Details | Last modification | View Log | RSS feed
| Rev | Author | Line No. | Line |
|---|---|---|---|
| 4973 | right-hear | 1 | /* Copyright (C) 1995 DJ Delorie, see COPYING.DJ for details */ |
| 2 | #include |
||
| 3 | .file "memset.s" |
||
| 4 | .text |
||
| 5 | .align 4 |
||
| 6 | MK_C_SYM(memset) |
||
| 7 | pushl %ebp |
||
| 8 | movl %esp,%ebp |
||
| 9 | pushl %edi |
||
| 10 | movl 8(%ebp),%edi |
||
| 11 | movl 12(%ebp),%eax |
||
| 12 | movl 16(%ebp),%ecx |
||
| 13 | cld |
||
| 14 | |||
| 15 | # We will handle memsets of <= 15 bytes one byte at a time. |
||
| 16 | # This avoids some extra overhead for small memsets, and |
||
| 17 | # knowing we are setting > 15 bytes eliminates some annoying |
||
| 18 | # checks in the "long move" case. |
||
| 19 | cmpl $15,%ecx |
||
| 20 | jle L3 |
||
| 21 | |||
| 22 | # Otherwise, tile the byte value out into %eax. |
||
| 23 | # 0x41 -> 0x41414141, etc. |
||
| 24 | movb %al,%ah |
||
| 25 | movl %eax,%edx |
||
| 26 | sall $16,%eax |
||
| 27 | movw %dx,%ax |
||
| 28 | jmp L2 |
||
| 29 | |||
| 30 | # Handle any cruft necessary to get %edi long-aligned. |
||
| 31 | L1: stosb |
||
| 32 | decl %ecx |
||
| 33 | L2: testl $3,%edi |
||
| 34 | jnz L1 |
||
| 35 | |||
| 36 | # Now slam out all of the longs. |
||
| 37 | movl %ecx,%edx |
||
| 38 | shrl $2,%ecx |
||
| 39 | rep |
||
| 40 | stosl |
||
| 41 | |||
| 42 | # Finally, handle any trailing cruft. We know the high three bytes |
||
| 43 | # of %ecx must be zero, so just put the "slop count" in the low byte. |
||
| 44 | movb %dl,%cl |
||
| 45 | andb $3,%cl |
||
| 46 | L3: rep |
||
| 47 | stosb |
||
| 48 | popl %edi |
||
| 49 | movl 8(%ebp),%eax |
||
| 50 | leave |
||
| 51 | ret |