This is the mail archive of the
mailing list for the GCC project.
Re: [GSoC] generation of GCC expression trees from isl ast expressions
- From: Tobias Grosser <tobias at grosser dot es>
- To: Roman Gareev <gareevroman at gmail dot com>
- Cc: Mircea Namolaru <mircea dot namolaru at inria dot fr>, gcc at gcc dot gnu dot org
- Date: Fri, 27 Jun 2014 07:39:24 +0200
- Subject: Re: [GSoC] generation of GCC expression trees from isl ast expressions
- Authentication-results: sourceware.org; auth=none
- References: <CABGF_gdTc0FmYg-f9RyavU8a_bP6jAj5Y3u1SvCt0YZhU7=QyQ at mail dot gmail dot com> <53A7D77D dot 2050205 at grosser dot es> <CABGF_gcY9s-nZYXm-O8FGjB2XBuU9Vx45BAnuS7jC8894Jn0eg at mail dot gmail dot com> <CABGF_gehEe-GYROuN1nJoqU87iNQd5roaAZ3zVf5xCaxa4qLbw at mail dot gmail dot com> <53AB3478 dot 3060209 at grosser dot es> <CABGF_ge6XHTJ5dcuvxw5b7LNVziq7eMojzgxF81HW-bT3PVC6A at mail dot gmail dot com>
On 27/06/2014 07:31, Roman Gareev wrote:
Are you saying we should better not do unit testing at the moment? (This is
perfectly fine with me, I am just verifying what you said)
Yes, I think we should better not to do it. It seems that unit-testing
isn't supported in gcc.
If we don't have a convenient way to do unit-testing, we need to do testing
for larger programs (at least empty loop bodies). This is what we did
previously and even though not ideal should work as well.
I agree that generation of loops with empty bodies is required
functionality for testing (I wrote about this in previous message.
Sorry for being unclear.)
I did not fully understand them. Are you comparing the resulting GIMPLE
trees or are you statically evaluating the isl and gimple trees to check
if the computed value is the same?
I'm statically evaluating gimple tree expressions, which are generated
from isl ast expressions. After this the result of the evaluation is
being compared with assumed result. I haven't found functions, which
evaluate isl ast expressions. That is why it compares results of
evaluations with assumed result. Maybe there is a better way to
compare the semantics of the GCC and isl expressions.
Just leave it out for now.
I originally hoped we could just ast generate an isl_ast_expr and then
just compare if the resulting gimple code is identical to a
hand-written/provided one. (Just comparing the individual instructions).