T0b1/wasmtime
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Optimize 0.0 floating point constants. (#344)

Browse Source 
* Optimize 0.0 floating point constants. Rather than using the existing
process of emitting bit patterns and moving them into floating point
registers, use the `xorps` instruction to zero out the register.

* is_zero predicate function will not accept negative zero. Fixed formatting for encoding recipe and filetests.
This commit is contained in:
data-pup
2018-05-24 17:16:25 -04:00
committed by Dan Gohman
parent 4afb28ef59
commit 191bab162b
6 changed files with 134 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

16
lib/codegen/meta/isa/x86/encodings.py
View File

@@ -2,10 +2,12 @@
x86 Encodings.
"""
from __future__ import absolute_import
from cdsl.predicates import IsZero32BitFloat, IsZero64BitFloat
from cdsl.predicates import IsUnsignedInt, Not, And
from base.predicates import IsColocatedFunc, IsColocatedData, LengthEquals
from base import instructions as base
from base.formats import UnaryImm, FuncAddr, Call, LoadComplex, StoreComplex
from base.formats import UnaryIeee32, UnaryIeee64, UnaryImm
from base.formats import FuncAddr, Call, LoadComplex, StoreComplex
from .defs import X86_64, X86_32
from . import recipes as r
from . import settings as cfg
@@ -604,6 +606,18 @@ X86_64.enc(base.uextend.i64.i32, *r.umr(0x89))
# Floating point
#
# floating-point constants equal to 0.0 can be encoded using either
# `xorps` or `xorpd`, for 32-bit and 64-bit floats respectively.
X86_32.enc(base.f32const, *r.f32imm_z(0x0f, 0x57),
instp=IsZero32BitFloat(UnaryIeee32.imm))
X86_32.enc(base.f64const, *r.f64imm_z(0x66, 0x0f, 0x57),
instp=IsZero64BitFloat(UnaryIeee64.imm))
enc_x86_64_instp(base.f32const, r.f32imm_z,
IsZero32BitFloat(UnaryIeee32.imm), 0x0f, 0x57)
enc_x86_64_instp(base.f64const, r.f64imm_z,
IsZero64BitFloat(UnaryIeee64.imm), 0x66, 0x0f, 0x57)
# movd
enc_both(base.bitcast.f32.i32, r.frurm, 0x66, 0x0f, 0x6e)
enc_both(base.bitcast.i32.f32, r.rfumr, 0x66, 0x0f, 0x7e)