swab.S revision 1.8
1 1.1 cgd /* 2 1.8 jtc * Written by J.T. Conklin <jtc (at) netbsd.org>. 3 1.8 jtc * Public domain. 4 1.1 cgd */ 5 1.1 cgd 6 1.7 jtc #include <machine/asm.h> 7 1.7 jtc 8 1.2 jtc #if defined(LIBC_SCCS) 9 1.8 jtc RCSID("$Id: swab.S,v 1.8 1995/04/28 22:54:38 jtc Exp $") 10 1.1 cgd #endif 11 1.1 cgd 12 1.1 cgd /* 13 1.1 cgd * On the i486, this code is negligibly faster than the code generated 14 1.3 jtc * by gcc at about half the size. If my i386 databook is correct, it 15 1.1 cgd * should be considerably faster than the gcc code on a i386. 16 1.1 cgd */ 17 1.1 cgd 18 1.1 cgd ENTRY(swab) 19 1.1 cgd pushl %esi 20 1.1 cgd pushl %edi 21 1.1 cgd movl 12(%esp),%esi 22 1.1 cgd movl 16(%esp),%edi 23 1.1 cgd movl 20(%esp),%ecx 24 1.1 cgd 25 1.1 cgd cld # set direction forward 26 1.1 cgd 27 1.4 jtc shrl $1,%ecx 28 1.1 cgd testl $7,%ecx # copy first group of 1 to 7 words 29 1.4 jtc jz L2 # while swaping alternate bytes. 30 1.1 cgd .align 2,0x90 31 1.1 cgd L1: lodsw 32 1.6 jtc rorw $8,%ax 33 1.1 cgd stosw 34 1.1 cgd decl %ecx 35 1.1 cgd testl $7,%ecx 36 1.4 jtc jnz L1 37 1.3 jtc 38 1.1 cgd L2: shrl $3,%ecx # copy remainder 8 words at a time 39 1.1 cgd jz L4 # while swapping alternate bytes. 40 1.1 cgd .align 2,0x90 41 1.1 cgd L3: lodsw 42 1.6 jtc rorw $8,%ax 43 1.1 cgd stosw 44 1.1 cgd lodsw 45 1.6 jtc rorw $8,%ax 46 1.1 cgd stosw 47 1.1 cgd lodsw 48 1.6 jtc rorw $8,%ax 49 1.1 cgd stosw 50 1.1 cgd lodsw 51 1.6 jtc rorw $8,%ax 52 1.1 cgd stosw 53 1.1 cgd lodsw 54 1.6 jtc rorw $8,%ax 55 1.1 cgd stosw 56 1.1 cgd lodsw 57 1.6 jtc rorw $8,%ax 58 1.1 cgd stosw 59 1.1 cgd lodsw 60 1.6 jtc rorw $8,%ax 61 1.1 cgd stosw 62 1.1 cgd lodsw 63 1.6 jtc rorw $8,%ax 64 1.1 cgd stosw 65 1.1 cgd decl %ecx 66 1.1 cgd jnz L3 67 1.1 cgd 68 1.6 jtc L4: popl %edi 69 1.1 cgd popl %esi 70 1.1 cgd ret 71
Indexes created Mon Sep 22 13:09:51 GMT 2025