Skip to content

Commit

Permalink
cmd/compile: eliminate unnecessary type conversions in TrailingZeros(…
Browse files Browse the repository at this point in the history 
…16|8) for arm64

This CL eliminates unnecessary type conversion operations: OpZeroExt16to64 and OpZeroExt8to64.
If the input argrument is a nonzero value, then ORconst operation can also be eliminated.

Benchmarks:

name               old time/op  new time/op  delta
TrailingZeros-8    2.75ns ± 0%  2.75ns ± 0%     ~     (all equal)
TrailingZeros8-8   3.49ns ± 1%  2.93ns ± 0%  -16.00%  (p=0.000 n=10+10)
TrailingZeros16-8  3.49ns ± 1%  2.93ns ± 0%  -16.05%  (p=0.000 n=9+10)
TrailingZeros32-8  2.67ns ± 1%  2.68ns ± 1%     ~     (p=0.468 n=10+10)
TrailingZeros64-8  2.67ns ± 1%  2.65ns ± 0%   -0.62%  (p=0.022 n=10+9)

code:

func f16(x uint) { z = bits.TrailingZeros16(uint16(x)) }

Before:

"".f16 STEXT size=48 args=0x8 locals=0x0 leaf
        0x0000 00000 (test.go:7)        TEXT    "".f16(SB), LEAF|NOFRAME|ABIInternal, $0-8
        0x0000 00000 (test.go:7)        FUNCDATA        ZR, gclocals·33cdeccccebe80329f1fdbee7f5874cb(SB)
        0x0000 00000 (test.go:7)        FUNCDATA        $1, gclocals·33cdeccccebe80329f1fdbee7f5874cb(SB)
        0x0000 00000 (test.go:7)        FUNCDATA        $3, gclocals·33cdeccccebe80329f1fdbee7f5874cb(SB)
        0x0000 00000 (test.go:7)        PCDATA  $2, ZR
        0x0000 00000 (test.go:7)        PCDATA  ZR, ZR
        0x0000 00000 (test.go:7)        MOVD    "".x(FP), R0
        0x0004 00004 (test.go:7)        MOVHU   R0, R0
        0x0008 00008 (test.go:7)        ORR     $65536, R0, R0
        0x000c 00012 (test.go:7)        RBIT    R0, R0
        0x0010 00016 (test.go:7)        CLZ     R0, R0
        0x0014 00020 (test.go:7)        MOVD    R0, "".z(SB)
        0x0020 00032 (test.go:7)        RET     (R30)

This line of code is unnecessary:
        0x0004 00004 (test.go:7)        MOVHU   R0, R0

After:

"".f16 STEXT size=32 args=0x8 locals=0x0 leaf
        0x0000 00000 (test.go:7)        TEXT    "".f16(SB), LEAF|NOFRAME|ABIInternal, $0-8
        0x0000 00000 (test.go:7)        FUNCDATA        ZR, gclocals·33cdeccccebe80329f1fdbee7f5874cb(SB)
        0x0000 00000 (test.go:7)        FUNCDATA        $1, gclocals·33cdeccccebe80329f1fdbee7f5874cb(SB)
        0x0000 00000 (test.go:7)        FUNCDATA        $3, gclocals·33cdeccccebe80329f1fdbee7f5874cb(SB)
        0x0000 00000 (test.go:7)        PCDATA  $2, ZR
        0x0000 00000 (test.go:7)        PCDATA  ZR, ZR
        0x0000 00000 (test.go:7)        MOVD    "".x(FP), R0
        0x0004 00004 (test.go:7)        ORR     $65536, R0, R0
        0x0008 00008 (test.go:7)        RBITW   R0, R0
        0x000c 00012 (test.go:7)        CLZW    R0, R0
        0x0010 00016 (test.go:7)        MOVD    R0, "".z(SB)
        0x001c 00028 (test.go:7)        RET     (R30)

The situation of TrailingZeros8 is similar to TrailingZeros16.

Change-Id: I473bdca06be8460a0be87abbae6fe640017e4c9d
Reviewed-on: https://go-review.googlesource.com/c/go/+/156999
Reviewed-by: Cherry Zhang <[email protected]>
Run-TryBot: Cherry Zhang <[email protected]>
TryBot-Result: Gobot Gobot <[email protected]>
  • Loading branch information
@cherrymui
erifan01 authored and cherrymui committed Mar 7, 2019
1 parent fee84cc commit 4e2b0dd
Show file tree
Hide file tree
Showing 4 changed files with 89 additions and 4 deletions.
8 changes: 4 additions & 4 deletions src/cmd/compile/internal/gc/ssa.go
View file
Original file line number Diff line number Diff line change
Expand Up @@ -3290,15 +3290,15 @@ func init() {
func(s *state, n *Node, args []*ssa.Value) *ssa.Value {
return s.newValue1(ssa.OpCtz16, types.Types[TINT], args[0])
},
sys.AMD64)
sys.AMD64, sys.ARM64)
addF("math/bits", "TrailingZeros16",
func(s *state, n *Node, args []*ssa.Value) *ssa.Value {
x := s.newValue1(ssa.OpZeroExt16to64, types.Types[TUINT64], args[0])
c := s.constInt64(types.Types[TUINT64], 1<<16)
y := s.newValue2(ssa.OpOr64, types.Types[TUINT64], x, c)
return s.newValue1(ssa.OpCtz64, types.Types[TINT], y)
},
sys.ARM64, sys.S390X, sys.PPC64)
sys.S390X, sys.PPC64)
addF("math/bits", "TrailingZeros8",
func(s *state, n *Node, args []*ssa.Value) *ssa.Value {
x := s.newValue1(ssa.OpZeroExt8to32, types.Types[TUINT32], args[0])
Expand All @@ -3311,15 +3311,15 @@ func init() {
func(s *state, n *Node, args []*ssa.Value) *ssa.Value {
return s.newValue1(ssa.OpCtz8, types.Types[TINT], args[0])
},
sys.AMD64)
sys.AMD64, sys.ARM64)
addF("math/bits", "TrailingZeros8",
func(s *state, n *Node, args []*ssa.Value) *ssa.Value {
x := s.newValue1(ssa.OpZeroExt8to64, types.Types[TUINT64], args[0])
c := s.constInt64(types.Types[TUINT64], 1<<8)
y := s.newValue2(ssa.OpOr64, types.Types[TUINT64], x, c)
return s.newValue1(ssa.OpCtz64, types.Types[TINT], y)
},
sys.ARM64, sys.S390X)
sys.S390X)
alias("math/bits", "ReverseBytes64", "runtime/internal/sys", "Bswap64", all...)
alias("math/bits", "ReverseBytes32", "runtime/internal/sys", "Bswap32", all...)
// ReverseBytes inlines correctly, no need to intrinsify it.
Expand Down
4 changes: 4 additions & 0 deletions src/cmd/compile/internal/ssa/gen/ARM64.rules
View file
Original file line number Diff line number Diff line change
Expand Up @@ -97,9 +97,13 @@

(Ctz64NonZero x) -> (Ctz64 x)
(Ctz32NonZero x) -> (Ctz32 x)
(Ctz16NonZero x) -> (Ctz32 x)
(Ctz8NonZero x) -> (Ctz32 x)

(Ctz64 <t> x) -> (CLZ (RBIT <t> x))
(Ctz32 <t> x) -> (CLZW (RBITW <t> x))
(Ctz16 <t> x) -> (CLZW <t> (RBITW <typ.UInt32> (ORconst <typ.UInt32> [0x10000] x)))
(Ctz8 <t> x) -> (CLZW <t> (RBITW <typ.UInt32> (ORconst <typ.UInt32> [0x100] x)))

(PopCount64 <t> x) -> (FMOVDfpgp <t> (VUADDLV <typ.Float64> (VCNT <typ.Float64> (FMOVDgpfp <typ.Float64> x))))
(PopCount32 <t> x) -> (FMOVDfpgp <t> (VUADDLV <typ.Float64> (VCNT <typ.Float64> (FMOVDgpfp <typ.Float64> (ZeroExt32to64 x)))))
Expand Down
74 changes: 74 additions & 0 deletions src/cmd/compile/internal/ssa/rewriteARM64.go
View file

Some generated files are not rendered by default. Learn more about how customized files appear on GitHub.

7 changes: 7 additions & 0 deletions test/codegen/mathbits.go
View file
Original file line number Diff line number Diff line change
Expand Up @@ -242,6 +242,7 @@ func RotateLeftVariable32(n uint32, m int) uint32 {

func TrailingZeros(n uint) int {
// amd64:"BSFQ","MOVL\t\\$64","CMOVQEQ"
// arm64:"RBIT","CLZ"
// s390x:"FLOGR"
// ppc64:"ANDN","POPCNTD"
// ppc64le:"ANDN","POPCNTD"
Expand All @@ -250,6 +251,7 @@ func TrailingZeros(n uint) int {

func TrailingZeros64(n uint64) int {
// amd64:"BSFQ","MOVL\t\\$64","CMOVQEQ"
// arm64:"RBIT","CLZ"
// s390x:"FLOGR"
// ppc64:"ANDN","POPCNTD"
// ppc64le:"ANDN","POPCNTD"
Expand All @@ -258,6 +260,7 @@ func TrailingZeros64(n uint64) int {

func TrailingZeros32(n uint32) int {
// amd64:"BTSQ\\t\\$32","BSFQ"
// arm64:"RBITW","CLZW"
// s390x:"FLOGR","MOVWZ"
// ppc64:"ANDN","POPCNTW"
// ppc64le:"ANDN","POPCNTW"
Expand All @@ -266,6 +269,7 @@ func TrailingZeros32(n uint32) int {

func TrailingZeros16(n uint16) int {
// amd64:"BSFL","BTSL\\t\\$16"
// arm64:"ORR\t\\$65536","RBITW","CLZW",-"MOVHU\tR",-"RBIT\t",-"CLZ\t"
// s390x:"FLOGR","OR\t\\$65536"
// ppc64:"POPCNTD","OR\\t\\$65536"
// ppc64le:"POPCNTD","OR\\t\\$65536"
Expand All @@ -274,6 +278,7 @@ func TrailingZeros16(n uint16) int {

func TrailingZeros8(n uint8) int {
// amd64:"BSFL","BTSL\\t\\$8"
// arm64:"ORR\t\\$256","RBITW","CLZW",-"MOVBU\tR",-"RBIT\t",-"CLZ\t"
// s390x:"FLOGR","OR\t\\$256"
return bits.TrailingZeros8(n)
}
Expand Down Expand Up @@ -314,6 +319,7 @@ func IterateBits16(n uint16) int {
i := 0
for n != 0 {
// amd64:"BSFL",-"BTSL"
// arm64:"RBITW","CLZW",-"ORR"
i += bits.TrailingZeros16(n)
n &= n - 1
}
Expand All @@ -324,6 +330,7 @@ func IterateBits8(n uint8) int {
i := 0
for n != 0 {
// amd64:"BSFL",-"BTSL"
// arm64:"RBITW","CLZW",-"ORR"
i += bits.TrailingZeros8(n)
n &= n - 1
}
Expand Down

0 comments on commit 4e2b0dd

Please sign in to comment.