06:01 ce7Thohh: Could someone point me to where I could modify the amdgpu vulkan implementation to disable the behaviour where system ram is used instead of gpu vram during vram overcommitment? In this scenario I'd just like to return VK_ERROR_OUT_OF_HOST_MEMORY instead of allocating on the heap. Sry if this is the wrong place to ask
07:44 tzimmermann: jfalempe, hi. i intent to clean up the TX-chip detection in ast, but i'd like to here your option first.
11:06 jfalempe: tzimmermann: we can discuss that tomorrow, I'm not really available today.
11:07 tzimmermann: ok
16:45 enunes: looks like our test-docs job is broken for MRs which update docs https://gitlab.freedesktop.org/mesa/mesa/-/jobs/63163181 I guess https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30952 made all those deps also be checked for docs and they are not in the image?
17:07 enunes: ok this may resolve it https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/31025
18:40 tail_hunter: hello i have old intel i5 480m. i want to use opengl2.1's fragment shader for process data. i got egl working with drm/gbm/bo. i map dma/bo framebuffer to ram and i can read it fast. my question is when i dont have `GBM_BO_USE_LINEAR` in`gbm_surface_create()` i get 2.5x more fps, which is a lot. but framebuffer data gets tiled/not readable easily. do i need mesa's intel/isl for this?
18:51 glehmann: why does NIR use two different types of linked lists in the IR
19:24 Company: I've been looking at Vulkan swapchains because of the rpi allocation issue
19:24 Company: and GTK sets VkSwapchainCreateInfoKHR.minImageCount = 4
19:25 Company: I was wondering why we don't set it to 2 and it turns out Mesa interprets minImageCount not as minimum but as actual number of images to allocate
19:26 Company: and I was taught 4 is required sometimes with direct scanout etc
19:26 Company: but usually I just need 2
19:27 Company: so I was wondering why Mesa doesn't allocate more images in such situations which would allow me to pass 2 instead of 4