[444] | 1 | /* |
---|
| 2 | * (c) Copyright 1986 HEWLETT-PACKARD COMPANY |
---|
| 3 | * |
---|
| 4 | * To anyone who acknowledges that this file is provided "AS IS" |
---|
| 5 | * without any express or implied warranty: |
---|
| 6 | * permission to use, copy, modify, and distribute this file |
---|
| 7 | * for any purpose is hereby granted without fee, provided that |
---|
| 8 | * the above copyright notice and this notice appears in all |
---|
| 9 | * copies, and that the name of Hewlett-Packard Company not be |
---|
| 10 | * used in advertising or publicity pertaining to distribution |
---|
| 11 | * of the software without specific, written prior permission. |
---|
| 12 | * Hewlett-Packard Company makes no representations about the |
---|
| 13 | * suitability of this software for any purpose. |
---|
| 14 | */ |
---|
| 15 | |
---|
| 16 | /* HPUX_ID: @(#) $Revision$ */ |
---|
| 17 | /* |
---|
| 18 | * memcpy(s1, s2, n) |
---|
| 19 | * |
---|
| 20 | * Copy n characters from s2 to s1; returns s1. |
---|
| 21 | */ |
---|
| 22 | |
---|
| 23 | #define d_addr arg0 |
---|
| 24 | #define s_addr arg1 |
---|
| 25 | #define count arg2 |
---|
| 26 | #define tmp5 arg3 |
---|
| 27 | #define tmp1 r19 |
---|
| 28 | #define tmp2 r20 |
---|
| 29 | #define tmp3 r21 |
---|
| 30 | #define tmp4 r22 |
---|
| 31 | #define tmp6 r31 |
---|
| 32 | |
---|
| 33 | #include "DEFS.h" |
---|
| 34 | |
---|
| 35 | ENTRY(memcpy) |
---|
| 36 | comib,>= 5,count,byteloop /* If count is <= 6 don't get fancy.*/ |
---|
| 37 | movb,=,n d_addr,ret0,done /* The return value is defined to be the value of d_addr. DELAY SLOT */ |
---|
| 38 | /* if d_addr is null then exit */ |
---|
| 39 | extru s_addr,31,2,tmp1 /* Extract the low two bits of the source address. */ |
---|
| 40 | extru d_addr,31,2,tmp2 /* Extract the low two bits of the destination address. */ |
---|
| 41 | add count,tmp2,count /* pre increment the count to adjust for alignment of s1 */ |
---|
| 42 | comb,<> tmp2,tmp1,not_aligned /* see if s1 is aligned w.r.t. s2. */ |
---|
| 43 | dep 0,31,2,s_addr /* Compute the word address of the source. DELAY SLOT. */ |
---|
| 44 | |
---|
| 45 | /* aligned */ |
---|
| 46 | |
---|
| 47 | /* We will now begin the 16 byte at a time word move if count >= 16 ! */ |
---|
| 48 | /* Else we will branch to the 4 byte-at-a time word move ! */ |
---|
| 49 | |
---|
| 50 | addibt,<,n -16,count,chekchunk /* If count < 16 then we can't move 16 byte chunks ! */ |
---|
| 51 | /* actually we can legally move 13 or more bytes on the first loop. */ |
---|
| 52 | /* These loads and stores are done so as to prevent processor interlock. */ |
---|
| 53 | chunks: |
---|
| 54 | ldwm 16(0,s_addr),tmp1 /* tmp1 = *s_addr s_addr += 16 */ |
---|
| 55 | ldw -12(0,s_addr),tmp2 /* tmp2 = 2nd word */ |
---|
| 56 | ldw -8(0,s_addr),tmp3 /* tmp3 = 3rd word */ |
---|
| 57 | ldw -4(0,s_addr),tmp4 /* tmp4 = 4th word */ |
---|
| 58 | /* Now store the results ! */ |
---|
| 59 | stbys,b,m tmp1,4(0,d_addr) /* tmp1 = 1st word stored d_addr += 16 also take care of front porch. */ |
---|
| 60 | stwm tmp2,4(0,d_addr) /* tmp2 = 2nd word stored. */ |
---|
| 61 | stwm tmp3,4(0,d_addr) /* tmp3 = 3rd word stored. */ |
---|
| 62 | addibf,< -16,count,chunks /* If count is still >= 16 do another loop. */ |
---|
| 63 | stwm tmp4,4(0,d_addr) /* tmp4 = 4th word stored. DELAY SLOT */ |
---|
| 64 | |
---|
| 65 | chekchunk: |
---|
| 66 | addibt,<,n 12,count,back_porch /* since the count is already decremented by -16 we're testing */ |
---|
| 67 | /* to see if there are at least 4 bytes left ? */ |
---|
| 68 | subchunk: |
---|
| 69 | ldws,ma 4(s_addr),tmp1 /* tmp1 = *s_addr++ */ |
---|
| 70 | addibf,< -4,count,subchunk /* count -= 4 */ |
---|
| 71 | stbys,b,m tmp1,4(d_addr) /* *d_addr++ = tmp1 */ |
---|
| 72 | |
---|
| 73 | |
---|
| 74 | back_porch: |
---|
| 75 | addibt,=,n 4,count,done /* if count = 0 we're, of course, done ! */ |
---|
| 76 | ldws 0(s_addr),tmp1 /* load up the back_porch */ |
---|
| 77 | add d_addr,count,d_addr/* final store address is +1 too high ! */ |
---|
| 78 | bv 0(r2) /* return--were done. */ |
---|
| 79 | stbys,e tmp1,0(d_addr) /* kerplunk! whew ! */ |
---|
| 80 | |
---|
| 81 | /* Begin non_aligned code. (no refrence to politics) */ |
---|
| 82 | not_aligned: |
---|
| 83 | sub,>= tmp2,tmp1,tmp3 /* compute the shift quantity again and skip the load if tmp2 > tmp1. */ |
---|
| 84 | ldwm 4(0,s_addr),tmp1 /* load up the first word from the source. tmp1 = *s_addr++ */ |
---|
| 85 | zdep tmp3,28,29,tmp4 /* compute the number of bits to shift based on the number of bytes above. */ |
---|
| 86 | mtctl tmp4,11 /* load the shift count into cr11 = shift count register. */ |
---|
| 87 | |
---|
| 88 | addibt,<,n -16,count,chkchnk2 /* first step in pre adjustment of count for looping. */ |
---|
| 89 | |
---|
| 90 | chunk2: |
---|
| 91 | ldwm 16(0,s_addr),tmp2 /* get either first or second word . tmp2 = *s_addr++ */ |
---|
| 92 | ldw -12(s_addr),tmp3 |
---|
| 93 | ldw -8(s_addr),tmp4 |
---|
| 94 | ldw -4(s_addr),tmp5 |
---|
| 95 | vshd tmp1,tmp2,tmp6 /* position data ! */ |
---|
| 96 | stbys,b,m tmp6,4(0,d_addr) /* store ! */ |
---|
| 97 | |
---|
| 98 | vshd tmp2,tmp3,tmp6 /* position data ! */ |
---|
| 99 | stwm tmp6,4(0,d_addr) /* store ! */ |
---|
| 100 | |
---|
| 101 | vshd tmp3,tmp4,tmp6 /* position data ! */ |
---|
| 102 | stwm tmp6,4(0,d_addr) /* store ! */ |
---|
| 103 | |
---|
| 104 | vshd tmp4,tmp5,tmp6 /* position data ! */ |
---|
| 105 | stwm tmp6,4(0,d_addr) /* store the data ! */ |
---|
| 106 | addibf,< -16,count,chunk2 /* If count is still >= 16 do another loop. */ |
---|
| 107 | copy tmp5,tmp1 |
---|
| 108 | |
---|
| 109 | |
---|
| 110 | chkchnk2: |
---|
| 111 | addibt,<,n 12,count,bp_0 /* if we don't have 4 bytes left then do the back porch (bp_0) */ |
---|
| 112 | |
---|
| 113 | subchnk2: |
---|
| 114 | ldwm 4(0,s_addr),tmp2 /* get next word ! */ |
---|
| 115 | vshd tmp1,tmp2,tmp3 /* position data ! */ |
---|
| 116 | addibt,< -4,count,bp_1 /* decrement count and when count < 4 goto back_porch (bp_1) */ |
---|
| 117 | stbys,b,m tmp3,4(0,d_addr) /* store ! */ |
---|
| 118 | |
---|
| 119 | ldwm 4(0,s_addr),tmp1 /* get 4th word ! */ |
---|
| 120 | vshd tmp2,tmp1,tmp3 /* position data ! */ |
---|
| 121 | addib,>= -4,count,subchnk2 /* decrement count and when count <= 4 go to back porch (bp_2) */ |
---|
| 122 | stbys,b,m tmp3,4(0,d_addr) /* store the data ! */ |
---|
| 123 | |
---|
| 124 | bp_0: copy tmp1,tmp2 /* switch registers used in the shift process. */ |
---|
| 125 | bp_1: addibt,<=,n 4,count,done /* if count = -4 this implies that count = 0 -> done */ |
---|
| 126 | add d_addr,count,d_addr /* bump destination address to be +1 too high ! */ |
---|
| 127 | mfctl sar,tmp3 /* suppress final ldwm unless result used */ |
---|
| 128 | extru tmp3,28,2,tmp3 /* convert bitshift to byteshift */ |
---|
| 129 | sub,<= count,tmp3,r0 /* bytes unused if (count-byteshift <= 0*/ |
---|
| 130 | |
---|
| 131 | ldwm 4(0,s_addr),tmp1 /* get final word ! */ |
---|
| 132 | vshd tmp2,tmp1,tmp3 /* position data ! */ |
---|
| 133 | bv 0(r2) /* return */ |
---|
| 134 | stbys,e tmp3,0(0,d_addr) /* store the data ! */ |
---|
| 135 | |
---|
| 136 | /* here we do ye old byte-at-a-time moves. */ |
---|
| 137 | byteloop: |
---|
| 138 | comb,>=,n 0,count,done |
---|
| 139 | |
---|
| 140 | encore: |
---|
| 141 | ldbs,ma 1(s_addr),tmp1 |
---|
| 142 | addibf,= -1,count,encore |
---|
| 143 | stbs,ma tmp1,1(d_addr) |
---|
| 144 | |
---|
| 145 | done: |
---|
| 146 | EXIT(memcpy) |
---|