replica
/
hanchenye-llvm-project
2
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

Mark several instructions SSE2 instead of SSE3 as they should be. 

llvm-svn: 158049
This commit is contained in:
Craig Topper 2012-06-06 06:45:27 +00:00
parent f470a4c2ca
commit bf2409e8aa
2 changed files with 11 additions and 9 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
llvm/lib/Target/X86/X86InstrFormats.td
View File

@ -378,6 +378,9 @@ class SDI<bits<8> o, Format F, dag outs, dag ins, string asm,
class SDIi8<bits<8> o, Format F, dag outs, dag ins, string asm,
list<dag> pattern, InstrItinClass itin = IIC_DEFAULT>
: Ii8<o, F, outs, ins, asm, pattern, itin>, XD, Requires<[HasSSE2]>;
class SSDI<bits<8> o, Format F, dag outs, dag ins, string asm,
list<dag> pattern, InstrItinClass itin = IIC_DEFAULT>
: I<o, F, outs, ins, asm, pattern, itin>, XS, Requires<[HasSSE2]>;
class SSDIi8<bits<8> o, Format F, dag outs, dag ins, string asm,
list<dag> pattern, InstrItinClass itin = IIC_DEFAULT>
: Ii8<o, F, outs, ins, asm, pattern>, XS, Requires<[HasSSE2]>;

17
llvm/lib/Target/X86/X86InstrSSE.td
View File

@ -1813,7 +1813,6 @@ def Int_CVTDQ2PSrm : I<0x5B, MRMSrcMem, (outs VR128:$dst), (ins i128mem:$src),
IIC_SSE_CVT_PS_RM>,
TB, Requires<[HasSSE2]>;
// FIXME: why the non-intrinsic version is described as SSE3?
// SSE2 instructions with XS prefix
def Int_VCVTDQ2PDrr : I<0xE6, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
"vcvtdq2pd\t{$src, $dst|$dst, $src}",
@ -4603,7 +4602,7 @@ def MOVPQIto64rr : RPDI<0x7E, MRMDestReg, (outs GR64:$dst), (ins VR128:$src),
// Bitcast FR64 <-> GR64
//
let Predicates = [HasAVX] in
def VMOV64toSDrm : S3SI<0x7E, MRMSrcMem, (outs FR64:$dst), (ins i64mem:$src),
def VMOV64toSDrm : SSDI<0x7E, MRMSrcMem, (outs FR64:$dst), (ins i64mem:$src),
"vmovq\t{$src, $dst|$dst, $src}",
[(set FR64:$dst, (bitconvert (loadi64 addr:$src)))]>,
VEX;
@ -4616,7 +4615,7 @@ def VMOVSDto64mr : VRPDI<0x7E, MRMDestMem, (outs), (ins i64mem:$dst, FR64:$src),
[(store (i64 (bitconvert FR64:$src)), addr:$dst)],
IIC_SSE_MOVDQ>, VEX;
def MOV64toSDrm : S3SI<0x7E, MRMSrcMem, (outs FR64:$dst), (ins i64mem:$src),
def MOV64toSDrm : SSDI<0x7E, MRMSrcMem, (outs FR64:$dst), (ins i64mem:$src),
"movq\t{$src, $dst|$dst, $src}",
[(set FR64:$dst, (bitconvert (loadi64 addr:$src)))],
IIC_SSE_MOVDQ>;
@ -4930,20 +4929,20 @@ def : Pat<(v4i32 (fp_to_sint (memopv4f64 addr:$src))),
// Convert Packed DW Integers to Packed Double FP
let Predicates = [HasAVX] in {
def VCVTDQ2PDrm : S3SI<0xE6, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
def VCVTDQ2PDrm : SSDI<0xE6, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
"vcvtdq2pd\t{$src, $dst|$dst, $src}", []>, VEX;
def VCVTDQ2PDrr : S3SI<0xE6, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
def VCVTDQ2PDrr : SSDI<0xE6, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
"vcvtdq2pd\t{$src, $dst|$dst, $src}", []>, VEX;
def VCVTDQ2PDYrm : S3SI<0xE6, MRMSrcMem, (outs VR256:$dst), (ins f128mem:$src),
def VCVTDQ2PDYrm : SSDI<0xE6, MRMSrcMem, (outs VR256:$dst), (ins f128mem:$src),
"vcvtdq2pd\t{$src, $dst|$dst, $src}", []>, VEX;
def VCVTDQ2PDYrr : S3SI<0xE6, MRMSrcReg, (outs VR256:$dst), (ins VR128:$src),
def VCVTDQ2PDYrr : SSDI<0xE6, MRMSrcReg, (outs VR256:$dst), (ins VR128:$src),
"vcvtdq2pd\t{$src, $dst|$dst, $src}", []>, VEX;
}
def CVTDQ2PDrm : S3SI<0xE6, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
def CVTDQ2PDrm : SSDI<0xE6, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
"cvtdq2pd\t{$src, $dst|$dst, $src}", [],
IIC_SSE_CVT_PD_RR>;
def CVTDQ2PDrr : S3SI<0xE6, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
def CVTDQ2PDrr : SSDI<0xE6, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
"cvtdq2pd\t{$src, $dst|$dst, $src}", [],
IIC_SSE_CVT_PD_RM>;