public inbox for gcc-cvs@sourceware.org help / color / mirror / Atom feed
From: Matthew Malcomson <matmal01@gcc.gnu.org> To: gcc-cvs@gcc.gnu.org Subject: [gcc(refs/vendors/ARM/heads/morello)] morello: Drop capability metadata when extracting a MEM_REF offset Date: Fri, 10 Dec 2021 16:49:18 +0000 (GMT) [thread overview] Message-ID: <20211210164918.7F82D385801D@sourceware.org> (raw) https://gcc.gnu.org/g:7a28f6ce56b5b02340f29c9862bfcda67f2f7e7a commit 7a28f6ce56b5b02340f29c9862bfcda67f2f7e7a Author: Stam Markianos-Wright <stam.markianos-wright@arm.com> Date: Tue Nov 2 11:38:15 2021 +0000 morello: Drop capability metadata when extracting a MEM_REF offset `mem_ref_offset` is used throughout the compiler to extract the constant offset second parameter to of a MEM_REF as a `poly_offset_int`. This is then always used as part of arithmetic offset calculations. The fact that `mem_ref_offset` returns a 128-bit constant capability value (with null metadata: all the upper bits are zero) sometimes causes an issue in the arithmetic if, e.g. the offset value is involved in a shift left, where the upper bits of the constant get shifted into the upper metadata bits instead of overflowing. Similar issues, but more rarely, can also happen with all other arithmetic that can cause such an overflow. This patch ensures that `mem_ref_offset` only returns the lower constant capability value bits, thus making the arithmetic safer. It also asserts that all MEM_REF second operands accessed through `mem_ref_offset` must have null metadata: any failure of this would indicate an arithmetic overflow into the metadata caused somewhere earlier in the compiler. It also fixes a minor sign-extension bug in the metadata extraction utility function. Diff: --- gcc/tree.c | 7 +++++-- gcc/tree.h | 5 ++++- 2 files changed, 9 insertions(+), 3 deletions(-) diff --git a/gcc/tree.c b/gcc/tree.c index 4b4f4fc3e61..982f198fd95 100644 --- a/gcc/tree.c +++ b/gcc/tree.c @@ -5041,8 +5041,11 @@ build_simple_mem_ref_loc (location_t loc, tree ptr) poly_offset_int mem_ref_offset (const_tree t) { - return poly_offset_int::from (wi::to_poly_wide (TREE_OPERAND (t, 1)), - SIGNED); + if (capability_type_p (TREE_TYPE (TREE_OPERAND (t, 1)))) + gcc_assert (tree_constant_capability_metadata (TREE_OPERAND (t, 1)) == 0); + return poly_offset_int::from (wi::to_poly_wide + (fold_drop_capability (TREE_OPERAND (t, 1))), + SIGNED); } /* Return an invariant ADDR_EXPR of type TYPE taking the address of BASE diff --git a/gcc/tree.h b/gcc/tree.h index 4a9e35361af..3519219927e 100644 --- a/gcc/tree.h +++ b/gcc/tree.h @@ -6373,7 +6373,10 @@ inline wide_int tree_constant_capability_metadata (const_tree t) { gcc_assert (TREE_CODE (t) == INTEGER_CST); - return wi::bit_and (wi::to_wide (t, TYPE_CAP_PRECISION (TREE_TYPE(t))), + return wi::bit_and (wide_int_storage::from + (wi::to_wide (t, TYPE_NONCAP_PRECISION (TREE_TYPE(t))), + TYPE_CAP_PRECISION (TREE_TYPE(t)), + UNSIGNED), wi::mask (TYPE_NONCAP_PRECISION (TREE_TYPE(t)), 1, TYPE_CAP_PRECISION (TREE_TYPE(t)))); }
reply other threads:[~2021-12-10 16:49 UTC|newest] Thread overview: [no followups] expand[flat|nested] mbox.gz Atom feed
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=20211210164918.7F82D385801D@sourceware.org \ --to=matmal01@gcc.gnu.org \ --cc=gcc-cvs@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).