TMS320C6743

Sebastien Philippe-Auguste said:

1) I want to confirm that the TMS320C6743 can (albeit

inefficiently) execute code from Flash.  My customer suspects that their software may exceed the size of the internal memory (by a bit) but would like to avoid external dynamic memory.  Our thought is that if they can run from flash then the rarely-used instructions can be read from there into the cache should they need to execute them. Is that possible?

This is certainly possible, however you can only execute out of generic asynchronous memories like NOR flash, it is not possible to execute directly out of NAND flash. Also note that as you suggest this would be low on performance, and also that there is limited space for NOR memories due to the number of address lines available (15 so 32k EMIF words per chip enable).

Sebastien Philippe-Auguste said:

2) Can we stash into the TMS320C6743 L1 cache?  If not how could we implement 1) above?

I need some ideas to do this: Certain microcontrollers will allow you to run the code in overlay; so automatically code can be swapped in and out to make space for excess code to run

I am not familiar with the operation of 'stash' in regards to the cache, but the L1P cache does allow you to "freeze" the cache, which means you can put it in a state where it will not evict lines or allocate new lines, though other statistics are still collected. This freeze functionality is designed for such cases as you describe, where you want to keep the contents of L1P in tact when executing some other code that you plan on running only once, it is discussed further in the cache user's guide at http://focus.ti.com/lit/ug/sprug82a/sprug82a.pdf.