runtime: remove unused prefetch functions

The only non test user of the assembler prefetch functions is the
heapBits.prefetch function which is itself unused.

The runtime prefetch functions have no functionality on most platforms
and are not inlineable since they are written in assembler. The function
call overhead eliminates the performance gains that could be achieved with
prefetching and would degrade performance for platforms where the functions
are no-ops.

If prefetch functions are needed back again later they can be improved
by avoiding the function call overhead and implementing them as intrinsics.

Change-Id: I52c553cf3607ffe09f0441c6e7a0a818cb21117d
Reviewed-on: https://go-review.googlesource.com/44370
Run-TryBot: Martin Möhrmann <moehrmann@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Austin Clements <austin@google.com>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>

src/runtime/asm_386.s

@@ -1637,19 +1637,6 @@ TEXT runtime·goexit(SB),NOSPLIT,$0-0
 	// traceback from goexit1 must hit code range of goexit
 	BYTE	$0x90	// NOP
 
-// Prefetching doesn't seem to help.
-TEXT runtime·prefetcht0(SB),NOSPLIT,$0-4
-	RET
-
-TEXT runtime·prefetcht1(SB),NOSPLIT,$0-4
-	RET
-
-TEXT runtime·prefetcht2(SB),NOSPLIT,$0-4
-	RET
-
-TEXT runtime·prefetchnta(SB),NOSPLIT,$0-4
-	RET
-
 // Add a module's moduledata to the linked list of moduledata objects. This
 // is called from .init_array by a function generated in the linker and so
 // follows the platform ABI wrt register preservation -- it only touches AX,

src/runtime/asm_amd64.s

@@ -2339,26 +2339,6 @@ TEXT runtime·goexit(SB),NOSPLIT,$0-0
 	// traceback from goexit1 must hit code range of goexit
 	BYTE	$0x90	// NOP
 
-TEXT runtime·prefetcht0(SB),NOSPLIT,$0-8
-	MOVQ	addr+0(FP), AX
-	PREFETCHT0	(AX)
-	RET
-
-TEXT runtime·prefetcht1(SB),NOSPLIT,$0-8
-	MOVQ	addr+0(FP), AX
-	PREFETCHT1	(AX)
-	RET
-
-TEXT runtime·prefetcht2(SB),NOSPLIT,$0-8
-	MOVQ	addr+0(FP), AX
-	PREFETCHT2	(AX)
-	RET
-
-TEXT runtime·prefetchnta(SB),NOSPLIT,$0-8
-	MOVQ	addr+0(FP), AX
-	PREFETCHNTA	(AX)
-	RET
-
 // This is called from .init_array and follows the platform, not Go, ABI.
 TEXT runtime·addmoduledata(SB),NOSPLIT,$0-0
 	PUSHQ	R15 // The access to global variables below implicitly uses R15, which is callee-save

src/runtime/asm_amd64p32.s

@@ -1042,27 +1042,6 @@ TEXT runtime·goexit(SB),NOSPLIT,$0-0
 	// traceback from goexit1 must hit code range of goexit
 	BYTE	$0x90	// NOP
 
-TEXT runtime·prefetcht0(SB),NOSPLIT,$0-4
-	MOVL	addr+0(FP), AX
-	PREFETCHT0	(AX)
-	RET
-
-TEXT runtime·prefetcht1(SB),NOSPLIT,$0-4
-	MOVL	addr+0(FP), AX
-	PREFETCHT1	(AX)
-	RET
-
-
-TEXT runtime·prefetcht2(SB),NOSPLIT,$0-4
-	MOVL	addr+0(FP), AX
-	PREFETCHT2	(AX)
-	RET
-
-TEXT runtime·prefetchnta(SB),NOSPLIT,$0-4
-	MOVL	addr+0(FP), AX
-	PREFETCHNTA	(AX)
-	RET
-
 TEXT ·checkASM(SB),NOSPLIT,$0-1
 	MOVB	$1, ret+0(FP)
 	RET

src/runtime/asm_arm.s

@@ -973,18 +973,6 @@ TEXT runtime·goexit(SB),NOSPLIT,$-4-0
 	// traceback from goexit1 must hit code range of goexit
 	MOVW	R0, R0	// NOP
 
-TEXT runtime·prefetcht0(SB),NOSPLIT,$0-4
-	RET
-
-TEXT runtime·prefetcht1(SB),NOSPLIT,$0-4
-	RET
-
-TEXT runtime·prefetcht2(SB),NOSPLIT,$0-4
-	RET
-
-TEXT runtime·prefetchnta(SB),NOSPLIT,$0-4
-	RET
-
 // x -> x/1000000, x%1000000, called from Go with args, results on stack.
 TEXT runtime·usplit(SB),NOSPLIT,$0-12
 	MOVW	x+0(FP), R0

src/runtime/asm_arm64.s

@@ -931,19 +931,6 @@ TEXT runtime·goexit(SB),NOSPLIT,$-8-0
 	MOVD	R0, R0	// NOP
 	BL	runtime·goexit1(SB)	// does not return
 
-// TODO(aram): use PRFM here.
-TEXT runtime·prefetcht0(SB),NOSPLIT,$0-8
-	RET
-
-TEXT runtime·prefetcht1(SB),NOSPLIT,$0-8
-	RET
-
-TEXT runtime·prefetcht2(SB),NOSPLIT,$0-8
-	RET
-
-TEXT runtime·prefetchnta(SB),NOSPLIT,$0-8
-	RET
-
 TEXT runtime·sigreturn(SB),NOSPLIT,$0-0
 	RET
 

src/runtime/asm_mips64x.s

@@ -823,18 +823,6 @@ TEXT runtime·goexit(SB),NOSPLIT,$-8-0
 	// traceback from goexit1 must hit code range of goexit
 	NOR	R0, R0	// NOP
 
-TEXT runtime·prefetcht0(SB),NOSPLIT,$0-8
-	RET
-
-TEXT runtime·prefetcht1(SB),NOSPLIT,$0-8
-	RET
-
-TEXT runtime·prefetcht2(SB),NOSPLIT,$0-8
-	RET
-
-TEXT runtime·prefetchnta(SB),NOSPLIT,$0-8
-	RET
-
 TEXT ·checkASM(SB),NOSPLIT,$0-1
 	MOVW	$1, R1
 	MOVB	R1, ret+0(FP)

src/runtime/asm_mipsx.s

@@ -903,18 +903,6 @@ TEXT runtime·goexit(SB),NOSPLIT,$-4-0
 	// traceback from goexit1 must hit code range of goexit
 	NOR	R0, R0	// NOP
 
-TEXT runtime·prefetcht0(SB),NOSPLIT,$0-4
-	RET
-
-TEXT runtime·prefetcht1(SB),NOSPLIT,$0-4
-	RET
-
-TEXT runtime·prefetcht2(SB),NOSPLIT,$0-4
-	RET
-
-TEXT runtime·prefetchnta(SB),NOSPLIT,$0-4
-	RET
-
 TEXT ·checkASM(SB),NOSPLIT,$0-1
 	MOVW	$1, R1
 	MOVB	R1, ret+0(FP)

src/runtime/asm_ppc64x.s

@@ -1353,18 +1353,6 @@ TEXT runtime·goexit(SB),NOSPLIT|NOFRAME,$0-0
 	// traceback from goexit1 must hit code range of goexit
 	MOVD	R0, R0	// NOP
 
-TEXT runtime·prefetcht0(SB),NOSPLIT,$0-8
-	RET
-
-TEXT runtime·prefetcht1(SB),NOSPLIT,$0-8
-	RET
-
-TEXT runtime·prefetcht2(SB),NOSPLIT,$0-8
-	RET
-
-TEXT runtime·prefetchnta(SB),NOSPLIT,$0-8
-	RET
-
 TEXT runtime·sigreturn(SB),NOSPLIT,$0-0
 	RET
 

src/runtime/asm_s390x.s

@@ -949,18 +949,6 @@ TEXT runtime·goexit(SB),NOSPLIT|NOFRAME,$0-0
 	// traceback from goexit1 must hit code range of goexit
 	BYTE $0x07; BYTE $0x00; // 2-byte nop
 
-TEXT runtime·prefetcht0(SB),NOSPLIT,$0-8
-	RET
-
-TEXT runtime·prefetcht1(SB),NOSPLIT,$0-8
-	RET
-
-TEXT runtime·prefetcht2(SB),NOSPLIT,$0-8
-	RET
-
-TEXT runtime·prefetchnta(SB),NOSPLIT,$0-8
-	RET
-
 TEXT runtime·sigreturn(SB),NOSPLIT,$0-0
 	RET
 

src/runtime/mbitmap.go

@@ -449,11 +449,6 @@ func heapBitsForObject(p, refBase, refOff uintptr) (base uintptr, hbits heapBits
 	return
 }
 
-// prefetch the bits.
-func (h heapBits) prefetch() {
-	prefetchnta(uintptr(unsafe.Pointer((h.bitp))))
-}
-
 // next returns the heapBits describing the next pointer-sized word in memory.
 // That is, if h describes address p, h.next() describes p+ptrSize.
 // Note that next does not modify h. The caller must record the result.

src/runtime/runtime1.go

@@ -99,10 +99,6 @@ var test_z64, test_x64 uint64
 func testAtomic64() {
 	test_z64 = 42
 	test_x64 = 0
-	prefetcht0(uintptr(unsafe.Pointer(&test_z64)))
-	prefetcht1(uintptr(unsafe.Pointer(&test_z64)))
-	prefetcht2(uintptr(unsafe.Pointer(&test_z64)))
-	prefetchnta(uintptr(unsafe.Pointer(&test_z64)))
 	if atomic.Cas64(&test_z64, test_x64, 1) {
 		throw("cas64 failed")
 	}

src/runtime/stubs.go

@@ -276,11 +276,6 @@ func call1073741824(typ, fn, arg unsafe.Pointer, n, retoffset uint32)
 
 func systemstack_switch()
 
-func prefetcht0(addr uintptr)
-func prefetcht1(addr uintptr)
-func prefetcht2(addr uintptr)
-func prefetchnta(addr uintptr)
-
 // round n up to a multiple of a.  a must be a power of 2.
 func round(n, a uintptr) uintptr {
 	return (n + a - 1) &^ (a - 1)