diff options
| author | Craig Topper <craig.topper@gmail.com> | 2013-10-14 00:24:33 +0000 |
|---|---|---|
| committer | Craig Topper <craig.topper@gmail.com> | 2013-10-14 00:24:33 +0000 |
| commit | c4a5a3f65d6b2c850f66f3496b1892a0673e4b1e (patch) | |
| tree | b51bf0e7d0549d9d5d33e77d3d494387c5397ec1 | |
| parent | 7af18578f8501eaf91953f10d987ce1aa776a0f9 (diff) | |
| download | bcm5719-llvm-c4a5a3f65d6b2c850f66f3496b1892a0673e4b1e.tar.gz bcm5719-llvm-c4a5a3f65d6b2c850f66f3496b1892a0673e4b1e.zip | |
Don't use 64-bit versions of MOVMSKPD in CodeGen. The instructions only produce a 1-bit result so we can just use SUBREG_TO_REG to extend the 32-bit versions.
llvm-svn: 192562
| -rw-r--r-- | llvm/lib/Target/X86/X86InstrSSE.td | 12 |
1 files changed, 8 insertions, 4 deletions
diff --git a/llvm/lib/Target/X86/X86InstrSSE.td b/llvm/lib/Target/X86/X86InstrSSE.td index 6d00c543388..6f7e9131580 100644 --- a/llvm/lib/Target/X86/X86InstrSSE.td +++ b/llvm/lib/Target/X86/X86InstrSSE.td @@ -2731,11 +2731,13 @@ let Predicates = [HasAVX] in { def : Pat<(i32 (X86fgetsign FR32:$src)), (VMOVMSKPSrr32 (COPY_TO_REGCLASS FR32:$src, VR128))>; def : Pat<(i64 (X86fgetsign FR32:$src)), - (VMOVMSKPSrr64 (COPY_TO_REGCLASS FR32:$src, VR128))>; + (SUBREG_TO_REG (i64 0), + (VMOVMSKPSrr32 (COPY_TO_REGCLASS FR32:$src, VR128)), sub_32bit)>; def : Pat<(i32 (X86fgetsign FR64:$src)), (VMOVMSKPDrr32 (COPY_TO_REGCLASS FR64:$src, VR128))>; def : Pat<(i64 (X86fgetsign FR64:$src)), - (VMOVMSKPDrr64 (COPY_TO_REGCLASS FR64:$src, VR128))>; + (SUBREG_TO_REG (i64 0), + (VMOVMSKPDrr32 (COPY_TO_REGCLASS FR64:$src, VR128)), sub_32bit)>; } defm MOVMSKPS : sse12_extr_sign_mask<VR128, int_x86_sse_movmsk_ps, "movmskps", @@ -2747,13 +2749,15 @@ def : Pat<(i32 (X86fgetsign FR32:$src)), (MOVMSKPSrr32 (COPY_TO_REGCLASS FR32:$src, VR128))>, Requires<[UseSSE1]>; def : Pat<(i64 (X86fgetsign FR32:$src)), - (MOVMSKPSrr64 (COPY_TO_REGCLASS FR32:$src, VR128))>, + (SUBREG_TO_REG (i64 0), + (MOVMSKPSrr32 (COPY_TO_REGCLASS FR32:$src, VR128)), sub_32bit)>, Requires<[UseSSE1]>; def : Pat<(i32 (X86fgetsign FR64:$src)), (MOVMSKPDrr32 (COPY_TO_REGCLASS FR64:$src, VR128))>, Requires<[UseSSE2]>; def : Pat<(i64 (X86fgetsign FR64:$src)), - (MOVMSKPDrr64 (COPY_TO_REGCLASS FR64:$src, VR128))>, + (SUBREG_TO_REG (i64 0), + (MOVMSKPDrr32 (COPY_TO_REGCLASS FR64:$src, VR128)), sub_32bit)>, Requires<[UseSSE2]>; //===---------------------------------------------------------------------===// |
