summaryrefslogtreecommitdiff
diff options
context:
space:
mode:
Diffstat
-rw-r--r--docs/LangRef.rst8
-rw-r--r--lib/Bitcode/Reader/BitcodeReader.cpp34
-rw-r--r--lib/Bitcode/Writer/BitcodeWriter.cpp6
-rw-r--r--lib/IR/Constants.cpp6
-rw-r--r--lib/IR/Instructions.cpp4
-rw-r--r--test/CodeGen/X86/vec_splat.ll16
-rw-r--r--test/Feature/instructions.ll2
7 files changed, 57 insertions, 19 deletions
diff --git a/docs/LangRef.rst b/docs/LangRef.rst
index a6595f7fd0..3d99a0e79b 100644
--- a/docs/LangRef.rst
+++ b/docs/LangRef.rst
@@ -4470,7 +4470,7 @@ Syntax:
::
- <result> = extractelement <n x <ty>> <val>, i32 <idx> ; yields <ty>
+ <result> = extractelement <n x <ty>> <val>, <ty2> <idx> ; yields <ty>
Overview:
"""""""""
@@ -4484,7 +4484,7 @@ Arguments:
The first operand of an '``extractelement``' instruction is a value of
:ref:`vector <t_vector>` type. The second operand is an index indicating
the position from which to extract the element. The index may be a
-variable.
+variable of any integer type.
Semantics:
""""""""""
@@ -4510,7 +4510,7 @@ Syntax:
::
- <result> = insertelement <n x <ty>> <val>, <ty> <elt>, i32 <idx> ; yields <n x <ty>>
+ <result> = insertelement <n x <ty>> <val>, <ty> <elt>, <ty2> <idx> ; yields <n x <ty>>
Overview:
"""""""""
@@ -4525,7 +4525,7 @@ The first operand of an '``insertelement``' instruction is a value of
:ref:`vector <t_vector>` type. The second operand is a scalar value whose
type must equal the element type of the first operand. The third operand
is an index indicating the position at which to insert the value. The
-index may be a variable.
+index may be a variable of any integer type.
Semantics:
""""""""""
diff --git a/lib/Bitcode/Reader/BitcodeReader.cpp b/lib/Bitcode/Reader/BitcodeReader.cpp
index 74e8439141..1b2cf76f3e 100644
--- a/lib/Bitcode/Reader/BitcodeReader.cpp
+++ b/lib/Bitcode/Reader/BitcodeReader.cpp
@@ -1418,7 +1418,8 @@ error_code BitcodeReader::ParseConstants() {
ValueList.getConstantFwdRef(Record[2],CurTy));
break;
}
- case bitc::CST_CODE_CE_EXTRACTELT: { // CE_EXTRACTELT: [opty, opval, opval]
+ case bitc::CST_CODE_CE_EXTRACTELT
+ : { // CE_EXTRACTELT: [opty, opval, opty, opval]
if (Record.size() < 3)
return Error(InvalidRecord);
VectorType *OpTy =
@@ -1426,20 +1427,37 @@ error_code BitcodeReader::ParseConstants() {
if (!OpTy)
return Error(InvalidRecord);
Constant *Op0 = ValueList.getConstantFwdRef(Record[1], OpTy);
- Constant *Op1 = ValueList.getConstantFwdRef(Record[2],
- Type::getInt32Ty(Context));
+ Constant *Op1 = nullptr;
+ if (Record.size() == 4) {
+ Type *IdxTy = getTypeByID(Record[2]);
+ if (!IdxTy)
+ return Error(InvalidRecord);
+ Op1 = ValueList.getConstantFwdRef(Record[3], IdxTy);
+ } else // TODO: Remove with llvm 4.0
+ Op1 = ValueList.getConstantFwdRef(Record[2], Type::getInt32Ty(Context));
+ if (!Op1)
+ return Error(InvalidRecord);
V = ConstantExpr::getExtractElement(Op0, Op1);
break;
}
- case bitc::CST_CODE_CE_INSERTELT: { // CE_INSERTELT: [opval, opval, opval]
+ case bitc::CST_CODE_CE_INSERTELT
+ : { // CE_INSERTELT: [opval, opval, opty, opval]
VectorType *OpTy = dyn_cast<VectorType>(CurTy);
if (Record.size() < 3 || !OpTy)
return Error(InvalidRecord);
Constant *Op0 = ValueList.getConstantFwdRef(Record[0], OpTy);
Constant *Op1 = ValueList.getConstantFwdRef(Record[1],
OpTy->getElementType());
- Constant *Op2 = ValueList.getConstantFwdRef(Record[2],
- Type::getInt32Ty(Context));
+ Constant *Op2 = nullptr;
+ if (Record.size() == 4) {
+ Type *IdxTy = getTypeByID(Record[2]);
+ if (!IdxTy)
+ return Error(InvalidRecord);
+ Op2 = ValueList.getConstantFwdRef(Record[3], IdxTy);
+ } else // TODO: Remove with llvm 4.0
+ Op2 = ValueList.getConstantFwdRef(Record[2], Type::getInt32Ty(Context));
+ if (!Op2)
+ return Error(InvalidRecord);
V = ConstantExpr::getInsertElement(Op0, Op1, Op2);
break;
}
@@ -2460,7 +2478,7 @@ error_code BitcodeReader::ParseFunctionBody(Function *F) {
unsigned OpNum = 0;
Value *Vec, *Idx;
if (getValueTypePair(Record, OpNum, NextValueNo, Vec) ||
- popValue(Record, OpNum, NextValueNo, Type::getInt32Ty(Context), Idx))
+ getValueTypePair(Record, OpNum, NextValueNo, Idx))
return Error(InvalidRecord);
I = ExtractElementInst::Create(Vec, Idx);
InstructionList.push_back(I);
@@ -2473,7 +2491,7 @@ error_code BitcodeReader::ParseFunctionBody(Function *F) {
if (getValueTypePair(Record, OpNum, NextValueNo, Vec) ||
popValue(Record, OpNum, NextValueNo,
cast<VectorType>(Vec->getType())->getElementType(), Elt) ||
- popValue(Record, OpNum, NextValueNo, Type::getInt32Ty(Context), Idx))
+ getValueTypePair(Record, OpNum, NextValueNo, Idx))
return Error(InvalidRecord);
I = InsertElementInst::Create(Vec, Elt, Idx);
InstructionList.push_back(I);
diff --git a/lib/Bitcode/Writer/BitcodeWriter.cpp b/lib/Bitcode/Writer/BitcodeWriter.cpp
index 92965fa7e4..23374872df 100644
--- a/lib/Bitcode/Writer/BitcodeWriter.cpp
+++ b/lib/Bitcode/Writer/BitcodeWriter.cpp
@@ -1087,12 +1087,14 @@ static void WriteConstants(unsigned FirstVal, unsigned LastVal,
Code = bitc::CST_CODE_CE_EXTRACTELT;
Record.push_back(VE.getTypeID(C->getOperand(0)->getType()));
Record.push_back(VE.getValueID(C->getOperand(0)));
+ Record.push_back(VE.getTypeID(C->getOperand(1)->getType()));
Record.push_back(VE.getValueID(C->getOperand(1)));
break;
case Instruction::InsertElement:
Code = bitc::CST_CODE_CE_INSERTELT;
Record.push_back(VE.getValueID(C->getOperand(0)));
Record.push_back(VE.getValueID(C->getOperand(1)));
+ Record.push_back(VE.getTypeID(C->getOperand(2)->getType()));
Record.push_back(VE.getValueID(C->getOperand(2)));
break;
case Instruction::ShuffleVector:
@@ -1253,13 +1255,13 @@ static void WriteInstruction(const Instruction &I, unsigned InstID,
case Instruction::ExtractElement:
Code = bitc::FUNC_CODE_INST_EXTRACTELT;
PushValueAndType(I.getOperand(0), InstID, Vals, VE);
- pushValue(I.getOperand(1), InstID, Vals, VE);
+ PushValueAndType(I.getOperand(1), InstID, Vals, VE);
break;
case Instruction::InsertElement:
Code = bitc::FUNC_CODE_INST_INSERTELT;
PushValueAndType(I.getOperand(0), InstID, Vals, VE);
pushValue(I.getOperand(1), InstID, Vals, VE);
- pushValue(I.getOperand(2), InstID, Vals, VE);
+ PushValueAndType(I.getOperand(2), InstID, Vals, VE);
break;
case Instruction::ShuffleVector:
Code = bitc::FUNC_CODE_INST_SHUFFLEVEC;
diff --git a/lib/IR/Constants.cpp b/lib/IR/Constants.cpp
index 54be980292..bb8d60b234 100644
--- a/lib/IR/Constants.cpp
+++ b/lib/IR/Constants.cpp
@@ -1937,8 +1937,8 @@ ConstantExpr::getFCmp(unsigned short pred, Constant *LHS, Constant *RHS) {
Constant *ConstantExpr::getExtractElement(Constant *Val, Constant *Idx) {
assert(Val->getType()->isVectorTy() &&
"Tried to create extractelement operation on non-vector type!");
- assert(Idx->getType()->isIntegerTy(32) &&
- "Extractelement index must be i32 type!");
+ assert(Idx->getType()->isIntegerTy() &&
+ "Extractelement index must be an integer type!");
if (Constant *FC = ConstantFoldExtractElementInstruction(Val, Idx))
return FC; // Fold a few common cases.
@@ -1958,7 +1958,7 @@ Constant *ConstantExpr::getInsertElement(Constant *Val, Constant *Elt,
"Tried to create insertelement operation on non-vector type!");
assert(Elt->getType() == Val->getType()->getVectorElementType() &&
"Insertelement types must match!");
- assert(Idx->getType()->isIntegerTy(32) &&
+ assert(Idx->getType()->isIntegerTy() &&
"Insertelement index must be i32 type!");
if (Constant *FC = ConstantFoldInsertElementInstruction(Val, Elt, Idx))
diff --git a/lib/IR/Instructions.cpp b/lib/IR/Instructions.cpp
index e8bcddbd99..8aebb8c7c2 100644
--- a/lib/IR/Instructions.cpp
+++ b/lib/IR/Instructions.cpp
@@ -1479,7 +1479,7 @@ ExtractElementInst::ExtractElementInst(Value *Val, Value *Index,
bool ExtractElementInst::isValidOperands(const Value *Val, const Value *Index) {
- if (!Val->getType()->isVectorTy() || !Index->getType()->isIntegerTy(32))
+ if (!Val->getType()->isVectorTy() || !Index->getType()->isIntegerTy())
return false;
return true;
}
@@ -1526,7 +1526,7 @@ bool InsertElementInst::isValidOperands(const Value *Vec, const Value *Elt,
if (Elt->getType() != cast<VectorType>(Vec->getType())->getElementType())
return false;// Second operand of insertelement must be vector element type.
- if (!Index->getType()->isIntegerTy(32))
+ if (!Index->getType()->isIntegerTy())
return false; // Third operand of insertelement must be i32.
return true;
}
diff --git a/test/CodeGen/X86/vec_splat.ll b/test/CodeGen/X86/vec_splat.ll
index 543c96ef3d..a02e383607 100644
--- a/test/CodeGen/X86/vec_splat.ll
+++ b/test/CodeGen/X86/vec_splat.ll
@@ -32,3 +32,19 @@ define void @test_v2sd(<2 x double>* %P, <2 x double>* %Q, double %X) nounwind {
; SSE3-LABEL: test_v2sd:
; SSE3: movddup
}
+
+; Fold extract of a load into the load's address computation. This avoids spilling to the stack.
+define <4 x float> @load_extract_splat(<4 x float>* nocapture readonly %ptr, i64 %i, i64 %j) nounwind {
+ %1 = getelementptr inbounds <4 x float>* %ptr, i64 %i
+ %2 = load <4 x float>* %1, align 16
+ %3 = extractelement <4 x float> %2, i64 %j
+ %4 = insertelement <4 x float> undef, float %3, i32 0
+ %5 = insertelement <4 x float> %4, float %3, i32 1
+ %6 = insertelement <4 x float> %5, float %3, i32 2
+ %7 = insertelement <4 x float> %6, float %3, i32 3
+ ret <4 x float> %7
+
+; AVX-LABEL: load_extract_splat
+; AVX-NOT: movs
+; AVX: vbroadcastss
+}
diff --git a/test/Feature/instructions.ll b/test/Feature/instructions.ll
index d0c303d719..aa962948a1 100644
--- a/test/Feature/instructions.ll
+++ b/test/Feature/instructions.ll
@@ -4,11 +4,13 @@
define i32 @test_extractelement(<4 x i32> %V) {
%R = extractelement <4 x i32> %V, i32 1 ; <i32> [#uses=1]
+ %S = extractelement <4 x i32> %V, i64 1 ; <i32> [#uses=0]
ret i32 %R
}
define <4 x i32> @test_insertelement(<4 x i32> %V) {
%R = insertelement <4 x i32> %V, i32 0, i32 0 ; <<4 x i32>> [#uses=1]
+ %S = insertelement <4 x i32> %V, i32 0, i64 0 ; <<4 x i32>> [#uses=0]
ret <4 x i32> %R
}
generated by cgit v1.2.3 (git 2.25.1) at 2024-05-23 01:50:17 +0000