variance_media.asm 13.7 KB
Newer Older
Johann's avatar
Johann committed
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
;
;  Copyright (c) 2011 The WebM project authors. All Rights Reserved.
;
;  Use of this source code is governed by a BSD-style license
;  that can be found in the LICENSE file in the root of the source
;  tree. An additional intellectual property rights grant can be found
;  in the file PATENTS.  All contributing project authors may
;  be found in the AUTHORS file in the root of the source tree.
;


    EXPORT  |vpx_variance16x16_media|
    EXPORT  |vpx_variance8x8_media|
    EXPORT  |vpx_mse16x16_media|

    ARM
    REQUIRE8
    PRESERVE8

    AREA ||.text||, CODE, READONLY, ALIGN=2

; r0    unsigned char *src_ptr
; r1    int source_stride
; r2    unsigned char *ref_ptr
; r3    int  recon_stride
; stack unsigned int *sse
|vpx_variance16x16_media| PROC

    stmfd   sp!, {r4-r12, lr}

    pld     [r0, r1, lsl #0]
    pld     [r2, r3, lsl #0]

    mov     r8, #0              ; initialize sum = 0
    mov     r11, #0             ; initialize sse = 0
    mov     r12, #16            ; set loop counter to 16 (=block height)

loop16x16
    ; 1st 4 pixels
    ldr     r4, [r0, #0]        ; load 4 src pixels
    ldr     r5, [r2, #0]        ; load 4 ref pixels

    mov     lr, #0              ; constant zero

    usub8   r6, r4, r5          ; calculate difference
    pld     [r0, r1, lsl #1]
    sel     r7, r6, lr          ; select bytes with positive difference
    usub8   r9, r5, r4          ; calculate difference with reversed operands
    pld     [r2, r3, lsl #1]
    sel     r6, r9, lr          ; select bytes with negative difference

    ; calculate partial sums
    usad8   r4, r7, lr          ; calculate sum of positive differences
    usad8   r5, r6, lr          ; calculate sum of negative differences
    orr     r6, r6, r7          ; differences of all 4 pixels
    ; calculate total sum
    adds    r8, r8, r4          ; add positive differences to sum
    subs    r8, r8, r5          ; subtract negative differences from sum

    ; calculate sse
    uxtb16  r5, r6              ; byte (two pixels) to halfwords
    uxtb16  r10, r6, ror #8     ; another two pixels to halfwords
    smlad   r11, r5, r5, r11    ; dual signed multiply, add and accumulate (1)

    ; 2nd 4 pixels
    ldr     r4, [r0, #4]        ; load 4 src pixels
    ldr     r5, [r2, #4]        ; load 4 ref pixels
    smlad   r11, r10, r10, r11  ; dual signed multiply, add and accumulate (2)

    usub8   r6, r4, r5          ; calculate difference
    sel     r7, r6, lr          ; select bytes with positive difference
    usub8   r9, r5, r4          ; calculate difference with reversed operands
    sel     r6, r9, lr          ; select bytes with negative difference

    ; calculate partial sums
    usad8   r4, r7, lr          ; calculate sum of positive differences
    usad8   r5, r6, lr          ; calculate sum of negative differences
    orr     r6, r6, r7          ; differences of all 4 pixels

    ; calculate total sum
    add     r8, r8, r4          ; add positive differences to sum
    sub     r8, r8, r5          ; subtract negative differences from sum

    ; calculate sse
    uxtb16  r5, r6              ; byte (two pixels) to halfwords
    uxtb16  r10, r6, ror #8     ; another two pixels to halfwords
    smlad   r11, r5, r5, r11    ; dual signed multiply, add and accumulate (1)

    ; 3rd 4 pixels
    ldr     r4, [r0, #8]        ; load 4 src pixels
    ldr     r5, [r2, #8]        ; load 4 ref pixels
    smlad   r11, r10, r10, r11  ; dual signed multiply, add and accumulate (2)

    usub8   r6, r4, r5          ; calculate difference
    sel     r7, r6, lr          ; select bytes with positive difference
    usub8   r9, r5, r4          ; calculate difference with reversed operands
    sel     r6, r9, lr          ; select bytes with negative difference

    ; calculate partial sums
    usad8   r4, r7, lr          ; calculate sum of positive differences
    usad8   r5, r6, lr          ; calculate sum of negative differences
    orr     r6, r6, r7          ; differences of all 4 pixels

    ; calculate total sum
    add     r8, r8, r4          ; add positive differences to sum
    sub     r8, r8, r5          ; subtract negative differences from sum

    ; calculate sse
    uxtb16  r5, r6              ; byte (two pixels) to halfwords
    uxtb16  r10, r6, ror #8     ; another two pixels to halfwords
    smlad   r11, r5, r5, r11    ; dual signed multiply, add and accumulate (1)

    ; 4th 4 pixels
    ldr     r4, [r0, #12]       ; load 4 src pixels
    ldr     r5, [r2, #12]       ; load 4 ref pixels
    smlad   r11, r10, r10, r11  ; dual signed multiply, add and accumulate (2)

    usub8   r6, r4, r5          ; calculate difference
    add     r0, r0, r1          ; set src_ptr to next row
    sel     r7, r6, lr          ; select bytes with positive difference
    usub8   r9, r5, r4          ; calculate difference with reversed operands
    add     r2, r2, r3          ; set dst_ptr to next row
    sel     r6, r9, lr          ; select bytes with negative difference

    ; calculate partial sums
    usad8   r4, r7, lr          ; calculate sum of positive differences
    usad8   r5, r6, lr          ; calculate sum of negative differences
    orr     r6, r6, r7          ; differences of all 4 pixels

    ; calculate total sum
    add     r8, r8, r4          ; add positive differences to sum
    sub     r8, r8, r5          ; subtract negative differences from sum

    ; calculate sse
    uxtb16  r5, r6              ; byte (two pixels) to halfwords
    uxtb16  r10, r6, ror #8     ; another two pixels to halfwords
    smlad   r11, r5, r5, r11    ; dual signed multiply, add and accumulate (1)
    smlad   r11, r10, r10, r11  ; dual signed multiply, add and accumulate (2)


    subs    r12, r12, #1

    bne     loop16x16

    ; return stuff
    ldr     r6, [sp, #40]       ; get address of sse
    mul     r0, r8, r8          ; sum * sum
    str     r11, [r6]           ; store sse
    sub     r0, r11, r0, lsr #8 ; return (sse - ((sum * sum) >> 8))

    ldmfd   sp!, {r4-r12, pc}

    ENDP

; r0    unsigned char *src_ptr
; r1    int source_stride
; r2    unsigned char *ref_ptr
; r3    int  recon_stride
; stack unsigned int *sse
|vpx_variance8x8_media| PROC

    push    {r4-r10, lr}

    pld     [r0, r1, lsl #0]
    pld     [r2, r3, lsl #0]

    mov     r12, #8             ; set loop counter to 8 (=block height)
    mov     r4, #0              ; initialize sum = 0
    mov     r5, #0              ; initialize sse = 0

loop8x8
    ; 1st 4 pixels
    ldr     r6, [r0, #0x0]      ; load 4 src pixels
    ldr     r7, [r2, #0x0]      ; load 4 ref pixels

    mov     lr, #0              ; constant zero

    usub8   r8, r6, r7          ; calculate difference
    pld     [r0, r1, lsl #1]
    sel     r10, r8, lr         ; select bytes with positive difference
    usub8   r9, r7, r6          ; calculate difference with reversed operands
    pld     [r2, r3, lsl #1]
    sel     r8, r9, lr          ; select bytes with negative difference

    ; calculate partial sums
    usad8   r6, r10, lr         ; calculate sum of positive differences
    usad8   r7, r8, lr          ; calculate sum of negative differences
    orr     r8, r8, r10         ; differences of all 4 pixels
    ; calculate total sum
    add    r4, r4, r6           ; add positive differences to sum
    sub    r4, r4, r7           ; subtract negative differences from sum

    ; calculate sse
    uxtb16  r7, r8              ; byte (two pixels) to halfwords
    uxtb16  r10, r8, ror #8     ; another two pixels to halfwords
    smlad   r5, r7, r7, r5      ; dual signed multiply, add and accumulate (1)

    ; 2nd 4 pixels
    ldr     r6, [r0, #0x4]      ; load 4 src pixels
    ldr     r7, [r2, #0x4]      ; load 4 ref pixels
    smlad   r5, r10, r10, r5    ; dual signed multiply, add and accumulate (2)

    usub8   r8, r6, r7          ; calculate difference
    add     r0, r0, r1          ; set src_ptr to next row
    sel     r10, r8, lr         ; select bytes with positive difference
    usub8   r9, r7, r6          ; calculate difference with reversed operands
    add     r2, r2, r3          ; set dst_ptr to next row
    sel     r8, r9, lr          ; select bytes with negative difference

    ; calculate partial sums
    usad8   r6, r10, lr         ; calculate sum of positive differences
    usad8   r7, r8, lr          ; calculate sum of negative differences
    orr     r8, r8, r10         ; differences of all 4 pixels

    ; calculate total sum
    add     r4, r4, r6          ; add positive differences to sum
    sub     r4, r4, r7          ; subtract negative differences from sum

    ; calculate sse
    uxtb16  r7, r8              ; byte (two pixels) to halfwords
    uxtb16  r10, r8, ror #8     ; another two pixels to halfwords
    smlad   r5, r7, r7, r5      ; dual signed multiply, add and accumulate (1)
    subs    r12, r12, #1        ; next row
    smlad   r5, r10, r10, r5    ; dual signed multiply, add and accumulate (2)

    bne     loop8x8

    ; return stuff
    ldr     r8, [sp, #32]       ; get address of sse
    mul     r1, r4, r4          ; sum * sum
    str     r5, [r8]            ; store sse
    sub     r0, r5, r1, ASR #6  ; return (sse - ((sum * sum) >> 6))

    pop     {r4-r10, pc}

    ENDP

; r0    unsigned char *src_ptr
; r1    int source_stride
; r2    unsigned char *ref_ptr
; r3    int  recon_stride
; stack unsigned int *sse
;
;note: Based on vpx_variance16x16_media. In this function, sum is never used.
;      So, we can remove this part of calculation.

|vpx_mse16x16_media| PROC

    push    {r4-r9, lr}

    pld     [r0, r1, lsl #0]
    pld     [r2, r3, lsl #0]

    mov     r12, #16            ; set loop counter to 16 (=block height)
    mov     r4, #0              ; initialize sse = 0

loopmse
    ; 1st 4 pixels
    ldr     r5, [r0, #0x0]      ; load 4 src pixels
    ldr     r6, [r2, #0x0]      ; load 4 ref pixels

    mov     lr, #0              ; constant zero

    usub8   r8, r5, r6          ; calculate difference
    pld     [r0, r1, lsl #1]
    sel     r7, r8, lr          ; select bytes with positive difference
    usub8   r9, r6, r5          ; calculate difference with reversed operands
    pld     [r2, r3, lsl #1]
    sel     r8, r9, lr          ; select bytes with negative difference

    ; calculate partial sums
    usad8   r5, r7, lr          ; calculate sum of positive differences
    usad8   r6, r8, lr          ; calculate sum of negative differences
    orr     r8, r8, r7          ; differences of all 4 pixels

    ldr     r5, [r0, #0x4]      ; load 4 src pixels

    ; calculate sse
    uxtb16  r6, r8              ; byte (two pixels) to halfwords
    uxtb16  r7, r8, ror #8      ; another two pixels to halfwords
    smlad   r4, r6, r6, r4      ; dual signed multiply, add and accumulate (1)

    ; 2nd 4 pixels
    ldr     r6, [r2, #0x4]      ; load 4 ref pixels
    smlad   r4, r7, r7, r4      ; dual signed multiply, add and accumulate (2)

    usub8   r8, r5, r6          ; calculate difference
    sel     r7, r8, lr          ; select bytes with positive difference
    usub8   r9, r6, r5          ; calculate difference with reversed operands
    sel     r8, r9, lr          ; select bytes with negative difference

    ; calculate partial sums
    usad8   r5, r7, lr          ; calculate sum of positive differences
    usad8   r6, r8, lr          ; calculate sum of negative differences
    orr     r8, r8, r7          ; differences of all 4 pixels
    ldr     r5, [r0, #0x8]      ; load 4 src pixels
    ; calculate sse
    uxtb16  r6, r8              ; byte (two pixels) to halfwords
    uxtb16  r7, r8, ror #8      ; another two pixels to halfwords
    smlad   r4, r6, r6, r4      ; dual signed multiply, add and accumulate (1)

    ; 3rd 4 pixels
    ldr     r6, [r2, #0x8]      ; load 4 ref pixels
    smlad   r4, r7, r7, r4      ; dual signed multiply, add and accumulate (2)

    usub8   r8, r5, r6          ; calculate difference
    sel     r7, r8, lr          ; select bytes with positive difference
    usub8   r9, r6, r5          ; calculate difference with reversed operands
    sel     r8, r9, lr          ; select bytes with negative difference

    ; calculate partial sums
    usad8   r5, r7, lr          ; calculate sum of positive differences
    usad8   r6, r8, lr          ; calculate sum of negative differences
    orr     r8, r8, r7          ; differences of all 4 pixels

    ldr     r5, [r0, #0xc]      ; load 4 src pixels

    ; calculate sse
    uxtb16  r6, r8              ; byte (two pixels) to halfwords
    uxtb16  r7, r8, ror #8      ; another two pixels to halfwords
    smlad   r4, r6, r6, r4      ; dual signed multiply, add and accumulate (1)

    ; 4th 4 pixels
    ldr     r6, [r2, #0xc]      ; load 4 ref pixels
    smlad   r4, r7, r7, r4      ; dual signed multiply, add and accumulate (2)

    usub8   r8, r5, r6          ; calculate difference
    add     r0, r0, r1          ; set src_ptr to next row
    sel     r7, r8, lr          ; select bytes with positive difference
    usub8   r9, r6, r5          ; calculate difference with reversed operands
    add     r2, r2, r3          ; set dst_ptr to next row
    sel     r8, r9, lr          ; select bytes with negative difference

    ; calculate partial sums
    usad8   r5, r7, lr          ; calculate sum of positive differences
    usad8   r6, r8, lr          ; calculate sum of negative differences
    orr     r8, r8, r7          ; differences of all 4 pixels

    subs    r12, r12, #1        ; next row

    ; calculate sse
    uxtb16  r6, r8              ; byte (two pixels) to halfwords
    uxtb16  r7, r8, ror #8      ; another two pixels to halfwords
    smlad   r4, r6, r6, r4      ; dual signed multiply, add and accumulate (1)
    smlad   r4, r7, r7, r4      ; dual signed multiply, add and accumulate (2)

    bne     loopmse

    ; return stuff
    ldr     r1, [sp, #28]       ; get address of sse
    mov     r0, r4              ; return sse
    str     r4, [r1]            ; store sse

    pop     {r4-r9, pc}

    ENDP

    END