diff options
| author | Dan Gohman <gohman@apple.com> | 2008-12-03 05:21:24 +0000 |
|---|---|---|
| committer | Dan Gohman <gohman@apple.com> | 2008-12-03 05:21:24 +0000 |
| commit | 62c939d7d5572e57963a5f26fb6fe802e13dc0bf (patch) | |
| tree | 40550321fe3caaaa7a081dc7d7ce4faa4339e234 /lib/Target/X86/X86InstrSSE.td | |
| parent | b3cf03cee1ce45a0b774bc8a8057ff7931a53215 (diff) | |
Mark x86's V_SET0 and V_SETALLONES with isSimpleLoad, and teach X86's
foldMemoryOperand how to "fold" them, by converting them into constant-pool
loads. When they aren't folded, they use xorps/cmpeqd, but for example when
register pressure is high, they may now be folded as memory operands, which
reduces register pressure.
Also, mark V_SET0 isAsCheapAsAMove so that two-address-elimination will
remat it instead of copying zeros around (V_SETALLONES was already marked).
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@60461 91177308-0d34-0410-b5e6-96231b3b80d8
Diffstat (limited to 'lib/Target/X86/X86InstrSSE.td')
| -rw-r--r-- | lib/Target/X86/X86InstrSSE.td | 8 |
1 files changed, 6 insertions, 2 deletions
diff --git a/lib/Target/X86/X86InstrSSE.td b/lib/Target/X86/X86InstrSSE.td index 426bddcea0..58e923d6df 100644 --- a/lib/Target/X86/X86InstrSSE.td +++ b/lib/Target/X86/X86InstrSSE.td @@ -987,7 +987,9 @@ def STMXCSR : PSI<0xAE, MRM3m, (outs), (ins i32mem:$dst), "stmxcsr\t$dst", [(int_x86_sse_stmxcsr addr:$dst)]>; // Alias instructions that map zero vector to pxor / xorp* for sse. -let isReMaterializable = 1 in +// We set isSimpleLoad because this can be converted to a constant-pool +// load of an all-zeros value if folding it would be beneficial. +let isReMaterializable = 1, isAsCheapAsAMove = 1, isSimpleLoad = 1 in def V_SET0 : PSI<0x57, MRMInitReg, (outs VR128:$dst), (ins), "xorps\t$dst, $dst", [(set VR128:$dst, (v4i32 immAllZerosV))]>; @@ -2253,7 +2255,9 @@ def : Pat<(membarrier (i8 imm:$ll), (i8 imm:$ls), (i8 imm:$sl), (i8 imm:$ss), (i8 1)), (MFENCE)>; // Alias instructions that map zero vector to pxor / xorp* for sse. -let isReMaterializable = 1, isAsCheapAsAMove = 1 in +// We set isSimpleLoad because this can be converted to a constant-pool +// load of an all-ones value if folding it would be beneficial. +let isReMaterializable = 1, isAsCheapAsAMove = 1, isSimpleLoad = 1 in def V_SETALLONES : PDI<0x76, MRMInitReg, (outs VR128:$dst), (ins), "pcmpeqd\t$dst, $dst", [(set VR128:$dst, (v4i32 immAllOnesV))]>; |