Added optimization that narrow load / op / store and the 'op' is a bit twiddling instruction and its second operand is an immediate. If bits that are touched by 'op' can be done with a narrower instruction, reduce the width of the load and store as well. This happens a lot with bitfield manipulation code.
e.g.
orl $65536, 8(%rax)
=>
orb $1, 10(%rax)
Since narrowing is not always a win, e.g. i32 -> i16 is a loss on x86, dag combiner consults with the target before performing the optimization.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@72507 91177308-0d34-0410-b5e6-96231b3b80d8
diff --git a/lib/Target/X86/X86ISelLowering.cpp b/lib/Target/X86/X86ISelLowering.cpp
index b89eef0..0136f90 100644
--- a/lib/Target/X86/X86ISelLowering.cpp
+++ b/lib/Target/X86/X86ISelLowering.cpp
@@ -6877,6 +6877,11 @@
return VT1 == MVT::i32 && VT2 == MVT::i64 && Subtarget->is64Bit();
}
+bool X86TargetLowering::isNarrowingProfitable(MVT VT1, MVT VT2) const {
+ // i16 instructions are longer (0x66 prefix) and potentially slower.
+ return !(VT1 == MVT::i32 && VT2 == MVT::i16);
+}
+
/// isShuffleMaskLegal - Targets can use this to indicate that they only
/// support *some* VECTOR_SHUFFLE operations, those with specific masks.
/// By default, if a target supports the VECTOR_SHUFFLE node, all mask values
diff --git a/lib/Target/X86/X86ISelLowering.h b/lib/Target/X86/X86ISelLowering.h
index badbd24..550f8bd 100644
--- a/lib/Target/X86/X86ISelLowering.h
+++ b/lib/Target/X86/X86ISelLowering.h
@@ -466,6 +466,11 @@
virtual bool isZExtFree(const Type *Ty1, const Type *Ty2) const;
virtual bool isZExtFree(MVT VT1, MVT VT2) const;
+ /// isNarrowingProfitable - Return true if it's profitable to narrow
+ /// operations of type VT1 to VT2. e.g. on x86, it's profitable to narrow
+ /// from i32 to i8 but not from i32 to i16.
+ virtual bool isNarrowingProfitable(MVT VT1, MVT VT2) const;
+
/// isShuffleMaskLegal - Targets can use this to indicate that they only
/// support *some* VECTOR_SHUFFLE operations, those with specific masks.
/// By default, if a target supports the VECTOR_SHUFFLE node, all mask