diff options
Diffstat (limited to 'libavcodec/x86/videodsp.asm')
-rw-r--r-- | libavcodec/x86/videodsp.asm | 134 |
1 files changed, 74 insertions, 60 deletions
diff --git a/libavcodec/x86/videodsp.asm b/libavcodec/x86/videodsp.asm index 59f19378ca..1ac02574d6 100644 --- a/libavcodec/x86/videodsp.asm +++ b/libavcodec/x86/videodsp.asm @@ -2,20 +2,20 @@ ;* Core video DSP functions ;* Copyright (c) 2012 Ronald S. Bultje <rsbultje@gmail.com> ;* -;* This file is part of Libav. +;* This file is part of FFmpeg. ;* -;* Libav is free software; you can redistribute it and/or +;* FFmpeg is free software; you can redistribute it and/or ;* modify it under the terms of the GNU Lesser General Public ;* License as published by the Free Software Foundation; either ;* version 2.1 of the License, or (at your option) any later version. ;* -;* Libav is distributed in the hope that it will be useful, +;* FFmpeg is distributed in the hope that it will be useful, ;* but WITHOUT ANY WARRANTY; without even the implied warranty of ;* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU ;* Lesser General Public License for more details. ;* ;* You should have received a copy of the GNU Lesser General Public -;* License along with Libav; if not, write to the Free Software +;* License along with FFmpeg; if not, write to the Free Software ;* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA ;****************************************************************************** @@ -54,13 +54,13 @@ SECTION .text ; | | <- bottom is copied from last line in body of source ; '----' <- bh %if ARCH_X86_64 -cglobal emu_edge_vvar, 7, 8, 1, dst, src, dst_stride, src_stride, \ +cglobal emu_edge_vvar, 7, 8, 1, dst, dst_stride, src, src_stride, \ start_y, end_y, bh, w %else ; x86-32 cglobal emu_edge_vvar, 1, 6, 1, dst, src, start_y, end_y, bh, w %define src_strideq r3mp -%define dst_strideq r2mp - mov srcq, r1mp +%define dst_strideq r1mp + mov srcq, r2mp mov start_yq, r4mp mov end_yq, r5mp mov bhq, r6mp @@ -102,8 +102,8 @@ cglobal emu_edge_hvar, 5, 6, 1, dst, dst_stride, start_x, n_words, h, w imul wd, 0x01010101 ; w *= 0x01010101 movd m0, wd mov wq, n_wordsq ; initialize w -%if cpuflag(sse) - shufps m0, m0, q0000 ; splat +%if cpuflag(sse2) + pshufd m0, m0, q0000 ; splat %else ; mmx punpckldq m0, m0 ; splat %endif ; mmx/sse @@ -124,7 +124,7 @@ INIT_MMX mmx hvar_fn %endif -INIT_XMM sse +INIT_XMM sse2 hvar_fn ; macro to read/write a horizontal number of pixels (%2) to/from registers @@ -137,42 +137,49 @@ hvar_fn ; - if (%2 & 3) fills 1, 2 or 4 bytes in eax ; writing data out is in the same way %macro READ_NUM_BYTES 2 -%assign %%off 0 ; offset in source buffer -%assign %%idx 0 ; mmx/xmm register index +%assign %%off 0 ; offset in source buffer +%assign %%mmx_idx 0 ; mmx register index +%assign %%xmm_idx 0 ; xmm register index %rep %2/mmsize - movu m %+ %%idx, [srcq+%%off] +%if mmsize == 16 + movu xmm %+ %%xmm_idx, [srcq+%%off] +%assign %%xmm_idx %%xmm_idx+1 +%else ; mmx + movu mm %+ %%mmx_idx, [srcq+%%off] +%assign %%mmx_idx %%mmx_idx+1 +%endif %assign %%off %%off+mmsize -%assign %%idx %%idx+1 %endrep ; %2/mmsize %if mmsize == 16 %if (%2-%%off) >= 8 %if %2 > 16 && (%2-%%off) > 8 - movu m %+ %%idx, [srcq+%2-16] + movu xmm %+ %%xmm_idx, [srcq+%2-16] +%assign %%xmm_idx %%xmm_idx+1 %assign %%off %2 %else - movq m %+ %%idx, [srcq+%%off] + movq mm %+ %%mmx_idx, [srcq+%%off] +%assign %%mmx_idx %%mmx_idx+1 %assign %%off %%off+8 %endif -%assign %%idx %%idx+1 %endif ; (%2-%%off) >= 8 %endif %if (%2-%%off) >= 4 %if %2 > 8 && (%2-%%off) > 4 - movq m %+ %%idx, [srcq+%2-8] + movq mm %+ %%mmx_idx, [srcq+%2-8] %assign %%off %2 %else - movd m %+ %%idx, [srcq+%%off] + movd mm %+ %%mmx_idx, [srcq+%%off] %assign %%off %%off+4 %endif -%assign %%idx %%idx+1 +%assign %%mmx_idx %%mmx_idx+1 %endif ; (%2-%%off) >= 4 %if (%2-%%off) >= 1 %if %2 >= 4 - movd m %+ %%idx, [srcq+%2-4] + movd mm %+ %%mmx_idx, [srcq+%2-4] %elif (%2-%%off) == 1 mov valb, [srcq+%2-1] %elif (%2-%%off) == 2 @@ -180,48 +187,55 @@ hvar_fn %elifidn %1, body mov vald, [srcq+%2-3] %else - movd m %+ %%idx, [srcq+%2-3] + movd mm %+ %%mmx_idx, [srcq+%2-3] %endif %endif ; (%2-%%off) >= 1 %endmacro ; READ_NUM_BYTES %macro WRITE_NUM_BYTES 2 -%assign %%off 0 ; offset in destination buffer -%assign %%idx 0 ; mmx/xmm register index +%assign %%off 0 ; offset in destination buffer +%assign %%mmx_idx 0 ; mmx register index +%assign %%xmm_idx 0 ; xmm register index %rep %2/mmsize - movu [dstq+%%off], m %+ %%idx +%if mmsize == 16 + movu [dstq+%%off], xmm %+ %%xmm_idx +%assign %%xmm_idx %%xmm_idx+1 +%else ; mmx + movu [dstq+%%off], mm %+ %%mmx_idx +%assign %%mmx_idx %%mmx_idx+1 +%endif %assign %%off %%off+mmsize -%assign %%idx %%idx+1 %endrep ; %2/mmsize %if mmsize == 16 %if (%2-%%off) >= 8 %if %2 > 16 && (%2-%%off) > 8 - movu [dstq+%2-16], m %+ %%idx + movu [dstq+%2-16], xmm %+ %%xmm_idx +%assign %%xmm_idx %%xmm_idx+1 %assign %%off %2 %else - movq [dstq+%%off], m %+ %%idx + movq [dstq+%%off], mm %+ %%mmx_idx +%assign %%mmx_idx %%mmx_idx+1 %assign %%off %%off+8 %endif -%assign %%idx %%idx+1 %endif ; (%2-%%off) >= 8 %endif %if (%2-%%off) >= 4 %if %2 > 8 && (%2-%%off) > 4 - movq [dstq+%2-8], m %+ %%idx + movq [dstq+%2-8], mm %+ %%mmx_idx %assign %%off %2 %else - movd [dstq+%%off], m %+ %%idx + movd [dstq+%%off], mm %+ %%mmx_idx %assign %%off %%off+4 %endif -%assign %%idx %%idx+1 +%assign %%mmx_idx %%mmx_idx+1 %endif ; (%2-%%off) >= 4 %if (%2-%%off) >= 1 %if %2 >= 4 - movd [dstq+%2-4], m %+ %%idx + movd [dstq+%2-4], mm %+ %%mmx_idx %elif (%2-%%off) == 1 mov [dstq+%2-1], valb %elif (%2-%%off) == 2 @@ -231,7 +245,7 @@ hvar_fn shr vald, 16 mov [dstq+%2-1], valb %else - movd vald, m %+ %%idx + movd vald, mm %+ %%mmx_idx mov [dstq+%2-3], valw shr vald, 16 mov [dstq+%2-1], valb @@ -248,30 +262,30 @@ hvar_fn %rep 1+%2-%1 %if %%n <= 3 %if ARCH_X86_64 -cglobal emu_edge_vfix %+ %%n, 6, 8, 0, dst, src, dst_stride, src_stride, \ +cglobal emu_edge_vfix %+ %%n, 6, 8, 0, dst, dst_stride, src, src_stride, \ start_y, end_y, val, bh mov bhq, r6mp ; r6mp = bhmp %else ; x86-32 cglobal emu_edge_vfix %+ %%n, 0, 6, 0, val, dst, src, start_y, end_y, bh mov dstq, r0mp - mov srcq, r1mp + mov srcq, r2mp mov start_yq, r4mp mov end_yq, r5mp mov bhq, r6mp -%define dst_strideq r2mp +%define dst_strideq r1mp %define src_strideq r3mp %endif ; x86-64/32 %else %if ARCH_X86_64 -cglobal emu_edge_vfix %+ %%n, 7, 7, 1, dst, src, dst_stride, src_stride, \ +cglobal emu_edge_vfix %+ %%n, 7, 7, 1, dst, dst_stride, src, src_stride, \ start_y, end_y, bh %else ; x86-32 cglobal emu_edge_vfix %+ %%n, 1, 5, 1, dst, src, start_y, end_y, bh - mov srcq, r1mp + mov srcq, r2mp mov start_yq, r4mp mov end_yq, r5mp mov bhq, r6mp -%define dst_strideq r2mp +%define dst_strideq r1mp %define src_strideq r3mp %endif ; x86-64/32 %endif @@ -330,25 +344,23 @@ VERTICAL_EXTEND 16, 22 ; obviously not the same on both sides. %macro READ_V_PIXEL 2 -%if %1 == 2 - movzx valw, byte %2 - imul valw, 0x0101 -%else movzx vald, byte %2 imul vald, 0x01010101 %if %1 >= 8 movd m0, vald %if mmsize == 16 - shufps m0, m0, q0000 + pshufd m0, m0, q0000 %else punpckldq m0, m0 -%endif -%endif ; %1 >= 8 -%endif +%endif ; mmsize == 16 +%endif ; %1 > 16 %endmacro ; READ_V_PIXEL %macro WRITE_V_PIXEL 2 %assign %%off 0 + +%if %1 >= 8 + %rep %1/mmsize movu [%2+%%off], m0 %assign %%off %%off+mmsize @@ -364,27 +376,29 @@ VERTICAL_EXTEND 16, 22 %assign %%off %%off+8 %endif %endif ; %1-%%off >= 8 -%endif +%endif ; mmsize == 16 %if %1-%%off >= 4 -%if %1 > 8 %% %1-%%off > 4 +%if %1 > 8 && %1-%%off > 4 movq [%2+%1-8], m0 %assign %%off %1 -%elif %1 >= 8 && %1-%%off >= 4 - movd [%2+%%off], m0 -%assign %%off %%off+4 %else - mov [%2+%%off], vald + movd [%2+%%off], m0 %assign %%off %%off+4 %endif %endif ; %1-%%off >= 4 -%if %1-%%off >= 2 -%if %1 >= 8 - movd [%2+%1-4], m0 -%else +%else ; %1 < 8 + +%rep %1/4 + mov [%2+%%off], vald +%assign %%off %%off+4 +%endrep ; %1/4 + +%endif ; %1 >=/< 8 + +%if %1-%%off == 2 mov [%2+%%off], valw -%endif %endif ; (%1-%%off)/2 %endmacro ; WRITE_V_PIXEL @@ -409,7 +423,7 @@ H_EXTEND 2, 14 H_EXTEND 16, 22 %endif -INIT_XMM sse +INIT_XMM sse2 H_EXTEND 16, 22 %macro PREFETCH_FN 1 |