[llvm-commits] [llvm] r160346 - in /llvm/trunk: lib/CodeGen/SelectionDAG/TargetLowering.cpp lib/Target/X86/X86ISelLowering.cpp lib/Target/X86/X86ISelLowering.h test/CodeGen/X86/cmp.ll
Evan Cheng
evan.cheng at apple.com
Mon Jul 16 23:53:40 PDT 2012
Author: evancheng
Date: Tue Jul 17 01:53:39 2012
New Revision: 160346
URL: http://llvm.org/viewvc/llvm-project?rev=160346&view=rev
Log:
This is another case where instcombine demanded bits optimization created
large immediates. Add dag combine logic to recover in case the large
immediates doesn't fit in cmp immediate operand field.
int foo(unsigned long l) {
return (l>> 47) == 1;
}
we produce
%shr.mask = and i64 %l, -140737488355328
%cmp = icmp eq i64 %shr.mask, 140737488355328
%conv = zext i1 %cmp to i32
ret i32 %conv
which codegens to
movq $0xffff800000000000,%rax
andq %rdi,%rax
movq $0x0000800000000000,%rcx
cmpq %rcx,%rax
sete %al
movzbl %al,%eax
ret
TargetLowering::SimplifySetCC would transform
(X & -256) == 256 -> (X >> 8) == 1
if the immediate fails the isLegalICmpImmediate() test. For x86,
that's immediates which are not a signed 32-bit immediate.
Based on a patch by Eli Friedman.
PR10328
rdar://9758774
Modified:
llvm/trunk/lib/CodeGen/SelectionDAG/TargetLowering.cpp
llvm/trunk/lib/Target/X86/X86ISelLowering.cpp
llvm/trunk/lib/Target/X86/X86ISelLowering.h
llvm/trunk/test/CodeGen/X86/cmp.ll
Modified: llvm/trunk/lib/CodeGen/SelectionDAG/TargetLowering.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/TargetLowering.cpp?rev=160346&r1=160345&r2=160346&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/SelectionDAG/TargetLowering.cpp (original)
+++ llvm/trunk/lib/CodeGen/SelectionDAG/TargetLowering.cpp Tue Jul 17 01:53:39 2012
@@ -2322,6 +2322,27 @@
}
}
}
+
+ if (!isLegalICmpImmediate(C1.getSExtValue())) {
+ // (X & -256) == 256 -> (X >> 8) == 1
+ if ((Cond == ISD::SETEQ || Cond == ISD::SETNE) &&
+ N0.getOpcode() == ISD::AND && N0.hasOneUse()) {
+ if (ConstantSDNode *AndRHS =
+ dyn_cast<ConstantSDNode>(N0.getOperand(1))) {
+ const APInt &AndRHSC = AndRHS->getAPIntValue();
+ if ((-AndRHSC).isPowerOf2() && (AndRHSC & C1) == C1) {
+ unsigned ShiftBits = AndRHSC.countTrailingZeros();
+ EVT ShiftTy = DCI.isBeforeLegalize() ?
+ getPointerTy() : getShiftAmountTy(N0.getValueType());
+ EVT CmpTy = N0.getValueType();
+ SDValue Shift = DAG.getNode(ISD::SRL, dl, CmpTy, N0.getOperand(0),
+ DAG.getConstant(ShiftBits, ShiftTy));
+ SDValue CmpRHS = DAG.getConstant(C1.lshr(ShiftBits), CmpTy);
+ return DAG.getSetCC(dl, VT, Shift, CmpRHS, Cond);
+ }
+ }
+ }
+ }
}
if (isa<ConstantFPSDNode>(N0.getNode())) {
Modified: llvm/trunk/lib/Target/X86/X86ISelLowering.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86ISelLowering.cpp?rev=160346&r1=160345&r2=160346&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86ISelLowering.cpp (original)
+++ llvm/trunk/lib/Target/X86/X86ISelLowering.cpp Tue Jul 17 01:53:39 2012
@@ -11384,6 +11384,14 @@
return true;
}
+bool X86TargetLowering::isLegalICmpImmediate(int64_t Imm) const {
+ return Imm == (int32_t)Imm;
+}
+
+bool X86TargetLowering::isLegalAddImmediate(int64_t Imm) const {
+ return Imm == (int32_t)Imm;
+}
+
bool X86TargetLowering::isTruncateFree(EVT VT1, EVT VT2) const {
if (!VT1.isInteger() || !VT2.isInteger())
return false;
Modified: llvm/trunk/lib/Target/X86/X86ISelLowering.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86ISelLowering.h?rev=160346&r1=160345&r2=160346&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86ISelLowering.h (original)
+++ llvm/trunk/lib/Target/X86/X86ISelLowering.h Tue Jul 17 01:53:39 2012
@@ -568,6 +568,18 @@
/// by AM is legal for this target, for a load/store of the specified type.
virtual bool isLegalAddressingMode(const AddrMode &AM, Type *Ty)const;
+ /// isLegalICmpImmediate - Return true if the specified immediate is legal
+ /// icmp immediate, that is the target has icmp instructions which can
+ /// compare a register against the immediate without having to materialize
+ /// the immediate into a register.
+ virtual bool isLegalICmpImmediate(int64_t Imm) const;
+
+ /// isLegalAddImmediate - Return true if the specified immediate is legal
+ /// add immediate, that is the target has add instructions which can
+ /// add a register and the immediate without having to materialize
+ /// the immediate into a register.
+ virtual bool isLegalAddImmediate(int64_t Imm) const;
+
/// isTruncateFree - Return true if it's free to truncate a value of
/// type Ty1 to type Ty2. e.g. On x86 it's free to truncate a i32 value in
/// register EAX to i16 by referencing its sub-register AX.
Modified: llvm/trunk/test/CodeGen/X86/cmp.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/cmp.ll?rev=160346&r1=160345&r2=160346&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/X86/cmp.ll (original)
+++ llvm/trunk/test/CodeGen/X86/cmp.ll Tue Jul 17 01:53:39 2012
@@ -91,7 +91,7 @@
}
; rdar://11866926
-define i32 @test7(i64 %res) nounwind uwtable readnone ssp {
+define i32 @test7(i64 %res) nounwind {
entry:
; CHECK: test7:
; CHECK-NOT: movabsq
@@ -103,7 +103,7 @@
ret i32 %lnot.ext
}
-define i32 @test8(i64 %res) nounwind uwtable readnone ssp {
+define i32 @test8(i64 %res) nounwind {
entry:
; CHECK: test8:
; CHECK-NOT: movabsq
@@ -114,7 +114,7 @@
ret i32 %lnot.ext
}
-define i32 @test9(i64 %res) nounwind uwtable readnone ssp {
+define i32 @test9(i64 %res) nounwind {
entry:
; CHECK: test9:
; CHECK-NOT: movabsq
@@ -126,7 +126,7 @@
ret i32 %lnot.ext
}
-define i32 @test10(i64 %res) nounwind uwtable readnone ssp {
+define i32 @test10(i64 %res) nounwind {
entry:
; CHECK: test10:
; CHECK-NOT: movabsq
@@ -138,3 +138,16 @@
ret i32 %lnot.ext
}
+; rdar://9758774
+define i32 @test11(i64 %l) nounwind {
+entry:
+; CHECK: test11:
+; CHECK-NOT: movabsq
+; CHECK-NOT: andq
+; CHECK: shrq $47, %rdi
+; CHECK: cmpq $1, %rdi
+ %shr.mask = and i64 %l, -140737488355328
+ %cmp = icmp eq i64 %shr.mask, 140737488355328
+ %conv = zext i1 %cmp to i32
+ ret i32 %conv
+}
More information about the llvm-commits
mailing list