mirror of
https://github.com/golang/go
synced 2024-11-23 06:40:05 -07:00
cmd/compile: use ellipses in AMD64 rules
file before after Δ % compile 20801800 20743944 -57856 -0.278% total 131542652 131484796 -57856 -0.044% file before after Δ % cmd/compile/internal/ssa.a 23321562 23177930 -143632 -0.616% total 125602774 125459142 -143632 -0.114% Change-Id: I3bdcff87e76d95a4367738a55316bd561c719c5c Reviewed-on: https://go-review.googlesource.com/c/go/+/220349 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org>
This commit is contained in:
Josh Bleecher Snyder
parent
8484d409ac
commit
a37bbcecca
@ -3,29 +3,29 @@
|
||||
// license that can be found in the LICENSE file.
|
||||
|
||||
// Lowering arithmetic
|
||||
(Add(64|32|16|8) x y) -> (ADD(Q|L|L|L) x y)
|
||||
(AddPtr x y) -> (ADDQ x y)
|
||||
(Add(32|64)F x y) -> (ADDS(S|D) x y)
|
||||
(Add(64|32|16|8) ...) -> (ADD(Q|L|L|L) ...)
|
||||
(AddPtr ...) -> (ADDQ ...)
|
||||
(Add(32|64)F ...) -> (ADDS(S|D) ...)
|
||||
|
||||
(Sub(64|32|16|8) x y) -> (SUB(Q|L|L|L) x y)
|
||||
(SubPtr x y) -> (SUBQ x y)
|
||||
(Sub(32|64)F x y) -> (SUBS(S|D) x y)
|
||||
(Sub(64|32|16|8) ...) -> (SUB(Q|L|L|L) ...)
|
||||
(SubPtr ...) -> (SUBQ ...)
|
||||
(Sub(32|64)F ...) -> (SUBS(S|D) ...)
|
||||
|
||||
(Mul(64|32|16|8) x y) -> (MUL(Q|L|L|L) x y)
|
||||
(Mul(32|64)F x y) -> (MULS(S|D) x y)
|
||||
(Mul(64|32|16|8) ...) -> (MUL(Q|L|L|L) ...)
|
||||
(Mul(32|64)F ...) -> (MULS(S|D) ...)
|
||||
|
||||
(Select0 (Mul64uover x y)) -> (Select0 <typ.UInt64> (MULQU x y))
|
||||
(Select0 (Mul32uover x y)) -> (Select0 <typ.UInt32> (MULLU x y))
|
||||
(Select1 (Mul(64|32)uover x y)) -> (SETO (Select1 <types.TypeFlags> (MUL(Q|L)U x y)))
|
||||
|
||||
(Hmul(64|32) x y) -> (HMUL(Q|L) x y)
|
||||
(Hmul(64|32)u x y) -> (HMUL(Q|L)U x y)
|
||||
(Hmul(64|32) ...) -> (HMUL(Q|L) ...)
|
||||
(Hmul(64|32)u ...) -> (HMUL(Q|L)U ...)
|
||||
|
||||
(Div(64|32|16) [a] x y) -> (Select0 (DIV(Q|L|W) [a] x y))
|
||||
(Div8 x y) -> (Select0 (DIVW (SignExt8to16 x) (SignExt8to16 y)))
|
||||
(Div(64|32|16)u x y) -> (Select0 (DIV(Q|L|W)U x y))
|
||||
(Div8u x y) -> (Select0 (DIVWU (ZeroExt8to16 x) (ZeroExt8to16 y)))
|
||||
(Div(32|64)F x y) -> (DIVS(S|D) x y)
|
||||
(Div(32|64)F ...) -> (DIVS(S|D) ...)
|
||||
|
||||
(Select0 (Add64carry x y c)) ->
|
||||
(Select0 <typ.UInt64> (ADCQ x y (Select1 <types.TypeFlags> (NEGLflags c))))
|
||||
@ -49,28 +49,28 @@
|
||||
(Select1 (NEGLflags (NEGQ (SBBQcarrymask x)))) -> x
|
||||
|
||||
|
||||
(Mul64uhilo x y) -> (MULQU2 x y)
|
||||
(Div128u xhi xlo y) -> (DIVQU2 xhi xlo y)
|
||||
(Mul64uhilo ...) -> (MULQU2 ...)
|
||||
(Div128u ...) -> (DIVQU2 ...)
|
||||
|
||||
(Avg64u x y) -> (AVGQU x y)
|
||||
(Avg64u ...) -> (AVGQU ...)
|
||||
|
||||
(Mod(64|32|16) [a] x y) -> (Select1 (DIV(Q|L|W) [a] x y))
|
||||
(Mod8 x y) -> (Select1 (DIVW (SignExt8to16 x) (SignExt8to16 y)))
|
||||
(Mod(64|32|16)u x y) -> (Select1 (DIV(Q|L|W)U x y))
|
||||
(Mod8u x y) -> (Select1 (DIVWU (ZeroExt8to16 x) (ZeroExt8to16 y)))
|
||||
|
||||
(And(64|32|16|8) x y) -> (AND(Q|L|L|L) x y)
|
||||
(Or(64|32|16|8) x y) -> (OR(Q|L|L|L) x y)
|
||||
(Xor(64|32|16|8) x y) -> (XOR(Q|L|L|L) x y)
|
||||
(Com(64|32|16|8) x) -> (NOT(Q|L|L|L) x)
|
||||
(And(64|32|16|8) ...) -> (AND(Q|L|L|L) ...)
|
||||
(Or(64|32|16|8) ...) -> (OR(Q|L|L|L) ...)
|
||||
(Xor(64|32|16|8) ...) -> (XOR(Q|L|L|L) ...)
|
||||
(Com(64|32|16|8) ...) -> (NOT(Q|L|L|L) ...)
|
||||
|
||||
(Neg(64|32|16|8) x) -> (NEG(Q|L|L|L) x)
|
||||
(Neg(64|32|16|8) ...) -> (NEG(Q|L|L|L) ...)
|
||||
(Neg32F x) -> (PXOR x (MOVSSconst <typ.Float32> [auxFrom32F(float32(math.Copysign(0, -1)))]))
|
||||
(Neg64F x) -> (PXOR x (MOVSDconst <typ.Float64> [auxFrom64F(math.Copysign(0, -1))]))
|
||||
|
||||
// Lowering boolean ops
|
||||
(AndB x y) -> (ANDL x y)
|
||||
(OrB x y) -> (ORL x y)
|
||||
(AndB ...) -> (ANDL ...)
|
||||
(OrB ...) -> (ORL ...)
|
||||
(Not x) -> (XORLconst [1] x)
|
||||
|
||||
// Lowering pointer arithmetic
|
||||
@ -84,9 +84,9 @@
|
||||
(Ctz8 x) -> (BSFL (BTSLconst <typ.UInt32> [ 8] x))
|
||||
|
||||
(Ctz64NonZero x) -> (Select0 (BSFQ x))
|
||||
(Ctz32NonZero x) -> (BSFL x)
|
||||
(Ctz16NonZero x) -> (BSFL x)
|
||||
(Ctz8NonZero x) -> (BSFL x)
|
||||
(Ctz32NonZero ...) -> (BSFL ...)
|
||||
(Ctz16NonZero ...) -> (BSFL ...)
|
||||
(Ctz8NonZero ...) -> (BSFL ...)
|
||||
|
||||
// BitLen64 of a 64 bit value x requires checking whether x == 0, since BSRQ is undefined when x == 0.
|
||||
// However, for zero-extended values, we can cheat a bit, and calculate
|
||||
@ -97,14 +97,13 @@
|
||||
(BitLen16 x) -> (BSRL (LEAL1 <typ.UInt32> [1] (MOVWQZX <typ.UInt32> x) (MOVWQZX <typ.UInt32> x)))
|
||||
(BitLen8 x) -> (BSRL (LEAL1 <typ.UInt32> [1] (MOVBQZX <typ.UInt32> x) (MOVBQZX <typ.UInt32> x)))
|
||||
|
||||
(Bswap(64|32) x) -> (BSWAP(Q|L) x)
|
||||
(Bswap(64|32) ...) -> (BSWAP(Q|L) ...)
|
||||
|
||||
(PopCount64 x) -> (POPCNTQ x)
|
||||
(PopCount32 x) -> (POPCNTL x)
|
||||
(PopCount(64|32) ...) -> (POPCNT(Q|L) ...)
|
||||
(PopCount16 x) -> (POPCNTL (MOVWQZX <typ.UInt32> x))
|
||||
(PopCount8 x) -> (POPCNTL (MOVBQZX <typ.UInt32> x))
|
||||
|
||||
(Sqrt x) -> (SQRTSD x)
|
||||
(Sqrt ...) -> (SQRTSD ...)
|
||||
|
||||
(RoundToEven x) -> (ROUNDSD [0] x)
|
||||
(Floor x) -> (ROUNDSD [1] x)
|
||||
@ -114,46 +113,46 @@
|
||||
|
||||
// Lowering extension
|
||||
// Note: we always extend to 64 bits even though some ops don't need that many result bits.
|
||||
(SignExt8to16 x) -> (MOVBQSX x)
|
||||
(SignExt8to32 x) -> (MOVBQSX x)
|
||||
(SignExt8to64 x) -> (MOVBQSX x)
|
||||
(SignExt16to32 x) -> (MOVWQSX x)
|
||||
(SignExt16to64 x) -> (MOVWQSX x)
|
||||
(SignExt32to64 x) -> (MOVLQSX x)
|
||||
(SignExt8to16 ...) -> (MOVBQSX ...)
|
||||
(SignExt8to32 ...) -> (MOVBQSX ...)
|
||||
(SignExt8to64 ...) -> (MOVBQSX ...)
|
||||
(SignExt16to32 ...) -> (MOVWQSX ...)
|
||||
(SignExt16to64 ...) -> (MOVWQSX ...)
|
||||
(SignExt32to64 ...) -> (MOVLQSX ...)
|
||||
|
||||
(ZeroExt8to16 x) -> (MOVBQZX x)
|
||||
(ZeroExt8to32 x) -> (MOVBQZX x)
|
||||
(ZeroExt8to64 x) -> (MOVBQZX x)
|
||||
(ZeroExt16to32 x) -> (MOVWQZX x)
|
||||
(ZeroExt16to64 x) -> (MOVWQZX x)
|
||||
(ZeroExt32to64 x) -> (MOVLQZX x)
|
||||
(ZeroExt8to16 ...) -> (MOVBQZX ...)
|
||||
(ZeroExt8to32 ...) -> (MOVBQZX ...)
|
||||
(ZeroExt8to64 ...) -> (MOVBQZX ...)
|
||||
(ZeroExt16to32 ...) -> (MOVWQZX ...)
|
||||
(ZeroExt16to64 ...) -> (MOVWQZX ...)
|
||||
(ZeroExt32to64 ...) -> (MOVLQZX ...)
|
||||
|
||||
(Slicemask <t> x) -> (SARQconst (NEGQ <t> x) [63])
|
||||
|
||||
// Lowering truncation
|
||||
// Because we ignore high parts of registers, truncates are just copies.
|
||||
(Trunc16to8 x) -> x
|
||||
(Trunc32to8 x) -> x
|
||||
(Trunc32to16 x) -> x
|
||||
(Trunc64to8 x) -> x
|
||||
(Trunc64to16 x) -> x
|
||||
(Trunc64to32 x) -> x
|
||||
(Trunc16to8 ...) -> (Copy ...)
|
||||
(Trunc32to8 ...) -> (Copy ...)
|
||||
(Trunc32to16 ...) -> (Copy ...)
|
||||
(Trunc64to8 ...) -> (Copy ...)
|
||||
(Trunc64to16 ...) -> (Copy ...)
|
||||
(Trunc64to32 ...) -> (Copy ...)
|
||||
|
||||
// Lowering float <-> int
|
||||
(Cvt32to32F x) -> (CVTSL2SS x)
|
||||
(Cvt32to64F x) -> (CVTSL2SD x)
|
||||
(Cvt64to32F x) -> (CVTSQ2SS x)
|
||||
(Cvt64to64F x) -> (CVTSQ2SD x)
|
||||
(Cvt32to32F ...) -> (CVTSL2SS ...)
|
||||
(Cvt32to64F ...) -> (CVTSL2SD ...)
|
||||
(Cvt64to32F ...) -> (CVTSQ2SS ...)
|
||||
(Cvt64to64F ...) -> (CVTSQ2SD ...)
|
||||
|
||||
(Cvt32Fto32 x) -> (CVTTSS2SL x)
|
||||
(Cvt32Fto64 x) -> (CVTTSS2SQ x)
|
||||
(Cvt64Fto32 x) -> (CVTTSD2SL x)
|
||||
(Cvt64Fto64 x) -> (CVTTSD2SQ x)
|
||||
(Cvt32Fto32 ...) -> (CVTTSS2SL ...)
|
||||
(Cvt32Fto64 ...) -> (CVTTSS2SQ ...)
|
||||
(Cvt64Fto32 ...) -> (CVTTSD2SL ...)
|
||||
(Cvt64Fto64 ...) -> (CVTTSD2SQ ...)
|
||||
|
||||
(Cvt32Fto64F x) -> (CVTSS2SD x)
|
||||
(Cvt64Fto32F x) -> (CVTSD2SS x)
|
||||
(Cvt32Fto64F ...) -> (CVTSS2SD ...)
|
||||
(Cvt64Fto32F ...) -> (CVTSD2SS ...)
|
||||
|
||||
(Round(32|64)F x) -> x
|
||||
(Round(32|64)F ...) -> (Copy ...)
|
||||
|
||||
// Lowering shifts
|
||||
// Unsigned shifts need to return 0 if shift amount is >= width of shifted value.
|
||||
@ -409,19 +408,19 @@
|
||||
(REPSTOSQ destptr (MOVQconst [s/8]) (MOVQconst [0]) mem)
|
||||
|
||||
// Lowering constants
|
||||
(Const8 [val]) -> (MOVLconst [val])
|
||||
(Const16 [val]) -> (MOVLconst [val])
|
||||
(Const32 [val]) -> (MOVLconst [val])
|
||||
(Const64 [val]) -> (MOVQconst [val])
|
||||
(Const32F [val]) -> (MOVSSconst [val])
|
||||
(Const64F [val]) -> (MOVSDconst [val])
|
||||
(ConstNil) -> (MOVQconst [0])
|
||||
(ConstBool [b]) -> (MOVLconst [b])
|
||||
(Const8 ...) -> (MOVLconst ...)
|
||||
(Const16 ...) -> (MOVLconst ...)
|
||||
(Const32 ...) -> (MOVLconst ...)
|
||||
(Const64 ...) -> (MOVQconst ...)
|
||||
(Const32F ...) -> (MOVSSconst ...)
|
||||
(Const64F ...) -> (MOVSDconst ...)
|
||||
(ConstNil ...) -> (MOVQconst ...)
|
||||
(ConstBool ...) -> (MOVLconst ...)
|
||||
|
||||
// Lowering calls
|
||||
(StaticCall [argwid] {target} mem) -> (CALLstatic [argwid] {target} mem)
|
||||
(ClosureCall [argwid] entry closure mem) -> (CALLclosure [argwid] entry closure mem)
|
||||
(InterCall [argwid] entry mem) -> (CALLinter [argwid] entry mem)
|
||||
(StaticCall ...) -> (CALLstatic ...)
|
||||
(ClosureCall ...) -> (CALLclosure ...)
|
||||
(InterCall ...) -> (CALLinter ...)
|
||||
|
||||
// Lowering conditional moves
|
||||
// If the condition is a SETxx, we can just run a CMOV from the comparison that was
|
||||
@ -473,12 +472,12 @@
|
||||
(IsNonNil p) -> (SETNE (TESTQ p p))
|
||||
(IsInBounds idx len) -> (SETB (CMPQ idx len))
|
||||
(IsSliceInBounds idx len) -> (SETBE (CMPQ idx len))
|
||||
(NilCheck ptr mem) -> (LoweredNilCheck ptr mem)
|
||||
(GetG mem) -> (LoweredGetG mem)
|
||||
(GetClosurePtr) -> (LoweredGetClosurePtr)
|
||||
(GetCallerPC) -> (LoweredGetCallerPC)
|
||||
(GetCallerSP) -> (LoweredGetCallerSP)
|
||||
(Addr {sym} base) -> (LEAQ {sym} base)
|
||||
(NilCheck ...) -> (LoweredNilCheck ...)
|
||||
(GetG ...) -> (LoweredGetG ...)
|
||||
(GetClosurePtr ...) -> (LoweredGetClosurePtr ...)
|
||||
(GetCallerPC ...) -> (LoweredGetCallerPC ...)
|
||||
(GetCallerSP ...) -> (LoweredGetCallerSP ...)
|
||||
(Addr ...) -> (LEAQ ...)
|
||||
(LocalAddr {sym} base _) -> (LEAQ {sym} base)
|
||||
|
||||
(MOVBstore [off] {sym} ptr y:(SETL x) mem) && y.Uses == 1 -> (SETLstore [off] {sym} ptr x mem)
|
||||
@ -514,10 +513,10 @@
|
||||
(If cond yes no) -> (NE (TESTB cond cond) yes no)
|
||||
|
||||
// Atomic loads. Other than preserving their ordering with respect to other loads, nothing special here.
|
||||
(AtomicLoad8 ptr mem) -> (MOVBatomicload ptr mem)
|
||||
(AtomicLoad32 ptr mem) -> (MOVLatomicload ptr mem)
|
||||
(AtomicLoad64 ptr mem) -> (MOVQatomicload ptr mem)
|
||||
(AtomicLoadPtr ptr mem) -> (MOVQatomicload ptr mem)
|
||||
(AtomicLoad8 ...) -> (MOVBatomicload ...)
|
||||
(AtomicLoad32 ...) -> (MOVLatomicload ...)
|
||||
(AtomicLoad64 ...) -> (MOVQatomicload ...)
|
||||
(AtomicLoadPtr ...) -> (MOVQatomicload ...)
|
||||
|
||||
// Atomic stores. We use XCHG to prevent the hardware reordering a subsequent load.
|
||||
// TODO: most runtime uses of atomic stores don't need that property. Use normal stores for those?
|
||||
@ -539,15 +538,15 @@
|
||||
(Select1 (AddTupleFirst64 _ tuple)) -> (Select1 tuple)
|
||||
|
||||
// Atomic compare and swap.
|
||||
(AtomicCompareAndSwap32 ptr old new_ mem) -> (CMPXCHGLlock ptr old new_ mem)
|
||||
(AtomicCompareAndSwap64 ptr old new_ mem) -> (CMPXCHGQlock ptr old new_ mem)
|
||||
(AtomicCompareAndSwap32 ...) -> (CMPXCHGLlock ...)
|
||||
(AtomicCompareAndSwap64 ...) -> (CMPXCHGQlock ...)
|
||||
|
||||
// Atomic memory updates.
|
||||
(AtomicAnd8 ptr val mem) -> (ANDBlock ptr val mem)
|
||||
(AtomicOr8 ptr val mem) -> (ORBlock ptr val mem)
|
||||
(AtomicAnd8 ...) -> (ANDBlock ...)
|
||||
(AtomicOr8 ...) -> (ORBlock ...)
|
||||
|
||||
// Write barrier.
|
||||
(WB {fn} destptr srcptr mem) -> (LoweredWB {fn} destptr srcptr mem)
|
||||
(WB ...) -> (LoweredWB ...)
|
||||
|
||||
(PanicBounds [kind] x y mem) && boundsABI(kind) == 0 -> (LoweredPanicBoundsA [kind] x y mem)
|
||||
(PanicBounds [kind] x y mem) && boundsABI(kind) == 1 -> (LoweredPanicBoundsB [kind] x y mem)
|
||||
@ -807,10 +806,10 @@
|
||||
(ROLWconst [c] (ROLWconst [d] x)) -> (ROLWconst [(c+d)&15] x)
|
||||
(ROLBconst [c] (ROLBconst [d] x)) -> (ROLBconst [(c+d)& 7] x)
|
||||
|
||||
(RotateLeft8 a b) -> (ROLB a b)
|
||||
(RotateLeft16 a b) -> (ROLW a b)
|
||||
(RotateLeft32 a b) -> (ROLL a b)
|
||||
(RotateLeft64 a b) -> (ROLQ a b)
|
||||
(RotateLeft8 ...) -> (ROLB ...)
|
||||
(RotateLeft16 ...) -> (ROLW ...)
|
||||
(RotateLeft32 ...) -> (ROLL ...)
|
||||
(RotateLeft64 ...) -> (ROLQ ...)
|
||||
|
||||
// Non-constant rotates.
|
||||
// We want to issue a rotate when the Go source contains code like
|
||||
|
||||
File diff suppressed because it is too large
Load Diff
Loading…
Reference in New Issue
Block a user