/* Wrap a function for processing a single block to process an entire buffer in
* parallel. */
int64_t bshuf_blocked_wrap_fun(bshufBlockFunDef fun, void* in, void* out,
const size_t size, const size_t elem_size, size_t block_size) {
size_t ii;
ioc_chain C;
ioc_init(&C, in, out);
int64_t err = 0, count, cum_count = 0;
size_t last_block_size;
if (block_size == 0) {
block_size = bshuf_default_block_size(elem_size);
}
if (block_size < 0 || block_size % BSHUF_BLOCKED_MULT) return -81;
#if defined(_OPENMP)
#pragma omp parallel for schedule(dynamic, 1) \
private(count) reduction(+ : cum_count)
#endif
for (ii = 0; ii < size / block_size; ii ++) {
count = fun(&C, block_size, elem_size);
if (count < 0) err = count;
cum_count += count;
}
last_block_size = size % block_size;
last_block_size = last_block_size - last_block_size % BSHUF_BLOCKED_MULT;
if (last_block_size) {
count = fun(&C, last_block_size, elem_size);
if (count < 0) err = count;
cum_count += count;
}
if (err < 0) return err;
size_t leftover_bytes = size % BSHUF_BLOCKED_MULT * elem_size;
size_t this_iter;
char *last_in = (char *) ioc_get_in(&C, &this_iter);
ioc_set_next_in(&C, &this_iter, (void *) (last_in + leftover_bytes));
char *last_out = (char *) ioc_get_out(&C, &this_iter);
ioc_set_next_out(&C, &this_iter, (void *) (last_out + leftover_bytes));
memcpy(last_out, last_in, leftover_bytes);
ioc_destroy(&C);
return cum_count + leftover_bytes;
}
size_t bshuf_compress_lz4_bound(const size_t size,
const size_t elem_size, size_t block_size) {
size_t bound, leftover;
if (block_size == 0) {
block_size = bshuf_default_block_size(elem_size);
}
if (block_size < 0 || block_size % BSHUF_BLOCKED_MULT) return -81;
// Note that each block gets a 4 byte header.
// Size of full blocks.
bound = (LZ4_compressBound(block_size * elem_size) + 4) * (size / block_size);
// Size of partial blocks, if any.
leftover = ((size % block_size) / BSHUF_BLOCKED_MULT) * BSHUF_BLOCKED_MULT;
if (leftover) bound += LZ4_compressBound(leftover * elem_size) + 4;
// Size of uncompressed data not fitting into any blocks.
bound += (size % BSHUF_BLOCKED_MULT) * elem_size;
return bound;
}
