Set a higher value for maxStoresPerMemcpy in MipsISelLowering.cpp.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@158414 91177308-0d34-0410-b5e6-96231b3b80d8
diff --git a/lib/Target/Mips/MipsISelLowering.cpp b/lib/Target/Mips/MipsISelLowering.cpp
index d6a8b8e..aa777f7 100644
--- a/lib/Target/Mips/MipsISelLowering.cpp
+++ b/lib/Target/Mips/MipsISelLowering.cpp
@@ -303,6 +303,8 @@
setExceptionPointerRegister(IsN64 ? Mips::A0_64 : Mips::A0);
setExceptionSelectorRegister(IsN64 ? Mips::A1_64 : Mips::A1);
+
+ maxStoresPerMemcpy = 16;
}
bool MipsTargetLowering::allowsUnalignedMemoryAccesses(EVT VT) const {
@@ -3455,6 +3457,16 @@
return false;
}
+EVT MipsTargetLowering::getOptimalMemOpType(uint64_t Size, unsigned DstAlign,
+ unsigned SrcAlign, bool IsZeroVal,
+ bool MemcpyStrSrc,
+ MachineFunction &MF) const {
+ if (Subtarget->hasMips64())
+ return MVT::i64;
+
+ return MVT::i32;
+}
+
bool MipsTargetLowering::isFPImmLegal(const APFloat &Imm, EVT VT) const {
if (VT != MVT::f32 && VT != MVT::f64)
return false;
diff --git a/lib/Target/Mips/MipsISelLowering.h b/lib/Target/Mips/MipsISelLowering.h
index 740cf06..e45d352 100644
--- a/lib/Target/Mips/MipsISelLowering.h
+++ b/lib/Target/Mips/MipsISelLowering.h
@@ -195,6 +195,11 @@
virtual bool isOffsetFoldingLegal(const GlobalAddressSDNode *GA) const;
+ virtual EVT getOptimalMemOpType(uint64_t Size, unsigned DstAlign,
+ unsigned SrcAlign, bool IsZeroVal,
+ bool MemcpyStrSrc,
+ MachineFunction &MF) const;
+
/// isFPImmLegal - Returns true if the target can instruction select the
/// specified FP immediate natively. If false, the legalizer will
/// materialize the FP immediate as a load from a constant pool.
diff --git a/test/CodeGen/Mips/memcpy.ll b/test/CodeGen/Mips/memcpy.ll
new file mode 100644
index 0000000..39764a9
--- /dev/null
+++ b/test/CodeGen/Mips/memcpy.ll
@@ -0,0 +1,19 @@
+; RUN: llc -march=mipsel < %s | FileCheck %s
+
+%struct.S1 = type { i32, [41 x i8] }
+
+@.str = private unnamed_addr constant [31 x i8] c"abcdefghijklmnopqrstuvwxyzABCD\00", align 1
+
+define void @foo1(%struct.S1* %s1, i8 signext %n) nounwind {
+entry:
+; CHECK-NOT: call16(memcpy
+
+ %arraydecay = getelementptr inbounds %struct.S1* %s1, i32 0, i32 1, i32 0
+ tail call void @llvm.memcpy.p0i8.p0i8.i32(i8* %arraydecay, i8* getelementptr inbounds ([31 x i8]* @.str, i32 0, i32 0), i32 31, i32 1, i1 false)
+ %arrayidx = getelementptr inbounds %struct.S1* %s1, i32 0, i32 1, i32 40
+ store i8 %n, i8* %arrayidx, align 1
+ ret void
+}
+
+declare void @llvm.memcpy.p0i8.p0i8.i32(i8* nocapture, i8* nocapture, i32, i32, i1) nounwind
+