Re: [PATCH v2] cputlb: Make store_helper less fragile to compiler optimizations

2020-07-28 Thread Alex Bennée
Richard Henderson writes: > On 7/24/20 12:51 PM, Shu-Chun Weng wrote: >> This change has no functional change. >> >> There is a potential link error with "undefined symbol: >> qemu_build_not_reached" due to how `store_helper` is structured. >> This does not produce at current QEMU head, but wa

Re: [PATCH v2] cputlb: Make store_helper less fragile to compiler optimizations

2020-07-27 Thread Shu-Chun Weng
That sounds like a much better solution -- I'm all for code size reduction. I still don't understand the function enough to feel confident about more invasive change thus the no-op change here. I'd be happy to see outlining unaligned access like you suggested. Shu-Chun On Mon, Jul 27, 2020 at 11

Re: [PATCH v2] cputlb: Make store_helper less fragile to compiler optimizations

2020-07-27 Thread Richard Henderson
On 7/24/20 12:51 PM, Shu-Chun Weng wrote: > This change has no functional change. > > There is a potential link error with "undefined symbol: > qemu_build_not_reached" due to how `store_helper` is structured. > This does not produce at current QEMU head, but was reproducible at > v4.2.0 with `clan

[PATCH v2] cputlb: Make store_helper less fragile to compiler optimizations

2020-07-24 Thread Shu-Chun Weng
This change has no functional change. There is a potential link error with "undefined symbol: qemu_build_not_reached" due to how `store_helper` is structured. This does not produce at current QEMU head, but was reproducible at v4.2.0 with `clang-10 -O2 -fexperimental-new-pass-manager`. The curren