#Meshy consumes 140 MB on startup, not too bad for a Python app using #GTK 4 and #libadwaita running in Flatpak.
But once you open the emoji picker it jumps to 650 MB with the vulkan renderer. As if it loaded all 3000 UTF emojis into memory at once. With the gl renderer the memory consumption jumps by ~30 MB and then grows as you browse the emojis. Cairo is even better at it, but there is a performance penalty.
@sesivany it does load all the emojis (in an idle loop, to avoid blocking the UI). It should use the same amount of memory in both GL and Vulkan, though
@GTK You're right, the original profiling was done on an Intel machine, I just did it on an AMD and ARM64 machines and GL and Vulkan behaves the same memory consumption wise. And the jump isn't so drastic. So I guess it may be some Intel GPU driver bug after all.