Hi,
I have a large vector that resides in STM32F407 Core Coupled Memory (CCM) while another vector in SRAM.
Am looking for a better way to copy from one to another besides the traditional memcpy because DMA transfer is not an option anymore.
Does CMSIS Vector Copy function offer any advantage compared to memcpy in this context? Or, is there any better idea?
Thanks in advance.
Seng Tak
Well, the documentation says,
"Copies sample by sample from source vector to destination vector." (my emphasis) www.keil.com/.../group__copy.html
Which sounds unlikely to be faster than a raw block copy?
But, surely, there's an easy way to find out - just try it?!
Or look at the source?
Why can't you use DMA?
Hi Andrew,
Thanks for your prompt reply.
Yes I have just tried it and it proven what the document says, its much slower than memcpy...but just wandering and interested to gather ideas as on why is the trouble of offering this CMSIS DSP function while memcpy does a better job on block copying...
In STM32F407, CCM is directly accessible only by the Cortex core via D-bus and no other D-bus masters...unfortunately I can't put both vectors on CCM because I have exhausted it.
Use LDM/STM via scratch registers.