sizeof() of DSP

Nov 29, 2024#生活笔记69

AI Translation

This post is translated from Chinese into English through AI.View Original

AI-generated summary

The author discusses the differences between DSP and ARM architectures, particularly how DSP uses 16-bit bytes, affecting structure alignment and the results of `sizeof()`. They express strong opposition to using standard library functions like `memset()` and `memcpy()`, preferring manual loops for batch operations. This approach can lead to issues when porting software to ARM, which may not occur with certain compilers. They note that using `sizeof()` and standard library functions would avoid these problems, as they reference bytes. Despite challenges in a company project, the author uses these functions cautiously in their own open-source library to enhance compatibility.

Finally, I caught a glimpse of the leader's little tail. Due to the different architectures of DSP and ARM, the byte of DSP is 16 bits, which leads to different alignment situations related to structures and different results for sizeof().

Additionally, he firmly opposes standard library functions like memset() and memcpy(), insisting on manually looping for batch operations. This causes problems when porting software to ARM chip platforms to reduce costs. This issue does not even occur with the AC6 compiler (Os), but it manifests when optimizing with GCC.

If one honestly uses standard library combinations like sizeof() and memset(), such problems would not arise, as they all reference bytes as the unit. The requirements in projects that make the company a slave are still useless, but I have used my own open-source library very cautiously, which actually enhances the compatibility of the software library.

This article is synchronized and updated to xLog by Mix Space. The original link is https://www.yono233.cn/notes/26