6502/llvm-6502
1
0
Fork 0
mirror of https://github.com/c64scene-ar/llvm-6502.git synced 2025-09-28 07:17:32 +00:00
Code Issues Projects Releases Wiki Activity

Give priority to AVX over SSE for 128-bit floating point unpck instructions.

Browse Source 
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@148233 91177308-0d34-0410-b5e6-96231b3b80d8
This commit is contained in:
Craig Topper
2012-01-16 09:56:42 +00:00
parent 9223822b85
commit b7ab7fe053
1 changed files with 34 additions and 34 deletions
Download Patch File Download Diff File Expand all files Collapse all files

68
lib/Target/X86/X86InstrSSE.td
View File

@@ -2529,40 +2529,7 @@ let AddedComplexity = 10 in {
} // Constraints = "$src1 = $dst" } // Constraints = "$src1 = $dst"
} // AddedComplexity } // AddedComplexity
let Predicates = [HasSSE1] in { let Predicates = [HasAVX], AddedComplexity = 1 in {
def : Pat<(v4f32 (X86Unpckl VR128:$src1, (memopv4f32 addr:$src2))),
(UNPCKLPSrm VR128:$src1, addr:$src2)>;
def : Pat<(v4f32 (X86Unpckl VR128:$src1, VR128:$src2)),
(UNPCKLPSrr VR128:$src1, VR128:$src2)>;
def : Pat<(v4f32 (X86Unpckh VR128:$src1, (memopv4f32 addr:$src2))),
(UNPCKHPSrm VR128:$src1, addr:$src2)>;
def : Pat<(v4f32 (X86Unpckh VR128:$src1, VR128:$src2)),
(UNPCKHPSrr VR128:$src1, VR128:$src2)>;
}
let Predicates = [HasSSE2] in {
def : Pat<(v2f64 (X86Unpckl VR128:$src1, (memopv2f64 addr:$src2))),
(UNPCKLPDrm VR128:$src1, addr:$src2)>;
def : Pat<(v2f64 (X86Unpckl VR128:$src1, VR128:$src2)),
(UNPCKLPDrr VR128:$src1, VR128:$src2)>;
def : Pat<(v2f64 (X86Unpckh VR128:$src1, (memopv2f64 addr:$src2))),
(UNPCKHPDrm VR128:$src1, addr:$src2)>;
def : Pat<(v2f64 (X86Unpckh VR128:$src1, VR128:$src2)),
(UNPCKHPDrr VR128:$src1, VR128:$src2)>;
// FIXME: Instead of X86Movddup, there should be a X86Unpckl here, the
// problem is during lowering, where it's not possible to recognize the load
// fold cause it has two uses through a bitcast. One use disappears at isel
// time and the fold opportunity reappears.
def : Pat<(v2f64 (X86Movddup VR128:$src)),
(UNPCKLPDrr VR128:$src, VR128:$src)>;
let AddedComplexity = 10 in
def : Pat<(splat_lo (v2f64 VR128:$src), (undef)),
(UNPCKLPDrr VR128:$src, VR128:$src)>;
}
let Predicates = [HasAVX] in {
def : Pat<(v4f32 (X86Unpckl VR128:$src1, (memopv4f32 addr:$src2))), def : Pat<(v4f32 (X86Unpckl VR128:$src1, (memopv4f32 addr:$src2))),
(VUNPCKLPSrm VR128:$src1, addr:$src2)>; (VUNPCKLPSrm VR128:$src1, addr:$src2)>;
def : Pat<(v4f32 (X86Unpckl VR128:$src1, VR128:$src2)), def : Pat<(v4f32 (X86Unpckl VR128:$src1, VR128:$src2)),
@@ -2610,6 +2577,39 @@ let Predicates = [HasAVX] in {
(VUNPCKLPDrr VR128:$src, VR128:$src)>; (VUNPCKLPDrr VR128:$src, VR128:$src)>;
} }
let Predicates = [HasSSE1] in {
def : Pat<(v4f32 (X86Unpckl VR128:$src1, (memopv4f32 addr:$src2))),
(UNPCKLPSrm VR128:$src1, addr:$src2)>;
def : Pat<(v4f32 (X86Unpckl VR128:$src1, VR128:$src2)),
(UNPCKLPSrr VR128:$src1, VR128:$src2)>;
def : Pat<(v4f32 (X86Unpckh VR128:$src1, (memopv4f32 addr:$src2))),
(UNPCKHPSrm VR128:$src1, addr:$src2)>;
def : Pat<(v4f32 (X86Unpckh VR128:$src1, VR128:$src2)),
(UNPCKHPSrr VR128:$src1, VR128:$src2)>;
}
let Predicates = [HasSSE2] in {
def : Pat<(v2f64 (X86Unpckl VR128:$src1, (memopv2f64 addr:$src2))),
(UNPCKLPDrm VR128:$src1, addr:$src2)>;
def : Pat<(v2f64 (X86Unpckl VR128:$src1, VR128:$src2)),
(UNPCKLPDrr VR128:$src1, VR128:$src2)>;
def : Pat<(v2f64 (X86Unpckh VR128:$src1, (memopv2f64 addr:$src2))),
(UNPCKHPDrm VR128:$src1, addr:$src2)>;
def : Pat<(v2f64 (X86Unpckh VR128:$src1, VR128:$src2)),
(UNPCKHPDrr VR128:$src1, VR128:$src2)>;
// FIXME: Instead of X86Movddup, there should be a X86Unpckl here, the
// problem is during lowering, where it's not possible to recognize the load
// fold cause it has two uses through a bitcast. One use disappears at isel
// time and the fold opportunity reappears.
def : Pat<(v2f64 (X86Movddup VR128:$src)),
(UNPCKLPDrr VR128:$src, VR128:$src)>;
let AddedComplexity = 10 in
def : Pat<(splat_lo (v2f64 VR128:$src), (undef)),
(UNPCKLPDrr VR128:$src, VR128:$src)>;
}
//===----------------------------------------------------------------------===// //===----------------------------------------------------------------------===//
// SSE 1 & 2 - Extract Floating-Point Sign mask // SSE 1 & 2 - Extract Floating-Point Sign mask
//===----------------------------------------------------------------------===// //===----------------------------------------------------------------------===//