internal/bytealg: add assembly implementation of Count/CountString for mips64x

Add a simple assembly implementation of Count/CountString for mips64x.

name            old sec/op    new sec/op     vs base
CountSingle/10-4   31.16n ±  0%    41.69n ±  0%  +33.79% (p=0.000 n=11)
CountSingle/32-4   69.58n ±  0%    59.61n ±  0%  -14.33% (p=0.000 n=11)
CountSingle/4K-4   7.428µ ±  0%    5.153µ ±  0%  -30.63% (p=0.000 n=11)
CountSingle/4M-4   7.634m ±  0%    5.300m ±  0%  -30.58% (p=0.000 n=11)
CountSingle/64M-4   134.4m ±  0%    100.8m ±  3%  -24.99% (p=0.000 n=11)

name             old B/s      new B/s       vs base
CountSingle/10-4   306.1Mi ±  0%    228.8Mi ±  0%  -25.25% (p=0.000 n=11)
CountSingle/32-4   438.6Mi ±  0%    512.0Mi ±  0%  +16.74% (p=0.000 n=11)
CountSingle/4K-4   525.9Mi ±  0%    758.0Mi ±  0%  +44.15% (p=0.000 n=11)
CountSingle/4M-4   523.9Mi ±  0%    754.7Mi ±  0%  +44.05% (p=0.000 n=11)
CountSingle/64M-4  476.3Mi ±  0%    635.0Mi ±  0%  +33.31% (p=0.000 n=11)

Change-Id: Id5ddbea0d080e2903156ef8dc86c030a8179115b
Reviewed-on: https://go-review.googlesource.com/c/go/+/650995
Reviewed-by: Keith Randall <[email protected]>
LUCI-TryBot-Result: Go LUCI <[email protected]>
Reviewed-by: Michael Knyszek <[email protected]>
Reviewed-by: Keith Randall <[email protected]>
Auto-Submit: Keith Randall <[email protected]>

Author: HeliC829

src/internal/bytealg/count_generic.go

@@ -2,7 +2,7 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
-//go:build !amd64 && !arm && !arm64 && !ppc64le && !ppc64 && !riscv64 && !s390x
+//go:build !amd64 && !arm && !arm64 && !mips64le && !mips64 && !ppc64le && !ppc64 && !riscv64 && !s390x
 
 package bytealg
 

src/internal/bytealg/count_mips64x.s

@@ -0,0 +1,52 @@
+// Copyright 2025 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+//go:build mips64 || mips64le
+
+#include "go_asm.h"
+#include "textflag.h"
+
+TEXT ·Count(SB),NOSPLIT,$0-40
+	// R1 = b_base
+	// R2 = b_len
+	// R3 = byte to count
+	MOVV	b_base+0(FP), R1
+	MOVV	b_len+8(FP), R2
+	MOVBU	c+24(FP), R3
+	MOVV	R0, R5	// count
+	ADDV	R1, R2	// end
+
+loop:
+	BEQ	R1, R2, done
+	MOVBU	(R1), R6
+	ADDV	$1, R1
+	BNE	R3, R6, loop
+	ADDV	$1, R5
+	JMP	loop
+
+done:
+	MOVV	R5, ret+32(FP)
+	RET
+
+TEXT ·CountString(SB),NOSPLIT,$0-32
+	// R1 = s_base
+	// R2 = s_len
+	// R3 = byte to count
+	MOVV	s_base+0(FP), R1
+	MOVV	s_len+8(FP), R2
+	MOVBU	c+16(FP), R3
+	MOVV	R0, R5	// count
+	ADDV	R1, R2	// end
+
+loop:
+	BEQ	R1, R2, done
+	MOVBU	(R1), R6
+	ADDV	$1, R1
+	BNE	R3, R6, loop
+	ADDV	$1, R5
+	JMP	loop
+
+done:
+	MOVV	R5, ret+24(FP)
+	RET

src/internal/bytealg/count_native.go

@@ -2,7 +2,7 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
-//go:build amd64 || arm || arm64 || ppc64le || ppc64 || riscv64 || s390x
+//go:build amd64 || arm || arm64 || mips64le || mips64 || ppc64le || ppc64 || riscv64 || s390x
 
 package bytealg