| commit | 0102263645b73422d84dbe7fac3bdee65c7447a7 | |
| author | Vlad Brezae <brezaevlad@gmail.com> | |
| Wed, 16 Oct 2019 20:01:54 +0000 (16 23:01 +0300) | ||
| committer | Vlad Brezae <brezaevlad@gmail.com> | |
| Thu, 17 Oct 2019 09:53:19 +0000 (17 12:53 +0300) | ||
| tree | f08e1d473c0a0143f1ab08e5959d3175896c345a | treesnapshot (tar.gz zip) |
| parent | 4f8a82c9fdebd7e9dcc5392a8b3a0fd129789bc0 | commitdiff |
[interp] Avoid using full barriers for volatile operations
This commit avoid unnecessary memory barriers on x86 and amd64.
Volatile writes have release semantics, volatile reads have acquire semantics. Organize the code a little bit to indicate this. On x86 and amd64 we don't require barrier for acquire and release semantics since they are provided by the arch, so we can skip them. Ideally we should have 3 memory barrier interp instructions for each type, but, currently, they would all seem to call mono_memory_barrier so it is useless at this point.
This commit avoid unnecessary memory barriers on x86 and amd64.
Volatile writes have release semantics, volatile reads have acquire semantics. Organize the code a little bit to indicate this. On x86 and amd64 we don't require barrier for acquire and release semantics since they are provided by the arch, so we can skip them. Ideally we should have 3 memory barrier interp instructions for each type, but, currently, they would all seem to call mono_memory_barrier so it is useless at this point.
| mono/mini/interp/transform.c | diffblobblamehistory |