variance_halfpixvar16x16_v_media.asm 6.61 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11
;
;  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.
;


Johann's avatar
Johann committed
12
    EXPORT  |vpx_variance_halfpixvar16x16_v_media|
13 14 15 16 17 18 19 20 21 22 23 24

    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
Johann's avatar
Johann committed
25
|vpx_variance_halfpixvar16x16_v_media| PROC
26 27

    stmfd   sp!, {r4-r12, lr}
28 29 30 31

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

32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49
    mov     r8, #0              ; initialize sum = 0
    ldr     r10, c80808080
    mov     r11, #0             ; initialize sse = 0
    mov     r12, #16            ; set loop counter to 16 (=block height)
    mov     lr, #0              ; constant zero
loop
    add     r9, r0, r1          ; set src pointer to next row
    ; 1st 4 pixels
    ldr     r4, [r0, #0]        ; load 4 src pixels
    ldr     r6, [r9, #0]        ; load 4 src pixels from next row
    ldr     r5, [r2, #0]        ; load 4 ref pixels

    ; bilinear interpolation
    mvn     r6, r6
    uhsub8  r4, r4, r6
    eor     r4, r4, r10

    usub8   r6, r4, r5          ; calculate difference
50
    pld     [r0, r1, lsl #1]
51 52
    sel     r7, r6, lr          ; select bytes with positive difference
    usub8   r6, r5, r4          ; calculate difference with reversed operands
53
    pld     [r2, r3, lsl #1]
54 55 56 57 58 59 60 61
    sel     r6, r6, 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
62
    subs    r8, r8, r5          ; subtract negative differences from sum
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

    ; calculate sse
    uxtb16  r5, r6              ; byte (two pixels) to halfwords
    uxtb16  r7, 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     r6, [r9, #4]        ; load 4 src pixels from next row
    ldr     r5, [r2, #4]        ; load 4 ref pixels

    ; bilinear interpolation
    mvn     r6, r6
    uhsub8  r4, r4, r6
    eor     r4, r4, r10

    smlad   r11, r7, r7, r11    ; dual signed multiply, add and accumulate (2)

    usub8   r6, r4, r5          ; calculate difference
    sel     r7, r6, lr          ; select bytes with positive difference
    usub8   r6, r5, r4          ; calculate difference with reversed operands
    sel     r6, r6, 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
93
    sub     r8, r8, r5          ; subtract negative differences from sum
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

    ; calculate sse
    uxtb16  r5, r6              ; byte (two pixels) to halfwords
    uxtb16  r7, 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     r6, [r9, #8]        ; load 4 src pixels from next row
    ldr     r5, [r2, #8]        ; load 4 ref pixels

    ; bilinear interpolation
    mvn     r6, r6
    uhsub8  r4, r4, r6
    eor     r4, r4, r10

    smlad   r11, r7, r7, r11    ; dual signed multiply, add and accumulate (2)

    usub8   r6, r4, r5          ; calculate difference
    sel     r7, r6, lr          ; select bytes with positive difference
    usub8   r6, r5, r4          ; calculate difference with reversed operands
    sel     r6, r6, 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
124
    sub     r8, r8, r5          ; subtract negative differences from sum
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

    ; calculate sse
    uxtb16  r5, r6              ; byte (two pixels) to halfwords
    uxtb16  r7, 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     r6, [r9, #12]       ; load 4 src pixels from next row
    ldr     r5, [r2, #12]       ; load 4 ref pixels

    ; bilinear interpolation
    mvn     r6, r6
    uhsub8  r4, r4, r6
    eor     r4, r4, r10

    smlad   r11, r7, r7, 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   r6, r5, r4          ; calculate difference with reversed operands
    add     r2, r2, r3          ; set dst_ptr to next row
    sel     r6, r6, 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
157
    sub     r8, r8, r5          ; subtract negative differences from sum
158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173

    ; calculate sse
    uxtb16  r5, r6              ; byte (two pixels) to halfwords
    uxtb16  r7, r6, ror #8      ; another two pixels to halfwords
    smlad   r11, r5, r5, r11    ; dual signed multiply, add and accumulate (1)
    smlad   r11, r7, r7, r11    ; dual signed multiply, add and accumulate (2)


    subs    r12, r12, #1

    bne     loop

    ; return stuff
    ldr     r6, [sp, #40]       ; get address of sse
    mul     r0, r8, r8          ; sum * sum
    str     r11, [r6]           ; store sse
Johann's avatar
Johann committed
174
    sub     r0, r11, r0, lsr #8 ; return (sse - ((sum * sum) >> 8))
175 176 177 178 179 180 181 182 183 184

    ldmfd   sp!, {r4-r12, pc}

    ENDP

c80808080
    DCD     0x80808080

    END