XMOS Static Timing Analyser - Designing With XMOS

[jonathan]

Hmm is it just me or does it give up on all while loops?

[andrew]

The XTA doesn't currently work out the loop iterations itself, it relies on the compiler to tell it the iterations on loops that the compiler knows. As all the tools improve this will improve.

[lilltroll]

I found the XTA very easy to work with so far.
I have rewritten the blocks from the start until the block with the black frame, in such a way that the branches almost takes the same time, reducing the worst time case.
I will do so for the rest as well, saving the total time for my TDFII BiQuad with "extended float"
:D :D :D

The XTA reports that: For the moment I'm able to run the IIR-filter @ 0.70 MHz, making it possible to do an IIR filter of 28:th order with HiEnd* Audio at 48 kHz in only one thread.

XTA.jpg

*Filter with a very low roundoff noise - much lower compared to an implementation with IEEE double.

[andrew]

You'll like the next release then. It's even better.

[shawn]

Slip this into the monitor ; )
Geeeze, this Xmos seems to spawn rich tools for software/hardware documentation.
This should be very usefull for the hardcore STL like analysis and debug, that I am
sure complements XC. I really hope this tool is fully realized, I think the software people
who get it, will be compelled to write beautifull code that really scales.

looking forward never straight,
Shawn

[lilltroll]

If you write code in XC - remember to compile it in debug mode before XTA.

That way it gets really nice in XTA, and it is very easy to follow your code :!:

OK I'm not used to see the Code from the CPUs perspective, so I often find cases where I can change the code in XC, (whitout the need of writing ASM by hand) so it runs faster. For example: If I do 3 lines of XC code in a different order, the CPU doesn't need to put data on the stack, instead the registers can be used directly.