TMS320F28377D: C2000 ASSEMBLER HELP: Adding an Address in one instruction.

Pete,

Unfortunately there's no indirect addressing mode which will do this with a 32-bit move. I can think of a couple of things which may help.

1) Given that two instructions are required, you can eliminate the looping overhead by using the RPTB with, say, four instruction pairs. Something like:

RPTB LOOP01, #33
MOVL *XAR2, XAR6
ADDL XAR6, ACC
MOVL *XAR2, XAR6
ADDL XAR6, ACC
MOVL *XAR2, XAR6
ADDL XAR6, ACC
MOVL *XAR2, XAR6
ADDL XAR6, ACC

LOOP01:

Obviously this would implement a multiple of four moves, but you could follow the loop with 1 - 3 additional move pairs if you have a non-multiple of four to do. This would eliminate the branching overhead.

2) If you have a long sequence of moves to do, you could set up a DMA channel to do this. You'd just have to initialize and trigger the channel and all the overhead would be gone. I can't point you towards an example which does exactly what you want, but there is a DMA channel example in C2000Ware for this device which may help.

Regards,

Richard

Hi Richard, thanks for your help. I tried the RPTB instruction and my assembler says its out of block range 8...127 even tho I chose #33 for testing. So that didn't work. I have managed to speed up the code by replacing ADDL XAR6,ACC with ADDB XAR6,#1. Now its one cycle per instruction. This sped the code up dramatically but it's still not fast enough for what I want it to do. Oh, I'm feeding a VGA DAC directly from the output of a SRAM chip, thats why I need to change the SRAM address using the 22bit XAR6 register.

I'll check out the DMA example you mentioned and see how that goes. If not I'll have to research for faster chip, maybe a RISC V. 

Thanks mate

Pete

Hi Pete,

You need either 8 or 9 words, depending on alignment, between the RPTB instruction and the label at the end of the block.  In my code there were 8 instructions and it built fine for me.  If you had a different alignment you might have got this error, in which case you could add another pair of instructions and it should go away.  I still think this is worth a try because RPTB provides zero overhead looping so it should make a big difference.

Anyway, I'll go ahead and close this thread.  Feel free to open another one if you run into further issues or need anything else.  Thanks.

Regards,

Richard

Thank you Richard, I'll try it again. Hopefully it'll speed up the process.

Oh, while I've got you. Does TI have a faster MPU with 32bit IO ports? If the F728377D is not fast enough I'll have to use a faster MPU and I would like to use TI MPU's with our product.

Thanks again for all your help.

Pete