2 * SIMD-optimized HuffYUV encoding functions
3 * Copyright (c) 2000, 2001 Fabrice Bellard
4 * Copyright (c) 2002-2004 Michael Niedermayer <michaelni@gmx.at>
6 * MMX optimization by Nick Kurshev <nickols_k@mail.ru>
8 * This file is part of FFmpeg.
10 * FFmpeg is free software; you can redistribute it and/or
11 * modify it under the terms of the GNU Lesser General Public
12 * License as published by the Free Software Foundation; either
13 * version 2.1 of the License, or (at your option) any later version.
15 * FFmpeg is distributed in the hope that it will be useful,
16 * but WITHOUT ANY WARRANTY; without even the implied warranty of
17 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
18 * Lesser General Public License for more details.
20 * You should have received a copy of the GNU Lesser General Public
21 * License along with FFmpeg; if not, write to the Free Software
22 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
25 #include "libavutil/attributes.h"
26 #include "libavutil/cpu.h"
27 #include "libavutil/x86/asm.h"
28 #include "libavutil/x86/cpu.h"
29 #include "libavcodec/huffyuvencdsp.h"
30 #include "libavcodec/mathops.h"
32 void ff_diff_bytes_mmx(uint8_t *dst, const uint8_t *src1, const uint8_t *src2,
34 void ff_diff_bytes_sse2(uint8_t *dst, const uint8_t *src1, const uint8_t *src2,
36 void ff_diff_bytes_avx2(uint8_t *dst, const uint8_t *src1, const uint8_t *src2,
41 static void sub_hfyu_median_pred_mmxext(uint8_t *dst, const uint8_t *src1,
42 const uint8_t *src2, intptr_t w,
43 int *left, int *left_top)
49 "movq (%1, %0), %%mm0 \n\t" // LT
50 "psllq $8, %%mm0 \n\t"
52 "movq (%1, %0), %%mm1 \n\t" // T
53 "movq -1(%2, %0), %%mm2 \n\t" // L
54 "movq (%2, %0), %%mm3 \n\t" // X
55 "movq %%mm2, %%mm4 \n\t" // L
56 "psubb %%mm0, %%mm2 \n\t"
57 "paddb %%mm1, %%mm2 \n\t" // L + T - LT
58 "movq %%mm4, %%mm5 \n\t" // L
59 "pmaxub %%mm1, %%mm4 \n\t" // max(T, L)
60 "pminub %%mm5, %%mm1 \n\t" // min(T, L)
61 "pminub %%mm2, %%mm4 \n\t"
62 "pmaxub %%mm1, %%mm4 \n\t"
63 "psubb %%mm4, %%mm3 \n\t" // dst - pred
64 "movq %%mm3, (%3, %0) \n\t"
66 "movq -1(%1, %0), %%mm0 \n\t" // LT
70 : "r" (src1), "r" (src2), "r" (dst), "r" ((x86_reg) w));
75 dst[0] = src2[0] - mid_pred(l, src1[0], (l + src1[0] - lt) & 0xFF);
77 *left_top = src1[w - 1];
81 #endif /* HAVE_INLINE_ASM */
83 av_cold void ff_huffyuvencdsp_init_x86(HuffYUVEncDSPContext *c)
85 av_unused int cpu_flags = av_get_cpu_flags();
87 if (ARCH_X86_32 && EXTERNAL_MMX(cpu_flags)) {
88 c->diff_bytes = ff_diff_bytes_mmx;
92 if (INLINE_MMXEXT(cpu_flags)) {
93 c->sub_hfyu_median_pred = sub_hfyu_median_pred_mmxext;
95 #endif /* HAVE_INLINE_ASM */
97 if (EXTERNAL_SSE2(cpu_flags)) {
98 c->diff_bytes = ff_diff_bytes_sse2;
101 if (EXTERNAL_AVX2_FAST(cpu_flags)) {
102 c->diff_bytes = ff_diff_bytes_avx2;