u3memcpy.s

来自「优龙2410linux2.6.8内核源代码」· S 代码 · 共 315 行

S
315 行
字号 
/* U3memcpy.S: UltraSparc-III optimized memcpy. * * Copyright (C) 1999, 2000, 2004 David S. Miller (davem@redhat.com) */#ifdef __KERNEL__#include <asm/visasm.h>#include <asm/asi.h>#include <asm/dcu.h>#include <asm/spitfire.h>#else#define ASI_BLK_P 0xf0#define FPRS_FEF  0x04#define VISEntryHalf rd %fprs, %o5; wr %g0, FPRS_FEF, %fprs#define VISExitHalf and %o5, FPRS_FEF, %o5; wr %o5, 0x0, %fprs#endif#ifndef XCC#define XCC xcc#endif	.register	%g2,#scratch	.register	%g3,#scratch	/* Special/non-trivial issues of this code:	 *	 * 1) %o5 is preserved from VISEntryHalf to VISExitHalf	 * 2) Only low 32 FPU registers are used so that only the	 *    lower half of the FPU register set is dirtied by this	 *    code.  This is especially important in the kernel.	 * 3) This code never prefetches cachelines past the end	 *    of the source buffer.	 */	.text	.align	32	/* The cheetah's flexible spine, oversized liver, enlarged heart,	 * slender muscular body, and claws make it the swiftest hunter	 * in Africa and the fastest animal on land.  Can reach speeds	 * of up to 2.4GB per second.	 */	.globl	U3memcpyU3memcpy:	/* %o0=dst, %o1=src, %o2=len */	mov		%o0, %g5	cmp		%o2, 0	be,pn		%XCC, 85f	 or		%o0, %o1, %o3	cmp		%o2, 16	bleu,a,pn	%XCC, 70f	 or		%o3, %o2, %o3	cmp		%o2, 256	blu,pt		%XCC, 80f	 andcc		%o3, 0x7, %g0	ba,pt		%xcc, 1f	 andcc		%o0, 0x3f, %g2	/* Here len >= 256 and condition codes reflect execution	 * of "andcc %o0, 0x7, %g2", done by caller.	 */	.align		641:	/* Is 'dst' already aligned on an 64-byte boundary? */	be,pt		%XCC, 2f	/* Compute abs((dst & 0x3f) - 0x40) into %g2.  This is the number	 * of bytes to copy to make 'dst' 64-byte aligned.  We pre-	 * subtract this from 'len'.	 */	 sub		%g2, 0x40, %g2	sub		%g0, %g2, %g2	sub		%o2, %g2, %o2	/* Copy %g2 bytes from src to dst, one byte at a time. */1:	ldub		[%o1 + 0x00], %o3	add		%o1, 0x1, %o1	add		%o0, 0x1, %o0	subcc		%g2, 0x1, %g2	bg,pt		%XCC, 1b	 stb		%o3, [%o0 + -1]2:	VISEntryHalf	and		%o1, 0x7, %g1	ba,pt		%xcc, 1f	 alignaddr	%o1, %g0, %o1	.align		641:	membar		#StoreLoad | #StoreStore | #LoadStore	prefetch	[%o1 + 0x000], #one_read	prefetch	[%o1 + 0x040], #one_read	andn		%o2, (0x40 - 1), %o4	prefetch	[%o1 + 0x080], #one_read	prefetch	[%o1 + 0x0c0], #one_read	ldd		[%o1 + 0x000], %f0	prefetch	[%o1 + 0x100], #one_read	ldd		[%o1 + 0x008], %f2	prefetch	[%o1 + 0x140], #one_read	ldd		[%o1 + 0x010], %f4	prefetch	[%o1 + 0x180], #one_read	faligndata	%f0, %f2, %f16	ldd		[%o1 + 0x018], %f6	faligndata	%f2, %f4, %f18	ldd		[%o1 + 0x020], %f8	faligndata	%f4, %f6, %f20	ldd		[%o1 + 0x028], %f10	faligndata	%f6, %f8, %f22	ldd		[%o1 + 0x030], %f12	faligndata	%f8, %f10, %f24	ldd		[%o1 + 0x038], %f14	faligndata	%f10, %f12, %f26	ldd		[%o1 + 0x040], %f0	sub		%o4, 0x80, %o4	add		%o1, 0x40, %o1	ba,pt		%xcc, 1f	 srl		%o4, 6, %o3	.align		641:	ldd		[%o1 + 0x008], %f2	faligndata	%f12, %f14, %f28	ldd		[%o1 + 0x010], %f4	faligndata	%f14, %f0, %f30	stda		%f16, [%o0] ASI_BLK_P	ldd		[%o1 + 0x018], %f6	faligndata	%f0, %f2, %f16	ldd		[%o1 + 0x020], %f8	faligndata	%f2, %f4, %f18	ldd		[%o1 + 0x028], %f10	faligndata	%f4, %f6, %f20	ldd		[%o1 + 0x030], %f12	faligndata	%f6, %f8, %f22	ldd		[%o1 + 0x038], %f14	faligndata	%f8, %f10, %f24	ldd		[%o1 + 0x040], %f0	prefetch	[%o1 + 0x180], #one_read	faligndata	%f10, %f12, %f26	subcc		%o3, 0x01, %o3	add		%o1, 0x40, %o1	bg,pt		%XCC, 1b	 add		%o0, 0x40, %o0	/* Finally we copy the last full 64-byte block. */	ldd		[%o1 + 0x008], %f2	faligndata	%f12, %f14, %f28	ldd		[%o1 + 0x010], %f4	faligndata	%f14, %f0, %f30	stda		%f16, [%o0] ASI_BLK_P	ldd		[%o1 + 0x018], %f6	faligndata	%f0, %f2, %f16	ldd		[%o1 + 0x020], %f8	faligndata	%f2, %f4, %f18	ldd		[%o1 + 0x028], %f10	faligndata	%f4, %f6, %f20	ldd		[%o1 + 0x030], %f12	faligndata	%f6, %f8, %f22	ldd		[%o1 + 0x038], %f14	faligndata	%f8, %f10, %f24	cmp		%g1, 0	be,pt		%XCC, 1f	 add		%o0, 0x40, %o0	ldd		[%o1 + 0x040], %f01:	faligndata	%f10, %f12, %f26	faligndata	%f12, %f14, %f28	faligndata	%f14, %f0, %f30	stda		%f16, [%o0] ASI_BLK_P	add		%o0, 0x40, %o0	add		%o1, 0x40, %o1	membar		#Sync	/* Now we copy the (len modulo 64) bytes at the end.	 * Note how we borrow the %f0 loaded above.	 *	 * Also notice how this code is careful not to perform a	 * load past the end of the src buffer.	 */	and		%o2, 0x3f, %o2	andcc		%o2, 0x38, %g2	be,pn		%XCC, 2f	 subcc		%g2, 0x8, %g2	be,pn		%XCC, 2f	 cmp		%g1, 0	be,a,pt		%XCC, 1f	 ldd		[%o1 + 0x00], %f01:	ldd		[%o1 + 0x08], %f2	add		%o1, 0x8, %o1	sub		%o2, 0x8, %o2	subcc		%g2, 0x8, %g2	faligndata	%f0, %f2, %f8	std		%f8, [%o0 + 0x00]	be,pn		%XCC, 2f	 add		%o0, 0x8, %o0	ldd		[%o1 + 0x08], %f0	add		%o1, 0x8, %o1	sub		%o2, 0x8, %o2	subcc		%g2, 0x8, %g2	faligndata	%f2, %f0, %f8	std		%f8, [%o0 + 0x00]	bne,pn		%XCC, 1b	 add		%o0, 0x8, %o0	/* If anything is left, we copy it one byte at a time.	 * Note that %g1 is (src & 0x3) saved above before the	 * alignaddr was performed.	 */2:	cmp		%o2, 0	add		%o1, %g1, %o1	VISExitHalf	be,pn		%XCC, 85f	 sub		%o0, %o1, %o3	andcc		%g1, 0x7, %g0	bne,pn		%icc, 90f	 andcc		%o2, 0x8, %g0	be,pt		%icc, 1f	 nop	ldx		[%o1], %o5	stx		%o5, [%o1 + %o3]	add		%o1, 0x8, %o11:	andcc		%o2, 0x4, %g0	be,pt		%icc, 1f	 nop	lduw		[%o1], %o5	stw		%o5, [%o1 + %o3]	add		%o1, 0x4, %o11:	andcc		%o2, 0x2, %g0	be,pt		%icc, 1f	 nop	lduh		[%o1], %o5	sth		%o5, [%o1 + %o3]	add		%o1, 0x2, %o11:	andcc		%o2, 0x1, %g0	be,pt		%icc, 85f	 nop	ldub		[%o1], %o5	ba,pt		%xcc, 85f	 stb		%o5, [%o1 + %o3]70: /* 16 < len <= 64 */	bne,pn		%XCC, 90f	 sub		%o0, %o1, %o3	andn		%o2, 0x7, %o4	and		%o2, 0x7, %o21:	subcc		%o4, 0x8, %o4	ldx		[%o1], %o5	stx		%o5, [%o1 + %o3]	bgu,pt		%XCC, 1b	 add		%o1, 0x8, %o1	andcc		%o2, 0x4, %g0	be,pt		%XCC, 1f	 nop	sub		%o2, 0x4, %o2	lduw		[%o1], %o5	stw		%o5, [%o1 + %o3]	add		%o1, 0x4, %o11:	cmp		%o2, 0	be,pt		%XCC, 85f	 nop	ba,pt		%xcc, 90f	 nop80: /* 0 < len <= 16 */	andcc		%o3, 0x3, %g0	bne,pn		%XCC, 90f	 sub		%o0, %o1, %o31:	subcc		%o2, 4, %o2	lduw		[%o1], %g1	stw		%g1, [%o1 + %o3]	bgu,pt		%XCC, 1b	 add		%o1, 4, %o185:	retl	 mov		%g5, %o0	.align	3290:	subcc		%o2, 1, %o2	ldub		[%o1], %g1	stb		%g1, [%o1 + %o3]	bgu,pt		%XCC, 90b	 add		%o1, 1, %o1	retl	 mov		%g5, %o0	/* Act like copy_{to,in}_user(), ie. return zero instead	 * of original destination pointer.  This is invoked when	 * copy_{to,in}_user() finds that %asi is kernel space.	 */	.globl	U3memcpy_user_stubU3memcpy_user_stub:	save		%sp, -192, %sp	mov		%i0, %o0	mov		%i1, %o1	call		U3memcpy	 mov		%i2, %o2	ret	 restore	%g0, %g0, %o0

u3memcpy.s - 源码说明

本页面展示了「优龙2410linux2.6.8内核源代码」中的 u3memcpy.s 源码文件，采用 S 编程语言编写，共 315 行代码。您可以在线阅读完整代码内容，也可以返回资源详情页下载完整源码包进行本地学习和开发。

虫虫下载站收录了大量与linux相关的技术资源，包括源代码、技术文档、电路图等，是电子工程师和嵌入式开发者的专业学习平台。

⌨️ 快捷键说明

复制代码Ctrl + C
搜索代码Ctrl + F
全屏模式F11
增大字号Ctrl + =
减小字号Ctrl + -
显示快捷键?