Bug 28144

Summary: JXRenderMark causes system freeze
Product: xorg Reporter: Clemens Eisserer <linuxhippy>
Component: Driver/RadeonAssignee: xf86-video-ati maintainers <xorg-driver-ati>
Status: RESOLVED FIXED QA Contact: Xorg Project Team <xorg-team>
Severity: normal    
Priority: medium    
Version: unspecified   
Hardware: Other   
OS: All   
Whiteboard:
i915 platform: i915 features:
Attachments:
Description Flags
xorg log
none
kernel output after boot
none
lspci output none

Description Clemens Eisserer 2010-05-17 07:18:37 UTC
Created attachment 35707 [details]
xorg log

Running the JXRendermark reliably freezes my system when executing the "Transformed Blit Billinear sharp edges"-test.
JXRenderMark can be downloaded from: http://93.83.133.214/downloads/JXRenderMark-1.0.1.zip

Basically the benchmark only executes a series of render-commands, like x11perf does, the test which causes the hang basically looks like the code at the end of the report.


System-Information:
X.Org X Server 1.8.0
Radeon 6.13.0
"ATI Radeon Mobility 9000 (M9) Lf (AGP)" (ChipID = 0x4c66)


void billinearTransformedBlit(int area, int x, int y)
{
  int maskSize = area/8;
  maskSize = maskSize == 0 ? 1 : area;
  area = maskSize*8;

  XRenderFillRectangle (display, PictOpSrc, alphaMask, &setAlpha, 0, 0, maskSize, maskSize);

  XTransform xf;
 //transform changes omited
 XRenderSetPictureTransform (display, sourceImage, &xf);
 //transform changes omited
 XRenderSetPictureTransform (display, alphaMask, &xf);

  XRenderComposite(display, PictOpOver, sourceImage, alphaMask, backBuffer, 0, 0, 0, 0, x, y, area, area);  

  XRenderFillRectangle (display, PictOpClear, alphaMask, &setAlpha, 0, 0, maskSize, maskSize);
  
   //transform changes omited
  XRenderSetPictureTransform (display, alphaMask, &xf);
}
Comment 1 Clemens Eisserer 2010-05-17 07:19:50 UTC
Created attachment 35708 [details]
kernel output after boot
Comment 2 Clemens Eisserer 2010-05-17 07:20:30 UTC
Created attachment 35709 [details]
lspci output
Comment 3 Michel Dänzer 2010-05-18 00:33:32 UTC
Do any of the usual suspects work around the problem?

Option "RenderAccel" "off", radeon.agpmode=1/2/-1, kernel 2.6.34, ...
Comment 4 Clemens Eisserer 2010-05-18 00:56:16 UTC
RenderAccel "off" helps
radeon.agpmode=-1 doesn't help
I'll try kernel 2.6.34 as soon as it is in fedora's updates-testing.

I don't think the upload/download hooks would make any difference, as far as I know no fallbacks are hit. What puzzles me is that the test works well for the 15x15 case, but crashes for 75x75.

Thanks, Clemens
Comment 5 Clemens Eisserer 2010-05-22 03:35:53 UTC
Would it be possible to just give the Benchmark a try on R200 class hardware?
Shouldn't take more than 5min, precompiled binaries are included in the archive, and its really not more than ./jxrendermark.
Comment 6 Clemens Eisserer 2010-06-04 03:36:44 UTC
Ok, just tried the benchmark on 2.6.34 and everything works fine there :)

Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.