Comments on: Effective C Tips #2 – Defining buffer sizes

By: Nigel Jones

Nigel Jones — Tue, 12 May 2009 12:17:00 +0000

Yevheniy – you are quite right. Thanks for pointing this out. I’ve modified the post and of course given you credit for the observation. Now I just have to go and modify my code!

By: Yevheniy

Yevheniy — Tue, 12 May 2009 10:45:00 +0000

Nigel, when RX_BUF_SIZE is 1 it will also be considered as a valid size by your code. Thus it’s better to check that RX_BUF_SIZE is not 1 as well.

By: Miro Samek

Miro Samek — Fri, 20 Mar 2009 18:39:00 +0000

Nigel,According to your own teaching (in “writing fast loops” post) the head counter should count down rather than up. Then the comparison for wrap-around would be against zero, which is special and probably as fast as masking the head counter. Additional bonus is that the buffer size does not need to be power of 2 anymore, so you can save some RAM.

By: Nigel Jones

Nigel Jones — Thu, 05 Mar 2009 00:07:00 +0000

Bruno it’s all about speed in this case. If you have a high frequency interrupt, then shaving a couple of op codes off the ISR handler can have a dramatic impact on the CPU utilization. BTW, the code you have suggested also has a potential problem if RX_BUF_SIZE is zero. I say ‘potential’ because most half decent compilers would flag it. The lousy ones would not, and then you’d have a problem.

By: Bruno Santiago

Bruno Santiago — Wed, 25 Feb 2009 17:58:00 +0000

Probably the code you posted is faster and the latency is determinisc, but why not just use a if?++RxHead;if( RxHead >= RX_BUF_SIZE ){ RxHead = 0;}Maybe it´s lack of insinght, but in this case I really don´t see the point of complicating things.