Home | History | Annotate | Download | only in Tremolo 
      1 @ Tremolo library
      2 @-----------------------------------------------------------------------
      3 @ Copyright (C) 2002-2009, Xiph.org Foundation
      4 @ Copyright (C) 2010, Robin Watts for Pinknoise Productions Ltd
      5 @ All rights reserved.
      6 
      7 @ Redistribution and use in source and binary forms, with or without
      8 @ modification, are permitted provided that the following conditions
      9 @ are met:
     10 
     11 @     * Redistributions of source code must retain the above copyright
     12 @ notice, this list of conditions and the following disclaimer.
     13 @     * Redistributions in binary form must reproduce the above
     14 @ copyright notice, this list of conditions and the following disclaimer
     15 @ in the documentation and/or other materials provided with the
     16 @ distribution.
     17 @     * Neither the names of the Xiph.org Foundation nor Pinknoise
     18 @ Productions Ltd nor the names of its contributors may be used to
     19 @ endorse or promote products derived from this software without
     20 @ specific prior written permission.
     21 @
     22 @ THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
     23 @ "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
     24 @ LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
     25 @ A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
     26 @ OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
     27 @ SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
     28 @ LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
     29 @ DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
     30 @ THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
     31 @ (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
     32 @ OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
     33 @ ----------------------------------------------------------------------
     34 
     35     .text
     36 
     37 	@ full accuracy version
     38 
     39 	.global mdct_backwardARM
     40 	.global mdct_shift_right
     41 	.global mdct_unroll_prelap
     42 	.global mdct_unroll_part2
     43 	.global mdct_unroll_part3
     44 	.global mdct_unroll_postlap
     45 
     46 	.extern	sincos_lookup0
     47 	.extern	sincos_lookup1
     48 
     49 mdct_unroll_prelap:
     50 	@ r0 = out
     51 	@ r1 = post
     52 	@ r2 = r
     53 	@ r3 = step
     54 	STMFD	r13!,{r4-r7,r14}
     55 	MVN	r4, #0x8000
     56 	MOV	r3, r3, LSL #1
     57 	SUB	r1, r2, r1		@ r1 = r - post
     58 	SUBS	r1, r1, #16		@ r1 = r - post - 16
     59 	BLT	unroll_over
     60 unroll_loop:
     61 	LDMDB	r2!,{r5,r6,r7,r12}
     62 
     63 	MOV	r5, r5, ASR #9		@ r5 = (*--r)>>9
     64 	MOV	r6, r6, ASR #9		@ r6 = (*--r)>>9
     65 	MOV	r7, r7, ASR #9		@ r7 = (*--r)>>9
     66 	MOV	r12,r12,ASR #9		@ r12= (*--r)>>9
     67 
     68 	MOV	r14,r12,ASR #15
     69 	TEQ	r14,r14,ASR #31		@ if r14==0 || r14==-1 then in range
     70 	EORNE	r12,r4, r14,ASR #31
     71 	STRH	r12,[r0], r3
     72 
     73 	MOV	r14,r7, ASR #15
     74 	TEQ	r14,r14,ASR #31		@ if r14==0 || r14==-1 then in range
     75 	EORNE	r7, r4, r14,ASR #31
     76 	STRH	r7, [r0], r3
     77 
     78 	MOV	r14,r6, ASR #15
     79 	TEQ	r14,r14,ASR #31		@ if r14==0 || r14==-1 then in range
     80 	EORNE	r6, r4, r14,ASR #31
     81 	STRH	r6, [r0], r3
     82 
     83 	MOV	r14,r5, ASR #15
     84 	TEQ	r14,r14,ASR #31		@ if r14==0 || r14==-1 then in range
     85 	EORNE	r5, r4, r14,ASR #31
     86 	STRH	r5, [r0], r3
     87 
     88 	SUBS	r1, r1, #16
     89 	BGE	unroll_loop
     90 
     91 unroll_over:
     92 	ADDS	r1, r1, #16
     93 	BLE	unroll_end
     94 unroll_loop2:
     95 	LDR	r5,[r2,#-4]!
     96 	@ stall
     97 	@ stall (Xscale)
     98 	MOV	r5, r5, ASR #9		@ r5 = (*--r)>>9
     99 	MOV	r14,r5, ASR #15
    100 	TEQ	r14,r14,ASR #31		@ if r14==0 || r14==-1 then in range
    101 	EORNE	r5, r4, r14,ASR #31
    102 	STRH	r5, [r0], r3
    103 	SUBS	r1, r1, #4
    104 	BGT	unroll_loop2
    105 unroll_end:
    106 	LDMFD	r13!,{r4-r7,PC}
    107 
    108 mdct_unroll_postlap:
    109 	@ r0 = out
    110 	@ r1 = post
    111 	@ r2 = l
    112 	@ r3 = step
    113 	STMFD	r13!,{r4-r7,r14}
    114 	MVN	r4, #0x8000
    115 	MOV	r3, r3, LSL #1
    116 	SUB	r1, r1, r2		@ r1 = post - l
    117 	MOV	r1, r1, ASR #1		@ r1 = (post - l)>>1
    118 	SUBS	r1, r1, #16		@ r1 = ((post - l)>>1) - 4
    119 	BLT	unroll_over3
    120 unroll_loop3:
    121 	LDR	r12,[r2],#8
    122 	LDR	r7, [r2],#8
    123 	LDR	r6, [r2],#8
    124 	LDR	r5, [r2],#8
    125 
    126 	RSB	r12,r12,#0
    127 	RSB	r5, r5, #0
    128 	RSB	r6, r6, #0
    129 	RSB	r7, r7, #0
    130 
    131 	MOV	r12, r12,ASR #9		@ r12= (-*l)>>9
    132 	MOV	r5,  r5, ASR #9		@ r5 = (-*l)>>9
    133 	MOV	r6,  r6, ASR #9		@ r6 = (-*l)>>9
    134 	MOV	r7,  r7, ASR #9		@ r7 = (-*l)>>9
    135 
    136 	MOV	r14,r12,ASR #15
    137 	TEQ	r14,r14,ASR #31		@ if r14==0 || r14==-1 then in range
    138 	EORNE	r12,r4, r14,ASR #31
    139 	STRH	r12,[r0], r3
    140 
    141 	MOV	r14,r7, ASR #15
    142 	TEQ	r14,r14,ASR #31		@ if r14==0 || r14==-1 then in range
    143 	EORNE	r7, r4, r14,ASR #31
    144 	STRH	r7, [r0], r3
    145 
    146 	MOV	r14,r6, ASR #15
    147 	TEQ	r14,r14,ASR #31		@ if r14==0 || r14==-1 then in range
    148 	EORNE	r6, r4, r14,ASR #31
    149 	STRH	r6, [r0], r3
    150 
    151 	MOV	r14,r5, ASR #15
    152 	TEQ	r14,r14,ASR #31		@ if r14==0 || r14==-1 then in range
    153 	EORNE	r5, r4, r14,ASR #31
    154 	STRH	r5, [r0], r3
    155 
    156 	SUBS	r1, r1, #16
    157 	BGE	unroll_loop3
    158 
    159 unroll_over3:
    160 	ADDS	r1, r1, #16
    161 	BLE	unroll_over4
    162 unroll_loop4:
    163 	LDR	r5,[r2], #8
    164 	@ stall
    165 	@ stall (Xscale)
    166 	RSB	r5, r5, #0
    167 	MOV	r5, r5, ASR #9		@ r5 = (-*l)>>9
    168 	MOV	r14,r5, ASR #15
    169 	TEQ	r14,r14,ASR #31		@ if r14==0 || r14==-1 then in range
    170 	EORNE	r5, r4, r14,ASR #31
    171 	STRH	r5, [r0], r3
    172 	SUBS	r1, r1, #4
    173 	BGT	unroll_loop4
    174 unroll_over4:
    175 	LDMFD	r13!,{r4-r7,PC}
    176 
    177 mdct_unroll_part2:
    178 	@ r0 = out
    179 	@ r1 = post
    180 	@ r2 = l
    181 	@ r3 = r
    182 	@ <> = step
    183 	@ <> = wL
    184 	@ <> = wR
    185 	MOV	r12,r13
    186 	STMFD	r13!,{r4,r6-r11,r14}
    187 	LDMFD	r12,{r8,r9,r10}		@ r8 = step
    188 					@ r9 = wL
    189 					@ r10= wR
    190 	MVN	r4, #0x8000
    191 	MOV	r8, r8, LSL #1
    192 	SUBS	r1, r3, r1		@ r1 = (r - post)
    193 	BLE	unroll_over5
    194 unroll_loop5:
    195 	LDR	r12,[r2, #-8]!		@ r12= *l       (but l -= 2 first)
    196 	LDR	r11,[r9],#4		@ r11= *wL++
    197 	LDR	r7, [r3, #-4]!		@ r7 = *--r
    198 	LDR	r6, [r10,#-4]!		@ r6 = *--wR
    199 
    200 	@ Can save a cycle here, at the cost of 1bit errors in rounding
    201 	SMULL	r14,r11,r12,r11		@ (r14,r11)  = *l   * *wL++
    202 	SMULL	r14,r6, r7, r6		@ (r14,r6)   = *--r * *--wR
    203 	ADD	r6, r6, r11
    204 	MOV	r6, r6, ASR #8
    205 	MOV	r14,r6, ASR #15
    206 	TEQ	r14,r14,ASR #31		@ if r14==0 || r14==-1 then in range
    207 	EORNE	r6, r4, r14,ASR #31
    208 	STRH	r6, [r0], r8
    209 
    210 	SUBS	r1, r1, #4
    211 	BGT	unroll_loop5
    212 
    213 unroll_over5:
    214 	LDMFD	r13!,{r4,r6-r11,PC}
    215 
    216 mdct_unroll_part3:
    217 	@ r0 = out
    218 	@ r1 = post
    219 	@ r2 = l
    220 	@ r3 = r
    221 	@ <> = step
    222 	@ <> = wL
    223 	@ <> = wR
    224 	MOV	r12,r13
    225 	STMFD	r13!,{r4,r6-r11,r14}
    226 	LDMFD	r12,{r8,r9,r10}		@ r8 = step
    227 					@ r9 = wL
    228 					@ r10= wR
    229 	MVN	r4, #0x8000
    230 	MOV	r8, r8, LSL #1
    231 	SUBS	r1, r1, r3		@ r1 = (post - r)
    232 	BLE	unroll_over6
    233 unroll_loop6:
    234 	LDR	r12,[r2],#8		@ r12= *l       (but l += 2 first)
    235 	LDR	r11,[r9],#4		@ r11= *wL++
    236 	LDR	r7, [r3],#4		@ r7 = *r++
    237 	LDR	r6, [r10,#-4]!		@ r6 = *--wR
    238 
    239 	@ Can save a cycle here, at the cost of 1bit errors in rounding
    240 	SMULL	r14,r11,r12,r11		@ (r14,r11)  = *l   * *wL++
    241 	SMULL	r14,r6, r7, r6		@ (r14,r6)   = *--r * *--wR
    242 	SUB	r6, r6, r11
    243 	MOV	r6, r6, ASR #8
    244 	MOV	r14,r6, ASR #15
    245 	TEQ	r14,r14,ASR #31		@ if r14==0 || r14==-1 then in range
    246 	EORNE	r6, r4, r14,ASR #31
    247 	STRH	r6, [r0], r8
    248 
    249 	SUBS	r1, r1, #4
    250 	BGT	unroll_loop6
    251 
    252 unroll_over6:
    253 	LDMFD	r13!,{r4,r6-r11,PC}
    254 
    255 mdct_shift_right:
    256 	@ r0 = n
    257 	@ r1 = in
    258 	@ r2 = right
    259 	STMFD	r13!,{r4-r11,r14}
    260 
    261 	MOV	r0, r0, LSR #2		@ n >>= 2
    262 	ADD	r1, r1, #4
    263 
    264 	SUBS	r0, r0,	#8
    265 	BLT	sr_less_than_8
    266 sr_loop:
    267 	LDR	r3, [r1], #8
    268 	LDR	r4, [r1], #8
    269 	LDR	r5, [r1], #8
    270 	LDR	r6, [r1], #8
    271 	LDR	r7, [r1], #8
    272 	LDR	r8, [r1], #8
    273 	LDR	r12,[r1], #8
    274 	LDR	r14,[r1], #8
    275 	SUBS	r0, r0, #8
    276 	STMIA	r2!,{r3,r4,r5,r6,r7,r8,r12,r14}
    277 	BGE	sr_loop
    278 sr_less_than_8:
    279 	ADDS	r0, r0, #8
    280 	BEQ	sr_end
    281 sr_loop2:
    282 	LDR	r3, [r1], #8
    283 	SUBS	r0, r0, #1
    284 	STR	r3, [r2], #4
    285 	BGT	sr_loop2
    286 sr_end:
    287 	LDMFD	r13!,{r4-r11,PC}
    288 
    289 mdct_backwardARM:
    290 	@ r0 = n
    291 	@ r1 = in
    292 	STMFD	r13!,{r4-r11,r14}
    293 
    294 	MOV	r2,#1<<4	@ r2 = 1<<shift
    295 	MOV	r3,#13-4	@ r3 = 13-shift
    296 find_shift_loop:
    297 	TST	r0,r2		@ if (n & (1<<shift)) == 0
    298 	MOV	r2,r2,LSL #1
    299 	SUBEQ	r3,r3,#1	@ shift--
    300 	BEQ	find_shift_loop
    301 	MOV	r2,#2
    302 	MOV	r2,r2,LSL r3	@ r2 = step = 2<<shift
    303 
    304 	@ presymmetry
    305 	@ r0 = n (a multiple of 4)
    306 	@ r1 = in
    307 	@ r2 = step
    308 	@ r3 = shift
    309 
    310 	ADD	r4, r1, r0, LSL #1	@ r4 = aX = in+(n>>1)
    311 	ADD	r14,r1, r0		@ r14= in+(n>>2)
    312 	SUB	r4, r4, #3*4		@ r4 = aX = in+n2-3
    313 	ADRL	r7, .Lsincos_lookup
    314 	LDR	r5, [r7]		@ r5 = T=sincos_lookup0
    315 	ADD	r5, r7
    316 
    317 presymmetry_loop1:
    318 	LDR	r7, [r4,#8]		@ r6 = s2 = aX[2]
    319 	LDR	r11,[r5,#4]		@ r11= T[1]
    320 	LDR	r6, [r4]		@ r6 = s0 = aX[0]
    321 	LDR	r10,[r5],r2,LSL #2	@ r10= T[0]   T += step
    322 
    323 	@ XPROD31(s0, s2, T[0], T[1], 0xaX[0], &ax[2])
    324 	SMULL	r8, r9, r7, r11		@ (r8, r9)   = s2*T[1]
    325 	@ stall
    326 	@ stall ?
    327 	SMLAL	r8, r9, r6, r10		@ (r8, r9)  += s0*T[0]
    328 	RSB	r6, r6, #0
    329 	@ stall ?
    330 	SMULL	r8, r12,r7, r10		@ (r8, r12)  = s2*T[0]
    331 	MOV	r9, r9, LSL #1
    332 	@ stall ?
    333 	SMLAL	r8, r12,r6, r11		@ (r8, r12) -= s0*T[1]
    334 	STR	r9, [r4],#-16		@ aX[0] = r9
    335 	CMP	r4,r14
    336 	MOV	r12,r12,LSL #1
    337 	STR	r12,[r4,#8+16]		@ aX[2] = r12
    338 
    339 	BGE	presymmetry_loop1	@ while (aX >= in+n4)
    340 
    341 presymmetry_loop2:
    342 	LDR	r6,[r4]			@ r6 = s0 = aX[0]
    343 	LDR	r10,[r5,#4]		@ r10= T[1]
    344 	LDR	r7,[r4,#8]		@ r6 = s2 = aX[2]
    345 	LDR	r11,[r5],-r2,LSL #2	@ r11= T[0]   T -= step
    346 
    347 	@ XPROD31(s0, s2, T[1], T[0], 0xaX[0], &ax[2])
    348 	SMULL	r8, r9, r6, r10		@ (r8, r9)   = s0*T[1]
    349 	@ stall
    350 	@ stall ?
    351 	SMLAL	r8, r9, r7, r11		@ (r8, r9)  += s2*T[0]
    352 	RSB	r6, r6, #0
    353 	@ stall ?
    354 	SMULL	r8, r12,r7, r10		@ (r8, r12)  = s2*T[1]
    355 	MOV	r9, r9, LSL #1
    356 	@ stall ?
    357 	SMLAL	r8, r12,r6, r11		@ (r8, r12) -= s0*T[0]
    358 	STR	r9, [r4],#-16		@ aX[0] = r9
    359 	CMP	r4,r1
    360 	MOV	r12,r12,LSL #1
    361 	STR	r12,[r4,#8+16]		@ aX[2] = r12
    362 
    363 	BGE	presymmetry_loop2	@ while (aX >= in)
    364 
    365 	@ r0 = n
    366 	@ r1 = in
    367 	@ r2 = step
    368 	@ r3 = shift
    369 	STMFD	r13!,{r3}
    370 	ADRL	r4, .Lsincos_lookup
    371 	LDR	r5, [r4]		@ r5 = T=sincos_lookup0
    372 	ADD	r5, r4
    373 	ADD	r4, r1, r0, LSL #1	@ r4 = aX = in+(n>>1)
    374 	SUB	r4, r4, #4*4		@ r4 = aX = in+(n>>1)-4
    375 	LDR	r11,[r5,#4]		@ r11= T[1]
    376 	LDR	r10,[r5],r2, LSL #2	@ r10= T[0]    T += step
    377 presymmetry_loop3:
    378 	LDR	r8,[r1],#16 		@ r8 = ro0 = bX[0]
    379 	LDR	r9,[r1,#8-16]		@ r9 = ro2 = bX[2]
    380 	LDR	r6,[r4]			@ r6 = ri0 = aX[0]
    381 
    382 	@ XNPROD31( ro2, ro0, T[1], T[0], 0xaX[0], &aX[2] )
    383 	@ aX[0] = (ro2*T[1] - ro0*T[0])>>31 aX[2] = (ro0*T[1] + ro2*T[0])>>31
    384 	SMULL	r14,r12,r8, r11		@ (r14,r12)  = ro0*T[1]
    385 	RSB	r8,r8,#0		@ r8 = -ro0
    386 	@ Stall ?
    387 	SMLAL	r14,r12,r9, r10		@ (r14,r12) += ro2*T[0]
    388 	LDR	r7,[r4,#8]		@ r7 = ri2 = aX[2]
    389 	@ Stall ?
    390 	SMULL	r14,r3, r9, r11		@ (r14,r3)   = ro2*T[1]
    391 	MOV	r12,r12,LSL #1
    392 	LDR	r11,[r5,#4]		@ r11= T[1]
    393 	SMLAL	r14,r3, r8, r10		@ (r14,r3)  -= ro0*T[0]
    394 	LDR	r10,[r5],r2, LSL #2	@ r10= T[0]    T += step
    395 	STR	r12,[r4,#8]
    396 	MOV	r3, r3, LSL #1
    397 	STR	r3, [r4],#-16
    398 
    399 	@ XNPROD31( ri2, ri0, T[0], T[1], 0xbX[0], &bX[2] )
    400 	@ bX[0] = (ri2*T[0] - ri0*T[1])>>31 bX[2] = (ri0*T[0] + ri2*T[1])>>31
    401 	SMULL	r14,r12,r6, r10		@ (r14,r12)  = ri0*T[0]
    402 	RSB	r6,r6,#0		@ r6 = -ri0
    403 	@ stall ?
    404 	SMLAL	r14,r12,r7, r11		@ (r14,r12) += ri2*T[1]
    405 	@ stall ?
    406 	@ stall ?
    407 	SMULL	r14,r3, r7, r10		@ (r14,r3)   = ri2*T[0]
    408 	MOV	r12,r12,LSL #1
    409 	@ stall ?
    410 	SMLAL	r14,r3, r6, r11		@ (r14,r3)  -= ri0*T[1]
    411 	CMP	r4,r1
    412 	STR	r12,[r1,#8-16]
    413 	MOV	r3, r3, LSL #1
    414 	STR	r3, [r1,#-16]
    415 
    416 	BGE	presymmetry_loop3
    417 
    418 	SUB	r1,r1,r0		@ r1 = in -= n>>2 (i.e. restore in)
    419 
    420 	LDR	r3,[r13]
    421 	STR	r2,[r13,#-4]!
    422 
    423 	@ mdct_butterflies
    424 	@ r0 = n  = (points * 2)
    425 	@ r1 = in = x
    426 	@ r2 = i
    427 	@ r3 = shift
    428 	STMFD	r13!,{r0-r1}
    429 	ADRL	r4, .Lsincos_lookup
    430 	LDR	r5, [r4]
    431 	ADD	r5, r4
    432 	RSBS	r4,r3,#6		@ r4 = stages = 7-shift then --stages
    433 	BLE	no_generics
    434 	MOV	r14,#4			@ r14= 4               (i=0)
    435 	MOV	r6, r14,LSL r3		@ r6 = (4<<i)<<shift
    436 mdct_butterflies_loop1:
    437 	MOV	r0, r0, LSR #1		@ r0 = points>>i = POINTS
    438 	MOV	r2, r14,LSR #2		@ r2 = (1<<i)-j        (j=0)
    439 	STMFD	r13!,{r4,r14}
    440 mdct_butterflies_loop2:
    441 
    442 	@ mdct_butterfly_generic(x+POINTS*j, POINTS, 4<<(i+shift))
    443 	@ mdct_butterfly_generic(r1, r0, r6)
    444 	@ r0 = points
    445 	@ r1 = x
    446 	@ preserve r2 (external loop counter)
    447 	@ preserve r3
    448 	@ preserve r4 (external loop counter)
    449 	@ r5 = T = sincos_lookup0
    450 	@ r6 = step
    451 	@ preserve r14
    452 
    453 	STR	r2,[r13,#-4]!		@ stack r2
    454 	ADD	r1,r1,r0,LSL #1		@ r1 = x2+4 = x + (POINTS>>1)
    455 	ADD	r7,r1,r0,LSL #1		@ r7 = x1+4 = x + POINTS
    456 	ADD	r12,r5,#1024*4		@ r12= sincos_lookup0+1024
    457 
    458 mdct_bufferfly_generic_loop1:
    459 	LDMDB	r7!,{r2,r3,r8,r11}	@ r2 = x1[0]
    460 					@ r3 = x1[1]
    461 					@ r8 = x1[2]
    462 					@ r11= x1[3]    x1 -= 4
    463 	LDMDB	r1!,{r4,r9,r10,r14}	@ r4 = x2[0]
    464 					@ r9 = x2[1]
    465 					@ r10= x2[2]
    466 					@ r14= x2[3]    x2 -= 4
    467 
    468 	SUB	r2, r2, r3		@ r2 = s0 = x1[0] - x1[1]
    469 	ADD	r3, r2, r3, LSL #1	@ r3 =      x1[0] + x1[1] (-> x1[0])
    470 	SUB	r11,r11,r8		@ r11= s1 = x1[3] - x1[2]
    471 	ADD	r8, r11,r8, LSL #1	@ r8 =      x1[3] + x1[2] (-> x1[2])
    472 	SUB	r9, r9, r4		@ r9 = s2 = x2[1] - x2[0]
    473 	ADD	r4, r9, r4, LSL #1	@ r4 =      x2[1] + x2[0] (-> x1[1])
    474 	SUB	r14,r14,r10		@ r14= s3 = x2[3] - x2[2]
    475 	ADD	r10,r14,r10,LSL #1	@ r10=      x2[3] + x2[2] (-> x1[3])
    476 	STMIA	r7,{r3,r4,r8,r10}
    477 
    478 	@ r0 = points
    479 	@ r1 = x2
    480 	@ r2 = s0
    481 	@ r3 free
    482 	@ r4 free
    483 	@ r5 = T
    484 	@ r6 = step
    485 	@ r7 = x1
    486 	@ r8 free
    487 	@ r9 = s2
    488 	@ r10 free
    489 	@ r11= s1
    490 	@ r12= limit
    491 	@ r14= s3
    492 
    493 	LDR	r8, [r5,#4]		@ r8 = T[1]
    494 	LDR	r10,[r5],r6,LSL #2	@ r10= T[0]		T += step
    495 
    496 	@ XPROD31(s1, s0, T[0], T[1], &x2[0], &x2[2])
    497 	@ x2[0] = (s1*T[0] + s0*T[1])>>31     x2[2] = (s0*T[0] - s1*T[1])>>31
    498 	@ stall Xscale
    499 	SMULL	r4, r3, r2, r8		@ (r4, r3)   = s0*T[1]
    500 	SMLAL	r4, r3, r11,r10		@ (r4, r3)  += s1*T[0]
    501 	RSB	r11,r11,#0
    502 	SMULL	r11,r4, r8, r11		@ (r11,r4)   = -s1*T[1]
    503 	SMLAL	r11,r4, r2, r10		@ (r11,r4)  += s0*T[0]
    504 	MOV	r2, r3, LSL #1		@ r2 = r3<<1 = Value for x2[0]
    505 
    506 	@ XPROD31(s2, s3, T[0], T[1], &x2[1], &x2[3])
    507 	@ x2[1] = (s2*T[0] + s3*T[1])>>31     x2[3] = (s3*T[0] - s2*T[1])>>31
    508 	SMULL	r11,r3, r9, r10		@ (r11,r3)   = s2*T[0]
    509 	MOV	r4, r4, LSL #1		@ r4 = r4<<1 = Value for x2[2]
    510 	SMLAL	r11,r3, r14,r8		@ (r11,r3)  += s3*T[1]
    511 	RSB	r9, r9, #0
    512 	SMULL	r10,r11,r14,r10		@ (r10,r11)  = s3*T[0]
    513 	MOV	r3, r3, LSL #1		@ r3 = r3<<1 = Value for x2[1]
    514 	SMLAL	r10,r11,r9,r8		@ (r10,r11) -= s2*T[1]
    515 	CMP	r5, r12
    516 	MOV	r11,r11,LSL #1		@ r11= r11<<1 = Value for x2[3]
    517 
    518 	STMIA	r1,{r2,r3,r4,r11}
    519 
    520 	BLT	mdct_bufferfly_generic_loop1
    521 
    522 	SUB	r12,r12,#1024*4
    523 mdct_bufferfly_generic_loop2:
    524 	LDMDB	r7!,{r2,r3,r9,r10}	@ r2 = x1[0]
    525 					@ r3 = x1[1]
    526 					@ r9 = x1[2]
    527 					@ r10= x1[3]    x1 -= 4
    528 	LDMDB	r1!,{r4,r8,r11,r14}	@ r4 = x2[0]
    529 					@ r8 = x2[1]
    530 					@ r11= x2[2]
    531 					@ r14= x2[3]    x2 -= 4
    532 
    533 	SUB	r2, r2, r3		@ r2 = s0 = x1[0] - x1[1]
    534 	ADD	r3, r2, r3, LSL #1	@ r3 =      x1[0] + x1[1] (-> x1[0])
    535 	SUB	r9, r9,r10		@ r9 = s1 = x1[2] - x1[3]
    536 	ADD	r10,r9,r10, LSL #1	@ r10=      x1[2] + x1[3] (-> x1[2])
    537 	SUB	r4, r4, r8		@ r4 = s2 = x2[0] - x2[1]
    538 	ADD	r8, r4, r8, LSL #1	@ r8 =      x2[0] + x2[1] (-> x1[1])
    539 	SUB	r14,r14,r11		@ r14= s3 = x2[3] - x2[2]
    540 	ADD	r11,r14,r11,LSL #1	@ r11=      x2[3] + x2[2] (-> x1[3])
    541 	STMIA	r7,{r3,r8,r10,r11}
    542 
    543 	@ r0 = points
    544 	@ r1 = x2
    545 	@ r2 = s0
    546 	@ r3 free
    547 	@ r4 = s2
    548 	@ r5 = T
    549 	@ r6 = step
    550 	@ r7 = x1
    551 	@ r8 free
    552 	@ r9 = s1
    553 	@ r10 free
    554 	@ r11 free
    555 	@ r12= limit
    556 	@ r14= s3
    557 
    558 	LDR	r8, [r5,#4]		@ r8 = T[1]
    559 	LDR	r10,[r5],-r6,LSL #2	@ r10= T[0]		T -= step
    560 
    561 	@ XNPROD31(s0, s1, T[0], T[1], &x2[0], &x2[2])
    562 	@ x2[0] = (s0*T[0] - s1*T[1])>>31     x2[2] = (s1*T[0] + s0*T[1])>>31
    563 	@ stall Xscale
    564 	SMULL	r3, r11,r2, r8		@ (r3, r11)  = s0*T[1]
    565 	SMLAL	r3, r11,r9, r10		@ (r3, r11) += s1*T[0]
    566 	RSB	r9, r9, #0
    567 	SMULL	r3, r2, r10,r2		@ (r3, r2)   = s0*T[0]
    568 	SMLAL	r3, r2, r9, r8		@ (r3, r2)  += -s1*T[1]
    569 	MOV	r9, r11,LSL #1		@ r9 = r11<<1 = Value for x2[2]
    570 
    571 	@ XNPROD31(s3, s2, T[0], T[1], &x2[1], &x2[3])
    572 	@ x2[1] = (s3*T[0] - s2*T[1])>>31     x2[3] = (s2*T[0] + s3*T[1])>>31
    573 	SMULL	r3, r11,r4, r10		@ (r3,r11)   = s2*T[0]
    574 	MOV	r2, r2, LSL #1		@ r2 = r2<<1  = Value for x2[0]
    575 	SMLAL	r3, r11,r14,r8		@ (r3,r11)  += s3*T[1]
    576 	RSB	r4, r4, #0
    577 	SMULL	r10,r3,r14,r10		@ (r10,r3)   = s3*T[0]
    578 	MOV	r11,r11,LSL #1		@ r11= r11<<1 = Value for x2[3]
    579 	SMLAL	r10,r3, r4, r8		@ (r10,r3)  -= s2*T[1]
    580 	CMP	r5, r12
    581 	MOV	r3, r3, LSL #1		@ r3 = r3<<1  = Value for x2[1]
    582 
    583 	STMIA	r1,{r2,r3,r9,r11}
    584 
    585 	BGT	mdct_bufferfly_generic_loop2
    586 
    587 	LDR	r2,[r13],#4		@ unstack r2
    588 	ADD	r1, r1, r0, LSL #2	@ r1 = x+POINTS*j
    589 	@ stall Xscale
    590 	SUBS	r2, r2, #1		@ r2--                 (j++)
    591 	BGT	mdct_butterflies_loop2
    592 
    593 	LDMFD	r13!,{r4,r14}
    594 
    595 	LDR	r1,[r13,#4]
    596 
    597 	SUBS	r4, r4, #1		@ stages--
    598 	MOV	r14,r14,LSL #1		@ r14= 4<<i            (i++)
    599 	MOV	r6, r6, LSL #1		@ r6 = step <<= 1      (i++)
    600 	BGE	mdct_butterflies_loop1
    601 	LDMFD	r13,{r0-r1}
    602 no_generics:
    603 	@ mdct_butterflies part2 (loop around mdct_bufferfly_32)
    604 	@ r0 = points
    605 	@ r1 = in
    606 	@ r2 = step
    607 	@ r3 = shift
    608 
    609 mdct_bufferflies_loop3:
    610 	@ mdct_bufferfly_32
    611 
    612 	@ block1
    613 	ADD	r4, r1, #16*4		@ r4 = &in[16]
    614 	LDMIA	r4,{r5,r6,r9,r10}	@ r5 = x[16]
    615 					@ r6 = x[17]
    616 					@ r9 = x[18]
    617 					@ r10= x[19]
    618 	LDMIA	r1,{r7,r8,r11,r12}	@ r7 = x[0]
    619 					@ r8 = x[1]
    620 					@ r11= x[2]
    621 					@ r12= x[3]
    622 	SUB	r5, r5, r6		@ r5 = s0 = x[16] - x[17]
    623 	ADD	r6, r5, r6, LSL #1	@ r6 =      x[16] + x[17]  -> x[16]
    624 	SUB	r9, r9, r10		@ r9 = s1 = x[18] - x[19]
    625 	ADD	r10,r9, r10,LSL #1	@ r10=      x[18] + x[19]  -> x[18]
    626 	SUB	r8, r8, r7		@ r8 = s2 = x[ 1] - x[ 0]
    627 	ADD	r7, r8, r7, LSL #1	@ r7 =      x[ 1] + x[ 0]  -> x[17]
    628 	SUB	r12,r12,r11		@ r12= s3 = x[ 3] - x[ 2]
    629 	ADD	r11,r12,r11, LSL #1	@ r11=      x[ 3] + x[ 2]  -> x[19]
    630 	STMIA	r4!,{r6,r7,r10,r11}
    631 
    632 	LDR	r6,cPI1_8
    633 	LDR	r7,cPI3_8
    634 
    635 	@ XNPROD31( s0, s1, cPI3_8, cPI1_8, &x[ 0], &x[ 2] )
    636 	@ x[0] = s0*cPI3_8 - s1*cPI1_8     x[2] = s1*cPI3_8 + s0*cPI1_8
    637 	@ stall Xscale
    638 	SMULL	r14,r11,r5, r6		@ (r14,r11)  = s0*cPI1_8
    639 	SMLAL	r14,r11,r9, r7		@ (r14,r11) += s1*cPI3_8
    640 	RSB	r9, r9, #0
    641 	SMULL	r14,r5, r7, r5		@ (r14,r5)   = s0*cPI3_8
    642 	SMLAL	r14,r5, r9, r6		@ (r14,r5)  -= s1*cPI1_8
    643 	MOV	r11,r11,LSL #1
    644 	MOV	r5, r5, LSL #1
    645 
    646 	@ XPROD31 ( s2, s3, cPI1_8, cPI3_8, &x[ 1], &x[ 3] )
    647 	@ x[1] = s2*cPI1_8 + s3*cPI3_8     x[3] = s3*cPI1_8 - s2*cPI3_8
    648 	SMULL	r14,r9, r8, r6		@ (r14,r9)   = s2*cPI1_8
    649 	SMLAL	r14,r9, r12,r7		@ (r14,r9)  += s3*cPI3_8
    650 	RSB	r8,r8,#0
    651 	SMULL	r14,r12,r6, r12		@ (r14,r12)  = s3*cPI1_8
    652 	SMLAL	r14,r12,r8, r7		@ (r14,r12) -= s2*cPI3_8
    653 	MOV	r9, r9, LSL #1
    654 	MOV	r12,r12,LSL #1
    655 	STMIA	r1!,{r5,r9,r11,r12}
    656 
    657 	@ block2
    658 	LDMIA	r4,{r5,r6,r9,r10}	@ r5 = x[20]
    659 					@ r6 = x[21]
    660 					@ r9 = x[22]
    661 					@ r10= x[23]
    662 	LDMIA	r1,{r7,r8,r11,r12}	@ r7 = x[4]
    663 					@ r8 = x[5]
    664 					@ r11= x[6]
    665 					@ r12= x[7]
    666 	SUB	r5, r5, r6		@ r5 = s0 = x[20] - x[21]
    667 	ADD	r6, r5, r6, LSL #1	@ r6 =      x[20] + x[21]  -> x[20]
    668 	SUB	r9, r9, r10		@ r9 = s1 = x[22] - x[23]
    669 	ADD	r10,r9, r10,LSL #1	@ r10=      x[22] + x[23]  -> x[22]
    670 	SUB	r8, r8, r7		@ r8 = s2 = x[ 5] - x[ 4]
    671 	ADD	r7, r8, r7, LSL #1	@ r7 =      x[ 5] + x[ 4]  -> x[21]
    672 	SUB	r12,r12,r11		@ r12= s3 = x[ 7] - x[ 6]
    673 	ADD	r11,r12,r11, LSL #1	@ r11=      x[ 7] + x[ 6]  -> x[23]
    674 	LDR	r14,cPI2_8
    675 	STMIA	r4!,{r6,r7,r10,r11}
    676 
    677 	SUB	r5, r5, r9		@ r5 = s0 - s1
    678 	ADD	r9, r5, r9, LSL #1	@ r9 = s0 + s1
    679 	SMULL	r6, r5, r14,r5		@ (r6,r5)  = (s0-s1)*cPI2_8
    680 	SUB	r12,r12,r8		@ r12= s3 - s2
    681 	ADD	r8, r12,r8, LSL #1	@ r8 = s3 + s2
    682 
    683 	SMULL	r6, r8, r14,r8		@ (r6,r8)  = (s3+s2)*cPI2_8
    684 	MOV	r5, r5, LSL #1
    685 	SMULL	r6, r9, r14,r9		@ (r6,r9)  = (s0+s1)*cPI2_8
    686 	MOV	r8, r8, LSL #1
    687 	SMULL	r6, r12,r14,r12		@ (r6,r12) = (s3-s2)*cPI2_8
    688 	MOV	r9, r9, LSL #1
    689 	MOV	r12,r12,LSL #1
    690 	STMIA	r1!,{r5,r8,r9,r12}
    691 
    692 	@ block3
    693 	LDMIA	r4,{r5,r6,r9,r10}	@ r5 = x[24]
    694 					@ r6 = x[25]
    695 					@ r9 = x[25]
    696 					@ r10= x[26]
    697 	LDMIA	r1,{r7,r8,r11,r12}	@ r7 = x[8]
    698 					@ r8 = x[9]
    699 					@ r11= x[10]
    700 					@ r12= x[11]
    701 	SUB	r5, r5, r6		@ r5 = s0 = x[24] - x[25]
    702 	ADD	r6, r5, r6, LSL #1	@ r6 =      x[24] + x[25]  -> x[25]
    703 	SUB	r9, r9, r10		@ r9 = s1 = x[26] - x[27]
    704 	ADD	r10,r9, r10,LSL #1	@ r10=      x[26] + x[27]  -> x[26]
    705 	SUB	r8, r8, r7		@ r8 = s2 = x[ 9] - x[ 8]
    706 	ADD	r7, r8, r7, LSL #1	@ r7 =      x[ 9] + x[ 8]  -> x[25]
    707 	SUB	r12,r12,r11		@ r12= s3 = x[11] - x[10]
    708 	ADD	r11,r12,r11, LSL #1	@ r11=      x[11] + x[10]  -> x[27]
    709 	STMIA	r4!,{r6,r7,r10,r11}
    710 
    711 	LDR	r6,cPI3_8
    712 	LDR	r7,cPI1_8
    713 
    714 	@ XNPROD31( s0, s1, cPI1_8, cPI3_8, &x[ 8], &x[10] )
    715 	@ x[8] = s0*cPI1_8 - s1*cPI3_8     x[10] = s1*cPI1_8 + s0*cPI3_8
    716 	@ stall Xscale
    717 	SMULL	r14,r11,r5, r6		@ (r14,r11)  = s0*cPI3_8
    718 	SMLAL	r14,r11,r9, r7		@ (r14,r11) += s1*cPI1_8
    719 	RSB	r9, r9, #0
    720 	SMULL	r14,r5, r7, r5		@ (r14,r5)   = s0*cPI1_8
    721 	SMLAL	r14,r5, r9, r6		@ (r14,r5)  -= s1*cPI3_8
    722 	MOV	r11,r11,LSL #1
    723 	MOV	r5, r5, LSL #1
    724 
    725 	@ XPROD31 ( s2, s3, cPI3_8, cPI1_8, &x[ 9], &x[11] )
    726 	@ x[9] = s2*cPI3_8 + s3*cPI1_8     x[11] = s3*cPI3_8 - s2*cPI1_8
    727 	SMULL	r14,r9, r8, r6		@ (r14,r9)   = s2*cPI3_8
    728 	SMLAL	r14,r9, r12,r7		@ (r14,r9)  += s3*cPI1_8
    729 	RSB	r8,r8,#0
    730 	SMULL	r14,r12,r6, r12		@ (r14,r12)  = s3*cPI3_8
    731 	SMLAL	r14,r12,r8, r7		@ (r14,r12) -= s2*cPI1_8
    732 	MOV	r9, r9, LSL #1
    733 	MOV	r12,r12,LSL #1
    734 	STMIA	r1!,{r5,r9,r11,r12}
    735 
    736 	@ block4
    737 	LDMIA	r4,{r5,r6,r10,r11}	@ r5 = x[28]
    738 					@ r6 = x[29]
    739 					@ r10= x[30]
    740 					@ r11= x[31]
    741 	LDMIA	r1,{r8,r9,r12,r14}	@ r8 = x[12]
    742 					@ r9 = x[13]
    743 					@ r12= x[14]
    744 					@ r14= x[15]
    745 	SUB	r5, r5, r6		@ r5 = s0 = x[28] - x[29]
    746 	ADD	r6, r5, r6, LSL #1	@ r6 =      x[28] + x[29]  -> x[28]
    747 	SUB	r7, r14,r12		@ r7 = s3 = x[15] - x[14]
    748 	ADD	r12,r7, r12, LSL #1	@ r12=      x[15] + x[14]  -> x[31]
    749 	SUB	r10,r10,r11		@ r10= s1 = x[30] - x[31]
    750 	ADD	r11,r10,r11,LSL #1	@ r11=      x[30] + x[31]  -> x[30]
    751 	SUB	r14, r8, r9		@ r14= s2 = x[12] - x[13]
    752 	ADD	r9, r14, r9, LSL #1	@ r9 =      x[12] + x[13]  -> x[29]
    753 	STMIA	r4!,{r6,r9,r11,r12}
    754 	STMIA	r1!,{r5,r7,r10,r14}
    755 
    756 	@ mdct_butterfly16 (1st version)
    757 	@ block 1
    758 	SUB	r1,r1,#16*4
    759 	ADD	r4,r1,#8*4
    760 	LDMIA	r4,{r5,r6,r9,r10}	@ r5 = x[ 8]
    761 					@ r6 = x[ 9]
    762 					@ r9 = x[10]
    763 					@ r10= x[11]
    764 	LDMIA	r1,{r7,r8,r11,r12}	@ r7 = x[0]
    765 					@ r8 = x[1]
    766 					@ r11= x[2]
    767 					@ r12= x[3]
    768 	SUB	r5, r5, r6		@ r5 = s0 = x[ 8] - x[ 9]
    769 	ADD	r6, r5, r6, LSL #1	@ r6 =      x[ 8] + x[ 9]  -> x[ 8]
    770 	SUB	r9, r9, r10		@ r9 = s1 = x[10] - x[11]
    771 	ADD	r10,r9, r10,LSL #1	@ r10=      x[10] + x[11]  -> x[10]
    772 	SUB	r8, r8, r7		@ r8 = s2 = x[ 1] - x[ 0]
    773 	ADD	r7, r8, r7, LSL #1	@ r7 =      x[ 1] + x[ 0]  -> x[ 9]
    774 	SUB	r12,r12,r11		@ r12= s3 = x[ 3] - x[ 2]
    775 	ADD	r11,r12,r11, LSL #1	@ r11=      x[ 3] + x[ 2]  -> x[11]
    776 	LDR	r14,cPI2_8
    777 	STMIA	r4!,{r6,r7,r10,r11}
    778 
    779 	SUB	r5, r5, r9		@ r5 = s0 - s1
    780 	ADD	r9, r5, r9, LSL #1	@ r9 = s0 + s1
    781 	SMULL	r6, r5, r14,r5		@ (r6,r5)  = (s0-s1)*cPI2_8
    782 	SUB	r12,r12,r8		@ r12= s3 - s2
    783 	ADD	r8, r12,r8, LSL #1	@ r8 = s3 + s2
    784 
    785 	SMULL	r6, r8, r14,r8		@ (r6,r8)  = (s3+s2)*cPI2_8
    786 	MOV	r5, r5, LSL #1
    787 	SMULL	r6, r9, r14,r9		@ (r6,r9)  = (s0+s1)*cPI2_8
    788 	MOV	r8, r8, LSL #1
    789 	SMULL	r6, r12,r14,r12		@ (r6,r12) = (s3-s2)*cPI2_8
    790 	MOV	r9, r9, LSL #1
    791 	MOV	r12,r12,LSL #1
    792 	STMIA	r1!,{r5,r8,r9,r12}
    793 
    794 	@ block4
    795 	LDMIA	r4,{r5,r6,r9,r10}	@ r5 = x[12]
    796 					@ r6 = x[13]
    797 					@ r9 = x[14]
    798 					@ r10= x[15]
    799 	LDMIA	r1,{r7,r8,r11,r12}	@ r7 = x[ 4]
    800 					@ r8 = x[ 5]
    801 					@ r11= x[ 6]
    802 					@ r12= x[ 7]
    803 	SUB	r14,r7, r8		@ r14= s0 = x[ 4] - x[ 5]
    804 	ADD	r8, r14,r8, LSL #1	@ r8 =      x[ 4] + x[ 5]  -> x[13]
    805 	SUB	r7, r12,r11		@ r7 = s1 = x[ 7] - x[ 6]
    806 	ADD	r11,r7, r11, LSL #1	@ r11=      x[ 7] + x[ 6]  -> x[15]
    807 	SUB	r5, r5, r6		@ r5 = s2 = x[12] - x[13]
    808 	ADD	r6, r5, r6, LSL #1	@ r6 =      x[12] + x[13]  -> x[12]
    809 	SUB	r12,r9, r10		@ r12= s3 = x[14] - x[15]
    810 	ADD	r10,r12,r10,LSL #1	@ r10=      x[14] + x[15]  -> x[14]
    811 	STMIA	r4!,{r6,r8,r10,r11}
    812 	STMIA	r1!,{r5,r7,r12,r14}
    813 
    814 	@ mdct_butterfly_8
    815 	LDMDB	r1,{r6,r7,r8,r9,r10,r11,r12,r14}
    816 					@ r6 = x[0]
    817 					@ r7 = x[1]
    818 					@ r8 = x[2]
    819 					@ r9 = x[3]
    820 					@ r10= x[4]
    821 					@ r11= x[5]
    822 					@ r12= x[6]
    823 					@ r14= x[7]
    824 	ADD	r6, r6, r7		@ r6 = s0 = x[0] + x[1]
    825 	SUB	r7, r6, r7, LSL #1	@ r7 = s1 = x[0] - x[1]
    826 	ADD	r8, r8, r9		@ r8 = s2 = x[2] + x[3]
    827 	SUB	r9, r8, r9, LSL #1	@ r9 = s3 = x[2] - x[3]
    828 	ADD	r10,r10,r11		@ r10= s4 = x[4] + x[5]
    829 	SUB	r11,r10,r11,LSL #1	@ r11= s5 = x[4] - x[5]
    830 	ADD	r12,r12,r14		@ r12= s6 = x[6] + x[7]
    831 	SUB	r14,r12,r14,LSL #1	@ r14= s7 = x[6] - x[7]
    832 
    833 	ADD	r2, r11,r9		@ r2 = x[0] = s5 + s3
    834 	SUB	r4, r2, r9, LSL #1	@ r4 = x[2] = s5 - s3
    835 	SUB	r3, r14,r7		@ r3 = x[1] = s7 - s1
    836 	ADD	r5, r3, r7, LSL #1	@ r5 = x[3] = s7 + s1
    837 	SUB	r10,r10,r6		@ r10= x[4] = s4 - s0
    838 	SUB	r11,r12,r8		@ r11= x[5] = s6 - s2
    839 	ADD	r12,r10,r6, LSL #1	@ r12= x[6] = s4 + s0
    840 	ADD	r14,r11,r8, LSL #1	@ r14= x[7] = s6 + s2
    841 	STMDB	r1,{r2,r3,r4,r5,r10,r11,r12,r14}
    842 
    843 	@ mdct_butterfly_8
    844 	LDMIA	r1,{r6,r7,r8,r9,r10,r11,r12,r14}
    845 					@ r6 = x[0]
    846 					@ r7 = x[1]
    847 					@ r8 = x[2]
    848 					@ r9 = x[3]
    849 					@ r10= x[4]
    850 					@ r11= x[5]
    851 					@ r12= x[6]
    852 					@ r14= x[7]
    853 	ADD	r6, r6, r7		@ r6 = s0 = x[0] + x[1]
    854 	SUB	r7, r6, r7, LSL #1	@ r7 = s1 = x[0] - x[1]
    855 	ADD	r8, r8, r9		@ r8 = s2 = x[2] + x[3]
    856 	SUB	r9, r8, r9, LSL #1	@ r9 = s3 = x[2] - x[3]
    857 	ADD	r10,r10,r11		@ r10= s4 = x[4] + x[5]
    858 	SUB	r11,r10,r11,LSL #1	@ r11= s5 = x[4] - x[5]
    859 	ADD	r12,r12,r14		@ r12= s6 = x[6] + x[7]
    860 	SUB	r14,r12,r14,LSL #1	@ r14= s7 = x[6] - x[7]
    861 
    862 	ADD	r2, r11,r9		@ r2 = x[0] = s5 + s3
    863 	SUB	r4, r2, r9, LSL #1	@ r4 = x[2] = s5 - s3
    864 	SUB	r3, r14,r7		@ r3 = x[1] = s7 - s1
    865 	ADD	r5, r3, r7, LSL #1	@ r5 = x[3] = s7 + s1
    866 	SUB	r10,r10,r6		@ r10= x[4] = s4 - s0
    867 	SUB	r11,r12,r8		@ r11= x[5] = s6 - s2
    868 	ADD	r12,r10,r6, LSL #1	@ r12= x[6] = s4 + s0
    869 	ADD	r14,r11,r8, LSL #1	@ r14= x[7] = s6 + s2
    870 	STMIA	r1,{r2,r3,r4,r5,r10,r11,r12,r14}
    871 
    872 	@ block 2
    873 	ADD	r1,r1,#16*4-8*4
    874 	ADD	r4,r1,#8*4
    875 	LDMIA	r4,{r5,r6,r9,r10}	@ r5 = x[ 8]
    876 					@ r6 = x[ 9]
    877 					@ r9 = x[10]
    878 					@ r10= x[11]
    879 	LDMIA	r1,{r7,r8,r11,r12}	@ r7 = x[0]
    880 					@ r8 = x[1]
    881 					@ r11= x[2]
    882 					@ r12= x[3]
    883 	SUB	r5, r5, r6		@ r5 = s0 = x[ 8] - x[ 9]
    884 	ADD	r6, r5, r6, LSL #1	@ r6 =      x[ 8] + x[ 9]  -> x[ 8]
    885 	SUB	r9, r9, r10		@ r9 = s1 = x[10] - x[11]
    886 	ADD	r10,r9, r10,LSL #1	@ r10=      x[10] + x[11]  -> x[10]
    887 	SUB	r8, r8, r7		@ r8 = s2 = x[ 1] - x[ 0]
    888 	ADD	r7, r8, r7, LSL #1	@ r7 =      x[ 1] + x[ 0]  -> x[ 9]
    889 	SUB	r12,r12,r11		@ r12= s3 = x[ 3] - x[ 2]
    890 	ADD	r11,r12,r11, LSL #1	@ r11=      x[ 3] + x[ 2]  -> x[11]
    891 	LDR	r14,cPI2_8
    892 	STMIA	r4!,{r6,r7,r10,r11}
    893 
    894 	SUB	r5, r5, r9		@ r5 = s0 - s1
    895 	ADD	r9, r5, r9, LSL #1	@ r9 = s0 + s1
    896 	SMULL	r6, r5, r14,r5		@ (r6,r5)  = (s0-s1)*cPI2_8
    897 	SUB	r12,r12,r8		@ r12= s3 - s2
    898 	ADD	r8, r12,r8, LSL #1	@ r8 = s3 + s2
    899 
    900 	SMULL	r6, r8, r14,r8		@ (r6,r8)  = (s3+s2)*cPI2_8
    901 	MOV	r5, r5, LSL #1
    902 	SMULL	r6, r9, r14,r9		@ (r6,r9)  = (s0+s1)*cPI2_8
    903 	MOV	r8, r8, LSL #1
    904 	SMULL	r6, r12,r14,r12		@ (r6,r12) = (s3-s2)*cPI2_8
    905 	MOV	r9, r9, LSL #1
    906 	MOV	r12,r12,LSL #1
    907 	STMIA	r1!,{r5,r8,r9,r12}
    908 
    909 	@ block4
    910 	LDMIA	r4,{r5,r6,r9,r10}	@ r5 = x[12]
    911 					@ r6 = x[13]
    912 					@ r9 = x[14]
    913 					@ r10= x[15]
    914 	LDMIA	r1,{r7,r8,r11,r12}	@ r7 = x[ 4]
    915 					@ r8 = x[ 5]
    916 					@ r11= x[ 6]
    917 					@ r12= x[ 7]
    918 	SUB	r5, r5, r6		@ r5 = s2 = x[12] - x[13]
    919 	ADD	r6, r5, r6, LSL #1	@ r6 =      x[12] + x[13]  -> x[12]
    920 	SUB	r9, r9, r10		@ r9 = s3 = x[14] - x[15]
    921 	ADD	r10,r9, r10,LSL #1	@ r10=      x[14] + x[15]  -> x[14]
    922 	SUB	r14,r7, r8		@ r14= s0 = x[ 4] - x[ 5]
    923 	ADD	r8, r14,r8, LSL #1	@ r8 =      x[ 4] + x[ 5]  -> x[13]
    924 	SUB	r7, r12,r11		@ r7 = s1 = x[ 7] - x[ 6]
    925 	ADD	r11,r7, r11, LSL #1	@ r11=      x[ 7] + x[ 6]  -> x[15]
    926 	STMIA	r4!,{r6,r8,r10,r11}
    927 	STMIA	r1!,{r5,r7,r9,r14}
    928 
    929 	@ mdct_butterfly_8
    930 	LDMDB	r1,{r6,r7,r8,r9,r10,r11,r12,r14}
    931 					@ r6 = x[0]
    932 					@ r7 = x[1]
    933 					@ r8 = x[2]
    934 					@ r9 = x[3]
    935 					@ r10= x[4]
    936 					@ r11= x[5]
    937 					@ r12= x[6]
    938 					@ r14= x[7]
    939 	ADD	r6, r6, r7		@ r6 = s0 = x[0] + x[1]
    940 	SUB	r7, r6, r7, LSL #1	@ r7 = s1 = x[0] - x[1]
    941 	ADD	r8, r8, r9		@ r8 = s2 = x[2] + x[3]
    942 	SUB	r9, r8, r9, LSL #1	@ r9 = s3 = x[2] - x[3]
    943 	ADD	r10,r10,r11		@ r10= s4 = x[4] + x[5]
    944 	SUB	r11,r10,r11,LSL #1	@ r11= s5 = x[4] - x[5]
    945 	ADD	r12,r12,r14		@ r12= s6 = x[6] + x[7]
    946 	SUB	r14,r12,r14,LSL #1	@ r14= s7 = x[6] - x[7]
    947 
    948 	ADD	r2, r11,r9		@ r2 = x[0] = s5 + s3
    949 	SUB	r4, r2, r9, LSL #1	@ r4 = x[2] = s5 - s3
    950 	SUB	r3, r14,r7		@ r3 = x[1] = s7 - s1
    951 	ADD	r5, r3, r7, LSL #1	@ r5 = x[3] = s7 + s1
    952 	SUB	r10,r10,r6		@ r10= x[4] = s4 - s0
    953 	SUB	r11,r12,r8		@ r11= x[5] = s6 - s2
    954 	ADD	r12,r10,r6, LSL #1	@ r12= x[6] = s4 + s0
    955 	ADD	r14,r11,r8, LSL #1	@ r14= x[7] = s6 + s2
    956 	STMDB	r1,{r2,r3,r4,r5,r10,r11,r12,r14}
    957 
    958 	@ mdct_butterfly_8
    959 	LDMIA	r1,{r6,r7,r8,r9,r10,r11,r12,r14}
    960 					@ r6 = x[0]
    961 					@ r7 = x[1]
    962 					@ r8 = x[2]
    963 					@ r9 = x[3]
    964 					@ r10= x[4]
    965 					@ r11= x[5]
    966 					@ r12= x[6]
    967 					@ r14= x[7]
    968 	ADD	r6, r6, r7		@ r6 = s0 = x[0] + x[1]
    969 	SUB	r7, r6, r7, LSL #1	@ r7 = s1 = x[0] - x[1]
    970 	ADD	r8, r8, r9		@ r8 = s2 = x[2] + x[3]
    971 	SUB	r9, r8, r9, LSL #1	@ r9 = s3 = x[2] - x[3]
    972 	ADD	r10,r10,r11		@ r10= s4 = x[4] + x[5]
    973 	SUB	r11,r10,r11,LSL #1	@ r11= s5 = x[4] - x[5]
    974 	ADD	r12,r12,r14		@ r12= s6 = x[6] + x[7]
    975 	SUB	r14,r12,r14,LSL #1	@ r14= s7 = x[6] - x[7]
    976 
    977 	ADD	r2, r11,r9		@ r2 = x[0] = s5 + s3
    978 	SUB	r4, r2, r9, LSL #1	@ r4 = x[2] = s5 - s3
    979 	SUB	r3, r14,r7		@ r3 = x[1] = s7 - s1
    980 	ADD	r5, r3, r7, LSL #1	@ r5 = x[3] = s7 + s1
    981 	SUB	r10,r10,r6		@ r10= x[4] = s4 - s0
    982 	SUB	r11,r12,r8		@ r11= x[5] = s6 - s2
    983 	ADD	r12,r10,r6, LSL #1	@ r12= x[6] = s4 + s0
    984 	ADD	r14,r11,r8, LSL #1	@ r14= x[7] = s6 + s2
    985 	STMIA	r1,{r2,r3,r4,r5,r10,r11,r12,r14}
    986 
    987 	ADD	r1,r1,#8*4
    988 	SUBS	r0,r0,#64
    989 	BGT	mdct_bufferflies_loop3
    990 
    991 	LDMFD	r13,{r0-r3}
    992 
    993 mdct_bitreverseARM:
    994 	@ r0 = points = n
    995 	@ r1 = in
    996 	@ r2 = step
    997 	@ r3 = shift
    998 
    999 	MOV	r4, #0			@ r4 = bit = 0
   1000 	ADD	r5, r1, r0, LSL #1	@ r5 = w = x + (n>>1)
   1001 	ADR	r6, bitrev
   1002 	SUB	r5, r5, #8
   1003 brev_lp:
   1004 	LDRB	r7, [r6, r4, LSR #6]
   1005 	AND	r8, r4, #0x3f
   1006 	LDRB	r8, [r6, r8]
   1007 	ADD	r4, r4, #1		@ bit++
   1008 	@ stall XScale
   1009 	ORR	r7, r7, r8, LSL #6	@ r7 = bitrev[bit]
   1010 	MOV	r7, r7, LSR r3
   1011 	ADD	r9, r1, r7, LSL #2	@ r9 = xx = x + (b>>shift)
   1012 	CMP	r5, r9			@ if (w > xx)
   1013 	LDR	r10,[r5],#-8		@   r10 = w[0]		w -= 2
   1014 	LDRGT	r11,[r5,#12]		@   r11 = w[1]
   1015 	LDRGT	r12,[r9]		@   r12 = xx[0]
   1016 	LDRGT	r14,[r9,#4]		@   r14 = xx[1]
   1017 	STRGT	r10,[r9]		@   xx[0]= w[0]
   1018 	STRGT	r11,[r9,#4]		@   xx[1]= w[1]
   1019 	STRGT	r12,[r5,#8]		@   w[0] = xx[0]
   1020 	STRGT	r14,[r5,#12]		@   w[1] = xx[1]
   1021 	CMP	r5,r1
   1022 	BGT	brev_lp
   1023 
   1024 	@ mdct_step7
   1025 	@ r0 = points
   1026 	@ r1 = in
   1027 	@ r2 = step
   1028 	@ r3 = shift
   1029 
   1030 	CMP	r2, #4			@ r5 = T = (step>=4) ?
   1031 	ADR	r7, .Lsincos_lookup	@          sincos_lookup0 +
   1032 	ADDLT	r7, #4			@          sincos_lookup1
   1033 	LDR	r5, [r7]
   1034 	ADD	r5, r7
   1035 	ADD	r7, r1, r0, LSL #1	@ r7 = w1 = x + (n>>1)
   1036 	ADDGE	r5, r5, r2, LSL #1	@		            (step>>1)
   1037 	ADD	r8, r5, #1024*4		@ r8 = Ttop
   1038 step7_loop1:
   1039 	LDR	r6, [r1]		@ r6 = w0[0]
   1040 	LDR	r9, [r1,#4]		@ r9 = w0[1]
   1041 	LDR	r10,[r7,#-8]!		@ r10= w1[0]	w1 -= 2
   1042 	LDR	r11,[r7,#4]		@ r11= w1[1]
   1043 	LDR	r14,[r5,#4]		@ r14= T[1]
   1044 	LDR	r12,[r5],r2,LSL #2	@ r12= T[0]	T += step
   1045 
   1046 	ADD	r6, r6, r10		@ r6 = s0 = w0[0] + w1[0]
   1047 	SUB	r10,r6, r10,LSL #1	@ r10= s1b= w0[0] - w1[0]
   1048 	SUB	r11,r11,r9		@ r11= s1 = w1[1] - w0[1]
   1049 	ADD	r9, r11,r9, LSL #1	@ r9 = s0b= w1[1] + w0[1]
   1050 
   1051 	@ Can save 1 cycle by using SMULL SMLAL - at the cost of being
   1052 	@ 1 off.
   1053 	SMULL	r0, r3, r6, r14		@ (r0,r3)   = s0*T[1]
   1054 	SMULL	r0, r4, r11,r12		@ (r0,r4)  += s1*T[0] = s2
   1055 	ADD	r3, r3, r4
   1056 	SMULL	r0, r14,r11,r14		@ (r0,r14)  = s1*T[1]
   1057 	SMULL	r0, r12,r6, r12		@ (r0,r12) += s0*T[0] = s3
   1058 	SUB	r14,r14,r12
   1059 
   1060 	@ r9 = s0b<<1
   1061 	@ r10= s1b<<1
   1062 	ADD	r9, r3, r9, ASR #1	@ r9 = s0b + s2
   1063 	SUB	r3, r9, r3, LSL #1	@ r3 = s0b - s2
   1064 
   1065 	SUB	r12,r14,r10,ASR #1	@ r12= s3  - s1b
   1066 	ADD	r10,r14,r10,ASR #1	@ r10= s3  + s1b
   1067 	STR	r9, [r1],#4
   1068 	STR	r10,[r1],#4		@ w0 += 2
   1069 	STR	r3, [r7]
   1070 	STR	r12,[r7,#4]
   1071 
   1072 	CMP	r5,r8
   1073 	BLT	step7_loop1
   1074 
   1075 step7_loop2:
   1076 	LDR	r6, [r1]		@ r6 = w0[0]
   1077 	LDR	r9, [r1,#4]		@ r9 = w0[1]
   1078 	LDR	r10,[r7,#-8]!		@ r10= w1[0]	w1 -= 2
   1079 	LDR	r11,[r7,#4]		@ r11= w1[1]
   1080 	LDR	r14,[r5,-r2,LSL #2]!	@ r12= T[1]	T -= step
   1081 	LDR	r12,[r5,#4]		@ r14= T[0]
   1082 
   1083 	ADD	r6, r6, r10		@ r6 = s0 = w0[0] + w1[0]
   1084 	SUB	r10,r6, r10,LSL #1	@ r10= s1b= w0[0] - w1[0]
   1085 	SUB	r11,r11,r9		@ r11= s1 = w1[1] - w0[1]
   1086 	ADD	r9, r11,r9, LSL #1	@ r9 = s0b= w1[1] + w0[1]
   1087 
   1088 	@ Can save 1 cycle by using SMULL SMLAL - at the cost of being
   1089 	@ 1 off.
   1090 	SMULL	r0, r3, r6, r14		@ (r0,r3)   = s0*T[0]
   1091 	SMULL	r0, r4, r11,r12		@ (r0,r4)  += s1*T[1] = s2
   1092 	ADD	r3, r3, r4
   1093 	SMULL	r0, r14,r11,r14		@ (r0,r14)  = s1*T[0]
   1094 	SMULL	r0, r12,r6, r12		@ (r0,r12) += s0*T[1] = s3
   1095 	SUB	r14,r14,r12
   1096 
   1097 	@ r9 = s0b<<1
   1098 	@ r10= s1b<<1
   1099 	ADD	r9, r3, r9, ASR #1	@ r9 = s0b + s2
   1100 	SUB	r3, r9, r3, LSL #1	@ r3 = s0b - s2
   1101 
   1102 	SUB	r12,r14,r10,ASR #1	@ r12= s3  - s1b
   1103 	ADD	r10,r14,r10,ASR #1	@ r10= s3  + s1b
   1104 	STR	r9, [r1],#4
   1105 	STR	r10,[r1],#4		@ w0 += 2
   1106 	STR	r3, [r7]
   1107 	STR	r12,[r7,#4]
   1108 
   1109 	CMP	r1,r7
   1110 	BLT	step7_loop2
   1111 
   1112 	LDMFD	r13!,{r0-r3}
   1113 
   1114 	@ r0 = points
   1115 	@ r1 = in
   1116 	@ r2 = step
   1117 	@ r3 = shift
   1118 	MOV	r2, r2, ASR #2		@ r2 = step >>= 2
   1119 	CMP	r2, #0
   1120 	CMPNE	r2, #1
   1121 	BEQ	mdct_end
   1122 
   1123 	@ step > 1 (default case)
   1124 	CMP	r2, #4			@ r5 = T = (step>=4) ?
   1125 	ADR	r7, .Lsincos_lookup	@          sincos_lookup0 +
   1126 	ADDLT	r7, #4			@          sincos_lookup1
   1127 	LDR	r5, [r7]
   1128 	ADD	r5, r7
   1129 	ADD	r7, r1, r0, LSL #1	@ r7 = iX = x + (n>>1)
   1130 	ADDGE	r5, r5, r2, LSL #1	@		            (step>>1)
   1131 mdct_step8_default:
   1132 	LDR	r6, [r1],#4		@ r6 =  s0 = x[0]
   1133 	LDR	r8, [r1],#4		@ r8 = -s1 = x[1]
   1134 	LDR	r12,[r5,#4]       	@ r12= T[1]
   1135 	LDR	r14,[r5],r2,LSL #2	@ r14= T[0]	T += step
   1136 	RSB	r8, r8, #0		@ r8 = s1
   1137 
   1138 	@ XPROD31(s0, s1, T[0], T[1], x, x+1)
   1139 	@ x[0] = s0 * T[0] + s1 * T[1]      x[1] = s1 * T[0] - s0 * T[1]
   1140 	SMULL	r9, r10, r8, r12	@ (r9,r10)  = s1 * T[1]
   1141 	CMP	r1, r7
   1142 	SMLAL	r9, r10, r6, r14	@ (r9,r10) += s0 * T[0]
   1143 	RSB	r6, r6, #0		@ r6 = -s0
   1144 	SMULL	r9, r11, r8, r14	@ (r9,r11)  = s1 * T[0]
   1145 	MOV	r10,r10,LSL #1
   1146 	SMLAL	r9, r11, r6, r12	@ (r9,r11) -= s0 * T[1]
   1147 	STR	r10,[r1,#-8]
   1148 	MOV	r11,r11,LSL #1
   1149 	STR	r11,[r1,#-4]
   1150 	BLT	mdct_step8_default
   1151 
   1152 mdct_end:
   1153 	MOV	r0, r2
   1154 	LDMFD	r13!,{r4-r11,PC}
   1155 
   1156 cPI1_8:
   1157 	.word	0x7641af3d
   1158 cPI2_8:
   1159 	.word	0x5a82799a
   1160 cPI3_8:
   1161 	.word	0x30fbc54d
   1162 bitrev:
   1163 	.byte	0
   1164 	.byte	32
   1165 	.byte	16
   1166 	.byte	48
   1167 	.byte	8
   1168 	.byte	40
   1169 	.byte	24
   1170 	.byte	56
   1171 	.byte	4
   1172 	.byte	36
   1173 	.byte	20
   1174 	.byte	52
   1175 	.byte	12
   1176 	.byte	44
   1177 	.byte	28
   1178 	.byte	60
   1179 	.byte	2
   1180 	.byte	34
   1181 	.byte	18
   1182 	.byte	50
   1183 	.byte	10
   1184 	.byte	42
   1185 	.byte	26
   1186 	.byte	58
   1187 	.byte	6
   1188 	.byte	38
   1189 	.byte	22
   1190 	.byte	54
   1191 	.byte	14
   1192 	.byte	46
   1193 	.byte	30
   1194 	.byte	62
   1195 	.byte	1
   1196 	.byte	33
   1197 	.byte	17
   1198 	.byte	49
   1199 	.byte	9
   1200 	.byte	41
   1201 	.byte	25
   1202 	.byte	57
   1203 	.byte	5
   1204 	.byte	37
   1205 	.byte	21
   1206 	.byte	53
   1207 	.byte	13
   1208 	.byte	45
   1209 	.byte	29
   1210 	.byte	61
   1211 	.byte	3
   1212 	.byte	35
   1213 	.byte	19
   1214 	.byte	51
   1215 	.byte	11
   1216 	.byte	43
   1217 	.byte	27
   1218 	.byte	59
   1219 	.byte	7
   1220 	.byte	39
   1221 	.byte	23
   1222 	.byte	55
   1223 	.byte	15
   1224 	.byte	47
   1225 	.byte	31
   1226 	.byte	63
   1227 
   1228 .Lsincos_lookup:
   1229 	.word	sincos_lookup0-.Lsincos_lookup
   1230 	.word	sincos_lookup1-(.Lsincos_lookup+4)
   1231 
   1232 	@ END
   1233