cmd/compile: eliminate unnecessary type conversions in TrailingZeros(…

…16|8) for arm This follows CL 156999 which did the same for arm64. name old time/op new time/op delta TrailingZeros-4 7.30ns ± 1% 7.30ns ± 0% ~ (p=0.413 n=9+9) TrailingZeros8-4 8.32ns ± 0% 7.17ns ± 0% -13.77% (p=0.000 n=10+9) TrailingZeros16-4 8.30ns ± 0% 7.18ns ± 0% -13.50% (p=0.000 n=9+10) TrailingZeros32-4 6.46ns ± 1% 6.47ns ± 1% ~ (p=0.325 n=10+10) TrailingZeros64-4 16.3ns ± 0% 16.2ns ± 0% -0.61% (p=0.000 n=7+10) Change-Id: I7e9e1abf7e30d811aa474d272b2824ec7cbbaa98 Reviewed-on: https://go-review.googlesource.com/c/go/+/167797 Run-TryBot: Tobias Klauser <[email protected]> TryBot-Result: Gobot Gobot <[email protected]> Reviewed-by: Cherry Zhang <[email protected]>
nsajko · Mar 15, 2019 · 156c830 · 156c830
1 parent d9db9e3
commit 156c830
Show file tree

Hide file tree

Showing 4 changed files with 158 additions and 5 deletions.
diff --git a/src/cmd/compile/internal/gc/ssa.go b/src/cmd/compile/internal/gc/ssa.go
@@ -3288,12 +3288,12 @@ func init() {
 			y := s.newValue2(ssa.OpOr32, types.Types[TUINT32], x, c)
 			return s.newValue1(ssa.OpCtz32, types.Types[TINT], y)
 		},
-		sys.ARM, sys.MIPS)
+		sys.MIPS)
 	addF("math/bits", "TrailingZeros16",
 		func(s *state, n *Node, args []*ssa.Value) *ssa.Value {
 			return s.newValue1(ssa.OpCtz16, types.Types[TINT], args[0])
 		},
-		sys.AMD64, sys.ARM64, sys.Wasm)
+		sys.AMD64, sys.ARM, sys.ARM64, sys.Wasm)
 	addF("math/bits", "TrailingZeros16",
 		func(s *state, n *Node, args []*ssa.Value) *ssa.Value {
 			x := s.newValue1(ssa.OpZeroExt16to64, types.Types[TUINT64], args[0])
@@ -3309,12 +3309,12 @@ func init() {
 			y := s.newValue2(ssa.OpOr32, types.Types[TUINT32], x, c)
 			return s.newValue1(ssa.OpCtz32, types.Types[TINT], y)
 		},
-		sys.ARM, sys.MIPS)
+		sys.MIPS)
 	addF("math/bits", "TrailingZeros8",
 		func(s *state, n *Node, args []*ssa.Value) *ssa.Value {
 			return s.newValue1(ssa.OpCtz8, types.Types[TINT], args[0])
 		},
-		sys.AMD64, sys.ARM64, sys.Wasm)
+		sys.AMD64, sys.ARM, sys.ARM64, sys.Wasm)
 	addF("math/bits", "TrailingZeros8",
 		func(s *state, n *Node, args []*ssa.Value) *ssa.Value {
 			x := s.newValue1(ssa.OpZeroExt8to64, types.Types[TUINT64], args[0])

diff --git a/src/cmd/compile/internal/ssa/gen/ARM.rules b/src/cmd/compile/internal/ssa/gen/ARM.rules
@@ -59,13 +59,22 @@
 
 // TODO: optimize this for ARMv5 and ARMv6
 (Ctz32NonZero x) -> (Ctz32 x)
+(Ctz16NonZero x) -> (Ctz32 x)
+(Ctz8NonZero x) -> (Ctz32 x)
 
 // count trailing zero for ARMv5 and ARMv6
 // 32 - CLZ(x&-x - 1)
-(Ctz32 <t> x) && objabi.GOARM<=6 -> (RSBconst [32] (CLZ <t> (SUBconst <t> (AND <t> x (RSBconst <t> [0] x)) [1])))
+(Ctz32 <t> x) && objabi.GOARM<=6 ->
+	(RSBconst [32] (CLZ <t> (SUBconst <t> (AND <t> x (RSBconst <t> [0] x)) [1])))
+(Ctz16 <t> x) && objabi.GOARM<=6 ->
+	(RSBconst [32] (CLZ <t> (SUBconst <typ.UInt32> (AND <typ.UInt32> (ORconst <typ.UInt32> [0x10000] x) (RSBconst <typ.UInt32> [0] (ORconst <typ.UInt32> [0x10000] x))) [1])))
+(Ctz8 <t> x) && objabi.GOARM<=6 ->
+	(RSBconst [32] (CLZ <t> (SUBconst <typ.UInt32> (AND <typ.UInt32> (ORconst <typ.UInt32> [0x100] x) (RSBconst <typ.UInt32> [0] (ORconst <typ.UInt32> [0x100] x))) [1])))
 
 // count trailing zero for ARMv7
 (Ctz32 <t> x) && objabi.GOARM==7 -> (CLZ <t> (RBIT <t> x))
+(Ctz16 <t> x) && objabi.GOARM==7 -> (CLZ <t> (RBIT <typ.UInt32> (ORconst <typ.UInt32> [0x10000] x)))
+(Ctz8 <t> x) && objabi.GOARM==7 -> (CLZ <t> (RBIT <typ.UInt32> (ORconst <typ.UInt32> [0x100] x)))
 
 // bit length
 (BitLen32 <t> x) -> (RSBconst [32] (CLZ <t> x))

diff --git a/src/cmd/compile/internal/ssa/rewriteARM.go b/src/cmd/compile/internal/ssa/rewriteARM.go
diff --git a/test/codegen/mathbits.go b/test/codegen/mathbits.go
@@ -258,6 +258,7 @@ func RotateLeftVariable32(n uint32, m int) uint32 {
 
 func TrailingZeros(n uint) int {
 	// amd64:"BSFQ","MOVL\t\\$64","CMOVQEQ"
+	// arm:"CLZ"
 	// arm64:"RBIT","CLZ"
 	// s390x:"FLOGR"
 	// ppc64:"ANDN","POPCNTD"
@@ -278,6 +279,7 @@ func TrailingZeros64(n uint64) int {
 
 func TrailingZeros32(n uint32) int {
 	// amd64:"BTSQ\\t\\$32","BSFQ"
+	// arm:"CLZ"
 	// arm64:"RBITW","CLZW"
 	// s390x:"FLOGR","MOVWZ"
 	// ppc64:"ANDN","POPCNTW"
@@ -288,6 +290,7 @@ func TrailingZeros32(n uint32) int {
 
 func TrailingZeros16(n uint16) int {
 	// amd64:"BSFL","BTSL\\t\\$16"
+	// arm:"ORR\t\\$65536","CLZ",-"MOVHU\tR"
 	// arm64:"ORR\t\\$65536","RBITW","CLZW",-"MOVHU\tR",-"RBIT\t",-"CLZ\t"
 	// s390x:"FLOGR","OR\t\\$65536"
 	// ppc64:"POPCNTD","OR\\t\\$65536"
@@ -298,6 +301,7 @@ func TrailingZeros16(n uint16) int {
 
 func TrailingZeros8(n uint8) int {
 	// amd64:"BSFL","BTSL\\t\\$8"
+	// arm:"ORR\t\\$256","CLZ",-"MOVBU\tR"
 	// arm64:"ORR\t\\$256","RBITW","CLZW",-"MOVBU\tR",-"RBIT\t",-"CLZ\t"
 	// s390x:"FLOGR","OR\t\\$256"
 	// wasm:"I64Ctz"