https://gcc.gnu.org/bugzilla/show_bug.cgi?id=90765
Bug ID: 90765
Summary: preferred_stack_boundary is updated for callee
Product: gcc
Version: 10.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: rtl-optimization
Assignee: unassigned at gcc dot gnu.org
Reporter: hjl.tools at gmail dot com
CC: skpgkp1 at gmail dot com
Target Milestone: ---
locate_and_pad_parm is called when expanding function call from
initialize_argument_information and when generating function body
from assign_parm_find_entry_rtl:
/* Remember if the outgoing parameter requires extra alignment on the
calling function side. */
if (crtl->stack_alignment_needed < boundary)
crtl->stack_alignment_needed = boundary;
if (crtl->preferred_stack_boundary < boundary)
crtl->preferred_stack_boundary = boundary;
preferred_stack_boundary should be updated only when expanding function
call, not when generating function body. In this testcase,
crtl->preferred_stack_boundary is to set 512 when not needed:
[hjl@gnu-cfl-1 lea-3]$ cat 1.i
typedef int __v16si __attribute__ ((__vector_size__ (64)));
void
foo (__v16si x, int i0, int i1, int i2, int i3, int i4, int i5, __v16si *p)
{
*p = x;
}
[hjl@gnu-cfl-1 lea-3]$ make 1.s
/export/build/gnu/tools-build/gcc-debug/build-x86_64-linux/gcc/xgcc
-B/export/build/gnu/tools-build/gcc-debug/build-x86_64-linux/gcc/ -mavx512f -O2
-Wa,-mx86-used-note=yes -S 1.i
[hjl@gnu-cfl-1 lea-3]$ cat 1.s
.file "1.i"
.text
.p2align 4
.globl foo
.type foo, @function
foo:
.LFB0:
.cfi_startproc
pushq %rbp
.cfi_def_cfa_offset 16
.cfi_offset 6, -16
movq %rsp, %rbp
.cfi_def_cfa_register 6
andq $-64, %rsp
movq 16(%rbp), %rax
vmovdqa64 %zmm0, (%rax)
leave
.cfi_def_cfa 7, 8
ret
.cfi_endproc
.LFE0:
.size foo, .-foo
.ident "GCC: (GNU) 10.0.0 20190605 (experimental)"
.section .note.GNU-stack,"",@progbits
[hjl@gnu-cfl-1 lea-3]$
