[RISCV] Collapse fast unaligned access into a single feature [nfc-ish] #73971
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
When we'd originally added unaligned-scalar-mem and unaligned-vector-mem, they were separated into two parts under the theory that some processor might implement one, but not the other. At the moment, we don't have evidence of such a processor. The C/C++ level interface, and the clang driver command lines have settled on a single unaligned flag which indicates both scalar and vector support unaligned. Given that, let's remove the test matrix complexity for a set of configurations which don't appear useful. Given these are internal feature names, I don't think we need to provide any forward compatibility. Anyone disagree? Note: The immediate trigger for this patch was finding another case where the unaligned-vector-mem wasn't being properly serialized to IR from clang which resulted in problems reproducing assembly from clang's -emit-llvm feature. Instead of fixing this, I decided getting rid of the complexity was the better approach.
llvmbot
added
clang
|
@llvm/pr-subscribers-clang @llvm/pr-subscribers-clang-driver Author: Philip Reames (preames) ChangesWhen we'd originally added unaligned-scalar-mem and unaligned-vector-mem, they were separated into two parts under the theory that some processor might implement one, but not the other. At the moment, we don't have evidence of such a processor. The C/C++ level interface, and the clang driver command lines have settled on a single unaligned flag which indicates both scalar and vector support unaligned. Given that, let's remove the test matrix complexity for a set of configurations which don't appear useful. Given these are internal feature names, I don't think we need to provide any forward compatibility. Anyone disagree? Note: The immediate trigger for this patch was finding another case where the unaligned-vector-mem wasn't being properly serialized to IR from clang which resulted in problems reproducing assembly from clang's -emit-llvm feature. Instead of fixing this, I decided getting rid of the complexity was the better approach. Patch is 21.41 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/73971.diff 20 Files Affected:
diff --git a/clang/lib/Basic/Targets/RISCV.cpp b/clang/lib/Basic/Targets/RISCV.cpp
index d1d9cc1c770e361..6ee82d6665501b9 100644
--- a/clang/lib/Basic/Targets/RISCV.cpp
+++ b/clang/lib/Basic/Targets/RISCV.cpp
@@ -377,7 +377,7 @@ bool RISCVTargetInfo::handleTargetFeatures(std::vector<std::string> &Features,
if (ISAInfo->hasExtension("zfh") || ISAInfo->hasExtension("zhinx"))
HasLegalHalfType = true;
- FastUnalignedAccess = llvm::is_contained(Features, "+unaligned-scalar-mem");
+ FastUnalignedAccess = llvm::is_contained(Features, "+fast-unaligned-access");
return true;
}
diff --git a/clang/lib/Driver/ToolChains/Arch/RISCV.cpp b/clang/lib/Driver/ToolChains/Arch/RISCV.cpp
index 346fb67ff277e9c..5db188fd7587daf 100644
--- a/clang/lib/Driver/ToolChains/Arch/RISCV.cpp
+++ b/clang/lib/Driver/ToolChains/Arch/RISCV.cpp
@@ -65,7 +65,7 @@ static void getRISCFeaturesFromMcpu(const Driver &D, const Arg *A,
}
if (llvm::RISCV::hasFastUnalignedAccess(Mcpu))
- Features.push_back("+unaligned-scalar-mem");
+ Features.push_back("+fast-unaligned-access");
}
void riscv::getRISCVTargetFeatures(const Driver &D, const llvm::Triple &Triple,
@@ -174,15 +174,10 @@ void riscv::getRISCVTargetFeatures(const Driver &D, const llvm::Triple &Triple,
bool HasV = llvm::is_contained(Features, "+zve32x");
if (const Arg *A = Args.getLastArg(options::OPT_munaligned_access,
options::OPT_mno_unaligned_access)) {
- if (A->getOption().matches(options::OPT_munaligned_access)) {
- Features.push_back("+unaligned-scalar-mem");
- if (HasV)
- Features.push_back("+unaligned-vector-mem");
- } else {
- Features.push_back("-unaligned-scalar-mem");
- if (HasV)
- Features.push_back("-unaligned-vector-mem");
- }
+ if (A->getOption().matches(options::OPT_munaligned_access))
+ Features.push_back("+fast-unaligned-access");
+ else
+ Features.push_back("-fast-unaligned-access");
}
// Now add any that the user explicitly requested on the command line,
diff --git a/clang/test/Driver/riscv-features.c b/clang/test/Driver/riscv-features.c
index 851a7c0507eb3a3..716f3f6da57b889 100644
--- a/clang/test/Driver/riscv-features.c
+++ b/clang/test/Driver/riscv-features.c
@@ -27,19 +27,13 @@
// DEFAULT: "-target-feature" "-save-restore"
// DEFAULT-NOT: "-target-feature" "+save-restore"
-// RUN: %clang --target=riscv32-unknown-elf -### %s -munaligned-access 2>&1 | FileCheck %s -check-prefix=UNALIGNED-SCALAR-MEM
-// RUN: %clang --target=riscv32-unknown-elf -### %s -mno-unaligned-access 2>&1 | FileCheck %s -check-prefix=NO-UNALIGNED-SCALAR-MEM
-// RUN: %clang --target=riscv32-unknown-elf -### %s -mno-strict-align 2>&1 | FileCheck %s -check-prefix=UNALIGNED-SCALAR-MEM
-// RUN: %clang --target=riscv32-unknown-elf -### %s -mstrict-align 2>&1 | FileCheck %s -check-prefix=NO-UNALIGNED-SCALAR-MEM
-// RUN: %clang --target=riscv32-unknown-elf -### %s -march=rv32gv -munaligned-access 2>&1 | FileCheck %s -check-prefix=UNALIGNED-VECTOR-MEM
-// RUN: %clang --target=riscv32-unknown-elf -### %s -march=rv32gv -mno-unaligned-access 2>&1 | FileCheck %s -check-prefix=NO-UNALIGNED-VECTOR-MEM
-// RUN: %clang --target=riscv32-unknown-elf -### %s -march=rv32gv -mno-strict-align 2>&1 | FileCheck %s -check-prefix=UNALIGNED-VECTOR-MEM
-// RUN: %clang --target=riscv32-unknown-elf -### %s -march=rv32gv -mstrict-align 2>&1 | FileCheck %s -check-prefix=NO-UNALIGNED-VECTOR-MEM
-
-// UNALIGNED-SCALAR-MEM: "-target-feature" "+unaligned-scalar-mem"
-// NO-UNALIGNED-SCALAR-MEM: "-target-feature" "-unaligned-scalar-mem"
-// UNALIGNED-VECTOR-MEM: "-target-feature" "+unaligned-vector-mem"
-// NO-UNALIGNED-VECTOR-MEM: "-target-feature" "-unaligned-vector-mem"
+// RUN: %clang --target=riscv32-unknown-elf -### %s -munaligned-access 2>&1 | FileCheck %s -check-prefix=FAST-UNALIGNED-ACCESS
+// RUN: %clang --target=riscv32-unknown-elf -### %s -mno-unaligned-access 2>&1 | FileCheck %s -check-prefix=NO-FAST-UNALIGNED-ACCESS
+// RUN: %clang --target=riscv32-unknown-elf -### %s -mno-strict-align 2>&1 | FileCheck %s -check-prefix=FAST-UNALIGNED-ACCESS
+// RUN: %clang --target=riscv32-unknown-elf -### %s -mstrict-align 2>&1 | FileCheck %s -check-prefix=NO-FAST-UNALIGNED-ACCESS
+
+// FAST-UNALIGNED-ACCESS: "-target-feature" "+fast-unaligned-access"
+// NO-FAST-UNALIGNED-ACCESS: "-target-feature" "-fast-unaligned-access"
// RUN: %clang --target=riscv32-linux -### %s -fsyntax-only 2>&1 \
// RUN: | FileCheck %s -check-prefix=DEFAULT-LINUX
diff --git a/llvm/lib/Target/RISCV/RISCVExpandPseudoInsts.cpp b/llvm/lib/Target/RISCV/RISCVExpandPseudoInsts.cpp
index ff35a8fc9c5a1e0..aa1bdbd8ba9c434 100644
--- a/llvm/lib/Target/RISCV/RISCVExpandPseudoInsts.cpp
+++ b/llvm/lib/Target/RISCV/RISCVExpandPseudoInsts.cpp
@@ -309,8 +309,8 @@ bool RISCVExpandPseudo::expandRV32ZdinxStore(MachineBasicBlock &MBB,
.addReg(MBBI->getOperand(1).getReg())
.add(MBBI->getOperand(2));
if (MBBI->getOperand(2).isGlobal() || MBBI->getOperand(2).isCPI()) {
- // FIXME: Zdinx RV32 can not work on unaligned scalar memory.
- assert(!STI->enableUnalignedScalarMem());
+ // FIXME: Zdinx RV32 can not work on unaligned memory.
+ assert(!STI->hasFastUnalignedAccess());
assert(MBBI->getOperand(2).getOffset() % 8 == 0);
MBBI->getOperand(2).setOffset(MBBI->getOperand(2).getOffset() + 4);
diff --git a/llvm/lib/Target/RISCV/RISCVFeatures.td b/llvm/lib/Target/RISCV/RISCVFeatures.td
index c5d88ca306969f4..7d142d38d0f9ade 100644
--- a/llvm/lib/Target/RISCV/RISCVFeatures.td
+++ b/llvm/lib/Target/RISCV/RISCVFeatures.td
@@ -946,15 +946,10 @@ def FeatureTrailingSeqCstFence : SubtargetFeature<"seq-cst-trailing-fence",
"true",
"Enable trailing fence for seq-cst store.">;
-def FeatureUnalignedScalarMem
- : SubtargetFeature<"unaligned-scalar-mem", "EnableUnalignedScalarMem",
- "true", "Has reasonably performant unaligned scalar "
- "loads and stores">;
-
-def FeatureUnalignedVectorMem
- : SubtargetFeature<"unaligned-vector-mem", "EnableUnalignedVectorMem",
- "true", "Has reasonably performant unaligned vector "
- "loads and stores">;
+def FeatureFastUnalignedAccess
+ : SubtargetFeature<"fast-unaligned-access", "HasFastUnalignedAccess",
+ "true", "Has reasonably performant unaligned "
+ "loads and stores (both scalar and vector)">;
def FeaturePostRAScheduler : SubtargetFeature<"use-postra-scheduler",
"UsePostRAScheduler", "true", "Schedule again after register allocation">;
diff --git a/llvm/lib/Target/RISCV/RISCVISelLowering.cpp b/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
index 18c6ca5348b6213..446938cf02c71cf 100644
--- a/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
+++ b/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
@@ -1873,7 +1873,7 @@ bool RISCVTargetLowering::shouldConvertConstantLoadToIntImm(const APInt &Imm,
// replace. If we don't support unaligned scalar mem, prefer the constant
// pool.
// TODO: Can the caller pass down the alignment?
- if (!Subtarget.enableUnalignedScalarMem())
+ if (!Subtarget.hasFastUnalignedAccess())
return true;
// Prefer to keep the load if it would require many instructions.
@@ -14619,7 +14619,7 @@ static bool matchIndexAsWiderOp(EVT VT, SDValue Index, SDValue Mask,
if (WiderElementSize > ST.getELen()/8)
return false;
- if (!ST.enableUnalignedVectorMem() && BaseAlign < WiderElementSize)
+ if (!ST.hasFastUnalignedAccess() && BaseAlign < WiderElementSize)
return false;
for (unsigned i = 0; i < Index->getNumOperands(); i++) {
@@ -19205,8 +19205,8 @@ bool RISCVTargetLowering::allowsMisalignedMemoryAccesses(
unsigned *Fast) const {
if (!VT.isVector()) {
if (Fast)
- *Fast = Subtarget.enableUnalignedScalarMem();
- return Subtarget.enableUnalignedScalarMem();
+ *Fast = Subtarget.hasFastUnalignedAccess();
+ return Subtarget.hasFastUnalignedAccess();
}
// All vector implementations must support element alignment
@@ -19222,8 +19222,8 @@ bool RISCVTargetLowering::allowsMisalignedMemoryAccesses(
// misaligned accesses. TODO: Work through the codegen implications of
// allowing such accesses to be formed, and considered fast.
if (Fast)
- *Fast = Subtarget.enableUnalignedVectorMem();
- return Subtarget.enableUnalignedVectorMem();
+ *Fast = Subtarget.hasFastUnalignedAccess();
+ return Subtarget.hasFastUnalignedAccess();
}
@@ -19258,7 +19258,7 @@ EVT RISCVTargetLowering::getOptimalMemOpType(const MemOp &Op,
// Do we have sufficient alignment for our preferred VT? If not, revert
// to largest size allowed by our alignment criteria.
- if (PreferredVT != MVT::i8 && !Subtarget.enableUnalignedVectorMem()) {
+ if (PreferredVT != MVT::i8 && !Subtarget.hasFastUnalignedAccess()) {
Align RequiredAlign(PreferredVT.getStoreSize());
if (Op.isFixedDstAlign())
RequiredAlign = std::min(RequiredAlign, Op.getDstAlign());
@@ -19450,7 +19450,7 @@ bool RISCVTargetLowering::isLegalStridedLoadStore(EVT DataType,
if (!isLegalElementTypeForRVV(ScalarType))
return false;
- if (!Subtarget.enableUnalignedVectorMem() &&
+ if (!Subtarget.hasFastUnalignedAccess() &&
Alignment < ScalarType.getStoreSize())
return false;
diff --git a/llvm/lib/Target/RISCV/RISCVTargetTransformInfo.h b/llvm/lib/Target/RISCV/RISCVTargetTransformInfo.h
index 75ae55e100e20ac..efc8350064a6e3f 100644
--- a/llvm/lib/Target/RISCV/RISCVTargetTransformInfo.h
+++ b/llvm/lib/Target/RISCV/RISCVTargetTransformInfo.h
@@ -199,7 +199,7 @@ class RISCVTTIImpl : public BasicTTIImplBase<RISCVTTIImpl> {
return false;
EVT ElemType = DataTypeVT.getScalarType();
- if (!ST->enableUnalignedVectorMem() && Alignment < ElemType.getStoreSize())
+ if (!ST->hasFastUnalignedAccess() && Alignment < ElemType.getStoreSize())
return false;
return TLI->isLegalElementTypeForRVV(ElemType);
@@ -224,7 +224,7 @@ class RISCVTTIImpl : public BasicTTIImplBase<RISCVTTIImpl> {
return false;
EVT ElemType = DataTypeVT.getScalarType();
- if (!ST->enableUnalignedVectorMem() && Alignment < ElemType.getStoreSize())
+ if (!ST->hasFastUnalignedAccess() && Alignment < ElemType.getStoreSize())
return false;
return TLI->isLegalElementTypeForRVV(ElemType);
diff --git a/llvm/test/CodeGen/RISCV/memcpy-inline.ll b/llvm/test/CodeGen/RISCV/memcpy-inline.ll
index 833e07351eec77d..343695ee37da84d 100644
--- a/llvm/test/CodeGen/RISCV/memcpy-inline.ll
+++ b/llvm/test/CodeGen/RISCV/memcpy-inline.ll
@@ -3,9 +3,9 @@
; RUN: | FileCheck %s --check-prefixes=RV32-BOTH,RV32
; RUN: llc < %s -mtriple=riscv64 \
; RUN: | FileCheck %s --check-prefixes=RV64-BOTH,RV64
-; RUN: llc < %s -mtriple=riscv32 -mattr=+unaligned-scalar-mem \
+; RUN: llc < %s -mtriple=riscv32 -mattr=+fast-unaligned-access \
; RUN: | FileCheck %s --check-prefixes=RV32-BOTH,RV32-FAST
-; RUN: llc < %s -mtriple=riscv64 -mattr=+unaligned-scalar-mem \
+; RUN: llc < %s -mtriple=riscv64 -mattr=+fast-unaligned-access \
; RUN: | FileCheck %s --check-prefixes=RV64-BOTH,RV64-FAST
; ----------------------------------------------------------------------
diff --git a/llvm/test/CodeGen/RISCV/memcpy.ll b/llvm/test/CodeGen/RISCV/memcpy.ll
index d831a2a002a3ede..26ad872509194f0 100644
--- a/llvm/test/CodeGen/RISCV/memcpy.ll
+++ b/llvm/test/CodeGen/RISCV/memcpy.ll
@@ -3,9 +3,9 @@
; RUN: | FileCheck %s --check-prefixes=RV32-BOTH,RV32
; RUN: llc < %s -mtriple=riscv64 \
; RUN: | FileCheck %s --check-prefixes=RV64-BOTH,RV64
-; RUN: llc < %s -mtriple=riscv32 -mattr=+unaligned-scalar-mem \
+; RUN: llc < %s -mtriple=riscv32 -mattr=+fast-unaligned-access \
; RUN: | FileCheck %s --check-prefixes=RV32-BOTH,RV32-FAST
-; RUN: llc < %s -mtriple=riscv64 -mattr=+unaligned-scalar-mem \
+; RUN: llc < %s -mtriple=riscv64 -mattr=+fast-unaligned-access \
; RUN: | FileCheck %s --check-prefixes=RV64-BOTH,RV64-FAST
%struct.x = type { i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8 }
diff --git a/llvm/test/CodeGen/RISCV/memset-inline.ll b/llvm/test/CodeGen/RISCV/memset-inline.ll
index 55fe81a58805edc..cc22b77c641e272 100644
--- a/llvm/test/CodeGen/RISCV/memset-inline.ll
+++ b/llvm/test/CodeGen/RISCV/memset-inline.ll
@@ -3,9 +3,9 @@
; RUN: | FileCheck %s --check-prefixes=RV32-BOTH,RV32
; RUN: llc < %s -mtriple=riscv64 -mattr=+m \
; RUN: | FileCheck %s --check-prefixes=RV64-BOTH,RV64
-; RUN: llc < %s -mtriple=riscv32 -mattr=+m,+unaligned-scalar-mem \
+; RUN: llc < %s -mtriple=riscv32 -mattr=+m,+fast-unaligned-access \
; RUN: | FileCheck %s --check-prefixes=RV32-BOTH,RV32-FAST
-; RUN: llc < %s -mtriple=riscv64 -mattr=+m,+unaligned-scalar-mem \
+; RUN: llc < %s -mtriple=riscv64 -mattr=+m,+fast-unaligned-access \
; RUN: | FileCheck %s --check-prefixes=RV64-BOTH,RV64-FAST
%struct.x = type { i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8 }
diff --git a/llvm/test/CodeGen/RISCV/pr56110.ll b/llvm/test/CodeGen/RISCV/pr56110.ll
index fa441f5fc3aef46..c795b17419f564e 100644
--- a/llvm/test/CodeGen/RISCV/pr56110.ll
+++ b/llvm/test/CodeGen/RISCV/pr56110.ll
@@ -1,6 +1,6 @@
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc < %s -mtriple=riscv32 | FileCheck %s
-; RUN: llc < %s -mtriple=riscv32 -mattr=+unaligned-scalar-mem | FileCheck %s
+; RUN: llc < %s -mtriple=riscv32 -mattr=+fast-unaligned-access | FileCheck %s
define void @foo_set(ptr nocapture noundef %a, i32 noundef %v) {
; CHECK-LABEL: foo_set:
diff --git a/llvm/test/CodeGen/RISCV/riscv-func-target-feature.ll b/llvm/test/CodeGen/RISCV/riscv-func-target-feature.ll
index d627ae9c90394eb..a03dadbc1d1160b 100644
--- a/llvm/test/CodeGen/RISCV/riscv-func-target-feature.ll
+++ b/llvm/test/CodeGen/RISCV/riscv-func-target-feature.ll
@@ -36,7 +36,7 @@ entry:
}
; CHECK-NOT: .option push
-define void @test5() "target-features"="+unaligned-scalar-mem" {
+define void @test5() "target-features"="+fast-unaligned-access" {
; CHECK-LABEL: test5
; CHECK-NOT: .option pop
entry:
diff --git a/llvm/test/CodeGen/RISCV/rvv/concat-vectors-constant-stride.ll b/llvm/test/CodeGen/RISCV/rvv/concat-vectors-constant-stride.ll
index ff35043dbd7e75e..f244810e739d936 100644
--- a/llvm/test/CodeGen/RISCV/rvv/concat-vectors-constant-stride.ll
+++ b/llvm/test/CodeGen/RISCV/rvv/concat-vectors-constant-stride.ll
@@ -1,7 +1,7 @@
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
-; RUN: llc -mtriple=riscv32 -mattr=+v,+unaligned-vector-mem -target-abi=ilp32 \
+; RUN: llc -mtriple=riscv32 -mattr=+v,+fast-unaligned-access -target-abi=ilp32 \
; RUN: -verify-machineinstrs < %s | FileCheck %s --check-prefixes=CHECK,RV32
-; RUN: llc -mtriple=riscv64 -mattr=+v,+unaligned-vector-mem -target-abi=lp64 \
+; RUN: llc -mtriple=riscv64 -mattr=+v,+fast-unaligned-access -target-abi=lp64 \
; RUN: -verify-machineinstrs < %s | FileCheck %s --check-prefixes=CHECK,RV64
define void @constant_forward_stride(ptr %s, ptr %d) {
diff --git a/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-strided-load-combine.ll b/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-strided-load-combine.ll
index f52ba6f51d5c897..ba5db552b8544fb 100644
--- a/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-strided-load-combine.ll
+++ b/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-strided-load-combine.ll
@@ -1,7 +1,7 @@
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 2
; RUN: llc -mtriple=riscv32 -mattr=+v,+zfh,+zvfh -verify-machineinstrs < %s | FileCheck %s -check-prefixes=CHECK,CHECK-NO-MISALIGN,RV32
; RUN: llc -mtriple=riscv64 -mattr=+v,+zfh,+zvfh -verify-machineinstrs < %s | FileCheck %s -check-prefixes=CHECK,CHECK-NO-MISALIGN,RV64
-; RUN: llc -mtriple=riscv64 -mattr=+v,+zfh,+zvfh,+unaligned-vector-mem -verify-machineinstrs < %s | FileCheck %s -check-prefixes=CHECK,RV64,RV64-MISALIGN
+; RUN: llc -mtriple=riscv64 -mattr=+v,+zfh,+zvfh,+fast-unaligned-access -verify-machineinstrs < %s | FileCheck %s -check-prefixes=CHECK,RV64,RV64-MISALIGN
; RUN: llc -mtriple=riscv64 -mattr=+f,+zfh,+zve64f,+zvl128b,+zvfh -verify-machineinstrs < %s | FileCheck %s -check-prefixes=CHECK,CHECK-NO-MISALIGN,ZVE64F
diff --git a/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-unaligned.ll b/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-unaligned.ll
index 1cc09b7f5eeb5f4..398ac70772378de 100644
--- a/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-unaligned.ll
+++ b/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-unaligned.ll
@@ -3,9 +3,9 @@
; RUN: | FileCheck %s --check-prefixes=SLOW,RV32-SLOW
; RUN: llc -mtriple=riscv64 -mattr=+m,+v -verify-machineinstrs < %s \
; RUN: | FileCheck %s --check-prefixes=SLOW,RV64-SLOW
-; RUN: llc -mtriple=riscv32 -mattr=+m,+v,+unaligned-vector-mem -verify-machineinstrs < %s \
+; RUN: llc -mtriple=riscv32 -mattr=+m,+v,+fast-unaligned-access -verify-machineinstrs < %s \
; RUN: | FileCheck %s --check-prefixes=FAST,RV32-FAST
-; RUN: llc -mtriple=riscv64 -mattr=+m,+v,+unaligned-vector-mem -verify-machineinstrs < %s \
+; RUN: llc -mtriple=riscv64 -mattr=+m,+v,+fast-unaligned-access -verify-machineinstrs < %s \
; RUN: | FileCheck %s --check-prefixes=FAST,RV64-FAST
define <4 x i32> @load_v4i32_align1(ptr %ptr) {
diff --git a/llvm/test/CodeGen/RISCV/rvv/memcpy-inline.ll b/llvm/test/CodeGen/RISCV/rvv/memcpy-inline.ll
index 53598c609107b0a..485f94ee2a10269 100644
--- a/llvm/test/CodeGen/RISCV/rvv/memcpy-inline.ll
+++ b/llvm/test/CodeGen/RISCV/rvv/memcpy-inline.ll
@@ -3,9 +3,9 @@
; RUN: | FileCheck %s --check-prefixes=RV32-BOTH,RV32
; RUN: llc < %s -mtriple=riscv64 -mattr=+v \
; RUN: | FileCheck %s --check-prefixes=RV64-BOTH,RV64
-; RUN: llc < %s -mtriple=riscv32 -mattr=+v,+unaligned-scalar-mem,+unaligned-vector-mem \
+; RUN: llc < %s -mtriple=riscv32 -mattr=+v,+fast-unaligned-access \
; RUN: | FileCheck %s --check-prefixes=RV32-BOTH,RV32-FAST
-; RUN: llc < %s -mtriple=riscv64 -mattr=+v,+unaligned-scalar-mem,+unaligned-vector-mem \
+; RUN: llc < %s -mtriple=riscv64 -mattr=+v,+fast-unaligned-access \
; RUN: | FileCheck %s --check-prefixes=RV64-BOTH,RV64-FAST
; ----------------------------------------------------------------------
diff --git a/llvm/test/CodeGen/RISCV/rvv/memset-inline.ll b/llvm/test/CodeGen/RISCV/rvv/memset-inline.ll
index accc18519d6260b..0e7e914cf68e8a1 100644
--- a/llvm/test/CodeGen/RISCV/rvv/memset-inline.ll
+++ b/llvm/test/CodeGen/RISCV/rvv/memset-inline.ll
@@ -3,9 +3,9 @@
; RUN: | FileCheck %s --check-prefixes=RV32-BOTH,RV32
; RUN: llc < %s -mtriple=riscv64 -mattr=+m,+v \
; RUN: | FileCheck %s --check-prefixes=RV64-BOTH,RV64
-; RUN: llc < %s -mtriple=riscv32 -mattr=+m,+v,+unaligned-scalar-mem,,+unaligned-vector-mem \
+; RUN: llc < %s -mtriple=riscv32 -mattr=+m,+v,+fast-unaligned-access \
; RUN: | FileCheck %s --check-prefixes=RV32-BOTH,RV32-FAST
-; RUN: llc < %s -mtriple=riscv64 -mattr=+m,+v,+unaligned-scalar-mem,+unaligned-vector-mem \
+; RUN: llc < %s -mtriple=riscv64 -mattr=+m,+v,+fast-unaligned-access \
; RUN: | FileCheck %s --check-prefixes=RV64-BOTH,RV64-FAST
%struct.x = type { i8, i8, i8, i8, i8, i8, i8, i8, i8, i8, i8 }
diff --git a/llvm/test/CodeGen/RISCV/rvv/unaligned-loads-stores.ll b/llvm/test/CodeGen/RISCV/rvv/unaligned-loads-stores.ll
index f76381bcdc6decb..a56ff463154e232 100644
--- a/llvm/test/CodeGen/RISCV/rvv/unaligned-loads-stores.ll
+++ b/llvm/test/CodeGen/RISCV/rvv/unaligned-loads-stores.ll
@@ -3,9 +3,9 @@
; RUN: -verify-machineinstrs | FileCheck %s
; RUN: llc -mtriple riscv64 -mattr=+d,+zfh,+zvfh,+v < %s \
; RUN: -verify-machineinstrs | FileCheck %s
-; RUN: llc -mtriple riscv32 -mattr=+d,+zfh,+zvfh,+v,+unaligned-vector-mem < %s \
+; RUN: llc -mtriple riscv32 -mattr=+d,+zfh,+zvfh,+v,+fast-unaligned-access < %s \
; RUN: -verify-machineinstrs | FileCheck --check-prefix=FAST %s
-; RUN: llc -mtriple riscv64 -mattr=+d,+zfh,+zvfh,+v,+unaligned-vector-mem < %s \
+; RUN: llc -mtriple riscv64 -mattr=+d,+zfh,+zvfh,+v,+fast-unaligned-access < %s \
; RUN: -verify-machineinstrs | FileCheck --check-prefix=FAST %s
diff --git a/llvm/test/CodeGen/RISCV/unaligned-load-store.ll b/llvm/test/CodeGen/RISCV/unaligned-load-store.ll
index ce0d8fedbfb88f2..599b0d08629eaff 100644
--- a/llvm/test/CodeGen/RISCV/unaligned-load-store.ll
+++ b/llvm/test/CodeGen/R...
[truncated]
|
topperc
reviewed
Nov 30, 2023
| @@ -174,15 +174,10 @@ void riscv::getRISCVTargetFeatures(const Driver &D, const llvm::Triple &Triple, | |||
| bool HasV = llvm::is_contained(Features, "+zve32x"); | |||
There was a problem hiding this comment.
Yep, change pending.
wangpc-pp
approved these changes
Dec 1, 2023
| @@ -171,18 +171,12 @@ void riscv::getRISCVTargetFeatures(const Driver &D, const llvm::Triple &Triple, | |||
| Features.push_back("-save-restore"); | |||
|
|
|||
| // -mno-unaligned-access is default, unless -munaligned-access is specified. | |||
| bool HasV = llvm::is_contained(Features, "+zve32x"); | |||
| if (const Arg *A = Args.getLastArg(options::OPT_munaligned_access, | |||
There was a problem hiding this comment.
This whole sippet can be replaced by tools::AddTargetFeature.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
When we'd originally added unaligned-scalar-mem and unaligned-vector-mem, they were separated into two parts under the theory that some processor might implement one, but not the other. At the moment, we don't have evidence of such a processor. The C/C++ level interface, and the clang driver command lines have settled on a single unaligned flag which indicates both scalar and vector support unaligned. Given that, let's remove the test matrix complexity for a set of configurations which don't appear useful.
Given these are internal feature names, I don't think we need to provide any forward compatibility. Anyone disagree?
Note: The immediate trigger for this patch was finding another case where the unaligned-vector-mem wasn't being properly serialized to IR from clang which resulted in problems reproducing assembly from clang's -emit-llvm feature. Instead of fixing this, I decided getting rid of the complexity was the better approach.