1.1 --- /dev/null Thu Jan 01 00:00:00 1970 +0000
1.2 +++ b/media/libtheora/lib/arm/armidct.s Wed Dec 31 06:09:35 2014 +0100
1.3 @@ -0,0 +1,1914 @@
1.4 +;********************************************************************
1.5 +;* *
1.6 +;* THIS FILE IS PART OF THE OggTheora SOFTWARE CODEC SOURCE CODE. *
1.7 +;* USE, DISTRIBUTION AND REPRODUCTION OF THIS LIBRARY SOURCE IS *
1.8 +;* GOVERNED BY A BSD-STYLE SOURCE LICENSE INCLUDED WITH THIS SOURCE *
1.9 +;* IN 'COPYING'. PLEASE READ THESE TERMS BEFORE DISTRIBUTING. *
1.10 +;* *
1.11 +;* THE Theora SOURCE CODE IS COPYRIGHT (C) 2002-2010 *
1.12 +;* by the Xiph.Org Foundation and contributors http://www.xiph.org/ *
1.13 +;* *
1.14 +;********************************************************************
1.15 +; Original implementation:
1.16 +; Copyright (C) 2009 Robin Watts for Pinknoise Productions Ltd
1.17 +; last mod: $Id: armidct.s 17481 2010-10-03 22:49:42Z tterribe $
1.18 +;********************************************************************
1.19 +
1.20 + AREA |.text|, CODE, READONLY
1.21 +
1.22 + ; Explicitly specifying alignment here because some versions of
1.23 + ; gas don't align code correctly. See
1.24 + ; http://lists.gnu.org/archive/html/bug-binutils/2011-06/msg00199.html
1.25 + ; https://bugzilla.mozilla.org/show_bug.cgi?id=920992
1.26 + ALIGN
1.27 +
1.28 + GET armopts.s
1.29 +
1.30 + EXPORT oc_idct8x8_1_arm
1.31 + EXPORT oc_idct8x8_arm
1.32 +
1.33 +oc_idct8x8_1_arm PROC
1.34 + ; r0 = ogg_int16_t *_y
1.35 + ; r1 = ogg_uint16_t _dc
1.36 + ORR r1, r1, r1, LSL #16
1.37 + MOV r2, r1
1.38 + MOV r3, r1
1.39 + MOV r12,r1
1.40 + STMIA r0!,{r1,r2,r3,r12}
1.41 + STMIA r0!,{r1,r2,r3,r12}
1.42 + STMIA r0!,{r1,r2,r3,r12}
1.43 + STMIA r0!,{r1,r2,r3,r12}
1.44 + STMIA r0!,{r1,r2,r3,r12}
1.45 + STMIA r0!,{r1,r2,r3,r12}
1.46 + STMIA r0!,{r1,r2,r3,r12}
1.47 + STMIA r0!,{r1,r2,r3,r12}
1.48 + MOV PC, r14
1.49 + ENDP
1.50 +
1.51 +oc_idct8x8_arm PROC
1.52 + ; r0 = ogg_int16_t *_y
1.53 + ; r1 = ogg_int16_t *_x
1.54 + ; r2 = int _last_zzi
1.55 + CMP r2, #3
1.56 + BLE oc_idct8x8_3_arm
1.57 + CMP r2, #6
1.58 + BLE oc_idct8x8_6_arm
1.59 + CMP r2, #10
1.60 + BLE oc_idct8x8_10_arm
1.61 +oc_idct8x8_slow_arm
1.62 + STMFD r13!,{r4-r11,r14}
1.63 + SUB r13,r13,#64*2
1.64 +; Row transforms
1.65 + STR r0, [r13,#-4]!
1.66 + ADD r0, r13, #4 ; Write to temp storage.
1.67 + BL idct8core_arm
1.68 + BL idct8core_arm
1.69 + BL idct8core_arm
1.70 + BL idct8core_arm
1.71 + BL idct8core_arm
1.72 + BL idct8core_arm
1.73 + BL idct8core_arm
1.74 + BL idct8core_arm
1.75 + LDR r0, [r13], #4 ; Write to the final destination.
1.76 + ; Clear input data for next block (decoder only).
1.77 + SUB r2, r1, #8*16
1.78 + CMP r0, r2
1.79 + MOV r1, r13 ; And read from temp storage.
1.80 + BEQ oc_idct8x8_slow_arm_cols
1.81 + MOV r4, #0
1.82 + MOV r5, #0
1.83 + MOV r6, #0
1.84 + MOV r7, #0
1.85 + STMIA r2!,{r4,r5,r6,r7}
1.86 + STMIA r2!,{r4,r5,r6,r7}
1.87 + STMIA r2!,{r4,r5,r6,r7}
1.88 + STMIA r2!,{r4,r5,r6,r7}
1.89 + STMIA r2!,{r4,r5,r6,r7}
1.90 + STMIA r2!,{r4,r5,r6,r7}
1.91 + STMIA r2!,{r4,r5,r6,r7}
1.92 + STMIA r2!,{r4,r5,r6,r7}
1.93 +oc_idct8x8_slow_arm_cols
1.94 +; Column transforms
1.95 + BL idct8core_down_arm
1.96 + BL idct8core_down_arm
1.97 + BL idct8core_down_arm
1.98 + BL idct8core_down_arm
1.99 + BL idct8core_down_arm
1.100 + BL idct8core_down_arm
1.101 + BL idct8core_down_arm
1.102 + BL idct8core_down_arm
1.103 + ADD r13,r13,#64*2
1.104 + LDMFD r13!,{r4-r11,PC}
1.105 + ENDP
1.106 +
1.107 +oc_idct8x8_10_arm PROC
1.108 + STMFD r13!,{r4-r11,r14}
1.109 + SUB r13,r13,#64*2
1.110 +; Row transforms
1.111 + MOV r2, r0
1.112 + MOV r0, r13 ; Write to temp storage.
1.113 + BL idct4core_arm
1.114 + BL idct3core_arm
1.115 + BL idct2core_arm
1.116 + BL idct1core_arm
1.117 + ; Clear input data for next block (decoder only).
1.118 + SUB r0, r1, #4*16
1.119 + CMP r0, r2
1.120 + MOV r1, r13 ; Read from temp storage.
1.121 + BEQ oc_idct8x8_10_arm_cols
1.122 + MOV r4, #0
1.123 + STR r4, [r0]
1.124 + STR r4, [r0,#4]
1.125 + STR r4, [r0,#16]
1.126 + STR r4, [r0,#20]
1.127 + STR r4, [r0,#32]
1.128 + STR r4, [r0,#48]
1.129 + MOV r0, r2 ; Write to the final destination
1.130 +oc_idct8x8_10_arm_cols
1.131 +; Column transforms
1.132 + BL idct4core_down_arm
1.133 + BL idct4core_down_arm
1.134 + BL idct4core_down_arm
1.135 + BL idct4core_down_arm
1.136 + BL idct4core_down_arm
1.137 + BL idct4core_down_arm
1.138 + BL idct4core_down_arm
1.139 + BL idct4core_down_arm
1.140 + ADD r13,r13,#64*2
1.141 + LDMFD r13!,{r4-r11,PC}
1.142 + ENDP
1.143 +
1.144 +oc_idct8x8_6_arm PROC
1.145 + STMFD r13!,{r4-r7,r9-r11,r14}
1.146 + SUB r13,r13,#64*2
1.147 +; Row transforms
1.148 + MOV r2, r0
1.149 + MOV r0, r13 ; Write to temp storage.
1.150 + BL idct3core_arm
1.151 + BL idct2core_arm
1.152 + BL idct1core_arm
1.153 + ; Clear input data for next block (decoder only).
1.154 + SUB r0, r1, #3*16
1.155 + CMP r0, r2
1.156 + MOV r1, r13 ; Read from temp storage.
1.157 + BEQ oc_idct8x8_6_arm_cols
1.158 + MOV r4, #0
1.159 + STR r4, [r0]
1.160 + STR r4, [r0,#4]
1.161 + STR r4, [r0,#16]
1.162 + STR r4, [r0,#32]
1.163 + MOV r0, r2 ; Write to the final destination
1.164 +oc_idct8x8_6_arm_cols
1.165 +; Column transforms
1.166 + BL idct3core_down_arm
1.167 + BL idct3core_down_arm
1.168 + BL idct3core_down_arm
1.169 + BL idct3core_down_arm
1.170 + BL idct3core_down_arm
1.171 + BL idct3core_down_arm
1.172 + BL idct3core_down_arm
1.173 + BL idct3core_down_arm
1.174 + ADD r13,r13,#64*2
1.175 + LDMFD r13!,{r4-r7,r9-r11,PC}
1.176 + ENDP
1.177 +
1.178 +oc_idct8x8_3_arm PROC
1.179 + STMFD r13!,{r4-r7,r9-r11,r14}
1.180 + SUB r13,r13,#64*2
1.181 +; Row transforms
1.182 + MOV r2, r0
1.183 + MOV r0, r13 ; Write to temp storage.
1.184 + BL idct2core_arm
1.185 + BL idct1core_arm
1.186 + ; Clear input data for next block (decoder only).
1.187 + SUB r0, r1, #2*16
1.188 + CMP r0, r2
1.189 + MOV r1, r13 ; Read from temp storage.
1.190 + MOVNE r4, #0
1.191 + STRNE r4, [r0]
1.192 + STRNE r4, [r0,#16]
1.193 + MOVNE r0, r2 ; Write to the final destination
1.194 +; Column transforms
1.195 + BL idct2core_down_arm
1.196 + BL idct2core_down_arm
1.197 + BL idct2core_down_arm
1.198 + BL idct2core_down_arm
1.199 + BL idct2core_down_arm
1.200 + BL idct2core_down_arm
1.201 + BL idct2core_down_arm
1.202 + BL idct2core_down_arm
1.203 + ADD r13,r13,#64*2
1.204 + LDMFD r13!,{r4-r7,r9-r11,PC}
1.205 + ENDP
1.206 +
1.207 +idct1core_arm PROC
1.208 + ; r0 = ogg_int16_t *_y (destination)
1.209 + ; r1 = const ogg_int16_t *_x (source)
1.210 + LDRSH r3, [r1], #16
1.211 + MOV r12,#0x05
1.212 + ORR r12,r12,#0xB500
1.213 + MUL r3, r12, r3
1.214 + ; Stall ?
1.215 + MOV r3, r3, ASR #16
1.216 + STRH r3, [r0], #2
1.217 + STRH r3, [r0, #14]
1.218 + STRH r3, [r0, #30]
1.219 + STRH r3, [r0, #46]
1.220 + STRH r3, [r0, #62]
1.221 + STRH r3, [r0, #78]
1.222 + STRH r3, [r0, #94]
1.223 + STRH r3, [r0, #110]
1.224 + MOV PC,R14
1.225 + ENDP
1.226 +
1.227 +idct2core_arm PROC
1.228 + ; r0 = ogg_int16_t *_y (destination)
1.229 + ; r1 = const ogg_int16_t *_x (source)
1.230 + LDRSH r9, [r1], #16 ; r9 = x[0]
1.231 + LDR r12,OC_C4S4
1.232 + LDRSH r11,[r1, #-14] ; r11= x[1]
1.233 + LDR r3, OC_C7S1
1.234 + MUL r9, r12,r9 ; r9 = t[0]<<16 = OC_C4S4*x[0]
1.235 + LDR r10,OC_C1S7
1.236 + MUL r3, r11,r3 ; r3 = t[4]<<16 = OC_C7S1*x[1]
1.237 + MOV r9, r9, ASR #16 ; r9 = t[0]
1.238 + MUL r11,r10,r11 ; r11= t[7]<<16 = OC_C1S7*x[1]
1.239 + MOV r3, r3, ASR #16 ; r3 = t[4]
1.240 + MUL r10,r12,r3 ; r10= t[5]<<16 = OC_C4S4*t[4]
1.241 + MOV r11,r11,ASR #16 ; r11= t[7]
1.242 + MUL r12,r11,r12 ; r12= t[6]<<16 = OC_C4S4*t[7]
1.243 + MOV r10,r10,ASR #16 ; r10= t[5]
1.244 + ADD r12,r9,r12,ASR #16 ; r12= t[0]+t[6]
1.245 + ADD r12,r12,r10 ; r12= t[0]+t2[6] = t[0]+t[6]+t[5]
1.246 + SUB r10,r12,r10,LSL #1 ; r10= t[0]+t2[5] = t[0]+t[6]-t[5]
1.247 + ADD r3, r3, r9 ; r3 = t[0]+t[4]
1.248 + ADD r11,r11,r9 ; r11= t[0]+t[7]
1.249 + STRH r11,[r0], #2 ; y[0] = t[0]+t[7]
1.250 + STRH r12,[r0, #14] ; y[1] = t[0]+t[6]
1.251 + STRH r10,[r0, #30] ; y[2] = t[0]+t[5]
1.252 + STRH r3, [r0, #46] ; y[3] = t[0]+t[4]
1.253 + RSB r3, r3, r9, LSL #1 ; r3 = t[0]*2-(t[0]+t[4])=t[0]-t[4]
1.254 + RSB r10,r10,r9, LSL #1 ; r10= t[0]*2-(t[0]+t[5])=t[0]-t[5]
1.255 + RSB r12,r12,r9, LSL #1 ; r12= t[0]*2-(t[0]+t[6])=t[0]-t[6]
1.256 + RSB r11,r11,r9, LSL #1 ; r1 = t[0]*2-(t[0]+t[7])=t[0]-t[7]
1.257 + STRH r3, [r0, #62] ; y[4] = t[0]-t[4]
1.258 + STRH r10,[r0, #78] ; y[5] = t[0]-t[5]
1.259 + STRH r12,[r0, #94] ; y[6] = t[0]-t[6]
1.260 + STRH r11,[r0, #110] ; y[7] = t[0]-t[7]
1.261 + MOV PC,r14
1.262 + ENDP
1.263 +
1.264 +idct2core_down_arm PROC
1.265 + ; r0 = ogg_int16_t *_y (destination)
1.266 + ; r1 = const ogg_int16_t *_x (source)
1.267 + LDRSH r9, [r1], #16 ; r9 = x[0]
1.268 + LDR r12,OC_C4S4
1.269 + LDRSH r11,[r1, #-14] ; r11= x[1]
1.270 + LDR r3, OC_C7S1
1.271 + MUL r9, r12,r9 ; r9 = t[0]<<16 = OC_C4S4*x[0]
1.272 + LDR r10,OC_C1S7
1.273 + MUL r3, r11,r3 ; r3 = t[4]<<16 = OC_C7S1*x[1]
1.274 + MOV r9, r9, ASR #16 ; r9 = t[0]
1.275 + MUL r11,r10,r11 ; r11= t[7]<<16 = OC_C1S7*x[1]
1.276 + ADD r9, r9, #8 ; r9 = t[0]+8
1.277 + MOV r3, r3, ASR #16 ; r3 = t[4]
1.278 + MUL r10,r12,r3 ; r10= t[5]<<16 = OC_C4S4*t[4]
1.279 + MOV r11,r11,ASR #16 ; r11= t[7]
1.280 + MUL r12,r11,r12 ; r12= t[6]<<16 = OC_C4S4*t[7]
1.281 + MOV r10,r10,ASR #16 ; r10= t[5]
1.282 + ADD r12,r9,r12,ASR #16 ; r12= t[0]+t[6]+8
1.283 + ADD r12,r12,r10 ; r12= t[0]+t2[6] = t[0]+t[6]+t[5]+8
1.284 + SUB r10,r12,r10,LSL #1 ; r10= t[0]+t2[5] = t[0]+t[6]-t[5]+8
1.285 + ADD r3, r3, r9 ; r3 = t[0]+t[4]+8
1.286 + ADD r11,r11,r9 ; r11= t[0]+t[7]+8
1.287 + ; TODO: This is wrong.
1.288 + ; The C code truncates to 16 bits by storing to RAM and doing the
1.289 + ; shifts later; we've got an extra 4 bits here.
1.290 + MOV r4, r11,ASR #4
1.291 + MOV r5, r12,ASR #4
1.292 + MOV r6, r10,ASR #4
1.293 + MOV r7, r3, ASR #4
1.294 + RSB r3, r3, r9, LSL #1 ;r3 =t[0]*2+8-(t[0]+t[4])=t[0]-t[4]+8
1.295 + RSB r10,r10,r9, LSL #1 ;r10=t[0]*2+8-(t[0]+t[5])=t[0]-t[5]+8
1.296 + RSB r12,r12,r9, LSL #1 ;r12=t[0]*2+8-(t[0]+t[6])=t[0]-t[6]+8
1.297 + RSB r11,r11,r9, LSL #1 ;r11=t[0]*2+8-(t[0]+t[7])=t[0]-t[7]+8
1.298 + MOV r3, r3, ASR #4
1.299 + MOV r10,r10,ASR #4
1.300 + MOV r12,r12,ASR #4
1.301 + MOV r11,r11,ASR #4
1.302 + STRH r4, [r0], #2 ; y[0] = t[0]+t[7]
1.303 + STRH r5, [r0, #14] ; y[1] = t[0]+t[6]
1.304 + STRH r6, [r0, #30] ; y[2] = t[0]+t[5]
1.305 + STRH r7, [r0, #46] ; y[3] = t[0]+t[4]
1.306 + STRH r3, [r0, #62] ; y[4] = t[0]-t[4]
1.307 + STRH r10,[r0, #78] ; y[5] = t[0]-t[5]
1.308 + STRH r12,[r0, #94] ; y[6] = t[0]-t[6]
1.309 + STRH r11,[r0, #110] ; y[7] = t[0]-t[7]
1.310 + MOV PC,r14
1.311 + ENDP
1.312 +
1.313 +idct3core_arm PROC
1.314 + LDRSH r9, [r1], #16 ; r9 = x[0]
1.315 + LDR r12,OC_C4S4 ; r12= OC_C4S4
1.316 + LDRSH r3, [r1, #-12] ; r3 = x[2]
1.317 + LDR r10,OC_C6S2 ; r10= OC_C6S2
1.318 + MUL r9, r12,r9 ; r9 = t[0]<<16 = OC_C4S4*x[0]
1.319 + LDR r4, OC_C2S6 ; r4 = OC_C2S6
1.320 + MUL r10,r3, r10 ; r10= t[2]<<16 = OC_C6S2*x[2]
1.321 + LDRSH r11,[r1, #-14] ; r11= x[1]
1.322 + MUL r3, r4, r3 ; r3 = t[3]<<16 = OC_C2S6*x[2]
1.323 + LDR r4, OC_C7S1 ; r4 = OC_C7S1
1.324 + LDR r5, OC_C1S7 ; r5 = OC_C1S7
1.325 + MOV r9, r9, ASR #16 ; r9 = t[0]
1.326 + MUL r4, r11,r4 ; r4 = t[4]<<16 = OC_C7S1*x[1]
1.327 + ADD r3, r9, r3, ASR #16 ; r3 = t[0]+t[3]
1.328 + MUL r11,r5, r11 ; r11= t[7]<<16 = OC_C1S7*x[1]
1.329 + MOV r4, r4, ASR #16 ; r4 = t[4]
1.330 + MUL r5, r12,r4 ; r5 = t[5]<<16 = OC_C4S4*t[4]
1.331 + MOV r11,r11,ASR #16 ; r11= t[7]
1.332 + MUL r12,r11,r12 ; r12= t[6]<<16 = OC_C4S4*t[7]
1.333 + ADD r10,r9, r10,ASR #16 ; r10= t[1] = t[0]+t[2]
1.334 + RSB r6, r10,r9, LSL #1 ; r6 = t[2] = t[0]-t[2]
1.335 + ; r3 = t2[0] = t[0]+t[3]
1.336 + RSB r9, r3, r9, LSL #1 ; r9 = t2[3] = t[0]-t[3]
1.337 + MOV r12,r12,ASR #16 ; r12= t[6]
1.338 + ADD r5, r12,r5, ASR #16 ; r5 = t2[6] = t[6]+t[5]
1.339 + RSB r12,r5, r12,LSL #1 ; r12= t2[5] = t[6]-t[5]
1.340 + ADD r11,r3, r11 ; r11= t2[0]+t[7]
1.341 + ADD r5, r10,r5 ; r5 = t[1]+t2[6]
1.342 + ADD r12,r6, r12 ; r12= t[2]+t2[5]
1.343 + ADD r4, r9, r4 ; r4 = t2[3]+t[4]
1.344 + STRH r11,[r0], #2 ; y[0] = t[0]+t[7]
1.345 + STRH r5, [r0, #14] ; y[1] = t[1]+t2[6]
1.346 + STRH r12,[r0, #30] ; y[2] = t[2]+t2[5]
1.347 + STRH r4, [r0, #46] ; y[3] = t2[3]+t[4]
1.348 + RSB r11,r11,r3, LSL #1 ; r11= t2[0] - t[7]
1.349 + RSB r5, r5, r10,LSL #1 ; r5 = t[1] - t2[6]
1.350 + RSB r12,r12,r6, LSL #1 ; r6 = t[2] - t2[5]
1.351 + RSB r4, r4, r9, LSL #1 ; r4 = t2[3] - t[4]
1.352 + STRH r4, [r0, #62] ; y[4] = t2[3]-t[4]
1.353 + STRH r12,[r0, #78] ; y[5] = t[2]-t2[5]
1.354 + STRH r5, [r0, #94] ; y[6] = t[1]-t2[6]
1.355 + STRH r11,[r0, #110] ; y[7] = t2[0]-t[7]
1.356 + MOV PC,R14
1.357 + ENDP
1.358 +
1.359 +idct3core_down_arm PROC
1.360 + LDRSH r9, [r1], #16 ; r9 = x[0]
1.361 + LDR r12,OC_C4S4 ; r12= OC_C4S4
1.362 + LDRSH r3, [r1, #-12] ; r3 = x[2]
1.363 + LDR r10,OC_C6S2 ; r10= OC_C6S2
1.364 + MUL r9, r12,r9 ; r9 = t[0]<<16 = OC_C4S4*x[0]
1.365 + LDR r4, OC_C2S6 ; r4 = OC_C2S6
1.366 + MUL r10,r3, r10 ; r10= t[2]<<16 = OC_C6S2*x[2]
1.367 + LDRSH r11,[r1, #-14] ; r11= x[1]
1.368 + MUL r3, r4, r3 ; r3 = t[3]<<16 = OC_C2S6*x[2]
1.369 + LDR r4, OC_C7S1 ; r4 = OC_C7S1
1.370 + LDR r5, OC_C1S7 ; r5 = OC_C1S7
1.371 + MOV r9, r9, ASR #16 ; r9 = t[0]
1.372 + MUL r4, r11,r4 ; r4 = t[4]<<16 = OC_C7S1*x[1]
1.373 + ADD r9, r9, #8 ; r9 = t[0]+8
1.374 + MUL r11,r5, r11 ; r11= t[7]<<16 = OC_C1S7*x[1]
1.375 + ADD r3, r9, r3, ASR #16 ; r3 = t[0]+t[3]+8
1.376 + MOV r4, r4, ASR #16 ; r4 = t[4]
1.377 + MUL r5, r12,r4 ; r5 = t[5]<<16 = OC_C4S4*t[4]
1.378 + MOV r11,r11,ASR #16 ; r11= t[7]
1.379 + MUL r12,r11,r12 ; r12= t[6]<<16 = OC_C4S4*t[7]
1.380 + ADD r10,r9, r10,ASR #16 ; r10= t[1]+8 = t[0]+t[2]+8
1.381 + RSB r6, r10,r9, LSL #1 ; r6 = t[2]+8 = t[0]-t[2]+8
1.382 + ; r3 = t2[0]+8 = t[0]+t[3]+8
1.383 + RSB r9, r3, r9, LSL #1 ; r9 = t2[3]+8 = t[0]-t[3]+8
1.384 + MOV r12,r12,ASR #16 ; r12= t[6]
1.385 + ADD r5, r12,r5, ASR #16 ; r5 = t2[6] = t[6]+t[5]
1.386 + RSB r12,r5, r12,LSL #1 ; r12= t2[5] = t[6]-t[5]
1.387 + ADD r11,r3, r11 ; r11= t2[0]+t[7] +8
1.388 + ADD r5, r10,r5 ; r5 = t[1] +t2[6]+8
1.389 + ADD r12,r6, r12 ; r12= t[2] +t2[5]+8
1.390 + ADD r4, r9, r4 ; r4 = t2[3]+t[4] +8
1.391 + RSB r3, r11,r3, LSL #1 ; r11= t2[0] - t[7] + 8
1.392 + RSB r10,r5, r10,LSL #1 ; r5 = t[1] - t2[6] + 8
1.393 + RSB r6, r12,r6, LSL #1 ; r6 = t[2] - t2[5] + 8
1.394 + RSB r9, r4, r9, LSL #1 ; r4 = t2[3] - t[4] + 8
1.395 + ; TODO: This is wrong.
1.396 + ; The C code truncates to 16 bits by storing to RAM and doing the
1.397 + ; shifts later; we've got an extra 4 bits here.
1.398 + MOV r11,r11,ASR #4
1.399 + MOV r5, r5, ASR #4
1.400 + MOV r12,r12,ASR #4
1.401 + MOV r4, r4, ASR #4
1.402 + MOV r9, r9, ASR #4
1.403 + MOV r6, r6, ASR #4
1.404 + MOV r10,r10,ASR #4
1.405 + MOV r3, r3, ASR #4
1.406 + STRH r11,[r0], #2 ; y[0] = t[0]+t[7]
1.407 + STRH r5, [r0, #14] ; y[1] = t[1]+t2[6]
1.408 + STRH r12,[r0, #30] ; y[2] = t[2]+t2[5]
1.409 + STRH r4, [r0, #46] ; y[3] = t2[3]+t[4]
1.410 + STRH r9, [r0, #62] ; y[4] = t2[3]-t[4]
1.411 + STRH r6, [r0, #78] ; y[5] = t[2]-t2[5]
1.412 + STRH r10,[r0, #94] ; y[6] = t[1]-t2[6]
1.413 + STRH r3, [r0, #110] ; y[7] = t2[0]-t[7]
1.414 + MOV PC,R14
1.415 + ENDP
1.416 +
1.417 +idct4core_arm PROC
1.418 + ; r0 = ogg_int16_t *_y (destination)
1.419 + ; r1 = const ogg_int16_t *_x (source)
1.420 + LDRSH r9, [r1], #16 ; r9 = x[0]
1.421 + LDR r10,OC_C4S4 ; r10= OC_C4S4
1.422 + LDRSH r12,[r1, #-12] ; r12= x[2]
1.423 + LDR r4, OC_C6S2 ; r4 = OC_C6S2
1.424 + MUL r9, r10,r9 ; r9 = t[0]<<16 = OC_C4S4*x[0]
1.425 + LDR r5, OC_C2S6 ; r5 = OC_C2S6
1.426 + MUL r4, r12,r4 ; r4 = t[2]<<16 = OC_C6S2*x[2]
1.427 + LDRSH r3, [r1, #-14] ; r3 = x[1]
1.428 + MUL r5, r12,r5 ; r5 = t[3]<<16 = OC_C2S6*x[2]
1.429 + LDR r6, OC_C7S1 ; r6 = OC_C7S1
1.430 + LDR r12,OC_C1S7 ; r12= OC_C1S7
1.431 + LDRSH r11,[r1, #-10] ; r11= x[3]
1.432 + MUL r6, r3, r6 ; r6 = t[4]<<16 = OC_C7S1*x[1]
1.433 + LDR r7, OC_C5S3 ; r7 = OC_C5S3
1.434 + MUL r3, r12,r3 ; r3 = t[7]<<16 = OC_C1S7*x[1]
1.435 + LDR r8, OC_C3S5 ; r8 = OC_C3S5
1.436 + MUL r7, r11,r7 ; r7 = -t[5]<<16 = OC_C5S3*x[3]
1.437 + MOV r9, r9, ASR #16 ; r9 = t[0]
1.438 + MUL r11,r8, r11 ; r11= t[6]<<16 = OC_C3S5*x[3]
1.439 + MOV r6, r6, ASR #16 ; r6 = t[4]
1.440 +; TODO: This is wrong; t[4]-t[5] and t[7]-t[6] need to be truncated to 16-bit
1.441 +; before multiplying, not after (this is not equivalent)
1.442 + SUB r7, r6, r7, ASR #16 ; r7 = t2[4]=t[4]+t[5] (as r7=-t[5])
1.443 + RSB r6, r7, r6, LSL #1 ; r6 = t[4]-t[5]
1.444 + MUL r6, r10,r6 ; r6 = t2[5]<<16 =OC_C4S4*(t[4]-t[5])
1.445 + MOV r3, r3, ASR #16 ; r3 = t[7]
1.446 + ADD r11,r3, r11,ASR #16 ; r11= t2[7]=t[7]+t[6]
1.447 + RSB r3, r11,r3, LSL #1 ; r3 = t[7]-t[6]
1.448 + MUL r3, r10,r3 ; r3 = t2[6]<<16 =OC_C4S4*(t[7]-t[6])
1.449 + ADD r4, r9, r4, ASR #16 ; r4 = t[1] = t[0] + t[2]
1.450 + RSB r10,r4, r9, LSL #1 ; r10= t[2] = t[0] - t[2]
1.451 + ADD r5, r9, r5, ASR #16 ; r5 = t[0] = t[0] + t[3]
1.452 + RSB r9, r5, r9, LSL #1 ; r9 = t[3] = t[0] - t[3]
1.453 + MOV r3, r3, ASR #16 ; r3 = t2[6]
1.454 + ADD r6, r3, r6, ASR #16 ; r6 = t3[6] = t2[6]+t2[5]
1.455 + RSB r3, r6, r3, LSL #1 ; r3 = t3[5] = t2[6]-t2[5]
1.456 + ADD r11,r5, r11 ; r11= t[0]+t2[7]
1.457 + ADD r6, r4, r6 ; r6 = t[1]+t3[6]
1.458 + ADD r3, r10,r3 ; r3 = t[2]+t3[5]
1.459 + ADD r7, r9, r7 ; r7 = t[3]+t2[4]
1.460 + STRH r11,[r0], #2 ; y[0] = t[0]+t[7]
1.461 + STRH r6, [r0, #14] ; y[1] = t[1]+t2[6]
1.462 + STRH r3, [r0, #30] ; y[2] = t[2]+t2[5]
1.463 + STRH r7, [r0, #46] ; y[3] = t2[3]+t[4]
1.464 + RSB r11,r11,r5, LSL #1 ; r11= t[0]-t2[7]
1.465 + RSB r6, r6, r4, LSL #1 ; r6 = t[1]-t3[6]
1.466 + RSB r3, r3, r10,LSL #1 ; r3 = t[2]-t3[5]
1.467 + RSB r7, r7, r9, LSL #1 ; r7 = t[3]-t2[4]
1.468 + STRH r7, [r0, #62] ; y[4] = t2[3]-t[4]
1.469 + STRH r3, [r0, #78] ; y[5] = t[2]-t2[5]
1.470 + STRH r6, [r0, #94] ; y[6] = t[1]-t2[6]
1.471 + STRH r11, [r0, #110] ; y[7] = t2[0]-t[7]
1.472 + MOV PC,r14
1.473 + ENDP
1.474 +
1.475 +idct4core_down_arm PROC
1.476 + ; r0 = ogg_int16_t *_y (destination)
1.477 + ; r1 = const ogg_int16_t *_x (source)
1.478 + LDRSH r9, [r1], #16 ; r9 = x[0]
1.479 + LDR r10,OC_C4S4 ; r10= OC_C4S4
1.480 + LDRSH r12,[r1, #-12] ; r12= x[2]
1.481 + LDR r4, OC_C6S2 ; r4 = OC_C6S2
1.482 + MUL r9, r10,r9 ; r9 = t[0]<<16 = OC_C4S4*x[0]
1.483 + LDR r5, OC_C2S6 ; r5 = OC_C2S6
1.484 + MUL r4, r12,r4 ; r4 = t[2]<<16 = OC_C6S2*x[2]
1.485 + LDRSH r3, [r1, #-14] ; r3 = x[1]
1.486 + MUL r5, r12,r5 ; r5 = t[3]<<16 = OC_C2S6*x[2]
1.487 + LDR r6, OC_C7S1 ; r6 = OC_C7S1
1.488 + LDR r12,OC_C1S7 ; r12= OC_C1S7
1.489 + LDRSH r11,[r1, #-10] ; r11= x[3]
1.490 + MUL r6, r3, r6 ; r6 = t[4]<<16 = OC_C7S1*x[1]
1.491 + LDR r7, OC_C5S3 ; r7 = OC_C5S3
1.492 + MUL r3, r12,r3 ; r3 = t[7]<<16 = OC_C1S7*x[1]
1.493 + LDR r8, OC_C3S5 ; r8 = OC_C3S5
1.494 + MUL r7, r11,r7 ; r7 = -t[5]<<16 = OC_C5S3*x[3]
1.495 + MOV r9, r9, ASR #16 ; r9 = t[0]
1.496 + MUL r11,r8, r11 ; r11= t[6]<<16 = OC_C3S5*x[3]
1.497 + MOV r6, r6, ASR #16 ; r6 = t[4]
1.498 +; TODO: This is wrong; t[4]-t[5] and t[7]-t[6] need to be truncated to 16-bit
1.499 +; before multiplying, not after (this is not equivalent)
1.500 + SUB r7, r6, r7, ASR #16 ; r7 = t2[4]=t[4]+t[5] (as r7=-t[5])
1.501 + RSB r6, r7, r6, LSL #1 ; r6 = t[4]-t[5]
1.502 + MUL r6, r10,r6 ; r6 = t2[5]<<16 =OC_C4S4*(t[4]-t[5])
1.503 + MOV r3, r3, ASR #16 ; r3 = t[7]
1.504 + ADD r11,r3, r11,ASR #16 ; r11= t2[7]=t[7]+t[6]
1.505 + RSB r3, r11,r3, LSL #1 ; r3 = t[7]-t[6]
1.506 + ADD r9, r9, #8 ; r9 = t[0]+8
1.507 + MUL r3, r10,r3 ; r3 = t2[6]<<16 =OC_C4S4*(t[7]-t[6])
1.508 + ADD r4, r9, r4, ASR #16 ; r4 = t[1] = t[0] + t[2] + 8
1.509 + RSB r10,r4, r9, LSL #1 ; r10= t[2] = t[0] - t[2] + 8
1.510 + ADD r5, r9, r5, ASR #16 ; r5 = t[0] = t[0] + t[3] + 8
1.511 + RSB r9, r5, r9, LSL #1 ; r9 = t[3] = t[0] - t[3] + 8
1.512 + MOV r3, r3, ASR #16 ; r3 = t2[6]
1.513 + ADD r6, r3, r6, ASR #16 ; r6 = t3[6] = t2[6]+t2[5]
1.514 + RSB r3, r6, r3, LSL #1 ; r3 = t3[5] = t2[6]-t2[5]
1.515 + ADD r5, r5, r11 ; r5 = t[0]+t2[7]+8
1.516 + ADD r4, r4, r6 ; r4 = t[1]+t3[6]+8
1.517 + ADD r10,r10,r3 ; r10= t[2]+t3[5]+8
1.518 + ADD r9, r9, r7 ; r9 = t[3]+t2[4]+8
1.519 + SUB r11,r5, r11,LSL #1 ; r11= t[0]-t2[7]+8
1.520 + SUB r6, r4, r6, LSL #1 ; r6 = t[1]-t3[6]+8
1.521 + SUB r3, r10,r3, LSL #1 ; r3 = t[2]-t3[5]+8
1.522 + SUB r7, r9, r7, LSL #1 ; r7 = t[3]-t2[4]+8
1.523 + ; TODO: This is wrong.
1.524 + ; The C code truncates to 16 bits by storing to RAM and doing the
1.525 + ; shifts later; we've got an extra 4 bits here.
1.526 + MOV r11,r11,ASR #4
1.527 + MOV r6, r6, ASR #4
1.528 + MOV r3, r3, ASR #4
1.529 + MOV r7, r7, ASR #4
1.530 + MOV r9, r9, ASR #4
1.531 + MOV r10,r10,ASR #4
1.532 + MOV r4, r4, ASR #4
1.533 + MOV r5, r5, ASR #4
1.534 + STRH r5,[r0], #2 ; y[0] = t[0]+t[7]
1.535 + STRH r4, [r0, #14] ; y[1] = t[1]+t2[6]
1.536 + STRH r10,[r0, #30] ; y[2] = t[2]+t2[5]
1.537 + STRH r9, [r0, #46] ; y[3] = t2[3]+t[4]
1.538 + STRH r7, [r0, #62] ; y[4] = t2[3]-t[4]
1.539 + STRH r3, [r0, #78] ; y[5] = t[2]-t2[5]
1.540 + STRH r6, [r0, #94] ; y[6] = t[1]-t2[6]
1.541 + STRH r11,[r0, #110] ; y[7] = t2[0]-t[7]
1.542 + MOV PC,r14
1.543 + ENDP
1.544 +
1.545 +idct8core_arm PROC
1.546 + ; r0 = ogg_int16_t *_y (destination)
1.547 + ; r1 = const ogg_int16_t *_x (source)
1.548 + LDRSH r2, [r1],#16 ; r2 = x[0]
1.549 + STMFD r13!,{r1,r14}
1.550 + LDRSH r6, [r1, #-8] ; r6 = x[4]
1.551 + LDR r12,OC_C4S4 ; r12= C4S4
1.552 + LDRSH r4, [r1, #-12] ; r4 = x[2]
1.553 + ADD r2, r2, r6 ; r2 = x[0] + x[4]
1.554 + SUB r6, r2, r6, LSL #1 ; r6 = x[0] - x[4]
1.555 + ; For spec compliance, these sums must be truncated to 16-bit precision
1.556 + ; _before_ the multiply (not after).
1.557 + ; Sadly, ARMv4 provides no simple way to do that.
1.558 + MOV r2, r2, LSL #16
1.559 + MOV r6, r6, LSL #16
1.560 + MOV r2, r2, ASR #16
1.561 + MOV r6, r6, ASR #16
1.562 + MUL r2, r12,r2 ; r2 = t[0]<<16 = C4S4*(x[0]+x[4])
1.563 + LDRSH r8, [r1, #-4] ; r8 = x[6]
1.564 + LDR r7, OC_C6S2 ; r7 = OC_C6S2
1.565 + MUL r6, r12,r6 ; r6 = t[1]<<16 = C4S4*(x[0]-x[4])
1.566 + LDR r14,OC_C2S6 ; r14= OC_C2S6
1.567 + MUL r3, r4, r7 ; r3 = OC_C6S2*x[2]
1.568 + LDR r5, OC_C7S1 ; r5 = OC_C7S1
1.569 + MUL r4, r14,r4 ; r4 = OC_C2S6*x[2]
1.570 + MOV r3, r3, ASR #16 ; r3 = OC_C6S2*x[2]>>16
1.571 + MUL r14,r8, r14 ; r14= OC_C2S6*x[6]
1.572 + MOV r4, r4, ASR #16 ; r4 = OC_C2S6*x[2]>>16
1.573 + MUL r8, r7, r8 ; r8 = OC_C6S2*x[6]
1.574 + LDR r7, OC_C1S7 ; r7 = OC_C1S7
1.575 + SUB r3, r3, r14,ASR #16 ; r3=t[2]=C6S2*x[2]>>16-C2S6*x[6]>>16
1.576 + LDRSH r14,[r1, #-14] ; r14= x[1]
1.577 + ADD r4, r4, r8, ASR #16 ; r4=t[3]=C2S6*x[2]>>16+C6S2*x[6]>>16
1.578 + LDRSH r8, [r1, #-2] ; r8 = x[7]
1.579 + MUL r9, r5, r14 ; r9 = OC_C7S1*x[1]
1.580 + LDRSH r10,[r1, #-6] ; r10= x[5]
1.581 + MUL r14,r7, r14 ; r14= OC_C1S7*x[1]
1.582 + MOV r9, r9, ASR #16 ; r9 = OC_C7S1*x[1]>>16
1.583 + MUL r7, r8, r7 ; r7 = OC_C1S7*x[7]
1.584 + MOV r14,r14,ASR #16 ; r14= OC_C1S7*x[1]>>16
1.585 + MUL r8, r5, r8 ; r8 = OC_C7S1*x[7]
1.586 + LDRSH r1, [r1, #-10] ; r1 = x[3]
1.587 + LDR r5, OC_C3S5 ; r5 = OC_C3S5
1.588 + LDR r11,OC_C5S3 ; r11= OC_C5S3
1.589 + ADD r8, r14,r8, ASR #16 ; r8=t[7]=C1S7*x[1]>>16+C7S1*x[7]>>16
1.590 + MUL r14,r5, r10 ; r14= OC_C3S5*x[5]
1.591 + SUB r9, r9, r7, ASR #16 ; r9=t[4]=C7S1*x[1]>>16-C1S7*x[7]>>16
1.592 + MUL r10,r11,r10 ; r10= OC_C5S3*x[5]
1.593 + MOV r14,r14,ASR #16 ; r14= OC_C3S5*x[5]>>16
1.594 + MUL r11,r1, r11 ; r11= OC_C5S3*x[3]
1.595 + MOV r10,r10,ASR #16 ; r10= OC_C5S3*x[5]>>16
1.596 + MUL r1, r5, r1 ; r1 = OC_C3S5*x[3]
1.597 + SUB r14,r14,r11,ASR #16 ;r14=t[5]=C3S5*x[5]>>16-C5S3*x[3]>>16
1.598 + ADD r10,r10,r1, ASR #16 ;r10=t[6]=C5S3*x[5]>>16+C3S5*x[3]>>16
1.599 + ; r2=t[0]<<16 r3=t[2] r4=t[3] r6=t[1]<<16 r8=t[7] r9=t[4]
1.600 + ; r10=t[6] r12=C4S4 r14=t[5]
1.601 +; TODO: This is wrong; t[4]-t[5] and t[7]-t[6] need to be truncated to 16-bit
1.602 +; before multiplying, not after (this is not equivalent)
1.603 + ; Stage 2
1.604 + ; 4-5 butterfly
1.605 + ADD r9, r9, r14 ; r9 = t2[4] = t[4]+t[5]
1.606 + SUB r14,r9, r14, LSL #1 ; r14= t[4]-t[5]
1.607 + MUL r14,r12,r14 ; r14= t2[5]<<16 = C4S4*(t[4]-t[5])
1.608 + ; 7-6 butterfly
1.609 + ADD r8, r8, r10 ; r8 = t2[7] = t[7]+t[6]
1.610 + SUB r10,r8, r10, LSL #1 ; r10= t[7]-t[6]
1.611 + MUL r10,r12,r10 ; r10= t2[6]<<16 = C4S4*(t[7]+t[6])
1.612 + ; r2=t[0]<<16 r3=t[2] r4=t[3] r6=t[1]<<16 r8=t2[7] r9=t2[4]
1.613 + ; r10=t2[6]<<16 r12=C4S4 r14=t2[5]<<16
1.614 + ; Stage 3
1.615 + ; 0-3 butterfly
1.616 + ADD r2, r4, r2, ASR #16 ; r2 = t2[0] = t[0] + t[3]
1.617 + SUB r4, r2, r4, LSL #1 ; r4 = t2[3] = t[0] - t[3]
1.618 + ; 1-2 butterfly
1.619 + ADD r6, r3, r6, ASR #16 ; r6 = t2[1] = t[1] + t[2]
1.620 + SUB r3, r6, r3, LSL #1 ; r3 = t2[2] = t[1] - t[2]
1.621 + ; 6-5 butterfly
1.622 + MOV r14,r14,ASR #16 ; r14= t2[5]
1.623 + ADD r10,r14,r10,ASR #16 ; r10= t3[6] = t[6] + t[5]
1.624 + SUB r14,r10,r14,LSL #1 ; r14= t3[5] = t[6] - t[5]
1.625 + ; r2=t2[0] r3=t2[2] r4=t2[3] r6=t2[1] r8=t2[7] r9=t2[4]
1.626 + ; r10=t3[6] r14=t3[5]
1.627 + ; Stage 4
1.628 + ADD r2, r2, r8 ; r2 = t[0] + t[7]
1.629 + ADD r6, r6, r10 ; r6 = t[1] + t[6]
1.630 + ADD r3, r3, r14 ; r3 = t[2] + t[5]
1.631 + ADD r4, r4, r9 ; r4 = t[3] + t[4]
1.632 + SUB r8, r2, r8, LSL #1 ; r8 = t[0] - t[7]
1.633 + SUB r10,r6, r10,LSL #1 ; r10= t[1] - t[6]
1.634 + SUB r14,r3, r14,LSL #1 ; r14= t[2] - t[5]
1.635 + SUB r9, r4, r9, LSL #1 ; r9 = t[3] - t[4]
1.636 + STRH r2, [r0], #2 ; y[0] = t[0]+t[7]
1.637 + STRH r6, [r0, #14] ; y[1] = t[1]+t[6]
1.638 + STRH r3, [r0, #30] ; y[2] = t[2]+t[5]
1.639 + STRH r4, [r0, #46] ; y[3] = t[3]+t[4]
1.640 + STRH r9, [r0, #62] ; y[4] = t[3]-t[4]
1.641 + STRH r14,[r0, #78] ; y[5] = t[2]-t[5]
1.642 + STRH r10,[r0, #94] ; y[6] = t[1]-t[6]
1.643 + STRH r8, [r0, #110] ; y[7] = t[0]-t[7]
1.644 + LDMFD r13!,{r1,PC}
1.645 + ENDP
1.646 +
1.647 +idct8core_down_arm PROC
1.648 + ; r0 = ogg_int16_t *_y (destination)
1.649 + ; r1 = const ogg_int16_t *_x (source)
1.650 + LDRSH r2, [r1],#16 ; r2 = x[0]
1.651 + STMFD r13!,{r1,r14}
1.652 + LDRSH r6, [r1, #-8] ; r6 = x[4]
1.653 + LDR r12,OC_C4S4 ; r12= C4S4
1.654 + LDRSH r4, [r1, #-12] ; r4 = x[2]
1.655 + ADD r2, r2, r6 ; r2 = x[0] + x[4]
1.656 + SUB r6, r2, r6, LSL #1 ; r6 = x[0] - x[4]
1.657 + ; For spec compliance, these sums must be truncated to 16-bit precision
1.658 + ; _before_ the multiply (not after).
1.659 + ; Sadly, ARMv4 provides no simple way to do that.
1.660 + MOV r2, r2, LSL #16
1.661 + MOV r6, r6, LSL #16
1.662 + MOV r2, r2, ASR #16
1.663 + MOV r6, r6, ASR #16
1.664 + MUL r2, r12,r2 ; r2 = t[0]<<16 = C4S4*(x[0]+x[4])
1.665 + LDRSH r8, [r1, #-4] ; r8 = x[6]
1.666 + LDR r7, OC_C6S2 ; r7 = OC_C6S2
1.667 + MUL r6, r12,r6 ; r6 = t[1]<<16 = C4S4*(x[0]-x[4])
1.668 + LDR r14,OC_C2S6 ; r14= OC_C2S6
1.669 + MUL r3, r4, r7 ; r3 = OC_C6S2*x[2]
1.670 + LDR r5, OC_C7S1 ; r5 = OC_C7S1
1.671 + MUL r4, r14,r4 ; r4 = OC_C2S6*x[2]
1.672 + MOV r3, r3, ASR #16 ; r3 = OC_C6S2*x[2]>>16
1.673 + MUL r14,r8, r14 ; r14= OC_C2S6*x[6]
1.674 + MOV r4, r4, ASR #16 ; r4 = OC_C2S6*x[2]>>16
1.675 + MUL r8, r7, r8 ; r8 = OC_C6S2*x[6]
1.676 + LDR r7, OC_C1S7 ; r7 = OC_C1S7
1.677 + SUB r3, r3, r14,ASR #16 ; r3=t[2]=C6S2*x[2]>>16-C2S6*x[6]>>16
1.678 + LDRSH r14,[r1, #-14] ; r14= x[1]
1.679 + ADD r4, r4, r8, ASR #16 ; r4=t[3]=C2S6*x[2]>>16+C6S2*x[6]>>16
1.680 + LDRSH r8, [r1, #-2] ; r8 = x[7]
1.681 + MUL r9, r5, r14 ; r9 = OC_C7S1*x[1]
1.682 + LDRSH r10,[r1, #-6] ; r10= x[5]
1.683 + MUL r14,r7, r14 ; r14= OC_C1S7*x[1]
1.684 + MOV r9, r9, ASR #16 ; r9 = OC_C7S1*x[1]>>16
1.685 + MUL r7, r8, r7 ; r7 = OC_C1S7*x[7]
1.686 + MOV r14,r14,ASR #16 ; r14= OC_C1S7*x[1]>>16
1.687 + MUL r8, r5, r8 ; r8 = OC_C7S1*x[7]
1.688 + LDRSH r1, [r1, #-10] ; r1 = x[3]
1.689 + LDR r5, OC_C3S5 ; r5 = OC_C3S5
1.690 + LDR r11,OC_C5S3 ; r11= OC_C5S3
1.691 + ADD r8, r14,r8, ASR #16 ; r8=t[7]=C1S7*x[1]>>16+C7S1*x[7]>>16
1.692 + MUL r14,r5, r10 ; r14= OC_C3S5*x[5]
1.693 + SUB r9, r9, r7, ASR #16 ; r9=t[4]=C7S1*x[1]>>16-C1S7*x[7]>>16
1.694 + MUL r10,r11,r10 ; r10= OC_C5S3*x[5]
1.695 + MOV r14,r14,ASR #16 ; r14= OC_C3S5*x[5]>>16
1.696 + MUL r11,r1, r11 ; r11= OC_C5S3*x[3]
1.697 + MOV r10,r10,ASR #16 ; r10= OC_C5S3*x[5]>>16
1.698 + MUL r1, r5, r1 ; r1 = OC_C3S5*x[3]
1.699 + SUB r14,r14,r11,ASR #16 ;r14=t[5]=C3S5*x[5]>>16-C5S3*x[3]>>16
1.700 + ADD r10,r10,r1, ASR #16 ;r10=t[6]=C5S3*x[5]>>16+C3S5*x[3]>>16
1.701 + ; r2=t[0]<<16 r3=t[2] r4=t[3] r6=t[1]<<16 r8=t[7] r9=t[4]
1.702 + ; r10=t[6] r12=C4S4 r14=t[5]
1.703 + ; Stage 2
1.704 +; TODO: This is wrong; t[4]-t[5] and t[7]-t[6] need to be truncated to 16-bit
1.705 +; before multiplying, not after (this is not equivalent)
1.706 + ; 4-5 butterfly
1.707 + ADD r9, r9, r14 ; r9 = t2[4] = t[4]+t[5]
1.708 + SUB r14,r9, r14, LSL #1 ; r14= t[4]-t[5]
1.709 + MUL r14,r12,r14 ; r14= t2[5]<<16 = C4S4*(t[4]-t[5])
1.710 + ; 7-6 butterfly
1.711 + ADD r8, r8, r10 ; r8 = t2[7] = t[7]+t[6]
1.712 + SUB r10,r8, r10, LSL #1 ; r10= t[7]-t[6]
1.713 + MUL r10,r12,r10 ; r10= t2[6]<<16 = C4S4*(t[7]+t[6])
1.714 + ; r2=t[0]<<16 r3=t[2] r4=t[3] r6=t[1]<<16 r8=t2[7] r9=t2[4]
1.715 + ; r10=t2[6]<<16 r12=C4S4 r14=t2[5]<<16
1.716 + ; Stage 3
1.717 + ADD r2, r2, #8<<16 ; r2 = t[0]+8<<16
1.718 + ADD r6, r6, #8<<16 ; r6 = t[1]+8<<16
1.719 + ; 0-3 butterfly
1.720 + ADD r2, r4, r2, ASR #16 ; r2 = t2[0] = t[0] + t[3] + 8
1.721 + SUB r4, r2, r4, LSL #1 ; r4 = t2[3] = t[0] - t[3] + 8
1.722 + ; 1-2 butterfly
1.723 + ADD r6, r3, r6, ASR #16 ; r6 = t2[1] = t[1] + t[2] + 8
1.724 + SUB r3, r6, r3, LSL #1 ; r3 = t2[2] = t[1] - t[2] + 8
1.725 + ; 6-5 butterfly
1.726 + MOV r14,r14,ASR #16 ; r14= t2[5]
1.727 + ADD r10,r14,r10,ASR #16 ; r10= t3[6] = t[6] + t[5]
1.728 + SUB r14,r10,r14,LSL #1 ; r14= t3[5] = t[6] - t[5]
1.729 + ; r2=t2[0] r3=t2[2] r4=t2[3] r6=t2[1] r8=t2[7] r9=t2[4]
1.730 + ; r10=t3[6] r14=t3[5]
1.731 + ; Stage 4
1.732 + ADD r2, r2, r8 ; r2 = t[0] + t[7] + 8
1.733 + ADD r6, r6, r10 ; r6 = t[1] + t[6] + 8
1.734 + ADD r3, r3, r14 ; r3 = t[2] + t[5] + 8
1.735 + ADD r4, r4, r9 ; r4 = t[3] + t[4] + 8
1.736 + SUB r8, r2, r8, LSL #1 ; r8 = t[0] - t[7] + 8
1.737 + SUB r10,r6, r10,LSL #1 ; r10= t[1] - t[6] + 8
1.738 + SUB r14,r3, r14,LSL #1 ; r14= t[2] - t[5] + 8
1.739 + SUB r9, r4, r9, LSL #1 ; r9 = t[3] - t[4] + 8
1.740 + ; TODO: This is wrong.
1.741 + ; The C code truncates to 16 bits by storing to RAM and doing the
1.742 + ; shifts later; we've got an extra 4 bits here.
1.743 + MOV r2, r2, ASR #4
1.744 + MOV r6, r6, ASR #4
1.745 + MOV r3, r3, ASR #4
1.746 + MOV r4, r4, ASR #4
1.747 + MOV r8, r8, ASR #4
1.748 + MOV r10,r10,ASR #4
1.749 + MOV r14,r14,ASR #4
1.750 + MOV r9, r9, ASR #4
1.751 + STRH r2, [r0], #2 ; y[0] = t[0]+t[7]
1.752 + STRH r6, [r0, #14] ; y[1] = t[1]+t[6]
1.753 + STRH r3, [r0, #30] ; y[2] = t[2]+t[5]
1.754 + STRH r4, [r0, #46] ; y[3] = t[3]+t[4]
1.755 + STRH r9, [r0, #62] ; y[4] = t[3]-t[4]
1.756 + STRH r14,[r0, #78] ; y[5] = t[2]-t[5]
1.757 + STRH r10,[r0, #94] ; y[6] = t[1]-t[6]
1.758 + STRH r8, [r0, #110] ; y[7] = t[0]-t[7]
1.759 + LDMFD r13!,{r1,PC}
1.760 + ENDP
1.761 +
1.762 + [ OC_ARM_ASM_MEDIA
1.763 + EXPORT oc_idct8x8_1_v6
1.764 + EXPORT oc_idct8x8_v6
1.765 +
1.766 +oc_idct8x8_1_v6 PROC
1.767 + ; r0 = ogg_int16_t *_y
1.768 + ; r1 = ogg_uint16_t _dc
1.769 + ORR r2, r1, r1, LSL #16
1.770 + ORR r3, r1, r1, LSL #16
1.771 + STRD r2, [r0], #8
1.772 + STRD r2, [r0], #8
1.773 + STRD r2, [r0], #8
1.774 + STRD r2, [r0], #8
1.775 + STRD r2, [r0], #8
1.776 + STRD r2, [r0], #8
1.777 + STRD r2, [r0], #8
1.778 + STRD r2, [r0], #8
1.779 + STRD r2, [r0], #8
1.780 + STRD r2, [r0], #8
1.781 + STRD r2, [r0], #8
1.782 + STRD r2, [r0], #8
1.783 + STRD r2, [r0], #8
1.784 + STRD r2, [r0], #8
1.785 + STRD r2, [r0], #8
1.786 + STRD r2, [r0], #8
1.787 + MOV PC, r14
1.788 + ENDP
1.789 +
1.790 +oc_idct8x8_v6 PROC
1.791 + ; r0 = ogg_int16_t *_y
1.792 + ; r1 = ogg_int16_t *_x
1.793 + ; r2 = int _last_zzi
1.794 + CMP r2, #3
1.795 + BLE oc_idct8x8_3_v6
1.796 + ;CMP r2, #6
1.797 + ;BLE oc_idct8x8_6_v6
1.798 + CMP r2, #10
1.799 + BLE oc_idct8x8_10_v6
1.800 +oc_idct8x8_slow_v6
1.801 + STMFD r13!,{r4-r11,r14}
1.802 + SUB r13,r13,#64*2
1.803 +; Row transforms
1.804 + STR r0, [r13,#-4]!
1.805 + ADD r0, r13, #4 ; Write to temp storage.
1.806 + BL idct8_8core_v6
1.807 + BL idct8_8core_v6
1.808 + BL idct8_8core_v6
1.809 + BL idct8_8core_v6
1.810 + LDR r0, [r13], #4 ; Write to the final destination.
1.811 + ; Clear input data for next block (decoder only).
1.812 + SUB r2, r1, #8*16
1.813 + CMP r0, r2
1.814 + MOV r1, r13 ; And read from temp storage.
1.815 + BEQ oc_idct8x8_slow_v6_cols
1.816 + MOV r4, #0
1.817 + MOV r5, #0
1.818 + STRD r4, [r2], #8
1.819 + STRD r4, [r2], #8
1.820 + STRD r4, [r2], #8
1.821 + STRD r4, [r2], #8
1.822 + STRD r4, [r2], #8
1.823 + STRD r4, [r2], #8
1.824 + STRD r4, [r2], #8
1.825 + STRD r4, [r2], #8
1.826 + STRD r4, [r2], #8
1.827 + STRD r4, [r2], #8
1.828 + STRD r4, [r2], #8
1.829 + STRD r4, [r2], #8
1.830 + STRD r4, [r2], #8
1.831 + STRD r4, [r2], #8
1.832 + STRD r4, [r2], #8
1.833 + STRD r4, [r2], #8
1.834 +oc_idct8x8_slow_v6_cols
1.835 +; Column transforms
1.836 + BL idct8_8core_down_v6
1.837 + BL idct8_8core_down_v6
1.838 + BL idct8_8core_down_v6
1.839 + BL idct8_8core_down_v6
1.840 + ADD r13,r13,#64*2
1.841 + LDMFD r13!,{r4-r11,PC}
1.842 + ENDP
1.843 +
1.844 +oc_idct8x8_10_v6 PROC
1.845 + STMFD r13!,{r4-r11,r14}
1.846 + SUB r13,r13,#64*2+4
1.847 +; Row transforms
1.848 + MOV r2, r13
1.849 + STR r0, [r13,#-4]!
1.850 + AND r0, r2, #4 ; Align the stack.
1.851 + ADD r0, r0, r2 ; Write to temp storage.
1.852 + BL idct4_3core_v6
1.853 + BL idct2_1core_v6
1.854 + LDR r0, [r13], #4 ; Write to the final destination.
1.855 + ; Clear input data for next block (decoder only).
1.856 + SUB r2, r1, #4*16
1.857 + CMP r0, r2
1.858 + AND r1, r13,#4 ; Align the stack.
1.859 + BEQ oc_idct8x8_10_v6_cols
1.860 + MOV r4, #0
1.861 + MOV r5, #0
1.862 + STRD r4, [r2]
1.863 + STRD r4, [r2,#16]
1.864 + STR r4, [r2,#32]
1.865 + STR r4, [r2,#48]
1.866 +oc_idct8x8_10_v6_cols
1.867 +; Column transforms
1.868 + ADD r1, r1, r13 ; And read from temp storage.
1.869 + BL idct4_4core_down_v6
1.870 + BL idct4_4core_down_v6
1.871 + BL idct4_4core_down_v6
1.872 + BL idct4_4core_down_v6
1.873 + ADD r13,r13,#64*2+4
1.874 + LDMFD r13!,{r4-r11,PC}
1.875 + ENDP
1.876 +
1.877 +oc_idct8x8_3_v6 PROC
1.878 + STMFD r13!,{r4-r8,r14}
1.879 + SUB r13,r13,#64*2
1.880 +; Row transforms
1.881 + MOV r8, r0
1.882 + MOV r0, r13 ; Write to temp storage.
1.883 + BL idct2_1core_v6
1.884 + ; Clear input data for next block (decoder only).
1.885 + SUB r0, r1, #2*16
1.886 + CMP r0, r8
1.887 + MOV r1, r13 ; Read from temp storage.
1.888 + MOVNE r4, #0
1.889 + STRNE r4, [r0]
1.890 + STRNE r4, [r0,#16]
1.891 + MOVNE r0, r8 ; Write to the final destination.
1.892 +; Column transforms
1.893 + BL idct2_2core_down_v6
1.894 + BL idct2_2core_down_v6
1.895 + BL idct2_2core_down_v6
1.896 + BL idct2_2core_down_v6
1.897 + ADD r13,r13,#64*2
1.898 + LDMFD r13!,{r4-r8,PC}
1.899 + ENDP
1.900 +
1.901 +idct2_1core_v6 PROC
1.902 + ; r0 = ogg_int16_t *_y (destination)
1.903 + ; r1 = const ogg_int16_t *_x (source)
1.904 +; Stage 1:
1.905 + LDR r2, [r1], #16 ; r2 = <x[0,1]|x[0,0]>
1.906 + LDR r3, OC_C4S4
1.907 + LDRSH r6, [r1], #16 ; r6 = x[1,0]
1.908 + SMULWB r12,r3, r2 ; r12= t[0,0]=OC_C4S4*x[0,0]>>16
1.909 + LDRD r4, OC_C7S1 ; r4 = OC_C7S1; r5 = OC_C1S7
1.910 + SMULWB r6, r3, r6 ; r6 = t[1,0]=OC_C4S4*x[1,0]>>16
1.911 + SMULWT r4, r4, r2 ; r4 = t[0,4]=OC_C7S1*x[0,1]>>16
1.912 + SMULWT r7, r5, r2 ; r7 = t[0,7]=OC_C1S7*x[0,1]>>16
1.913 +; Stage 2:
1.914 + SMULWB r5, r3, r4 ; r5 = t[0,5]=OC_C4S4*t[0,4]>>16
1.915 + PKHBT r12,r12,r6, LSL #16 ; r12= <t[1,0]|t[0,0]>
1.916 + SMULWB r6, r3, r7 ; r6 = t[0,6]=OC_C4S4*t[0,7]>>16
1.917 + PKHBT r7, r7, r3 ; r7 = <0|t[0,7]>
1.918 +; Stage 3:
1.919 + PKHBT r5, r6, r5, LSL #16 ; r5 = <t[0,5]|t[0,6]>
1.920 + PKHBT r4, r4, r3 ; r4 = <0|t[0,4]>
1.921 + SASX r5, r5, r5 ; r5 = <t[0,6]+t[0,5]|t[0,6]-t[0,5]>
1.922 +; Stage 4:
1.923 + PKHTB r6, r3, r5, ASR #16 ; r6 = <0|t[0,6]>
1.924 + PKHBT r5, r5, r3 ; r5 = <0|t[0,5]>
1.925 + SADD16 r3, r12,r7 ; r3 = t[0]+t[7]
1.926 + STR r3, [r0], #4 ; y[0<<3] = t[0]+t[7]
1.927 + SADD16 r3, r12,r6 ; r3 = t[0]+t[6]
1.928 + STR r3, [r0, #12] ; y[1<<3] = t[0]+t[6]
1.929 + SADD16 r3, r12,r5 ; r3 = t[0]+t[5]
1.930 + STR r3, [r0, #28] ; y[2<<3] = t[0]+t[5]
1.931 + SADD16 r3, r12,r4 ; r3 = t[0]+t[4]
1.932 + STR r3, [r0, #44] ; y[3<<3] = t[0]+t[4]
1.933 + SSUB16 r4, r12,r4 ; r4 = t[0]-t[4]
1.934 + STR r4, [r0, #60] ; y[4<<3] = t[0]-t[4]
1.935 + SSUB16 r5, r12,r5 ; r5 = t[0]-t[5]
1.936 + STR r5, [r0, #76] ; y[5<<3] = t[0]-t[5]
1.937 + SSUB16 r6, r12,r6 ; r6 = t[0]-t[6]
1.938 + STR r6, [r0, #92] ; y[6<<3] = t[0]-t[6]
1.939 + SSUB16 r7, r12,r7 ; r7 = t[0]-t[7]
1.940 + STR r7, [r0, #108] ; y[7<<3] = t[0]-t[7]
1.941 + MOV PC,r14
1.942 + ENDP
1.943 + ]
1.944 +
1.945 + ALIGN 8
1.946 +OC_C7S1
1.947 + DCD 12785 ; 31F1
1.948 +OC_C1S7
1.949 + DCD 64277 ; FB15
1.950 +OC_C6S2
1.951 + DCD 25080 ; 61F8
1.952 +OC_C2S6
1.953 + DCD 60547 ; EC83
1.954 +OC_C5S3
1.955 + DCD 36410 ; 8E3A
1.956 +OC_C3S5
1.957 + DCD 54491 ; D4DB
1.958 +OC_C4S4
1.959 + DCD 46341 ; B505
1.960 +
1.961 + [ OC_ARM_ASM_MEDIA
1.962 +idct2_2core_down_v6 PROC
1.963 + ; r0 = ogg_int16_t *_y (destination)
1.964 + ; r1 = const ogg_int16_t *_x (source)
1.965 +; Stage 1:
1.966 + LDR r2, [r1], #16 ; r2 = <x[0,1]|x[0,0]>
1.967 + LDR r3, OC_C4S4
1.968 + MOV r7 ,#8 ; r7 = 8
1.969 + LDR r6, [r1], #16 ; r6 = <x[1,1]|x[1,0]>
1.970 + SMLAWB r12,r3, r2, r7 ; r12= (t[0,0]=OC_C4S4*x[0,0]>>16)+8
1.971 + LDRD r4, OC_C7S1 ; r4 = OC_C7S1; r5 = OC_C1S7
1.972 + SMLAWB r7, r3, r6, r7 ; r7 = (t[1,0]=OC_C4S4*x[1,0]>>16)+8
1.973 + SMULWT r5, r5, r2 ; r2 = t[0,7]=OC_C1S7*x[0,1]>>16
1.974 + PKHBT r12,r12,r7, LSL #16 ; r12= <t[1,0]+8|t[0,0]+8>
1.975 + SMULWT r4, r4, r2 ; r4 = t[0,4]=OC_C7S1*x[0,1]>>16
1.976 +; Here we cheat: row 1 had just a DC, so x[0,1]==x[1,1] by definition.
1.977 + PKHBT r7, r5, r5, LSL #16 ; r7 = <t[0,7]|t[0,7]>
1.978 +; Stage 2:
1.979 + SMULWB r6, r3, r7 ; r6 = t[0,6]=OC_C4S4*t[0,7]>>16
1.980 + PKHBT r4, r4, r4, LSL #16 ; r4 = <t[0,4]|t[0,4]>
1.981 + SMULWT r2, r3, r7 ; r2 = t[1,6]=OC_C4S4*t[1,7]>>16
1.982 + SMULWB r5, r3, r4 ; r5 = t[0,5]=OC_C4S4*t[0,4]>>16
1.983 + PKHBT r6, r6, r2, LSL #16 ; r6 = <t[1,6]|t[0,6]>
1.984 + SMULWT r2, r3, r4 ; r2 = t[1,5]=OC_C4S4*t[1,4]>>16
1.985 + PKHBT r2, r5, r2, LSL #16 ; r2 = <t[1,5]|t[0,5]>
1.986 +; Stage 3:
1.987 + SSUB16 r5, r6, r2 ; r5 = <t[1,6]-t[1,5]|t[0,6]-t[0,5]>
1.988 + SADD16 r6, r6, r2 ; r6 = <t[1,6]+t[1,5]|t[0,6]+t[0,5]>
1.989 +; Stage 4:
1.990 + SADD16 r2, r12,r7 ; r2 = t[0]+t[7]+8
1.991 + MOV r3, r2, ASR #4
1.992 + MOV r2, r2, LSL #16
1.993 + PKHTB r3, r3, r2, ASR #20 ; r3 = t[0]+t[7]+8>>4
1.994 + STR r3, [r0], #4 ; y[0<<3] = t[0]+t[7]+8>>4
1.995 + SADD16 r2, r12,r6 ; r2 = t[0]+t[6]+8
1.996 + MOV r3, r2, ASR #4
1.997 + MOV r2, r2, LSL #16
1.998 + PKHTB r3, r3, r2, ASR #20 ; r3 = t[0]+t[6]+8>>4
1.999 + STR r3, [r0, #12] ; y[1<<3] = t[0]+t[6]+8>>4
1.1000 + SADD16 r2, r12,r5 ; r2 = t[0]+t[5]+8
1.1001 + MOV r3, r2, ASR #4
1.1002 + MOV r2, r2, LSL #16
1.1003 + PKHTB r3, r3, r2, ASR #20 ; r3 = t[0]+t[5]+8>>4
1.1004 + STR r3, [r0, #28] ; y[2<<3] = t[0]+t[5]+8>>4
1.1005 + SADD16 r2, r12,r4 ; r2 = t[0]+t[4]+8
1.1006 + MOV r3, r2, ASR #4
1.1007 + MOV r2, r2, LSL #16
1.1008 + PKHTB r3, r3, r2, ASR #20 ; r3 = t[0]+t[4]+8>>4
1.1009 + STR r3, [r0, #44] ; y[3<<3] = t[0]+t[4]+8>>4
1.1010 + SSUB16 r4, r12,r4 ; r4 = t[0]-t[4]+8
1.1011 + MOV r3, r4, ASR #4
1.1012 + MOV r4, r4, LSL #16
1.1013 + PKHTB r3, r3, r4, ASR #20 ; r3 = t[0]-t[4]+8>>4
1.1014 + STR r3, [r0, #60] ; y[4<<3] = t[0]-t[4]+8>>4
1.1015 + SSUB16 r5, r12,r5 ; r5 = t[0]-t[5]+8
1.1016 + MOV r3, r5, ASR #4
1.1017 + MOV r5, r5, LSL #16
1.1018 + PKHTB r3, r3, r5, ASR #20 ; r3 = t[0]-t[5]+8>>4
1.1019 + STR r3, [r0, #76] ; y[5<<3] = t[0]-t[5]+8>>4
1.1020 + SSUB16 r6, r12,r6 ; r6 = t[0]-t[6]+8
1.1021 + MOV r3, r6, ASR #4
1.1022 + MOV r6, r6, LSL #16
1.1023 + PKHTB r3, r3, r6, ASR #20 ; r3 = t[0]-t[6]+8>>4
1.1024 + STR r3, [r0, #92] ; y[6<<3] = t[0]-t[6]+8>>4
1.1025 + SSUB16 r7, r12,r7 ; r7 = t[0]-t[7]+8
1.1026 + MOV r3, r7, ASR #4
1.1027 + MOV r7, r7, LSL #16
1.1028 + PKHTB r3, r3, r7, ASR #20 ; r3 = t[0]-t[7]+8>>4
1.1029 + STR r3, [r0, #108] ; y[7<<3] = t[0]-t[7]+8>>4
1.1030 + MOV PC,r14
1.1031 + ENDP
1.1032 +
1.1033 +; In theory this should save ~75 cycles over oc_idct8x8_10, more than enough to
1.1034 +; pay for increased branch mis-prediction to get here, but in practice it
1.1035 +; doesn't seem to slow anything down to take it out, and it's less code this
1.1036 +; way.
1.1037 + [ 0
1.1038 +oc_idct8x8_6_v6 PROC
1.1039 + STMFD r13!,{r4-r8,r10,r11,r14}
1.1040 + SUB r13,r13,#64*2+4
1.1041 +; Row transforms
1.1042 + MOV r8, r0
1.1043 + AND r0, r13,#4 ; Align the stack.
1.1044 + ADD r0, r0, r13 ; Write to temp storage.
1.1045 + BL idct3_2core_v6
1.1046 + BL idct1core_v6
1.1047 + ; Clear input data for next block (decoder only).
1.1048 + SUB r0, r1, #3*16
1.1049 + CMP r0, r8
1.1050 + AND r1, r13,#4 ; Align the stack.
1.1051 + BEQ oc_idct8x8_6_v6_cols
1.1052 + MOV r4, #0
1.1053 + MOV r5, #0
1.1054 + STRD r4, [r0]
1.1055 + STR r4, [r0,#16]
1.1056 + STR r4, [r0,#32]
1.1057 + MOV r0, r8 ; Write to the final destination.
1.1058 +oc_idct8x8_6_v6_cols
1.1059 +; Column transforms
1.1060 + ADD r1, r1, r13 ; And read from temp storage.
1.1061 + BL idct3_3core_down_v6
1.1062 + BL idct3_3core_down_v6
1.1063 + BL idct3_3core_down_v6
1.1064 + BL idct3_3core_down_v6
1.1065 + ADD r13,r13,#64*2+4
1.1066 + LDMFD r13!,{r4-r8,r10,r11,PC}
1.1067 + ENDP
1.1068 +
1.1069 +idct1core_v6 PROC
1.1070 + ; r0 = ogg_int16_t *_y (destination)
1.1071 + ; r1 = const ogg_int16_t *_x (source)
1.1072 + LDRSH r3, [r1], #16
1.1073 + MOV r12,#0x05
1.1074 + ORR r12,r12,#0xB500
1.1075 + MUL r3, r12, r3
1.1076 + ; Stall ?
1.1077 + MOV r3, r3, ASR #16
1.1078 + ; Don't need to actually store the odd lines; they won't be read.
1.1079 + STRH r3, [r0], #2
1.1080 + STRH r3, [r0, #30]
1.1081 + STRH r3, [r0, #62]
1.1082 + STRH r3, [r0, #94]
1.1083 + MOV PC,R14
1.1084 + ENDP
1.1085 +
1.1086 +idct3_2core_v6 PROC
1.1087 + ; r0 = ogg_int16_t *_y (destination)
1.1088 + ; r1 = const ogg_int16_t *_x (source)
1.1089 +; Stage 1:
1.1090 + LDRD r4, [r1], #16 ; r4 = <x[0,1]|x[0,0]>; r5 = <*|x[0,2]>
1.1091 + LDRD r10,OC_C6S2_3_v6 ; r10= OC_C6S2; r11= OC_C2S6
1.1092 + ; Stall
1.1093 + SMULWB r3, r11,r5 ; r3 = t[0,3]=OC_C2S6*x[0,2]>>16
1.1094 + LDR r11,OC_C4S4
1.1095 + SMULWB r2, r10,r5 ; r2 = t[0,2]=OC_C6S2*x[0,2]>>16
1.1096 + LDR r5, [r1], #16 ; r5 = <x[1,1]|x[1,0]>
1.1097 + SMULWB r12,r11,r4 ; r12= (t[0,0]=OC_C4S4*x[0,0]>>16)
1.1098 + LDRD r6, OC_C7S1_3_v6 ; r6 = OC_C7S1; r7 = OC_C1S7
1.1099 + SMULWB r10,r11,r5 ; r10= (t[1,0]=OC_C4S4*x[1,0]>>16)
1.1100 + PKHBT r12,r12,r10,LSL #16 ; r12= <t[1,0]|t[0,0]>
1.1101 + SMULWT r10,r7, r5 ; r10= t[1,7]=OC_C1S7*x[1,1]>>16
1.1102 + PKHBT r2, r2, r11 ; r2 = <0|t[0,2]>
1.1103 + SMULWT r7, r7, r4 ; r7 = t[0,7]=OC_C1S7*x[0,1]>>16
1.1104 + PKHBT r3, r3, r11 ; r3 = <0|t[0,3]>
1.1105 + SMULWT r5, r6, r5 ; r10= t[1,4]=OC_C7S1*x[1,1]>>16
1.1106 + PKHBT r7, r7, r10,LSL #16 ; r7 = <t[1,7]|t[0,7]>
1.1107 + SMULWT r4, r6, r4 ; r4 = t[0,4]=OC_C7S1*x[0,1]>>16
1.1108 +; Stage 2:
1.1109 + SMULWB r6, r11,r7 ; r6 = t[0,6]=OC_C4S4*t[0,7]>>16
1.1110 + PKHBT r4, r4, r5, LSL #16 ; r4 = <t[1,4]|t[0,4]>
1.1111 + SMULWT r10,r11,r7 ; r10= t[1,6]=OC_C4S4*t[1,7]>>16
1.1112 + SMULWB r5, r11,r4 ; r5 = t[0,5]=OC_C4S4*t[0,4]>>16
1.1113 + PKHBT r6, r6, r10,LSL #16 ; r6 = <t[1,6]|t[0,6]>
1.1114 + SMULWT r10,r11,r4 ; r10= t[1,5]=OC_C4S4*t[1,4]>>16
1.1115 +; Stage 3:
1.1116 + B idct4_3core_stage3_v6
1.1117 + ENDP
1.1118 +
1.1119 +; Another copy so the LDRD offsets are less than +/- 255.
1.1120 + ALIGN 8
1.1121 +OC_C7S1_3_v6
1.1122 + DCD 12785 ; 31F1
1.1123 +OC_C1S7_3_v6
1.1124 + DCD 64277 ; FB15
1.1125 +OC_C6S2_3_v6
1.1126 + DCD 25080 ; 61F8
1.1127 +OC_C2S6_3_v6
1.1128 + DCD 60547 ; EC83
1.1129 +
1.1130 +idct3_3core_down_v6 PROC
1.1131 + ; r0 = ogg_int16_t *_y (destination)
1.1132 + ; r1 = const ogg_int16_t *_x (source)
1.1133 +; Stage 1:
1.1134 + LDRD r10,[r1], #16 ; r10= <x[0,1]|x[0,0]>; r11= <??|x[0,2]>
1.1135 + LDRD r6, OC_C6S2_3_v6 ; r6 = OC_C6S2; r7 = OC_C2S6
1.1136 + LDR r4, [r1], #16 ; r4 = <x[1,1]|x[1,0]>
1.1137 + SMULWB r3, r7, r11 ; r3 = t[0,3]=OC_C2S6*x[0,2]>>16
1.1138 + MOV r7,#8
1.1139 + SMULWB r2, r6, r11 ; r2 = t[0,2]=OC_C6S2*x[0,2]>>16
1.1140 + LDR r11,OC_C4S4
1.1141 + SMLAWB r12,r11,r10,r7 ; r12= t[0,0]+8=(OC_C4S4*x[0,0]>>16)+8
1.1142 +; Here we cheat: row 2 had just a DC, so x[0,2]==x[1,2] by definition.
1.1143 + PKHBT r3, r3, r3, LSL #16 ; r3 = <t[0,3]|t[0,3]>
1.1144 + SMLAWB r5, r11,r4, r7 ; r5 = t[1,0]+8=(OC_C4S4*x[1,0]>>16)+8
1.1145 + PKHBT r2, r2, r2, LSL #16 ; r2 = <t[0,2]|t[0,2]>
1.1146 + LDRD r6, OC_C7S1_3_v6 ; r6 = OC_C7S1; r7 = OC_C1S7
1.1147 + PKHBT r12,r12,r5, LSL #16 ; r12= <t[1,0]+8|t[0,0]+8>
1.1148 + SMULWT r5, r7, r4 ; r5 = t[1,7]=OC_C1S7*x[1,1]>>16
1.1149 + SMULWT r7, r7, r10 ; r7 = t[0,7]=OC_C1S7*x[0,1]>>16
1.1150 + SMULWT r10,r6, r10 ; r10= t[0,4]=OC_C7S1*x[0,1]>>16
1.1151 + PKHBT r7, r7, r5, LSL #16 ; r7 = <t[1,7]|t[0,7]>
1.1152 + SMULWT r4, r6, r4 ; r4 = t[1,4]=OC_C7S1*x[1,1]>>16
1.1153 +; Stage 2:
1.1154 + SMULWB r6, r11,r7 ; r6 = t[0,6]=OC_C4S4*t[0,7]>>16
1.1155 + PKHBT r4, r10,r4, LSL #16 ; r4 = <t[1,4]|t[0,4]>
1.1156 + SMULWT r10,r11,r7 ; r10= t[1,6]=OC_C4S4*t[1,7]>>16
1.1157 + SMULWB r5, r11,r4 ; r5 = t[0,5]=OC_C4S4*t[0,4]>>16
1.1158 + PKHBT r6, r6, r10,LSL #16 ; r6 = <t[1,6]|t[0,6]>
1.1159 + SMULWT r10,r11,r4 ; r10= t[1,5]=OC_C4S4*t[1,4]>>16
1.1160 +; Stage 3:
1.1161 + B idct4_4core_down_stage3_v6
1.1162 + ENDP
1.1163 + ]
1.1164 +
1.1165 +idct4_3core_v6 PROC
1.1166 + ; r0 = ogg_int16_t *_y (destination)
1.1167 + ; r1 = const ogg_int16_t *_x (source)
1.1168 +; Stage 1:
1.1169 + LDRD r10,[r1], #16 ; r10= <x[0,1]|x[0,0]>; r11= <x[0,3]|x[0,2]>
1.1170 + LDRD r2, OC_C5S3_4_v6 ; r2 = OC_C5S3; r3 = OC_C3S5
1.1171 + LDRD r4, [r1], #16 ; r4 = <x[1,1]|x[1,0]>; r5 = <??|x[1,2]>
1.1172 + SMULWT r9, r3, r11 ; r9 = t[0,6]=OC_C3S5*x[0,3]>>16
1.1173 + SMULWT r8, r2, r11 ; r8 = -t[0,5]=OC_C5S3*x[0,3]>>16
1.1174 + PKHBT r9, r9, r2 ; r9 = <0|t[0,6]>
1.1175 + LDRD r6, OC_C6S2_4_v6 ; r6 = OC_C6S2; r7 = OC_C2S6
1.1176 + PKHBT r8, r8, r2 ; r9 = <0|-t[0,5]>
1.1177 + SMULWB r3, r7, r11 ; r3 = t[0,3]=OC_C2S6*x[0,2]>>16
1.1178 + SMULWB r2, r6, r11 ; r2 = t[0,2]=OC_C6S2*x[0,2]>>16
1.1179 + LDR r11,OC_C4S4
1.1180 + SMULWB r12,r7, r5 ; r12= t[1,3]=OC_C2S6*x[1,2]>>16
1.1181 + SMULWB r5, r6, r5 ; r5 = t[1,2]=OC_C6S2*x[1,2]>>16
1.1182 + PKHBT r3, r3, r12,LSL #16 ; r3 = <t[1,3]|t[0,3]>
1.1183 + SMULWB r12,r11,r10 ; r12= t[0,0]=OC_C4S4*x[0,0]>>16
1.1184 + PKHBT r2, r2, r5, LSL #16 ; r2 = <t[1,2]|t[0,2]>
1.1185 + SMULWB r5, r11,r4 ; r5 = t[1,0]=OC_C4S4*x[1,0]>>16
1.1186 + LDRD r6, OC_C7S1_4_v6 ; r6 = OC_C7S1; r7 = OC_C1S7
1.1187 + PKHBT r12,r12,r5, LSL #16 ; r12= <t[1,0]|t[0,0]>
1.1188 + SMULWT r5, r7, r4 ; r5 = t[1,7]=OC_C1S7*x[1,1]>>16
1.1189 + SMULWT r7, r7, r10 ; r7 = t[0,7]=OC_C1S7*x[0,1]>>16
1.1190 + SMULWT r10,r6, r10 ; r10= t[0,4]=OC_C7S1*x[0,1]>>16
1.1191 + PKHBT r7, r7, r5, LSL #16 ; r7 = <t[1,7]|t[0,7]>
1.1192 + SMULWT r4, r6, r4 ; r4 = t[1,4]=OC_C7S1*x[1,1]>>16
1.1193 +; Stage 2:
1.1194 + SSUB16 r6, r7, r9 ; r6 = t[7]-t[6]
1.1195 + PKHBT r4, r10,r4, LSL #16 ; r4 = <t[1,4]|t[0,4]>
1.1196 + SADD16 r7, r7, r9 ; r7 = t[7]=t[7]+t[6]
1.1197 + SMULWT r9, r11,r6 ; r9 = t[1,6]=OC_C4S4*r6T>>16
1.1198 + SADD16 r5, r4, r8 ; r5 = t[4]-t[5]
1.1199 + SMULWB r6, r11,r6 ; r6 = t[0,6]=OC_C4S4*r6B>>16
1.1200 + SSUB16 r4, r4, r8 ; r4 = t[4]=t[4]+t[5]
1.1201 + SMULWT r10,r11,r5 ; r10= t[1,5]=OC_C4S4*r5T>>16
1.1202 + PKHBT r6, r6, r9, LSL #16 ; r6 = <t[1,6]|t[0,6]>
1.1203 + SMULWB r5, r11,r5 ; r5 = t[0,5]=OC_C4S4*r5B>>16
1.1204 +; Stage 3:
1.1205 +idct4_3core_stage3_v6
1.1206 + SADD16 r11,r12,r2 ; r11= t[1]=t[0]+t[2]
1.1207 + PKHBT r10,r5, r10,LSL #16 ; r10= <t[1,5]|t[0,5]>
1.1208 + SSUB16 r2, r12,r2 ; r2 = t[2]=t[0]-t[2]
1.1209 +idct4_3core_stage3_5_v6
1.1210 + SSUB16 r5, r6, r10 ; r5 = t[5]'=t[6]-t[5]
1.1211 + SADD16 r6, r6, r10 ; r6 = t[6]=t[6]+t[5]
1.1212 + SADD16 r10,r12,r3 ; r10= t[0]'=t[0]+t[3]
1.1213 + SSUB16 r3, r12,r3 ; r3 = t[3]=t[0]-t[3]
1.1214 +; Stage 4:
1.1215 + SADD16 r12,r10,r7 ; r12= t[0]+t[7]
1.1216 + STR r12,[r0], #4 ; y[0<<3] = t[0]+t[7]
1.1217 + SADD16 r12,r11,r6 ; r12= t[1]+t[6]
1.1218 + STR r12,[r0, #12] ; y[1<<3] = t[1]+t[6]
1.1219 + SADD16 r12,r2, r5 ; r12= t[2]+t[5]
1.1220 + STR r12,[r0, #28] ; y[2<<3] = t[2]+t[5]
1.1221 + SADD16 r12,r3, r4 ; r12= t[3]+t[4]
1.1222 + STR r12,[r0, #44] ; y[3<<3] = t[3]+t[4]
1.1223 + SSUB16 r4, r3, r4 ; r4 = t[3]-t[4]
1.1224 + STR r4, [r0, #60] ; y[4<<3] = t[3]-t[4]
1.1225 + SSUB16 r5, r2, r5 ; r5 = t[2]-t[5]
1.1226 + STR r5, [r0, #76] ; y[5<<3] = t[2]-t[5]
1.1227 + SSUB16 r6, r11,r6 ; r6 = t[1]-t[6]
1.1228 + STR r6, [r0, #92] ; y[6<<3] = t[1]-t[6]
1.1229 + SSUB16 r7, r10,r7 ; r7 = t[0]-t[7]
1.1230 + STR r7, [r0, #108] ; y[7<<3] = t[0]-t[7]
1.1231 + MOV PC,r14
1.1232 + ENDP
1.1233 +
1.1234 +; Another copy so the LDRD offsets are less than +/- 255.
1.1235 + ALIGN 8
1.1236 +OC_C7S1_4_v6
1.1237 + DCD 12785 ; 31F1
1.1238 +OC_C1S7_4_v6
1.1239 + DCD 64277 ; FB15
1.1240 +OC_C6S2_4_v6
1.1241 + DCD 25080 ; 61F8
1.1242 +OC_C2S6_4_v6
1.1243 + DCD 60547 ; EC83
1.1244 +OC_C5S3_4_v6
1.1245 + DCD 36410 ; 8E3A
1.1246 +OC_C3S5_4_v6
1.1247 + DCD 54491 ; D4DB
1.1248 +
1.1249 +idct4_4core_down_v6 PROC
1.1250 + ; r0 = ogg_int16_t *_y (destination)
1.1251 + ; r1 = const ogg_int16_t *_x (source)
1.1252 +; Stage 1:
1.1253 + LDRD r10,[r1], #16 ; r10= <x[0,1]|x[0,0]>; r11= <x[0,3]|x[0,2]>
1.1254 + LDRD r2, OC_C5S3_4_v6 ; r2 = OC_C5S3; r3 = OC_C3S5
1.1255 + LDRD r4, [r1], #16 ; r4 = <x[1,1]|x[1,0]>; r5 = <x[1,3]|x[1,2]>
1.1256 + SMULWT r9, r3, r11 ; r9 = t[0,6]=OC_C3S5*x[0,3]>>16
1.1257 + LDRD r6, OC_C6S2_4_v6 ; r6 = OC_C6S2; r7 = OC_C2S6
1.1258 + SMULWT r8, r2, r11 ; r8 = -t[0,5]=OC_C5S3*x[0,3]>>16
1.1259 +; Here we cheat: row 3 had just a DC, so x[0,3]==x[1,3] by definition.
1.1260 + PKHBT r9, r9, r9, LSL #16 ; r9 = <t[0,6]|t[0,6]>
1.1261 + SMULWB r3, r7, r11 ; r3 = t[0,3]=OC_C2S6*x[0,2]>>16
1.1262 + PKHBT r8, r8, r8, LSL #16 ; r8 = <-t[0,5]|-t[0,5]>
1.1263 + SMULWB r2, r6, r11 ; r2 = t[0,2]=OC_C6S2*x[0,2]>>16
1.1264 + LDR r11,OC_C4S4
1.1265 + SMULWB r12,r7, r5 ; r12= t[1,3]=OC_C2S6*x[1,2]>>16
1.1266 + MOV r7,#8
1.1267 + SMULWB r5, r6, r5 ; r5 = t[1,2]=OC_C6S2*x[1,2]>>16
1.1268 + PKHBT r3, r3, r12,LSL #16 ; r3 = <t[1,3]|t[0,3]>
1.1269 + SMLAWB r12,r11,r10,r7 ; r12= t[0,0]+8=(OC_C4S4*x[0,0]>>16)+8
1.1270 + PKHBT r2, r2, r5, LSL #16 ; r2 = <t[1,2]|t[0,2]>
1.1271 + SMLAWB r5, r11,r4 ,r7 ; r5 = t[1,0]+8=(OC_C4S4*x[1,0]>>16)+8
1.1272 + LDRD r6, OC_C7S1_4_v6 ; r6 = OC_C7S1; r7 = OC_C1S7
1.1273 + PKHBT r12,r12,r5, LSL #16 ; r12= <t[1,0]+8|t[0,0]+8>
1.1274 + SMULWT r5, r7, r4 ; r5 = t[1,7]=OC_C1S7*x[1,1]>>16
1.1275 + SMULWT r7, r7, r10 ; r7 = t[0,7]=OC_C1S7*x[0,1]>>16
1.1276 + SMULWT r10,r6, r10 ; r10= t[0,4]=OC_C7S1*x[0,1]>>16
1.1277 + PKHBT r7, r7, r5, LSL #16 ; r7 = <t[1,7]|t[0,7]>
1.1278 + SMULWT r4, r6, r4 ; r4 = t[1,4]=OC_C7S1*x[1,1]>>16
1.1279 +; Stage 2:
1.1280 + SSUB16 r6, r7, r9 ; r6 = t[7]-t[6]
1.1281 + PKHBT r4, r10,r4, LSL #16 ; r4 = <t[1,4]|t[0,4]>
1.1282 + SADD16 r7, r7, r9 ; r7 = t[7]=t[7]+t[6]
1.1283 + SMULWT r9, r11,r6 ; r9 = t[1,6]=OC_C4S4*r6T>>16
1.1284 + SADD16 r5, r4, r8 ; r5 = t[4]-t[5]
1.1285 + SMULWB r6, r11,r6 ; r6 = t[0,6]=OC_C4S4*r6B>>16
1.1286 + SSUB16 r4, r4, r8 ; r4 = t[4]=t[4]+t[5]
1.1287 + SMULWT r10,r11,r5 ; r10= t[1,5]=OC_C4S4*r5T>>16
1.1288 + PKHBT r6, r6, r9, LSL #16 ; r6 = <t[1,6]|t[0,6]>
1.1289 + SMULWB r5, r11,r5 ; r5 = t[0,5]=OC_C4S4*r5B>>16
1.1290 +; Stage 3:
1.1291 +idct4_4core_down_stage3_v6
1.1292 + SADD16 r11,r12,r2 ; r11= t[1]+8=t[0]+t[2]+8
1.1293 + PKHBT r10,r5, r10,LSL #16 ; r10= <t[1,5]|t[0,5]>
1.1294 + SSUB16 r2, r12,r2 ; r2 = t[2]+8=t[0]-t[2]+8
1.1295 + B idct8_8core_down_stage3_5_v6
1.1296 + ENDP
1.1297 +
1.1298 +idct8_8core_v6 PROC
1.1299 + STMFD r13!,{r0,r14}
1.1300 +; Stage 1:
1.1301 + ;5-6 rotation by 3pi/16
1.1302 + LDRD r10,OC_C5S3_4_v6 ; r10= OC_C5S3, r11= OC_C3S5
1.1303 + LDR r4, [r1,#8] ; r4 = <x[0,5]|x[0,4]>
1.1304 + LDR r7, [r1,#24] ; r7 = <x[1,5]|x[1,4]>
1.1305 + SMULWT r5, r11,r4 ; r5 = OC_C3S5*x[0,5]>>16
1.1306 + LDR r0, [r1,#4] ; r0 = <x[0,3]|x[0,2]>
1.1307 + SMULWT r3, r11,r7 ; r3 = OC_C3S5*x[1,5]>>16
1.1308 + LDR r12,[r1,#20] ; r12= <x[1,3]|x[1,2]>
1.1309 + SMULWT r6, r11,r0 ; r6 = OC_C3S5*x[0,3]>>16
1.1310 + SMULWT r11,r11,r12 ; r11= OC_C3S5*x[1,3]>>16
1.1311 + SMLAWT r6, r10,r4, r6 ; r6 = t[0,6]=r6+(OC_C5S3*x[0,5]>>16)
1.1312 + PKHBT r5, r5, r3, LSL #16 ; r5 = <r3|r5>
1.1313 + SMLAWT r11,r10,r7, r11 ; r11= t[1,6]=r11+(OC_C5S3*x[1,5]>>16)
1.1314 + PKHBT r4, r4, r7, LSL #16 ; r4 = <x[1,4]|x[0,4]>
1.1315 + SMULWT r3, r10,r0 ; r3 = OC_C5S3*x[0,3]>>16
1.1316 + PKHBT r6, r6, r11,LSL #16 ; r6 = <t[1,6]|t[0,6]>
1.1317 + SMULWT r8, r10,r12 ; r8 = OC_C5S3*x[1,3]>>16
1.1318 + ;2-3 rotation by 6pi/16
1.1319 + LDRD r10,OC_C6S2_4_v6 ; r10= OC_C6S2, r11= OC_C2S6
1.1320 + PKHBT r3, r3, r8, LSL #16 ; r3 = <r8|r3>
1.1321 + LDR r8, [r1,#12] ; r8 = <x[0,7]|x[0,6]>
1.1322 + SMULWB r2, r10,r0 ; r2 = OC_C6S2*x[0,2]>>16
1.1323 + SSUB16 r5, r5, r3 ; r5 = <t[1,5]|t[0,5]>
1.1324 + SMULWB r9, r10,r12 ; r9 = OC_C6S2*x[1,2]>>16
1.1325 + LDR r7, [r1,#28] ; r7 = <x[1,7]|x[1,6]>
1.1326 + SMULWB r3, r10,r8 ; r3 = OC_C6S2*x[0,6]>>16
1.1327 + SMULWB r10,r10,r7 ; r10= OC_C6S2*x[1,6]>>16
1.1328 + PKHBT r2, r2, r9, LSL #16 ; r2 = <r2|r9>
1.1329 + SMLAWB r3, r11,r0, r3 ; r3 = t[0,3]=r3+(OC_C2S6*x[0,2]>>16)
1.1330 + SMLAWB r10,r11,r12,r10 ; r10= t[1,3]=r10+(OC_C2S6*x[1,2]>>16)
1.1331 + SMULWB r9, r11,r8 ; r9 = OC_C2S6*x[0,6]>>16
1.1332 + PKHBT r3, r3, r10,LSL #16 ; r3 = <t[1,6]|t[0,6]>
1.1333 + SMULWB r12,r11,r7 ; r12= OC_C2S6*x[1,6]>>16
1.1334 + ;4-7 rotation by 7pi/16
1.1335 + LDRD r10,OC_C7S1_8_v6 ; r10= OC_C7S1, r11= OC_C1S7
1.1336 + PKHBT r9, r9, r12,LSL #16 ; r9 = <r9|r12>
1.1337 + LDR r0, [r1],#16 ; r0 = <x[0,1]|x[0,0]>
1.1338 + PKHTB r7, r7, r8, ASR #16 ; r7 = <x[1,7]|x[0,7]>
1.1339 + SSUB16 r2, r2, r9 ; r2 = <t[1,2]|t[0,2]>
1.1340 + SMULWB r9, r10,r7 ; r9 = OC_C7S1*x[0,7]>>16
1.1341 + LDR r14,[r1],#16 ; r14= <x[1,1]|x[1,0]>
1.1342 + SMULWT r12,r10,r7 ; r12= OC_C7S1*x[1,7]>>16
1.1343 + SMULWT r8, r10,r0 ; r8 = OC_C7S1*x[0,1]>>16
1.1344 + SMULWT r10,r10,r14 ; r10= OC_C7S1*x[1,1]>>16
1.1345 + SMLAWT r9, r11,r0, r9 ; r9 = t[0,7]=r9+(OC_C1S7*x[0,1]>>16)
1.1346 + PKHBT r8, r8, r10,LSL #16 ; r8 = <r12|r8>
1.1347 + SMLAWT r12,r11,r14,r12 ; r12= t[1,7]=r12+(OC_C1S7*x[1,1]>>16)
1.1348 + PKHBT r0, r0, r14,LSL #16 ; r0 = <x[1,0]|x[0,0]>
1.1349 + SMULWB r10,r11,r7 ; r10= OC_C1S7*x[0,6]>>16
1.1350 + PKHBT r9, r9, r12,LSL #16 ; r9 = <t[1,7]|t[0,7]>
1.1351 + SMULWT r12,r11,r7 ; r12= OC_C1S7*x[1,6]>>16
1.1352 + ;0-1 butterfly
1.1353 + LDR r11,OC_C4S4
1.1354 + PKHBT r10,r10,r12,LSL #16 ; r10= <r12|r10>
1.1355 + SADD16 r7, r0, r4 ; r7 = x[0]+x[4]
1.1356 + SSUB16 r10,r8, r10 ; r10= <t[1,4]|t[0,4]>
1.1357 + SSUB16 r4, r0, r4 ; r4 = x[0]-x[4]
1.1358 + SMULWB r8, r11,r7 ; r8 = t[0,0]=OC_C4S4*r7B>>16
1.1359 + SMULWT r12,r11,r7 ; r12= t[1,0]=OC_C4S4*r7T>>16
1.1360 + SMULWB r7, r11,r4 ; r7 = t[0,1]=OC_C4S4*r4B>>16
1.1361 + PKHBT r12,r8, r12,LSL #16 ; r12= <t[1,0]|t[0,0]>
1.1362 + SMULWT r8, r11,r4 ; r8 = t[1,1]=OC_C4S4*r4T>>16
1.1363 +; Stage 2:
1.1364 + SADD16 r4, r10,r5 ; r4 = t[4]'=t[4]+t[5]
1.1365 + PKHBT r8, r7, r8, LSL #16 ; r8 = <t[1,0]|t[0,0]>
1.1366 + SSUB16 r5, r10,r5 ; r5 = t[4]-t[5]
1.1367 + SMULWB r10,r11,r5 ; r10= t[0,5]=OC_C4S4*r5B>>16
1.1368 + SADD16 r7, r9, r6 ; r7 = t[7]'=t[7]+t[6]
1.1369 + SMULWT r5, r11,r5 ; r5 = t[1,5]=OC_C4S4*r5T>>16
1.1370 + SSUB16 r6, r9, r6 ; r6 = t[7]-t[6]
1.1371 + SMULWB r9, r11,r6 ; r9 = t[0,6]=OC_C4S4*r6B>>16
1.1372 + PKHBT r10,r10,r5, LSL #16 ; r10= <t[1,5]|t[0,5]>
1.1373 + SMULWT r6, r11,r6 ; r6 = t[1,6]=OC_C4S4*r6T>>16
1.1374 +; Stage 3:
1.1375 + SADD16 r11,r8, r2 ; r11= t[1]'=t[1]+t[2]
1.1376 + PKHBT r6, r9, r6, LSL #16 ; r6 = <t[1,6]|t[0,6]>
1.1377 + SSUB16 r2, r8, r2 ; r2 = t[2]=t[1]-t[2]
1.1378 + LDMFD r13!,{r0,r14}
1.1379 + B idct4_3core_stage3_5_v6
1.1380 + ENDP
1.1381 +
1.1382 +; Another copy so the LDRD offsets are less than +/- 255.
1.1383 + ALIGN 8
1.1384 +OC_C7S1_8_v6
1.1385 + DCD 12785 ; 31F1
1.1386 +OC_C1S7_8_v6
1.1387 + DCD 64277 ; FB15
1.1388 +OC_C6S2_8_v6
1.1389 + DCD 25080 ; 61F8
1.1390 +OC_C2S6_8_v6
1.1391 + DCD 60547 ; EC83
1.1392 +OC_C5S3_8_v6
1.1393 + DCD 36410 ; 8E3A
1.1394 +OC_C3S5_8_v6
1.1395 + DCD 54491 ; D4DB
1.1396 +
1.1397 +idct8_8core_down_v6 PROC
1.1398 + STMFD r13!,{r0,r14}
1.1399 +; Stage 1:
1.1400 + ;5-6 rotation by 3pi/16
1.1401 + LDRD r10,OC_C5S3_8_v6 ; r10= OC_C5S3, r11= OC_C3S5
1.1402 + LDR r4, [r1,#8] ; r4 = <x[0,5]|x[0,4]>
1.1403 + LDR r7, [r1,#24] ; r7 = <x[1,5]|x[1,4]>
1.1404 + SMULWT r5, r11,r4 ; r5 = OC_C3S5*x[0,5]>>16
1.1405 + LDR r0, [r1,#4] ; r0 = <x[0,3]|x[0,2]>
1.1406 + SMULWT r3, r11,r7 ; r3 = OC_C3S5*x[1,5]>>16
1.1407 + LDR r12,[r1,#20] ; r12= <x[1,3]|x[1,2]>
1.1408 + SMULWT r6, r11,r0 ; r6 = OC_C3S5*x[0,3]>>16
1.1409 + SMULWT r11,r11,r12 ; r11= OC_C3S5*x[1,3]>>16
1.1410 + SMLAWT r6, r10,r4, r6 ; r6 = t[0,6]=r6+(OC_C5S3*x[0,5]>>16)
1.1411 + PKHBT r5, r5, r3, LSL #16 ; r5 = <r3|r5>
1.1412 + SMLAWT r11,r10,r7, r11 ; r11= t[1,6]=r11+(OC_C5S3*x[1,5]>>16)
1.1413 + PKHBT r4, r4, r7, LSL #16 ; r4 = <x[1,4]|x[0,4]>
1.1414 + SMULWT r3, r10,r0 ; r3 = OC_C5S3*x[0,3]>>16
1.1415 + PKHBT r6, r6, r11,LSL #16 ; r6 = <t[1,6]|t[0,6]>
1.1416 + SMULWT r8, r10,r12 ; r8 = OC_C5S3*x[1,3]>>16
1.1417 + ;2-3 rotation by 6pi/16
1.1418 + LDRD r10,OC_C6S2_8_v6 ; r10= OC_C6S2, r11= OC_C2S6
1.1419 + PKHBT r3, r3, r8, LSL #16 ; r3 = <r8|r3>
1.1420 + LDR r8, [r1,#12] ; r8 = <x[0,7]|x[0,6]>
1.1421 + SMULWB r2, r10,r0 ; r2 = OC_C6S2*x[0,2]>>16
1.1422 + SSUB16 r5, r5, r3 ; r5 = <t[1,5]|t[0,5]>
1.1423 + SMULWB r9, r10,r12 ; r9 = OC_C6S2*x[1,2]>>16
1.1424 + LDR r7, [r1,#28] ; r7 = <x[1,7]|x[1,6]>
1.1425 + SMULWB r3, r10,r8 ; r3 = OC_C6S2*x[0,6]>>16
1.1426 + SMULWB r10,r10,r7 ; r10= OC_C6S2*x[1,6]>>16
1.1427 + PKHBT r2, r2, r9, LSL #16 ; r2 = <r2|r9>
1.1428 + SMLAWB r3, r11,r0, r3 ; r3 = t[0,3]=r3+(OC_C2S6*x[0,2]>>16)
1.1429 + SMLAWB r10,r11,r12,r10 ; r10= t[1,3]=r10+(OC_C2S6*x[1,2]>>16)
1.1430 + SMULWB r9, r11,r8 ; r9 = OC_C2S6*x[0,6]>>16
1.1431 + PKHBT r3, r3, r10,LSL #16 ; r3 = <t[1,6]|t[0,6]>
1.1432 + SMULWB r12,r11,r7 ; r12= OC_C2S6*x[1,6]>>16
1.1433 + ;4-7 rotation by 7pi/16
1.1434 + LDRD r10,OC_C7S1_8_v6 ; r10= OC_C7S1, r11= OC_C1S7
1.1435 + PKHBT r9, r9, r12,LSL #16 ; r9 = <r9|r12>
1.1436 + LDR r0, [r1],#16 ; r0 = <x[0,1]|x[0,0]>
1.1437 + PKHTB r7, r7, r8, ASR #16 ; r7 = <x[1,7]|x[0,7]>
1.1438 + SSUB16 r2, r2, r9 ; r2 = <t[1,2]|t[0,2]>
1.1439 + SMULWB r9, r10,r7 ; r9 = OC_C7S1*x[0,7]>>16
1.1440 + LDR r14,[r1],#16 ; r14= <x[1,1]|x[1,0]>
1.1441 + SMULWT r12,r10,r7 ; r12= OC_C7S1*x[1,7]>>16
1.1442 + SMULWT r8, r10,r0 ; r8 = OC_C7S1*x[0,1]>>16
1.1443 + SMULWT r10,r10,r14 ; r10= OC_C7S1*x[1,1]>>16
1.1444 + SMLAWT r9, r11,r0, r9 ; r9 = t[0,7]=r9+(OC_C1S7*x[0,1]>>16)
1.1445 + PKHBT r8, r8, r10,LSL #16 ; r8 = <r12|r8>
1.1446 + SMLAWT r12,r11,r14,r12 ; r12= t[1,7]=r12+(OC_C1S7*x[1,1]>>16)
1.1447 + PKHBT r0, r0, r14,LSL #16 ; r0 = <x[1,0]|x[0,0]>
1.1448 + SMULWB r10,r11,r7 ; r10= OC_C1S7*x[0,6]>>16
1.1449 + PKHBT r9, r9, r12,LSL #16 ; r9 = <t[1,7]|t[0,7]>
1.1450 + SMULWT r12,r11,r7 ; r12= OC_C1S7*x[1,6]>>16
1.1451 + ;0-1 butterfly
1.1452 + LDR r11,OC_C4S4
1.1453 + MOV r14,#8
1.1454 + PKHBT r10,r10,r12,LSL #16 ; r10= <r12|r10>
1.1455 + SADD16 r7, r0, r4 ; r7 = x[0]+x[4]
1.1456 + SSUB16 r10,r8, r10 ; r10= <t[1,4]|t[0,4]>
1.1457 + SMLAWB r8, r11,r7, r14 ; r8 = t[0,0]+8=(OC_C4S4*r7B>>16)+8
1.1458 + SSUB16 r4, r0, r4 ; r4 = x[0]-x[4]
1.1459 + SMLAWT r12,r11,r7, r14 ; r12= t[1,0]+8=(OC_C4S4*r7T>>16)+8
1.1460 + SMLAWB r7, r11,r4, r14 ; r7 = t[0,1]+8=(OC_C4S4*r4B>>16)+8
1.1461 + PKHBT r12,r8, r12,LSL #16 ; r12= <t[1,0]+8|t[0,0]+8>
1.1462 + SMLAWT r8, r11,r4, r14 ; r8 = t[1,1]+8=(OC_C4S4*r4T>>16)+8
1.1463 +; Stage 2:
1.1464 + SADD16 r4, r10,r5 ; r4 = t[4]'=t[4]+t[5]
1.1465 + PKHBT r8, r7, r8, LSL #16 ; r8 = <t[1,0]+8|t[0,0]+8>
1.1466 + SSUB16 r5, r10,r5 ; r5 = t[4]-t[5]
1.1467 + SMULWB r10,r11,r5 ; r10= t[0,5]=OC_C4S4*r5B>>16
1.1468 + SADD16 r7, r9, r6 ; r7 = t[7]'=t[7]+t[6]
1.1469 + SMULWT r5, r11,r5 ; r5 = t[1,5]=OC_C4S4*r5T>>16
1.1470 + SSUB16 r6, r9, r6 ; r6 = t[7]-t[6]
1.1471 + SMULWB r9, r11,r6 ; r9 = t[0,6]=OC_C4S4*r6B>>16
1.1472 + PKHBT r10,r10,r5, LSL #16 ; r10= <t[1,5]|t[0,5]>
1.1473 + SMULWT r6, r11,r6 ; r6 = t[1,6]=OC_C4S4*r6T>>16
1.1474 +; Stage 3:
1.1475 + SADD16 r11,r8, r2 ; r11= t[1]'+8=t[1]+t[2]+8
1.1476 + PKHBT r6, r9, r6, LSL #16 ; r6 = <t[1,6]|t[0,6]>
1.1477 + SSUB16 r2, r8, r2 ; r2 = t[2]+8=t[1]-t[2]+8
1.1478 + LDMFD r13!,{r0,r14}
1.1479 +idct8_8core_down_stage3_5_v6
1.1480 + SSUB16 r5, r6, r10 ; r5 = t[5]'=t[6]-t[5]
1.1481 + SADD16 r6, r6, r10 ; r6 = t[6]=t[6]+t[5]
1.1482 + SADD16 r10,r12,r3 ; r10= t[0]'+8=t[0]+t[3]+8
1.1483 + SSUB16 r3, r12,r3 ; r3 = t[3]+8=t[0]-t[3]+8
1.1484 +; Stage 4:
1.1485 + SADD16 r12,r10,r7 ; r12= t[0]+t[7]+8
1.1486 + SSUB16 r7, r10,r7 ; r7 = t[0]-t[7]+8
1.1487 + MOV r10,r12,ASR #4
1.1488 + MOV r12,r12,LSL #16
1.1489 + PKHTB r10,r10,r12,ASR #20 ; r10= t[0]+t[7]+8>>4
1.1490 + STR r10,[r0], #4 ; y[0<<3] = t[0]+t[7]+8>>4
1.1491 + SADD16 r12,r11,r6 ; r12= t[1]+t[6]+8
1.1492 + SSUB16 r6, r11,r6 ; r6 = t[1]-t[6]+8
1.1493 + MOV r10,r12,ASR #4
1.1494 + MOV r12,r12,LSL #16
1.1495 + PKHTB r10,r10,r12,ASR #20 ; r10= t[1]+t[6]+8>>4
1.1496 + STR r10,[r0, #12] ; y[1<<3] = t[1]+t[6]+8>>4
1.1497 + SADD16 r12,r2, r5 ; r12= t[2]+t[5]+8
1.1498 + SSUB16 r5, r2, r5 ; r5 = t[2]-t[5]+8
1.1499 + MOV r10,r12,ASR #4
1.1500 + MOV r12,r12,LSL #16
1.1501 + PKHTB r10,r10,r12,ASR #20 ; r10= t[2]+t[5]+8>>4
1.1502 + STR r10,[r0, #28] ; y[2<<3] = t[2]+t[5]+8>>4
1.1503 + SADD16 r12,r3, r4 ; r12= t[3]+t[4]+8
1.1504 + SSUB16 r4, r3, r4 ; r4 = t[3]-t[4]+8
1.1505 + MOV r10,r12,ASR #4
1.1506 + MOV r12,r12,LSL #16
1.1507 + PKHTB r10,r10,r12,ASR #20 ; r10= t[3]+t[4]+8>>4
1.1508 + STR r10,[r0, #44] ; y[3<<3] = t[3]+t[4]+8>>4
1.1509 + MOV r10,r4, ASR #4
1.1510 + MOV r4, r4, LSL #16
1.1511 + PKHTB r10,r10,r4, ASR #20 ; r10= t[3]-t[4]+8>>4
1.1512 + STR r10,[r0, #60] ; y[4<<3] = t[3]-t[4]+8>>4
1.1513 + MOV r10,r5, ASR #4
1.1514 + MOV r5, r5, LSL #16
1.1515 + PKHTB r10,r10,r5, ASR #20 ; r10= t[2]-t[5]+8>>4
1.1516 + STR r10,[r0, #76] ; y[5<<3] = t[2]-t[5]+8>>4
1.1517 + MOV r10,r6, ASR #4
1.1518 + MOV r6, r6, LSL #16
1.1519 + PKHTB r10,r10,r6, ASR #20 ; r10= t[1]-t[6]+8>>4
1.1520 + STR r10,[r0, #92] ; y[6<<3] = t[1]-t[6]+8>>4
1.1521 + MOV r10,r7, ASR #4
1.1522 + MOV r7, r7, LSL #16
1.1523 + PKHTB r10,r10,r7, ASR #20 ; r10= t[0]-t[7]+8>>4
1.1524 + STR r10,[r0, #108] ; y[7<<3] = t[0]-t[7]+8>>4
1.1525 + MOV PC,r14
1.1526 + ENDP
1.1527 + ]
1.1528 +
1.1529 + [ OC_ARM_ASM_NEON
1.1530 + EXPORT oc_idct8x8_1_neon
1.1531 + EXPORT oc_idct8x8_neon
1.1532 +
1.1533 + ALIGN 16
1.1534 +OC_IDCT_CONSTS_NEON
1.1535 + DCW 8
1.1536 + DCW 64277 ; FB15 (C1S7)
1.1537 + DCW 60547 ; EC83 (C2S6)
1.1538 + DCW 54491 ; D4DB (C3S5)
1.1539 + DCW 46341 ; B505 (C4S4)
1.1540 + DCW 36410 ; 471D (C5S3)
1.1541 + DCW 25080 ; 30FC (C6S2)
1.1542 + DCW 12785 ; 31F1 (C7S1)
1.1543 +
1.1544 +oc_idct8x8_1_neon PROC
1.1545 + ; r0 = ogg_int16_t *_y
1.1546 + ; r1 = ogg_uint16_t _dc
1.1547 + VDUP.S16 Q0, r1
1.1548 + VMOV Q1, Q0
1.1549 + VST1.64 {D0, D1, D2, D3}, [r0@128]!
1.1550 + VST1.64 {D0, D1, D2, D3}, [r0@128]!
1.1551 + VST1.64 {D0, D1, D2, D3}, [r0@128]!
1.1552 + VST1.64 {D0, D1, D2, D3}, [r0@128]
1.1553 + MOV PC, r14
1.1554 + ENDP
1.1555 +
1.1556 +oc_idct8x8_neon PROC
1.1557 + ; r0 = ogg_int16_t *_y
1.1558 + ; r1 = ogg_int16_t *_x
1.1559 + ; r2 = int _last_zzi
1.1560 + CMP r2, #10
1.1561 + BLE oc_idct8x8_10_neon
1.1562 +oc_idct8x8_slow_neon
1.1563 + VPUSH {D8-D15}
1.1564 + MOV r2, r1
1.1565 + ADR r3, OC_IDCT_CONSTS_NEON
1.1566 + ; Row transforms (input is pre-transposed)
1.1567 + VLD1.64 {D16,D17,D18,D19}, [r2@128]!
1.1568 + VLD1.64 {D20,D21,D22,D23}, [r2@128]!
1.1569 + VLD1.64 {D24,D25,D26,D27}, [r2@128]!
1.1570 + VSUB.S16 Q1, Q8, Q12 ; Q8 = x[0]-x[4]
1.1571 + VLD1.64 {D28,D29,D30,D31}, [r2@128]
1.1572 + VADD.S16 Q8, Q8, Q12 ; Q1 = x[0]+x[4]
1.1573 + VLD1.64 {D0,D1}, [r3@128]
1.1574 + MOV r12, r14
1.1575 + BL oc_idct8x8_stage123_neon
1.1576 +; Stage 4
1.1577 + VSUB.S16 Q15,Q8, Q7 ; Q15 = y[7]=t[0]'-t[7]'
1.1578 + VADD.S16 Q8, Q8, Q7 ; Q8 = y[0]=t[0]'+t[7]'
1.1579 + VSUB.S16 Q14,Q9, Q3 ; Q14 = y[6]=t[1]'-t[6]''
1.1580 + VADD.S16 Q9, Q9, Q3 ; Q9 = y[1]=t[1]'+t[6]''
1.1581 + VSUB.S16 Q13,Q10,Q5 ; Q13 = y[5]=t[2]'-t[5]''
1.1582 + VADD.S16 Q10,Q10,Q5 ; Q10 = y[2]=t[2]'+t[5]''
1.1583 + VTRN.16 Q14,Q15
1.1584 + VSUB.S16 Q12,Q11,Q4 ; Q12 = y[4]=t[3]'-t[4]'
1.1585 + VADD.S16 Q11,Q11,Q4 ; Q11 = y[3]=t[3]'+t[4]'
1.1586 + ; 8x8 Transpose
1.1587 + VTRN.16 Q8, Q9
1.1588 + VTRN.16 Q10,Q11
1.1589 + VTRN.16 Q12,Q13
1.1590 + VTRN.32 Q8, Q10
1.1591 + VTRN.32 Q9, Q11
1.1592 + VTRN.32 Q12,Q14
1.1593 + VTRN.32 Q13,Q15
1.1594 + VSWP D17,D24
1.1595 + VSUB.S16 Q1, Q8, Q12 ; Q8 = x[0]-x[4]
1.1596 + VSWP D19,D26
1.1597 + VADD.S16 Q8, Q8, Q12 ; Q1 = x[0]+x[4]
1.1598 + VSWP D21,D28
1.1599 + VSWP D23,D30
1.1600 + ; Column transforms
1.1601 + BL oc_idct8x8_stage123_neon
1.1602 + CMP r0,r1
1.1603 + ; We have to put the return address back in the LR, or the branch
1.1604 + ; predictor will not recognize the function return and mis-predict the
1.1605 + ; entire call stack.
1.1606 + MOV r14, r12
1.1607 +; Stage 4
1.1608 + VSUB.S16 Q15,Q8, Q7 ; Q15 = y[7]=t[0]'-t[7]'
1.1609 + VADD.S16 Q8, Q8, Q7 ; Q8 = y[0]=t[0]'+t[7]'
1.1610 + VSUB.S16 Q14,Q9, Q3 ; Q14 = y[6]=t[1]'-t[6]''
1.1611 + VADD.S16 Q9, Q9, Q3 ; Q9 = y[1]=t[1]'+t[6]''
1.1612 + VSUB.S16 Q13,Q10,Q5 ; Q13 = y[5]=t[2]'-t[5]''
1.1613 + VADD.S16 Q10,Q10,Q5 ; Q10 = y[2]=t[2]'+t[5]''
1.1614 + VSUB.S16 Q12,Q11,Q4 ; Q12 = y[4]=t[3]'-t[4]'
1.1615 + VADD.S16 Q11,Q11,Q4 ; Q11 = y[3]=t[3]'+t[4]'
1.1616 + BEQ oc_idct8x8_slow_neon_noclear
1.1617 + VMOV.I8 Q2,#0
1.1618 + VPOP {D8-D15}
1.1619 + VMOV.I8 Q3,#0
1.1620 + VRSHR.S16 Q8, Q8, #4 ; Q8 = y[0]+8>>4
1.1621 + VST1.64 {D4, D5, D6, D7}, [r1@128]!
1.1622 + VRSHR.S16 Q9, Q9, #4 ; Q9 = y[1]+8>>4
1.1623 + VRSHR.S16 Q10,Q10,#4 ; Q10 = y[2]+8>>4
1.1624 + VST1.64 {D4, D5, D6, D7}, [r1@128]!
1.1625 + VRSHR.S16 Q11,Q11,#4 ; Q11 = y[3]+8>>4
1.1626 + VRSHR.S16 Q12,Q12,#4 ; Q12 = y[4]+8>>4
1.1627 + VST1.64 {D4, D5, D6, D7}, [r1@128]!
1.1628 + VRSHR.S16 Q13,Q13,#4 ; Q13 = y[5]+8>>4
1.1629 + VRSHR.S16 Q14,Q14,#4 ; Q14 = y[6]+8>>4
1.1630 + VST1.64 {D4, D5, D6, D7}, [r1@128]
1.1631 + VRSHR.S16 Q15,Q15,#4 ; Q15 = y[7]+8>>4
1.1632 + VSTMIA r0, {D16-D31}
1.1633 + MOV PC, r14
1.1634 +
1.1635 +oc_idct8x8_slow_neon_noclear
1.1636 + VPOP {D8-D15}
1.1637 + VRSHR.S16 Q8, Q8, #4 ; Q8 = y[0]+8>>4
1.1638 + VRSHR.S16 Q9, Q9, #4 ; Q9 = y[1]+8>>4
1.1639 + VRSHR.S16 Q10,Q10,#4 ; Q10 = y[2]+8>>4
1.1640 + VRSHR.S16 Q11,Q11,#4 ; Q11 = y[3]+8>>4
1.1641 + VRSHR.S16 Q12,Q12,#4 ; Q12 = y[4]+8>>4
1.1642 + VRSHR.S16 Q13,Q13,#4 ; Q13 = y[5]+8>>4
1.1643 + VRSHR.S16 Q14,Q14,#4 ; Q14 = y[6]+8>>4
1.1644 + VRSHR.S16 Q15,Q15,#4 ; Q15 = y[7]+8>>4
1.1645 + VSTMIA r0, {D16-D31}
1.1646 + MOV PC, r14
1.1647 + ENDP
1.1648 +
1.1649 +oc_idct8x8_stage123_neon PROC
1.1650 +; Stages 1 & 2
1.1651 + VMULL.S16 Q4, D18,D1[3]
1.1652 + VMULL.S16 Q5, D19,D1[3]
1.1653 + VMULL.S16 Q7, D30,D1[3]
1.1654 + VMULL.S16 Q6, D31,D1[3]
1.1655 + VMULL.S16 Q2, D30,D0[1]
1.1656 + VMULL.S16 Q3, D31,D0[1]
1.1657 + VSHRN.S32 D8, Q4, #16
1.1658 + VSHRN.S32 D9, Q5, #16 ; Q4 = (OC_C7S1*x[1]>>16)
1.1659 + VSHRN.S32 D14,Q7, #16
1.1660 + VSHRN.S32 D15,Q6, #16 ; Q7 = (OC_C7S1*x[7]>>16)
1.1661 + VSHRN.S32 D4, Q2, #16
1.1662 + VSHRN.S32 D5, Q3, #16 ; Q2 = (OC_C1S7*x[7]>>16)-x[7]
1.1663 + VSUB.S16 Q4, Q4, Q15
1.1664 + VADD.S16 Q7, Q7, Q9
1.1665 + VSUB.S16 Q4, Q4, Q2 ; Q4 = t[4]
1.1666 + VMULL.S16 Q2, D18,D0[1]
1.1667 + VMULL.S16 Q9, D19,D0[1]
1.1668 + VMULL.S16 Q5, D26,D0[3]
1.1669 + VMULL.S16 Q3, D27,D0[3]
1.1670 + VMULL.S16 Q6, D22,D0[3]
1.1671 + VMULL.S16 Q12,D23,D0[3]
1.1672 + VSHRN.S32 D4, Q2, #16
1.1673 + VSHRN.S32 D5, Q9, #16 ; Q2 = (OC_C1S7*x[1]>>16)-x[1]
1.1674 + VSHRN.S32 D10,Q5, #16
1.1675 + VSHRN.S32 D11,Q3, #16 ; Q5 = (OC_C3S5*x[5]>>16)-x[5]
1.1676 + VSHRN.S32 D12,Q6, #16
1.1677 + VSHRN.S32 D13,Q12,#16 ; Q6 = (OC_C3S5*x[3]>>16)-x[3]
1.1678 + VADD.S16 Q7, Q7, Q2 ; Q7 = t[7]
1.1679 + VSUB.S16 Q5, Q5, Q11
1.1680 + VADD.S16 Q6, Q6, Q11
1.1681 + VADD.S16 Q5, Q5, Q13
1.1682 + VADD.S16 Q6, Q6, Q13
1.1683 + VMULL.S16 Q9, D22,D1[1]
1.1684 + VMULL.S16 Q11,D23,D1[1]
1.1685 + VMULL.S16 Q15,D26,D1[1]
1.1686 + VMULL.S16 Q13,D27,D1[1]
1.1687 + VMULL.S16 Q2, D20,D1[2]
1.1688 + VMULL.S16 Q12,D21,D1[2]
1.1689 + VSHRN.S32 D18,Q9, #16
1.1690 + VSHRN.S32 D19,Q11,#16 ; Q9 = (OC_C5S3*x[3]>>16)-x[3]
1.1691 + VSHRN.S32 D30,Q15,#16
1.1692 + VSHRN.S32 D31,Q13,#16 ; Q15= (OC_C5S3*x[5]>>16)-x[5]
1.1693 + VSHRN.S32 D4, Q2, #16
1.1694 + VSHRN.S32 D5, Q12,#16 ; Q2 = (OC_C6S2*x[2]>>16)
1.1695 + VSUB.S16 Q5, Q5, Q9 ; Q5 = t[5]
1.1696 + VADD.S16 Q6, Q6, Q15 ; Q6 = t[6]
1.1697 + VSUB.S16 Q2, Q2, Q14
1.1698 + VMULL.S16 Q3, D28,D1[2]
1.1699 + VMULL.S16 Q11,D29,D1[2]
1.1700 + VMULL.S16 Q12,D28,D0[2]
1.1701 + VMULL.S16 Q9, D29,D0[2]
1.1702 + VMULL.S16 Q13,D20,D0[2]
1.1703 + VMULL.S16 Q15,D21,D0[2]
1.1704 + VSHRN.S32 D6, Q3, #16
1.1705 + VSHRN.S32 D7, Q11,#16 ; Q3 = (OC_C6S2*x[6]>>16)
1.1706 + VSHRN.S32 D24,Q12,#16
1.1707 + VSHRN.S32 D25,Q9, #16 ; Q12= (OC_C2S6*x[6]>>16)-x[6]
1.1708 + VSHRN.S32 D26,Q13,#16
1.1709 + VSHRN.S32 D27,Q15,#16 ; Q13= (OC_C2S6*x[2]>>16)-x[2]
1.1710 + VSUB.S16 Q9, Q4, Q5 ; Q9 = t[4]-t[5]
1.1711 + VSUB.S16 Q11,Q7, Q6 ; Q11= t[7]-t[6]
1.1712 + VADD.S16 Q3, Q3, Q10
1.1713 + VADD.S16 Q4, Q4, Q5 ; Q4 = t[4]'=t[4]+t[5]
1.1714 + VADD.S16 Q7, Q7, Q6 ; Q7 = t[7]'=t[7]+t[6]
1.1715 + VSUB.S16 Q2, Q2, Q12 ; Q2 = t[2]
1.1716 + VADD.S16 Q3, Q3, Q13 ; Q3 = t[3]
1.1717 + VMULL.S16 Q12,D16,D1[0]
1.1718 + VMULL.S16 Q13,D17,D1[0]
1.1719 + VMULL.S16 Q14,D2, D1[0]
1.1720 + VMULL.S16 Q15,D3, D1[0]
1.1721 + VMULL.S16 Q5, D18,D1[0]
1.1722 + VMULL.S16 Q6, D22,D1[0]
1.1723 + VSHRN.S32 D24,Q12,#16
1.1724 + VSHRN.S32 D25,Q13,#16
1.1725 + VSHRN.S32 D28,Q14,#16
1.1726 + VSHRN.S32 D29,Q15,#16
1.1727 + VMULL.S16 Q13,D19,D1[0]
1.1728 + VMULL.S16 Q15,D23,D1[0]
1.1729 + VADD.S16 Q8, Q8, Q12 ; Q8 = t[0]
1.1730 + VADD.S16 Q1, Q1, Q14 ; Q1 = t[1]
1.1731 + VSHRN.S32 D10,Q5, #16
1.1732 + VSHRN.S32 D12,Q6, #16
1.1733 + VSHRN.S32 D11,Q13,#16
1.1734 + VSHRN.S32 D13,Q15,#16
1.1735 + VADD.S16 Q5, Q5, Q9 ; Q5 = t[5]'=OC_C4S4*(t[4]-t[5])>>16
1.1736 + VADD.S16 Q6, Q6, Q11 ; Q6 = t[6]'=OC_C4S4*(t[7]-t[6])>>16
1.1737 +; Stage 3
1.1738 + VSUB.S16 Q11,Q8, Q3 ; Q11 = t[3]''=t[0]-t[3]
1.1739 + VADD.S16 Q8, Q8, Q3 ; Q8 = t[0]''=t[0]+t[3]
1.1740 + VADD.S16 Q9, Q1, Q2 ; Q9 = t[1]''=t[1]+t[2]
1.1741 + VADD.S16 Q3, Q6, Q5 ; Q3 = t[6]''=t[6]'+t[5]'
1.1742 + VSUB.S16 Q10,Q1, Q2 ; Q10 = t[2]''=t[1]-t[2]
1.1743 + VSUB.S16 Q5, Q6, Q5 ; Q5 = t[5]''=t[6]'-t[5]'
1.1744 + MOV PC, r14
1.1745 + ENDP
1.1746 +
1.1747 +oc_idct8x8_10_neon PROC
1.1748 + ADR r3, OC_IDCT_CONSTS_NEON
1.1749 + VLD1.64 {D0,D1}, [r3@128]
1.1750 + MOV r2, r1
1.1751 + ; Row transforms (input is pre-transposed)
1.1752 +; Stage 1
1.1753 + VLD1.64 {D16,D17,D18,D19},[r2@128]!
1.1754 + MOV r12, #16
1.1755 + VMULL.S16 Q15,D16,D1[0] ; Q15= OC_C4S4*x[0]-(x[0]<<16)
1.1756 + VLD1.64 {D17}, [r2@64], r12
1.1757 + VMULL.S16 Q2, D18,D0[1] ; Q2 = OC_C1S7*x[1]-(x[1]<<16)
1.1758 + VLD1.64 {D19}, [r2@64]
1.1759 + VMULL.S16 Q14,D17,D0[2] ; Q14= OC_C2S6*x[2]-(x[2]<<16)
1.1760 + VMULL.S16 Q3, D19,D0[3] ; Q3 = OC_C3S5*x[3]-(x[3]<<16)
1.1761 + VMULL.S16 Q13,D19,D1[1] ; Q13= OC_C5S3*x[3]-(x[3]<<16)
1.1762 + VMULL.S16 Q12,D18,D1[3] ; Q12= OC_C7S1*x[1]
1.1763 + VMULL.S16 Q1, D17,D1[2] ; Q1 = OC_C6S2*x[2]
1.1764 + VSHRN.S32 D30,Q15,#16 ; D30= t[0]-x[0]
1.1765 + VSHRN.S32 D4, Q2, #16 ; D4 = t[7]-x[1]
1.1766 + VSHRN.S32 D31,Q14,#16 ; D31= t[3]-x[2]
1.1767 + VSHRN.S32 D6, Q3, #16 ; D6 = t[6]-x[3]
1.1768 + VSHRN.S32 D7, Q13,#16 ; D7 = -t[5]-x[3]
1.1769 + VSHRN.S32 D5, Q12,#16 ; D5 = t[4]
1.1770 + VSHRN.S32 D2, Q1, #16 ; D2 = t[2]
1.1771 + VADD.S16 D4, D4, D18 ; D4 = t[7]
1.1772 + VADD.S16 D6, D6, D19 ; D6 = t[6]
1.1773 + VADD.S16 D7, D7, D19 ; D7 = -t[5]
1.1774 + VADD.S16 Q15,Q15,Q8 ; D30= t[0]
1.1775 + ; D31= t[3]
1.1776 +; Stages 2 & 3
1.1777 + VSUB.S16 Q12,Q2, Q3 ; D24= t[7]-t[6]
1.1778 + ; D25= t[4]'=t[4]+t[5]
1.1779 + VADD.S16 Q13,Q2, Q3 ; D26= t[7]'=t[7]+t[6]
1.1780 + ; D27= t[4]-t[5]
1.1781 + VMULL.S16 Q11,D24,D1[0] ; Q11= OC_C4S4*(t[7]-t[6])
1.1782 + ; -(t[7]-t[6]<<16)
1.1783 + VMULL.S16 Q14,D27,D1[0] ; Q14= OC_C4S4*(t[4]-t[5])
1.1784 + ; -(t[4]-t[5]<<16)
1.1785 + VADD.S16 D16,D30,D31 ; D16= t[0]'=t[0]+t[3]
1.1786 + VSUB.S16 D17,D30,D2 ; D17= t[2]'=t[0]-t[2]
1.1787 + VADD.S16 D18,D30,D2 ; D18= t[1]'=t[0]+t[2]
1.1788 + VSHRN.S32 D22,Q11,#16 ; D22= (OC_C4S4*(t[7]-t[6])>>16)
1.1789 + ; -(t[7]-t[6])
1.1790 + VSHRN.S32 D23,Q14,#16 ; D23= (OC_C4S4*(t[4]-t[5])>>16)
1.1791 + ; -(t[4]-t[5])
1.1792 + VSUB.S16 D19,D30,D31 ; D19= t[3]'=t[0]-t[3]
1.1793 + VADD.S16 D22,D22,D24 ; D22= t[6]'=OC_C4S4*(t[7]-t[6])>>16
1.1794 + VADD.S16 D23,D23,D27 ; D23= t[5]'=OC_C4S4*(t[4]-t[5])>>16
1.1795 + VSUB.S16 D27,D22,D23 ; D27= t[5]''=t[6]'-t[5]'
1.1796 + VADD.S16 D24,D22,D23 ; D24= t[6]''=t[6]'+t[5]'
1.1797 +; Stage 4
1.1798 + VSUB.S16 Q11,Q8, Q13 ; D22= y[7]=t[0]'-t[7]'
1.1799 + ; D23= y[5]=t[2]'-t[5]''
1.1800 + VSUB.S16 Q10,Q9, Q12 ; D20= y[6]=t[1]'-t[6]'
1.1801 + ; D21= y[4]=t[3]'-t[4]''
1.1802 + VADD.S16 Q8, Q8, Q13 ; D16= y[0]=t[0]'+t[7]'
1.1803 + ; D17= y[2]=t[2]'+t[5]''
1.1804 + VADD.S16 Q9, Q9, Q12 ; D18= y[1]=t[1]'-t[6]'
1.1805 + ; D19= y[3]=t[3]'-t[4]''
1.1806 + ; 8x4 transpose
1.1807 + VTRN.16 Q10,Q11 ; Q10= c5c4a5a4 c7c6a7a6
1.1808 + ; Q11= d5d4b5b4 d7d6b7b6
1.1809 + VTRN.16 Q8, Q9 ; Q8 = c3c2a3a2 c1c0a1a0
1.1810 + ; Q9 = d3d2b3b2 d1d0b1b0
1.1811 + VSWP D20,D21 ; Q10= c7c6a7a6 c5c4a5a4
1.1812 + VSWP D22,D23 ; Q11= d7d6b7b6 d5d4b5b4
1.1813 + VUZP.32 Q9, Q11 ; Q9 = b7b6b5b4 b3b2b1b0
1.1814 + ; Q11= d7d6d5d4 d3d2d1d0
1.1815 + VMULL.S16 Q15,D18,D0[1]
1.1816 + VMULL.S16 Q13,D22,D1[1]
1.1817 + VUZP.32 Q8, Q10 ; Q8 = a7a6a5a4 a3a2a1a0
1.1818 + ; Q10= c7c6c5c4 c3c2c1c0
1.1819 + ; Column transforms
1.1820 +; Stages 1, 2, & 3
1.1821 + VMULL.S16 Q14,D19,D0[1] ; Q14:Q15= OC_C1S7*x[1]-(x[1]<<16)
1.1822 + VMULL.S16 Q12,D23,D1[1] ; Q12:Q13= OC_C5S3*x[3]-(x[3]<<16)
1.1823 + VMULL.S16 Q3, D22,D0[3]
1.1824 + VMULL.S16 Q2, D23,D0[3] ; Q2:Q3 = OC_C3S5*x[3]-(x[3]<<16)
1.1825 + VSHRN.S32 D30,Q15,#16
1.1826 + VSHRN.S32 D31,Q14,#16 ; Q15= (OC_C1S7*x[1]>>16)-x[1]
1.1827 + VSHRN.S32 D26,Q13,#16
1.1828 + VSHRN.S32 D27,Q12,#16 ; Q13= (OC_C5S3*x[3]>>16)-x[3]
1.1829 + VSHRN.S32 D28,Q3, #16
1.1830 + VSHRN.S32 D29,Q2, #16 ; Q14= (OC_C3S5*x[3]>>16)-x[3]
1.1831 + VADD.S16 Q15,Q15,Q9 ; Q15= t[7]
1.1832 + VADD.S16 Q13,Q13,Q11 ; Q13= -t[5]
1.1833 + VADD.S16 Q14,Q14,Q11 ; Q14= t[6]
1.1834 + VMULL.S16 Q12,D18,D1[3]
1.1835 + VMULL.S16 Q2, D19,D1[3] ; Q2:Q12= OC_C7S1*x[1]
1.1836 + VMULL.S16 Q1, D16,D1[0]
1.1837 + VMULL.S16 Q11,D17,D1[0] ; Q11:Q1 = OC_C4S4*x[0]-(x[0]<<16)
1.1838 + VMULL.S16 Q3, D20,D0[2]
1.1839 + VMULL.S16 Q9, D21,D0[2] ; Q9:Q3 = OC_C2S6*x[2]-(x[2]<<16)
1.1840 + VSHRN.S32 D24,Q12,#16
1.1841 + VSHRN.S32 D25,Q2, #16 ; Q12= t[4]
1.1842 + VMULL.S16 Q2, D20,D1[2]
1.1843 + VSHRN.S32 D2, Q1, #16
1.1844 + VSHRN.S32 D3, Q11,#16 ; Q1 = (OC_C4S4*x[0]>>16)-x[0]
1.1845 + VMULL.S16 Q11,D21,D1[2] ; Q2:Q11= OC_C6S2*x[2]
1.1846 + VSHRN.S32 D6, Q3, #16
1.1847 + VSHRN.S32 D7, Q9, #16 ; Q3 = (OC_C2S6*x[2]>>16)-x[2]
1.1848 + VSUB.S16 Q9, Q15,Q14 ; Q9 = t[7]-t[6]
1.1849 + VADD.S16 Q15,Q15,Q14 ; Q15= t[7]'=t[7]+t[6]
1.1850 + VSHRN.S32 D4, Q2, #16
1.1851 + VSHRN.S32 D5, Q11,#16 ; Q2 = t[2]
1.1852 + VADD.S16 Q1, Q1, Q8 ; Q1 = t[0]
1.1853 + VADD.S16 Q8, Q12,Q13 ; Q8 = t[4]-t[5]
1.1854 + VADD.S16 Q3, Q3, Q10 ; Q3 = t[3]
1.1855 + VMULL.S16 Q10,D16,D1[0]
1.1856 + VMULL.S16 Q11,D17,D1[0] ; Q11:Q10= OC_C4S4*(t[4]-t[5])
1.1857 + ; -(t[4]-t[5]<<16)
1.1858 + VSUB.S16 Q12,Q12,Q13 ; Q12= t[4]'=t[4]+t[5]
1.1859 + VMULL.S16 Q14,D18,D1[0]
1.1860 + VMULL.S16 Q13,D19,D1[0] ; Q13:Q14= OC_C4S4*(t[6]-t[7])
1.1861 + ; -(t[6]-t[7]<<16)
1.1862 + VSHRN.S32 D20,Q10,#16
1.1863 + VSHRN.S32 D21,Q11,#16 ; Q10= (OC_C4S4*(t[4]-t[5])>>16)
1.1864 + ; -(t[4]-t[5])
1.1865 + VADD.S16 Q11,Q1, Q3 ; Q11= t[0]'=t[0]+t[3]
1.1866 + VSUB.S16 Q3, Q1, Q3 ; Q3 = t[3]'=t[0]-t[3]
1.1867 + VSHRN.S32 D28,Q14,#16
1.1868 + VSHRN.S32 D29,Q13,#16 ; Q14= (OC_C4S4*(t[7]-t[6])>>16)
1.1869 + ; -(t[7]-t[6])
1.1870 + VADD.S16 Q10,Q10,Q8 ; Q10=t[5]'
1.1871 + VADD.S16 Q14,Q14,Q9 ; Q14=t[6]'
1.1872 + VSUB.S16 Q13,Q14,Q10 ; Q13=t[5]''=t[6]'-t[5]'
1.1873 + VADD.S16 Q14,Q14,Q10 ; Q14=t[6]''=t[6]'+t[5]'
1.1874 + VADD.S16 Q10,Q1, Q2 ; Q10= t[1]'=t[0]+t[2]
1.1875 + VSUB.S16 Q2, Q1, Q2 ; Q2 = t[2]'=t[0]-t[2]
1.1876 +; Stage 4
1.1877 + CMP r0, r1
1.1878 + VADD.S16 Q8, Q11,Q15 ; Q8 = y[0]=t[0]'+t[7]'
1.1879 + VADD.S16 Q9, Q10,Q14 ; Q9 = y[1]=t[1]'+t[6]''
1.1880 + VSUB.S16 Q15,Q11,Q15 ; Q15 = y[7]=t[0]'-t[7]'
1.1881 + VSUB.S16 Q14,Q10,Q14 ; Q14 = y[6]=t[1]'-t[6]''
1.1882 + VADD.S16 Q10,Q2, Q13 ; Q10 = y[2]=t[2]'+t[5]''
1.1883 + VADD.S16 Q11,Q3, Q12 ; Q11 = y[3]=t[3]'+t[4]'
1.1884 + VSUB.S16 Q12,Q3, Q12 ; Q12 = y[4]=t[3]'-t[4]'
1.1885 + VSUB.S16 Q13,Q2, Q13 ; Q13 = y[5]=t[2]'-t[5]''
1.1886 + BEQ oc_idct8x8_10_neon_noclear
1.1887 + VMOV.I8 D2, #0
1.1888 + VRSHR.S16 Q8, Q8, #4 ; Q8 = y[0]+8>>4
1.1889 + VST1.64 {D2}, [r1@64], r12
1.1890 + VRSHR.S16 Q9, Q9, #4 ; Q9 = y[1]+8>>4
1.1891 + VRSHR.S16 Q10,Q10,#4 ; Q10 = y[2]+8>>4
1.1892 + VST1.64 {D2}, [r1@64], r12
1.1893 + VRSHR.S16 Q11,Q11,#4 ; Q11 = y[3]+8>>4
1.1894 + VRSHR.S16 Q12,Q12,#4 ; Q12 = y[4]+8>>4
1.1895 + VST1.64 {D2}, [r1@64], r12
1.1896 + VRSHR.S16 Q13,Q13,#4 ; Q13 = y[5]+8>>4
1.1897 + VRSHR.S16 Q14,Q14,#4 ; Q14 = y[6]+8>>4
1.1898 + VST1.64 {D2}, [r1@64]
1.1899 + VRSHR.S16 Q15,Q15,#4 ; Q15 = y[7]+8>>4
1.1900 + VSTMIA r0, {D16-D31}
1.1901 + MOV PC, r14
1.1902 +
1.1903 +oc_idct8x8_10_neon_noclear
1.1904 + VRSHR.S16 Q8, Q8, #4 ; Q8 = y[0]+8>>4
1.1905 + VRSHR.S16 Q9, Q9, #4 ; Q9 = y[1]+8>>4
1.1906 + VRSHR.S16 Q10,Q10,#4 ; Q10 = y[2]+8>>4
1.1907 + VRSHR.S16 Q11,Q11,#4 ; Q11 = y[3]+8>>4
1.1908 + VRSHR.S16 Q12,Q12,#4 ; Q12 = y[4]+8>>4
1.1909 + VRSHR.S16 Q13,Q13,#4 ; Q13 = y[5]+8>>4
1.1910 + VRSHR.S16 Q14,Q14,#4 ; Q14 = y[6]+8>>4
1.1911 + VRSHR.S16 Q15,Q15,#4 ; Q15 = y[7]+8>>4
1.1912 + VSTMIA r0, {D16-D31}
1.1913 + MOV PC, r14
1.1914 + ENDP
1.1915 + ]
1.1916 +
1.1917 + END
