topic Great stuff, thanks Michael. in Developing Games on Intel Graphics

Replacing expression e with (false ? ... : e) causes dramatic change in rendering

Alastair_D_1 — Wed, 24 Feb 2016 16:32:15 GMT

I have a fragment shader for which changing:

color = vec3(1.0, 0.1, 0.1) * diffuse;

to:

color = vec3(1.0, 0.1, 0.1) * (false ? mix(targetDepth, time, false) : diffuse);

causes a dramatic change in what is rendered. However, I would have expected this change to have no effect since (false ? ... : diffuse) should evaluate to diffuse.

To reproduce the issue, rename geexlab.txt (attached) to geexlab.xml and load it in GeeXLab. The file as is should render something similar to good.png (attached). This is the expected behaviour of the shader, which originates from here:

http://glslsandbox.com/e#29059

Changing line 99 of geexlab.xml to line 101 should render something similar to bad.png (attached), which is dramatically different.

I am using Iris Graphics 6100, driver version 20.19.15.4352, under Windows 10.

Are others able to reproduce this issue?

Thanks

Ally

Hi Alastair,

Michael_C_Intel2 — Wed, 24 Feb 2016 21:28:06 GMT

Hi Alastair,

I got the ticket filed and let the team know.

-Michael

Hi Alastair,

Michael_C_Intel2 — Tue, 10 May 2016 19:45:12 GMT

Hi Alastair,

This issue is fixed and should be available in the next driver release.

-Michael

Great stuff, thanks Michael.

Alastair_D_1 — Tue, 10 May 2016 22:02:41 GMT

Great stuff, thanks Michael.

Ally

Quote:Michael C. (Intel)

Stefan3D — Fri, 17 Jun 2016 19:04:33 GMT

Michael C. (Intel) wrote:

This issue is fixed and should be available in the next driver release.

Driver v4463 is out.

Looks like you forgot to commit the fixes to the Broadwell branch?

4463 was released 4/29 before

Michael_C_Intel2 — Mon, 20 Jun 2016 14:09:09 GMT

4463 was released 4/29 before we had a fix for this issue.

Re-checked with v4474, issue

Stefan3D — Wed, 29 Jun 2016 01:51:31 GMT

Re-checked with v4474, issue persists

Hello Michael,,

Islam_A_ — Sat, 09 Jul 2016 22:49:44 GMT

Hello Michael,,

this is great stuff...thanks Michael.

Issue still persists with 20

Stefan3D — Wed, 07 Sep 2016 15:13:01 GMT

Issue still persists with 20.19.15.4501 (8-11-2016)