math: remove asm version of Dim

Dim performance has regressed by 14% vs 1.9 on amd64.
Current pure go version of Dim is faster and,
what is even more important for performance, is inlinable, so
instead of tweaking asm implementation, just remove it.
I had to update BenchmarkDim, because it was simply reloading
constant(answer) in a loop.
Perf data below:

name   old time/op  new time/op  delta
Dim-6  6.79ns ± 0%  1.60ns ± 1%  -76.39%  (p=0.000 n=7+10)

If I modify benchmark to be the same as in this CL results are even better:

name   old time/op  new time/op  delta
Dim-6  10.2ns ± 0%   1.6ns ± 1%  -84.27%  (p=0.000 n=8+10)

Updates #21913

Change-Id: I00e23c8affc293531e1d9f0e0e49f3a525634f53
Reviewed-on: https://go-review.googlesource.com/80695
Run-TryBot: Ilya Tocar <ilya.tocar@intel.com>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>

src/math/all_test.go

@@ -3210,7 +3210,7 @@ func BenchmarkAbs(b *testing.B) {
 func BenchmarkDim(b *testing.B) {
 	x := 0.0
 	for i := 0; i < b.N; i++ {
-		x = Dim(10, 3)
+		x = Dim(GlobalF, x)
 	}
 	GlobalF = x
 }

src/math/dim.go

@@ -10,9 +10,7 @@ package math
 //	Dim(+Inf, +Inf) = NaN
 //	Dim(-Inf, -Inf) = NaN
 //	Dim(x, NaN) = Dim(NaN, x) = NaN
-func Dim(x, y float64) float64
-
-func dim(x, y float64) float64 {
+func Dim(x, y float64) float64 {
 	// The special cases result in NaN after the subtraction:
 	//      +Inf - +Inf = NaN
 	//      -Inf - -Inf = NaN

src/math/dim_386.s

@@ -4,9 +4,6 @@
 
 #include "textflag.h"
 
-TEXT ·Dim(SB),NOSPLIT,$0
-	JMP ·dim(SB)
-
 TEXT ·Max(SB),NOSPLIT,$0
 	JMP ·max(SB)
 

src/math/dim_amd64.s

@@ -8,44 +8,6 @@
 #define NaN    0x7FF8000000000001
 #define NegInf 0xFFF0000000000000
 
-// func Dim(x, y float64) float64
-TEXT ·Dim(SB),NOSPLIT,$0
-	// (+Inf, +Inf) special case
-	MOVQ    x+0(FP), BX
-	MOVQ    y+8(FP), CX
-	MOVQ    $PosInf, AX
-	CMPQ    AX, BX
-	JNE     dim2
-	CMPQ    AX, CX
-	JEQ     bothInf
-dim2:	// (-Inf, -Inf) special case
-	MOVQ    $NegInf, AX
-	CMPQ    AX, BX
-	JNE     dim3
-	CMPQ    AX, CX
-	JEQ     bothInf
-dim3:	// (NaN, x) or (x, NaN)
-	MOVQ    $~(1<<63), DX
-	MOVQ    $PosInf, AX
-	ANDQ    DX, BX // x = |x|
-	CMPQ    AX, BX
-	JLT     isDimNaN
-	ANDQ    DX, CX // y = |y|
-	CMPQ    AX, CX
-	JLT     isDimNaN
-
-	MOVSD x+0(FP), X0
-	SUBSD y+8(FP), X0
-	MOVSD $(0.0), X1
-	MAXSD X1, X0
-	MOVSD X0, ret+16(FP)
-	RET
-bothInf: // Dim(-Inf, -Inf) or Dim(+Inf, +Inf)
-isDimNaN:
-	MOVQ    $NaN, AX
-	MOVQ    AX, ret+16(FP)
-	RET
-
 // func ·Max(x, y float64) float64
 TEXT ·Max(SB),NOSPLIT,$0
 	// +Inf special cases

src/math/dim_arm.s

@@ -4,9 +4,6 @@
 
 #include "textflag.h"
 
-TEXT ·Dim(SB),NOSPLIT,$0
-	B ·dim(SB)
-
 TEXT ·Min(SB),NOSPLIT,$0
 	B ·min(SB)
 

src/math/dim_arm64.s

@@ -8,35 +8,6 @@
 #define NaN    0x7FF8000000000001
 #define NegInf 0xFFF0000000000000
 
-// func Dim(x, y float64) float64
-TEXT ·Dim(SB),NOSPLIT,$0
-	// (+Inf, +Inf) special case
-	MOVD	$PosInf, R0
-	MOVD	x+0(FP), R1
-	MOVD	y+8(FP), R2
-	CMP	R0, R1
-	BNE	dim2
-	CMP	R0, R2
-	BEQ	bothInf
-dim2:	// (-Inf, -Inf) special case
-	MOVD	$NegInf, R0
-	CMP	R0, R1
-	BNE	dim3
-	CMP	R0, R2
-	BEQ	bothInf
-dim3:	// normal case
-	FMOVD	R1, F0
-	FMOVD	R2, F1
-	FMOVD	$0.0, F2
-	FSUBD	F1, F0
-	FMAXD	F0, F2, F0
-	FMOVD	F0, ret+16(FP)
-	RET
-bothInf:
-	MOVD	$NaN, R0
-	MOVD	R0, ret+16(FP)
-	RET
-
 // func ·Max(x, y float64) float64
 TEXT ·Max(SB),NOSPLIT,$0
 	// +Inf special cases

src/math/stubs_mips64x.s

@@ -27,9 +27,6 @@ TEXT ·Atan(SB),NOSPLIT,$0
 TEXT ·Atanh(SB),NOSPLIT,$0
 	JMP ·atanh(SB)
 
-TEXT ·Dim(SB),NOSPLIT,$0
-	JMP ·dim(SB)
-
 TEXT ·Min(SB),NOSPLIT,$0
 	JMP ·min(SB)
 

src/math/stubs_mipsx.s

@@ -27,9 +27,6 @@ TEXT ·Atan(SB),NOSPLIT,$0
 TEXT ·Atanh(SB),NOSPLIT,$0
 	JMP	·atanh(SB)
 
-TEXT ·Dim(SB),NOSPLIT,$0
-	JMP	·dim(SB)
-
 TEXT ·Min(SB),NOSPLIT,$0
 	JMP	·min(SB)
 

src/math/stubs_ppc64x.s

@@ -27,9 +27,6 @@ TEXT ·Atan(SB),NOSPLIT,$0
 TEXT ·Atanh(SB),NOSPLIT,$0
 	BR ·atanh(SB)
 
-TEXT ·Dim(SB),NOSPLIT,$0
-	BR ·dim(SB)
-
 TEXT ·Min(SB),NOSPLIT,$0
 	BR ·min(SB)
 

src/math/stubs_s390x.s

@@ -4,9 +4,6 @@
 
 #include "textflag.h"
 
-TEXT ·Dim(SB),NOSPLIT,$0
-	BR ·dim(SB)
-
 TEXT ·Exp2(SB),NOSPLIT,$0
 	BR ·exp2(SB)