This is the mail archive of the
mailing list for the GCC project.
Re: [0/6] nvptx testsuite patches
- From: Jeff Law <law at redhat dot com>
- To: Bernd Schmidt <bernds at codesourcery dot com>, GCC Patches <gcc-patches at gcc dot gnu dot org>
- Date: Tue, 21 Oct 2014 14:46:43 +0000
- Subject: Re: [0/6] nvptx testsuite patches
- Authentication-results: sourceware.org; auth=none
- References: <5446694A dot 1010705 at codesourcery dot com>
On 10/21/14 14:10, Bernd Schmidt wrote:
Yea. When I first looked at PTX, my thought was to use the existing
testsuite, to the extent possible, to shake out the initial code
generation issues. There's just some things that would require heroic
effort to make work and they aren't really a priority for PTX.
This series modifies a large number of tests in order to clean up
testsuite results on nvptx. The goal here was never really to get an
entirely clean run - the target is just too different from conventional
ones - but to be able to test the compiler sufficiently to be sure that
it's in good shape for use in offloading. Most of the patches here add
annotations for use of features like alloca or indirect jumps that are
unsupported on the target.
Examples of things that still cause failures are things like dots in
identifiers, use of constructors (which is something I want to look
into), certain constructs that trigger bugs in the ptxas tool, and lots
of undefined C library functions.
So I've got not problem conceptually with the direction this work is taking.