public inbox for gcc-bugs@sourceware.org help / color / mirror / Atom feed
From: "zhaoweiliew at gmail dot com" <gcc-bugzilla@gcc.gnu.org> To: gcc-bugs@gcc.gnu.org Subject: [Bug tree-optimization/103855] Missed optimization: 64bit division used instead of 32bit division Date: Wed, 29 Dec 2021 03:16:24 +0000 [thread overview] Message-ID: <bug-103855-4-n6sGQZW5LV@http.gcc.gnu.org/bugzilla/> (raw) In-Reply-To: <bug-103855-4@http.gcc.gnu.org/bugzilla/> https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103855 --- Comment #1 from Zhao Wei Liew <zhaoweiliew at gmail dot com> --- After some research, I decided to look into value range propagation (VRP). I compiled with `-fdump-tree-vrp` and the VRP files contained the following for `optimized()`: For the file ending in .vrp1, ``` ;; Function unoptimized (_Z11unoptimizedjj, funcdef_no=1, decl_uid=2121, cgraph_uid=2, symbol_order=1) ;; 1 loops found ;; ;; Loop 0 ;; header 0, latch 1 ;; depth 0, outer -1 ;; nodes: 0 1 2 ;; 2 succs { 1 } Value ranges after VRP: _1: long long unsigned int [0, 4294967295] _2: long long unsigned int [0, 4294967295] a_3(D): unsigned int VARYING all_4: long long unsigned int [0, 4294967295] b_5(D): unsigned int VARYING _6: unsigned int VARYING unsigned int unoptimized (unsigned int a, unsigned int b) { long long unsigned int all; long long unsigned int _1; long long unsigned int _2; unsigned int _6; <bb 2> [local count: 1073741824]: all_4 = (long long unsigned int) a_3(D); _1 = (long long unsigned int) b_5(D); _2 = all_4 / _1; _6 = (unsigned int) _2; return _6; } ``` and for the file ending in .vrp2, ``` ;; Function unoptimized (_Z11unoptimizedjj, funcdef_no=1, decl_uid=2121, cgraph_uid=2, symbol_order=1) ;; 1 loops found ;; ;; Loop 0 ;; header 0, latch 1 ;; depth 0, outer -1 ;; nodes: 0 1 2 ;; 2 succs { 1 } Exported global range table: ============================ _1 : long long unsigned int [0, 4294967295] _2 : long long unsigned int [0, 4294967295] all_4 : long long unsigned int [0, 4294967295] unsigned int unoptimized (unsigned int a, unsigned int b) { long long unsigned int all; long long unsigned int _1; long long unsigned int _2; unsigned int _6; <bb 2> [local count: 1073741824]: all_4 = (long long unsigned int) a_3(D); _1 = (long long unsigned int) b_5(D); _2 = all_4 / _1; _6 = (unsigned int) _2; return _6; } ``` If I'm interpreting the above correctly, the optimizer seems to think all_4 can be in the range [0, 4294967295], which is correct. Hence, I suppose it's safe to conclude that VRP isn't the issue here? Please correct me if I'm wrong.
next prev parent reply other threads:[~2021-12-29 3:16 UTC|newest] Thread overview: 8+ messages / expand[flat|nested] mbox.gz Atom feed top 2021-12-29 3:00 [Bug tree-optimization/103855] New: " zhaoweiliew at gmail dot com 2021-12-29 3:16 ` zhaoweiliew at gmail dot com [this message] 2021-12-29 3:34 ` [Bug tree-optimization/103855] " pinskia at gcc dot gnu.org 2021-12-29 3:46 ` pinskia at gcc dot gnu.org 2021-12-29 3:53 ` pinskia at gcc dot gnu.org 2021-12-29 8:21 ` zhaoweiliew at gmail dot com 2021-12-29 8:32 ` zhaoweiliew at gmail dot com 2021-12-29 9:41 ` jakub 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-103855-4-n6sGQZW5LV@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).