compress AST in memory, not just on disk #20933
Conversation
|
@nanosoldier |
base/inference.jl
Outdated
| @@ -3916,16 +3908,27 @@ function inlineable(f::ANY, ft::ANY, e::Expr, atypes::Vector{Any}, sv::Inference | |||
| rettype = linfo.rettype | |||
| end | |||
|
|
|||
| inferred === nothing && return invoke_NF(argexprs0, e.typ, atypes, sv, | |||
| atype_unlimited, invoke_data) | |||
There was a problem hiding this comment.
Kinda weird indentation here. atype_unlimited should probably be aligned beneath argexprs0 for consistency with the invoke_NF call a few lines down.
|
Looks like this did something weird in the inference code: |
|
Your benchmark job has completed - possible performance regressions were detected. A full report can be found here. cc @jrevels |
|
@nanosoldier runbenchmarks("dates", vs=":master") |
|
You missed some backticks. @nanosoldier |
|
Your benchmark job has completed - no performance regressions were detected. A full report can be found here. cc @jrevels |
|
Thanks. OK, good. That confirms that the "dates"-"parse" benchmark was just noise (as I was seeing locally also). |
vtjnash
force-pushed
the
jn/compress
branch
2 times, most recently
from
70583c4 to
3ee9df2
Compare
sysimg size will be exactly the same, but seems to be pretty significant savings in memory and initial GC time (about 10% on each)
|
|
||
| ios_putc('\0', s.s); |
sysimg size will be exactly the same, but seems to be pretty significant savings in memory and initial GC time (about 10% on each, estimated)