https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112510
--- Comment #6 from Vladimir Sadovnikov <sadko4u at gmail dot com> --- Here is disassembly of `test.cpp` function for GCC 7.5.0: ``` 0000000000400727 <_Z13gate_x1_curvePfPKfPKN3dsp11gate_knee_tEm>: 400727: 4c 8d 54 24 08 lea 0x8(%rsp),%r10 40072c: 48 83 e4 c0 and $0xffffffffffffffc0,%rsp 400730: 41 ff 72 f8 push -0x8(%r10) 400734: 55 push %rbp 400735: 48 89 e5 mov %rsp,%rbp 400738: 41 57 push %r15 40073a: 41 56 push %r14 40073c: 41 55 push %r13 40073e: 41 54 push %r12 400740: 41 52 push %r10 400742: 53 push %rbx 400743: 48 81 ec 80 03 00 00 sub $0x380,%rsp 40074a: 49 89 fc mov %rdi,%r12 40074d: 49 89 f5 mov %rsi,%r13 400750: 49 89 d7 mov %rdx,%r15 400753: 49 89 ce mov %rcx,%r14 400756: 48 8d 9d 50 fc ff ff lea -0x3b0(%rbp),%rbx 40075d: 83 3d 1c 19 00 00 00 cmpl $0x0,0x191c(%rip) # 402080 <__asan_option_detect_stack_use_after_return@@Base> 400764: 0f 85 32 01 00 00 jne 40089c <_Z13gate_x1_curvePfPKfPKN3dsp11gate_knee_tEm+0x175> 40076a: 48 c7 03 b3 8a b5 41 movq $0x41b58ab3,(%rbx) 400771: 48 c7 43 08 f8 0c 40 movq $0x400cf8,0x8(%rbx) 400778: 00 400779: 48 c7 43 10 27 07 40 movq $0x400727,0x10(%rbx) 400780: 00 400781: 48 89 df mov %rbx,%rdi 400784: 48 c1 ef 03 shr $0x3,%rdi 400788: c7 87 00 80 ff 7f f1 movl $0xf1f1f1f1,0x7fff8000(%rdi) 40078f: f1 f1 f1 400792: c7 87 04 80 ff 7f f1 movl $0xf1f1f1f1,0x7fff8004(%rdi) 400799: f1 f1 f1 40079c: c7 87 0c 80 ff 7f f2 movl $0xf2f2f2f2,0x7fff800c(%rdi) 4007a3: f2 f2 f2 4007a6: c7 87 20 80 ff 7f f2 movl $0xf2f2f2f2,0x7fff8020(%rdi) 4007ad: f2 f2 f2 4007b0: c7 87 24 80 ff 7f f2 movl $0xf2f2f2f2,0x7fff8024(%rdi) 4007b7: f2 f2 f2 4007ba: c7 87 68 80 ff 7f f3 movl $0xf3f3f3f3,0x7fff8068(%rdi) 4007c1: f3 f3 f3 4007c4: c7 87 6c 80 ff 7f f3 movl $0xf3f3f3f3,0x7fff806c(%rdi) 4007cb: f3 f3 f3 4007ce: 62 d2 7d 48 18 07 vbroadcastss (%r15),%zmm0 4007d4: 62 d2 7d 48 18 4f 01 vbroadcastss 0x4(%r15),%zmm1 4007db: 62 d2 7d 48 18 57 02 vbroadcastss 0x8(%r15),%zmm2 4007e2: 62 d2 7d 48 18 5f 03 vbroadcastss 0xc(%r15),%zmm3 4007e9: 62 d2 7d 48 18 67 04 vbroadcastss 0x10(%r15),%zmm4 4007f0: 62 d2 7d 48 18 6f 05 vbroadcastss 0x14(%r15),%zmm5 4007f7: 62 d2 7d 48 18 77 06 vbroadcastss 0x18(%r15),%zmm6 4007fe: 62 d2 7d 48 18 7f 07 vbroadcastss 0x1c(%r15),%zmm7 400805: 62 f1 7c 48 29 43 05 vmovaps %zmm0,0x140(%rbx) 40080c: 62 f1 7c 48 29 4b 06 vmovaps %zmm1,0x180(%rbx) 400813: 62 f1 7c 48 29 53 07 vmovaps %zmm2,0x1c0(%rbx) 40081a: 62 f1 7c 48 29 5b 08 vmovaps %zmm3,0x200(%rbx) 400821: 62 f1 7c 48 29 63 09 vmovaps %zmm4,0x240(%rbx) 400828: 62 f1 7c 48 29 6b 0a vmovaps %zmm5,0x280(%rbx) 40082f: 62 f1 7c 48 29 73 0b vmovaps %zmm6,0x2c0(%rbx) 400836: 62 f1 7c 48 29 7b 0c vmovaps %zmm7,0x300(%rbx) ``` gcc --version gcc (SUSE Linux) 7.5.0 Copyright (C) 2017 Free Software Foundation, Inc. This is free software; see the source for copying conditions. There is NO warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. As we see, there is no load of %rbx from stack and all offsets are multiple of 0x40.
