6502/llvm-6502
1
0
Fork 0
mirror of https://github.com/c64scene-ar/llvm-6502.git synced 2024-12-13 20:32:21 +00:00
Code Issues Projects Releases Wiki Activity

Use inbounds GEPs for memcpy and memset lowering

Browse Source 
Follow-up on discussion in http://reviews.llvm.org/D11220


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@242542 91177308-0d34-0410-b5e6-96231b3b80d8
This commit is contained in:
Eli Bendersky 2015-07-17 16:42:33 +00:00
parent f82375c972
commit 27bd1ca0d9
2 changed files with 15 additions and 13 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

14
lib/Target/NVPTX/NVPTXLowerAggrCopies.cpp
View File

@ -87,13 +87,14 @@ void convertMemCpyToLoop(Instruction *ConvertedInst, Value *SrcAddr,
// load from SrcAddr+LoopIndex // load from SrcAddr+LoopIndex
// TODO: we can leverage the align parameter of llvm.memcpy for more efficient // TODO: we can leverage the align parameter of llvm.memcpy for more efficient
// word-sized loads and stores. // word-sized loads and stores.
Value *Element = LoopBuilder.CreateLoad( Value *Element =
LoopBuilder.CreateGEP(LoopBuilder.getInt8Ty(), SrcAddr, LoopIndex), LoopBuilder.CreateLoad(LoopBuilder.CreateInBoundsGEP(
LoopBuilder.getInt8Ty(), SrcAddr, LoopIndex),
SrcIsVolatile); SrcIsVolatile);
// store at DstAddr+LoopIndex // store at DstAddr+LoopIndex
LoopBuilder.CreateStore( LoopBuilder.CreateStore(Element,
Element, LoopBuilder.CreateInBoundsGEP(LoopBuilder.getInt8Ty(),
LoopBuilder.CreateGEP(LoopBuilder.getInt8Ty(), DstAddr, LoopIndex), DstAddr, LoopIndex),
DstIsVolatile); DstIsVolatile);
// The value for LoopIndex coming from backedge is (LoopIndex + 1) // The value for LoopIndex coming from backedge is (LoopIndex + 1)
@ -225,7 +226,8 @@ void convertMemSetToLoop(Instruction *ConvertedInst, Value *DstAddr,
LoopIndex->addIncoming(ConstantInt::get(CopyLen->getType(), 0), OrigBB); LoopIndex->addIncoming(ConstantInt::get(CopyLen->getType(), 0), OrigBB);
LoopBuilder.CreateStore( LoopBuilder.CreateStore(
SetValue, LoopBuilder.CreateGEP(SetValue->getType(), DstAddr, LoopIndex), SetValue,
LoopBuilder.CreateInBoundsGEP(SetValue->getType(), DstAddr, LoopIndex),
false); false);
Value *NewIndex = Value *NewIndex =

10
test/CodeGen/NVPTX/lower-aggr-copies.ll
View File

@ -18,9 +18,9 @@ entry:
; IR-LABEL: @memcpy_caller ; IR-LABEL: @memcpy_caller
; IR: loadstoreloop: ; IR: loadstoreloop:
; IR: [[LOADPTR:%[0-9]+]] = getelementptr i8, i8* %src, i64 ; IR: [[LOADPTR:%[0-9]+]] = getelementptr inbounds i8, i8* %src, i64
; IR-NEXT: [[VAL:%[0-9]+]] = load i8, i8* [[LOADPTR]] ; IR-NEXT: [[VAL:%[0-9]+]] = load i8, i8* [[LOADPTR]]
; IR-NEXT: [[STOREPTR:%[0-9]+]] = getelementptr i8, i8* %dst, i64 ; IR-NEXT: [[STOREPTR:%[0-9]+]] = getelementptr inbounds i8, i8* %dst, i64
; IR-NEXT: store i8 [[VAL]], i8* [[STOREPTR]] ; IR-NEXT: store i8 [[VAL]], i8* [[STOREPTR]]
; PTX-LABEL: .visible .func (.param .b64 func_retval0) memcpy_caller ; PTX-LABEL: .visible .func (.param .b64 func_retval0) memcpy_caller
@ -61,8 +61,8 @@ entry:
; IR-LABEL: @memcpy_casting_caller ; IR-LABEL: @memcpy_casting_caller
; IR: [[DSTCAST:%[0-9]+]] = bitcast i32* %dst to i8* ; IR: [[DSTCAST:%[0-9]+]] = bitcast i32* %dst to i8*
; IR: [[SRCCAST:%[0-9]+]] = bitcast i32* %src to i8* ; IR: [[SRCCAST:%[0-9]+]] = bitcast i32* %src to i8*
; IR: getelementptr i8, i8* [[SRCCAST]] ; IR: getelementptr inbounds i8, i8* [[SRCCAST]]
; IR: getelementptr i8, i8* [[DSTCAST]] ; IR: getelementptr inbounds i8, i8* [[DSTCAST]]
} }
define i8* @memset_caller(i8* %dst, i32 %c, i64 %n) #0 { define i8* @memset_caller(i8* %dst, i32 %c, i64 %n) #0 {
@ -74,7 +74,7 @@ entry:
; IR-LABEL: @memset_caller ; IR-LABEL: @memset_caller
; IR: [[VAL:%[0-9]+]] = trunc i32 %c to i8 ; IR: [[VAL:%[0-9]+]] = trunc i32 %c to i8
; IR: loadstoreloop: ; IR: loadstoreloop:
; IR: [[STOREPTR:%[0-9]+]] = getelementptr i8, i8* %dst, i64 ; IR: [[STOREPTR:%[0-9]+]] = getelementptr inbounds i8, i8* %dst, i64
; IR-NEXT: store i8 [[VAL]], i8* [[STOREPTR]] ; IR-NEXT: store i8 [[VAL]], i8* [[STOREPTR]]
; PTX-LABEL: .visible .func (.param .b64 func_retval0) memset_caller( ; PTX-LABEL: .visible .func (.param .b64 func_retval0) memset_caller(