public inbox for gcc-bugs@sourceware.org
help / color / mirror / Atom feed
* [Bug middle-end/114824] New: gcc doesn't respect __attribute__((zero_call_used_regs("used"))) when performing a tail call optimization
@ 2024-04-23 10:24 mikulas at artax dot karlin.mff.cuni.cz
  0 siblings, 0 replies; only message in thread
From: mikulas at artax dot karlin.mff.cuni.cz @ 2024-04-23 10:24 UTC (permalink / raw)
  To: gcc-bugs

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114824

            Bug ID: 114824
           Summary: gcc doesn't respect
                    __attribute__((zero_call_used_regs("used"))) when
                    performing a tail call optimization
           Product: gcc
           Version: 14.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: middle-end
          Assignee: unassigned at gcc dot gnu.org
          Reporter: mikulas at artax dot karlin.mff.cuni.cz
  Target Milestone: ---

If a function has the attribute __attribute__((zero_call_used_regs("used")))
and the function ends with a sibling call, the used registers are not zeroed.

See this function:
#include <stdio.h>
__attribute__((zero_call_used_regs("used"))) void f(void)
{
        __asm__ volatile("" ::: "x0", "x1", "x2", "x3", "x4", "x5", "x6",
"x7");
        printf("Hello World!\n");
}
Compile it with "gcc -S -o - -O2 call.c -Wall -Wextra" - the result is:
f:
        adrp    x0, .LC0
        add     x0, x0, :lo12:.LC0
        b       puts

If we disable the tail call optimization, the registers x0 - x7 are cleared
properly:
gcc -S -o - -O2 call.c -Wall -Wextra -fno-optimize-sibling-calls
f:
        stp     x29, x30, [sp, -16]!
        mov     x29, sp
        adrp    x0, .LC0
        add     x0, x0, :lo12:.LC0
        bl      puts
        ldp     x29, x30, [sp], 16
        mov     x0, 0
        mov     x1, 0
        mov     x2, 0
        mov     x3, 0
        mov     x4, 0
        mov     x5, 0
        mov     x6, 0
        mov     x7, 0
        mov     x16, 0
        mov     x17, 0
        ret

I tested it on arm64 and x86-64 and both of them suffer from this problem.

^ permalink raw reply	[flat|nested] only message in thread

only message in thread, other threads:[~2024-04-23 10:24 UTC | newest]

Thread overview: (only message) (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2024-04-23 10:24 [Bug middle-end/114824] New: gcc doesn't respect __attribute__((zero_call_used_regs("used"))) when performing a tail call optimization mikulas at artax dot karlin.mff.cuni.cz

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).