Only add big objarray to remset once

[gbaraldi]

Fixes #49205

[d-netto]

What's the performance impact of this PR on #49205?

[gbaraldi]

Makes it faster than on 1.8

[d-netto]

The logic for decide whether we need to chunk or push the array into the remset could be a bit cleaner. I'd refactor:

size_t too_big = (obj_end - obj_begin) / MAX_REFS_AT_ONCE > step; // use this order of operations to avoid idiv
jl_value_t **scan_end = obj_end;
if (too_big) {
    scan_end = obj_begin + step * MAX_REFS_AT_ONCE;
    if ((nptr & 0x3) != 0x2) {
        jl_gc_chunk_t c = {GC_objary_chunk, obj_parent, scan_end,
                           obj_end,      NULL,       NULL,
                           step,         nptr};
        gc_chunkqueue_push(mq, &c);
        obj_end = scan_end;
    }
}
for (; obj_begin < scan_end; obj_begin += step) {
    new_obj = *obj_begin;
    if (new_obj != NULL) {
        verify_parent2("obj array", obj_parent, obj_begin, "elem(%d)",
                       gc_slot_to_arrayidx(obj_parent, obj_begin));
        gc_try_claim_and_push(mq, new_obj, &nptr);
        gc_heap_snapshot_record_array_edge(obj_parent, &new_obj);
    }
}
if (too_big) {
    if (obj_end != scan_end) {
        jl_gc_chunk_t c = {GC_objary_chunk, obj_parent, scan_end,
                           obj_end,      NULL,       NULL,
                           step,         nptr};
        gc_chunkqueue_push(mq, &c);
    }
}
else {
    gc_mark_push_remset(ptls, obj_parent, nptr);
}

into:

size_t too_big = (obj_end - obj_begin) / MAX_REFS_AT_ONCE > step; // use this order of operations to avoid idiv
jl_value_t **scan_end = obj_end;
if (too_big) {
    scan_end = obj_begin + step * MAX_REFS_AT_ONCE;
}
for (; obj_begin < scan_end; obj_begin += step) {
    new_obj = *obj_begin;
    if (new_obj != NULL) {
        verify_parent2("obj array", obj_parent, obj_begin, "elem(%d)",
                       gc_slot_to_arrayidx(obj_parent, obj_begin));
        gc_try_claim_and_push(mq, new_obj, &nptr);
        gc_heap_snapshot_record_array_edge(obj_parent, &new_obj);
    }
}
if (too_big) {
    jl_gc_chunk_t c = {GC_objary_chunk, obj_parent, scan_end,
                        obj_end,      NULL,       NULL,
                        step,         nptr};
    gc_chunkqueue_push(mq, &c);
}
else {
    gc_mark_push_remset(ptls, obj_parent, nptr);
}

Same change for gc_mark_array8/16.

[vtjnash]

As long as you don't care that your change eliminates the possibility of intra-array thread-stealing and thus slightly reduces parallelism of threaded GC, yeah, that rearrangement is fine.

[d-netto]

slightly reduces parallelism of threaded GC

This should be fine. If an array is large enough to be chunked then it will expose a lot of jl_value_t * that could potentially be stolen.

[d-netto]

Also, in #48600 we prioritize stealing pointers before chunks.

[d-netto]

Replace NULL, NULL here with elem_begin, elem_end. Same for array16.