Chromium Code Reviews
chromiumcodereview-hr@appspot.gserviceaccount.com (chromiumcodereview-hr) | Please choose your nickname with Settings | Help | Chromium Project | Gerrit Changes | Sign out
(438)

Issues Search
    My Issues | Starred     Open | Closed | All

Side by Side Diff: source/libvpx/vp9/common/mips/msa/vp9_convolve_msa.h

Issue 1124333011: libvpx: Pull from upstream (Closed) Base URL: https://chromium.googlesource.com/chromium/deps/libvpx.git@master
Patch Set: only update to last nights LKGR Created 5 years, 7 months ago
Use n/p to move between diff chunks; N/P to move between comments. Draft comments are only viewable by you.
Jump to:
View unified diff | Download patch
« no previous file with comments | « source/libvpx/vp9/common/mips/msa/vp9_convolve_copy_msa.c ('k') | source/libvpx/vp9/common/mips/msa/vp9_idct16x16_msa.c » ('j') | no next file with comments »
Toggle Intra-line Diffs ('i') | Expand Comments ('e') | Collapse Comments ('c') | Show Comments Hide Comments ('s')
OLDNEW
(Empty)
1 /*
2 * Copyright (c) 2015 The WebM project authors. All Rights Reserved.
3 *
4 * Use of this source code is governed by a BSD-style license
5 * that can be found in the LICENSE file in the root of the source
6 * tree. An additional intellectual property rights grant can be found
7 * in the file PATENTS. All contributing project authors may
8 * be found in the AUTHORS file in the root of the source tree.
9 */
10
11 #ifndef VP9_COMMON_MIPS_MSA_VP9_CONVOLVE_MSA_H_
12 #define VP9_COMMON_MIPS_MSA_VP9_CONVOLVE_MSA_H_
13
14 #include "vp9/common/vp9_filter.h"
15 #include "vp9/common/mips/msa/vp9_macros_msa.h"
16
17 extern const uint8_t mc_filt_mask_arr[16 * 3];
18
19 #define HORIZ_8TAP_FILT(src, mask0, mask1, mask2, mask3, \
20 filt_h0, filt_h1, filt_h2, filt_h3) ({ \
21 v8i16 vec0, vec1, vec2, vec3, horiz_out; \
22 \
23 vec0 = (v8i16)__msa_vshf_b((v16i8)(mask0), (v16i8)(src), (v16i8)(src)); \
24 vec0 = __msa_dotp_s_h((v16i8)vec0, (v16i8)(filt_h0)); \
25 vec1 = (v8i16)__msa_vshf_b((v16i8)(mask1), (v16i8)(src), (v16i8)(src)); \
26 vec0 = __msa_dpadd_s_h(vec0, (v16i8)(filt_h1), (v16i8)vec1); \
27 vec2 = (v8i16)__msa_vshf_b((v16i8)(mask2), (v16i8)(src), (v16i8)(src)); \
28 vec2 = __msa_dotp_s_h((v16i8)vec2, (v16i8)(filt_h2)); \
29 vec3 = (v8i16)__msa_vshf_b((v16i8)(mask3), (v16i8)(src), (v16i8)(src)); \
30 vec2 = __msa_dpadd_s_h(vec2, (v16i8)(filt_h3), (v16i8)vec3); \
31 vec0 = __msa_adds_s_h(vec0, vec2); \
32 horiz_out = SRARI_SATURATE_SIGNED_H(vec0, FILTER_BITS, 7); \
33 \
34 horiz_out; \
35 })
36
37 #define HORIZ_8TAP_FILT_2VECS(src0, src1, mask0, mask1, mask2, mask3, \
38 filt_h0, filt_h1, filt_h2, filt_h3) ({ \
39 v8i16 vec0, vec1, vec2, vec3, horiz_out; \
40 \
41 vec0 = (v8i16)__msa_vshf_b((v16i8)(mask0), (v16i8)(src1), (v16i8)(src0)); \
42 vec0 = __msa_dotp_s_h((v16i8)vec0, (v16i8)(filt_h0)); \
43 vec1 = (v8i16)__msa_vshf_b((v16i8)(mask1), (v16i8)(src1), (v16i8)(src0)); \
44 vec0 = __msa_dpadd_s_h(vec0, (v16i8)(filt_h1), (v16i8)vec1); \
45 vec2 = (v8i16)__msa_vshf_b((v16i8)(mask2), (v16i8)(src1), (v16i8)(src0)); \
46 vec2 = __msa_dotp_s_h((v16i8)vec2, (v16i8)(filt_h2)); \
47 vec3 = (v8i16)__msa_vshf_b((v16i8)(mask3), (v16i8)(src1), (v16i8)(src0)); \
48 vec2 = __msa_dpadd_s_h(vec2, ((v16i8)filt_h3), (v16i8)vec3); \
49 vec0 = __msa_adds_s_h(vec0, vec2); \
50 horiz_out = (v8i16)SRARI_SATURATE_SIGNED_H(vec0, FILTER_BITS, 7); \
51 \
52 horiz_out; \
53 })
54
55 #define FILT_8TAP_DPADD_S_H(vec0, vec1, vec2, vec3, \
56 filt0, filt1, filt2, filt3) ({ \
57 v8i16 tmp0, tmp1; \
58 \
59 tmp0 = __msa_dotp_s_h((v16i8)(vec0), (v16i8)(filt0)); \
60 tmp0 = __msa_dpadd_s_h(tmp0, (v16i8)(vec1), (v16i8)(filt1)); \
61 tmp1 = __msa_dotp_s_h((v16i8)(vec2), (v16i8)(filt2)); \
62 tmp1 = __msa_dpadd_s_h(tmp1, (v16i8)(vec3), ((v16i8)filt3)); \
63 tmp0 = __msa_adds_s_h(tmp0, tmp1); \
64 \
65 tmp0; \
66 })
67
68 #define HORIZ_8TAP_4WID_4VECS_FILT(src0, src1, src2, src3, \
69 mask0, mask1, mask2, mask3, \
70 filt0, filt1, filt2, filt3, \
71 out0, out1) { \
72 v8i16 vec0_m, vec1_m, vec2_m, vec3_m, vec4_m, vec5_m, vec6_m, vec7_m; \
73 v8i16 res0_m, res1_m, res2_m, res3_m; \
74 \
75 vec0_m = (v8i16)__msa_vshf_b((v16i8)(mask0), (v16i8)(src1), (v16i8)(src0)); \
76 vec1_m = (v8i16)__msa_vshf_b((v16i8)(mask0), (v16i8)(src3), (v16i8)(src2)); \
77 \
78 res0_m = __msa_dotp_s_h((v16i8)vec0_m, (v16i8)(filt0)); \
79 res1_m = __msa_dotp_s_h((v16i8)vec1_m, (v16i8)(filt0)); \
80 \
81 vec2_m = (v8i16)__msa_vshf_b((v16i8)(mask1), (v16i8)(src1), (v16i8)(src0)); \
82 vec3_m = (v8i16)__msa_vshf_b((v16i8)(mask1), (v16i8)(src3), (v16i8)(src2)); \
83 \
84 res0_m = __msa_dpadd_s_h(res0_m, (filt1), (v16i8)vec2_m); \
85 res1_m = __msa_dpadd_s_h(res1_m, (filt1), (v16i8)vec3_m); \
86 \
87 vec4_m = (v8i16)__msa_vshf_b((v16i8)(mask2), (v16i8)(src1), (v16i8)(src0)); \
88 vec5_m = (v8i16)__msa_vshf_b((v16i8)(mask2), (v16i8)(src3), (v16i8)(src2)); \
89 \
90 res2_m = __msa_dotp_s_h((v16i8)(filt2), (v16i8)vec4_m); \
91 res3_m = __msa_dotp_s_h((v16i8)(filt2), (v16i8)vec5_m); \
92 \
93 vec6_m = (v8i16)__msa_vshf_b((v16i8)(mask3), (v16i8)(src1), (v16i8)(src0)); \
94 vec7_m = (v8i16)__msa_vshf_b((v16i8)(mask3), (v16i8)(src3), (v16i8)(src2)); \
95 \
96 res2_m = __msa_dpadd_s_h(res2_m, (v16i8)(filt3), (v16i8)vec6_m); \
97 res3_m = __msa_dpadd_s_h(res3_m, (v16i8)(filt3), (v16i8)vec7_m); \
98 \
99 out0 = __msa_adds_s_h(res0_m, res2_m); \
100 out1 = __msa_adds_s_h(res1_m, res3_m); \
101 }
102
103 #define HORIZ_8TAP_8WID_4VECS_FILT(src0, src1, src2, src3, \
104 mask0, mask1, mask2, mask3, \
105 filt0, filt1, filt2, filt3, \
106 out0, out1, out2, out3) { \
107 v8i16 vec0_m, vec1_m, vec2_m, vec3_m; \
108 v8i16 vec4_m, vec5_m, vec6_m, vec7_m; \
109 v8i16 res0_m, res1_m, res2_m, res3_m; \
110 v8i16 res4_m, res5_m, res6_m, res7_m; \
111 \
112 vec0_m = (v8i16)__msa_vshf_b((v16i8)(mask0), (v16i8)(src0), (v16i8)(src0)); \
113 vec1_m = (v8i16)__msa_vshf_b((v16i8)(mask0), (v16i8)(src1), (v16i8)(src1)); \
114 vec2_m = (v8i16)__msa_vshf_b((v16i8)(mask0), (v16i8)(src2), (v16i8)(src2)); \
115 vec3_m = (v8i16)__msa_vshf_b((v16i8)(mask0), (v16i8)(src3), (v16i8)(src3)); \
116 \
117 res0_m = __msa_dotp_s_h((v16i8)vec0_m, (v16i8)(filt0)); \
118 res1_m = __msa_dotp_s_h((v16i8)vec1_m, (v16i8)(filt0)); \
119 res2_m = __msa_dotp_s_h((v16i8)vec2_m, (v16i8)(filt0)); \
120 res3_m = __msa_dotp_s_h((v16i8)vec3_m, (v16i8)(filt0)); \
121 \
122 vec0_m = (v8i16)__msa_vshf_b((v16i8)(mask2), (v16i8)(src0), (v16i8)(src0)); \
123 vec1_m = (v8i16)__msa_vshf_b((v16i8)(mask2), (v16i8)(src1), (v16i8)(src1)); \
124 vec2_m = (v8i16)__msa_vshf_b((v16i8)(mask2), (v16i8)(src2), (v16i8)(src2)); \
125 vec3_m = (v8i16)__msa_vshf_b((v16i8)(mask2), (v16i8)(src3), (v16i8)(src3)); \
126 \
127 res4_m = __msa_dotp_s_h((v16i8)vec0_m, (v16i8)(filt2)); \
128 res5_m = __msa_dotp_s_h((v16i8)vec1_m, (v16i8)(filt2)); \
129 res6_m = __msa_dotp_s_h((v16i8)vec2_m, (v16i8)(filt2)); \
130 res7_m = __msa_dotp_s_h((v16i8)vec3_m, (v16i8)(filt2)); \
131 \
132 vec4_m = (v8i16)__msa_vshf_b((v16i8)(mask1), (v16i8)(src0), (v16i8)(src0)); \
133 vec5_m = (v8i16)__msa_vshf_b((v16i8)(mask1), (v16i8)(src1), (v16i8)(src1)); \
134 vec6_m = (v8i16)__msa_vshf_b((v16i8)(mask1), (v16i8)(src2), (v16i8)(src2)); \
135 vec7_m = (v8i16)__msa_vshf_b((v16i8)(mask1), (v16i8)(src3), (v16i8)(src3)); \
136 \
137 res0_m = __msa_dpadd_s_h(res0_m, (v16i8)(filt1), (v16i8)vec4_m); \
138 res1_m = __msa_dpadd_s_h(res1_m, (v16i8)(filt1), (v16i8)vec5_m); \
139 res2_m = __msa_dpadd_s_h(res2_m, (v16i8)(filt1), (v16i8)vec6_m); \
140 res3_m = __msa_dpadd_s_h(res3_m, (v16i8)(filt1), (v16i8)vec7_m); \
141 \
142 vec4_m = (v8i16)__msa_vshf_b((v16i8)(mask3), (v16i8)(src0), (v16i8)(src0)); \
143 vec5_m = (v8i16)__msa_vshf_b((v16i8)(mask3), (v16i8)(src1), (v16i8)(src1)); \
144 vec6_m = (v8i16)__msa_vshf_b((v16i8)(mask3), (v16i8)(src2), (v16i8)(src2)); \
145 vec7_m = (v8i16)__msa_vshf_b((v16i8)(mask3), (v16i8)(src3), (v16i8)(src3)); \
146 \
147 res4_m = __msa_dpadd_s_h(res4_m, (v16i8)(filt3), (v16i8)vec4_m); \
148 res5_m = __msa_dpadd_s_h(res5_m, (v16i8)(filt3), (v16i8)vec5_m); \
149 res6_m = __msa_dpadd_s_h(res6_m, (v16i8)(filt3), (v16i8)vec6_m); \
150 res7_m = __msa_dpadd_s_h(res7_m, (v16i8)(filt3), (v16i8)vec7_m); \
151 \
152 out0 = __msa_adds_s_h(res0_m, res4_m); \
153 out1 = __msa_adds_s_h(res1_m, res5_m); \
154 out2 = __msa_adds_s_h(res2_m, res6_m); \
155 out3 = __msa_adds_s_h(res3_m, res7_m); \
156 }
157 #endif /* VP9_COMMON_MIPS_MSA_VP9_CONVOLVE_MSA_H_ */
OLDNEW
« no previous file with comments | « source/libvpx/vp9/common/mips/msa/vp9_convolve_copy_msa.c ('k') | source/libvpx/vp9/common/mips/msa/vp9_idct16x16_msa.c » ('j') | no next file with comments »

Powered by Google App Engine
This is Rietveld 408576698