有用户在使用SF32-lb52开发板进行图像运算时出现了卡顿的现象，这里建议用户在进行图形旋转的运算时，尽量将源素材放在psram或者sram，而不是直接从flash读取运算。因为flash的延时和带宽对于现有GPU的旋转运算非常不友好，会导致运算速率特别慢。
具体可以参考55x的内存使用指南，后续也会提供52x相关的内存使用指南。
2 Likes
So you mean the resource file needs to be moved to memory first before performing the rotation operation, and this is faster than rotating directly?
What I mean is, the resource file will definitely reside in flash, due to the large amount of resources. Even if placed in (P)SRAM, it can only be temporarily loaded during computation. Can there still be an acceleration effect despite the extra step?
Is there a size limit for the original image?