Optimized lcd_yuv_blit() (used by mpegplayer) for the Sansa E200. This is just a copy-and-paste of the Gigabeat version by Michael Sevakis with no attempt to adapt to the different CPU in the E200 - so there could be room for further improvement. Reported to increase the FPS for the 224x128 version of Elephants Dream from 21fps to 27fps.

git-svn-id: svn://svn.rockbox.org/rockbox/trunk@13042 a1c6a512-1295-4272-9138-f99709370657
author: Dave Chapman <dave@dchapman.com> 2007-04-06 09:34:13 +0000
committer: Dave Chapman <dave@dchapman.com> 2007-04-06 09:34:13 +0000
commit: 4303ab02a3efbd6a77e809641617c53082f9366f (patch)
tree: 73b2bde0cb847785981b69dc749d7223d2ae7a12
parent: 31d8068b331af435c50cacfbb0ed2de1c2221f4d (diff)
download: rockbox-4303ab02a3efbd6a77e809641617c53082f9366f.tar.gz
rockbox-4303ab02a3efbd6a77e809641617c53082f9366f.zip
3 files changed, 257 insertions, 104 deletions
diff --git a/firmware/SOURCES b/firmware/SOURCES
index 8b1d3cf896..9f90919817 100644
--- a/firmware/SOURCES
+++ b/firmware/SOURCES
@@ -353,6 +353,7 @@ target/sh/archos/ondio/button-ondio.c
 #ifndef SIMULATOR
 target/arm/sandisk/sansa-e200/ata-e200.c
 target/arm/sandisk/sansa-e200/lcd-e200.c
+target/arm/sandisk/sansa-e200/lcd-as-e200.S
 target/arm/sandisk/sansa-e200/adc-e200.c
 target/arm/sandisk/sansa-e200/backlight-e200.c
 target/arm/usb-pp.c
diff --git a/firmware/target/arm/sandisk/sansa-e200/lcd-as-e200.S b/firmware/target/arm/sandisk/sansa-e200/lcd-as-e200.S
new file mode 100644
index 0000000000..9e130cf54b
--- /dev/null
+++ b/firmware/target/arm/sandisk/sansa-e200/lcd-as-e200.S
@@ -0,0 +1,222 @@
+/***************************************************************************
+ *             __________               __   ___.
+ *   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_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
+    stmdb       sp!, { r4-r12, lr }     @ save non-scratch
+    stmdb       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
+                                        @
+    ldmia       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       @
+                                        @
+    ldmia       sp!, { r4-r12, pc }     @ restore registers and return
+    .size   lcd_write_yuv420_lines, .-lcd_write_yuv420_lines
diff --git a/firmware/target/arm/sandisk/sansa-e200/lcd-e200.c b/firmware/target/arm/sandisk/sansa-e200/lcd-e200.c
index 6b51531056..3e8246b367 100644
--- a/firmware/target/arm/sandisk/sansa-e200/lcd-e200.c
+++ b/firmware/target/arm/sandisk/sansa-e200/lcd-e200.c
@@ -329,115 +329,45 @@ void lcd_blit(const fb_data* data, int x, int by, int width,
    (void)stride;
 }
-#define CSUB_X 2
+/* Line write helper function for lcd_yuv_blit. Write two lines of yuv420. */
-#define CSUB_Y 2
+extern void lcd_write_yuv420_lines(fb_data *dst,
+                                   unsigned char chroma_buf[LCD_HEIGHT/2*3],
-#define RYFAC (31*257)
+                                   unsigned char const * const src[3],
-#define GYFAC (63*257)
+                                   int width,
-#define BYFAC (31*257)
+                                   int stride);
-#define RVFAC 11170     /* 31 * 257 *  1.402    */
+/* Performance function to blit a YUV bitmap directly to the LCD */
-#define GVFAC (-11563)  /* 63 * 257 * -0.714136 */
+/* For the e200 - show it rotated */
-#define GUFAC (-5572)   /* 63 * 257 * -0.344136 */
+/* So the LCD_WIDTH is now the height */
-#define BUFAC 14118     /* 31 * 257 *  1.772    */
-#define ROUNDOFFS (127*257)
-/* Performance function to blit a YUV bitmap directly to the LCD
-   Actually this code is from gigabeat, because this target is also
-   writing direct to a buffer. */
 void lcd_yuv_blit(unsigned char * const src[3],
                  int src_x, int src_y, int stride,
-                  int _x, int _y, int width, int height)
+                  int x, int y, int width, int height)
 {
-    const unsigned char *usrc;
+    /* Caches for chroma data so it only need be recaculated every other
-    const unsigned char *vsrc;
+       line */
-    const unsigned char *ysrc;
+    unsigned char chroma_buf[LCD_HEIGHT/2*3]; /* 480 bytes */
-    int xphase;
+    unsigned char const * yuv_src[3];
-    int rc, gc, bc;
+    off_t z;
-    int y, u, v;
-    int red, green, blue;
+    /* Sorry, but width and height must be >= 2 or else */
-    unsigned rbits, gbits, bbits;
+    width &= ~1;
-    int count;
+    height >>= 1;
-    fb_data *dst_row;
-    width = (width + 1) & ~1;
    fb_data *dst = (fb_data*)lcd_driver_framebuffer + 
-                   _x * LCD_WIDTH + (LCD_WIDTH - _y) - 1;
+                   x * LCD_WIDTH + (LCD_WIDTH - y) - 1;
-    fb_data *dst_last = dst - (height - 1);
+    z = stride*src_y;
+    yuv_src[0] = src[0] + z + src_x;
+    yuv_src[1] = src[1] + (z >> 2) + (src_x >> 1);
+    yuv_src[2] = src[2] + (yuv_src[1] - src[1]);
    do
    {
-        dst_row = dst;
+        lcd_write_yuv420_lines(dst, chroma_buf, yuv_src, width,
-        count = width;
+                               stride);
-        ysrc = src[0] + stride * src_y + src_x;
+        yuv_src[0] += stride << 1; /* Skip down two luma lines */
+        yuv_src[1] += stride >> 1; /* Skip down one chroma line */
-        /* upsampling, YUV->RGB conversion and reduction to RGB565 in one go */
+        yuv_src[2] += stride >> 1;
-        usrc = src[1] + (stride/CSUB_X) * (src_y/CSUB_Y)
+        dst -= 2;
-                                           + (src_x/CSUB_X);
+    }
-        vsrc = src[2] + (stride/CSUB_X) * (src_y/CSUB_Y)
+    while (--height > 0);
-                                           + (src_x/CSUB_X);
-        xphase = src_x % CSUB_X;
-        u = *usrc++ - 128;
-        v = *vsrc++ - 128;
-        rc = RVFAC * v + ROUNDOFFS;
-        gc = GVFAC * v + GUFAC * u + ROUNDOFFS;
-        bc = BUFAC * u + ROUNDOFFS;
-        do
-        {
-            y = *ysrc++;
-            red   = RYFAC * y + rc;
-            green = GYFAC * y + gc;
-            blue  = BYFAC * y + bc;
-            if ((unsigned)red > (RYFAC*255+ROUNDOFFS))
-            {
-                if (red < 0)
-                    red = 0;
-                else
-                    red = (RYFAC*255+ROUNDOFFS);
-            }
-            if ((unsigned)green > (GYFAC*255+ROUNDOFFS))
-            {
-                if (green < 0)
-                    green = 0;
-                else
-                    green = (GYFAC*255+ROUNDOFFS);
-            }
-            if ((unsigned)blue > (BYFAC*255+ROUNDOFFS))
-            {
-                if (blue < 0)
-                    blue = 0;
-                else
-                    blue = (BYFAC*255+ROUNDOFFS);
-            }
-            rbits = ((unsigned)red) >> 16 ;
-            gbits = ((unsigned)green) >> 16 ;
-            bbits = ((unsigned)blue) >> 16 ;
-            *dst_row = (rbits << 11) | (gbits << 5) | bbits;
-            /* next pixel - since rotated, add WIDTH */
-            dst_row += LCD_WIDTH;
-            if (++xphase >= CSUB_X)
-            {
-                u = *usrc++ - 128;
-                v = *vsrc++ - 128;
-                rc = RVFAC * v + ROUNDOFFS;
-                gc = GVFAC * v + GUFAC * u + ROUNDOFFS;
-                bc = BUFAC * u + ROUNDOFFS;
-                xphase = 0;
-            }
-        }
-        while (--count);
-        if (dst == dst_last) break;
-        dst--;
-        src_y++;
-    } while( 1);
 }
author	Dave Chapman <dave@dchapman.com>	2007-04-06 09:34:13 +0000
committer	Dave Chapman <dave@dchapman.com>	2007-04-06 09:34:13 +0000
commit	4303ab02a3efbd6a77e809641617c53082f9366f (patch)
tree	73b2bde0cb847785981b69dc749d7223d2ae7a12
parent	31d8068b331af435c50cacfbb0ed2de1c2221f4d (diff)
download	rockbox-4303ab02a3efbd6a77e809641617c53082f9366f.tar.gz rockbox-4303ab02a3efbd6a77e809641617c53082f9366f.zip

diff --git a/firmware/SOURCES b/firmware/SOURCES index 8b1d3cf896..9f90919817 100644 --- a/firmware/SOURCES +++ b/firmware/SOURCES
@@ -353,6 +353,7 @@ target/sh/archos/ondio/button-ondio.c
353	#ifndef SIMULATOR	353	#ifndef SIMULATOR
354	target/arm/sandisk/sansa-e200/ata-e200.c	354	target/arm/sandisk/sansa-e200/ata-e200.c
355	target/arm/sandisk/sansa-e200/lcd-e200.c	355	target/arm/sandisk/sansa-e200/lcd-e200.c
		356	target/arm/sandisk/sansa-e200/lcd-as-e200.S
356	target/arm/sandisk/sansa-e200/adc-e200.c	357	target/arm/sandisk/sansa-e200/adc-e200.c
357	target/arm/sandisk/sansa-e200/backlight-e200.c	358	target/arm/sandisk/sansa-e200/backlight-e200.c
358	target/arm/usb-pp.c	359	target/arm/usb-pp.c


diff --git a/firmware/target/arm/sandisk/sansa-e200/lcd-as-e200.S b/firmware/target/arm/sandisk/sansa-e200/lcd-as-e200.S new file mode 100644 index 0000000000..9e130cf54b --- /dev/null +++ b/firmware/target/arm/sandisk/sansa-e200/lcd-as-e200.S
@@ -0,0 +1,222 @@
		1	/***************************************************************************
		2	* __________ __ ___.
		3	* Open \______ \ ____ ____ \| \| _\_ \|__ _______ ___
		4	* Source \| _// _ \_/ ___\\| \|/ /\| __ \ / _ \ \/ /
		5	* Jukebox \| \| ( <_> ) \___\| < \| \_\ ( <_> > < <
		6	* Firmware \|____\|_ /\____/ \___ >__\|_ \\|___ /\____/__/\_ \
		7	* \/ \/ \/ \/ \/
		8	* $Id$
		9	*
		10	* Copyright (C) 2007 by Michael Sevakis
		11	*
		12	* All files in this archive are subject to the GNU General Public License.
		13	* See the file COPYING in the source tree root for full license agreement.
		14	*
		15	* This software is distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY
		16	* KIND, either express or implied.
		17	*
		18	****************************************************************************/
		19
		20	#include "config.h"
		21	#include "cpu.h"
		22
		23	/****************************************************************************
		24	* void lcd_write_yuv_420_lines(fb_data *dst,
		25	* unsigned char chroma_buf[LCD_HEIGHT/2*3],
		26	unsigned char const * const src[3],
		27	* int width,
		28	* int stride);
		29	*
		30	* \|R\| \|1.000000 -0.000001 1.402000\| \|Y'\|
		31	* \|G\| = \|1.000000 -0.334136 -0.714136\| \|Pb\|
		32	* \|B\| \|1.000000 1.772000 0.000000\| \|Pr\|
		33	* Scaled, normalized, rounded and tweaked to yield RGB 565:
		34	* \|R\| \|74 0 101\| \|Y' - 16\| >> 9
		35	* \|G\| = \|74 -24 -51\| \|Cb - 128\| >> 8
		36	* \|B\| \|74 128 0\| \|Cr - 128\| >> 9
		37	*/
		38	.section .icode, "ax", %progbits
		39	.align 2
		40	.global lcd_write_yuv420_lines
		41	.type lcd_write_yuv420_lines, %function
		42	lcd_write_yuv420_lines:
		43	@ r0 = dst
		44	@ r1 = chroma_buf
		45	@ r2 = yuv_src
		46	@ r3 = width
		47	@ [sp] = stride
		48	stmdb sp!, { r4-r12, lr } @ save non-scratch
		49	stmdb sp!, { r0, r3 } @ save dst and width
		50	mov r14, #74 @ r14 = Y factor
		51	ldmia r2, { r4, r5, r6 } @ r4 = yuv_src[0] = Y'_p
		52	@ r5 = yuv_src[1] = Cb_p
		53	@ r6 = yuv_src[2] = Cr_p
		54	10: @ loop line 1 @
		55	ldrb r2, [r4], #1 @ r2 = *Y'_p++;
		56	ldrb r8, [r5], #1 @ r8 = *Cb_p++;
		57	ldrb r11, [r6], #1 @ r11 = *Cr_p++;
		58	@
		59	@ compute Y
		60	sub r2, r2, #16 @ r7 = Y = (Y' - 16)*74
		61	mul r7, r2, r14 @
		62	@
		63	sub r8, r8, #128 @ Cb -= 128
		64	sub r11, r11, #128 @ Cr -= 128
		65	@
		66	mvn r2, #24 @ compute guv
		67	mul r10, r2, r8 @ r10 = Cb*-24
		68	mvn r2, #51 @
		69	mla r10, r2, r11, r10 @ r10 = r10 + Cr*-51
		70	@
		71	mov r2, #101 @ compute rv
		72	mul r9, r11, r2 @ r9 = rv = Cr*101
		73	@
		74	@ store chromas in line buffer
		75	add r8, r8, #2 @ bu = (Cb + 2) >> 2
		76	mov r8, r8, asr #2 @
		77	strb r8, [r1], #1 @
		78	add r9, r9, #256 @ rv = (Cr + 256) >> 9
		79	mov r9, r9, asr #9 @
		80	strb r9, [r1], #1 @
		81	mov r10, r10, asr #8 @ guv >>= 8
		82	strb r10, [r1], #1 @
		83	@ compute R, G, and B
		84	add r2, r8, r7, asr #9 @ r2 = b = (Y >> 9) + bu
		85	add r11, r9, r7, asr #9 @ r11 = r = (Y >> 9) + rv
		86	add r7, r10, r7, asr #8 @ r7 = g = (Y >> 8) + guv
		87	@
		88	orr r12, r2, r11 @ check if clamping is needed...
		89	orr r12, r12, r7, asr #1 @ ...at all
		90	cmp r12, #31 @
		91	bls 15f @ no clamp @
		92	mov r12, #31 @
		93	cmp r12, r2 @ clamp b
		94	andlo r2, r12, r2, asr #31 @
		95	eorlo r2, r2, r12 @
		96	cmp r12, r11 @ clamp r
		97	andlo r11, r12, r11, asr #31 @
		98	eorlo r11, r11, r12 @
		99	cmp r12, r7, asr #1 @ clamp g
		100	andlo r7, r12, r7, asr #31 @
		101	eorlo r7, r7, r12 @
		102	orrlo r7, r7, r7, asl #1 @
		103	15: @ no clamp @
		104	@
		105	orr r12, r2, r7, lsl #5 @ r4 \|= (g << 5)
		106	ldrb r2, [r4], #1 @ r2 = Y' = *Y'_p++
		107	orr r12, r12, r11, lsl #11 @ r4 = b \| (r << 11)
		108	strh r12, [r0], #LCD_WIDTH @ store pixel
		109	@
		110	sub r2, r2, #16 @ r7 = Y = (Y' - 16)*74
		111	mul r7, r2, r14 @ next Y
		112	@ compute R, G, and B
		113	add r2, r8, r7, asr #9 @ r2 = b = (Y >> 9) + bu
		114	add r11, r9, r7, asr #9 @ r11 = r = (Y >> 9) + rv
		115	add r7, r10, r7, asr #8 @ r7 = g = (Y >> 8) + guv
		116	@
		117	orr r12, r2, r11 @ check if clamping is needed...
		118	orr r12, r12, r7, asr #1 @ ...at all
		119	cmp r12, #31 @
		120	bls 15f @ no clamp @
		121	mov r12, #31 @
		122	cmp r12, r2 @ clamp b
		123	andlo r2, r12, r2, asr #31 @
		124	eorlo r2, r2, r12 @
		125	cmp r12, r11 @ clamp r
		126	andlo r11, r12, r11, asr #31 @
		127	eorlo r11, r11, r12 @
		128	cmp r12, r7, asr #1 @ clamp g
		129	andlo r7, r12, r7, asr #31 @
		130	eorlo r7, r7, r12 @
		131	orrlo r7, r7, r7, asl #1 @
		132	15: @ no clamp @
		133	@
		134	orr r12, r2, r11, lsl #11 @ r4 = b \| (r << 11)
		135	orr r12, r12, r7, lsl #5 @ r4 \|= (g << 5)
		136	strh r12, [r0, #LCD_WIDTH]! @ store pixel
		137	add r0, r0, #2*LCD_WIDTH @
		138	@
		139	subs r3, r3, #2 @
		140	bgt 10b @ loop line 1 @
		141	@ do second line
		142	@
		143	ldmia sp!, { r0, r3 } @ pop dst and width
		144	sub r0, r0, #2 @ set dst to start of next line
		145	sub r1, r1, r3, asl #1 @ rewind chroma pointer...
		146	ldr r2, [sp, #40] @ r2 = stride
		147	add r1, r1, r3, asr #1 @ ... (r1 -= width/2*3)
		148	@ move sources to start of next line
		149	sub r2, r2, r3 @ r2 = skip = stride - width
		150	add r4, r4, r2 @ r4 = Y'_p + skip
		151	@
		152	20: @ loop line 2 @
		153	ldrb r2, [r4], #1 @ r7 = Y' = *Y'_p++
		154	ldrsb r8, [r1], #1 @ reload saved chromas
		155	ldrsb r9, [r1], #1 @
		156	ldrsb r10, [r1], #1 @
		157	@
		158	sub r2, r2, #16 @ r2 = Y = (Y' - 16)*74
		159	mul r7, r2, r14 @
		160	@ compute R, G, and B
		161	add r2, r8, r7, asr #9 @ r2 = b = (Y >> 9) + bu
		162	add r11, r9, r7, asr #9 @ r11 = r = (Y >> 9) + rv
		163	add r7, r10, r7, asr #8 @ r7 = g = (Y >> 8) + guv
		164	@
		165	orr r12, r2, r11 @ check if clamping is needed...
		166	orr r12, r12, r7, asr #1 @ ...at all
		167	cmp r12, #31 @
		168	bls 25f @ no clamp @
		169	mov r12, #31 @
		170	cmp r12, r2 @ clamp b
		171	andlo r2, r12, r2, asr #31 @
		172	eorlo r2, r2, r12 @
		173	cmp r12, r11 @ clamp r
		174	andlo r11, r12, r11, asr #31 @
		175	eorlo r11, r11, r12 @
		176	cmp r12, r7, asr #1 @ clamp g
		177	andlo r7, r12, r7, asr #31 @
		178	eorlo r7, r7, r12 @
		179	orrlo r7, r7, r7, asl #1 @
		180	25: @ no clamp @
		181	@
		182	orr r12, r2, r11, lsl #11 @ r4 = b \| (r << 11)
		183	ldrb r2, [r4], #1 @ r2 = Y' = *Y'_p++
		184	orr r12, r12, r7, lsl #5 @ r4 \|= (g << 5)
		185	strh r12, [r0], #LCD_WIDTH @ store pixel
		186	@
		187	@ do second pixel
		188	@
		189	sub r2, r2, #16 @ r2 = Y = (Y' - 16)*74
		190	mul r7, r2, r14 @
		191	@ compute R, G, and B
		192	add r2, r8, r7, asr #9 @ r2 = b = (Y >> 9) + bu
		193	add r11, r9, r7, asr #9 @ r11 = r = (Y >> 9) + rv
		194	add r7, r10, r7, asr #8 @ r7 = g = (Y >> 8) + guv
		195	@
		196	orr r12, r2, r11 @ check if clamping is needed...
		197	orr r12, r12, r7, asr #1 @ ...at all
		198	cmp r12, #31 @
		199	bls 25f @ no clamp @
		200	mov r12, #31 @
		201	cmp r12, r2 @ clamp b
		202	andlo r2, r12, r2, asr #31 @
		203	eorlo r2, r2, r12 @
		204	cmp r12, r11 @ clamp r
		205	andlo r11, r12, r11, asr #31 @
		206	eorlo r11, r11, r12 @
		207	cmp r12, r7, asr #1 @ clamp g
		208	andlo r7, r12, r7, asr #31 @
		209	eorlo r7, r7, r12 @
		210	orrlo r7, r7, r7, asl #1 @
		211	25: @ no clamp @
		212	@
		213	orr r12, r2, r11, lsl #11 @ r4 = b \| (r << 11)
		214	orr r12, r12, r7, lsl #5 @ r4 \|= (g << 5)
		215	strh r12, [r0, #LCD_WIDTH]! @ store pixel
		216	add r0, r0, #2*LCD_WIDTH @
		217	@
		218	subs r3, r3, #2 @
		219	bgt 20b @ loop line 2 @
		220	@
		221	ldmia sp!, { r4-r12, pc } @ restore registers and return
		222	.size lcd_write_yuv420_lines, .-lcd_write_yuv420_lines


diff --git a/firmware/target/arm/sandisk/sansa-e200/lcd-e200.c b/firmware/target/arm/sandisk/sansa-e200/lcd-e200.c index 6b51531056..3e8246b367 100644 --- a/firmware/target/arm/sandisk/sansa-e200/lcd-e200.c +++ b/firmware/target/arm/sandisk/sansa-e200/lcd-e200.c
@@ -329,115 +329,45 @@ void lcd_blit(const fb_data* data, int x, int by, int width,
329	(void)stride;	329	(void)stride;
330	}	330	}
331		331
332	#define CSUB_X 2	332	/* Line write helper function for lcd_yuv_blit. Write two lines of yuv420. */
333	#define CSUB_Y 2	333	extern void lcd_write_yuv420_lines(fb_data *dst,
334		334	unsigned char chroma_buf[LCD_HEIGHT/2*3],
335	#define RYFAC (31*257)	335	unsigned char const * const src[3],
336	#define GYFAC (63*257)	336	int width,
337	#define BYFAC (31*257)	337	int stride);
338	#define RVFAC 11170 /* 31 * 257 * 1.402 */	338	/* Performance function to blit a YUV bitmap directly to the LCD */
339	#define GVFAC (-11563) /* 63 * 257 * -0.714136 */	339	/* For the e200 - show it rotated */
340	#define GUFAC (-5572) /* 63 * 257 * -0.344136 */	340	/* So the LCD_WIDTH is now the height */
341	#define BUFAC 14118 /* 31 * 257 * 1.772 */
342
343	#define ROUNDOFFS (127*257)
344
345	/* Performance function to blit a YUV bitmap directly to the LCD
346	Actually this code is from gigabeat, because this target is also
347	writing direct to a buffer. */
348	void lcd_yuv_blit(unsigned char * const src[3],	341	void lcd_yuv_blit(unsigned char * const src[3],
349	int src_x, int src_y, int stride,	342	int src_x, int src_y, int stride,
350	int _x, int _y, int width, int height)	343	int x, int y, int width, int height)
351	{	344	{
352	const unsigned char *usrc;	345	/* Caches for chroma data so it only need be recaculated every other
353	const unsigned char *vsrc;	346	line */
354	const unsigned char *ysrc;	347	unsigned char chroma_buf[LCD_HEIGHT/23]; / 480 bytes */
355	int xphase;	348	unsigned char const * yuv_src[3];
356	int rc, gc, bc;	349	off_t z;
357	int y, u, v;	350
358	int red, green, blue;	351	/* Sorry, but width and height must be >= 2 or else */
359	unsigned rbits, gbits, bbits;	352	width &= ~1;
360	int count;	353	height >>= 1;
361	fb_data *dst_row;	354
362
363
364	width = (width + 1) & ~1;
365	fb_data dst = (fb_data)lcd_driver_framebuffer +	355	fb_data dst = (fb_data)lcd_driver_framebuffer +
366	_x * LCD_WIDTH + (LCD_WIDTH - _y) - 1;	356	x * LCD_WIDTH + (LCD_WIDTH - y) - 1;
367	fb_data *dst_last = dst - (height - 1);	357
		358	z = stride*src_y;
		359	yuv_src[0] = src[0] + z + src_x;
		360	yuv_src[1] = src[1] + (z >> 2) + (src_x >> 1);
		361	yuv_src[2] = src[2] + (yuv_src[1] - src[1]);
368		362
369	do	363	do
370	{	364	{
371	dst_row = dst;	365	lcd_write_yuv420_lines(dst, chroma_buf, yuv_src, width,
372	count = width;	366	stride);
373	ysrc = src[0] + stride * src_y + src_x;	367	yuv_src[0] += stride << 1; /* Skip down two luma lines */
374		368	yuv_src[1] += stride >> 1; /* Skip down one chroma line */
375	/* upsampling, YUV->RGB conversion and reduction to RGB565 in one go */	369	yuv_src[2] += stride >> 1;
376	usrc = src[1] + (stride/CSUB_X) * (src_y/CSUB_Y)	370	dst -= 2;
377	+ (src_x/CSUB_X);	371	}
378	vsrc = src[2] + (stride/CSUB_X) * (src_y/CSUB_Y)	372	while (--height > 0);
379	+ (src_x/CSUB_X);
380	xphase = src_x % CSUB_X;
381
382	u = *usrc++ - 128;
383	v = *vsrc++ - 128;
384	rc = RVFAC * v + ROUNDOFFS;
385	gc = GVFAC * v + GUFAC * u + ROUNDOFFS;
386	bc = BUFAC * u + ROUNDOFFS;
387
388	do
389	{
390	y = *ysrc++;
391	red = RYFAC * y + rc;
392	green = GYFAC * y + gc;
393	blue = BYFAC * y + bc;
394
395	if ((unsigned)red > (RYFAC*255+ROUNDOFFS))
396	{
397	if (red < 0)
398	red = 0;
399	else
400	red = (RYFAC*255+ROUNDOFFS);
401	}
402	if ((unsigned)green > (GYFAC*255+ROUNDOFFS))
403	{
404	if (green < 0)
405	green = 0;
406	else
407	green = (GYFAC*255+ROUNDOFFS);
408	}
409	if ((unsigned)blue > (BYFAC*255+ROUNDOFFS))
410	{
411	if (blue < 0)
412	blue = 0;
413	else
414	blue = (BYFAC*255+ROUNDOFFS);
415	}
416	rbits = ((unsigned)red) >> 16 ;
417	gbits = ((unsigned)green) >> 16 ;
418	bbits = ((unsigned)blue) >> 16 ;
419
420	*dst_row = (rbits << 11) \| (gbits << 5) \| bbits;
421
422	/* next pixel - since rotated, add WIDTH */
423	dst_row += LCD_WIDTH;
424
425	if (++xphase >= CSUB_X)
426	{
427	u = *usrc++ - 128;
428	v = *vsrc++ - 128;
429	rc = RVFAC * v + ROUNDOFFS;
430	gc = GVFAC * v + GUFAC * u + ROUNDOFFS;
431	bc = BUFAC * u + ROUNDOFFS;
432	xphase = 0;
433	}
434	}
435	while (--count);
436
437	if (dst == dst_last) break;
438
439	dst--;
440	src_y++;
441	} while( 1);
442	}	373	}
443