Simply enlarge the window, and the memory will increase exponentially.

[LeeeSe]

more_memory.mp4

If this is normal, please close this issue.

[LeeeSe]

slint = "1.5"
macOS = "14.5 (23F79)"
default_backend ( slint? )

[tronical]

I suspect that this is similar to #4079 , except for macOS. Could you try enabling the renderer-skia feature? That would flip to metal layering - I wonder if that reproduces it.

[LeeeSe]

During the waiting period for a reply, I also noticed that with the increase in the number of conversation turns, the memory usage of the application increases rapidly. When I ask the LLM to “tell me a 1000-word story” three times, the memory usage directly increases by 60MB.

However, when I enable the renderer-skia feature, this issue basically disappears, and the memory usage only increases slightly with each conversation turn. But the memory change caused by zooming remains the same.

---

I mentioned in another issue that my application is an LLM client. Based on the information you provided, I conducted the above tests and obtained these results. The complete code can be found at dali.