Framebuffer work is coming back.

Framebuffer work is coming back.

[James Simmons]

Hi folks!!

  Sorry I have been quit for the last few weeks. I just finished moving 
closer to work. I had a two hour commute each way from work for the last 6 
months. Now I live 30 minutes from work. So now I will have more free time 
to work on the framebuffer layer. I have alot of email to cover so be 
patient with me. 




-------------------------------------------------------
This SF.net email is sponsored by: IBM Linux Tutorials.
Become an expert in LINUX or just sharpen your skills.  Sign up for IBM's
Free Linux Tutorials.  Learn everything from the bash shell to sys admin.
Click now! http://ads.osdn.com/?ad_id=1278&alloc_id=3371&op=click

Re: Framebuffer work is coming back.

[Alexander Kern]

[-- Attachment #1: Type: text/plain, Size: 488 bytes --]

Am Montag, 5. Januar 2004 18:41 schrieb James Simmons:
> Hi folks!!
>
>   Sorry I have been quit for the last few weeks. I just finished moving
> closer to work. I had a two hour commute each way from work for the last 6
> months. Now I live 30 minutes from work. So now I will have more free time
> to work on the framebuffer layer. I have alot of email to cover so be
> patient with me.
Great to hear,

here is a one more patch for MACH64. This is an accelerated imgblit.

Regards Alex

[-- Attachment #2: mach64_accel_img_blit.patch --]
[-- Type: text/x-diff, Size: 7217 bytes --]

diff -r -u -X /data/patches/exclude /usr/src/linux-2.6.orig/drivers/video/aty/mach64_accel.c /usr/src/linux-2.6.dev/drivers/video/aty/mach64_accel.c
--- /usr/src/linux-2.6.orig/drivers/video/aty/mach64_accel.c	2003-12-18 03:58:04.000000000 +0100
+++ /usr/src/linux-2.6.dev/drivers/video/aty/mach64_accel.c	2004-01-07 23:14:00.000000000 +0100
@@ -12,23 +12,30 @@
     /*
      *  Generic Mach64 routines
      */
+     
+/* this is for DMA GUI engine! to be continue */
+typedef struct {
+	u32 frame_buf_offset;
+	u32 system_mem_addr;
+	u32 command;
+	u32 reserved;
+} BM_DESCRIPTOR_ENTRY;
+
+#define LAST_DESCRIPTOR (1 << 31)
+#define SYSTEM_TO_FRAME_BUFFER 0
 
 void aty_reset_engine(const struct atyfb_par *par)
 {
 	/* reset engine */
 	aty_st_le32(GEN_TEST_CNTL,
-		    aty_ld_le32(GEN_TEST_CNTL, par) & ~GUI_ENGINE_ENABLE,
-		    par);
+		aty_ld_le32(GEN_TEST_CNTL, par) & ~GUI_ENGINE_ENABLE, par);
 	/* enable engine */
 	aty_st_le32(GEN_TEST_CNTL,
-		    aty_ld_le32(GEN_TEST_CNTL, par) | GUI_ENGINE_ENABLE,
-		    par);
+		aty_ld_le32(GEN_TEST_CNTL, par) | GUI_ENGINE_ENABLE, par);
 	/* ensure engine is not locked up by clearing any FIFO or */
 	/* HOST errors */
 	aty_st_le32(BUS_CNTL,
-		    aty_ld_le32(BUS_CNTL,
-				par) | BUS_HOST_ERR_ACK | BUS_FIFO_ERR_ACK,
-		    par);
+		aty_ld_le32(BUS_CNTL, par) | BUS_HOST_ERR_ACK | BUS_FIFO_ERR_ACK, par);
 }
 
 static void reset_GTC_3D_engine(const struct atyfb_par *par)
@@ -51,7 +58,7 @@
 	if (info->var.bits_per_pixel == 24) {
 		/* In 24 bpp, the engine is in 8 bpp - this requires that all */
 		/* horizontal coordinates and widths must be adjusted */
-		pitch_value = pitch_value * 3;
+		pitch_value *= 3;
 	}
 
 	/* On GTC (RagePro), we need to reset the 3D engine before */
@@ -146,7 +153,7 @@
 	aty_st_le32(DP_CHAIN_MASK, par->crtc.dp_chain_mask, par);
 
 	wait_for_fifo(5, par);
-	aty_st_le32(SCALE_3D_CNTL, 0, par);
+ 	aty_st_le32(SCALE_3D_CNTL, 0, par);
 	aty_st_le32(Z_CNTL, 0, par);
 	aty_st_le32(CRTC_INT_CNTL, aty_ld_le32(CRTC_INT_CNTL, par) & ~0x20,
 		    par);
@@ -174,8 +181,7 @@
 {
 	struct atyfb_par *par = (struct atyfb_par *) info->par;
 	u32 dy = area->dy, sy = area->sy, direction = DST_LAST_PEL;
-	u32 sx = area->sx, dx = area->dx, width = area->width;	
-	u32 pitch_value;
+	u32 sx = area->sx, dx = area->dx, width = area->width;
 
 	if (!area->width || !area->height)
 		return;
@@ -186,11 +192,9 @@
 		return;
 	}
 
-	pitch_value = info->var.xres_virtual;
 	if (info->var.bits_per_pixel == 24) {
 		/* In 24 bpp, the engine is in 8 bpp - this requires that all */
 		/* horizontal coordinates and widths must be adjusted */
-		pitch_value *= 3;
 		sx *= 3;
 		dx *= 3;
 		width *= 3;
@@ -254,8 +258,104 @@
 void atyfb_imageblit(struct fb_info *info, const struct fb_image *image)
 {
 	struct atyfb_par *par = (struct atyfb_par *) info->par;
-    
-	if (par->blitter_may_be_busy)
-		wait_for_idle(par);
-	cfb_imageblit(info, image);
+	u32 src_writes, src_width, *pbitmap, tmp, pix_width_save;
+	u32 dx = image->dx, dy = image->dy, width = image->width;
+	u32 src_pix_width = 0, src_pix_mask;
+
+	if (!image->width || !image->height)
+		return;
+	if (!par->accel_flags || 
+	    (image->depth != 1 && info->var.bits_per_pixel != image->depth)) {
+		if (par->blitter_may_be_busy)
+			wait_for_idle(par);
+
+		cfb_imageblit(info, image);
+		return;
+	}
+
+	if (info->var.bits_per_pixel == 24) {
+		/* In 24 bpp, the engine is in 8 bpp - this requires that all */
+		/* horizontal coordinates and widths must be adjusted */
+		dx *= 3;
+		width *= 3;
+	}
+
+	src_pix_mask = DST_MASK;
+	switch (image->depth) {
+	case 1:
+	    src_pix_width = BYTE_ORDER_MSB_TO_LSB | HOST_1BPP | SRC_1BPP;
+	    break;
+	case 4:
+	    src_pix_width = BYTE_ORDER_MSB_TO_LSB | HOST_4BPP | SRC_4BPP;
+	    break;
+	case 8:
+	    src_pix_width = HOST_8BPP | SRC_8BPP;
+	    src_pix_mask |= BYTE_ORDER_MASK;
+	    break;
+	case 15:
+	    src_pix_width = HOST_15BPP | SRC_15BPP;
+	    src_pix_mask |= BYTE_ORDER_MASK;
+	    break;
+	case 16:
+	    src_pix_width = HOST_16BPP | SRC_16BPP;
+	    src_pix_mask |= BYTE_ORDER_MASK;
+	    break;
+	case 24:
+	    src_pix_width = HOST_24BPP | SRC_24BPP;
+	    src_pix_mask |= BYTE_ORDER_MASK;
+	    break;
+	case 32:
+	    src_pix_width = HOST_32BPP | SRC_32BPP;
+	    src_pix_mask |= BYTE_ORDER_MASK;
+	    break;
+	}
+	src_width = image->width * image->height * image->depth;
+	src_writes = src_width / 32;
+	if(src_width != (src_writes * 32))
+	    src_writes++;
+
+	pix_width_save = aty_ld_le32(DP_PIX_WIDTH, par);
+	tmp = pix_width_save & src_pix_mask;
+	aty_st_le32(DP_PIX_WIDTH, src_pix_width | tmp, par);
+
+	if(image->depth == 1) {
+		u32 fg, bg;
+		if (info->fix.visual == FB_VISUAL_TRUECOLOR ||
+		    info->fix.visual == FB_VISUAL_DIRECTCOLOR) {
+			fg = ((u32*)(info->pseudo_palette))[image->fg_color];
+			bg = ((u32*)(info->pseudo_palette))[image->bg_color];
+		} else {
+			fg = image->fg_color;
+			bg = image->bg_color;
+		}
+
+		wait_for_fifo(4, par);
+		aty_st_le32(DP_SRC, MONO_SRC_HOST | FRGD_SRC_FRGD_CLR | BKGD_SRC_BKGD_CLR, par);
+		aty_st_le32(DP_MIX, FRGD_MIX_S | BKGD_MIX_S, par);
+		aty_st_le32(DP_BKGD_CLR, bg, par);
+		aty_st_le32(DP_FRGD_CLR, fg, par);
+	} else {
+		wait_for_fifo(2, par);
+		aty_st_le32(DP_SRC, MONO_SRC_ONE | FRGD_SRC_HOST, par);
+		aty_st_le32(DP_MIX, FRGD_MIX_D_XOR_S | BKGD_MIX_D, par);
+	}
+	
+	wait_for_fifo(5, par);
+	aty_st_le32(DST_CNTL, DST_Y_TOP_TO_BOTTOM | DST_X_LEFT_TO_RIGHT, par);
+	aty_st_le32(DST_X, dx, par);
+	aty_st_le32(DST_Y, dy, par);
+	aty_st_le32(DST_HEIGHT, image->height, par);
+	aty_st_le32(DST_WIDTH, width, par);
+
+	/* copy host data */
+	pbitmap = (u32*)(image->data);
+	for(; src_writes; src_writes--) {
+	    wait_for_fifo(1, par);
+	    aty_st_le32(HOST_DATA0, *pbitmap, par);
+	    pbitmap++;
+	}
+
+	wait_for_idle(par);
+	/* restore pix_width */
+	aty_st_le32(DP_PIX_WIDTH, pix_width_save, par);
 }
diff -r -u -X /data/patches/exclude /usr/src/linux-2.6.orig/include/video/mach64.h /usr/src/linux-2.6.dev/include/video/mach64.h
--- /usr/src/linux-2.6.orig/include/video/mach64.h	2004-01-07 14:41:43.000000000 +0100
+++ /usr/src/linux-2.6.dev/include/video/mach64.h	2004-01-07 19:30:33.000000000 +0100
@@ -983,13 +983,14 @@
 #define DP_CHAIN_32BPP		0x8080
 
 /* DP_PIX_WIDTH register constants */
-#define DST_1BPP		0
-#define DST_4BPP		1
-#define DST_8BPP		2
-#define DST_15BPP		3
-#define DST_16BPP		4
-#define DST_24BPP		5
-#define DST_32BPP		6
+#define DST_1BPP		0x0
+#define DST_4BPP		0x1
+#define DST_8BPP		0x2
+#define DST_15BPP		0x3
+#define DST_16BPP		0x4
+#define DST_24BPP		0x5
+#define DST_32BPP		0x6
+#define DST_MASK		0xF
 #define SRC_1BPP		0x000
 #define SRC_4BPP		0x100
 #define SRC_8BPP		0x200
@@ -997,6 +998,7 @@
 #define SRC_16BPP		0x400
 #define SRC_24BPP		0x500
 #define SRC_32BPP		0x600
+#define SRC_MASK		0xF00
 #define HOST_1BPP		0x00000
 #define HOST_4BPP		0x10000
 #define HOST_8BPP		0x20000
@@ -1004,8 +1006,10 @@
 #define HOST_16BPP		0x40000
 #define HOST_24BPP		0x50000
 #define HOST_32BPP		0x60000
+#define HOST_MASK		0xF0000
 #define BYTE_ORDER_MSB_TO_LSB	0
 #define BYTE_ORDER_LSB_TO_MSB	0x1000000
+#define BYTE_ORDER_MASK		0x1000000
 
 /* DP_MIX register constants */
 #define BKGD_MIX_NOT_D			0

Re: Framebuffer work is coming back.

[James Simmons]

 
> here is a one more patch for MACH64. This is an accelerated imgblit.

Nice !!! ;-) Only one problem I found when testing. If you pick the sun 
12x22 fonts at boot up some of the fonts are distorted. Once X starts 
and you go back to the console this problem goes away completely. This 
is using a non accelerated X server. 

P.S
Also I noticed that in the accelerated function you have HOST_XBPP. I like 
to give it a try for color expension. This would be cool sine this means 
we could have the 4 bpp penguin shown on any depth display!!!




-------------------------------------------------------
This SF.net email is sponsored by: Perforce Software.
Perforce is the Fast Software Configuration Management System offering
advanced branching capabilities and atomic changes on 50+ platforms.
Free Eval! http://www.perforce.com/perforce/loadprog.html

Re: Framebuffer work is coming back.

[Alexander Kern]

Am Mittwoch, 7. Januar 2004 23:58 schrieben Sie:
> > here is a one more patch for MACH64. This is an accelerated imgblit.
>
> Nice !!! ;-) Only one problem I found when testing. If you pick the sun
> 12x22 fonts at boot up some of the fonts are distorted. Once X starts
> and you go back to the console this problem goes away completely. This
> is using a non accelerated X server.
>
> P.S
> Also I noticed that in the accelerated function you have HOST_XBPP. I like
> to give it a try for color expension. This would be cool sine this means
> we could have the 4 bpp penguin shown on any depth display!!!
I trying to investigate this problem, but found nothing. I suspect the 
bytealigning of images, are they all bytealigned, 12x22 too?

Cheers Alex

P.S. 4bpp penguin on any bpp will not work! Mach64 supports only monoblit
extension, colorblit only if source and destination have the same depth. 



-------------------------------------------------------
This SF.net email is sponsored by: Perforce Software.
Perforce is the Fast Software Configuration Management System offering
advanced branching capabilities and atomic changes on 50+ platforms.
Free Eval! http://www.perforce.com/perforce/loadprog.html