diff options
| author | Michael Liao <michael.liao@intel.com> | 2012-09-26 08:26:25 +0000 |
|---|---|---|
| committer | Michael Liao <michael.liao@intel.com> | 2012-09-26 08:26:25 +0000 |
| commit | 4fa2ddbb94bb7b7f67e4f4d0aac292998fbf00ed (patch) | |
| tree | 02f208c5c9f1e8d1620137e6ea020395e7006348 /lib/Target | |
| parent | 6bcdb5b9034a02901175e94e89b4815f28f2f141 (diff) | |
Add SARX/SHRX/SHLX code generation support
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@164675 91177308-0d34-0410-b5e6-96231b3b80d8
Diffstat (limited to 'lib/Target')
| -rw-r--r-- | lib/Target/X86/X86InstrInfo.cpp | 6 | ||||
| -rw-r--r-- | lib/Target/X86/X86InstrShiftRotate.td | 55 |
2 files changed, 61 insertions, 0 deletions
diff --git a/lib/Target/X86/X86InstrInfo.cpp b/lib/Target/X86/X86InstrInfo.cpp index 2fb2ed2104..95d2ad4a24 100644 --- a/lib/Target/X86/X86InstrInfo.cpp +++ b/lib/Target/X86/X86InstrInfo.cpp @@ -565,6 +565,12 @@ X86InstrInfo::X86InstrInfo(X86TargetMachine &tm) // BMI/BMI2 foldable instructions { X86::RORX32ri, X86::RORX32mi, 0 }, { X86::RORX64ri, X86::RORX64mi, 0 }, + { X86::SARX32rr, X86::SARX32rm, 0 }, + { X86::SARX64rr, X86::SARX64rm, 0 }, + { X86::SHRX32rr, X86::SHRX32rm, 0 }, + { X86::SHRX64rr, X86::SHRX64rm, 0 }, + { X86::SHLX32rr, X86::SHLX32rm, 0 }, + { X86::SHLX64rr, X86::SHLX64rm, 0 }, }; for (unsigned i = 0, e = array_lengthof(OpTbl1); i != e; ++i) { diff --git a/lib/Target/X86/X86InstrShiftRotate.td b/lib/Target/X86/X86InstrShiftRotate.td index fe7d0ecf89..893488c159 100644 --- a/lib/Target/X86/X86InstrShiftRotate.td +++ b/lib/Target/X86/X86InstrShiftRotate.td @@ -896,4 +896,59 @@ let Predicates = [HasBMI2] in { (RORX32mi addr:$src, (ROT32L2R_imm8 imm:$shamt))>; def : Pat<(rotl (loadi64 addr:$src), (i8 imm:$shamt)), (RORX64mi addr:$src, (ROT64L2R_imm8 imm:$shamt))>; + + // Prefer SARX/SHRX/SHLX over SAR/SHR/SHL with variable shift BUT not + // immedidate shift, i.e. the following code is considered better + // + // mov %edi, %esi + // shl $imm, %esi + // ... %edi, ... + // + // than + // + // movb $imm, %sil + // shlx %sil, %edi, %esi + // ... %edi, ... + // + let AddedComplexity = 1 in { + def : Pat<(sra GR32:$src1, GR8:$src2), + (SARX32rr GR32:$src1, + (INSERT_SUBREG + (i32 (IMPLICIT_DEF)), GR8:$src2, sub_8bit))>; + def : Pat<(sra GR64:$src1, GR8:$src2), + (SARX64rr GR64:$src1, + (INSERT_SUBREG + (i64 (IMPLICIT_DEF)), GR8:$src2, sub_8bit))>; + + def : Pat<(srl GR32:$src1, GR8:$src2), + (SHRX32rr GR32:$src1, + (INSERT_SUBREG + (i32 (IMPLICIT_DEF)), GR8:$src2, sub_8bit))>; + def : Pat<(srl GR64:$src1, GR8:$src2), + (SHRX64rr GR64:$src1, + (INSERT_SUBREG + (i64 (IMPLICIT_DEF)), GR8:$src2, sub_8bit))>; + + def : Pat<(shl GR32:$src1, GR8:$src2), + (SHLX32rr GR32:$src1, + (INSERT_SUBREG + (i32 (IMPLICIT_DEF)), GR8:$src2, sub_8bit))>; + def : Pat<(shl GR64:$src1, GR8:$src2), + (SHLX64rr GR64:$src1, + (INSERT_SUBREG + (i64 (IMPLICIT_DEF)), GR8:$src2, sub_8bit))>; + } + + // Patterns on SARXrm/SHRXrm/SHLXrm are explicitly omitted to favor + // + // mov (%ecx), %esi + // shl $imm, $esi + // + // over + // + // movb $imm %al + // shlx %al, (%ecx), %esi + // + // As SARXrr/SHRXrr/SHLXrr is favored on variable shift, the peephole + // optimization will fold them into SARXrm/SHRXrm/SHLXrm if possible. } |