6502/llvm-6502
1
0
Fork 0
mirror of https://github.com/c64scene-ar/llvm-6502.git synced 2024-08-17 21:29:20 +00:00
Code Issues Projects Releases Wiki Activity

Prefer {p}unpack* and mov*dup over {p}shuf* as well.

Browse Source 
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@27844 91177308-0d34-0410-b5e6-96231b3b80d8
This commit is contained in:
Evan Cheng 2006-04-19 21:15:24 +00:00
parent f5e1dc278b
commit fd111b5be5
1 changed files with 13 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

16
lib/Target/X86/X86InstrSSE.td
View File

@ -778,7 +778,7 @@ def MOVUPDmr : PDI<0x11, MRMDestMem, (ops f128mem:$dst, VR128:$src),
[(int_x86_sse2_storeu_pd addr:$dst, VR128:$src)]>;
let isTwoAddress = 1 in {
let AddedComplexity = 10 in {
let AddedComplexity = 20 in {
def MOVLPSrm : PSI<0x12, MRMSrcMem, (ops VR128:$dst, VR128:$src1, f64mem:$src2),
"movlps {$src2, $dst|$dst, $src2}",
[(set VR128:$dst,
@ -832,7 +832,7 @@ def MOVHPDmr : PDI<0x17, MRMDestMem, (ops f64mem:$dst, VR128:$src),
addr:$dst)]>;
let isTwoAddress = 1 in {
let AddedComplexity = 10 in {
let AddedComplexity = 20 in {
def MOVLHPSrr : PSI<0x16, MRMSrcReg, (ops VR128:$dst, VR128:$src1, VR128:$src2),
"movlhps {$src2, $dst|$dst, $src2}",
[(set VR128:$dst,
@ -1245,6 +1245,7 @@ def SHUFPDrmi : PDIi8<0xC6, MRMSrcMem,
VR128:$src1, (load addr:$src2),
SHUFP_shuffle_mask:$src3)))]>;
let AddedComplexity = 10 in {
def UNPCKHPSrr : PSI<0x15, MRMSrcReg,
(ops VR128:$dst, VR128:$src1, VR128:$src2),
"unpckhps {$src2, $dst|$dst, $src2}",
@ -1294,6 +1295,7 @@ def UNPCKLPDrm : PDI<0x14, MRMSrcMem,
[(set VR128:$dst, (v2f64 (vector_shuffle
VR128:$src1, (load addr:$src2),
UNPCKL_shuffle_mask)))]>;
} // AddedComplexity
}
// Horizontal ops
@ -2195,6 +2197,7 @@ def MOVLSD2PDrr : SDI<0x10, MRMSrcReg, (ops VR128:$dst, VR128:$src1, FR64:$src2)
def MOVLDI2PDIrr : PDI<0x6E, MRMSrcReg, (ops VR128:$dst, VR128:$src1, R32:$src2),
"movd {$src2, $dst|$dst, $src2}", []>;
let AddedComplexity = 20 in {
def MOVLPSrr : SSI<0x10, MRMSrcReg, (ops VR128:$dst, VR128:$src1, VR128:$src2),
"movss {$src2, $dst|$dst, $src2}",
[(set VR128:$dst,
@ -2206,6 +2209,7 @@ def MOVLPDrr : SDI<0x10, MRMSrcReg, (ops VR128:$dst, VR128:$src1, VR128:$src2),
(v2f64 (vector_shuffle VR128:$src1, VR128:$src2,
MOVS_shuffle_mask)))]>;
}
}
// Store / copy lower 64-bits of a XMM register.
def MOVLQ128mr : PDI<0xD6, MRMDestMem, (ops i64mem:$dst, VR128:$src),
@ -2350,10 +2354,12 @@ def : Pat<(v16i8 (X86zexts2vec R8:$src)),
(MOVLDI2PDIrr (V_SET0_PI), (MOVZX32rr8 R8:$src))>, Requires<[HasSSE2]>;
// Splat v2f64 / v2i64
let AddedComplexity = 10 in {
def : Pat<(vector_shuffle (v2f64 VR128:$src), (undef), SSE_splat_v2_mask:$sm),
(v2f64 (UNPCKLPDrr VR128:$src, VR128:$src))>, Requires<[HasSSE2]>;
def : Pat<(vector_shuffle (v2i64 VR128:$src), (undef), SSE_splat_v2_mask:$sm),
(v2i64 (PUNPCKLQDQrr VR128:$src, VR128:$src))>, Requires<[HasSSE2]>;
}
// Splat v4f32
def : Pat<(vector_shuffle (v4f32 VR128:$src), (undef), SSE_splat_mask:$sm),
@ -2382,6 +2388,7 @@ def : Pat<(vector_shuffle (v4i32 VR128:$src1),
PSHUFD_binary_shuffle_mask:$sm))>, Requires<[HasSSE2]>;
// vector_shuffle v1, <undef>, <0, 0, 1, 1, ...>
let AddedComplexity = 10 in {
def : Pat<(v4f32 (vector_shuffle VR128:$src, (undef),
UNPCKL_v_undef_shuffle_mask)),
(UNPCKLPSrr VR128:$src, VR128:$src)>, Requires<[HasSSE2]>;
@ -2394,7 +2401,9 @@ def : Pat<(v8i16 (vector_shuffle VR128:$src, (undef),
def : Pat<(v4i32 (vector_shuffle VR128:$src, (undef),
UNPCKL_v_undef_shuffle_mask)),
(PUNPCKLDQrr VR128:$src, VR128:$src)>, Requires<[HasSSE1]>;
}
let AddedComplexity = 20 in {
// vector_shuffle v1, <undef> <1, 1, 3, 3>
def : Pat<(v4i32 (vector_shuffle VR128:$src, (undef),
MOVSHDUP_shuffle_mask)),
@ -2410,8 +2419,9 @@ def : Pat<(v4i32 (vector_shuffle VR128:$src, (undef),
def : Pat<(v4i32 (vector_shuffle (bc_v4i32 (loadv2i64 addr:$src)), (undef),
MOVSLDUP_shuffle_mask)),
(MOVSLDUPrm addr:$src)>, Requires<[HasSSE3]>;
}
let AddedComplexity = 10 in {
let AddedComplexity = 20 in {
// vector_shuffle v1, v2 <0, 1, 4, 5> using MOVLHPS
def : Pat<(v4i32 (vector_shuffle VR128:$src1, VR128:$src2,
MOVHP_shuffle_mask)),