To do a more effiency memcpy in C674x DSP?

SuitJune Young

HI,TIer.

Right now we wanna develop our own algorithm in DSP,and in the alg,there are lots of memcpy.

I have do a test in which memcpy is placed in for loop,the when I read the .asm file,I found that only two parallel.

for (c = 0; c < 10; c++)
{
memcpy(fff, ttt, 2);
}

So,I think memcpy in CCS is not effiency software pipleined.And,do I need to write a more fast memcpy_dsp function in CCS to make algorithm fun faster?

BR!

over 10 years ago

George Mock over 10 years ago

TI__Guru**** 252710 points

This does not seem like a typical use case. This loop iterates 10 times, and each iteration calls memcpy to copy over 2 bytes. Given the unusual nature of the operation, I think the compiler has generated reasonable output. Consider changing the call to memcpy to copy over much more than 2 bytes.

Thanks and regards,

-George

Archaeologist over 10 years ago in reply to George Mock

TI__Guru* 84285 points

Both the compiler and the RTS provide optimized memcpy code for pointers that are aligned and of sufficient size.

SuitJune Young over 10 years ago in reply to Archaeologist

Genius 3985 points

OK,I know that.Thank you.

Code Composer Studio™︎

Code Composer Studio forum

To do a more effiency memcpy in C674x DSP?