void
CardTableModRefBS::
process_stride(Space* sp,
MemRegion used,
jint stride, int n_strides,
DirtyCardToOopClosure* dcto_cl,
MemRegionClosure* cl,
bool clear,
jbyte** lowest_non_clean,
uintptr_t lowest_non_clean_base_chunk_index,
size_t lowest_non_clean_chunk_size) {
// We don't have to go downwards here; it wouldn't help anyway,
// because of parallelism.
// Find the first card address of the first chunk in the stride that is
// at least "bottom" of the used region.
jbyte* start_card = byte_for(used.start());
jbyte* end_card = byte_after(used.last());
uintptr_t start_chunk = addr_to_chunk_index(used.start());
uintptr_t start_chunk_stride_num = start_chunk % n_strides;
jbyte* chunk_card_start;
if ((uintptr_t)stride >= start_chunk_stride_num) {
chunk_card_start = (jbyte*)(start_card +
(stride - start_chunk_stride_num) *
CardsPerStrideChunk);
} else {
// Go ahead to the next chunk group boundary, then to the requested stride.
chunk_card_start = (jbyte*)(start_card +
(n_strides - start_chunk_stride_num + stride) *
CardsPerStrideChunk);
}
while (chunk_card_start < end_card) {
// We don't have to go downwards here; it wouldn't help anyway,
// because of parallelism. (We take care with "min_done"; see below.)
// Invariant: chunk_mr should be fully contained within the "used" region.
jbyte* chunk_card_end = chunk_card_start + CardsPerStrideChunk;
MemRegion chunk_mr = MemRegion(addr_for(chunk_card_start),
chunk_card_end >= end_card ?
used.end() : addr_for(chunk_card_end));
assert(chunk_mr.word_size() > 0, "[chunk_card_start > used_end)");
assert(used.contains(chunk_mr), "chunk_mr should be subset of used");
// Process the chunk.
process_chunk_boundaries(sp,
dcto_cl,
chunk_mr,
used,
lowest_non_clean,
lowest_non_clean_base_chunk_index,
lowest_non_clean_chunk_size);
non_clean_card_iterate_work(chunk_mr, cl, clear);
// Find the next chunk of the stride.
chunk_card_start += CardsPerStrideChunk * n_strides;
}
}
void
CardTableModRefBS::
process_stride(Space* sp,
MemRegion used,
jint stride, int n_strides,
OopsInGenClosure* cl,
CardTableRS* ct,
jbyte** lowest_non_clean,
uintptr_t lowest_non_clean_base_chunk_index,
size_t lowest_non_clean_chunk_size) {
// We go from higher to lower addresses here; it wouldn't help that much
// because of the strided parallelism pattern used here.
// Find the first card address of the first chunk in the stride that is
// at least "bottom" of the used region.
jbyte* start_card = byte_for(used.start());
jbyte* end_card = byte_after(used.last());
uintptr_t start_chunk = addr_to_chunk_index(used.start());
uintptr_t start_chunk_stride_num = start_chunk % n_strides;
jbyte* chunk_card_start;
if ((uintptr_t)stride >= start_chunk_stride_num) {
chunk_card_start = (jbyte*)(start_card +
(stride - start_chunk_stride_num) *
ParGCCardsPerStrideChunk);
} else {
// Go ahead to the next chunk group boundary, then to the requested stride.
chunk_card_start = (jbyte*)(start_card +
(n_strides - start_chunk_stride_num + stride) *
ParGCCardsPerStrideChunk);
}
while (chunk_card_start < end_card) {
// Even though we go from lower to higher addresses below, the
// strided parallelism can interleave the actual processing of the
// dirty pages in various ways. For a specific chunk within this
// stride, we take care to avoid double scanning or missing a card
// by suitably initializing the "min_done" field in process_chunk_boundaries()
// below, together with the dirty region extension accomplished in
// DirtyCardToOopClosure::do_MemRegion().
jbyte* chunk_card_end = chunk_card_start + ParGCCardsPerStrideChunk;
// Invariant: chunk_mr should be fully contained within the "used" region.
MemRegion chunk_mr = MemRegion(addr_for(chunk_card_start),
chunk_card_end >= end_card ?
used.end() : addr_for(chunk_card_end));
assert(chunk_mr.word_size() > 0, "[chunk_card_start > used_end)");
assert(used.contains(chunk_mr), "chunk_mr should be subset of used");
DirtyCardToOopClosure* dcto_cl = sp->new_dcto_cl(cl, precision(),
cl->gen_boundary());
ClearNoncleanCardWrapper clear_cl(dcto_cl, ct);
// Process the chunk.
process_chunk_boundaries(sp,
dcto_cl,
chunk_mr,
used,
lowest_non_clean,
lowest_non_clean_base_chunk_index,
lowest_non_clean_chunk_size);
// We want the LNC array updates above in process_chunk_boundaries
// to be visible before any of the card table value changes as a
// result of the dirty card iteration below.
OrderAccess::storestore();
// We do not call the non_clean_card_iterate_serial() version because
// we want to clear the cards: clear_cl here does the work of finding
// contiguous dirty ranges of cards to process and clear.
clear_cl.do_MemRegion(chunk_mr);
// Find the next chunk of the stride.
chunk_card_start += ParGCCardsPerStrideChunk * n_strides;
}
}
