RE: [avr-gcc-list] Trouble with division

[Dean Ferreyra]

Ulrich,

Larry has it right when he writes "relative times".

You can look at a version of the old source for my program here

  http://bourbonstreetsoftware.com/community/avr/types.c

to give you an idea of where these number came from.  For example, the test
for addition looks like this

  volatile uint8_t x_u8 = 67;
  volatile uint8_t y_u8 = 3;
  volatile uint8_t z_u8 = 0;
  //...
  TEST("x+y", z_u8 = x_u8 + y_u8);

where TEST is a macro that runs the second argument in a tight loop.  So the
"Cycles" number you see includes reading the bytes out of memory, adding
them, and assigning the result back into memory.

Dean

-----Original Message-----
From: address@hidden
[mailto:address@hidden Behalf Of Ulrich Schweitzer
Sent: Sunday, June 15, 2003 7:50 AM
To: avr-gcc List
Subject: Re: [avr-gcc-list] Trouble with division


On Sunday 15 June 2003 15:42, Larry Barello wrote:
> Here is an old email with relative times for various math operations in
> avr-gcc.

Thanks a lot, the list is very helpful.
I'm a bit confused though. Doesn't Atmel brag about "Up to x MIPS
throughput at x MHz" for all of their AVRs? How can this be when a simple
add operation takes 7 cycles?

Ulrich
--
PGP key ID: 0xDF6FC4FA


_______________________________________________
avr-gcc-list mailing list
address@hidden
http://www.avr1.org/mailman/listinfo/avr-gcc-list