The guy interviewing here today apparently reads my blog.
I'm kind of a big deal.
"asks about vulkan concept that relates to performance on the vulkan subreddit"
"gets lectured about premature optimization instead of answered"
Looks like NVidia hardware limits your uniform buffers to 64k ...hmm... that's a bummer.
I'm going to test it out - but does anyone have a good ballpark for the performance hit of using an SSBO vs a UBO in a shader on current desktop gpus?
After my post last week about using arrays of textures to centralize texture bindings in vulkan, I wanted to see if I could do the same for uniform data.
It turns out you can - even if your shaders use different uniform buffer data (you just have to keep the uniform buffers the same size)
I wrote a tutorial/blog post thing about it here: http://kylehalladay.com/blog/tutorial/vulkan/2018/02/05/Bind-Once-Uniform-Data-Vulkan.html
Spend all morning trying to figure out what seemed like an alignment issue in a vulkan experiment I'm doing.
Turns out I had the wrong shader bound the whole time.
I had a bunch of trouble figuring out how to set up an array of textures in a vulkan shader, so even though the reality of it is very easy, I made a quick blog post in case anyone else gets stuck
There's too many interesting problems to solve in gamedev. Sometimes staying focused on one is hard. Especially when what you're doing at work keeps exposing you to new really interesting things.
Something something chasing two rabbits.
In other news... I ignored my inner child wanting to start a bold new project, and instead got sampler arrays working in my material system. Which is pretty nifty.
Got nerd sniped at work and now its taking all my will power to not abandon my vulkan personal project for awhile and write an ios .crash symbolicator that runs on windows.
Im sure there already is one, it just sounds like so much fun to make
It's a bit outside my usual development box, but I'm pretty excited to try out zero mq.
Anyone have any experience getting it compiling against the latest ios sdk? Or is that going to be step one for me? :)
Does anyone have a good resource for doing the "bind everything" approach to textures (a la the "Vulkan Fast Paths" document)?
Or rather - how does one set up an array of textures in a vulkan descriptor set? (not an array texture)
So... not a bug in glslang, powershell just defaults to saving text in UTF-16... which means that if you generate a configuration file for glslangValidator like this:
glslangValidator -c > myconf.conf
in powershell, you end up with a .conf file that glslangValidator can't read, because it wants UTF8
( /s )
Woo nothing says fun like starting your day off with a bug in glslslang!
This bug would be very interesting if I didn't have an impending deadline that it needs to be fixed by.
Last year I made a rule for myself that I wouldn't write any new classes in personal projects (using classes from external libs and stdlib was still fine), and it was a great learning experience.
This year I'm considering making all my personal projects C only. Anyone out there working in primarily C? I'd love some recommendations for a decent math lib, since right now I'm using GLM (which is super template heavy...)
Re-read through Mike Acton's "Typical C++ Bullshit" last night, and noticed a weird comment near the end:
"References? Useless cruft"
So... what's the problem with c++ references? Is it just that you don't get to control when the pointer deref happens?
Oh god it's the beginning of the new year, let the recruiter emailpocalypse begin.
Over the years I find that instead of the oft repeated "break big functions into small ones", I'm writing very large functions and writing long comments throughout that describe what (and why) is going on. (this is when there's a lot of logic that is never needed elsewhere)
I'm then wrapping logic blocks in braces so I hide the parts of the function I don't need for readability.
Anyone else doing this? It feels like it's easier to read, but I'm wondering if there's a good argument against it.