diff options
Diffstat (limited to 'firmware/target/arm/ipod/lcd-as-color-nano.S')
| -rw-r--r-- | firmware/target/arm/ipod/lcd-as-color-nano.S | 287 |
1 files changed, 287 insertions, 0 deletions
diff --git a/firmware/target/arm/ipod/lcd-as-color-nano.S b/firmware/target/arm/ipod/lcd-as-color-nano.S new file mode 100644 index 0000000000..f6f9cc5be3 --- /dev/null +++ b/firmware/target/arm/ipod/lcd-as-color-nano.S @@ -0,0 +1,287 @@ +/*************************************************************************** + * __________ __ ___. + * Open \______ \ ____ ____ | | _\_ |__ _______ ___ + * Source | _// _ \_/ ___\| |/ /| __ \ / _ \ \/ / + * Jukebox | | ( <_> ) \___| < | \_\ ( <_> > < < + * Firmware |____|_ /\____/ \___ >__|_ \|___ /\____/__/\_ \ + * \/ \/ \/ \/ \/ + * $Id:$ + * + * Copyright (C) 2010-2011 by Andree Buschmann + * + * Generic asm helper function used by YUV blitting. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation; either version 2 + * of the License, or (at your option) any later version. + * + * This software is distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY + * KIND, either express or implied. + * + ****************************************************************************/ + +#include "config.h" +#include "cpu.h" + +/**************************************************************************** + * #define FORCE_FIFO_WAIT + * + * This is not needed in YUV blitting when the LCD IF is fast enough. In this + * case YUV-to-RGB conversion per pixel needs longer than the transfer of a + * pixel via the LCD IF. + ****************************************************************************/ + +#include "config.h" + +/* Set FIFO wait for both iPod Color and iPod nano1G until we know for which + * devices we can switch this off. */ +#define FORCE_FIFO_WAIT + + .section .icode, "ax", %progbits + +/**************************************************************************** + * extern void lcd_write_yuv420_lines(unsigned char const * const src[3], + * const unsigned LCD_BASE, + * int width, + * int stride); + * + * Conversion from Motion JPEG and MPEG Y'PbPr to RGB is: + * |R| |1.164 0.000 1.596| |Y' - 16| + * |G| = |1.164 -0.391 -0.813| |Pb - 128| + * |B| |1.164 2.018 0.000| |Pr - 128| + * + * Scaled, normalized, rounded and tweaked to yield RGB 565: + * |R| |74 0 101| |Y' - 16| >> 9 + * |G| = |74 -24 -51| |Cb - 128| >> 8 + * |B| |74 128 0| |Cr - 128| >> 9 + * + * Converts two lines from YUV to RGB565 and writes to LCD at once. First loop + * loads Cb/Cr, calculates the chroma offset and saves them to buffer. Within + * the second loop these chroma offset are reloaded from buffer. Within each + * loop two pixels are calculated and written to LCD. + */ + .align 2 + .global lcd_write_yuv420_lines + .type lcd_write_yuv420_lines, %function +lcd_write_yuv420_lines: + /* r0 = src = yuv_src */ + /* r1 = dst = LCD_BASE */ + /* r2 = width */ + /* r3 = stride */ + stmfd sp!, { r4-r10, lr } /* save non-scratch */ + ldmia r0, { r9, r10, r12 } /* r9 = yuv_src[0] = Y'_p */ + /* r10 = yuv_src[1] = Cb_p */ + /* r12 = yuv_src[2] = Cr_p */ + add r3, r9, r3 /* r3 = &ysrc[stride] */ + add r4, r2, r2, asr #1 /* chroma buffer lenght = width/2 *3 */ + mov r4, r4, asl #2 /* use words for str/ldm possibility */ + add r4, r4, #19 /* plus room for 4 additional words, */ + bic r4, r4, #3 /* rounded up to multiples of 4 byte */ + sub sp, sp, r4 /* and allocate on stack */ + stmia sp, {r1-r4} /* LCD_BASE, width, &ysrc[stride], stack_alloc */ + + mov r7, r2 /* r7 = loop count */ + add r8, sp, #16 /* chroma buffer */ + add lr, r1, #0x100 /* LCD data port = LCD2_BASE + 0x100 */ + + /* 1st loop start */ +10: /* loop start */ + + ldrb r0, [r10], #1 /* r0 = *usrc++ = *Cb_p++ */ + ldrb r1, [r12], #1 /* r1 = *vsrc++ = *Cr_p++ */ + + sub r0, r0, #128 /* r0 = Cb-128 */ + sub r1, r1, #128 /* r1 = Cr-128 */ + + add r2, r1, r1, asl #1 /* r2 = Cr*51 + Cb*24 */ + add r2, r2, r2, asl #4 + add r2, r2, r0, asl #3 + add r2, r2, r0, asl #4 + + add r4, r1, r1, asl #2 /* r1 = Cr*101 */ + add r4, r4, r1, asl #5 + add r1, r4, r1, asl #6 + + add r1, r1, #256 /* r1 = rv = (r1 + 256) >> 9 */ + mov r1, r1, asr #9 + rsb r2, r2, #128 /* r2 = guv = (-r2 + 128) >> 8 */ + mov r2, r2, asr #8 + add r0, r0, #2 /* r0 = bu = (Cb*128 + 256) >> 9 */ + mov r0, r0, asr #2 + stmia r8!, {r0-r2} /* store r0, r1 and r2 to chroma buffer */ + + /* 1st loop, first pixel */ + ldrb r5, [r9], #1 /* r5 = *ysrc++ = *Y'_p++ */ + sub r5, r5, #16 /* r5 = (Y'-16) * 74 */ + add r3, r5, r5, asl #2 + add r5, r3, r5, asl #5 + + add r6, r1, r5, asr #8 /* r6 = r = (Y >> 9) + rv */ + add r3, r2, r5, asr #7 /* r3 = g = (Y >> 8) + guv */ + add r4, r0, r5, asr #8 /* r4 = b = (Y >> 9) + bu */ + + orr r5, r6, r4 /* check if clamping is needed... */ + orr r5, r5, r3, asr #1 /* ...at all */ + cmp r5, #31 + bls 15f /* -> no clamp */ + cmp r6, #31 /* clamp r */ + mvnhi r6, r6, asr #31 + andhi r6, r6, #31 + cmp r3, #63 /* clamp g */ + mvnhi r3, r3, asr #31 + andhi r3, r3, #63 + cmp r4, #31 /* clamp b */ + mvnhi r4, r4, asr #31 + andhi r4, r4, #31 +15: /* no clamp */ + + /* calculate pixel_1 and save to r4 for later pixel packing */ + orr r4, r4, r3, lsl #5 /* pixel_1 = r<<11 | g<<5 | b */ + orr r4, r4, r6, lsl #11 /* r4 = pixel_1 */ + + /* 1st loop, second pixel */ + ldrb r5, [r9], #1 /* r5 = *ysrc++ = *Y'_p++ */ + sub r5, r5, #16 /* r5 = (Y'-16) * 74 */ + add r3, r5, r5, asl #2 + add r5, r3, r5, asl #5 + + add r6, r1, r5, asr #8 /* r6 = r = (Y >> 9) + rv */ + add r3, r2, r5, asr #7 /* r3 = g = (Y >> 8) + guv */ + add r5, r0, r5, asr #8 /* r5 = b = (Y >> 9) + bu */ + + orr r0, r6, r5 /* check if clamping is needed... */ + orr r0, r0, r3, asr #1 /* ...at all */ + cmp r0, #31 + bls 15f /* -> no clamp */ + cmp r6, #31 /* clamp r */ + mvnhi r6, r6, asr #31 + andhi r6, r6, #31 + cmp r3, #63 /* clamp g */ + mvnhi r3, r3, asr #31 + andhi r3, r3, #63 + cmp r5, #31 /* clamp b */ + mvnhi r5, r5, asr #31 + andhi r5, r5, #31 +15: /* no clamp */ + + /* calculate pixel_2 and pack with pixel_1 before writing */ + orr r5, r5, r3, lsl #5 /* pixel_2 = r<<11 | g<<5 | b */ + orr r5, r5, r6, lsl #11 /* r5 = pixel_2 */ +#ifdef FORCE_FIFO_WAIT + /* wait for FIFO half full */ +.fifo_wait1: + ldr r3, [lr, #-0xE0] /* while !(LCD2_BLOCK_CTRL & 0x1000000); */ + tst r3, #0x1000000 + beq .fifo_wait1 +#endif + + mov r3, r4, lsl #8 /* swap pixel_1 */ + and r3, r3, #0xff00 + add r4, r3, r4, lsr #8 + + orr r4, r4, r5, lsl #24 /* swap pixel_2 and pack with pixel_1 */ + mov r5, r5, lsr #8 + orr r4, r4, r5, lsl #16 + + str r4, [lr] /* write pixel_1 and pixel_2 */ + + subs r7, r7, #2 /* check for loop end */ + bgt 10b /* back to beginning */ + /* 1st loop end */ + + /* Reload several registers for pointer rewinding for next loop */ + add r8, sp, #16 /* chroma buffer */ + ldmia sp, { r1, r7, r9} /* r1 = LCD_BASE */ + /* r7 = loop count */ + /* r9 = &ysrc[stride] */ + + /* 2nd loop start */ +20: /* loop start */ + /* restore r0 (bu), r1 (rv) and r2 (guv) from chroma buffer */ + ldmia r8!, {r0-r2} + + /* 2nd loop, first pixel */ + ldrb r5, [r9], #1 /* r5 = *ysrc++ = *Y'_p++ */ + sub r5, r5, #16 /* r5 = (Y'-16) * 74 */ + add r3, r5, r5, asl #2 + add r5, r3, r5, asl #5 + + add r6, r1, r5, asr #8 /* r6 = r = (Y >> 9) + rv */ + add r3, r2, r5, asr #7 /* r3 = g = (Y >> 8) + guv */ + add r4, r0, r5, asr #8 /* r4 = b = (Y >> 9) + bu */ + + orr r5, r6, r4 /* check if clamping is needed... */ + orr r5, r5, r3, asr #1 /* ...at all */ + cmp r5, #31 + bls 15f /* -> no clamp */ + cmp r6, #31 /* clamp r */ + mvnhi r6, r6, asr #31 + andhi r6, r6, #31 + cmp r3, #63 /* clamp g */ + mvnhi r3, r3, asr #31 + andhi r3, r3, #63 + cmp r4, #31 /* clamp b */ + mvnhi r4, r4, asr #31 + andhi r4, r4, #31 +15: /* no clamp */ + /* calculate pixel_1 and save to r4 for later pixel packing */ + orr r4, r4, r3, lsl #5 /* pixel_1 = r<<11 | g<<5 | b */ + orr r4, r4, r6, lsl #11 /* r4 = pixel_1 */ + + /* 2nd loop, second pixel */ + ldrb r5, [r9], #1 /* r5 = *ysrc++ = *Y'_p++ */ + sub r5, r5, #16 /* r5 = (Y'-16) * 74 */ + add r3, r5, r5, asl #2 + add r5, r3, r5, asl #5 + + add r6, r1, r5, asr #8 /* r6 = r = (Y >> 9) + rv */ + add r3, r2, r5, asr #7 /* r3 = g = (Y >> 8) + guv */ + add r5, r0, r5, asr #8 /* r5 = b = (Y >> 9) + bu */ + + orr r0, r6, r5 /* check if clamping is needed... */ + orr r0, r0, r3, asr #1 /* ...at all */ + cmp r0, #31 + bls 15f /* -> no clamp */ + cmp r6, #31 /* clamp r */ + mvnhi r6, r6, asr #31 + andhi r6, r6, #31 + cmp r3, #63 /* clamp g */ + mvnhi r3, r3, asr #31 + andhi r3, r3, #63 + cmp r5, #31 /* clamp b */ + mvnhi r5, r5, asr #31 + andhi r5, r5, #31 +15: /* no clamp */ + + /* calculate pixel_2 and pack with pixel_1 before writing */ + orr r5, r5, r3, lsl #5 /* pixel_2 = r<<11 | g<<5 | b */ + orr r5, r5, r6, lsl #11 /* r5 = pixel_2 */ +#ifdef FORCE_FIFO_WAIT + /* wait for FIFO half full */ +.fifo_wait2: + ldr r3, [lr, #-0xE0] /* while !(LCD2_BLOCK_CTRL & 0x1000000); */ + tst r3, #0x1000000 + beq .fifo_wait2 +#endif + + mov r3, r4, lsl #8 /* swap pixel_1 */ + and r3, r3, #0xff00 + add r4, r3, r4, lsr #8 + + orr r4, r4, r5, lsl #24 /* swap pixel_2 and pack with pixel_1 */ + mov r5, r5, lsr #8 + orr r4, r4, r5, lsl #16 + + str r4, [lr] /* write pixel_1 and pixel_2 */ + + subs r7, r7, #2 /* check for loop end */ + bgt 20b /* back to beginning */ + /* 2nd loop end */ + + ldr r3, [sp, #12] + add sp, sp, r3 /* deallocate buffer */ + ldmpc regs=r4-r10 /* restore registers */ + + .ltorg + .size lcd_write_yuv420_lines, .-lcd_write_yuv420_lines |