pack more fields in datatype #41018
pack more fields in datatype #41018
Conversation
|
@nanosoldier Just curious about the "damage". |
|
Your package evaluation job has completed - possible new issues were detected. A full report can be found here. cc @maleadt |
|
Looks like a couple "key" packages (SymbolServer, AxisArrays, JuliaInterpreter), but not a major impact. |
| function datatype_min_ninitialized(t::DataType) | ||
| isabstracttype(t) && return 0 | ||
| if t.name === NamedTuple_typename | ||
| names, types = t.parameters[1], t.parameters[2] |
There was a problem hiding this comment.
We could just call fieldcount in this case.
There was a problem hiding this comment.
Hm, I guess not, since we want 0 instead of an error for abstract types...
There was a problem hiding this comment.
yes, that is why it is reimplemented here
| STATIC_INLINE jl_sym_t *jl_field_name(jl_datatype_t *st, size_t i) JL_NOTSAFEPOINT | ||
| { | ||
| return (jl_sym_t*)jl_svecref(jl_field_names(st), i); | ||
| return st->name->names; |
There was a problem hiding this comment.
Maybe assert here that it's not a named tuple?
There was a problem hiding this comment.
I figured it would assert due to the null pointer anyways, so it was unnecessary to enforce it (edit: and it is also applicable to Tuple)
As Jeff noted recently, more fields can be moved to reduce the data size (I think removing
.nameshad the biggest effect on sysimg size, but all seemed helpful). I've also packed some Bool fields too.