|
1 // Copyright (c) 2002-2009 Nokia Corporation and/or its subsidiary(-ies). |
|
2 // All rights reserved. |
|
3 // This component and the accompanying materials are made available |
|
4 // under the terms of the License "Eclipse Public License v1.0" |
|
5 // which accompanies this distribution, and is available |
|
6 // at the URL "http://www.eclipse.org/legal/epl-v10.html". |
|
7 // |
|
8 // Initial Contributors: |
|
9 // Nokia Corporation - initial contribution. |
|
10 // |
|
11 // Contributors: |
|
12 // |
|
13 // Description: |
|
14 // e32\euser\epoc\arm\uc_huffman.cia |
|
15 // |
|
16 // |
|
17 |
|
18 #include "e32huffman.h" |
|
19 #include <cpudefs.h> |
|
20 |
|
21 #ifdef __HUFFMAN_MACHINE_CODED__ |
|
22 |
|
23 // NB The EABI layout of TBitInput is different from GCC. Its vtbl is at |
|
24 // offset 0 rather than offset 16 |
|
25 |
|
26 EXPORT_C __NAKED__ TUint TBitInput::ReadL() |
|
27 /** Read a single bit from the input |
|
28 |
|
29 Return the next bit in the input stream. This will call UnderflowL() if |
|
30 there are no more bits available. |
|
31 |
|
32 @return The next bit in the stream |
|
33 |
|
34 @leave "UnderflowL()" It the bit stream is exhausted more UnderflowL is called |
|
35 to get more data |
|
36 */ |
|
37 { |
|
38 asm(".code 32"); |
|
39 #ifdef __EABI__ |
|
40 asm("ldmib r0!, {r2,r3}"); |
|
41 #else |
|
42 asm("ldmia r0!, {r2,r3}"); |
|
43 #endif |
|
44 // r0 = this+8 |
|
45 asm("subs r2, r2, #1"); |
|
46 asm("movpl r12, r3, asl #1"); |
|
47 #ifdef __EABI__ |
|
48 asm("stmplda r0, {r2,r12}"); |
|
49 #else |
|
50 asm("stmpldb r0, {r2,r12}"); |
|
51 #endif |
|
52 asm("movpl r0, r3, lsr #31"); |
|
53 __JUMP(pl,lr); |
|
54 asm("mov r12, #0"); |
|
55 asm("b __readmorebits"); // r2==-1, r12==0 -> so all set |
|
56 } |
|
57 |
|
58 EXPORT_C __NAKED__ TUint TBitInput::ReadL(TInt aSize) |
|
59 /** Read a multi-bit value from the input |
|
60 |
|
61 Return the next few bits as an unsigned integer. The last bit read is |
|
62 the least significant bit of the returned value, and the value is |
|
63 zero extended to return a 32-bit result. |
|
64 |
|
65 A read of zero bits will always reaturn zero. |
|
66 |
|
67 This will call UnderflowL() if there are not enough bits available. |
|
68 |
|
69 @param "TInt aSize" The number of bits to read |
|
70 |
|
71 @return The bits read from the stream |
|
72 |
|
73 @leave "UnderflowL()" It the bit stream is exhausted more UnderflowL is called |
|
74 to get more data |
|
75 */ |
|
76 { |
|
77 asm(".code 32"); |
|
78 #ifdef __EABI__ |
|
79 asm(" ldmib r0!, {r2,r3}"); |
|
80 #else |
|
81 asm(" ldmia r0!, {r2,r3}"); |
|
82 #endif |
|
83 asm(" subs r2, r2, r1"); |
|
84 asm(" movpl r12, r3, asl r1"); |
|
85 #ifdef __EABI__ |
|
86 asm(" stmplda r0, {r2,r12}"); |
|
87 #else |
|
88 asm(" stmpldb r0, {r2,r12}"); |
|
89 #endif |
|
90 asm(" rsb r1, r1, #32"); |
|
91 asm(" movpl r0, r3, lsr r1"); |
|
92 __JUMP(pl,lr); |
|
93 // r0=this+8, r1=32-#reqd, r2=count-#reqd (<0), r3=bits |
|
94 asm(" sub r1, r1, r2"); // r1 = 32-iCount |
|
95 asm(" mov r3, r3, lsr r1"); // scrub existing bits |
|
96 asm(" rsb r1, r2, #0"); // r1 = #reqd-iCount |
|
97 asm(" mov r12, r3, lsl r1"); // r12 has bits in final position |
|
98 asm("__readmorebits:"); |
|
99 //r0=this+8 |
|
100 asm(" stmfd sp!, {r12,lr}"); |
|
101 |
|
102 #ifdef __EABI__ |
|
103 asm(" ldmib r0!, {r1,r14}"); // r1=remain, r14=ptr |
|
104 #else |
|
105 asm(" ldmia r0!, {r1,r14}"); // r1=remain, r14=ptr |
|
106 #endif |
|
107 //r0=this+16 |
|
108 asm(" cmp r1, #0"); |
|
109 asm(" ble __readunderflow"); |
|
110 asm(" ldr r12, [r14], #4"); |
|
111 asm(" subs r3, r1, #32"); |
|
112 #ifdef __EABI__ |
|
113 asm(" stmda r0!, {r3,r14}"); |
|
114 #else |
|
115 asm(" stmdb r0!, {r3,r14}"); |
|
116 #endif |
|
117 // r0=this+8 |
|
118 asm(" addmi r1, r2, r1"); |
|
119 asm(" addpl r1, r2, #32"); // r1 = bits left in |
|
120 asm(" eor r3, r12, r12, ror #16"); |
|
121 asm(" bic r3, r3, #16711680"); |
|
122 asm(" mov r3, r3, lsr #8"); |
|
123 asm(" eor r3, r3, r12, ror #8"); // r3=bits |
|
124 // r0=this+8, r1=count-#reqd, r2=-#reqd, r3=bits, sp[0]=accum, sp[4]=return-pc |
|
125 asm(" cmp r1, #0"); |
|
126 asm(" bmi __readgoroundagain"); |
|
127 asm("__readhavebits:"); |
|
128 asm(" rsb r12, r2, #0"); |
|
129 asm(" mov r12, r3, asl r12"); |
|
130 #ifdef __EABI__ |
|
131 asm(" stmda r0, {r1,r12}"); |
|
132 #else |
|
133 asm(" stmdb r0, {r1,r12}"); |
|
134 #endif |
|
135 asm(" ldmfd sp!, {r12,lr}"); |
|
136 asm(" add r2, r2, #32"); |
|
137 asm(" orr r0, r12, r3, lsr r2"); |
|
138 __JUMP(,lr); |
|
139 asm("__readunderflow:"); |
|
140 #ifdef __EABI__ |
|
141 asm(" ldr r3, [r0, #-16]!"); // restore r0 to this |
|
142 #else |
|
143 asm(" ldr r3, [r0], #-16"); // restore r0 to this |
|
144 #endif |
|
145 asm(" stmfd sp!, {r0,r2}"); |
|
146 // at this point the stack has grown by 16 bytes from the CFA |
|
147 __EH_FRAME_ADDRESS(sp,16) |
|
148 // The stack looks like this |
|
149 // CFA-4 -> lr |
|
150 // CFA-8 -> r12 |
|
151 // CFA-12 -> r2 |
|
152 // CFA-16 -> r0 |
|
153 // So we need to record lr @ CFA-4 |
|
154 // Nothing else needs to be restored since there's no barrier here |
|
155 __EH_FRAME_SAVE1(lr,-4) |
|
156 asm(" adr lr, 1f "); |
|
157 #ifdef __EABI__ |
|
158 asm(" ldr ip, [r3, #0]"); |
|
159 #else |
|
160 asm(" ldr ip, [r3, #8]"); |
|
161 #endif |
|
162 __JUMP(,ip); |
|
163 asm("1: "); |
|
164 asm(" ldmfd sp!, {r0,r2}"); |
|
165 #ifdef __EABI__ |
|
166 asm(" ldmib r0!, {r1,r3}"); // r1=count, r3=bits |
|
167 #else |
|
168 asm(" ldmia r0!, {r1,r3}"); // r1=count, r3=bits |
|
169 #endif |
|
170 asm(" adds r1, r1, r2"); // adjust count for bits to extract |
|
171 asm(" bpl __readhavebits"); |
|
172 asm("__readgoroundagain:"); |
|
173 // r0=this+8, r1=count-#reqd (<0), r2=-#reqd, r3=bits, sp[0]=accum, sp[4]=return-pc |
|
174 asm(" sub r12, r1, r2"); // r12 = iCount |
|
175 asm(" rsb r12, r12, #32"); // r12= 32-iCount |
|
176 asm(" mov r3, r3, lsr r12"); // scrub existing bits |
|
177 asm(" ldmfd sp!, {r12,lr}"); |
|
178 asm(" mov r2, r1"); // r2 = iCount-#reqd |
|
179 asm(" rsb r1, r1, #0"); // r1 = #reqd-iCount |
|
180 asm(" orr r12, r12, r3, lsl r1"); // r12 has bits in final position |
|
181 asm(" b __readmorebits"); |
|
182 } |
|
183 |
|
184 EXPORT_C __NAKED__ TUint TBitInput::HuffmanL(const TUint32* aTree) |
|
185 /** Read and decode a Huffman Code |
|
186 |
|
187 Interpret the next bits in the input as a Huffman code in the specified |
|
188 decoding. The decoding tree should be the output from Huffman::Decoding(). |
|
189 |
|
190 @param "const TUint32* aTree" The huffman decoding tree |
|
191 |
|
192 @return The symbol that was decoded |
|
193 |
|
194 @leave "UnderflowL()" It the bit stream is exhausted more UnderflowL is called |
|
195 to get more data |
|
196 */ |
|
197 { |
|
198 asm(".code 32"); |
|
199 asm("__huffman:"); |
|
200 #ifdef __EABI__ |
|
201 asm(" ldmib r0!, {r2,r3}"); |
|
202 #else |
|
203 asm(" ldmia r0!, {r2,r3}"); |
|
204 #endif |
|
205 asm("__decode:"); |
|
206 asm(" ldr r12, [r1]"); |
|
207 asm(" subs r2, r2, #4"); |
|
208 asm(" bmi __fewbits"); |
|
209 asm("__fastloop:"); |
|
210 asm(" movs r3, r3, asl #1"); |
|
211 asm(" movcc r12, r12, asl #16"); |
|
212 asm(" tst r12, #65536"); |
|
213 asm(" bne __fastdecoded"); |
|
214 asm(" ldr r12, [r1, r12, lsr #16]!"); |
|
215 asm(" sub r2, r2, #1"); |
|
216 asm(" movs r3, r3, asl #1"); |
|
217 asm(" movcc r12, r12, asl #16"); |
|
218 asm(" tst r12, #65536"); |
|
219 asm(" bne __fastdecoded"); |
|
220 asm(" ldr r12, [r1, r12, lsr #16]!"); |
|
221 asm(" sub r2, r2, #1"); |
|
222 asm(" movs r3, r3, asl #1"); |
|
223 asm(" movcc r12, r12, asl #16"); |
|
224 asm(" tst r12, #65536"); |
|
225 asm(" bne __fastdecoded"); |
|
226 asm(" ldr r12, [r1, r12, lsr #16]!"); |
|
227 asm(" sub r2, r2, #1"); |
|
228 asm(" movs r3, r3, asl #1"); |
|
229 asm(" movcc r12, r12, asl #16"); |
|
230 asm(" tst r12, #65536"); |
|
231 asm(" bne __fastdecoded"); |
|
232 asm(" ldr r12, [r1, r12, lsr #16]!"); |
|
233 asm(" subs r2, r2, #1"); |
|
234 asm(" bpl __fastloop"); |
|
235 asm("__fewbits:"); |
|
236 asm(" adds r2, r2, #3"); |
|
237 asm(" bmi __nobits"); |
|
238 asm("__slowloop:"); |
|
239 asm(" movs r3, r3, asl #1"); |
|
240 asm(" movcc r12, r12, asl #16"); |
|
241 asm(" tst r12, #65536"); |
|
242 asm(" bne __decoded"); |
|
243 asm(" ldr r12, [r1, r12, lsr #16]!"); |
|
244 asm(" subs r2, r2, #1"); |
|
245 asm(" bpl __slowloop"); |
|
246 asm("__nobits:"); |
|
247 // r0=this, r1=tree, r2=count (-1), r3=bits (0), r12=node=*tree |
|
248 // must preserve r0,r1 |
|
249 #ifdef __EABI__ |
|
250 asm(" ldmib r0!, {r2,r3}"); // r2=remain, r3=ptr |
|
251 #else |
|
252 asm(" ldmia r0!, {r2,r3}"); // r2=remain, r3=ptr |
|
253 #endif |
|
254 asm(" cmp r2, #0"); |
|
255 asm(" ble __underflow"); |
|
256 asm(" ldr r12, [r3], #4"); |
|
257 asm(" subs r2, r2, #32"); |
|
258 #ifdef __EABI__ |
|
259 asm(" stmda r0!, {r2,r3}"); |
|
260 #else |
|
261 asm(" stmdb r0!, {r2,r3}"); |
|
262 #endif |
|
263 asm(" addmi r2, r2, #32"); |
|
264 asm(" movpl r2, #32"); |
|
265 asm(" eor r3, r12, r12, ror #16"); |
|
266 asm(" bic r3, r3, #16711680"); |
|
267 asm(" mov r3, r3, lsr #8"); |
|
268 asm(" eor r3, r3, r12, ror #8"); |
|
269 asm(" b __decode"); |
|
270 asm("__underflow:"); |
|
271 #ifdef __EABI__ |
|
272 asm(" ldr r3, [r0, #-16]!"); // restore r0 to this |
|
273 #else |
|
274 asm(" ldr r3, [r0], #-16"); // restore r0 to this |
|
275 #endif |
|
276 asm(" stmfd sp!, {r0,r1,r2,lr}"); |
|
277 // sp is 16 bytes from CFA |
|
278 __EH_FRAME_ADDRESS(sp,16) |
|
279 // The stack looks like this: |
|
280 // CFA-4 -> lr |
|
281 // CFA-8 -> r2 |
|
282 // CFA-12 -> r1 |
|
283 // CFA-16 -> r0 |
|
284 // We saved lr @ CFA-4 |
|
285 // Nothing else needs to be restored since there's no barrier here |
|
286 __EH_FRAME_SAVE1(lr,-4) |
|
287 asm(" adr lr, 1f "); |
|
288 #ifdef __EABI__ |
|
289 asm(" ldr ip, [r3, #0]"); |
|
290 #else |
|
291 asm(" ldr ip, [r3, #8]"); |
|
292 #endif |
|
293 __JUMP(,ip); |
|
294 asm("1: "); |
|
295 asm(" ldmfd sp!, {r0,r1,r2,lr}"); |
|
296 asm(" b __huffman"); |
|
297 asm("__fastdecoded:"); |
|
298 asm(" add r2, r2, #3"); |
|
299 asm("__decoded:"); |
|
300 #ifdef __EABI__ |
|
301 asm(" stmda r0, {r2, r3}"); |
|
302 #else |
|
303 asm(" stmdb r0, {r2, r3}"); |
|
304 #endif |
|
305 asm(" mov r0, r12, lsr #17"); |
|
306 __JUMP(,lr); |
|
307 } |
|
308 |
|
309 #endif |
|
310 |