diff options
| author | Craig Topper <craig.topper@gmail.com> | 2013-09-18 03:55:53 +0000 |
|---|---|---|
| committer | Craig Topper <craig.topper@gmail.com> | 2013-09-18 03:55:53 +0000 |
| commit | 98064b9f4d5da172c8d09625edb60f3cc1e9a210 (patch) | |
| tree | 8ce171be4b01b62a7fec0a9b71f9797d0aa7769d /llvm/lib | |
| parent | c4d7c82c7f7356ac9aca70d67593479897ad7f2b (diff) | |
| download | bcm5719-llvm-98064b9f4d5da172c8d09625edb60f3cc1e9a210.tar.gz bcm5719-llvm-98064b9f4d5da172c8d09625edb60f3cc1e9a210.zip | |
Lift alignment restrictions for load/store folding on VINSERTF128/VEXTRACTF128. Fixes PR17268.
llvm-svn: 190916
Diffstat (limited to 'llvm/lib')
| -rw-r--r-- | llvm/lib/Target/X86/X86InstrSSE.td | 20 | ||||
| -rw-r--r-- | llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp | 2 |
2 files changed, 11 insertions, 11 deletions
diff --git a/llvm/lib/Target/X86/X86InstrSSE.td b/llvm/lib/Target/X86/X86InstrSSE.td index 2b271e92c59..81ea2752e69 100644 --- a/llvm/lib/Target/X86/X86InstrSSE.td +++ b/llvm/lib/Target/X86/X86InstrSSE.td @@ -7716,11 +7716,11 @@ def : Pat<(vinsert128_insert:$ins (v4f64 VR256:$src1), (v2f64 VR128:$src2), (VINSERTF128rr VR256:$src1, VR128:$src2, (INSERT_get_vinsert128_imm VR256:$ins))>; -def : Pat<(vinsert128_insert:$ins (v8f32 VR256:$src1), (memopv4f32 addr:$src2), +def : Pat<(vinsert128_insert:$ins (v8f32 VR256:$src1), (loadv4f32 addr:$src2), (iPTR imm)), (VINSERTF128rm VR256:$src1, addr:$src2, (INSERT_get_vinsert128_imm VR256:$ins))>; -def : Pat<(vinsert128_insert:$ins (v4f64 VR256:$src1), (memopv2f64 addr:$src2), +def : Pat<(vinsert128_insert:$ins (v4f64 VR256:$src1), (loadv2f64 addr:$src2), (iPTR imm)), (VINSERTF128rm VR256:$src1, addr:$src2, (INSERT_get_vinsert128_imm VR256:$ins))>; @@ -7744,22 +7744,22 @@ def : Pat<(vinsert128_insert:$ins (v16i16 VR256:$src1), (v8i16 VR128:$src2), (VINSERTF128rr VR256:$src1, VR128:$src2, (INSERT_get_vinsert128_imm VR256:$ins))>; -def : Pat<(vinsert128_insert:$ins (v4i64 VR256:$src1), (memopv2i64 addr:$src2), +def : Pat<(vinsert128_insert:$ins (v4i64 VR256:$src1), (loadv2i64 addr:$src2), (iPTR imm)), (VINSERTF128rm VR256:$src1, addr:$src2, (INSERT_get_vinsert128_imm VR256:$ins))>; def : Pat<(vinsert128_insert:$ins (v8i32 VR256:$src1), - (bc_v4i32 (memopv2i64 addr:$src2)), + (bc_v4i32 (loadv2i64 addr:$src2)), (iPTR imm)), (VINSERTF128rm VR256:$src1, addr:$src2, (INSERT_get_vinsert128_imm VR256:$ins))>; def : Pat<(vinsert128_insert:$ins (v32i8 VR256:$src1), - (bc_v16i8 (memopv2i64 addr:$src2)), + (bc_v16i8 (loadv2i64 addr:$src2)), (iPTR imm)), (VINSERTF128rm VR256:$src1, addr:$src2, (INSERT_get_vinsert128_imm VR256:$ins))>; def : Pat<(vinsert128_insert:$ins (v16i16 VR256:$src1), - (bc_v8i16 (memopv2i64 addr:$src2)), + (bc_v8i16 (loadv2i64 addr:$src2)), (iPTR imm)), (VINSERTF128rm VR256:$src1, addr:$src2, (INSERT_get_vinsert128_imm VR256:$ins))>; @@ -7791,12 +7791,12 @@ def : Pat<(vextract128_extract:$ext VR256:$src1, (iPTR imm)), (v4f64 VR256:$src1), (EXTRACT_get_vextract128_imm VR128:$ext)))>; -def : Pat<(alignedstore (v4f32 (vextract128_extract:$ext (v8f32 VR256:$src1), - (iPTR imm))), addr:$dst), +def : Pat<(store (v4f32 (vextract128_extract:$ext (v8f32 VR256:$src1), + (iPTR imm))), addr:$dst), (VEXTRACTF128mr addr:$dst, VR256:$src1, (EXTRACT_get_vextract128_imm VR128:$ext))>; -def : Pat<(alignedstore (v2f64 (vextract128_extract:$ext (v4f64 VR256:$src1), - (iPTR imm))), addr:$dst), +def : Pat<(store (v2f64 (vextract128_extract:$ext (v4f64 VR256:$src1), + (iPTR imm))), addr:$dst), (VEXTRACTF128mr addr:$dst, VR256:$src1, (EXTRACT_get_vextract128_imm VR128:$ext))>; } diff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp index b287ca7c8d5..c2427e72d84 100644 --- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp +++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp @@ -51,7 +51,7 @@ static cl::opt<int> "number ")); namespace { -static const unsigned MinVecRegSize = 128; +static const unsigned MinVecRegSize = 256; static const unsigned RecursionMaxDepth = 12; |
