15:18 <alyssa> zmike: Is ZINK_DEBUG=shaderdb expected to work with radv?

15:19 <alyssa> Looking at RADV_DEBUG=shaders, it looks like the entire fragment shader is getting optimized out

15:19 <alyssa> maybe because we need to add some pretend render target formats to the pipeline

15:20 <alyssa> (if shader objects aren't used, anyway)

15:20 <alyssa> (If that's it, patch incoming if I can figure out Zink)

15:20 <zmike> uh

15:20 <zmike> typically it does work?

15:20 <alyssa> Hmm

15:21 <zmike> but it can only work on stuff that gets precompiled

15:21 <alyssa> Trying to use shader-db's ./run

15:21 <alyssa> https://rosenzweig.io/bad.txt

15:21 <alyssa> Notice that by the time we get to ACO, the fragment shader has been optimized out

15:22 <zmike> I'd guess it's a separate shader object or something

15:22 <zmike> only shaders linked through zink_link_gfx_shader get output for shaderdb

15:23 <zmike> EXT_shader_object can't help there since there's no way to get properties from a shader object

15:23 <alyssa> glLinkProgram is called

15:23 <zmike> then idk

15:23 <alyssa> alright. I will eyes harder

15:23 <zmike> print_pipeline_stats prints the pipeline stats 🤷

15:24 <alyssa> more interested in the assembly than the stats but yeah

15:24 <alyssa> (Wanted to see how a certain piece of GLSL gets compiled by ACO, and Zink + RADV + shader-db ./run + amdgpu drm-shim seemed like the obvious way to do it)

15:25 <zmike> thanks in advance for the bug fixes

15:25 <alyssa> look at me im a zink developer now

15:26 <zmike> smh you didn't believe me

15:27 <alyssa> ?

15:36 <alyssa> got it

15:36 <alyssa> patch incoming

15:46 <alyssa> patch sent

15:46 <alyssa> with that yak shave over, back to torturing ACO

15:56 <zmike> fwiw I'm not sure anyone has ever used zink with shaderdb, so you're officially an adventurer

15:59 <alyssa> Hooray!