/*************************************************************************** * __________ __ ___. * Open \______ \ ____ ____ | | _\_ |__ _______ ___ * Source | _// _ \_/ ___\| |/ /| __ \ / _ \ \/ / * Jukebox | | ( <_> ) \___| < | \_\ ( <_> > < < * Firmware |____|_ /\____/ \___ >__|_ \|___ /\____/__/\_ \ * \/ \/ \/ \/ \/ * $Id$ * * Copyright (C) 2007 by Michael Sevakis * * 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_copy_buffer_rect(fb_data *dst, fb_data *src, int width, * int height); */ .section .icode, "ax", %progbits .align 2 .global lcd_copy_buffer_rect .type lcd_copy_buffer_rect, %function @ r0 = dst @ r1 = src @ r2 = width @ r3 = height lcd_copy_buffer_rect: @ stmfd sp!, { r4-r12, lr } @ save non-scratch regs mov r5, r2 @ r5 = cached width rsb r4, r2, #LCD_WIDTH @ r4 = LCD_WIDTH - width 10: @ copy line @ subs r2, r5, #1 @ r2 = width - 1 beq 40f @ finish line @ one halfword? skip to trailing copy tst r0, #2 @ word aligned? beq 20f @ rem copy @ yes? skip to word copy ldrh r6, [r1], #2 @ copy leading halfword subs r2, r2, #1 @ strh r6, [r0], #2 @ ble 40f @ finish line @ next line if lt or finish @ trailing halfword if eq 20: @ rem copy @ add r14, r2, #1 @ get remaining width mod 16 after word @ align (rw) and r14, r14, #0xe @ r14 = 0 (16), 2, 4, 6, 8, 10, 12, 14 add pc, pc, r14, lsl #3 @ branch to 32-byte align nop @ b 30f @ rw % 16 = 0 or 1? use octword loop nop @ nop @ nop @ ldr r6, [r1], #4 @ rw % 16 = 2 or 3 subs r2, r2, #2 @ str r6, [r0], #4 @ b 25f @ copy up done @ ldmia r1!, { r6-r7 } @ rw % 16 = 4 or 5 subs r2, r2, #4 @ stmia r0!, { r6-r7 } @ b 25f @ copy up done @ ldmia r1!, { r6-r8 } @ rw % 16 = 6 or 7 subs r2, r2, #6 @ stmia r0!, { r6-r8 } @ b 25f @ copy up done @ ldmia r1!, { r6-r9 } @ rw % 16 = 8 or 9 subs r2, r2, #8 @ stmia r0!, { r6-r9 } @ b 25f @ copy up done @ ldmia r1!, { r6-r10 } @ rw % 16 = 10 or 11 subs r2, r2, #10 @ stmia r0!, { r6-r10 } @ b 25f @ copy up done @ ldmia r1!, { r6-r11 } @ rw % 16 = 12 or 13 subs r2, r2, #12 @ stmia r0!, { r6-r11 } @ b 25f @ copy up done @ ldmia r1!, { r6-r12 } @ rw % 16 = 14 or 15 subs r2, r2, #14 @ stmia r0!, { r6-r12 } @ 25: @ copy up done @ ble 40f @ finish line @ no 32-byte segments remaining? 30: @ octword loop @ copy 16 pixels per loop ldmia r1!, { r6-r12, r14 } @ subs r2, r2, #16 @ stmia r0!, { r6-r12, r14 } @ bgt 30b @ octword loop @ 40: @ finish line @ ldreqh r6, [r1], #2 @ finish last halfword if eq ... add r1, r1, r4, lsl #1 @ streqh r6, [r0], #2 @ ... add r0, r0, r4, lsl #1 @ subs r3, r3, #1 @ next line bgt 10b @ copy line @ ldmfd sp!, { r4-r12, pc } @ restore regs and return .size lcd_copy_buffer_rect, .-lcd_copy_buffer_rect /**************************************************************************** * void lcd_write_yuv_420_lines(fb_data *dst, * unsigned char chroma_buf[LCD_HEIGHT/2*3], 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 */ .section .icode, "ax", %progbits .align 2 .global lcd_write_yuv420_lines .type lcd_write_yuv420_lines, %function lcd_write_yuv420_lines: @ r0 = dst @ r1 = chroma_buf @ r2 = yuv_src @ r3 = width @ [sp] = stride stmfd sp!, { r4-r12, lr } @ save non-scratch stmfd sp!, { r0, r3 } @ save dst and width mov r14, #74 @ r14 = Y factor ldmia r2, { r4, r5, r6 } @ r4 = yuv_src[0] = Y'_p @ r5 = yuv_src[1] = Cb_p @ r6 = yuv_src[2] = Cr_p 10: @ loop line 1 @ ldrb r2, [r4], #1 @ r2 = *Y'_p++; ldrb r8, [r5], #1 @ r8 = *Cb_p++; ldrb r11, [r6], #1 @ r11 = *Cr_p++; @ @ compute Y sub r2, r2, #16 @ r7 = Y = (Y' - 16)*74 mul r7, r2, r14 @ @ sub r8, r8, #128 @ Cb -= 128 sub r11, r11, #128 @ Cr -= 128 @ mvn r2, #24 @ compute guv mul r10, r2, r8 @ r10 = Cb*-24 mvn r2, #51 @ mla r10, r2, r11, r10 @ r10 = r10 + Cr*-51 @ mov r2, #101 @ compute rv mul r9, r11, r2 @ r9 = rv = Cr*101 @ @ store chromas in line buffer add r8, r8, #2 @ bu = (Cb + 2) >> 2 mov r8, r8, asr #2 @ strb r8, [r1], #1 @ add r9, r9, #256 @ rv = (Cr + 256) >> 9 mov r9, r9, asr #9 @ strb r9, [r1], #1 @ mov r10, r10, asr #8 @ guv >>= 8 strb r10, [r1], #1 @ @ compute R, G, and B add r2, r8, r7, asr #9 @ r2 = b = (Y >> 9) + bu add r11, r9, r7, asr #9 @ r11 = r = (Y >> 9) + rv add r7, r10, r7, asr #8 @ r7 = g = (Y >> 8) + guv @ orr r12, r2, r11 @ check if clamping is needed... orr r12, r12, r7, asr #1 @ ...at all cmp r12, #31 @ bls 15f @ no clamp @ mov r12, #31 @ cmp r12, r2 @ clamp b andlo r2, r12, r2, asr #31 @ eorlo r2, r2, r12 @ cmp r12, r11 @ clamp r andlo r11, r12, r11, asr #31 @ eorlo r11, r11, r12 @ cmp r12, r7, asr #1 @ clamp g andlo r7, r12, r7, asr #31 @ eorlo r7, r7, r12 @ orrlo r7, r7, r7, asl #1 @ 15: @ no clamp @ @ orr r12, r2, r7, lsl #5 @ r4 |= (g << 5) ldrb r2, [r4], #1 @ r2 = Y' = *Y'_p++ orr r12, r12, r11, lsl #11 @ r4 = b | (r << 11) strh r12, [r0], #LCD_WIDTH @ store pixel @ sub r2, r2, #16 @ r7 = Y = (Y' - 16)*74 mul r7, r2, r14 @ next Y @ compute R, G, and B add r2, r8, r7, asr #9 @ r2 = b = (Y >> 9) + bu add r11, r9, r7, asr #9 @ r11 = r = (Y >> 9) + rv add r7, r10, r7, asr #8 @ r7 = g = (Y >> 8) + guv @ orr r12, r2, r11 @ check if clamping is needed... orr r12, r12, r7, asr #1 @ ...at all cmp r12, #31 @ bls 15f @ no clamp @ mov r12, #31 @ cmp r12, r2 @ clamp b andlo r2, r12, r2, asr #31 @ eorlo r2, r2, r12 @ cmp r12, r11 @ clamp r andlo r11, r12, r11, asr #31 @ eorlo r11, r11, r12 @ cmp r12, r7, asr #1 @ clamp g andlo r7, r12, r7, asr #31 @ eorlo r7, r7, r12 @ orrlo r7, r7, r7, asl #1 @ 15: @ no clamp @ @ orr r12, r2, r11, lsl #11 @ r4 = b | (r << 11) orr r12, r12, r7, lsl #5 @ r4 |= (g << 5) strh r12, [r0, #LCD_WIDTH]! @ store pixel add r0, r0, #2*LCD_WIDTH @ @ subs r3, r3, #2 @ bgt 10b @ loop line 1 @ @ do second line @ ldmfd sp!, { r0, r3 } @ pop dst and width sub r0, r0, #2 @ set dst to start of next line sub r1, r1, r3, asl #1 @ rewind chroma pointer... ldr r2, [sp, #40] @ r2 = stride add r1, r1, r3, asr #1 @ ... (r1 -= width/2*3) @ move sources to start of next line sub r2, r2, r3 @ r2 = skip = stride - width add r4, r4, r2 @ r4 = Y'_p + skip @ 20: @ loop line 2 @ ldrb r2, [r4], #1 @ r7 = Y' = *Y'_p++ ldrsb r8, [r1], #1 @ reload saved chromas ldrsb r9, [r1], #1 @ ldrsb r10, [r1], #1 @ @ sub r2, r2, #16 @ r2 = Y = (Y' - 16)*74 mul r7, r2, r14 @ @ compute R, G, and B add r2, r8, r7, asr #9 @ r2 = b = (Y >> 9) + bu add r11, r9, r7, asr #9 @ r11 = r = (Y >> 9) + rv add r7, r10, r7, asr #8 @ r7 = g = (Y >> 8) + guv @ orr r12, r2, r11 @ check if clamping is needed... orr r12, r12, r7, asr #1 @ ...at all cmp r12, #31 @ bls 25f @ no clamp @ mov r12, #31 @ cmp r12, r2 @ clamp b andlo r2, r12, r2, asr #31 @ eorlo r2, r2, r12 @ cmp r12, r11 @ clamp r andlo r11, r12, r11, asr #31 @ eorlo r11, r11, r12 @ cmp r12, r7, asr #1 @ clamp g andlo r7, r12, r7, asr #31 @ eorlo r7, r7, r12 @ orrlo r7, r7, r7, asl #1 @ 25: @ no clamp @ @ orr r12, r2, r11, lsl #11 @ r4 = b | (r << 11) ldrb r2, [r4], #1 @ r2 = Y' = *Y'_p++ orr r12, r12, r7, lsl #5 @ r4 |= (g << 5) strh r12, [r0], #LCD_WIDTH @ store pixel @ @ do second pixel @ sub r2, r2, #16 @ r2 = Y = (Y' - 16)*74 mul r7, r2, r14 @ @ compute R, G, and B add r2, r8, r7, asr #9 @ r2 = b = (Y >> 9) + bu add r11, r9, r7, asr #9 @ r11 = r = (Y >> 9) + rv add r7, r10, r7, asr #8 @ r7 = g = (Y >> 8) + guv @ orr r12, r2, r11 @ check if clamping is needed... orr r12, r12, r7, asr #1 @ ...at all cmp r12, #31 @ bls 25f @ no clamp @ mov r12, #31 @ cmp r12, r2 @ clamp b andlo r2, r12, r2, asr #31 @ eorlo r2, r2, r12 @ cmp r12, r11 @ clamp r andlo r11, r12, r11, asr #31 @ eorlo r11, r11, r12 @ cmp r12, r7, asr #1 @ clamp g andlo r7, r12, r7, asr #31 @ eorlo r7, r7, r12 @ orrlo r7, r7, r7, asl #1 @ 25: @ no clamp @ @ orr r12, r2, r11, lsl #11 @ r4 = b | (r << 11) orr r12, r12, r7, lsl #5 @ r4 |= (g << 5) strh r12, [r0, #LCD_WIDTH]! @ store pixel add r0, r0, #2*LCD_WIDTH @ @ subs r3, r3, #2 @ bgt 20b @ loop line 2 @ @ ldmfd sp!, { r4-r12, pc } @ restore registers and return .size lcd_write_yuv420_lines, .-lcd_write_yuv420_lines