On Thu, Mar 3, 2011 at 4:23 AM, Andrew Stubbs ams@codesourcery.com wrote:
Hi All,
Up until now, I have had no choice but to test toolchain correctness on A8 hardware. It made sense to use the same -mfpu settings as the Linaro/Ubuntu package builds use. This did not match the policy that the interesting platform was A9-NEON, but I didn't have that option.
That's changed now - our Panda boards have arrived! Yay! :)
But, it seems to me that if I change to using the Pandas for correctness testing (not performance testing) then I won't be testing what Ubuntu will actually use.
So what should I test on?
There's a list of the combinations I want to monitor at: https://wiki.linaro.org/WorkingGroups/ToolChain/Flyer
The A8 and A9 versions are built as part of the continuous build. I'm not there yet with the other combinations.
For the A8 and A9, I use the configuration files at: http://bazaar.launchpad.net/~linaro-toolchain-dev/cbuild/trunk/files/head:/c...
such as the current A9 r1 configuration at: http://bazaar.launchpad.net/~linaro-toolchain-dev/cbuild/trunk/view/head:/co...
which means that GCC is built with '--with-mode=thumb --with-arch=armv7-a --with-tune=cortex-a9 --with-float=softfp --with-fpu=neon'.
If there are no objections then I'd like to standardise on these two configurations. The A8 one is the same as what Ubuntu uses.
I'd rather not double my test load by testing on both, but that is an option
Test for the A9 with NEON first. Do the A8 if you have the time but don't worry too much as any significant failures will be picked up when the RC goes through the continuous build system.
-- Michael