public inbox for gcc-bugs@sourceware.org help / color / mirror / Atom feed
From: "danglin at gcc dot gnu.org" <gcc-bugzilla@gcc.gnu.org> To: gcc-bugs@gcc.gnu.org Subject: [Bug target/53789] ICE in gen_reg_rtx, at emit-rtl.c:864/865 when compiling GNU MPFR on parisc Date: Sun, 23 Dec 2012 18:31:00 -0000 [thread overview] Message-ID: <bug-53789-4-d40Lojv7LJ@http.gcc.gnu.org/bugzilla/> (raw) In-Reply-To: <bug-53789-4@http.gcc.gnu.org/bugzilla/> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=53789 --- Comment #10 from John David Anglin <danglin at gcc dot gnu.org> 2012-12-23 18:31:30 UTC --- I implemented a reload pattern to try to do the reload for the following TLS symbol reference: "(symbol_ref:SI ("__gmpfr_cache_const_pi") [flags 0x60] <var_decl 0x405e55a0 __gmpfr_cache_const_pi>)" However, this approach fails here: /* ??? It would be useful to be able to handle only two, or more than three, operands, but for now we can only handle the case of having exactly three: output, input and one temp/scratch. */ gcc_assert (insn_data[(int) icode].n_operands == 3); (gdb) p insn_data[(int) icode].n_operands $3 = 6 '\006' I don't see any simple way to reduce the number of scratch registers needed for the reload. So, the only possible solution that I can see is to bump the cost for TLS symbol references to try to avoid the reload. The reload comes about because we have a pseudo register copy of the initial address load. It needs to be spilled rather than using the REG_EQUAL note.
next prev parent reply other threads:[~2012-12-23 18:31 UTC|newest] Thread overview: 18+ messages / expand[flat|nested] mbox.gz Atom feed top 2012-06-27 20:34 [Bug target/53789] New: " vincent-gcc at vinc17 dot net 2012-06-28 10:01 ` [Bug target/53789] " rguenth at gcc dot gnu.org 2012-07-05 14:56 ` vincent-gcc at vinc17 dot net 2012-07-05 15:12 ` redi at gcc dot gnu.org 2012-07-05 15:21 ` redi at gcc dot gnu.org 2012-07-07 0:30 ` vincent-gcc at vinc17 dot net 2012-12-23 3:15 ` danglin at gcc dot gnu.org 2012-12-23 3:58 ` danglin at gcc dot gnu.org 2012-12-23 5:07 ` danglin at gcc dot gnu.org 2012-12-23 15:16 ` danglin at gcc dot gnu.org 2012-12-23 18:31 ` danglin at gcc dot gnu.org [this message] 2012-12-25 17:59 ` danglin at gcc dot gnu.org 2012-12-25 21:05 ` danglin at gcc dot gnu.org 2012-12-26 17:29 ` danglin at gcc dot gnu.org 2012-12-26 19:47 ` danglin at gcc dot gnu.org 2013-01-04 3:56 ` danglin at gcc dot gnu.org 2013-01-04 4:04 ` danglin at gcc dot gnu.org 2013-01-05 16:56 ` danglin at gcc dot gnu.org
Reply instructions: You may reply publicly to this message via plain-text email using any one of the following methods: * Save the following mbox file, import it into your mail client, and reply-to-all from there: mbox Avoid top-posting and favor interleaved quoting: https://en.wikipedia.org/wiki/Posting_style#Interleaved_style * Reply using the --to, --cc, and --in-reply-to switches of git-send-email(1): git send-email \ --in-reply-to=bug-53789-4-d40Lojv7LJ@http.gcc.gnu.org/bugzilla/ \ --to=gcc-bugzilla@gcc.gnu.org \ --cc=gcc-bugs@gcc.gnu.org \ /path/to/YOUR_REPLY https://kernel.org/pub/software/scm/git/docs/git-send-email.html * If your mail client supports setting the In-Reply-To header via mailto: links, try the mailto: linkBe sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions for how to clone and mirror all data and code used for this inbox; as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).