public inbox for gcc-bugs@sourceware.org help / color / mirror / Atom feed
From: "rguenth at gcc dot gnu.org" <gcc-bugzilla@gcc.gnu.org> To: gcc-bugs@gcc.gnu.org Subject: [Bug target/107257] [13 Regression] Wrong code at -O2 on x86_64-linux-gnu since r13-857-gf1652e3343b1ec47 Date: Mon, 17 Oct 2022 11:09:55 +0000 [thread overview] Message-ID: <bug-107257-4-eqEiQqEBMe@http.gcc.gnu.org/bugzilla/> (raw) In-Reply-To: <bug-107257-4@http.gcc.gnu.org/bugzilla/> https://gcc.gnu.org/bugzilla/show_bug.cgi?id=107257 --- Comment #6 from Richard Biener <rguenth at gcc dot gnu.org> --- With -fno-tree-slp-vectorize the failure goes away. diff in optimized: __attribute__((noipa, noinline, noclone, no_icf)) void q () { - <bb 2> [local count: 1073742492]: - MEM <unsigned long> [(int *)&d] = 18380021091030725580; + int b_lsm.15; + int f; + _Bool _5; + + <bb 2> [local count: 1073741824]: + MEM <vector(2) int> [(int *)&d] = { -15535156, -15535156 }; d[2] = -15535156; + + <bb 3> [local count: 8687551919]: + # f_30 = PHI <f_25(5), 0(2)> + if (f_30 == 0) + goto <bb 5>; [1.43%] + else + goto <bb 4>; [98.57%] + + <bb 4> [local count: 4698759900]: + if (f_30 != 2) + goto <bb 5>; [79.66%] + else + goto <bb 6>; [20.34%] + + <bb 5> [local count: 7731921808]: + # f_25 = PHI <2(4), 1(3)> + goto <bb 3>; [100.00%] + + <bb 6> [local count: 1073742492]: a = 2; - b = -1; + _5 = f_30 == 0; + b_lsm.15_33 = _5 ? 2 : -1; + b = b_lsm.15_33; return; } both variants look OK so instead RTL expansion looks fishy(?) ;; b = b_lsm.15_33; (insn 28 26 29 (parallel [ (set (reg:CCC 17 flags) (ne:CCC (reg/v:SI 83 [ f ]) (const_int 0 [0]))) (set (reg:SI 89) (neg:SI (reg/v:SI 83 [ f ]))) ]) -1 (nil)) (insn 29 28 30 (parallel [ (set (reg:SI 90) (neg:SI (ltu:SI (reg:CCC 17 flags) (const_int 0 [0])))) (clobber (reg:CC 17 flags)) ]) -1 (nil)) (insn 30 29 31 (parallel [ (set (reg:SI 90) (ior:SI (reg:SI 90) (const_int 2 [0x2]))) (clobber (reg:CC 17 flags)) ]) -1 (nil)) (insn 31 30 32 (set (reg:SI 88 [ b_lsm.15 ]) (reg:SI 90)) -1 (nil)) (insn 32 31 0 (set (mem/c:SI (symbol_ref:DI ("b") [flags 0x2] <var_decl 0x7ffff6526d80 b>) [1 b+0 S4 A32]) (reg:SI 88 [ b_lsm.15 ])) -1 (nil)) and we eventually optimize this to q: .LFB2: .cfi_startproc movq .LC0(%rip), %rax movl $-15535156, d+8(%rip) movl $2, a(%rip) movq %rax, d(%rip) movl $2, b(%rip) ret
next prev parent reply other threads:[~2022-10-17 11:09 UTC|newest] Thread overview: 11+ messages / expand[flat|nested] mbox.gz Atom feed top 2022-10-14 8:32 [Bug c/107257] New: Wrong code at -O2 on x86_64-linux-gnu shaohua.li at inf dot ethz.ch 2022-10-14 9:27 ` [Bug c/107257] " rguenth at gcc dot gnu.org 2022-10-14 10:00 ` marxin at gcc dot gnu.org 2022-10-14 10:03 ` marxin at gcc dot gnu.org 2022-10-14 12:52 ` rguenth at gcc dot gnu.org 2022-10-17 10:27 ` shaohua.li at inf dot ethz.ch 2022-10-17 11:05 ` [Bug target/107257] [13 Regression] Wrong code at -O2 on x86_64-linux-gnu since r13-857-gf1652e3343b1ec47 marxin at gcc dot gnu.org 2022-10-17 11:09 ` rguenth at gcc dot gnu.org [this message] 2022-10-20 17:25 ` pinskia at gcc dot gnu.org 2023-07-27 8:01 ` shaohua.li at inf dot ethz.ch 2023-07-27 14:47 ` shaohua.li at inf dot ethz.ch
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-107257-4-eqEiQqEBMe@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).