From f8709ac7e25386017cee762666c12e2ac6d71a7a Mon Sep 17 00:00:00 2001 From: Michael Sevakis Date: Sat, 26 Jan 2008 14:24:09 +0000 Subject: H10 20GB: Assembly YUV blitting and dithering. git-svn-id: svn://svn.rockbox.org/rockbox/trunk@16171 a1c6a512-1295-4272-9138-f99709370657 --- firmware/target/arm/iriver/h10/lcd-as-h10.S | 528 ++++++++++++++++++++++++++++ 1 file changed, 528 insertions(+) create mode 100644 firmware/target/arm/iriver/h10/lcd-as-h10.S (limited to 'firmware/target/arm/iriver/h10/lcd-as-h10.S') diff --git a/firmware/target/arm/iriver/h10/lcd-as-h10.S b/firmware/target/arm/iriver/h10/lcd-as-h10.S new file mode 100644 index 0000000000..f4f7b9398e --- /dev/null +++ b/firmware/target/arm/iriver/h10/lcd-as-h10.S @@ -0,0 +1,528 @@ +/*************************************************************************** + * __________ __ ___. + * Open \______ \ ____ ____ | | _\_ |__ _______ ___ + * Source | _// _ \_/ ___\| |/ /| __ \ / _ \ \/ / + * Jukebox | | ( <_> ) \___| < | \_\ ( <_> > < < + * Firmware |____|_ /\____/ \___ >__|_ \|___ /\____/__/\_ \ + * \/ \/ \/ \/ \/ + * $Id$ + * + * Copyright (C) 2007-2008 by Michael Sevakis + * + * H10 20GB LCD assembly routines + * + * All files in this archive are subject to the GNU General Public License. + * See the file COPYING in the source tree root for full license agreement. + * + * This software is distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY + * KIND, either express or implied. + * + ****************************************************************************/ + +#include "config.h" +#include "cpu.h" + +/**************************************************************************** + * void lcd_write_yuv_420_lines(unsigned char const * const src[3], + * int width, + * int stride); + * + * |R| |1.000000 -0.000001 1.402000| |Y'| + * |G| = |1.000000 -0.334136 -0.714136| |Pb| + * |B| |1.000000 1.772000 0.000000| |Pr| + * 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 + * + * Write four RGB565 pixels in the following order on each loop: + * 1 3 + > down + * 2 4 \/ left + */ + .section .icode, "ax", %progbits + .align 2 + .global lcd_write_yuv420_lines + .type lcd_write_yuv420_lines, %function +lcd_write_yuv420_lines: + @ r0 = yuv_src + @ r1 = width + @ r2 = stride + stmfd sp!, { r4-r12, lr } @ save non-scratch + ldmia r0, { r4, r5, r6 } @ r4 = yuv_src[0] = Y'_p + @ r5 = yuv_src[1] = Cb_p + @ r6 = yuv_src[2] = Cr_p + @ + mov r0, #0x7000000c @ r0 = &LCD2_PORT = 0x70008a0c + add r0, r0, #0x8a00 @ + mov r14, #LCD2_DATA_MASK @ + @ + sub r2, r2, #1 @ Adjust stride because of increment +10: @ loop line @ + ldrb r7, [r4], #1 @ r7 = *Y'_p++; + ldrb r8, [r5], #1 @ r8 = *Cb_p++; + ldrb r9, [r6], #1 @ r9 = *Cr_p++; + @ + sub r7, r7, #16 @ r7 = Y = (Y' - 16)*74 + add r12, r7, r7, asl #2 @ actually (Y' - 16)*37 and shift right + add r7, r12, r7, asl #5 @ by one less when adding - same for all + @ + sub r8, r8, #128 @ Cb -= 128 + sub r9, r9, #128 @ Cr -= 128 + @ + add r10, r9, r9, asl #1 @ r10 = Cr*51 + Cb*24 + add r10, r10, r10, asl #4 @ + add r10, r10, r8, asl #3 @ + add r10, r10, r8, asl #4 @ + @ + add r11, r9, r9, asl #2 @ r9 = Cr*101 + add r11, r11, r9, asl #5 @ + add r9, r11, r9, asl #6 @ + @ + add r8, r8, #2 @ r8 = bu = (Cb*128 + 128) >> 8 + mov r8, r8, asr #2 @ + add r9, r9, #256 @ r9 = rv = (r8 + 256) >> 9 + mov r9, r9, asr #9 @ + rsb r10, r10, #128 @ r10 = guv = (-r9 + 128) >> 8 + mov r10, r10, asr #8 @ + @ compute R, G, and B + add r3, r8, r7, asr #8 @ r3 = b = (Y >> 9) + bu + add r11, r9, r7, asr #8 @ r11 = r = (Y >> 9) + rv + add r7, r10, r7, asr #7 @ r7 = g = (Y >> 8) + guv + @ + orr r12, r3, r11 @ check if clamping is needed... + orr r12, r12, r7, asr #1 @ ...at all + cmp r12, #31 @ + bls 15f @ no clamp @ + cmp r3, #31 @ clamp b + mvnhi r3, r3, asr #31 @ + andhi r3, r3, #31 @ + cmp r11, #31 @ clamp r + mvnhi r11, r11, asr #31 @ + andhi r11, r11, #31 @ + cmp r7, #63 @ clamp g + mvnhi r7, r7, asr #31 @ + andhi r7, r7, #63 @ +15: @ no clamp @ + @ + ldrb r12, [r4, r2] @ r12 = Y' = *(Y'_p + stride) + @ + orr r3, r3, r11, lsl #11 @ r3 = b | (r << 11) + orr r3, r3, r7, lsl #5 @ r3 |= (g << 5) + @ + orr r7, r14, r3, lsr #8 @ store pixel + orr r11, r14, r3 @ +20: @ + ldr r3, [r0] @ + tst r3, #LCD2_BUSY_MASK @ + bne 20b @ + stmia r0, { r7, r11 } @ + @ + sub r7, r12, #16 @ r7 = Y = (Y' - 16)*74 + add r12, r7, r7, asl #2 @ + add r7, r12, r7, asl #5 @ + @ compute R, G, and B + add r3, r8, r7, asr #8 @ r3 = b = (Y >> 9) + bu + add r11, r9, r7, asr #8 @ r11 = r = (Y >> 9) + rv + add r7, r10, r7, asr #7 @ r7 = g = (Y >> 8) + guv + @ + orr r12, r3, r11 @ check if clamping is needed... + orr r12, r12, r7, asr #1 @ ...at all + cmp r12, #31 @ + bls 15f @ no clamp @ + cmp r3, #31 @ clamp b + mvnhi r3, r3, asr #31 @ + andhi r3, r3, #31 @ + cmp r11, #31 @ clamp r + mvnhi r11, r11, asr #31 @ + andhi r11, r11, #31 @ + cmp r7, #63 @ clamp g + mvnhi r7, r7, asr #31 @ + andhi r7, r7, #63 @ +15: @ no clamp @ + @ + ldrb r12, [r4], #1 @ r12 = Y' = *(Y'_p++) + @ + orr r3, r3, r11, lsl #11 @ r3 = b | (r << 11) + orr r3, r3, r7, lsl #5 @ r3 |= (g << 5) + @ + orr r7, r14, r3, lsr #8 @ store pixel + orr r11, r14, r3 @ +20: @ + ldr r3, [r0] @ + tst r3, #LCD2_BUSY_MASK @ + bne 20b @ + stmia r0, { r7, r11 } @ + @ + sub r7, r12, #16 @ r7 = Y = (Y' - 16)*74 + add r12, r7, r7, asl #2 @ + add r7, r12, r7, asl #5 @ + @ compute R, G, and B + add r3, r8, r7, asr #8 @ r3 = b = (Y >> 9) + bu + add r11, r9, r7, asr #8 @ r11 = r = (Y >> 9) + rv + add r7, r10, r7, asr #7 @ r7 = g = (Y >> 8) + guv + @ + orr r12, r3, r11 @ check if clamping is needed... + orr r12, r12, r7, asr #1 @ ...at all + cmp r12, #31 @ + bls 15f @ no clamp @ + cmp r3, #31 @ clamp b + mvnhi r3, r3, asr #31 @ + andhi r3, r3, #31 @ + cmp r11, #31 @ clamp r + mvnhi r11, r11, asr #31 @ + andhi r11, r11, #31 @ + cmp r7, #63 @ clamp g + mvnhi r7, r7, asr #31 @ + andhi r7, r7, #63 @ +15: @ no clamp @ + @ + ldrb r12, [r4, r2] @ r12 = Y' = *(Y'_p + stride) + @ + orr r3, r3, r7, lsl #5 @ r3 = b | (g << 5) + orr r3, r3, r11, lsl #11 @ r3 |= (r << 11) + @ + orr r7, r14, r3, lsr #8 @ store pixel + orr r11, r14, r3 @ +20: @ + ldr r3, [r0] @ + tst r3, #LCD2_BUSY_MASK @ + bne 20b @ + stmia r0, { r7, r11 } @ + @ + sub r7, r12, #16 @ r7 = Y = (Y' - 16)*74 + add r12, r7, r7, asl #2 @ + add r7, r12, r7, asl #5 @ + @ compute R, G, and B + add r3, r8, r7, asr #8 @ r3 = b = (Y >> 9) + bu + add r11, r9, r7, asr #8 @ r11 = r = (Y >> 9) + rv + add r7, r10, r7, asr #7 @ r7 = g = (Y >> 8) + guv + @ + orr r12, r3, r11 @ check if clamping is needed... + orr r12, r12, r7, asr #1 @ ...at all + cmp r12, #31 @ + bls 15f @ no clamp @ + cmp r3, #31 @ clamp b + mvnhi r3, r3, asr #31 @ + andhi r3, r3, #31 @ + cmp r11, #31 @ clamp r + mvnhi r11, r11, asr #31 @ + andhi r11, r11, #31 @ + cmp r7, #63 @ clamp g + mvnhi r7, r7, asr #31 @ + andhi r7, r7, #63 @ +15: @ no clamp @ + @ + orr r3, r3, r11, lsl #11 @ r3 = b | (r << 11) + orr r3, r3, r7, lsl #5 @ r3 |= (g << 5) + @ + orr r7, r14, r3, lsr #8 @ store pixel + orr r11, r14, r3 @ +20: @ + ldr r3, [r0] @ + tst r3, #LCD2_BUSY_MASK @ + bne 20b @ + stmia r0, { r7, r11 } @ + @ + subs r1, r1, #2 @ subtract block from width + bgt 10b @ loop line @ + @ + ldmfd sp!, { r4-r12, pc } @ restore registers and return + .ltorg @ dump constant pool + .size lcd_write_yuv420_lines, .-lcd_write_yuv420_lines + + +/**************************************************************************** + * void lcd_write_yuv_420_lines_odither(unsigned char const * const src[3], + * int width, + * int stride, + * int x_screen, + * int y_screen); + * + * |R| |1.000000 -0.000001 1.402000| |Y'| + * |G| = |1.000000 -0.334136 -0.714136| |Pb| + * |B| |1.000000 1.772000 0.000000| |Pr| + * Red scaled at twice g & b but at same precision to place it in correct + * bit position after multiply and leave instruction count lower. + * |R| |258 0 408| |Y' - 16| + * |G| = |149 -49 -104| |Cb - 128| + * |B| |149 258 0| |Cr - 128| + * + * Write four RGB565 pixels in the following order on each loop: + * 1 3 + > down + * 2 4 \/ left + * + * Kernel pattern (raw|use order): + * 5 3 4 2 row0 row2 > down + * 1 7 0 6 | 5 1 3 7 4 0 2 6 col0 left + * 4 2 5 3 | 4 0 2 6 5 1 3 7 col2 \/ + * 0 6 1 7 + */ + .section .icode, "ax", %progbits + .align 2 + .global lcd_write_yuv420_lines_odither + .type lcd_write_yuv420_lines_odither, %function +lcd_write_yuv420_lines_odither: + @ r0 = yuv_src + @ r1 = width + @ r2 = stride + @ r3 = x_screen + @ [sp] = y_screen + stmfd sp!, { r4-r12, lr } @ save non-scratch + ldmia r0, { r4, r5, r6 } @ r4 = yuv_src[0] = Y'_p + @ r5 = yuv_src[1] = Cb_p + @ r6 = yuv_src[2] = Cr_p + @ + ldr r0, [sp, #40] @ Line up pattern and kernel quadrant + eor r14, r3, r0 @ + and r14, r14, #0x2 @ + mov r14, r14, lsl #6 @ 0x00 or 0x80 + @ + mov r0, #0x7000000c @ r0 = &LCD2_PORT = 0x70008a0c + add r0, r0, #0x8a00 @ + @ + sub r2, r2, #1 @ Adjust stride because of increment +10: @ loop line @ + @ + ldrb r7, [r4], #1 @ r7 = *Y'_p++; + ldrb r8, [r5], #1 @ r8 = *Cb_p++; + ldrb r9, [r6], #1 @ r9 = *Cr_p++; + @ + eor r14, r14, #0x80 @ flip pattern quadrant + @ + sub r7, r7, #16 @ r7 = Y = (Y' - 16)*149 + add r12, r7, r7, asl #2 @ + add r12, r12, r12, asl #4 @ + add r7, r12, r7, asl #6 @ + @ + sub r8, r8, #128 @ Cb -= 128 + sub r9, r9, #128 @ Cr -= 128 + @ + add r10, r8, r8, asl #4 @ r10 = guv = Cr*104 + Cb*49 + add r10, r10, r8, asl #5 @ + add r10, r10, r9, asl #3 @ + add r10, r10, r9, asl #5 @ + add r10, r10, r9, asl #6 @ + @ + mov r8, r8, asl #1 @ r8 = bu = Cb*258 + add r8, r8, r8, asl #7 @ + @ + add r9, r9, r9, asl #1 @ r9 = rv = Cr*408 + add r9, r9, r9, asl #4 @ + mov r9, r9, asl #3 @ + @ + @ compute R, G, and B + add r3, r8, r7 @ r3 = b' = Y + bu + add r11, r9, r7, asl #1 @ r11 = r' = Y*2 + rv + rsb r7, r10, r7 @ r7 = g' = Y + guv + @ + @ r8 = bu, r9 = rv, r10 = guv + @ + sub r12, r3, r3, lsr #5 @ r3 = 31/32*b + b/256 + add r3, r12, r3, lsr #8 @ + @ + sub r12, r11, r11, lsr #5 @ r11 = 31/32*r + r/256 + add r11, r12, r11, lsr #8 @ + @ + sub r12, r7, r7, lsr #6 @ r7 = 63/64*g + g/256 + add r7, r12, r7, lsr #8 @ + @ + add r12, r14, #0x200 @ + @ + add r3, r3, r12 @ b = r3 + delta + add r11, r11, r12, lsl #1 @ r = r11 + delta*2 + add r7, r7, r12, lsr #1 @ g = r7 + delta/2 + @ + orr r12, r3, r11, asr #1 @ check if clamping is needed... + orr r12, r12, r7 @ ...at all + movs r12, r12, asr #15 @ + beq 15f @ no clamp @ + movs r12, r3, asr #15 @ clamp b + mvnne r3, r12, lsr #15 @ + andne r3, r3, #0x7c00 @ mask b only if clamped + movs r12, r11, asr #16 @ clamp r + mvnne r11, r12, lsr #16 @ + movs r12, r7, asr #15 @ clamp g + mvnne r7, r12, lsr #15 @ +15: @ no clamp @ + @ + ldrb r12, [r4, r2] @ r12 = Y' = *(Y'_p + stride) + @ + and r11, r11, #0xf800 @ pack pixel + and r7, r7, #0x7e00 @ r3 = pixel = (r & 0xf800) | + orr r11, r11, r7, lsr #4 @ ((g & 0x7e00) >> 4) | + orr r3, r11, r3, lsr #10 @ (b >> 10) + @ + mov r11, #LCD2_DATA_MASK @ store pixel + orr r7, r11, r3, lsr #8 @ + orr r11, r11, r3 @ +20: @ + ldr r3, [r0] @ + tst r3, #LCD2_BUSY_MASK @ + bne 20b @ + stmia r0, { r7, r11 } @ + @ + sub r7, r12, #16 @ r7 = Y = (Y' - 16)*149 + add r12, r7, r7, asl #2 @ + add r12, r12, r12, asl #4 @ + add r7, r12, r7, asl #6 @ + @ compute R, G, and B + add r3, r8, r7 @ r3 = b' = Y + bu + add r11, r9, r7, asl #1 @ r11 = r' = Y*2 + rv + rsb r7, r10, r7 @ r7 = g' = Y + guv + @ + sub r12, r3, r3, lsr #5 @ r3 = 31/32*b' + b'/256 + add r3, r12, r3, lsr #8 @ + @ + sub r12, r11, r11, lsr #5 @ r11 = 31/32*r' + r'/256 + add r11, r12, r11, lsr #8 @ + @ + sub r12, r7, r7, lsr #6 @ r7 = 63/64*g' + g'/256 + add r7, r12, r7, lsr #8 @ + @ + @ This element is zero - use r14 @ + @ + add r3, r3, r14 @ b = r3 + delta + add r11, r11, r14, lsl #1 @ r = r11 + delta*2 + add r7, r7, r14, lsr #1 @ g = r7 + delta/2 + @ + orr r12, r3, r11, asr #1 @ check if clamping is needed... + orr r12, r12, r7 @ ...at all + movs r12, r12, asr #15 @ + beq 15f @ no clamp @ + movs r12, r3, asr #15 @ clamp b + mvnne r3, r12, lsr #15 @ + andne r3, r3, #0x7c00 @ mask b only if clamped + movs r12, r11, asr #16 @ clamp r + mvnne r11, r12, lsr #16 @ + movs r12, r7, asr #15 @ clamp g + mvnne r7, r12, lsr #15 @ +15: @ no clamp @ + @ + ldrb r12, [r4], #1 @ r12 = Y' = *(Y'_p++) + @ + and r11, r11, #0xf800 @ pack pixel + and r7, r7, #0x7e00 @ r3 = pixel = (r & 0xf800) | + orr r11, r11, r7, lsr #4 @ ((g & 0x7e00) >> 4) | + orr r3, r11, r3, lsr #10 @ (b >> 10) + @ + mov r11, #LCD2_DATA_MASK @ store pixel + orr r7, r11, r3, lsr #8 @ + orr r11, r11, r3 @ +20: @ + ldr r3, [r0] @ + tst r3, #LCD2_BUSY_MASK @ + bne 20b @ + stmia r0, { r7, r11 } @ + @ + sub r7, r12, #16 @ r7 = Y = (Y' - 16)*149 + add r12, r7, r7, asl #2 @ + add r12, r12, r12, asl #4 @ + add r7, r12, r7, asl #6 @ + @ compute R, G, and B + add r3, r8, r7 @ r3 = b' = Y + bu + add r11, r9, r7, asl #1 @ r11 = r' = Y*2 + rv + rsb r7, r10, r7 @ r7 = g' = Y + guv + @ + @ r8 = bu, r9 = rv, r10 = guv + @ + sub r12, r3, r3, lsr #5 @ r3 = 31/32*b' + b'/256 + add r3, r12, r3, lsr #8 @ + @ + sub r12, r11, r11, lsr #5 @ r11 = 31/32*r' + r'/256 + add r11, r12, r11, lsr #8 @ + @ + sub r12, r7, r7, lsr #6 @ r7 = 63/64*g' + g'/256 + add r7, r12, r7, lsr #8 @ + @ + add r12, r14, #0x100 @ + @ + add r3, r3, r12 @ b = r3 + delta + add r11, r11, r12, lsl #1 @ r = r11 + delta*2 + add r7, r7, r12, lsr #1 @ g = r7 + delta/2 + @ + orr r12, r3, r11, asr #1 @ check if clamping is needed... + orr r12, r12, r7 @ ...at all + movs r12, r12, asr #15 @ + beq 15f @ no clamp @ + movs r12, r3, asr #15 @ clamp b + mvnne r3, r12, lsr #15 @ + andne r3, r3, #0x7c00 @ mask b only if clamped + movs r12, r11, asr #16 @ clamp r + mvnne r11, r12, lsr #16 @ + movs r12, r7, asr #15 @ clamp g + mvnne r7, r12, lsr #15 @ +15: @ no clamp @ + @ + ldrb r12, [r4, r2] @ r12 = Y' = *(Y'_p + stride) + @ + and r11, r11, #0xf800 @ pack pixel + and r7, r7, #0x7e00 @ r3 = pixel = (r & 0xf800) | + orr r11, r11, r7, lsr #4 @ ((g & 0x7e00) >> 4) | + orr r3, r11, r3, lsr #10 @ (b >> 10) + @ + mov r11, #LCD2_DATA_MASK @ store pixel + orr r7, r11, r3, lsr #8 @ + orr r11, r11, r3 @ +20: @ + ldr r3, [r0] @ + tst r3, #LCD2_BUSY_MASK @ + bne 20b @ + stmia r0, { r7, r11 } @ + @ + sub r7, r12, #16 @ r7 = Y = (Y' - 16)*149 + add r12, r7, r7, asl #2 @ + add r12, r12, r12, asl #4 @ + add r7, r12, r7, asl #6 @ + @ compute R, G, and B + add r3, r8, r7 @ r3 = b' = Y + bu + add r11, r9, r7, asl #1 @ r11 = r' = Y*2 + rv + rsb r7, r10, r7 @ r7 = g' = Y + guv + @ + sub r12, r3, r3, lsr #5 @ r3 = 31/32*b + b/256 + add r3, r12, r3, lsr #8 @ + @ + sub r12, r11, r11, lsr #5 @ r11 = 31/32*r + r/256 + add r11, r12, r11, lsr #8 @ + @ + sub r12, r7, r7, lsr #6 @ r7 = 63/64*g + g/256 + add r7, r12, r7, lsr #8 @ + @ + add r12, r14, #0x300 @ + @ + add r3, r3, r12 @ b = r3 + delta + add r11, r11, r12, lsl #1 @ r = r11 + delta*2 + add r7, r7, r12, lsr #1 @ g = r7 + delta/2 + @ + orr r12, r3, r11, asr #1 @ check if clamping is needed... + orr r12, r12, r7 @ ...at all + movs r12, r12, asr #15 @ + beq 15f @ no clamp @ + movs r12, r3, asr #15 @ clamp b + mvnne r3, r12, lsr #15 @ + andne r3, r3, #0x7c00 @ mask b only if clamped + movs r12, r11, asr #16 @ clamp r + mvnne r11, r12, lsr #16 @ + movs r12, r7, asr #15 @ clamp g + mvnne r7, r12, lsr #15 @ +15: @ no clamp @ + @ + and r11, r11, #0xf800 @ pack pixel + and r7, r7, #0x7e00 @ r3 = pixel = (r & 0xf800) | + orr r11, r11, r7, lsr #4 @ ((g & 0x7e00) >> 4) | + orr r3, r11, r3, lsr #10 @ (b >> 10) + @ + mov r11, #LCD2_DATA_MASK @ store pixel + orr r7, r11, r3, lsr #8 @ + orr r11, r11, r3 @ +20: @ + ldr r3, [r0] @ + tst r3, #LCD2_BUSY_MASK @ + bne 20b @ + stmia r0, { r7, r11 } @ + @ + subs r1, r1, #2 @ subtract block from width + bgt 10b @ loop line @ + @ + ldmfd sp!, { r4-r12, pc } @ restore registers and return + .ltorg @ dump constant pool + .size lcd_write_yuv420_lines_odither, .-lcd_write_yuv420_lines_odither -- cgit v1.2.3