| 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 23 matching lines...) Expand all Loading... |
| 34 ; r10 = JDIMENSION out_width | 34 ; r10 = JDIMENSION out_width |
| 35 ; r11 = JSAMPIMAGE input_buf | 35 ; r11 = JSAMPIMAGE input_buf |
| 36 ; r12 = JDIMENSION input_row | 36 ; r12 = JDIMENSION input_row |
| 37 ; r13 = JSAMPARRAY output_buf | 37 ; r13 = JSAMPARRAY output_buf |
| 38 ; r14 = int num_rows | 38 ; r14 = int num_rows |
| 39 | 39 |
| 40 %define wk(i) rbp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM] | 40 %define wk(i) rbp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM] |
| 41 %define WK_NUM 2 | 41 %define WK_NUM 2 |
| 42 | 42 |
| 43 align 16 | 43 align 16 |
| 44 » global» EXTN(jsimd_ycc_rgb_convert_sse2) | 44 » global» EXTN(jsimd_ycc_rgb_convert_sse2) PRIVATE |
| 45 | 45 |
| 46 EXTN(jsimd_ycc_rgb_convert_sse2): | 46 EXTN(jsimd_ycc_rgb_convert_sse2): |
| 47 push rbp | 47 push rbp |
| 48 mov rax,rsp ; rax = original rbp | 48 mov rax,rsp ; rax = original rbp |
| 49 sub rsp, byte 4 | 49 sub rsp, byte 4 |
| 50 and rsp, byte (-SIZEOF_XMMWORD) ; align to 128 bits | 50 and rsp, byte (-SIZEOF_XMMWORD) ; align to 128 bits |
| 51 mov [rsp],rax | 51 mov [rsp],rax |
| 52 mov rbp,rsp ; rbp = aligned rbp | 52 mov rbp,rsp ; rbp = aligned rbp |
| 53 lea rsp, [wk(0)] | 53 lea rsp, [wk(0)] |
| 54 collect_args | 54 collect_args |
| (...skipping 423 matching lines...) Expand 10 before | Expand all | Expand 10 after Loading... |
| 478 pop rbx | 478 pop rbx |
| 479 uncollect_args | 479 uncollect_args |
| 480 mov rsp,rbp ; rsp <- aligned rbp | 480 mov rsp,rbp ; rsp <- aligned rbp |
| 481 pop rsp ; rsp <- original rbp | 481 pop rsp ; rsp <- original rbp |
| 482 pop rbp | 482 pop rbp |
| 483 ret | 483 ret |
| 484 | 484 |
| 485 ; For some reason, the OS X linker does not honor the request to align the | 485 ; For some reason, the OS X linker does not honor the request to align the |
| 486 ; segment unless we do this. | 486 ; segment unless we do this. |
| 487 align 16 | 487 align 16 |
| OLD | NEW |