[Bug rtl-optimization/101693] Terrible SIMD register allocation with a tight loop operating on 8 registers.

[ts.tomeksopel at gmail dot com via Gcc-bugs]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101693

--- Comment #1 from Tomasz Sobczyk <ts.tomeksopel at gmail dot com> ---
PS. when 

#define USE_VNNI

is commented out it exhibits similar behaviour to
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=80283