2 * Alpha optimized DSP utils
3 * Copyright (c) 2002 Falk Hueffner <falk@debian.org>
5 * This library is free software; you can redistribute it and/or
6 * modify it under the terms of the GNU Lesser General Public
7 * License as published by the Free Software Foundation; either
8 * version 2 of the License, or (at your option) any later version.
10 * This library is distributed in the hope that it will be useful,
11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
13 * Lesser General Public License for more details.
15 * You should have received a copy of the GNU Lesser General Public
16 * License along with this library; if not, write to the Free Software
17 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
21 #include "../dsputil.h"
22 #include "../mpegvideo.h"
24 static void dct_unquantize_h263_intra_axp(MpegEncContext *s, DCTELEM *block,
30 DCTELEM *orig_block = block;
31 DCTELEM block0; /* might not be used uninitialized */
33 qadd = WORD_VEC((qscale - 1) | 1);
35 /* This mask kills spill from negative subwords to the next subword. */
36 correction = WORD_VEC((qmul - 1) + 1); /* multiplication / addition */
40 block0 = block[0] * s->y_dc_scale;
42 block0 = block[0] * s->c_dc_scale;
46 n_coeffs = 63; // does not always use zigzag table
48 for(i = 0; i <= n_coeffs; block += 4, i += 4) {
49 uint64_t levels, negmask, zeros, add;
56 /* I don't think the speed difference justifies runtime
58 negmask = maxsw4(levels, -1); /* negative -> ffff (-1) */
59 negmask = minsw4(negmask, 0); /* positive -> 0000 (0) */
61 negmask = cmpbge(WORD_VEC(0x7fff), levels);
62 negmask &= (negmask >> 1) | (1 << 7);
63 negmask = zap(-1, negmask);
66 zeros = cmpbge(0, levels);
68 /* zeros |= zeros << 1 is not needed since qadd <= 255, so
69 zapping the lower byte suffices. */
72 levels -= correction & (negmask << 16);
74 /* Negate qadd for negative levels. */
76 add += WORD_VEC(0x0001) & negmask;
77 /* Set qadd to 0 for levels == 0. */
78 add = zap(add, zeros);
85 if (s->mb_intra && !s->h263_aic)
86 orig_block[0] = block0;
89 static void dct_unquantize_h263_inter_axp(MpegEncContext *s, DCTELEM *block,
96 qadd = WORD_VEC((qscale - 1) | 1);
98 /* This mask kills spill from negative subwords to the next subword. */
99 correction = WORD_VEC((qmul - 1) + 1); /* multiplication / addition */
101 n_coeffs = s->intra_scantable.raster_end[s->block_last_index[n]];
103 for(i = 0; i <= n_coeffs; block += 4, i += 4) {
104 uint64_t levels, negmask, zeros, add;
111 /* I don't think the speed difference justifies runtime
113 negmask = maxsw4(levels, -1); /* negative -> ffff (-1) */
114 negmask = minsw4(negmask, 0); /* positive -> 0000 (0) */
116 negmask = cmpbge(WORD_VEC(0x7fff), levels);
117 negmask &= (negmask >> 1) | (1 << 7);
118 negmask = zap(-1, negmask);
121 zeros = cmpbge(0, levels);
123 /* zeros |= zeros << 1 is not needed since qadd <= 255, so
124 zapping the lower byte suffices. */
127 levels -= correction & (negmask << 16);
129 /* Negate qadd for negative levels. */
130 add = qadd ^ negmask;
131 add += WORD_VEC(0x0001) & negmask;
132 /* Set qadd to 0 for levels == 0. */
133 add = zap(add, zeros);
141 void MPV_common_init_axp(MpegEncContext *s)
143 s->dct_unquantize_h263_intra = dct_unquantize_h263_intra_axp;
144 s->dct_unquantize_h263_inter = dct_unquantize_h263_inter_axp;