[llvm] r256818 - [X86] Make MOV32ri64 a post-RA pseudo instead of a CodeGenOnly instruction. It was only needed for rematerialization.
Craig Topper via llvm-commits
llvm-commits at lists.llvm.org
Mon Jan 4 23:44:14 PST 2016
Author: ctopper
Date: Tue Jan 5 01:44:14 2016
New Revision: 256818
URL: http://llvm.org/viewvc/llvm-project?rev=256818&view=rev
Log:
[X86] Make MOV32ri64 a post-RA pseudo instead of a CodeGenOnly instruction. It was only needed for rematerialization.
Modified:
llvm/trunk/lib/Target/X86/X86InstrCompiler.td
llvm/trunk/lib/Target/X86/X86InstrInfo.cpp
llvm/trunk/lib/Target/X86/X86MCInstLower.cpp
Modified: llvm/trunk/lib/Target/X86/X86InstrCompiler.td
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86InstrCompiler.td?rev=256818&r1=256817&r2=256818&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86InstrCompiler.td (original)
+++ llvm/trunk/lib/Target/X86/X86InstrCompiler.td Tue Jan 5 01:44:14 2016
@@ -281,10 +281,9 @@ let Predicates = [OptForSize, NotSlowInc
// Materialize i64 constant where top 32-bits are zero. This could theoretically
// use MOV32ri with a SUBREG_TO_REG to represent the zero-extension, however
// that would make it more difficult to rematerialize.
-let AddedComplexity = 1, isReMaterializable = 1, isAsCheapAsAMove = 1,
- isCodeGenOnly = 1, hasSideEffects = 0 in
-def MOV32ri64 : Ii32<0xb8, AddRegFrm, (outs GR32:$dst), (ins i64i32imm:$src),
- "", [], IIC_ALU_NONMEM>, Sched<[WriteALU]>;
+let isReMaterializable = 1, isAsCheapAsAMove = 1,
+ isPseudo = 1, hasSideEffects = 0 in
+def MOV32ri64 : I<0, Pseudo, (outs GR32:$dst), (ins i64i32imm:$src), "", []>;
// This 64-bit pseudo-move can be used for both a 64-bit constant that is
// actually the zero-extension of a 32-bit constant and for labels in the
Modified: llvm/trunk/lib/Target/X86/X86InstrInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86InstrInfo.cpp?rev=256818&r1=256817&r2=256818&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86InstrInfo.cpp (original)
+++ llvm/trunk/lib/Target/X86/X86InstrInfo.cpp Tue Jan 5 01:44:14 2016
@@ -5369,7 +5369,10 @@ bool X86InstrInfo::expandPostRAPseudo(Ma
case X86::TEST8ri_NOREX:
MI->setDesc(get(X86::TEST8ri));
return true;
-
+ case X86::MOV32ri64:
+ MI->setDesc(get(X86::MOV32ri));
+ return true;
+
// KNL does not recognize dependency-breaking idioms for mask registers,
// so kxnor %k1, %k1, %k2 has a RAW dependence on %k1.
// Using %k0 as the undef input register is a performance heuristic based
Modified: llvm/trunk/lib/Target/X86/X86MCInstLower.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86MCInstLower.cpp?rev=256818&r1=256817&r2=256818&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86MCInstLower.cpp (original)
+++ llvm/trunk/lib/Target/X86/X86MCInstLower.cpp Tue Jan 5 01:44:14 2016
@@ -455,10 +455,6 @@ ReSimplify:
"LEA has segment specified!");
break;
- case X86::MOV32ri64:
- OutMI.setOpcode(X86::MOV32ri);
- break;
-
// Commute operands to get a smaller encoding by using VEX.R instead of VEX.B
// if one of the registers is extended, but other isn't.
case X86::VMOVZPQILo2PQIrr:
More information about the llvm-commits
mailing list