The following one line program gives an ICE typedef float vsf __attribute__((vector_size(2048))); x.c:1: internal compiler error: in round_up, at fold-const.c:11646 I reproduced this on IA64 and HPPA platforms and I do not think it is platform specific.
Hmm, it works just fine on powerpc-darwin (I don't know why) but it ICEs on i686-pc-linux-gnu. And it worked just fine with "gcc version 3.5.0 20040909 (experimental)" I don't know if I can consider this a regression though, before 4.0.0 we gave an error: t.c:1: error: no vector mode with the size and type specified could be found
Subject: Bug 21969 CVSROOT: /cvs/gcc Module name: gcc Changes by: sje@gcc.gnu.org 2005-06-29 15:34:58 Modified files: gcc : ChangeLog tree.c tree.h tree-vect-transform.c Log message: PR middle-end/21969 * tree.h (TYPE_VECTOR_SUBPARTS): Change to shift expression. (SET_TYPE_VECTOR_SUBPARTS): New. * tree.c (make_vector_type): Replace TYPE_VECTOR_SUBPARTS with SET_TYPE_VECTOR_SUBPARTS. * tree-vect-transform.c (vect_transform_loop): Add cast. Patches: http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/ChangeLog.diff?cvsroot=gcc&r1=2.9275&r2=2.9276 http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/tree.c.diff?cvsroot=gcc&r1=1.489&r2=1.490 http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/tree.h.diff?cvsroot=gcc&r1=1.741&r2=1.742 http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/tree-vect-transform.c.diff?cvsroot=gcc&r1=2.33&r2=2.34
Subject: Bug 21969 CVSROOT: /cvs/gcc Module name: gcc Changes by: sje@gcc.gnu.org 2005-06-29 15:38:34 Modified files: gcc/testsuite : ChangeLog Added files: gcc/testsuite/gcc.dg/vect: pr21969.c Log message: PR middle-end/21969 * gcc.dg/vect/pr21969.c: New test. Patches: http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/ChangeLog.diff?cvsroot=gcc&r1=1.5703&r2=1.5704 http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/gcc.dg/vect/pr21969.c.diff?cvsroot=gcc&r1=NONE&r2=1.1
Subject: Bug 21969 CVSROOT: /cvs/gcc Module name: gcc Changes by: sje@gcc.gnu.org 2005-06-29 16:15:41 Modified files: gcc/testsuite : ChangeLog gcc/testsuite/gcc.dg/torture: pr21817-1.c Log message: PR testsuite/21969 * gcc.dg/torture/pr21817.c: Use -fno-common for hppa*-*-* targets. Patches: http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/ChangeLog.diff?cvsroot=gcc&r1=1.5704&r2=1.5705 http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/gcc.dg/torture/pr21817-1.c.diff?cvsroot=gcc&r1=1.2&r2=1.3
Ignore the second patch (it was commited with the wrong PR number in the comments). But the first patch does fix the problem.
Subject: Bug 21969 CVSROOT: /cvs/gcc Module name: gcc Branch: gcc-4_0-branch Changes by: sje@gcc.gnu.org 2005-07-19 20:28:00 Modified files: gcc/testsuite : ChangeLog gcc/testsuite/gcc.dg/torture: pr21817-1.c Log message: Backport from mainline. PR testsuite/21969 * gcc.dg/torture/pr21817.c: Use -fno-common for hppa*-*-* targets. Patches: http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/ChangeLog.diff?cvsroot=gcc&only_with_tag=gcc-4_0-branch&r1=1.5084.2.289&r2=1.5084.2.290 http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/gcc.dg/torture/pr21817-1.c.diff?cvsroot=gcc&only_with_tag=gcc-4_0-branch&r1=1.1.2.1&r2=1.1.2.2