Project Home
Project Home
Documents
Documents
Wiki
Wiki
Discussion Forums
Discussions
Project Information
Project Info
Forum Topic - Very slow graphics rendering: Page 1 of 2 (29 Items)
   
Very slow graphics rendering  
We are seeing some very slow widget rendering times, and are in need of some means of speeding things up.

In general, our situation is this:
We have some background tasks rendering moving waveforms via line draws.  When we display a dialog (over the waveforms),
 we are seeing such a large amount of time taken to draw the dialog, that the background waveform tasks are held off too
 long.
The dialogs consist of a background image and multiple button images and text drawn within the dialog.

With smaller dialogs (e.g. 120x240 background image and two or three button images in the 75x55 range) the rendering 
time is short enough.
It's when we display much larger dialogs that the overall time is too long.  Currently the larger dialog we are seeing 
the problems with is 487x634, and contains about 20 buttons, most of which also have text drawn on them using a PtLabel 
widget.
After we return from the Photon event callback that makes all of the realize calls for the widgets, the Photon server 
time to render all of these buttons is about 250ms.

Some additional context:
Our systems consists of multiple C++ classes, loosely mapped to various Photon widgets.  The images that represent 
buttons are not nested within Photon widgets, but rather nested within a list maintained in our Dialog class.
We are currently using offscreen pre-rendered images and blitting them onto the screen in the Draw() method of dialogs, 
buttons, etc.  This has improved the time needed to display the dialog, (versus just allowing the default Photon 
PtImageArea widget to draw the bitmaps) but the time is still rather long.

Our hardware is:
   1280x800 display
   Freescale iMX53 evaluation board


This is the Draw() method for our raw widget:

void Image::Draw( PtWidget_t *widget, PhTile_t *damage ) {
   Image *image = NULL;

	// Find the Image object from its widget member.
   for( uint32_t x = 0; x < m_image_list.size(); x++ )
   {
      if( m_image_list[x]->Widget_get() == widget )
      {
         image = m_image_list[x];
         break;
      }
   }

	// Did we find the Image object?
   if( image == NULL )
   {
      //TODO raise error
      return;
   }

   if( image->m_image != IMG_NULL )
   {
		// Does this Image object have an offscreen context that we pre-rendered the bitmap into?
		PdOffscreenContext_t *oscontext = IMG::GetOSContext( image->m_image );

		// This is required, otherwise the blitting does not work properly.
		PgFlush();
		
		PhRect_t src;
		src.ul.x = src.ul.y = 0;
		src.lr.x = oscontext->dim.w - 1;
		src.lr.y = oscontext->dim.h - 1;
		
		PgChromaOn();

		// Set the chroma operation to copy every pixel in the source to
		// the destination, except the color 0x0000FF00 (bright green)
		PgSetChroma( 0x0000FF00, Pg_CHROMA_SRC_MATCH | Pg_CHROMA_NODRAW );

		PgContextBlit( oscontext, &src, PhDCGetCurrent(), PtCalcCanvas(widget,NULL) );
		
		PgChromaOff();
		PgFlush();
   }
}

Any suggestions would be greatly appreciated.
Thanks
RE: Very slow graphics rendering  
Hi Mark,

Can you describe why the  flush calls are required, are both required?
Blits are immediate operations at the lowest level (unless IMX53 uses a drawlist), but not immediate at the Pg level.
What context are you PgXXXXX "render draws" targeted at, and what is your current context, is it oscontext?

I would suggest:

PgFlushCx(oscontext);

And eliminate the second PgFlush call and replace it with a PgWaitHWIdle().

Now, if that is all working, then I would suggest using a locking mutex, and two back buffers instead of one (only if 
you are re-rasterizing the entire scene).

I am also thinking a lot of these operations may be falling back to software libffb ... have you brought up this issue 
via a support channel?
What display bpp are you using, 32-bits?  Are your offscreen contexts the same bpp as the display?  (very important)

Regards,
-Derek

-----Original Message-----
From: Mark Anderson [mailto:community-noreply@qnx.com] 
Sent: June-21-12 1:08 PM
To: photon-graphics
Subject: Very slow graphics rendering

We are seeing some very slow widget rendering times, and are in need of some means of speeding things up.

In general, our situation is this:
We have some background tasks rendering moving waveforms via line draws.  When we display a dialog (over the waveforms),
 we are seeing such a large amount of time taken to draw the dialog, that the background waveform tasks are held off too
 long.
The dialogs consist of a background image and multiple button images and text drawn within the dialog.

With smaller dialogs (e.g. 120x240 background image and two or three button images in the 75x55 range) the rendering 
time is short enough.
It's when we display much larger dialogs that the overall time is too long.  Currently the larger dialog we are seeing 
the problems with is 487x634, and contains about 20 buttons, most of which also have text drawn on them using a PtLabel 
widget.
After we return from the Photon event callback that makes all of the realize calls for the widgets, the Photon server 
time to render all of these buttons is about 250ms.

Some additional context:
Our systems consists of multiple C++ classes, loosely mapped to various Photon widgets.  The images that represent 
buttons are not nested within Photon widgets, but rather nested within a list maintained in our Dialog class.
We are currently using offscreen pre-rendered images and blitting them onto the screen in the Draw() method of dialogs, 
buttons, etc.  This has improved the time needed to display the dialog, (versus just allowing the default Photon 
PtImageArea widget to draw the bitmaps) but the time is still rather long.

Our hardware is:
   1280x800 display
   Freescale iMX53 evaluation board


This is the Draw() method for our raw widget:

void Image::Draw( PtWidget_t *widget, PhTile_t *damage ) {
   Image *image = NULL;

	// Find the Image object from its widget member.
   for( uint32_t x = 0; x < m_image_list.size(); x++ )
   {
      if( m_image_list[x]->Widget_get() == widget )
      {
         image = m_image_list[x];
         break;
      }
   }

	// Did we find the Image object?
   if( image == NULL )
   {
      //TODO raise error
      return;
   }

   if( image->m_image != IMG_NULL )
   {
		// Does this Image object have an offscreen context that we pre-rendered the bitmap into?
		PdOffscreenContext_t *oscontext = IMG::GetOSContext( image->m_image );

		// This is required, otherwise the blitting does not work properly.
		PgFlush();
		
		PhRect_t src;
		src.ul.x = src.ul.y = 0;
		src.lr.x = oscontext->dim.w - 1;
		src.lr.y = oscontext->dim.h - 1;
		
		PgChromaOn();

		// Set the chroma operation to copy every pixel in the source to
		// the destination, except the color 0x0000FF00 (bright green)
		PgSetChroma( 0x0000FF00, Pg_CHROMA_SRC_MATCH | Pg_CHROMA_NODRAW );

		PgContextBlit( oscontext, &src, PhDCGetCurrent(), PtCalcCanvas(widget,NULL)...
View Full Message
Re: RE: Very slow graphics rendering  
Thanks for the lightning fast response! How do I know if it is falling back to software libffb vs HW?

Thanks
RE: RE: Very slow graphics rendering  
No way to easily tell from the API end ... what devg are you using?

pidin -Pio-display mem

I might be a good idea to file a support ticket on this once we exhaust simple efforts.

-----Original Message-----
From: Mark Anderson [mailto:community-noreply@qnx.com] 
Sent: June-21-12 1:43 PM
To: photon-graphics
Subject: Re: RE: Very slow graphics rendering

Thanks for the lightning fast response! How do I know if it is falling back to software libffb vs HW?

Thanks




_______________________________________________

Photon microGUI
http://community.qnx.com/sf/go/post93818
To cancel your subscription to this discussion, please e-mail photon-graphics-unsubscribe@community.qnx.com
Re: RE: RE: Very slow graphics rendering  
The BSP driver for the MX53 does not have the io-gpumgr which I saw in the file releases section.  In order to have HW 
acceleration do we have to use the patched driver?

Below is the result of the pidin cmd:

# pidin -Pio-graphics mem
     pid tid name               prio STATE            code  data        stack
5980183   1 on/bin/io-graphics  12r CONDVAR           72K 5952K   24K(516K)*
5980183   2 on/bin/io-graphics  10r RECEIVE           72K 5952K  8192(516K)
5980183   3 on/bin/io-graphics  12r REPLY             72K 5952K   12K(132K)
            libc.so.3          @ 1000000             460K   16K
            libgf.so.1         @78000000              96K  4096
            libfont.so.1       @78019000              28K  4096
            FCcore.so          @78021000              32K  8192
            libblkcache.so.2   @7802b000              12K  4096
            libphrender.so.2   @78030000             224K  8192
            phfont.so          @78070000              80K   16K
            PHFcore.so         @78088000              24K  8192
            libFF-T2K.so.3     @78090000             304K   16K
            ttfFFcore.so       @780e0000              32K  8192
            devg-imx51.so      @780ea000              44K  8192
            libph.so.3         @78100000            1036K   32K
            libffb.so.2        @78210000             148K  8192
            libm.so.2          @78240000             180K   12K
            m/Pf5980183.11a310 @28000000 (       0)         64K
            em/ctl-0000,0006,0 @28010000 (       0)        4096
            /dev/mem           @28011000 (       0)        8192
            /dev/mem           @28013000 (       0)         16K
            /dev/mem           @28020000 (       0)        300K
            /dev/mem           @28070000 (       0)        200K
            face-0000,0006,0:0 @28100000 (       0)       4000K


We have submitted a ticket I was just trying to get an answer ASAP.
RE: RE: RE: Very slow graphics rendering  
Hi Mark,

devg-imx51.so driver is all software (libffb) for draw ops.

File a support ticket, and we can see what we have in the toolbox.
I cannot provide binaries via support forum.

Check the display bpp versus the offscreen context bpp, make sure they match.
Change the PgFlush to the PgFlushCx and replace the second PgFlush with the PgWaitHWIdle().

Regards,
-Derek

-----Original Message-----
From: Mark Anderson [mailto:community-noreply@qnx.com] 
Sent: June-21-12 1:57 PM
To: photon-graphics
Subject: Re: RE: RE: Very slow graphics rendering

The BSP driver for the MX53 does not have the io-gpumgr which I saw in the file releases section.  In order to have HW 
acceleration do we have to use the patched driver?

Below is the result of the pidin cmd:

# pidin -Pio-graphics mem
     pid tid name               prio STATE            code  data        stack
5980183   1 on/bin/io-graphics  12r CONDVAR           72K 5952K   24K(516K)*
5980183   2 on/bin/io-graphics  10r RECEIVE           72K 5952K  8192(516K)
5980183   3 on/bin/io-graphics  12r REPLY             72K 5952K   12K(132K)
            libc.so.3          @ 1000000             460K   16K
            libgf.so.1         @78000000              96K  4096
            libfont.so.1       @78019000              28K  4096
            FCcore.so          @78021000              32K  8192
            libblkcache.so.2   @7802b000              12K  4096
            libphrender.so.2   @78030000             224K  8192
            phfont.so          @78070000              80K   16K
            PHFcore.so         @78088000              24K  8192
            libFF-T2K.so.3     @78090000             304K   16K
            ttfFFcore.so       @780e0000              32K  8192
            devg-imx51.so      @780ea000              44K  8192
            libph.so.3         @78100000            1036K   32K
            libffb.so.2        @78210000             148K  8192
            libm.so.2          @78240000             180K   12K
            m/Pf5980183.11a310 @28000000 (       0)         64K
            em/ctl-0000,0006,0 @28010000 (       0)        4096
            /dev/mem           @28011000 (       0)        8192
            /dev/mem           @28013000 (       0)         16K
            /dev/mem           @28020000 (       0)        300K
            /dev/mem           @28070000 (       0)        200K
            face-0000,0006,0:0 @28100000 (       0)       4000K


We have submitted a ticket I was just trying to get an answer ASAP.



_______________________________________________

Photon microGUI
http://community.qnx.com/sf/go/post93820
To cancel your subscription to this discussion, please e-mail photon-graphics-unsubscribe@community.qnx.com
Re: RE: RE: RE: Very slow graphics rendering  
Case: 00115395 logged by RBC who is developing the code for us.

Thanks
Mark
Re: RE: RE: RE: Very slow graphics rendering  
Derek,

Hi - I'm one of the RBC team working on this project.  Thanks for the suggestions and info!

Changing the second PgFlush() to PgWaitHWIdle() doesn't seem to have any noticeable impact on the timing.

Changing the first PgFlush() to  PgFlushCx( oscontext ) causes bitmaps to not be drawn.  Is the oscontext the correct 
argument for that call?  Since PgFlush() clears the current draw context, it seems like the only option.

For the bpp issue, we've tried creating the offscreen context with both:
	   oscontext = PdCreateOffscreenContext( image->type, image->size.w, image->size.h, 0 );
                           and
	   oscontext = PdCreateOffscreenContext( 0, image->size.w, image->size.h, 0 );
We get about the same timing for both.

Thanks,
Don
RE: RE: RE: RE: Very slow graphics rendering  
Hi Don,

That was part of my questions, as to whether the correct context was being flushed.
Whatever context that has the draw stream is the context that should be flushed.

Use image->type for the type.

The second PgFlush potentially causes for overhead, use the PgWaitHWIdle().


Maybe describe what is going on with this code.

Are you rendering Photon primitives into the offscreen context,
then blitting the results to the screen?

Thanks,
-Derek

-----Original Message-----
From: Jimmy Townsend [mailto:community-noreply@qnx.com] 
Sent: June-21-12 5:13 PM
To: photon-graphics
Subject: Re: RE: RE: RE: Very slow graphics rendering

Derek,

Hi - I'm one of the RBC team working on this project.  Thanks for the suggestions and info!

Changing the second PgFlush() to PgWaitHWIdle() doesn't seem to have any noticeable impact on the timing.

Changing the first PgFlush() to  PgFlushCx( oscontext ) causes bitmaps to not be drawn.  Is the oscontext the correct 
argument for that call?  Since PgFlush() clears the current draw context, it seems like the only option.

For the bpp issue, we've tried creating the offscreen context with both:
	   oscontext = PdCreateOffscreenContext( image->type, image->size.w, image->size.h, 0 );
                           and
	   oscontext = PdCreateOffscreenContext( 0, image->size.w, image->size.h, 0 ); We get about the same timing for both.

Thanks,
Don



_______________________________________________

Photon microGUI
http://community.qnx.com/sf/go/post93823
To cancel your subscription to this discussion, please e-mail photon-graphics-unsubscribe@community.qnx.com
Re: RE: RE: RE: RE: Very slow graphics rendering  
Derek,

The second flush is now PgWaitHWIdle(), and that works fine.

The first flush, we get these results:
Works: PgFlush();
Works:  PgFlushCx( PhDCGetCurrent() );
Does not work:  PgFlushCx( oscontext );

In our case, none of these changes seems to have made a noticeable improvement in our dialog drawing time.

Yes, we are creating an offscreen context for each image that we display (buttons, dialog backgrounds, etc.) and then 
the Draw() method of our raw widget is blitting them onto the screen.  (the code sample Mark posted initially).

This is the code we use to create those offscreen contexts at system startup:

	   oscontext = PdCreateOffscreenContext( image->type, image->size.w, image->size.h, 0 );
		if( oscontext )
		{
			PhDrawContext_t *dc = PhDCSetCurrent( oscontext );
			PgDrawPhImagemx( NULL, image, 0 );			
			PgFlush();
			PgWaitHWIdle();
			PhDCSetCurrent( dc );
		}

Thanks,
Don
RE: RE: RE: RE: RE: Very slow graphics rendering  
Hi Don,

Ok, now I get what you are doing.

There is still not been a reply as to what bpp is being used for the display and what bpp the images are.
Can you check into that?  To get optimum performance, they must all match.

You can tell your support contact to send me an email, if they are not planning to do so already.

Regards + good weekend,
-Derek

-----Original Message-----
From: Jimmy Townsend [mailto:community-noreply@qnx.com] 
Sent: June-22-12 3:25 PM
To: photon-graphics
Subject: Re: RE: RE: RE: RE: Very slow graphics rendering

Derek,

The second flush is now PgWaitHWIdle(), and that works fine.

The first flush, we get these results:
Works: PgFlush();
Works:  PgFlushCx( PhDCGetCurrent() );
Does not work:  PgFlushCx( oscontext );

In our case, none of these changes seems to have made a noticeable improvement in our dialog drawing time.

Yes, we are creating an offscreen context for each image that we display (buttons, dialog backgrounds, etc.) and then 
the Draw() method of our raw widget is blitting them onto the screen.  (the code sample Mark posted initially).

This is the code we use to create those offscreen contexts at system startup:

	   oscontext = PdCreateOffscreenContext( image->type, image->size.w, image->size.h, 0 );
		if( oscontext )
		{
			PhDrawContext_t *dc = PhDCSetCurrent( oscontext );
			PgDrawPhImagemx( NULL, image, 0 );			
			PgFlush();
			PgWaitHWIdle();
			PhDCSetCurrent( dc );
		}

Thanks,
Don




_______________________________________________

Photon microGUI
http://community.qnx.com/sf/go/post93842
To cancel your subscription to this discussion, please e-mail photon-graphics-unsubscribe@community.qnx.com
RE: RE: RE: RE: RE: Very slow graphics rendering  
Hi Don,

Just wanted to add, I appreciate that there has been no performance improvement as of yet, but we need to check all 
these points before proceeding further.

Also, what MHz rating is the CPU?

Regards,
Derek

-----Original Message-----
From: Jimmy Townsend [mailto:community-noreply@qnx.com] 
Sent: June-22-12 3:25 PM
To: photon-graphics
Subject: Re: RE: RE: RE: RE: Very slow graphics rendering

Derek,

The second flush is now PgWaitHWIdle(), and that works fine.

The first flush, we get these results:
Works: PgFlush();
Works:  PgFlushCx( PhDCGetCurrent() );
Does not work:  PgFlushCx( oscontext );

In our case, none of these changes seems to have made a noticeable improvement in our dialog drawing time.

Yes, we are creating an offscreen context for each image that we display (buttons, dialog backgrounds, etc.) and then 
the Draw() method of our raw widget is blitting them onto the screen.  (the code sample Mark posted initially).

This is the code we use to create those offscreen contexts at system startup:

	   oscontext = PdCreateOffscreenContext( image->type, image->size.w, image->size.h, 0 );
		if( oscontext )
		{
			PhDrawContext_t *dc = PhDCSetCurrent( oscontext );
			PgDrawPhImagemx( NULL, image, 0 );			
			PgFlush();
			PgWaitHWIdle();
			PhDCSetCurrent( dc );
		}

Thanks,
Don




_______________________________________________

Photon microGUI
http://community.qnx.com/sf/go/post93842
To cancel your subscription to this discussion, please e-mail photon-graphics-unsubscribe@community.qnx.com
Re: RE: RE: RE: RE: RE: Very slow graphics rendering  
Derek,

The driver configuration is  argb8888.  The images are 8-bit PNGs loaded via PxLoadImage(), and the image type after 
loading is mostly Pg_IMAGE_DIRECT_8888, with some being Pg_IMAGE_DIRECT_888.  (not sure why some are different types, 
due to how they were created by the graphic artist I assume)

Hope you have a good weekend as well!

Thanks,
Don
RE: RE: RE: RE: RE: RE: Very slow graphics rendering  
OK, great.

OK, you need to investigate the scope of the Pg_IMAGE_DIRECT_888 impact, because it is brutally slow to process those 
images out to anything other than a 24-bit hardware accelerated surface (which is not available for this chipset).  If 
they were loaded as Pg_IMAGE_DIRECT_8888, then the alpha bits are just blindly forced to 0xFF, which should produce the 
exact same visual representation.  Can you change those 888 to 8888 and recompile and see what happens versus display 
(hopefully there is no reliance on 24-bit 888 anywhere), and performance.  The underlying libimg will upconvert the 
image type on load.

Thanks,
Derek

-----Original Message-----
From: Jimmy Townsend [mailto:community-noreply@qnx.com] 
Sent: June-22-12 7:57 PM
To: photon-graphics
Subject: Re: RE: RE: RE: RE: RE: Very slow graphics rendering

Derek,

The driver configuration is  argb8888.  The images are 8-bit PNGs loaded via PxLoadImage(), and the image type after 
loading is mostly Pg_IMAGE_DIRECT_8888, with some being Pg_IMAGE_DIRECT_888.  (not sure why some are different types, 
due to how they were created by the graphic artist I assume)

Hope you have a good weekend as well!

Thanks,
Don



_______________________________________________

Photon microGUI
http://community.qnx.com/sf/go/post93847
To cancel your subscription to this discussion, please e-mail photon-graphics-unsubscribe@community.qnx.com
Re: RE: RE: RE: RE: RE: RE: Very slow graphics rendering  
Ok, that's very interesting info.

Just to see if I'm understanding correctly, since our driver is configured to use the alpha channel, our images should 
be also, correct?  (even though we don't use alpha in our application - we're using chroma for transparency.)

So, here's what I've tried:

	PhImage_t *new_img = PiConvertImage( opened_image, NULL, Pg_IMAGE_DIRECT_8888, ( Pi_FREE | Pi_CALC_PALETTE ) );

For the images that loaded as just 888.  This results in a black image being rendered.  I've also tried this without the
 Pi_CALC_PALETTE flag - same result)

I'm taking it to mean that we need to convert those image assets to have an alpha channel outside of Photon, so they are
 then just initially loaded as 8888.

Another question - is their any way programmatically or via the command shell to determine with 100% certainty what mode
 our video driver is in?  I know what I'm seeing in the config files, but I always like to be able to check the real 
running end result just in case.

Thanks,
Don

RE: RE: RE: RE: RE: RE: RE: Very slow graphics rendering  
Hi Don,

Ok, this is interesting ... I actually did not mean to use PiConvertImage() ... I meant to load it specifically as 8888 
via PxLoadImage() ... the underlying library, libimg has an API call img_load_file() which can do this directly, but it 
turns out that PxLoadImage does not exposes this capability.  I checked the documentation on PxLoadImage(), and it does 
not appear that we can override the format at load, this is unfortunate :(.  The PiConvertImage() also sets the upper 8-
bits to 0, so this is also unfortunate.

So the only way to get around this:

Call img_load_file() directly, and override the format to be 8888.
Use PhCreateImage() to wrap the img_t structure into a PhImage_t structure.

Support should be able to help with this part.

The other option, is to have the images generated in 32-bit 8888 for the get go, then you avoid all overhead in 
converting them.

-Derek

-----Original Message-----
From: Jimmy Townsend [mailto:community-noreply@qnx.com] 
Sent: June-26-12 4:50 PM
To: photon-graphics
Subject: Re: RE: RE: RE: RE: RE: RE: Very slow graphics rendering

Ok, that's very interesting info.

Just to see if I'm understanding correctly, since our driver is configured to use the alpha channel, our images should 
be also, correct?  (even though we don't use alpha in our application - we're using chroma for transparency.)

So, here's what I've tried:

	PhImage_t *new_img = PiConvertImage( opened_image, NULL, Pg_IMAGE_DIRECT_8888, ( Pi_FREE | Pi_CALC_PALETTE ) );

For the images that loaded as just 888.  This results in a black image being rendered.  I've also tried this without the
 Pi_CALC_PALETTE flag - same result)

I'm taking it to mean that we need to convert those image assets to have an alpha channel outside of Photon, so they are
 then just initially loaded as 8888.

Another question - is their any way programmatically or via the command shell to determine with 100% certainty what mode
 our video driver is in?  I know what I'm seeing in the config files, but I always like to be able to check the real 
running end result just in case.

Thanks,
Don





_______________________________________________

Photon microGUI
http://community.qnx.com/sf/go/post93880
To cancel your subscription to this discussion, please e-mail photon-graphics-unsubscribe@community.qnx.com
Re: RE: RE: RE: RE: RE: RE: Very slow graphics rendering  
Derek,

Also, we have gotten a reply via the support case we opened.  Bjorn Weber gave us a few suggestions, including matching 
the pixel format to the driver (which I think we've been discussing here), turning of font anti-aliasing, and looking 
into hardware acceleration.

I've replied to him and indicated we'd been discussing this with you on the forum as well.

As to the hardware acceleration - we have a patch "patch-650-2413-iMX53GPU.tar" that Mark found.  Is this the one we 
should be trying to incorporate to give us hardware acceleration?  And if so, where can we find a description of how to 
apply and enable this patch?

Thanks,
Don
RE: RE: RE: RE: RE: RE: RE: Very slow graphics rendering  
I do not know the contents of patches, what does it contain?  If it is just 3D HW accel, it will not help in the 
situation.
 
-----Original Message-----
From: Jimmy Townsend [mailto:community-noreply@qnx.com] 
Sent: June-26-12 4:51 PM
To: photon-graphics
Subject: Re: RE: RE: RE: RE: RE: RE: Very slow graphics rendering

Derek,

Also, we have gotten a reply via the support case we opened.  Bjorn Weber gave us a few suggestions, including matching 
the pixel format to the driver (which I think we've been discussing here), turning of font anti-aliasing, and looking 
into hardware acceleration.

I've replied to him and indicated we'd been discussing this with you on the forum as well.

As to the hardware acceleration - we have a patch "patch-650-2413-iMX53GPU.tar" that Mark found.  Is this the one we 
should be trying to incorporate to give us hardware acceleration?  And if so, where can we find a description of how to 
apply and enable this patch?

Thanks,
Don



_______________________________________________

Photon microGUI
http://community.qnx.com/sf/go/post93882
To cancel your subscription to this discussion, please e-mail photon-graphics-unsubscribe@community.qnx.com
Re: RE: RE: RE: RE: RE: RE: RE: Very slow graphics rendering  
The description on the download says "OpenGL ES 1.1, OpenGL ES 2.0, OpenVG 1.1 acceleration on iMX51/iMX53"

See: http://community.qnx.com/sf/frs/do/viewSummary/projects.graphics/frs

-Mark
RE: RE: RE: RE: RE: RE: RE: RE: Very slow graphics rendering  
Ok, that will not help this current situation.

-----Original Message-----
From: Mark Anderson [mailto:community-noreply@qnx.com] 
Sent: June-27-12 9:04 AM
To: photon-graphics
Subject: Re: RE: RE: RE: RE: RE: RE: RE: Very slow graphics rendering

The description on the download says "OpenGL ES 1.1, OpenGL ES 2.0, OpenVG 1.1 acceleration on iMX51/iMX53"

See: http://community.qnx.com/sf/frs/do/viewSummary/projects.graphics/frs

-Mark



_______________________________________________

Photon microGUI
http://community.qnx.com/sf/go/post93897
To cancel your subscription to this discussion, please e-mail photon-graphics-unsubscribe@community.qnx.com
Re: RE: RE: RE: RE: RE: RE: RE: RE: Very slow graphics rendering  
Ok, What are our options for getting hardware acceleration on the MX53?  If I were to invoke a priority support 
agreement what would that do for us?

Was this work already done for the QNX/Freescale/EB collaboration project? : http://www.qnx.com/news/pr_4633_1.html


Thanks,
Mark
RE: RE: RE: RE: RE: RE: RE: RE: RE: Very slow graphics rendering  
Hi Mark,

You would need to discuss this with support, and a sales rep.  I do not have scope into the available options.
In my opinion, the image bbp situation needs to be resolved before pursuing other options.

-Derek


-----Original Message-----
From: Mark Anderson [mailto:community-noreply@qnx.com] 
Sent: June-27-12 9:36 AM
To: photon-graphics
Subject: Re: RE: RE: RE: RE: RE: RE: RE: RE: Very slow graphics rendering

Ok, What are our options for getting hardware acceleration on the MX53?  If I were to invoke a priority support 
agreement what would that do for us?

Was this work already done for the QNX/Freescale/EB collaboration project? : http://www.qnx.com/news/pr_4633_1.html


Thanks,
Mark



_______________________________________________

Photon microGUI
http://community.qnx.com/sf/go/post93900
To cancel your subscription to this discussion, please e-mail photon-graphics-unsubscribe@community.qnx.com
Re: Very slow graphics rendering  
Hello,

from a support point of view I see the problem that officially there is 
no Photon for ARMv7. So anything in regards of increasing the 
performance would be a change on something that does not exist ;) This 
is also why my support is kind of restricted as this would be more like 
Prio Support or even services.

Greetings,

     bweb


Am 27.06.2012 15:40, schrieb Derek Leach:
> Hi Mark,
>
> You would need to discuss this with support, and a sales rep.  I do not have scope into the available options.
> In my opinion, the image bbp situation needs to be resolved before pursuing other options.
>
> -Derek
>
>
> -----Original Message-----
> From: Mark Anderson [mailto:community-noreply@qnx.com]
> Sent: June-27-12 9:36 AM
> To: photon-graphics
> Subject: Re: RE: RE: RE: RE: RE: RE: RE: RE: Very slow graphics rendering
>
> Ok, What are our options for getting hardware acceleration on the MX53?  If I were to invoke a priority support 
agreement what would that do for us?
>
> Was this work already done for the QNX/Freescale/EB collaboration project? : http://www.qnx.com/news/pr_4633_1.html
>
>
> Thanks,
> Mark
>
>
>
> _______________________________________________
>
> Photon microGUI
> http://community.qnx.com/sf/go/post93900
> To cancel your subscription to this discussion, please e-mail photon-graphics-unsubscribe@community.qnx.com
>

-- 
QNX Software Systems GmbH           Tel: +49 (511) 94091-475
Am Listholze 76                     Fax: +49 (511) 94091-199
30177 Hannover                      bweber@qnx.com
Germany                             http://www.qnx.com/


Re: Very slow graphics rendering  
The graphics acceleration in the MX53 appears to be the same as the MX35.  They are both use the AMD Z160 graphics 
accelerator.  Is there a driver from the mx35 that would allow us to get the graphics acceleration?

RE: Very slow graphics rendering  
That patch only supplies 3D.  Photon uses 2D acceleration for its main use, to draw the widgets, etc.

-----Original Message-----
From: Mark Anderson [mailto:community-noreply@qnx.com] 
Sent: June-27-12 4:52 PM
To: photon-graphics
Subject: Re: Very slow graphics rendering

The graphics acceleration in the MX53 appears to be the same as the MX35.  They are both use the AMD Z160 graphics 
accelerator.  Is there a driver from the mx35 that would allow us to get the graphics acceleration?





_______________________________________________

Photon microGUI
http://community.qnx.com/sf/go/post93909
To cancel your subscription to this discussion, please e-mail photon-graphics-unsubscribe@community.qnx.com