summaryrefslogtreecommitdiff
path: root/firmware/target/arm/ipod/video/lcd-as-video.S
diff options
context:
space:
mode:
Diffstat (limited to 'firmware/target/arm/ipod/video/lcd-as-video.S')
-rw-r--r--firmware/target/arm/ipod/video/lcd-as-video.S237
1 files changed, 237 insertions, 0 deletions
diff --git a/firmware/target/arm/ipod/video/lcd-as-video.S b/firmware/target/arm/ipod/video/lcd-as-video.S
index 1b982c75ce..47155b8c75 100644
--- a/firmware/target/arm/ipod/video/lcd-as-video.S
+++ b/firmware/target/arm/ipod/video/lcd-as-video.S
@@ -63,3 +63,240 @@ lcd_write_data: /* r1 = pixel count, must be even */
63 strne r3, [lr] 63 strne r3, [lr]
64 64
65 ldmpc regs=r4 65 ldmpc regs=r4
66
67/****************************************************************************
68 * extern void lcd_write_yuv420_lines(unsigned char const * const src[3],
69 * unsigned bcmaddr
70 * int width,
71 * int stride);
72 *
73 * Conversion from Motion JPEG and MPEG Y'PbPr to RGB is:
74 * |R| |1.164 0.000 1.596| |Y' - 16|
75 * |G| = |1.164 -0.391 -0.813| |Pb - 128|
76 * |B| |1.164 2.018 0.000| |Pr - 128|
77 *
78 * Scaled, normalized, rounded and tweaked to yield RGB 565:
79 * |R| |74 0 101| |Y' - 16| >> 9
80 * |G| = |74 -24 -51| |Cb - 128| >> 8
81 * |B| |74 128 0| |Cr - 128| >> 9
82 *
83 * Converts two lines from YUV to RGB565 and writes to BCM at once. First loop
84 * loads Cb/Cr, calculates the chroma offset and saves them to buffer. Within
85 * the second loop these chroma offset are reloaded from buffer.
86 * Within each loop two pixels are calculated and written to BCM. Before each
87 * loop the desired destination address is transmitted to BCM.
88 */
89 .align 2
90 .global lcd_write_yuv420_lines
91 .type lcd_write_yuv420_lines, %function
92lcd_write_yuv420_lines:
93 /* r0 = src = yuv_src */
94 /* r1 = dst = bcmaddr */
95 /* r2 = width */
96 /* r3 = stride */
97 stmfd sp!, { r4-r10, lr } /* save non-scratch */
98 ldmia r0, { r9, r10, r12 } /* r9 = yuv_src[0] = Y'_p */
99 /* r10 = yuv_src[1] = Cb_p */
100 /* r12 = yuv_src[2] = Cr_p */
101 add r3, r9, r3 /* r3 = &ysrc[stride] */
102 add r4, r2, r2, asr #1 /* chroma buffer lenght = width/2 *3 */
103 mov r4, r4, asl #2 /* use words for str/ldm possibility */
104 add r4, r4, #19 /* plus room for 4 additional words, */
105 bic r4, r4, #3 /* rounded up to multiples of 4 byte */
106 sub sp, sp, r4 /* and allocate on stack */
107 stmia sp, {r1-r4} /* bcmaddr, width, &ysrc[stride], stack_alloc */
108
109 mov r7, r2 /* r7 = loop count */
110 add r8, sp, #16 /* chroma buffer */
111 mov lr, #0x30000000 /* LCD data port */
112
113 /* The following writes dest address to BCM and waits for write ready */
114 orr r2, lr, #0x00010000 /* r2 = BCM_WR_ADDR32 */
115 orr r6, lr, #0x00030000 /* r6 = BCM_CONTROL */
116 str r1, [r2] /* BCM_WR_ADDR32 = bcmaddr */
117.busy_1:
118 ldrh r1, [r6] /* while (!(BCM_CONTROL & 0x2)) */
119 tst r1, #0x2
120 beq .busy_1
121
122 /* 1st loop start */
12310: /* loop start */
124
125 ldrb r0, [r10], #1 /* r0 = *usrc++ = *Cb_p++ */
126 ldrb r1, [r12], #1 /* r1 = *vsrc++ = *Cr_p++ */
127
128 sub r0, r0, #128 /* r0 = Cb-128 */
129 sub r1, r1, #128 /* r1 = Cr-128 */
130
131 add r2, r1, r1, asl #1 /* r2 = Cr*51 + Cb*24 */
132 add r2, r2, r2, asl #4
133 add r2, r2, r0, asl #3
134 add r2, r2, r0, asl #4
135
136 add r4, r1, r1, asl #2 /* r1 = Cr*101 */
137 add r4, r4, r1, asl #5
138 add r1, r4, r1, asl #6
139
140 add r1, r1, #256 /* r1 = rv = (r1 + 256) >> 9 */
141 mov r1, r1, asr #9
142 rsb r2, r2, #128 /* r2 = guv = (-r2 + 128) >> 8 */
143 mov r2, r2, asr #8
144 add r0, r0, #2 /* r0 = bu = (Cb*128 + 256) >> 9 */
145 mov r0, r0, asr #2
146 stmia r8!, {r0-r2} /* store r0, r1 and r2 to chroma buffer */
147
148 /* 1st loop, first pixel */
149 ldrb r5, [r9], #1 /* r5 = *ysrc++ = *Y'_p++ */
150 sub r5, r5, #16 /* r5 = (Y'-16) * 74 */
151 add r3, r5, r5, asl #2
152 add r5, r3, r5, asl #5
153
154 add r6, r1, r5, asr #8 /* r6 = r = (Y >> 9) + rv */
155 add r3, r2, r5, asr #7 /* r3 = g = (Y >> 8) + guv */
156 add r4, r0, r5, asr #8 /* r4 = b = (Y >> 9) + bu */
157
158 orr r5, r6, r4 /* check if clamping is needed... */
159 orr r5, r5, r3, asr #1 /* ...at all */
160 cmp r5, #31
161 bls 15f /* -> no clamp */
162 cmp r6, #31 /* clamp r */
163 mvnhi r6, r6, asr #31
164 andhi r6, r6, #31
165 cmp r3, #63 /* clamp g */
166 mvnhi r3, r3, asr #31
167 andhi r3, r3, #63
168 cmp r4, #31 /* clamp b */
169 mvnhi r4, r4, asr #31
170 andhi r4, r4, #31
17115: /* no clamp */
172
173 /* calculate pixel_1 and save to r5 for later pixel packing */
174 orr r4, r4, r3, lsl #5 /* pixel_1 = r<<11 | g<<5 | b */
175 orr r5, r4, r6, lsl #11 /* r5 = pixel_1 */
176
177 /* 1st loop, second pixel */
178 ldrb r4, [r9], #1 /* r4 = *ysrc++ = *Y'_p++ */
179 sub r4, r4, #16 /* r4 = (Y'-16) * 74 */
180 add r3, r4, r4, asl #2
181 add r4, r3, r4, asl #5
182
183 add r6, r1, r4, asr #8 /* r6 = r = (Y >> 9) + rv */
184 add r3, r2, r4, asr #7 /* r3 = g = (Y >> 8) + guv */
185 add r4, r0, r4, asr #8 /* r4 = b = (Y >> 9) + bu */
186
187 orr r0, r6, r4 /* check if clamping is needed... */
188 orr r0, r0, r3, asr #1 /* ...at all */
189 cmp r0, #31
190 bls 15f /* -> no clamp */
191 cmp r6, #31 /* clamp r */
192 mvnhi r6, r6, asr #31
193 andhi r6, r6, #31
194 cmp r3, #63 /* clamp g */
195 mvnhi r3, r3, asr #31
196 andhi r3, r3, #63
197 cmp r4, #31 /* clamp b */
198 mvnhi r4, r4, asr #31
199 andhi r4, r4, #31
20015: /* no clamp */
201
202 /* calculate pixel_2 and pack with pixel_1 before writing */
203 orr r4, r4, r3, lsl #5 /* pixel_2 = r<<11 | g<<5 | b */
204 orr r4, r4, r6, lsl #11 /* r4 = pixel_2 */
205 orr r4, r5, r4, lsl #16 /* r4 = pixel_2<<16 | pixel_1 */
206 str r4, [lr] /* write packed pixels */
207
208 subs r7, r7, #2 /* check for loop end */
209 bgt 10b /* back to beginning */
210 /* 1st loop end */
211
212 /* Reload several registers for pointer rewinding for next loop */
213 add r8, sp, #16 /* chroma buffer */
214 ldmia sp, { r1, r7, r9} /* r1 = bcmaddr */
215 /* r7 = loop count */
216 /* r9 = &ysrc[stride] */
217
218 /* The following writes dest address to BCM and waits for write ready */
219 orr r2, lr, #0x00010000 /* r2 = BCM_WR_ADDR32 */
220 orr r6, lr, #0x00030000 /* r6 = BCM_CONTROL */
221 add r1, r1, #640 /* dst += (LCD_WIDTH*2) */
222 str r1, [r2] /* BCM_WR_ADDR32 = dst */
223.busy_2:
224 ldrh r1, [r6] /* while (!(BCM_CONTROL & 0x2)) */
225 tst r1, #0x2
226 beq .busy_2
227
228
229 /* 2nd loop start */
23020: /* loop start */
231 /* restore r0 (bu), r1 (rv) and r2 (guv) from chroma buffer */
232 ldmia r8!, {r0-r2}
233
234 /* 2nd loop, first pixel */
235 ldrb r5, [r9], #1 /* r5 = *ysrc++ = *Y'_p++ */
236 sub r5, r5, #16 /* r5 = (Y'-16) * 74 */
237 add r3, r5, r5, asl #2
238 add r5, r3, r5, asl #5
239
240 add r6, r1, r5, asr #8 /* r6 = r = (Y >> 9) + rv */
241 add r3, r2, r5, asr #7 /* r3 = g = (Y >> 8) + guv */
242 add r4, r0, r5, asr #8 /* r4 = b = (Y >> 9) + bu */
243
244 orr r5, r6, r4 /* check if clamping is needed... */
245 orr r5, r5, r3, asr #1 /* ...at all */
246 cmp r5, #31
247 bls 15f /* -> no clamp */
248 cmp r6, #31 /* clamp r */
249 mvnhi r6, r6, asr #31
250 andhi r6, r6, #31
251 cmp r3, #63 /* clamp g */
252 mvnhi r3, r3, asr #31
253 andhi r3, r3, #63
254 cmp r4, #31 /* clamp b */
255 mvnhi r4, r4, asr #31
256 andhi r4, r4, #31
25715: /* no clamp */
258 /* calculate pixel_1 and save to r5 for later pixel packing */
259 orr r4, r4, r3, lsl #5 /* pixel_1 = r<<11 | g<<5 | b */
260 orr r5, r4, r6, lsl #11 /* r5 = pixel_1 */
261
262 /* 2nd loop, second pixel */
263 ldrb r4, [r9], #1 /* r4 = *ysrc++ = *Y'_p++ */
264 sub r4, r4, #16 /* r4 = (Y'-16) * 74 */
265 add r3, r4, r4, asl #2
266 add r4, r3, r4, asl #5
267
268 add r6, r1, r4, asr #8 /* r6 = r = (Y >> 9) + rv */
269 add r3, r2, r4, asr #7 /* r3 = g = (Y >> 8) + guv */
270 add r4, r0, r4, asr #8 /* r4 = b = (Y >> 9) + bu */
271
272 orr r0, r6, r4 /* check if clamping is needed... */
273 orr r0, r0, r3, asr #1 /* ...at all */
274 cmp r0, #31
275 bls 15f /* -> no clamp */
276 cmp r6, #31 /* clamp r */
277 mvnhi r6, r6, asr #31
278 andhi r6, r6, #31
279 cmp r3, #63 /* clamp g */
280 mvnhi r3, r3, asr #31
281 andhi r3, r3, #63
282 cmp r4, #31 /* clamp b */
283 mvnhi r4, r4, asr #31
284 andhi r4, r4, #31
28515: /* no clamp */
286
287 /* calculate pixel_2 and pack with pixel_1 before writing */
288 orr r4, r4, r3, lsl #5 /* pixel_2 = r<<11 | g<<5 | b */
289 orr r4, r4, r6, lsl #11 /* r4 = pixel_2 */
290 orr r4, r5, r4, lsl #16 /* r4 = pixel_2<<16 | pixel_1 */
291 str r4, [lr] /* write packed pixels */
292
293 subs r7, r7, #2 /* check for loop end */
294 bgt 20b /* back to beginning */
295 /* 2nd loop end */
296
297 ldr r3, [sp, #12]
298 add sp, sp, r3 /* deallocate buffer */
299 ldmpc regs=r4-r10 /* restore registers */
300
301 .ltorg
302 .size lcd_write_yuv420_lines, .-lcd_write_yuv420_lines
generated by cgit v1.2.3 (git 2.39.1) at 2024-07-07 17:21:37 +0000