* [PATCH] ATI Mach64 accelerated imgblit (sligthly improved)
@ 2004-01-17 0:20 Alexander Kern
2004-01-17 18:25 ` Geert Uytterhoeven
0 siblings, 1 reply; 6+ messages in thread
From: Alexander Kern @ 2004-01-17 0:20 UTC (permalink / raw)
To: James Simmons; +Cc: fbdev
[-- Attachment #1: Type: text/plain, Size: 186 bytes --]
Hi,
here is a patch. Now it works with SUN12x22 font and 8, 16 and 32 bpp. I have
the strange issues with 24bpp. Should we simply drop acceleration for this
resolution?
Regards Alex
[-- Attachment #2: accel_imgblit.patch --]
[-- Type: text/x-diff, Size: 10799 bytes --]
diff -r -u -X /data/patches/exclude /usr/src/linux-2.6.orig/drivers/video/aty/atyfb_base.c /usr/src/linux-2.6.dev/drivers/video/aty/atyfb_base.c
--- /usr/src/linux-2.6.orig/drivers/video/aty/atyfb_base.c 2004-01-17 00:48:16.000000000 +0100
+++ /usr/src/linux-2.6.dev/drivers/video/aty/atyfb_base.c 2004-01-13 17:38:03.000000000 +0100
@@ -1134,11 +1130,11 @@
v_total = v_sync_end + debug.upper_margin;
hSync = 1000000000 / (pixclock_in_ps * h_total);
- vRefresh = (hSync * 1000.0) / v_total;
+ vRefresh = (hSync * 1000) / v_total;
if (par->crtc.gen_cntl & CRTC_INTERLACE_EN)
- vRefresh *= 2.0;
+ vRefresh *= 2;
if (par->crtc.gen_cntl & CRTC_DBL_SCAN_EN)
- vRefresh /= 2.0;
+ vRefresh /= 2;
printk("atyfb: atyfb_set_par\n");
printk(" Set Visible Mode to %ix%i-%i\n", var->xres, var->yres, var->bits_per_pixel);
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 2004-01-09 07:59:04.000000000 +0100
+++ /usr/src/linux-2.6.dev/drivers/video/aty/mach64_accel.c 2004-01-17 00:23:23.000000000 +0100
@@ -12,23 +12,42 @@
/*
* Generic Mach64 routines
*/
+
+/* this is for DMA GUI engine! work 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
+
+static u32 rotation24bpp(u32 dx, u32 direction)
+{
+ u32 rotation;
+ if (direction & DST_X_LEFT_TO_RIGHT) {
+ rotation = (dx / 4) % 6;
+ } else {
+ rotation = ((dx + 2) / 4) % 6;
+ }
+
+ return ((rotation << 8) | DST_24_ROTATION_ENABLE);
+}
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 +70,7 @@
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 +165,7 @@
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 +193,7 @@
{
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, rotation = 0;
if (!area->width || !area->height)
return;
@@ -186,11 +204,9 @@
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;
@@ -208,18 +224,22 @@
} else
direction |= DST_X_LEFT_TO_RIGHT;
+ if (info->var.bits_per_pixel == 24) {
+ rotation = rotation24bpp(dx, direction);
+ }
+
wait_for_fifo(4, par);
aty_st_le32(DP_SRC, FRGD_SRC_BLIT, par);
aty_st_le32(SRC_Y_X, (sx << 16) | sy, par);
aty_st_le32(SRC_HEIGHT1_WIDTH1, (width << 16) | area->height, par);
- aty_st_le32(DST_CNTL, direction, par);
+ aty_st_le32(DST_CNTL, direction | rotation, par);
draw_rect(dx, dy, width, area->height, par);
}
void atyfb_fillrect(struct fb_info *info, const struct fb_fillrect *rect)
{
struct atyfb_par *par = (struct atyfb_par *) info->par;
- u32 color = rect->color, dx = rect->dx, width = rect->width;
+ u32 color = rect->color, dx = rect->dx, width = rect->width, rotation = 0;
if (!rect->width || !rect->height)
return;
@@ -238,6 +258,7 @@
/* horizontal coordinates and widths must be adjusted */
dx *= 3;
width *= 3;
+ rotation = rotation24bpp(dx, DST_X_LEFT_TO_RIGHT);
}
wait_for_fifo(3, par);
@@ -247,15 +268,160 @@
par);
aty_st_le32(DST_CNTL,
DST_LAST_PEL | DST_Y_TOP_TO_BOTTOM |
- DST_X_LEFT_TO_RIGHT, par);
+ DST_X_LEFT_TO_RIGHT | rotation, par);
draw_rect(dx, rect->dy, width, rect->height, par);
}
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 dwords, dx = image->dx, dy = image->dy, width = image->width;
+ u32 pix_width_save, pix_width, host_cntl, rotation = 0, src, mix;
+
+ 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;
+ }
+
+ wait_for_idle(par);
+ pix_width = pix_width_save = aty_ld_le32(DP_PIX_WIDTH, par);
+ host_cntl = aty_ld_le32(HOST_CNTL, par) | HOST_BYTE_ALIGN;
+
+ switch (image->depth) {
+ case 1:
+ pix_width &= ~(BYTE_ORDER_MASK | HOST_MASK);
+ pix_width |= (BYTE_ORDER_MSB_TO_LSB | HOST_1BPP);
+ break;
+ case 4:
+ pix_width &= ~(BYTE_ORDER_MASK | HOST_MASK);
+ pix_width |= (BYTE_ORDER_MSB_TO_LSB | HOST_4BPP);
+ break;
+ case 8:
+ pix_width &= ~HOST_MASK;
+ pix_width |= HOST_8BPP;
+ break;
+ case 15:
+ pix_width &= ~HOST_MASK;
+ pix_width |= HOST_15BPP;
+ break;
+ case 16:
+ pix_width &= ~HOST_MASK;
+ pix_width |= HOST_16BPP;
+ break;
+ case 24:
+ pix_width &= ~HOST_MASK;
+ pix_width |= HOST_24BPP;
+ break;
+ case 32:
+ pix_width &= ~HOST_MASK;
+ pix_width |= HOST_32BPP;
+ break;
+ }
+
+ 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;
+
+ rotation = rotation24bpp(dx, DST_X_LEFT_TO_RIGHT);
+
+ pix_width &= ~DST_MASK;
+ pix_width |= DST_8BPP;
+
+ /* ab IIC we have DP_HOST_TRIPLE_EN bit */
+ /* FIXME: 24bpp + SUN12x22 corruption
+ if (image->width % 8 == 0) */
+ pix_width |= DP_HOST_TRIPLE_EN;
+ }
+
+ dwords = (((((image->width * image->depth) + 7) / 8) * image->height * image->depth)+ 3) / 4 ;
+
+ 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(2, par);
+ aty_st_le32(DP_BKGD_CLR, bg, par);
+ aty_st_le32(DP_FRGD_CLR, fg, par);
+ src = MONO_SRC_HOST | FRGD_SRC_FRGD_CLR | BKGD_SRC_BKGD_CLR;
+ mix = FRGD_MIX_S | BKGD_MIX_S;
+ } else {
+ src = MONO_SRC_ONE | FRGD_SRC_HOST;
+ mix = FRGD_MIX_D_XOR_S | BKGD_MIX_D;
+ }
+
+ wait_for_fifo(6, par);
+ aty_st_le32(DP_WRITE_MASK, 0xFFFFFFFF, par);
+ aty_st_le32(DP_PIX_WIDTH, pix_width, par);
+ aty_st_le32(DP_MIX, mix, par);
+ aty_st_le32(DP_SRC, src, par);
+ aty_st_le32(HOST_CNTL, host_cntl, par);
+ aty_st_le32(DST_CNTL, DST_Y_TOP_TO_BOTTOM | DST_X_LEFT_TO_RIGHT | rotation, par);
+
+ draw_rect(dx, dy, width, image->height, par);
+
+ /* manual triple each pixel, freezes the box */
+ if (0 && info->var.bits_per_pixel == 24 && !(pix_width & DP_HOST_TRIPLE_EN)) {
+ int inbit, outbit, mult24, byte_id_in_dword, width;
+ u8 *pbitmapin, *pbitmapout;
+ u32 hostdword;
+
+ pbitmapin = (u8*)image->data;
+
+ for(width = image->width, inbit = 7, mult24 = 0; dwords; dwords--) {
+ for(hostdword = 0, pbitmapout = (u8*)&hostdword, byte_id_in_dword = 0;
+ byte_id_in_dword < 4;
+ byte_id_in_dword++, pbitmapout++) {
+ for(outbit = 7; outbit >= 0; outbit--) {
+ *pbitmapout |= (((*pbitmapin >> inbit) & 1) << outbit);
+ mult24++;
+ /* next bit */
+ if (mult24 == 3) {
+ mult24 = 0;
+ inbit--;
+ width--;
+ }
+
+ /* next byte */
+ if (inbit < 0 || width == 0) {
+ pbitmapin++;
+ inbit = 7;
+
+ if(width == 0) {
+ width = image->width;
+ outbit = 0;
+ }
+ }
+ }
+ }
+ wait_for_fifo(1, par);
+ aty_st_le32(HOST_DATA0, hostdword, par);
+ }
+ } else {
+ u32 *pbitmap;
+ for(pbitmap = (u32*)(image->data); dwords; dwords--, pbitmap++) {
+ wait_for_fifo(1, par);
+ aty_st_le32(HOST_DATA0, *pbitmap, par);
+ }
+ }
+
+ wait_for_idle(par);
+
+ /* restore pix_width */
+ wait_for_fifo(1, par);
+ 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 2004-01-17 00:48:16.000000000 +0100
+++ /usr/src/linux-2.6.dev/include/video/mach64.h 2004-01-15 00:27:29.000000000 +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 +998,8 @@
#define SRC_16BPP 0x400
#define SRC_24BPP 0x500
#define SRC_32BPP 0x600
+#define SRC_MASK 0xF00
+#define DP_HOST_TRIPLE_EN 0x2000
#define HOST_1BPP 0x00000
#define HOST_4BPP 0x10000
#define HOST_8BPP 0x20000
@@ -1004,8 +1007,10 @@
#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
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [PATCH] ATI Mach64 accelerated imgblit (sligthly improved)
2004-01-17 0:20 [PATCH] ATI Mach64 accelerated imgblit (sligthly improved) Alexander Kern
@ 2004-01-17 18:25 ` Geert Uytterhoeven
2004-01-17 19:35 ` Alexander Kern
0 siblings, 1 reply; 6+ messages in thread
From: Geert Uytterhoeven @ 2004-01-17 18:25 UTC (permalink / raw)
To: Alexander Kern; +Cc: James Simmons, fbdev
On Sat, 17 Jan 2004, Alexander Kern wrote:
> here is a patch. Now it works with SUN12x22 font and 8, 16 and 32 bpp. I have
> the strange issues with 24bpp. Should we simply drop acceleration for this
> resolution?
All acceleration, or just the imageblit? I don't think there's a problem with
rectangle fill/copy in 24 bpp, or is there?
Gr{oetje,eeting}s,
Geert
--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org
In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
-- Linus Torvalds
-------------------------------------------------------
The SF.Net email is sponsored by EclipseCon 2004
Premiere Conference on Open Tools Development and Integration
See the breadth of Eclipse activity. February 3-5 in Anaheim, CA.
http://www.eclipsecon.org/osdn
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [PATCH] ATI Mach64 accelerated imgblit (sligthly improved)
2004-01-17 18:25 ` Geert Uytterhoeven
@ 2004-01-17 19:35 ` Alexander Kern
2004-01-19 18:11 ` James Simmons
0 siblings, 1 reply; 6+ messages in thread
From: Alexander Kern @ 2004-01-17 19:35 UTC (permalink / raw)
To: Geert Uytterhoeven; +Cc: James Simmons, fbdev
Am Samstag, 17. Januar 2004 19:25 schrieb Geert Uytterhoeven:
> On Sat, 17 Jan 2004, Alexander Kern wrote:
> > here is a patch. Now it works with SUN12x22 font and 8, 16 and 32 bpp. I
> > have the strange issues with 24bpp. Should we simply drop acceleration
> > for this resolution?
>
> All acceleration, or just the imageblit? I don't think there's a problem
> with rectangle fill/copy in 24 bpp, or is there?
>
> Gr{oetje,eeting}s,
>
> Geert
No, it's only imageblit, which has issue.
Regards
Alex Kern
-------------------------------------------------------
The SF.Net email is sponsored by EclipseCon 2004
Premiere Conference on Open Tools Development and Integration
See the breadth of Eclipse activity. February 3-5 in Anaheim, CA.
http://www.eclipsecon.org/osdn
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [PATCH] ATI Mach64 accelerated imgblit (sligthly improved)
2004-01-17 19:35 ` Alexander Kern
@ 2004-01-19 18:11 ` James Simmons
2004-01-23 17:58 ` Alexander Kern
0 siblings, 1 reply; 6+ messages in thread
From: James Simmons @ 2004-01-19 18:11 UTC (permalink / raw)
To: Alexander Kern; +Cc: Geert Uytterhoeven, fbdev
> No, it's only imageblit, which has issue.
> Regards
> Alex Kern
Applied your patch. I tested it and it works wonderful. I haven't tested
in 24 bpp mode tho. I think that is a minor issue.
-------------------------------------------------------
The SF.Net email is sponsored by EclipseCon 2004
Premiere Conference on Open Tools Development and Integration
See the breadth of Eclipse activity. February 3-5 in Anaheim, CA.
http://www.eclipsecon.org/osdn
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [PATCH] ATI Mach64 accelerated imgblit (sligthly improved)
2004-01-19 18:11 ` James Simmons
@ 2004-01-23 17:58 ` Alexander Kern
2004-01-30 21:49 ` James Simmons
0 siblings, 1 reply; 6+ messages in thread
From: Alexander Kern @ 2004-01-23 17:58 UTC (permalink / raw)
To: James Simmons; +Cc: fbdev
[-- Attachment #1: Type: text/plain, Size: 526 bytes --]
Am Montag, 19. Januar 2004 19:11 schrieben Sie:
> > No, it's only imageblit, which has issue.
> > Regards
> > Alex Kern
>
> Applied your patch. I tested it and it works wonderful. I haven't tested
> in 24 bpp mode tho. I think that is a minor issue.
Hi James,
will you update your fbdev.diff.gz? What about resync and submit it again to
mm tree?
Here is one better patch for accelerated imageblit. It looks like mach64 has a
problems with buildin trippler and unaligned data. Manual tripple path works
now!
Regards Alex
[-- Attachment #2: 20040123-mach64-include.diff --]
[-- Type: text/x-diff, Size: 1337 bytes --]
diff -u /usr/src/linux-2.6.dev/include/video/mach64.h /usr/src/linux-2.6.dirty/include/video/mach64.h
--- /usr/src/linux-2.6.dev/include/video/mach64.h 2004-01-23 15:42:01.000000000 +0100
+++ /usr/src/linux-2.6.dirty/include/video/mach64.h 2004-01-15 00:27:29.000000000 +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 +998,8 @@
#define SRC_16BPP 0x400
#define SRC_24BPP 0x500
#define SRC_32BPP 0x600
+#define SRC_MASK 0xF00
+#define DP_HOST_TRIPLE_EN 0x2000
#define HOST_1BPP 0x00000
#define HOST_4BPP 0x10000
#define HOST_8BPP 0x20000
@@ -1004,8 +1007,10 @@
#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
[-- Attachment #3: 20040123-mach64.diff --]
[-- Type: text/x-diff, Size: 15164 bytes --]
diff -u /usr/src/linux-2.6.dev/drivers/video/aty/atyfb.h /usr/src/linux-2.6.dirty/drivers/video/aty/atyfb.h
--- /usr/src/linux-2.6.dev/drivers/video/aty/atyfb.h 2004-01-23 15:41:52.000000000 +0100
+++ /usr/src/linux-2.6.dirty/drivers/video/aty/atyfb.h 2004-01-23 14:54:39.000000000 +0100
@@ -169,6 +169,7 @@
#define M64F_LT_SLEEP 0x00040000
#define M64F_XL_DLL 0x00080000
#define M64F_MFB_FORCE_4 0x00100000
+#define M64F_NO_HW_TRIPLE 0x00200000
/*
* Register access
diff -u /usr/src/linux-2.6.dev/drivers/video/aty/atyfb_base.c /usr/src/linux-2.6.dirty/drivers/video/aty/atyfb_base.c
--- /usr/src/linux-2.6.dev/drivers/video/aty/atyfb_base.c 2004-01-23 15:41:52.000000000 +0100
+++ /usr/src/linux-2.6.dirty/drivers/video/aty/atyfb_base.c 2004-01-23 14:58:31.000000000 +0100
@@ -323,30 +323,29 @@
} aty_chips[] __initdata = {
#ifdef CONFIG_FB_ATY_GX
/* Mach64 GX */
- { 0x4758, 0x00d7, 0x00, 0x00, m64n_gx, 135, 50, 50, M64F_GX },
- { 0x4358, 0x0057, 0x00, 0x00, m64n_cx, 135, 50, 50, M64F_GX },
+ { 0x4758, 0x00d7, 0x00, 0x00, m64n_gx, 135, 50, 50, M64F_GX | M64F_NO_HW_TRIPLE },
+ { 0x4358, 0x0057, 0x00, 0x00, m64n_cx, 135, 50, 50, M64F_GX | M64F_NO_HW_TRIPLE },
#endif /* CONFIG_FB_ATY_GX */
#ifdef CONFIG_FB_ATY_CT
/* Mach64 CT */
- { 0x4354, 0x4354, 0x00, 0x00, m64n_ct, 135, 60, 60, M64F_CT | M64F_INTEGRATED | M64F_CT_BUS | M64F_MAGIC_FIFO },
- { 0x4554, 0x4554, 0x00, 0x00, m64n_et, 135, 60, 60, M64F_CT | M64F_INTEGRATED | M64F_CT_BUS | M64F_MAGIC_FIFO },
+ { 0x4354, 0x4354, 0x00, 0x00, m64n_ct, 135, 60, 60, M64F_CT | M64F_INTEGRATED | M64F_CT_BUS | M64F_MAGIC_FIFO | M64F_NO_HW_TRIPLE },
+ { 0x4554, 0x4554, 0x00, 0x00, m64n_et, 135, 60, 60, M64F_CT | M64F_INTEGRATED | M64F_CT_BUS | M64F_MAGIC_FIFO | M64F_NO_HW_TRIPLE },
/* Mach64 VT */
- { 0x5654, 0x5654, 0xc7, 0x00, m64n_vta3, 170, 67, 67, M64F_VT | M64F_INTEGRATED | M64F_VT_BUS | M64F_MAGIC_FIFO | M64F_FIFO_24 },
- { 0x5654, 0x5654, 0xc7, 0x40, m64n_vta4, 200, 67, 67, M64F_VT | M64F_INTEGRATED | M64F_VT_BUS | M64F_MAGIC_FIFO | M64F_FIFO_24 | M64F_MAGIC_POSTDIV },
- { 0x5654, 0x5654, 0x00, 0x00, m64n_vtb, 200, 67, 67, M64F_VT | M64F_INTEGRATED | M64F_VT_BUS | M64F_GTB_DSP | M64F_FIFO_24 },
- { 0x5655, 0x5655, 0x00, 0x00, m64n_vtb, 200, 67, 67, M64F_VT | M64F_INTEGRATED | M64F_VT_BUS | M64F_GTB_DSP | M64F_FIFO_24 | M64F_SDRAM_MAGIC_PLL },
- { 0x5656, 0x5656, 0x00, 0x00, m64n_vt4, 230, 83, 83, M64F_VT | M64F_INTEGRATED | M64F_GTB_DSP },
+ { 0x5654, 0x5654, 0xc7, 0x00, m64n_vta3, 170, 67, 67, M64F_VT | M64F_INTEGRATED | M64F_VT_BUS | M64F_MAGIC_FIFO | M64F_FIFO_24 | M64F_NO_HW_TRIPLE },
+ { 0x5654, 0x5654, 0xc7, 0x40, m64n_vta4, 200, 67, 67, M64F_VT | M64F_INTEGRATED | M64F_VT_BUS | M64F_MAGIC_FIFO | M64F_FIFO_24 | M64F_MAGIC_POSTDIV | M64F_NO_HW_TRIPLE },
+ { 0x5654, 0x5654, 0x00, 0x00, m64n_vtb, 200, 67, 67, M64F_VT | M64F_INTEGRATED | M64F_VT_BUS | M64F_GTB_DSP | M64F_FIFO_24 | M64F_NO_HW_TRIPLE },
+ { 0x5655, 0x5655, 0x00, 0x00, m64n_vtb, 200, 67, 67, M64F_VT | M64F_INTEGRATED | M64F_VT_BUS | M64F_GTB_DSP | M64F_FIFO_24 | M64F_SDRAM_MAGIC_PLL | M64F_NO_HW_TRIPLE },
+ { 0x5656, 0x5656, 0x00, 0x00, m64n_vt4, 230, 83, 83, M64F_VT | M64F_INTEGRATED | M64F_GTB_DSP | M64F_NO_HW_TRIPLE },
/* Mach64 GT (3D RAGE) */
- { 0x4754, 0x4754, 0x07, 0x00, m64n_gt, 135, 63, 63, M64F_GT | M64F_INTEGRATED | M64F_MAGIC_FIFO | M64F_FIFO_24 | M64F_EXTRA_BRIGHT },
- { 0x4754, 0x4754, 0x07, 0x01, m64n_gt, 170, 67, 67, M64F_GT | M64F_INTEGRATED | M64F_GTB_DSP | M64F_FIFO_24 | M64F_SDRAM_MAGIC_PLL | M64F_EXTRA_BRIGHT },
- { 0x4754, 0x4754, 0x07, 0x02, m64n_gt, 200, 67, 67, M64F_GT | M64F_INTEGRATED | M64F_GTB_DSP | M64F_FIFO_24 | M64F_SDRAM_MAGIC_PLL | M64F_EXTRA_BRIGHT },
- { 0x4755, 0x4755, 0x00, 0x00, m64n_gtb, 200, 67, 67, M64F_GT | M64F_INTEGRATED | M64F_GTB_DSP | M64F_FIFO_24 | M64F_SDRAM_MAGIC_PLL | M64F_EXTRA_BRIGHT },
+ { 0x4754, 0x4754, 0x07, 0x00, m64n_gt, 135, 63, 63, M64F_GT | M64F_INTEGRATED | M64F_MAGIC_FIFO | M64F_FIFO_24 | M64F_EXTRA_BRIGHT | M64F_NO_HW_TRIPLE },
+ { 0x4754, 0x4754, 0x07, 0x01, m64n_gt, 170, 67, 67, M64F_GT | M64F_INTEGRATED | M64F_GTB_DSP | M64F_FIFO_24 | M64F_SDRAM_MAGIC_PLL | M64F_EXTRA_BRIGHT | M64F_NO_HW_TRIPLE },
+ { 0x4754, 0x4754, 0x07, 0x02, m64n_gt, 200, 67, 67, M64F_GT | M64F_INTEGRATED | M64F_GTB_DSP | M64F_FIFO_24 | M64F_SDRAM_MAGIC_PLL | M64F_EXTRA_BRIGHT | M64F_NO_HW_TRIPLE },
+ { 0x4755, 0x4755, 0x00, 0x00, m64n_gtb, 200, 67, 67, M64F_GT | M64F_INTEGRATED | M64F_GTB_DSP | M64F_FIFO_24 | M64F_SDRAM_MAGIC_PLL | M64F_EXTRA_BRIGHT | M64F_NO_HW_TRIPLE },
{ 0x4756, 0x4756, 0x00, 0x00, m64n_iic_p, 230, 83, 83, M64F_GT | M64F_INTEGRATED | M64F_GTB_DSP | M64F_FIFO_24 | M64F_SDRAM_MAGIC_PLL | M64F_EXTRA_BRIGHT },
{ 0x4757, 0x4757, 0x00, 0x00, m64n_iic_a, 230, 83, 83, M64F_GT | M64F_INTEGRATED | M64F_GTB_DSP | M64F_FIFO_24 | M64F_SDRAM_MAGIC_PLL | M64F_EXTRA_BRIGHT },
- { 0x4759, 0x4759, 0x00, 0x00, m64n_iic_p, 230, 83, 83, M64F_GT | M64F_INTEGRATED | M64F_GTB_DSP | M64F_FIFO_24 | M64F_SDRAM_MAGIC_PLL | M64F_EXTRA_BRIGHT },
{ 0x475a, 0x475a, 0x00, 0x00, m64n_iic_a, 230, 83, 83, M64F_GT | M64F_INTEGRATED | M64F_GTB_DSP | M64F_FIFO_24 | M64F_SDRAM_MAGIC_PLL | M64F_EXTRA_BRIGHT },
/* Mach64 LT */
@@ -360,9 +359,6 @@
{ 0x4750, 0x4750, 0x00, 0x00, m64n_gtc_pp, 230, 100, 100, M64F_GT | M64F_INTEGRATED | M64F_RESET_3D | M64F_GTB_DSP | M64F_SDRAM_MAGIC_PLL | M64F_EXTRA_BRIGHT },
{ 0x4751, 0x4751, 0x00, 0x00, m64n_gtc_ppl, 230, 100, 100, M64F_GT | M64F_INTEGRATED | M64F_RESET_3D | M64F_GTB_DSP | M64F_SDRAM_MAGIC_PLL | M64F_EXTRA_BRIGHT },
-
- /* 3D RAGE XL Unknown model */
- { 0x474d, 0x474d, 0x00, 0x00, m64n_xl_66, 230, 83, 63, M64F_GT | M64F_INTEGRATED | M64F_RESET_3D | M64F_GTB_DSP | M64F_SDRAM_MAGIC_PLL | M64F_EXTRA_BRIGHT | M64F_XL_DLL | M64F_MFB_FORCE_4 },
/* 3D RAGE XL PCI-66/BGA */
{ 0x474f, 0x474f, 0x00, 0x00, m64n_xl_66, 230, 83, 63, M64F_GT | M64F_INTEGRATED | M64F_RESET_3D | M64F_GTB_DSP | M64F_SDRAM_MAGIC_PLL | M64F_EXTRA_BRIGHT | M64F_XL_DLL | M64F_MFB_FORCE_4 },
/* 3D RAGE XL PCI-33/BGA */
@@ -1134,11 +1130,11 @@
v_total = v_sync_end + debug.upper_margin;
hSync = 1000000000 / (pixclock_in_ps * h_total);
- vRefresh = (hSync * 1000.0) / v_total;
+ vRefresh = (hSync * 1000) / v_total;
if (par->crtc.gen_cntl & CRTC_INTERLACE_EN)
- vRefresh *= 2.0;
+ vRefresh *= 2;
if (par->crtc.gen_cntl & CRTC_DBL_SCAN_EN)
- vRefresh /= 2.0;
+ vRefresh /= 2;
printk("atyfb: atyfb_set_par\n");
printk(" Set Visible Mode to %ix%i-%i\n", var->xres, var->yres, var->bits_per_pixel);
diff -u /usr/src/linux-2.6.dev/drivers/video/aty/mach64_accel.c /usr/src/linux-2.6.dirty/drivers/video/aty/mach64_accel.c
--- /usr/src/linux-2.6.dev/drivers/video/aty/mach64_accel.c 2004-01-09 07:59:04.000000000 +0100
+++ /usr/src/linux-2.6.dirty/drivers/video/aty/mach64_accel.c 2004-01-23 15:08:29.000000000 +0100
@@ -12,23 +12,42 @@
/*
* Generic Mach64 routines
*/
+
+/* this is for DMA GUI engine! works 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
+
+static u32 rotation24bpp(u32 dx, u32 direction)
+{
+ u32 rotation;
+ if (direction & DST_X_LEFT_TO_RIGHT) {
+ rotation = (dx / 4) % 6;
+ } else {
+ rotation = ((dx + 2) / 4) % 6;
+ }
+
+ return ((rotation << 8) | DST_24_ROTATION_ENABLE);
+}
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 +70,7 @@
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 +165,7 @@
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 +193,7 @@
{
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, rotation = 0;
if (!area->width || !area->height)
return;
@@ -186,11 +204,9 @@
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;
@@ -208,18 +224,22 @@
} else
direction |= DST_X_LEFT_TO_RIGHT;
+ if (info->var.bits_per_pixel == 24) {
+ rotation = rotation24bpp(dx, direction);
+ }
+
wait_for_fifo(4, par);
aty_st_le32(DP_SRC, FRGD_SRC_BLIT, par);
aty_st_le32(SRC_Y_X, (sx << 16) | sy, par);
aty_st_le32(SRC_HEIGHT1_WIDTH1, (width << 16) | area->height, par);
- aty_st_le32(DST_CNTL, direction, par);
+ aty_st_le32(DST_CNTL, direction | rotation, par);
draw_rect(dx, dy, width, area->height, par);
}
void atyfb_fillrect(struct fb_info *info, const struct fb_fillrect *rect)
{
struct atyfb_par *par = (struct atyfb_par *) info->par;
- u32 color = rect->color, dx = rect->dx, width = rect->width;
+ u32 color = rect->color, dx = rect->dx, width = rect->width, rotation = 0;
if (!rect->width || !rect->height)
return;
@@ -238,6 +258,7 @@
/* horizontal coordinates and widths must be adjusted */
dx *= 3;
width *= 3;
+ rotation = rotation24bpp(dx, DST_X_LEFT_TO_RIGHT);
}
wait_for_fifo(3, par);
@@ -247,15 +268,160 @@
par);
aty_st_le32(DST_CNTL,
DST_LAST_PEL | DST_Y_TOP_TO_BOTTOM |
- DST_X_LEFT_TO_RIGHT, par);
+ DST_X_LEFT_TO_RIGHT | rotation, par);
draw_rect(dx, rect->dy, width, rect->height, par);
}
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_bytes, dx = image->dx, dy = image->dy, width = image->width;
+ u32 pix_width_save, pix_width, host_cntl, rotation = 0, src, mix;
+
+ 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;
+ }
+
+ wait_for_idle(par);
+ pix_width = pix_width_save = aty_ld_le32(DP_PIX_WIDTH, par);
+ host_cntl = aty_ld_le32(HOST_CNTL, par) | HOST_BYTE_ALIGN;
+
+ switch (image->depth) {
+ case 1:
+ pix_width &= ~(BYTE_ORDER_MASK | HOST_MASK);
+ pix_width |= (BYTE_ORDER_MSB_TO_LSB | HOST_1BPP);
+ break;
+ case 4:
+ pix_width &= ~(BYTE_ORDER_MASK | HOST_MASK);
+ pix_width |= (BYTE_ORDER_MSB_TO_LSB | HOST_4BPP);
+ break;
+ case 8:
+ pix_width &= ~HOST_MASK;
+ pix_width |= HOST_8BPP;
+ break;
+ case 15:
+ pix_width &= ~HOST_MASK;
+ pix_width |= HOST_15BPP;
+ break;
+ case 16:
+ pix_width &= ~HOST_MASK;
+ pix_width |= HOST_16BPP;
+ break;
+ case 24:
+ pix_width &= ~HOST_MASK;
+ pix_width |= HOST_24BPP;
+ break;
+ case 32:
+ pix_width &= ~HOST_MASK;
+ pix_width |= HOST_32BPP;
+ break;
+ }
+
+ 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;
+
+ rotation = rotation24bpp(dx, DST_X_LEFT_TO_RIGHT);
+
+ pix_width &= ~DST_MASK;
+ pix_width |= DST_8BPP;
+
+ /*
+ * ab IIC we have DP_HOST_TRIPLE_EN bit
+ * this hwaccelerated triple has an issue with not aligned data
+ */
+ if (!M64_HAS(NO_HW_TRIPLE) && image->width % 8 == 0)
+ pix_width |= DP_HOST_TRIPLE_EN;
+ }
+
+ 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(2, par);
+ aty_st_le32(DP_BKGD_CLR, bg, par);
+ aty_st_le32(DP_FRGD_CLR, fg, par);
+ src = MONO_SRC_HOST | FRGD_SRC_FRGD_CLR | BKGD_SRC_BKGD_CLR;
+ mix = FRGD_MIX_S | BKGD_MIX_S;
+ } else {
+ src = MONO_SRC_ONE | FRGD_SRC_HOST;
+ mix = FRGD_MIX_D_XOR_S | BKGD_MIX_D;
+ }
+
+ wait_for_fifo(6, par);
+ aty_st_le32(DP_WRITE_MASK, 0xFFFFFFFF, par);
+ aty_st_le32(DP_PIX_WIDTH, pix_width, par);
+ aty_st_le32(DP_MIX, mix, par);
+ aty_st_le32(DP_SRC, src, par);
+ aty_st_le32(HOST_CNTL, host_cntl, par);
+ aty_st_le32(DST_CNTL, DST_Y_TOP_TO_BOTTOM | DST_X_LEFT_TO_RIGHT | rotation, par);
+
+ draw_rect(dx, dy, width, image->height, par);
+ src_bytes = (((image->width * image->depth) + 7) / 8) * image->height;
+
+ /* manual triple each pixel */
+ if (info->var.bits_per_pixel == 24 && !(pix_width & DP_HOST_TRIPLE_EN)) {
+ int inbit, outbit, mult24, byte_id_in_dword, width;
+ u8 *pbitmapin = (u8*)image->data, *pbitmapout;
+ u32 hostdword;
+
+ for (width = image->width, inbit = 7, mult24 = 0; src_bytes; ) {
+ for (hostdword = 0, pbitmapout = (u8*)&hostdword, byte_id_in_dword = 0;
+ byte_id_in_dword < 4 && src_bytes;
+ byte_id_in_dword++, pbitmapout++) {
+ for (outbit = 7; outbit >= 0; outbit--) {
+ *pbitmapout |= (((*pbitmapin >> inbit) & 1) << outbit);
+ mult24++;
+ /* next bit */
+ if (mult24 == 3) {
+ mult24 = 0;
+ inbit--;
+ width--;
+ }
+
+ /* next byte */
+ if (inbit < 0 || width == 0) {
+ src_bytes--;
+ pbitmapin++;
+ inbit = 7;
+
+ if (width == 0) {
+ width = image->width;
+ outbit = 0;
+ }
+ }
+ }
+ }
+ wait_for_fifo(1, par);
+ aty_st_le32(HOST_DATA0, hostdword, par);
+ }
+ } else {
+ u32 *pbitmap, dwords = (src_bytes + 3) / 4;
+ for (pbitmap = (u32*)(image->data); dwords; dwords--, pbitmap++) {
+ wait_for_fifo(1, par);
+ aty_st_le32(HOST_DATA0, *pbitmap, par);
+ }
+ }
+
+ wait_for_idle(par);
+
+ /* restore pix_width */
+ wait_for_fifo(1, par);
+ aty_st_le32(DP_PIX_WIDTH, pix_width_save, par);
}
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [PATCH] ATI Mach64 accelerated imgblit (sligthly improved)
2004-01-23 17:58 ` Alexander Kern
@ 2004-01-30 21:49 ` James Simmons
0 siblings, 0 replies; 6+ messages in thread
From: James Simmons @ 2004-01-30 21:49 UTC (permalink / raw)
To: Alexander Kern; +Cc: fbdev
I have a new patch at
http://phoenix.infradead.org/~jsimmons/fbdev.diff.gz.
On Fri, 23 Jan 2004, Alexander Kern wrote:
> Am Montag, 19. Januar 2004 19:11 schrieben Sie:
> > > No, it's only imageblit, which has issue.
> > > Regards
> > > Alex Kern
> >
> > Applied your patch. I tested it and it works wonderful. I haven't tested
> > in 24 bpp mode tho. I think that is a minor issue.
> Hi James,
>
> will you update your fbdev.diff.gz? What about resync and submit it again to
> mm tree?
>
> Here is one better patch for accelerated imageblit. It looks like mach64 has a
> problems with buildin trippler and unaligned data. Manual tripple path works
> now!
>
> Regards Alex
>
-------------------------------------------------------
The SF.Net email is sponsored by EclipseCon 2004
Premiere Conference on Open Tools Development and Integration
See the breadth of Eclipse activity. February 3-5 in Anaheim, CA.
http://www.eclipsecon.org/osdn
^ permalink raw reply [flat|nested] 6+ messages in thread
end of thread, other threads:[~2004-01-30 21:49 UTC | newest]
Thread overview: 6+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2004-01-17 0:20 [PATCH] ATI Mach64 accelerated imgblit (sligthly improved) Alexander Kern
2004-01-17 18:25 ` Geert Uytterhoeven
2004-01-17 19:35 ` Alexander Kern
2004-01-19 18:11 ` James Simmons
2004-01-23 17:58 ` Alexander Kern
2004-01-30 21:49 ` James Simmons
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.