const uint8_t *out_end = data + (stream_idx * 64 + y * 8 + x + 1) * 1024;
const uint8_t *ptr = data + offsets[stream_idx * 64 + y * 8 + x];
uint32_t num_rans_bytes = out_end - ptr;
+ assert(num_rans_bytes <= 1024);
if (num_rans_bytes == last_block.size() &&
memcmp(last_block.data(), ptr, last_block.size()) == 0) {