg,
// Instead, we need to access a register class that contains Reg and
// get the size of that register class.
RC = getMinimalPhysRegClass(Reg);
- } else {
- LLT Ty = MRI.getType(Reg);
- unsigned RegSize = Ty.isValid() ? Ty.getSizeInBits() : 0;
- // If Reg is not a generic register, query the register class to
- // get its size.
- if (RegSize)
- return RegSize;
- // Since Reg is not a generic register, it must have a register class.
- RC = MRI.getRegClass(Reg);
+ assert(RC && "Unable to deduce the register class");
+ return getRegSizeInBits(*RC);
}
+ LLT Ty = MRI.getType(Reg);
+ if (Ty.isValid())
+ return Ty.getSizeInBits();
+
+ // Since Reg is not a generic register, it may have a register class.
+ RC = MRI.getRegClass(Reg);
assert(RC && "Unable to deduce the register class");
return getRegSizeInBits(*RC);
}
diff --git a/llvm/test/CodeGen/RISCV/GlobalISel/irtranslator/fallback.ll b/llvm/test/CodeGen/RISCV/GlobalISel/irtranslator/fallback.ll
index 5dd62de8a6bc415..a3a913d8ce02d83 100644
--- a/llvm/test/CodeGen/RISCV/GlobalISel/irtranslator/fallback.ll
+++ b/llvm/test/CodeGen/RISCV/GlobalISel/irtranslator/fallback.ll
@@ -22,7 +22,7 @@ entry:
ret <vscale x 1 x i8> %a
}
-; FALLBACK-WITH-REPORT-ERR: remark: <unknown>:0:0: unable to translate instruction{{.*}}scalable_inst
+; FALLBACK-WITH-REPORT-ERR: remark: <unknown>:0:0: unable to translate instruction: call:
; FALLBACK-WITH-REPORT-OUT-LABEL: scalable_inst
define <vscale x 1 x i8> @scalable_inst(i64 %0) nounwind {
entry:
@@ -35,7 +35,7 @@ entry:
ret <vscale x 1 x i8> %a
}
-; FALLBACK-WITH-REPORT-ERR: remark: <unknown>:0:0: unable to translate instruction{{.*}}scalable_alloca
+; FALLBACK-WITH-REPORT-ERR: remark: <unknown>:0:0: unable to translate instruction: alloca:
; FALLBACK-WITH-REPORT-OUT-LABEL: scalable_alloca
define void @scalable_alloca() #1 {
%local0 = alloca <vscale x 16 x i8>
diff --git a/llvm/test/MachineVerifier/copy-scalable.mir b/llvm/test/MachineVerifier/copy-scalable.mir
new file mode 100644
index 000000000000000..f4088f7aed34dde
--- /dev/null
+++ b/llvm/test/MachineVerifier/copy-scalable.mir
@@ -0,0 +1,23 @@
+# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py UTC_ARGS: --version 3
+# RUN: llc -mtriple=riscv64 -o - -global-isel -run-pass=none -verify-machineinstrs %s | FileCheck %s
+# REQUIRES: riscv64-registered-target
+
+---
+name: test_copy_fixed_to_scalable
+legalized: true
+regBankSelected: false
+selected: false
+tracksRegLiveness: true
+registers:
+ - { id: 0, class: _, preferred-register: '' }
+liveins:
+body: |
+ bb.0:
+ liveins: $v8
+
+ ; CHECK-LABEL: name: test_copy_fixed_to_scalable
+ ; CHECK: liveins: $v8
+ ; CHECK-NEXT: {{ $}}
+ ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 1 x s8>) = COPY $v8
+ %0:_(<vscale x 1 x s8>) = COPY $v8
+...
>From 54c7ee68e13861cda0666041200f5a9dabb4e93b Mon Sep 17 00:00:00 2001
From: jiahanxie353 <jx353 at cornell.edu>
Date: Sat, 4 Nov 2023 11:55:38 -0400
Subject: [PATCH 2/4] [RISCV][GISEL] Add support for lowerFormalArguments that
contain scalable vector types
---
llvm/lib/CodeGen/GlobalISel/CallLowering.cpp | 3 +-
.../CodeGen/GlobalISel/MachineIRBuilder.cpp | 6 +-
llvm/lib/CodeGen/LowLevelType.cpp | 2 +-
llvm/lib/CodeGen/MachineVerifier.cpp | 21 +-
.../Target/RISCV/GISel/RISCVCallLowering.cpp | 37 +-
.../irtranslator/vec-args-bf16-err.ll | 16 +
.../irtranslator/vec-args-f16-err.ll | 16 +
.../RISCV/GlobalISel/irtranslator/vec-args.ll | 909 ++++++++++++++++++
8 files changed, 997 insertions(+), 13 deletions(-)
create mode 100644 llvm/test/CodeGen/RISCV/GlobalISel/irtranslator/vec-args-bf16-err.ll
create mode 100644 llvm/test/CodeGen/RISCV/GlobalISel/irtranslator/vec-args-f16-err.ll
create mode 100644 llvm/test/CodeGen/RISCV/GlobalISel/irtranslator/vec-args.ll
diff --git a/llvm/lib/CodeGen/GlobalISel/CallLowering.cpp b/llvm/lib/CodeGen/GlobalISel/CallLowering.cpp
index 975787a8f5e734f..2527b1431289677 100644
--- a/llvm/lib/CodeGen/GlobalISel/CallLowering.cpp
+++ b/llvm/lib/CodeGen/GlobalISel/CallLowering.cpp
@@ -358,7 +358,7 @@ static void buildCopyFromRegs(MachineIRBuilder &B, ArrayRef<Register> OrigRegs,
if (PartLLT.isVector() == LLTy.isVector() &&
PartLLT.getScalarSizeInBits() > LLTy.getScalarSizeInBits() &&
(!PartLLT.isVector() ||
- PartLLT.getNumElements() == LLTy.getNumElements()) &&
+ PartLLT.getElementCount() == LLTy.getElementCount()) &&
OrigRegs.size() == 1 && Regs.size() == 1) {
Register SrcReg = Regs[0];
@@ -406,6 +406,7 @@ static void buildCopyFromRegs(MachineIRBuilder &B, ArrayRef<Register> OrigRegs,
// If PartLLT is a mismatched vector in both number of elements and element
// size, e.g. PartLLT == v2s64 and LLTy is v3s32, then first coerce it to
// have the same elt type, i.e. v4s32.
+ // TODO: Extend this coersion to element multiples other than just 2.
if (PartLLT.getSizeInBits() > LLTy.getSizeInBits() &&
PartLLT.getScalarSizeInBits() == LLTy.getScalarSizeInBits() * 2 &&
Regs.size() == 1) {
diff --git a/llvm/lib/CodeGen/GlobalISel/MachineIRBuilder.cpp b/llvm/lib/CodeGen/GlobalISel/MachineIRBuilder.cpp
index 5b4e2b725e1dd76..80e9c08e850b683 100644
--- a/llvm/lib/CodeGen/GlobalISel/MachineIRBuilder.cpp
+++ b/llvm/lib/CodeGen/GlobalISel/MachineIRBuilder.cpp
@@ -1065,16 +1065,16 @@ void MachineIRBuilder::validateTruncExt(const LLT DstTy, const LLT SrcTy,
#ifndef NDEBUG
if (DstTy.isVector()) {
assert(SrcTy.isVector() && "mismatched cast between vector and non-vector");
- assert(SrcTy.getNumElements() == DstTy.getNumElements() &&
+ assert(SrcTy.getElementCount() == DstTy.getElementCount() &&
"different number of elements in a trunc/ext");
} else
assert(DstTy.isScalar() && SrcTy.isScalar() && "invalid extend/trunc");
if (IsExtend)
- assert(DstTy.getSizeInBits() > SrcTy.getSizeInBits() &&
+ assert(TypeSize::isKnownGT(DstTy.getSizeInBits(), SrcTy.getSizeInBits()) &&
"invalid narrowing extend");
else
- assert(DstTy.getSizeInBits() < SrcTy.getSizeInBits() &&
+ assert(TypeSize::isKnownLT(DstTy.getSizeInBits(), SrcTy.getSizeInBits()) &&
"invalid widening trunc");
#endif
}
diff --git a/llvm/lib/CodeGen/LowLevelType.cpp b/llvm/lib/CodeGen/LowLevelType.cpp
index 24c30b756737b20..cd85bf606989f9e 100644
--- a/llvm/lib/CodeGen/LowLevelType.cpp
+++ b/llvm/lib/CodeGen/LowLevelType.cpp
@@ -17,7 +17,7 @@ using namespace llvm;
LLT::LLT(MVT VT) {
if (VT.isVector()) {
- bool asVector = VT.getVectorMinNumElements() > 1;
+ bool asVector = VT.getVectorMinNumElements() > 1 || VT.isScalableVector();
init(/*IsPointer=*/false, asVector, /*IsScalar=*/!asVector,
VT.getVectorElementCount(), VT.getVectorElementType().getSizeInBits(),
/*AddressSpace=*/0);
diff --git a/llvm/lib/CodeGen/MachineVerifier.cpp b/llvm/lib/CodeGen/MachineVerifier.cpp
index dc15f0d3b842304..9ccda9abc019ddf 100644
--- a/llvm/lib/CodeGen/MachineVerifier.cpp
+++ b/llvm/lib/CodeGen/MachineVerifier.cpp
@@ -965,7 +965,7 @@ bool MachineVerifier::verifyVectorElementMatch(LLT Ty0, LLT Ty1,
return false;
}
- if (Ty0.isVector() && Ty0.getNumElements() != Ty1.getNumElements()) {
+ if (Ty0.isVector() && Ty0.getElementCount() != Ty1.getElementCount()) {
report("operand types must preserve number of vector elements", MI);
return false;
}
@@ -1946,6 +1946,9 @@ void MachineVerifier::visitMachineInstrBefore(const MachineInstr *MI) {
SrcSize = TRI->getRegSizeInBits(*SrcRC);
}
+ if (SrcSize.isZero())
+ SrcSize = TRI->getRegSizeInBits(SrcReg, *MRI);
+
if (DstReg.isPhysical() && SrcTy.isValid()) {
const TargetRegisterClass *DstRC =
TRI->getMinimalPhysRegClassLLT(DstReg, SrcTy);
@@ -1953,13 +1956,19 @@ void MachineVerifier::visitMachineInstrBefore(const MachineInstr *MI) {
DstSize = TRI->getRegSizeInBits(*DstRC);
}
- // If this is a copy from physical register to virtual register, and if the
- // Dst is scalable and the Src is fixed, then the Dst can only hold the Src
- // if the minimum size Dst can hold is at least as big as Src.
- if (SrcReg.isPhysical() && DstReg.isVirtual() && DstSize.isScalable() &&
- !SrcSize.isScalable() &&
+ if (DstSize.isZero())
+ DstSize = TRI->getRegSizeInBits(DstReg, *MRI);
+
+ // If the Dst is scalable and the Src is fixed, then the Dst can only hold
+ // the Src if the minimum size Dst can hold is at least as big as Src.
+ if (DstSize.isScalable() && !SrcSize.isScalable() &&
DstSize.getKnownMinValue() <= SrcSize.getFixedValue())
break;
+ // If the Src is scalable and the Dst is fixed, then Dest can only hold
+ // the Src is known to fit in Dest
+ if (SrcSize.isScalable() && !DstSize.isScalable() &&
+ TypeSize::isKnownLE(DstSize, SrcSize))
+ break;
if (SrcSize.isNonZero() && DstSize.isNonZero() && SrcSize != DstSize) {
if (!DstOp.getSubReg() && !SrcOp.getSubReg()) {
diff --git a/llvm/lib/Target/RISCV/GISel/RISCVCallLowering.cpp b/llvm/lib/Target/RISCV/GISel/RISCVCallLowering.cpp
index a1dbc21ca364666..e73d8863963d0b2 100644
--- a/llvm/lib/Target/RISCV/GISel/RISCVCallLowering.cpp
+++ b/llvm/lib/Target/RISCV/GISel/RISCVCallLowering.cpp
@@ -14,6 +14,7 @@
#include "RISCVCallLowering.h"
#include "RISCVISelLowering.h"
+#include "RISCVMachineFunctionInfo.h"
#include "RISCVSubtarget.h"
#include "llvm/CodeGen/Analysis.h"
#include "llvm/CodeGen/GlobalISel/MachineIRBuilder.h"
@@ -185,6 +186,9 @@ struct RISCVIncomingValueAssigner : public CallLowering::IncomingValueAssigner {
const DataLayout &DL = MF.getDataLayout();
const RISCVSubtarget &Subtarget = MF.getSubtarget<RISCVSubtarget>();
+ if (LocVT.isScalableVector())
+ MF.getInfo<RISCVMachineFunctionInfo>()->setIsVectorCall();
+
if (RISCVAssignFn(DL, Subtarget.getTargetABI(), ValNo, ValVT, LocVT,
LocInfo, Flags, State, /*IsFixed=*/true, IsRet, Info.Ty,
*Subtarget.getTargetLowering(),
@@ -301,8 +305,31 @@ struct RISCVCallReturnHandler : public RISCVIncomingValueHandler {
RISCVCallLowering::RISCVCallLowering(const RISCVTargetLowering &TLI)
: CallLowering(&TLI) {}
+/// Return true if scalable vector with ScalarTy is legal for lowering.
+static bool isLegalElementTypeForRVV(Type *EltTy,
+ const RISCVSubtarget &Subtarget) {
+ if (EltTy->isPointerTy())
+ return Subtarget.is64Bit() ? Subtarget.hasVInstructionsI64() : true;
+ if (EltTy->isIntegerTy(1) || EltTy->isIntegerTy(8) ||
+ EltTy->isIntegerTy(16) || EltTy->isIntegerTy(32))
+ return true;
+ if (EltTy->isIntegerTy(64))
+ return Subtarget.hasVInstructionsI64();
+ if (EltTy->isHalfTy())
+ return Subtarget.hasVInstructionsF16();
+ if (EltTy->isBFloatTy())
+ return Subtarget.hasVInstructionsBF16();
+ if (EltTy->isFloatTy())
+ return Subtarget.hasVInstructionsF32();
+ if (EltTy->isDoubleTy())
+ return Subtarget.hasVInstructionsF64();
+ return false;
+}
+
// TODO: Support all argument types.
-static bool isSupportedArgumentType(Type *T, const RISCVSubtarget &Subtarget) {
+// TODO: Remove IsLowerArgs argument by adding support for vectors in lowerCall.
+static bool isSupportedArgumentType(Type *T, const RISCVSubtarget &Subtarget,
+ bool IsLowerArgs = false) {
// TODO: Integers larger than 2*XLen are passed indirectly which is not
// supported yet.
if (T->isIntegerTy())
@@ -311,6 +338,11 @@ static bool isSupportedArgumentType(Type *T, const RISCVSubtarget &Subtarget) {
return true;
if (T->isPointerTy())
return true;
+ // TODO: Support fixed vector types.
+ if (IsLowerArgs && T->isVectorTy() && Subtarget.hasVInstructions() &&
+ T->isScalableTy() &&
+ isLegalElementTypeForRVV(T->getScalarType(), Subtarget))
+ return true;
return false;
}
@@ -398,7 +430,8 @@ bool RISCVCallLowering::lowerFormalArguments(MachineIRBuilder &MIRBuilder,
const RISCVSubtarget &Subtarget =
MIRBuilder.getMF().getSubtarget<RISCVSubtarget>();
for (auto &Arg : F.args()) {
- if (!isSupportedArgumentType(Arg.getType(), Subtarget))
+ if (!isSupportedArgumentType(Arg.getType(), Subtarget,
+ /*IsLowerArgs=*/true))
return false;
}
diff --git a/llvm/test/CodeGen/RISCV/GlobalISel/irtranslator/vec-args-bf16-err.ll b/llvm/test/CodeGen/RISCV/GlobalISel/irtranslator/vec-args-bf16-err.ll
new file mode 100644
index 000000000000000..f39e7793e5d4f31
--- /dev/null
+++ b/llvm/test/CodeGen/RISCV/GlobalISel/irtranslator/vec-args-bf16-err.ll
@@ -0,0 +1,16 @@
+; RUN: not --crash llc -mtriple=riscv32 -mattr=+v -global-isel -stop-after=irtranslator \
+; RUN: -verify-machineinstrs < %s 2>&1 | FileCheck %s
+; RUN: not --crash llc -mtriple=riscv64 -mattr=+v -global-isel -stop-after=irtranslator \
+; RUN: -verify-machineinstrs < %s 2>&1 | FileCheck %s
+
+; The purpose of this test is to show that the compiler throws an error when
+; there is no support for bf16 vectors. If the compiler did not throw an error,
+; then it will try to scalarize the argument to an s32, which may drop elements.
+define void @test_args_nxv1bf16(<vscale x 1 x bfloat> %a) {
+entry:
+ ret void
+}
+
+; CHECK: LLVM ERROR: unable to lower arguments: ptr (in function: test_args_nxv1bf16)
+
+
diff --git a/llvm/test/CodeGen/RISCV/GlobalISel/irtranslator/vec-args-f16-err.ll b/llvm/test/CodeGen/RISCV/GlobalISel/irtranslator/vec-args-f16-err.ll
new file mode 100644
index 000000000000000..042b455bfb54754
--- /dev/null
+++ b/llvm/test/CodeGen/RISCV/GlobalISel/irtranslator/vec-args-f16-err.ll
@@ -0,0 +1,16 @@
+; RUN: not --crash llc -mtriple=riscv32 -mattr=+v -global-isel -stop-after=irtranslator \
+; RUN: -verify-machineinstrs < %s 2>&1 | FileCheck %s
+; RUN: not --crash llc -mtriple=riscv64 -mattr=+v -global-isel -stop-after=irtranslator \
+; RUN: -verify-machineinstrs < %s 2>&1 | FileCheck %s
+
+; The purpose of this test is to show that the compiler throws an error when
+; there is no support for f16 vectors. If the compiler did not throw an error,
+; then it will try to scalarize the argument to an s32, which may drop elements.
+define void @test_args_nxv1f16(<vscale x 1 x half> %a) {
+entry:
+ ret void
+}
+
+; CHECK: LLVM ERROR: unable to lower arguments: ptr (in function: test_args_nxv1f16)
+
+
diff --git a/llvm/test/CodeGen/RISCV/GlobalISel/irtranslator/vec-args.ll b/llvm/test/CodeGen/RISCV/GlobalISel/irtranslator/vec-args.ll
new file mode 100644
index 000000000000000..4df0a8f48cc8d0b
--- /dev/null
+++ b/llvm/test/CodeGen/RISCV/GlobalISel/irtranslator/vec-args.ll
@@ -0,0 +1,909 @@
+; NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
+; RUN: llc -mtriple=riscv32 -mattr=+v,+experimental-zvfbfmin,+zvfh -global-isel -stop-after=irtranslator \
+; RUN: -verify-machineinstrs < %s | FileCheck -check-prefixes=RV32 %s
+; RUN: llc -mtriple=riscv64 -mattr=+v,+experimental-zvfbfmin,+zvfh -global-isel -stop-after=irtranslator \
+; RUN: -verify-machineinstrs < %s | FileCheck -check-prefixes=RV64 %s
+
+; ==========================================================================
+; ============================= Scalable Types =============================
+; ==========================================================================
+
+define void @test_args_nxv1i8(<vscale x 1 x i8> %a) {
+ ; RV32-LABEL: name: test_args_nxv1i8
+ ; RV32: bb.1.entry:
+ ; RV32-NEXT: liveins: $v8
+ ; RV32-NEXT: {{ $}}
+ ; RV32-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 1 x s8>) = COPY $v8
+ ; RV32-NEXT: PseudoRET
+ ;
+ ; RV64-LABEL: name: test_args_nxv1i8
+ ; RV64: bb.1.entry:
+ ; RV64-NEXT: liveins: $v8
+ ; RV64-NEXT: {{ $}}
+ ; RV64-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 1 x s8>) = COPY $v8
+ ; RV64-NEXT: PseudoRET
+entry:
+ ret void
+}
+
+define void @test_args_nxv2i8(<vscale x 2 x i8> %a) {
+ ; RV32-LABEL: name: test_args_nxv2i8
+ ; RV32: bb.1.entry:
+ ; RV32-NEXT: liveins: $v8
+ ; RV32-NEXT: {{ $}}
+ ; RV32-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 2 x s8>) = COPY $v8
+ ; RV32-NEXT: PseudoRET
+ ;
+ ; RV64-LABEL: name: test_args_nxv2i8
+ ; RV64: bb.1.entry:
+ ; RV64-NEXT: liveins: $v8
+ ; RV64-NEXT: {{ $}}
+ ; RV64-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 2 x s8>) = COPY $v8
+ ; RV64-NEXT: PseudoRET
+entry:
+ ret void
+}
+
+define void @test_args_nxv4i8(<vscale x 4 x i8> %a) {
+ ; RV32-LABEL: name: test_args_nxv4i8
+ ; RV32: bb.1.entry:
+ ; RV32-NEXT: liveins: $v8
+ ; RV32-NEXT: {{ $}}
+ ; RV32-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 4 x s8>) = COPY $v8
+ ; RV32-NEXT: PseudoRET
+ ;
+ ; RV64-LABEL: name: test_args_nxv4i8
+ ; RV64: bb.1.entry:
+ ; RV64-NEXT: liveins: $v8
+ ; RV64-NEXT: {{ $}}
+ ; RV64-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 4 x s8>) = COPY $v8
+ ; RV64-NEXT: PseudoRET
+entry:
+ ret void
+}
+
+define void @test_args_nxv8i8(<vscale x 8 x i8> %a) {
+ ; RV32-LABEL: name: test_args_nxv8i8
+ ; RV32: bb.1.entry:
+ ; RV32-NEXT: liveins: $v8
+ ; RV32-NEXT: {{ $}}
+ ; RV32-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 8 x s8>) = COPY $v8
+ ; RV32-NEXT: PseudoRET
+ ;
+ ; RV64-LABEL: name: test_args_nxv8i8
+ ; RV64: bb.1.entry:
+ ; RV64-NEXT: liveins: $v8
+ ; RV64-NEXT: {{ $}}
+ ; RV64-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 8 x s8>) = COPY $v8
+ ; RV64-NEXT: PseudoRET
+entry:
+ ret void
+}
+
+define void @test_args_nxv16i8(<vscale x 16 x i8> %a) {
+ ; RV32-LABEL: name: test_args_nxv16i8
+ ; RV32: bb.1.entry:
+ ; RV32-NEXT: liveins: $v8m2
+ ; RV32-NEXT: {{ $}}
+ ; RV32-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 16 x s8>) = COPY $v8m2
+ ; RV32-NEXT: PseudoRET
+ ;
+ ; RV64-LABEL: name: test_args_nxv16i8
+ ; RV64: bb.1.entry:
+ ; RV64-NEXT: liveins: $v8m2
+ ; RV64-NEXT: {{ $}}
+ ; RV64-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 16 x s8>) = COPY $v8m2
+ ; RV64-NEXT: PseudoRET
+entry:
+ ret void
+}
+
+define void @test_args_nxv32i8(<vscale x 32 x i8> %a) {
+ ; RV32-LABEL: name: test_args_nxv32i8
+ ; RV32: bb.1.entry:
+ ; RV32-NEXT: liveins: $v8m4
+ ; RV32-NEXT: {{ $}}
+ ; RV32-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 32 x s8>) = COPY $v8m4
+ ; RV32-NEXT: PseudoRET
+ ;
+ ; RV64-LABEL: name: test_args_nxv32i8
+ ; RV64: bb.1.entry:
+ ; RV64-NEXT: liveins: $v8m4
+ ; RV64-NEXT: {{ $}}
+ ; RV64-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 32 x s8>) = COPY $v8m4
+ ; RV64-NEXT: PseudoRET
+entry:
+ ret void
+}
+
+define void @test_args_nxv64i8(<vscale x 64 x i8> %a) {
+ ; RV32-LABEL: name: test_args_nxv64i8
+ ; RV32: bb.1.entry:
+ ; RV32-NEXT: liveins: $v8m8
+ ; RV32-NEXT: {{ $}}
+ ; RV32-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 64 x s8>) = COPY $v8m8
+ ; RV32-NEXT: PseudoRET
+ ;
+ ; RV64-LABEL: name: test_args_nxv64i8
+ ; RV64: bb.1.entry:
+ ; RV64-NEXT: liveins: $v8m8
+ ; RV64-NEXT: {{ $}}
+ ; RV64-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 64 x s8>) = COPY $v8m8
+ ; RV64-NEXT: PseudoRET
+entry:
+ ret void
+}
+
+define void @test_args_nxv1i16(<vscale x 1 x i16> %a) {
+ ; RV32-LABEL: name: test_args_nxv1i16
+ ; RV32: bb.1.entry:
+ ; RV32-NEXT: liveins: $v8
+ ; RV32-NEXT: {{ $}}
+ ; RV32-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 1 x s16>) = COPY $v8
+ ; RV32-NEXT: PseudoRET
+ ;
+ ; RV64-LABEL: name: test_args_nxv1i16
+ ; RV64: bb.1.entry:
+ ; RV64-NEXT: liveins: $v8
+ ; RV64-NEXT: {{ $}}
+ ; RV64-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 1 x s16>) = COPY $v8
+ ; RV64-NEXT: PseudoRET
+entry:
+ ret void
+}
+
+define void @test_args_nxv2i16(<vscale x 2 x i16> %a) {
+ ; RV32-LABEL: name: test_args_nxv2i16
+ ; RV32: bb.1.entry:
+ ; RV32-NEXT: liveins: $v8
+ ; RV32-NEXT: {{ $}}
+ ; RV32-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 2 x s16>) = COPY $v8
+ ; RV32-NEXT: PseudoRET
+ ;
+ ; RV64-LABEL: name: test_args_nxv2i16
+ ; RV64: bb.1.entry:
+ ; RV64-NEXT: liveins: $v8
+ ; RV64-NEXT: {{ $}}
+ ; RV64-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 2 x s16>) = COPY $v8
+ ; RV64-NEXT: PseudoRET
+entry:
+ ret void
+}
+
+define void @test_args_nxv4i16(<vscale x 4 x i16> %a) {
+ ; RV32-LABEL: name: test_args_nxv4i16
+ ; RV32: bb.1.entry:
+ ; RV32-NEXT: liveins: $v8
+ ; RV32-NEXT: {{ $}}
+ ; RV32-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 4 x s16>) = COPY $v8
+ ; RV32-NEXT: PseudoRET
+ ;
+ ; RV64-LABEL: name: test_args_nxv4i16
+ ; RV64: bb.1.entry:
+ ; RV64-NEXT: liveins: $v8
+ ; RV64-NEXT: {{ $}}
+ ; RV64-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 4 x s16>) = COPY $v8
+ ; RV64-NEXT: PseudoRET
+entry:
+ ret void
+}
+
+define void @test_args_nxv8i16(<vscale x 8 x i16> %a) {
+ ; RV32-LABEL: name: test_args_nxv8i16
+ ; RV32: bb.1.entry:
+ ; RV32-NEXT: liveins: $v8m2
+ ; RV32-NEXT: {{ $}}
+ ; RV32-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 8 x s16>) = COPY $v8m2
+ ; RV32-NEXT: PseudoRET
+ ;
+ ; RV64-LABEL: name: test_args_nxv8i16
+ ; RV64: bb.1.entry:
+ ; RV64-NEXT: liveins: $v8m2
+ ; RV64-NEXT: {{ $}}
+ ; RV64-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 8 x s16>) = COPY $v8m2
+ ; RV64-NEXT: PseudoRET
+entry:
+ ret void
+}
+
+define void @test_args_nxv16i16(<vscale x 16 x i16> %a) {
+ ; RV32-LABEL: name: test_args_nxv16i16
+ ; RV32: bb.1.entry:
+ ; RV32-NEXT: liveins: $v8m4
+ ; RV32-NEXT: {{ $}}
+ ; RV32-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 16 x s16>) = COPY $v8m4
+ ; RV32-NEXT: PseudoRET
+ ;
+ ; RV64-LABEL: name: test_args_nxv16i16
+ ; RV64: bb.1.entry:
+ ; RV64-NEXT: liveins: $v8m4
+ ; RV64-NEXT: {{ $}}
+ ; RV64-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 16 x s16>) = COPY $v8m4
+ ; RV64-NEXT: PseudoRET
+entry:
+ ret void
+}
+
+define void @test_args_nxv32i16(<vscale x 32 x i16> %a) {
+ ; RV32-LABEL: name: test_args_nxv32i16
+ ; RV32: bb.1.entry:
+ ; RV32-NEXT: liveins: $v8m8
+ ; RV32-NEXT: {{ $}}
+ ; RV32-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 32 x s16>) = COPY $v8m8
+ ; RV32-NEXT: PseudoRET
+ ;
+ ; RV64-LABEL: name: test_args_nxv32i16
+ ; RV64: bb.1.entry:
+ ; RV64-NEXT: liveins: $v8m8
+ ; RV64-NEXT: {{ $}}
+ ; RV64-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 32 x s16>) = COPY $v8m8
+ ; RV64-NEXT: PseudoRET
+entry:
+ ret void
+}
+
+define void @test_args_nxv1i32(<vscale x 1 x i32> %a) {
+ ; RV32-LABEL: name: test_args_nxv1i32
+ ; RV32: bb.1.entry:
+ ; RV32-NEXT: liveins: $v8
+ ; RV32-NEXT: {{ $}}
+ ; RV32-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 1 x s32>) = COPY $v8
+ ; RV32-NEXT: PseudoRET
+ ;
+ ; RV64-LABEL: name: test_args_nxv1i32
+ ; RV64: bb.1.entry:
+ ; RV64-NEXT: liveins: $v8
+ ; RV64-NEXT: {{ $}}
+ ; RV64-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 1 x s32>) = COPY $v8
+ ; RV64-NEXT: PseudoRET
+entry:
+ ret void
+}
+
+define void @test_args_nxv2i32(<vscale x 2 x i32> %a) {
+ ; RV32-LABEL: name: test_args_nxv2i32
+ ; RV32: bb.1.entry:
+ ; RV32-NEXT: liveins: $v8
+ ; RV32-NEXT: {{ $}}
+ ; RV32-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 2 x s32>) = COPY $v8
+ ; RV32-NEXT: PseudoRET
+ ;
+ ; RV64-LABEL: name: test_args_nxv2i32
+ ; RV64: bb.1.entry:
+ ; RV64-NEXT: liveins: $v8
+ ; RV64-NEXT: {{ $}}
+ ; RV64-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 2 x s32>) = COPY $v8
+ ; RV64-NEXT: PseudoRET
+entry:
+ ret void
+}
+
+define void @test_args_nxv4i32(<vscale x 4 x i32> %a) {
+ ; RV32-LABEL: name: test_args_nxv4i32
+ ; RV32: bb.1.entry:
+ ; RV32-NEXT: liveins: $v8m2
+ ; RV32-NEXT: {{ $}}
+ ; RV32-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 4 x s32>) = COPY $v8m2
+ ; RV32-NEXT: PseudoRET
+ ;
+ ; RV64-LABEL: name: test_args_nxv4i32
+ ; RV64: bb.1.entry:
+ ; RV64-NEXT: liveins: $v8m2
+ ; RV64-NEXT: {{ $}}
+ ; RV64-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 4 x s32>) = COPY $v8m2
+ ; RV64-NEXT: PseudoRET
+entry:
+ ret void
+}
+
+define void @test_args_nxv8i32(<vscale x 8 x i32> %a) {
+ ; RV32-LABEL: name: test_args_nxv8i32
+ ; RV32: bb.1.entry:
+ ; RV32-NEXT: liveins: $v8m4
+ ; RV32-NEXT: {{ $}}
+ ; RV32-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 8 x s32>) = COPY $v8m4
+ ; RV32-NEXT: PseudoRET
+ ;
+ ; RV64-LABEL: name: test_args_nxv8i32
+ ; RV64: bb.1.entry:
+ ; RV64-NEXT: liveins: $v8m4
+ ; RV64-NEXT: {{ $}}
+ ; RV64-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 8 x s32>) = COPY $v8m4
+ ; RV64-NEXT: PseudoRET
+entry:
+ ret void
+}
+
+define void @test_args_nxv16i32(<vscale x 16 x i32> %a) {
+ ; RV32-LABEL: name: test_args_nxv16i32
+ ; RV32: bb.1.entry:
+ ; RV32-NEXT: liveins: $v8m8
+ ; RV32-NEXT: {{ $}}
+ ; RV32-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 16 x s32>) = COPY $v8m8
+ ; RV32-NEXT: PseudoRET
+ ;
+ ; RV64-LABEL: name: test_args_nxv16i32
+ ; RV64: bb.1.entry:
+ ; RV64-NEXT: liveins: $v8m8
+ ; RV64-NEXT: {{ $}}
+ ; RV64-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 16 x s32>) = COPY $v8m8
+ ; RV64-NEXT: PseudoRET
+entry:
+ ret void
+}
+
+define void @test_args_nxv1i64(<vscale x 1 x i64> %a) {
+ ; RV32-LABEL: name: test_args_nxv1i64
+ ; RV32: bb.1.entry:
+ ; RV32-NEXT: liveins: $v8
+ ; RV32-NEXT: {{ $}}
+ ; RV32-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 1 x s64>) = COPY $v8
+ ; RV32-NEXT: PseudoRET
+ ;
+ ; RV64-LABEL: name: test_args_nxv1i64
+ ; RV64: bb.1.entry:
+ ; RV64-NEXT: liveins: $v8
+ ; RV64-NEXT: {{ $}}
+ ; RV64-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 1 x s64>) = COPY $v8
+ ; RV64-NEXT: PseudoRET
+entry:
+ ret void
+}
+
+define void @test_args_nxv2i64(<vscale x 2 x i64> %a) {
+ ; RV32-LABEL: name: test_args_nxv2i64
+ ; RV32: bb.1.entry:
+ ; RV32-NEXT: liveins: $v8m2
+ ; RV32-NEXT: {{ $}}
+ ; RV32-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 2 x s64>) = COPY $v8m2
+ ; RV32-NEXT: PseudoRET
+ ;
+ ; RV64-LABEL: name: test_args_nxv2i64
+ ; RV64: bb.1.entry:
+ ; RV64-NEXT: liveins: $v8m2
+ ; RV64-NEXT: {{ $}}
+ ; RV64-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 2 x s64>) = COPY $v8m2
+ ; RV64-NEXT: PseudoRET
+entry:
+ ret void
+}
+
+define void @test_args_nxv4i64(<vscale x 4 x i64> %a) {
+ ; RV32-LABEL: name: test_args_nxv4i64
+ ; RV32: bb.1.entry:
+ ; RV32-NEXT: liveins: $v8m4
+ ; RV32-NEXT: {{ $}}
+ ; RV32-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 4 x s64>) = COPY $v8m4
+ ; RV32-NEXT: PseudoRET
+ ;
+ ; RV64-LABEL: name: test_args_nxv4i64
+ ; RV64: bb.1.entry:
+ ; RV64-NEXT: liveins: $v8m4
+ ; RV64-NEXT: {{ $}}
+ ; RV64-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 4 x s64>) = COPY $v8m4
+ ; RV64-NEXT: PseudoRET
+entry:
+ ret void
+}
+
+define void @test_args_nxv8i64(<vscale x 8 x i64> %a) {
+ ; RV32-LABEL: name: test_args_nxv8i64
+ ; RV32: bb.1.entry:
+ ; RV32-NEXT: liveins: $v8m8
+ ; RV32-NEXT: {{ $}}
+ ; RV32-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 8 x s64>) = COPY $v8m8
+ ; RV32-NEXT: PseudoRET
+ ;
+ ; RV64-LABEL: name: test_args_nxv8i64
+ ; RV64: bb.1.entry:
+ ; RV64-NEXT: liveins: $v8m8
+ ; RV64-NEXT: {{ $}}
+ ; RV64-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 8 x s64>) = COPY $v8m8
+ ; RV64-NEXT: PseudoRET
+entry:
+ ret void
+}
+
+define void @test_args_nxv64i1(<vscale x 64 x i1> %a) {
+ ; RV32-LABEL: name: test_args_nxv64i1
+ ; RV32: bb.1.entry:
+ ; RV32-NEXT: liveins: $v8
+ ; RV32-NEXT: {{ $}}
+ ; RV32-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 64 x s1>) = COPY $v8
+ ; RV32-NEXT: PseudoRET
+ ;
+ ; RV64-LABEL: name: test_args_nxv64i1
+ ; RV64: bb.1.entry:
+ ; RV64-NEXT: liveins: $v8
+ ; RV64-NEXT: {{ $}}
+ ; RV64-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 64 x s1>) = COPY $v8
+ ; RV64-NEXT: PseudoRET
+entry:
+ ret void
+}
+
+define void @test_args_nxv32i1(<vscale x 32 x i1> %a) {
+ ; RV32-LABEL: name: test_args_nxv32i1
+ ; RV32: bb.1.entry:
+ ; RV32-NEXT: liveins: $v8
+ ; RV32-NEXT: {{ $}}
+ ; RV32-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 32 x s1>) = COPY $v8
+ ; RV32-NEXT: PseudoRET
+ ;
+ ; RV64-LABEL: name: test_args_nxv32i1
+ ; RV64: bb.1.entry:
+ ; RV64-NEXT: liveins: $v8
+ ; RV64-NEXT: {{ $}}
+ ; RV64-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 32 x s1>) = COPY $v8
+ ; RV64-NEXT: PseudoRET
+entry:
+ ret void
+}
+
+define void @test_args_nxv16i1(<vscale x 16 x i1> %a) {
+ ; RV32-LABEL: name: test_args_nxv16i1
+ ; RV32: bb.1.entry:
+ ; RV32-NEXT: liveins: $v8
+ ; RV32-NEXT: {{ $}}
+ ; RV32-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 16 x s1>) = COPY $v8
+ ; RV32-NEXT: PseudoRET
+ ;
+ ; RV64-LABEL: name: test_args_nxv16i1
+ ; RV64: bb.1.entry:
+ ; RV64-NEXT: liveins: $v8
+ ; RV64-NEXT: {{ $}}
+ ; RV64-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 16 x s1>) = COPY $v8
+ ; RV64-NEXT: PseudoRET
+entry:
+ ret void
+}
+
+define void @test_args_nxv8i1(<vscale x 8 x i1> %a) {
+ ; RV32-LABEL: name: test_args_nxv8i1
+ ; RV32: bb.1.entry:
+ ; RV32-NEXT: liveins: $v8
+ ; RV32-NEXT: {{ $}}
+ ; RV32-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 8 x s1>) = COPY $v8
+ ; RV32-NEXT: PseudoRET
+ ;
+ ; RV64-LABEL: name: test_args_nxv8i1
+ ; RV64: bb.1.entry:
+ ; RV64-NEXT: liveins: $v8
+ ; RV64-NEXT: {{ $}}
+ ; RV64-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 8 x s1>) = COPY $v8
+ ; RV64-NEXT: PseudoRET
+entry:
+ ret void
+}
+
+define void @test_args_nxv4i1(<vscale x 4 x i1> %a) {
+ ; RV32-LABEL: name: test_args_nxv4i1
+ ; RV32: bb.1.entry:
+ ; RV32-NEXT: liveins: $v8
+ ; RV32-NEXT: {{ $}}
+ ; RV32-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 4 x s1>) = COPY $v8
+ ; RV32-NEXT: PseudoRET
+ ;
+ ; RV64-LABEL: name: test_args_nxv4i1
+ ; RV64: bb.1.entry:
+ ; RV64-NEXT: liveins: $v8
+ ; RV64-NEXT: {{ $}}
+ ; RV64-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 4 x s1>) = COPY $v8
+ ; RV64-NEXT: PseudoRET
+entry:
+ ret void
+}
+
+define void @test_args_nxv2i1(<vscale x 2 x i1> %a) {
+ ; RV32-LABEL: name: test_args_nxv2i1
+ ; RV32: bb.1.entry:
+ ; RV32-NEXT: liveins: $v8
+ ; RV32-NEXT: {{ $}}
+ ; RV32-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 2 x s1>) = COPY $v8
+ ; RV32-NEXT: PseudoRET
+ ;
+ ; RV64-LABEL: name: test_args_nxv2i1
+ ; RV64: bb.1.entry:
+ ; RV64-NEXT: liveins: $v8
+ ; RV64-NEXT: {{ $}}
+ ; RV64-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 2 x s1>) = COPY $v8
+ ; RV64-NEXT: PseudoRET
+entry:
+ ret void
+}
+
+define void @test_args_nxv1i1(<vscale x 1 x i1> %a) {
+ ; RV32-LABEL: name: test_args_nxv1i1
+ ; RV32: bb.1.entry:
+ ; RV32-NEXT: liveins: $v8
+ ; RV32-NEXT: {{ $}}
+ ; RV32-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 1 x s1>) = COPY $v8
+ ; RV32-NEXT: PseudoRET
+ ;
+ ; RV64-LABEL: name: test_args_nxv1i1
+ ; RV64: bb.1.entry:
+ ; RV64-NEXT: liveins: $v8
+ ; RV64-NEXT: {{ $}}
+ ; RV64-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 1 x s1>) = COPY $v8
+ ; RV64-NEXT: PseudoRET
+entry:
+ ret void
+}
+
+define void @test_args_nxv1f32(<vscale x 1 x float> %a) {
+ ; RV32-LABEL: name: test_args_nxv1f32
+ ; RV32: bb.1.entry:
+ ; RV32-NEXT: liveins: $v8
+ ; RV32-NEXT: {{ $}}
+ ; RV32-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 1 x s32>) = COPY $v8
+ ; RV32-NEXT: PseudoRET
+ ;
+ ; RV64-LABEL: name: test_args_nxv1f32
+ ; RV64: bb.1.entry:
+ ; RV64-NEXT: liveins: $v8
+ ; RV64-NEXT: {{ $}}
+ ; RV64-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 1 x s32>) = COPY $v8
+ ; RV64-NEXT: PseudoRET
+entry:
+ ret void
+}
+
+define void @test_args_nxv2f32(<vscale x 2 x float> %a) {
+ ; RV32-LABEL: name: test_args_nxv2f32
+ ; RV32: bb.1.entry:
+ ; RV32-NEXT: liveins: $v8
+ ; RV32-NEXT: {{ $}}
+ ; RV32-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 2 x s32>) = COPY $v8
+ ; RV32-NEXT: PseudoRET
+ ;
+ ; RV64-LABEL: name: test_args_nxv2f32
+ ; RV64: bb.1.entry:
+ ; RV64-NEXT: liveins: $v8
+ ; RV64-NEXT: {{ $}}
+ ; RV64-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 2 x s32>) = COPY $v8
+ ; RV64-NEXT: PseudoRET
+entry:
+ ret void
+}
+
+define void @test_args_nxv4f32(<vscale x 4 x float> %a) {
+ ; RV32-LABEL: name: test_args_nxv4f32
+ ; RV32: bb.1.entry:
+ ; RV32-NEXT: liveins: $v8m2
+ ; RV32-NEXT: {{ $}}
+ ; RV32-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 4 x s32>) = COPY $v8m2
+ ; RV32-NEXT: PseudoRET
+ ;
+ ; RV64-LABEL: name: test_args_nxv4f32
+ ; RV64: bb.1.entry:
+ ; RV64-NEXT: liveins: $v8m2
+ ; RV64-NEXT: {{ $}}
+ ; RV64-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 4 x s32>) = COPY $v8m2
+ ; RV64-NEXT: PseudoRET
+entry:
+ ret void
+}
+
+define void @test_args_nxv8f32(<vscale x 8 x float> %a) {
+ ; RV32-LABEL: name: test_args_nxv8f32
+ ; RV32: bb.1.entry:
+ ; RV32-NEXT: liveins: $v8m4
+ ; RV32-NEXT: {{ $}}
+ ; RV32-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 8 x s32>) = COPY $v8m4
+ ; RV32-NEXT: PseudoRET
+ ;
+ ; RV64-LABEL: name: test_args_nxv8f32
+ ; RV64: bb.1.entry:
+ ; RV64-NEXT: liveins: $v8m4
+ ; RV64-NEXT: {{ $}}
+ ; RV64-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 8 x s32>) = COPY $v8m4
+ ; RV64-NEXT: PseudoRET
+entry:
+ ret void
+}
+
+define void @test_args_nxv16f32(<vscale x 16 x float> %a) {
+ ; RV32-LABEL: name: test_args_nxv16f32
+ ; RV32: bb.1.entry:
+ ; RV32-NEXT: liveins: $v8m8
+ ; RV32-NEXT: {{ $}}
+ ; RV32-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 16 x s32>) = COPY $v8m8
+ ; RV32-NEXT: PseudoRET
+ ;
+ ; RV64-LABEL: name: test_args_nxv16f32
+ ; RV64: bb.1.entry:
+ ; RV64-NEXT: liveins: $v8m8
+ ; RV64-NEXT: {{ $}}
+ ; RV64-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 16 x s32>) = COPY $v8m8
+ ; RV64-NEXT: PseudoRET
+entry:
+ ret void
+}
+
+define void @test_args_nxv1f64(<vscale x 1 x double> %a) {
+ ; RV32-LABEL: name: test_args_nxv1f64
+ ; RV32: bb.1.entry:
+ ; RV32-NEXT: liveins: $v8
+ ; RV32-NEXT: {{ $}}
+ ; RV32-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 1 x s64>) = COPY $v8
+ ; RV32-NEXT: PseudoRET
+ ;
+ ; RV64-LABEL: name: test_args_nxv1f64
+ ; RV64: bb.1.entry:
+ ; RV64-NEXT: liveins: $v8
+ ; RV64-NEXT: {{ $}}
+ ; RV64-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 1 x s64>) = COPY $v8
+ ; RV64-NEXT: PseudoRET
+entry:
+ ret void
+}
+
+define void @test_args_nxv2f64(<vscale x 2 x double> %a) {
+ ; RV32-LABEL: name: test_args_nxv2f64
+ ; RV32: bb.1.entry:
+ ; RV32-NEXT: liveins: $v8m2
+ ; RV32-NEXT: {{ $}}
+ ; RV32-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 2 x s64>) = COPY $v8m2
+ ; RV32-NEXT: PseudoRET
+ ;
+ ; RV64-LABEL: name: test_args_nxv2f64
+ ; RV64: bb.1.entry:
+ ; RV64-NEXT: liveins: $v8m2
+ ; RV64-NEXT: {{ $}}
+ ; RV64-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 2 x s64>) = COPY $v8m2
+ ; RV64-NEXT: PseudoRET
+entry:
+ ret void
+}
+
+define void @test_args_nxv4f64(<vscale x 4 x double> %a) {
+ ; RV32-LABEL: name: test_args_nxv4f64
+ ; RV32: bb.1.entry:
+ ; RV32-NEXT: liveins: $v8m4
+ ; RV32-NEXT: {{ $}}
+ ; RV32-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 4 x s64>) = COPY $v8m4
+ ; RV32-NEXT: PseudoRET
+ ;
+ ; RV64-LABEL: name: test_args_nxv4f64
+ ; RV64: bb.1.entry:
+ ; RV64-NEXT: liveins: $v8m4
+ ; RV64-NEXT: {{ $}}
+ ; RV64-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 4 x s64>) = COPY $v8m4
+ ; RV64-NEXT: PseudoRET
+entry:
+ ret void
+}
+
+define void @test_args_nxv8f64(<vscale x 8 x double> %a) {
+ ; RV32-LABEL: name: test_args_nxv8f64
+ ; RV32: bb.1.entry:
+ ; RV32-NEXT: liveins: $v8m8
+ ; RV32-NEXT: {{ $}}
+ ; RV32-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 8 x s64>) = COPY $v8m8
+ ; RV32-NEXT: PseudoRET
+ ;
+ ; RV64-LABEL: name: test_args_nxv8f64
+ ; RV64: bb.1.entry:
+ ; RV64-NEXT: liveins: $v8m8
+ ; RV64-NEXT: {{ $}}
+ ; RV64-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 8 x s64>) = COPY $v8m8
+ ; RV64-NEXT: PseudoRET
+entry:
+ ret void
+}
+
+define void @test_args_nxv1f16(<vscale x 1 x half> %a) {
+ ; RV32-LABEL: name: test_args_nxv1f16
+ ; RV32: bb.1.entry:
+ ; RV32-NEXT: liveins: $v8
+ ; RV32-NEXT: {{ $}}
+ ; RV32-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 1 x s16>) = COPY $v8
+ ; RV32-NEXT: PseudoRET
+ ;
+ ; RV64-LABEL: name: test_args_nxv1f16
+ ; RV64: bb.1.entry:
+ ; RV64-NEXT: liveins: $v8
+ ; RV64-NEXT: {{ $}}
+ ; RV64-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 1 x s16>) = COPY $v8
+ ; RV64-NEXT: PseudoRET
+entry:
+ ret void
+}
+
+define void @test_args_nxv2f16(<vscale x 2 x half> %a) {
+ ; RV32-LABEL: name: test_args_nxv2f16
+ ; RV32: bb.1.entry:
+ ; RV32-NEXT: liveins: $v8
+ ; RV32-NEXT: {{ $}}
+ ; RV32-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 2 x s16>) = COPY $v8
+ ; RV32-NEXT: PseudoRET
+ ;
+ ; RV64-LABEL: name: test_args_nxv2f16
+ ; RV64: bb.1.entry:
+ ; RV64-NEXT: liveins: $v8
+ ; RV64-NEXT: {{ $}}
+ ; RV64-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 2 x s16>) = COPY $v8
+ ; RV64-NEXT: PseudoRET
+entry:
+ ret void
+}
+
+define void @test_args_nxv4f16(<vscale x 4 x half> %a) {
+ ; RV32-LABEL: name: test_args_nxv4f16
+ ; RV32: bb.1.entry:
+ ; RV32-NEXT: liveins: $v8
+ ; RV32-NEXT: {{ $}}
+ ; RV32-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 4 x s16>) = COPY $v8
+ ; RV32-NEXT: PseudoRET
+ ;
+ ; RV64-LABEL: name: test_args_nxv4f16
+ ; RV64: bb.1.entry:
+ ; RV64-NEXT: liveins: $v8
+ ; RV64-NEXT: {{ $}}
+ ; RV64-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 4 x s16>) = COPY $v8
+ ; RV64-NEXT: PseudoRET
+entry:
+ ret void
+}
+
+define void @test_args_nxv8f16(<vscale x 8 x half> %a) {
+ ; RV32-LABEL: name: test_args_nxv8f16
+ ; RV32: bb.1.entry:
+ ; RV32-NEXT: liveins: $v8m2
+ ; RV32-NEXT: {{ $}}
+ ; RV32-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 8 x s16>) = COPY $v8m2
+ ; RV32-NEXT: PseudoRET
+ ;
+ ; RV64-LABEL: name: test_args_nxv8f16
+ ; RV64: bb.1.entry:
+ ; RV64-NEXT: liveins: $v8m2
+ ; RV64-NEXT: {{ $}}
+ ; RV64-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 8 x s16>) = COPY $v8m2
+ ; RV64-NEXT: PseudoRET
+entry:
+ ret void
+}
+
+define void @test_args_nxv16f16(<vscale x 16 x half> %a) {
+ ; RV32-LABEL: name: test_args_nxv16f16
+ ; RV32: bb.1.entry:
+ ; RV32-NEXT: liveins: $v8m4
+ ; RV32-NEXT: {{ $}}
+ ; RV32-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 16 x s16>) = COPY $v8m4
+ ; RV32-NEXT: PseudoRET
+ ;
+ ; RV64-LABEL: name: test_args_nxv16f16
+ ; RV64: bb.1.entry:
+ ; RV64-NEXT: liveins: $v8m4
+ ; RV64-NEXT: {{ $}}
+ ; RV64-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 16 x s16>) = COPY $v8m4
+ ; RV64-NEXT: PseudoRET
+entry:
+ ret void
+}
+
+define void @test_args_nxv32f16(<vscale x 32 x half> %a) {
+ ; RV32-LABEL: name: test_args_nxv32f16
+ ; RV32: bb.1.entry:
+ ; RV32-NEXT: liveins: $v8m8
+ ; RV32-NEXT: {{ $}}
+ ; RV32-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 32 x s16>) = COPY $v8m8
+ ; RV32-NEXT: PseudoRET
+ ;
+ ; RV64-LABEL: name: test_args_nxv32f16
+ ; RV64: bb.1.entry:
+ ; RV64-NEXT: liveins: $v8m8
+ ; RV64-NEXT: {{ $}}
+ ; RV64-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 32 x s16>) = COPY $v8m8
+ ; RV64-NEXT: PseudoRET
+entry:
+ ret void
+}
+
+define void @test_args_nxv1b16(<vscale x 1 x bfloat> %a) {
+ ; RV32-LABEL: name: test_args_nxv1b16
+ ; RV32: bb.1.entry:
+ ; RV32-NEXT: liveins: $v8
+ ; RV32-NEXT: {{ $}}
+ ; RV32-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 1 x s16>) = COPY $v8
+ ; RV32-NEXT: PseudoRET
+ ;
+ ; RV64-LABEL: name: test_args_nxv1b16
+ ; RV64: bb.1.entry:
+ ; RV64-NEXT: liveins: $v8
+ ; RV64-NEXT: {{ $}}
+ ; RV64-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 1 x s16>) = COPY $v8
+ ; RV64-NEXT: PseudoRET
+entry:
+ ret void
+}
+
+define void @test_args_nxv2b16(<vscale x 2 x bfloat> %a) {
+ ; RV32-LABEL: name: test_args_nxv2b16
+ ; RV32: bb.1.entry:
+ ; RV32-NEXT: liveins: $v8
+ ; RV32-NEXT: {{ $}}
+ ; RV32-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 2 x s16>) = COPY $v8
+ ; RV32-NEXT: PseudoRET
+ ;
+ ; RV64-LABEL: name: test_args_nxv2b16
+ ; RV64: bb.1.entry:
+ ; RV64-NEXT: liveins: $v8
+ ; RV64-NEXT: {{ $}}
+ ; RV64-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 2 x s16>) = COPY $v8
+ ; RV64-NEXT: PseudoRET
+entry:
+ ret void
+}
+
+define void @test_args_nxv4b16(<vscale x 4 x bfloat> %a) {
+ ; RV32-LABEL: name: test_args_nxv4b16
+ ; RV32: bb.1.entry:
+ ; RV32-NEXT: liveins: $v8
+ ; RV32-NEXT: {{ $}}
+ ; RV32-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 4 x s16>) = COPY $v8
+ ; RV32-NEXT: PseudoRET
+ ;
+ ; RV64-LABEL: name: test_args_nxv4b16
+ ; RV64: bb.1.entry:
+ ; RV64-NEXT: liveins: $v8
+ ; RV64-NEXT: {{ $}}
+ ; RV64-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 4 x s16>) = COPY $v8
+ ; RV64-NEXT: PseudoRET
+entry:
+ ret void
+}
+
+define void @test_args_nxv8b16(<vscale x 8 x bfloat> %a) {
+ ; RV32-LABEL: name: test_args_nxv8b16
+ ; RV32: bb.1.entry:
+ ; RV32-NEXT: liveins: $v8m2
+ ; RV32-NEXT: {{ $}}
+ ; RV32-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 8 x s16>) = COPY $v8m2
+ ; RV32-NEXT: PseudoRET
+ ;
+ ; RV64-LABEL: name: test_args_nxv8b16
+ ; RV64: bb.1.entry:
+ ; RV64-NEXT: liveins: $v8m2
+ ; RV64-NEXT: {{ $}}
+ ; RV64-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 8 x s16>) = COPY $v8m2
+ ; RV64-NEXT: PseudoRET
+entry:
+ ret void
+}
+
+define void @test_args_nxv16b16(<vscale x 16 x bfloat> %a) {
+ ; RV32-LABEL: name: test_args_nxv16b16
+ ; RV32: bb.1.entry:
+ ; RV32-NEXT: liveins: $v8m4
+ ; RV32-NEXT: {{ $}}
+ ; RV32-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 16 x s16>) = COPY $v8m4
+ ; RV32-NEXT: PseudoRET
+ ;
+ ; RV64-LABEL: name: test_args_nxv16b16
+ ; RV64: bb.1.entry:
+ ; RV64-NEXT: liveins: $v8m4
+ ; RV64-NEXT: {{ $}}
+ ; RV64-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 16 x s16>) = COPY $v8m4
+ ; RV64-NEXT: PseudoRET
+entry:
+ ret void
+}
+
+define void @test_args_nxv32b16(<vscale x 32 x bfloat> %a) {
+ ; RV32-LABEL: name: test_args_nxv32b16
+ ; RV32: bb.1.entry:
+ ; RV32-NEXT: liveins: $v8m8
+ ; RV32-NEXT: {{ $}}
+ ; RV32-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 32 x s16>) = COPY $v8m8
+ ; RV32-NEXT: PseudoRET
+ ;
+ ; RV64-LABEL: name: test_args_nxv32b16
+ ; RV64: bb.1.entry:
+ ; RV64-NEXT: liveins: $v8m8
+ ; RV64-NEXT: {{ $}}
+ ; RV64-NEXT: [[COPY:%[0-9]+]]:_(<vscale x 32 x s16>) = COPY $v8m8
+ ; RV64-NEXT: PseudoRET
+entry:
+ ret void
+}
>From 05f3f4c27b4796eac39f87441aa6cb67662a6059 Mon Sep 17 00:00:00 2001
From: jiahanxie353 <jx353 at cornell.edu>
Date: Sat, 4 Nov 2023 11:33:20 -0400
Subject: [PATCH 3/4] [RISCV][GlobalISel] Vector Extension vadd Legalizer; test
on a mimimum example of vadd-vv
---
llvm/test/CodeGen/RISCV/rvv/vadd-vv.ll | 14 ++++++++++++++
1 file changed, 14 insertions(+)
create mode 100644 llvm/test/CodeGen/RISCV/rvv/vadd-vv.ll
diff --git a/llvm/test/CodeGen/RISCV/rvv/vadd-vv.ll b/llvm/test/CodeGen/RISCV/rvv/vadd-vv.ll
new file mode 100644
index 000000000000000..1e431707cd156e7
--- /dev/null
+++ b/llvm/test/CodeGen/RISCV/rvv/vadd-vv.ll
@@ -0,0 +1,14 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=riscv32 -mattr=+v -stop-after=irtranslator | FileCheck %s --check-prefixes=CHECK,RV32
+; RUN: llc -mtriple=riscv64 -mattr=+v --stop-after=irtranslator | FileCheck %s --check-prefixes=CHECK,RV64
+
+define <vscale x 1 x i8> @vadd_vv_mask_nxv1i8(<vscale x 1 x i8> %va, <vscale x 1 x i8> %vb, <vscale x 1 x i1> %mask) {
+; CHECK-LABEL: vadd_vv_mask_nxv1i8:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vsetvli a0, zero, e4, m2, ta, mu
+; CHECK-NEXT: vadd.vv v8, v8, v12, v0.t
+; CHECK-NEXT: ret
+ %vs = select <vscale x 1 x i1> %mask, <vscale x 1 x i8> %vb, <vscale x 1 x i8> zeroinitializer
+ %vc = add <vscale x 1 x i8> %va, %vs
+ ret <vscale x 1 x i8> %vc
+}
>From a029a29217a5170fb8fc50e379de95478f4eaaee Mon Sep 17 00:00:00 2001
From: jiahanxie353 <jx353 at cornell.edu>
Date: Mon, 6 Nov 2023 13:13:46 -0500
Subject: [PATCH 4/4] minimum exaple to start testing legalization
---
.../RISCV/GlobalISel/legalizer/rv32/legalize-vadd.mir | 9 +++++++++
1 file changed, 9 insertions(+)
create mode 100644 llvm/test/CodeGen/RISCV/GlobalISel/legalizer/rv32/legalize-vadd.mir
diff --git a/llvm/test/CodeGen/RISCV/GlobalISel/legalizer/rv32/legalize-vadd.mir b/llvm/test/CodeGen/RISCV/GlobalISel/legalizer/rv32/legalize-vadd.mir
new file mode 100644
index 000000000000000..57bd72e218571b3
--- /dev/null
+++ b/llvm/test/CodeGen/RISCV/GlobalISel/legalizer/rv32/legalize-vadd.mir
@@ -0,0 +1,9 @@
+# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
+# RUN: llc -mtriple=riscv32 -mattr=+v -run-pass=legalizer %s -o - \
+# RUN: | FileCheck %s
+
+---
+name: test_vadd
+body: |
+ bb.0.entry:
+ ; CHECK-LABEL: name: test_vadd
More information about the llvm-commits
mailing list