@ezrosent and I have verified that this change does not adversely affect performance. Unfortunately, getting to that state required the addition of the somewhat-unergonomic macro alloc_tls_fast_with - calling the with method directly significantly degraded performance. This probably reflects a missed optimization opportunity that could be reported to the rustc folks.
@ezrosent and I have verified that this change does not adversely affect performance. Unfortunately, getting to that state required the addition of the somewhat-unergonomic macro
alloc_tls_fast_with- calling thewithmethod directly significantly degraded performance. This probably reflects a missed optimization opportunity that could be reported to therustcfolks.