1. ffmpeg
  2. Commit fd6045bae5bf097e33906eac03ac44f1946f85e1
Browse code

Alpha: fix pix_abs16

Originally committed as revision 16675 to svn://svn.ffmpeg.org/ffmpeg/trunk

Måns Rullgård authored on 2009/01/19 03:31:52
Showing 2 changed files
libavcodec/alpha/dsputil_alpha.c
History View file @ fd6045b
... ... 
@@ -42,7 +42,7 @@ void get_pixels_mvi(DCTELEM *restrict block,
42 42 
 void diff_pixels_mvi(DCTELEM *block, const uint8_t *s1, const uint8_t *s2,
43 43 
                      int stride);
44 44 
 int pix_abs8x8_mvi(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int h);
45 
-int pix_abs16x16_mvi_asm(uint8_t *pix1, uint8_t *pix2, int line_size);
45 
+int pix_abs16x16_mvi_asm(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int h);
46 46 
 int pix_abs16x16_x2_mvi(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int h);
47 47 
 int pix_abs16x16_y2_mvi(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int h);
48 48 
 int pix_abs16x16_xy2_mvi(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int h);
... ... 
@@ -287,11 +287,6 @@ void put_pixels16_axp_asm(uint8_t *block, const uint8_t *pixels,
287 287 
     put_pixels_axp_asm(block + 8, pixels + 8, line_size, h);
288 288 
 }
289 289
290 
-static int sad16x16_mvi(void *s, uint8_t *a, uint8_t *b, int stride)
291 
-{
292 
-    return pix_abs16x16_mvi_asm(a, b, stride);
293 
-}
294 
-
295 290 
 void dsputil_init_alpha(DSPContext* c, AVCodecContext *avctx)
296 291 
 {
297 292 
     c->put_pixels_tab[0][0] = put_pixels16_axp_asm;
... ... 
@@ -343,10 +338,9 @@ void dsputil_init_alpha(DSPContext* c, AVCodecContext *avctx)
343 343
344 344 
         c->get_pixels       = get_pixels_mvi;
345 345 
         c->diff_pixels      = diff_pixels_mvi;
346 
-        c->sad[0]           = sad16x16_mvi;
346 
+        c->sad[0]           = pix_abs16x16_mvi_asm;
347 347 
         c->sad[1]           = pix_abs8x8_mvi;
348 
-//        c->pix_abs[0][0]    = pix_abs16x16_mvi_asm; //FIXME function arguments for the asm must be fixed
349 
-        c->pix_abs[0][0]    = sad16x16_mvi;
348 
+        c->pix_abs[0][0]    = pix_abs16x16_mvi_asm;
350 349 
         c->pix_abs[1][0]    = pix_abs8x8_mvi;
351 350 
         c->pix_abs[0][1]    = pix_abs16x16_x2_mvi;
352 351 
         c->pix_abs[0][2]    = pix_abs16x16_y2_mvi;
libavcodec/alpha/motion_est_mvi_asm.S
History View file @ fd6045b
... ... 
@@ -60,9 +60,8 @@ pix_abs16x16_mvi_asm:
60 60 
         jsr     AT, (AT), _mcount
61 61 
 #endif
62 62
63 
-        and     a1, 7, t0
63 
+        and     a2, 7, t0
64 64 
         clr     v0
65 
-        lda     a3, 16
66 65 
         beq     t0, $aligned
67 66 
         .align 4
68 67 
 $unaligned:
... ... 
@@ -86,80 +85,80 @@ $unaligned:
86 86 
            td:  error right  */
87 87
88 88 
         /* load line 0 */
89 
-        ldq_u   t0, 0(a1)       # left_u
90 
-        ldq_u   t1, 8(a1)       # mid
91 
-        ldq_u   t2, 16(a1)      # right_u
92 
-        ldq     t3, 0(a0)       # ref left
93 
-        ldq     t4, 8(a0)       # ref right
94 
-        addq    a0, a2, a0      # pix1
95 
-        addq    a1, a2, a1      # pix2
89 
+        ldq_u   t0, 0(a2)       # left_u
90 
+        ldq_u   t1, 8(a2)       # mid
91 
+        ldq_u   t2, 16(a2)      # right_u
92 
+        ldq     t3, 0(a1)       # ref left
93 
+        ldq     t4, 8(a1)       # ref right
94 
+        addq    a1, a3, a1      # pix1
95 
+        addq    a2, a3, a2      # pix2
96 96 
         /* load line 1 */
97 
-        ldq_u   t5, 0(a1)       # left_u
98 
-        ldq_u   t6, 8(a1)       # mid
99 
-        ldq_u   t7, 16(a1)      # right_u
100 
-        ldq     t8, 0(a0)       # ref left
101 
-        ldq     t9, 8(a0)       # ref right
102 
-        addq    a0, a2, a0      # pix1
103 
-        addq    a1, a2, a1      # pix2
97 
+        ldq_u   t5, 0(a2)       # left_u
98 
+        ldq_u   t6, 8(a2)       # mid
99 
+        ldq_u   t7, 16(a2)      # right_u
100 
+        ldq     t8, 0(a1)       # ref left
101 
+        ldq     t9, 8(a1)       # ref right
102 
+        addq    a1, a3, a1      # pix1
103 
+        addq    a2, a3, a2      # pix2
104 104 
         /* calc line 0 */
105 
-        extql   t0, a1, t0      # left lo
106 
-        extqh   t1, a1, ta      # left hi
107 
-        extql   t1, a1, tb      # right lo
105 
+        extql   t0, a2, t0      # left lo
106 
+        extqh   t1, a2, ta      # left hi
107 
+        extql   t1, a2, tb      # right lo
108 108 
         or      t0, ta, t0      # left
109 
-        extqh   t2, a1, t2      # right hi
109 
+        extqh   t2, a2, t2      # right hi
110 110 
         perr    t3, t0, tc      # error left
111 111 
         or      t2, tb, t2      # right
112 112 
         perr    t4, t2, td      # error right
113 113 
         addq    v0, tc, v0      # add error left
114 114 
         addq    v0, td, v0      # add error left
115 115 
         /* calc line 1 */
116 
-        extql   t5, a1, t5      # left lo
117 
-        extqh   t6, a1, ta      # left hi
118 
-        extql   t6, a1, tb      # right lo
116 
+        extql   t5, a2, t5      # left lo
117 
+        extqh   t6, a2, ta      # left hi
118 
+        extql   t6, a2, tb      # right lo
119 119 
         or      t5, ta, t5      # left
120 
-        extqh   t7, a1, t7      # right hi
120 
+        extqh   t7, a2, t7      # right hi
121 121 
         perr    t8, t5, tc      # error left
122 122 
         or      t7, tb, t7      # right
123 123 
         perr    t9, t7, td      # error right
124 124 
         addq    v0, tc, v0      # add error left
125 125 
         addq    v0, td, v0      # add error left
126 126 
         /* loop */
127 
-        subq    a3,  2, a3      # h -= 2
128 
-        bne     a3, $unaligned
127 
+        subq    a4,  2, a4      # h -= 2
128 
+        bne     a4, $unaligned
129 129 
         ret
130 130
131 131 
         .align 4
132 132 
 $aligned:
133 133 
         /* load line 0 */
134 
-        ldq     t0, 0(a1)       # left
135 
-        ldq     t1, 8(a1)       # right
136 
-        addq    a1, a2, a1      # pix2
137 
-        ldq     t2, 0(a0)       # ref left
138 
-        ldq     t3, 8(a0)       # ref right
139 
-        addq    a0, a2, a0      # pix1
134 
+        ldq     t0, 0(a2)       # left
135 
+        ldq     t1, 8(a2)       # right
136 
+        addq    a2, a3, a2      # pix2
137 
+        ldq     t2, 0(a1)       # ref left
138 
+        ldq     t3, 8(a1)       # ref right
139 
+        addq    a1, a3, a1      # pix1
140 140 
         /* load line 1 */
141 
-        ldq     t4, 0(a1)       # left
142 
-        ldq     t5, 8(a1)       # right
143 
-        addq    a1, a2, a1      # pix2
144 
-        ldq     t6, 0(a0)       # ref left
145 
-        ldq     t7, 8(a0)       # ref right
146 
-        addq    a0, a2, a0      # pix1
141 
+        ldq     t4, 0(a2)       # left
142 
+        ldq     t5, 8(a2)       # right
143 
+        addq    a2, a3, a2      # pix2
144 
+        ldq     t6, 0(a1)       # ref left
145 
+        ldq     t7, 8(a1)       # ref right
146 
+        addq    a1, a3, a1      # pix1
147 147 
         /* load line 2 */
148 
-        ldq     t8, 0(a1)       # left
149 
-        ldq     t9, 8(a1)       # right
150 
-        addq    a1, a2, a1      # pix2
151 
-        ldq     ta, 0(a0)       # ref left
152 
-        ldq     tb, 8(a0)       # ref right
153 
-        addq    a0, a2, a0      # pix1
148 
+        ldq     t8, 0(a2)       # left
149 
+        ldq     t9, 8(a2)       # right
150 
+        addq    a2, a3, a2      # pix2
151 
+        ldq     ta, 0(a1)       # ref left
152 
+        ldq     tb, 8(a1)       # ref right
153 
+        addq    a1, a3, a1      # pix1
154 154 
         /* load line 3 */
155 
-        ldq     tc, 0(a1)       # left
156 
-        ldq     td, 8(a1)       # right
157 
-        addq    a1, a2, a1      # pix2
158 
-        ldq     te, 0(a0)       # ref left
159 
-        ldq     tf, 8(a0)       # ref right
155 
+        ldq     tc, 0(a2)       # left
156 
+        ldq     td, 8(a2)       # right
157 
+        addq    a2, a3, a2      # pix2
158 
+        ldq     te, 0(a1)       # ref left
159 
+        ldq     a0, 8(a1)       # ref right
160 160 
         /* calc line 0 */
161 161 
         perr    t0, t2, t0      # error left
162 
-        addq    a0, a2, a0      # pix1
162 
+        addq    a1, a3, a1      # pix1
163 163 
         perr    t1, t3, t1      # error right
164 164 
         addq    v0, t0, v0      # add error left
165 165 
         /* calc line 1 */
... ... 
@@ -175,11 +174,11 @@ $aligned:
175 175 
         /* calc line 3 */
176 176 
         perr    tc, te, t0      # error left
177 177 
         addq    v0, t1, v0      # add error right
178 
-        perr    td, tf, t1      # error right
178 
+        perr    td, a0, t1      # error right
179 179 
         addq    v0, t0, v0      # add error left
180 180 
         addq    v0, t1, v0      # add error right
181 181 
         /* loop */
182 
-        subq    a3,  4, a3      # h -= 4
183 
-        bne     a3, $aligned
182 
+        subq    a4,  4, a4      # h -= 4
183 
+        bne     a4, $aligned
184 184 
         ret
185 185 
         .end pix_abs16x16_mvi_asm