| OLD | NEW |
| 1 ; | 1 ; |
| 2 ; jdclrss2-64.asm - colorspace conversion (64-bit SSE2) | 2 ; jdclrss2-64.asm - colorspace conversion (64-bit SSE2) |
| 3 ; | 3 ; |
| 4 ; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB | 4 ; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB |
| 5 ; Copyright 2009 D. R. Commander | 5 ; Copyright 2009 D. R. Commander |
| 6 ; | 6 ; |
| 7 ; Based on | 7 ; Based on |
| 8 ; x86 SIMD extension for IJG JPEG library | 8 ; x86 SIMD extension for IJG JPEG library |
| 9 ; Copyright (C) 1999-2006, MIYASAKA Masaru. | 9 ; Copyright (C) 1999-2006, MIYASAKA Masaru. |
| 10 ; For conditions of distribution and use, see copyright notice in jsimdext.inc | 10 ; For conditions of distribution and use, see copyright notice in jsimdext.inc |
| (...skipping 293 matching lines...) Expand 10 before | Expand all | Expand 10 after Loading... |
| 304 ; space. | 304 ; space. |
| 305 cmp rcx, byte SIZEOF_DWORD | 305 cmp rcx, byte SIZEOF_DWORD |
| 306 jb short .column_st3 | 306 jb short .column_st3 |
| 307 movd DWORD [rdi], xmmA | 307 movd DWORD [rdi], xmmA |
| 308 add rdi, byte SIZEOF_DWORD | 308 add rdi, byte SIZEOF_DWORD |
| 309 sub rcx, byte SIZEOF_DWORD | 309 sub rcx, byte SIZEOF_DWORD |
| 310 psrldq xmmA, SIZEOF_DWORD | 310 psrldq xmmA, SIZEOF_DWORD |
| 311 .column_st3: | 311 .column_st3: |
| 312 ; Store the lower 2 bytes of rax to the output when it has enough | 312 ; Store the lower 2 bytes of rax to the output when it has enough |
| 313 ; space. | 313 ; space. |
| 314 » movd» rax, xmmA | 314 » movd» eax, xmmA |
| 315 cmp rcx, byte SIZEOF_WORD | 315 cmp rcx, byte SIZEOF_WORD |
| 316 jb short .column_st1 | 316 jb short .column_st1 |
| 317 mov WORD [rdi], ax | 317 mov WORD [rdi], ax |
| 318 add rdi, byte SIZEOF_WORD | 318 add rdi, byte SIZEOF_WORD |
| 319 sub rcx, byte SIZEOF_WORD | 319 sub rcx, byte SIZEOF_WORD |
| 320 shr rax, 16 | 320 shr rax, 16 |
| 321 .column_st1: | 321 .column_st1: |
| 322 ; Store the lower 1 byte of rax to the output when it has enough | 322 ; Store the lower 1 byte of rax to the output when it has enough |
| 323 ; space. | 323 ; space. |
| 324 test rcx, rcx | 324 test rcx, rcx |
| (...skipping 204 matching lines...) Expand 10 before | Expand all | Expand 10 after Loading... |
| 529 pop rbx | 529 pop rbx |
| 530 uncollect_args | 530 uncollect_args |
| 531 mov rsp,rbp ; rsp <- aligned rbp | 531 mov rsp,rbp ; rsp <- aligned rbp |
| 532 pop rsp ; rsp <- original rbp | 532 pop rsp ; rsp <- original rbp |
| 533 pop rbp | 533 pop rbp |
| 534 ret | 534 ret |
| 535 | 535 |
| 536 ; For some reason, the OS X linker does not honor the request to align the | 536 ; For some reason, the OS X linker does not honor the request to align the |
| 537 ; segment unless we do this. | 537 ; segment unless we do this. |
| 538 align 16 | 538 align 16 |
| OLD | NEW |