Hi,
it seems that I found a problem concerning the Vivante GPU or driver stack when using OpenVG radial gradient.
The attached example code (based on the tiger example of the 0.5.0 OpenVG driver and MQX) shows an arc that is filled with a radial gradient. On Vybrid this gradient is displayed way too dark.
I compared the output with the win32 version of the RI code and on i.MX53 using QNX. RI and QNX shows the expected output (a bit different which is to be expected when using gradients but visually almost the same). I took some screenshots using vgReadPixels on Win32, QNX and MQX (for this I use MFS and store result as tga file to FAT formatted SD Card which is mounted as A: at startup)
I also compared the output with i.MX6 Linux and it shows the same dark gradient as Vybrid. On Vybrid the same problem exists with another RTOS that uses - according to the supplier - a vanilla unmodified Vivante stack.
Kind regards
Dirk Dörr
Original Attachment has been moved to: gradient.zip
解決済! 解決策の投稿を見る。
Hello dirkdoerr
Vivante created a patch and we are planning to include this on our next release of the drivers.
I haven't got the chance to look at this and test it because I am traveling until the 4th October.
Dirk, if you want to try it before me, I can send it directly to you.
Best Regards,
Ioseph
Hello dirkdoerr
Vivante created a patch and we are planning to include this on our next release of the drivers.
I haven't got the chance to look at this and test it because I am traveling until the 4th October.
Dirk, if you want to try it before me, I can send it directly to you.
Best Regards,
Ioseph
Hi,
I've applied the patch to the 0.5.0 VG driver of MQX and it seems to fix the problem.
Regards
Dirk
Hi Ioseph,
is it a patch that I could apply to 0.5.0 of your MQX VG driver?
In any case: Yes, I would like to test it.
Best regards
Dirk
I noticed the eglconfig is RGB888, how about RGB565?
I have submitted the issue. I will let you know when I have any updates.
ioseph_martinez do you have any update?
Earlier this morning escalated the issue in Vivante because I entered the ticket the 2 of September. I will let you know when they respond.
Perfect, thanks
Hi Dirk,
Interesting that you are seeing the same on both mx6 and Vybrid, that tells us a lot.
I did some futher testing by changing some parameters: I set the color ramps with alpha = 1.0f so is not transparent and changed the background to the same color as the initial ramp color. I also changed some of the gradient parameters.
Anyways, I could see that indeed there is some blackening on the gradient in Vybrid vs my computer emulation OpenVG, I was expecting very few variations between the path and the background and I can see that on the PC but no on the target. I think I will escalate this to Vivante.
Best Regards,
Ioseph