1;
2;  Copyright (c) 2011 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
12    EXPORT  |vp8_fast_quantize_b_armv6|
13
14    INCLUDE asm_enc_offsets.asm
15
16    ARM
17    REQUIRE8
18    PRESERVE8
19
20    AREA ||.text||, CODE, READONLY, ALIGN=2
21
22; r0    BLOCK *b
23; r1    BLOCKD *d
24|vp8_fast_quantize_b_armv6| PROC
25    stmfd   sp!, {r1, r4-r11, lr}
26
27    ldr     r3, [r0, #vp8_block_coeff]      ; coeff
28    ldr     r4, [r0, #vp8_block_quant_fast] ; quant_fast
29    ldr     r5, [r0, #vp8_block_round]      ; round
30    ldr     r6, [r1, #vp8_blockd_qcoeff]    ; qcoeff
31    ldr     r7, [r1, #vp8_blockd_dqcoeff]   ; dqcoeff
32    ldr     r8, [r1, #vp8_blockd_dequant]   ; dequant
33
34    ldr     r2, loop_count          ; loop_count=0x1000000. 'lsls' instruction
35                                    ; is used to update the counter so that
36                                    ; it can be used to mark nonzero
37                                    ; quantized coefficient pairs.
38
39    mov     r1, #0                  ; flags for quantized coeffs
40
41    ; PART 1: quantization and dequantization loop
42loop
43    ldr     r9, [r3], #4            ; [z1 | z0]
44    ldr     r10, [r5], #4           ; [r1 | r0]
45    ldr     r11, [r4], #4           ; [q1 | q0]
46
47    ssat16  lr, #1, r9              ; [sz1 | sz0]
48    eor     r9, r9, lr              ; [z1 ^ sz1 | z0 ^ sz0]
49    ssub16  r9, r9, lr              ; x = (z ^ sz) - sz
50    sadd16  r9, r9, r10             ; [x1+r1 | x0+r0]
51
52    ldr     r12, [r3], #4           ; [z3 | z2]
53
54    smulbb  r0, r9, r11             ; [(x0+r0)*q0]
55    smultt  r9, r9, r11             ; [(x1+r1)*q1]
56
57    ldr     r10, [r5], #4           ; [r3 | r2]
58
59    ssat16  r11, #1, r12            ; [sz3 | sz2]
60    eor     r12, r12, r11           ; [z3 ^ sz3 | z2 ^ sz2]
61    pkhtb   r0, r9, r0, asr #16     ; [y1 | y0]
62    ldr     r9, [r4], #4            ; [q3 | q2]
63    ssub16  r12, r12, r11           ; x = (z ^ sz) - sz
64
65    sadd16  r12, r12, r10           ; [x3+r3 | x2+r2]
66
67    eor     r0, r0, lr              ; [(y1 ^ sz1) | (y0 ^ sz0)]
68
69    smulbb  r10, r12, r9            ; [(x2+r2)*q2]
70    smultt  r12, r12, r9            ; [(x3+r3)*q3]
71
72    ssub16  r0, r0, lr              ; x = (y ^ sz) - sz
73
74    cmp     r0, #0                  ; check if zero
75    orrne   r1, r1, r2, lsr #24     ; add flag for nonzero coeffs
76
77    str     r0, [r6], #4            ; *qcoeff++ = x
78    ldr     r9, [r8], #4            ; [dq1 | dq0]
79
80    pkhtb   r10, r12, r10, asr #16  ; [y3 | y2]
81    eor     r10, r10, r11           ; [(y3 ^ sz3) | (y2 ^ sz2)]
82    ssub16  r10, r10, r11           ; x = (y ^ sz) - sz
83
84    cmp     r10, #0                 ; check if zero
85    orrne   r1, r1, r2, lsr #23     ; add flag for nonzero coeffs
86
87    str     r10, [r6], #4           ; *qcoeff++ = x
88    ldr     r11, [r8], #4           ; [dq3 | dq2]
89
90    smulbb  r12, r0, r9             ; [x0*dq0]
91    smultt  r0, r0, r9              ; [x1*dq1]
92
93    smulbb  r9, r10, r11            ; [x2*dq2]
94    smultt  r10, r10, r11           ; [x3*dq3]
95
96    lsls    r2, r2, #2              ; update loop counter
97    strh    r12, [r7, #0]           ; dqcoeff[0] = [x0*dq0]
98    strh    r0, [r7, #2]            ; dqcoeff[1] = [x1*dq1]
99    strh    r9, [r7, #4]            ; dqcoeff[2] = [x2*dq2]
100    strh    r10, [r7, #6]           ; dqcoeff[3] = [x3*dq3]
101    add     r7, r7, #8              ; dqcoeff += 8
102    bne     loop
103
104    ; PART 2: check position for eob...
105    mov     lr, #0                  ; init eob
106    cmp     r1, #0                  ; coeffs after quantization?
107    ldr     r11, [sp, #0]           ; restore BLOCKD pointer
108    beq     end                     ; skip eob calculations if all zero
109
110    ldr     r0, [r11, #vp8_blockd_qcoeff]
111
112    ; check shortcut for nonzero qcoeffs
113    tst    r1, #0x80
114    bne    quant_coeff_15_14
115    tst    r1, #0x20
116    bne    quant_coeff_13_11
117    tst    r1, #0x8
118    bne    quant_coeff_12_7
119    tst    r1, #0x40
120    bne    quant_coeff_10_9
121    tst    r1, #0x10
122    bne    quant_coeff_8_3
123    tst    r1, #0x2
124    bne    quant_coeff_6_5
125    tst    r1, #0x4
126    bne    quant_coeff_4_2
127    b      quant_coeff_1_0
128
129quant_coeff_15_14
130    ldrh    r2, [r0, #30]       ; rc=15, i=15
131    mov     lr, #16
132    cmp     r2, #0
133    bne     end
134
135    ldrh    r3, [r0, #28]       ; rc=14, i=14
136    mov     lr, #15
137    cmp     r3, #0
138    bne     end
139
140quant_coeff_13_11
141    ldrh    r2, [r0, #22]       ; rc=11, i=13
142    mov     lr, #14
143    cmp     r2, #0
144    bne     end
145
146quant_coeff_12_7
147    ldrh    r3, [r0, #14]       ; rc=7,  i=12
148    mov     lr, #13
149    cmp     r3, #0
150    bne     end
151
152    ldrh    r2, [r0, #20]       ; rc=10, i=11
153    mov     lr, #12
154    cmp     r2, #0
155    bne     end
156
157quant_coeff_10_9
158    ldrh    r3, [r0, #26]       ; rc=13, i=10
159    mov     lr, #11
160    cmp     r3, #0
161    bne     end
162
163    ldrh    r2, [r0, #24]       ; rc=12, i=9
164    mov     lr, #10
165    cmp     r2, #0
166    bne     end
167
168quant_coeff_8_3
169    ldrh    r3, [r0, #18]       ; rc=9,  i=8
170    mov     lr, #9
171    cmp     r3, #0
172    bne     end
173
174    ldrh    r2, [r0, #12]       ; rc=6,  i=7
175    mov     lr, #8
176    cmp     r2, #0
177    bne     end
178
179quant_coeff_6_5
180    ldrh    r3, [r0, #6]        ; rc=3,  i=6
181    mov     lr, #7
182    cmp     r3, #0
183    bne     end
184
185    ldrh    r2, [r0, #4]        ; rc=2,  i=5
186    mov     lr, #6
187    cmp     r2, #0
188    bne     end
189
190quant_coeff_4_2
191    ldrh    r3, [r0, #10]       ; rc=5,  i=4
192    mov     lr, #5
193    cmp     r3, #0
194    bne     end
195
196    ldrh    r2, [r0, #16]       ; rc=8,  i=3
197    mov     lr, #4
198    cmp     r2, #0
199    bne     end
200
201    ldrh    r3, [r0, #8]        ; rc=4,  i=2
202    mov     lr, #3
203    cmp     r3, #0
204    bne     end
205
206quant_coeff_1_0
207    ldrh    r2, [r0, #2]        ; rc=1,  i=1
208    mov     lr, #2
209    cmp     r2, #0
210    bne     end
211
212    mov     lr, #1              ; rc=0,  i=0
213
214end
215    str     lr, [r11, #vp8_blockd_eob]
216    ldmfd   sp!, {r1, r4-r11, pc}
217
218    ENDP
219
220loop_count
221    DCD     0x1000000
222
223    END
224
225