Skip to content

Instantly share code, notes, and snippets.

What would you like to do?
Ponderous older version of our "are we inside quotes" code
// Step 2: detect insides of quote pairs
u64 quote_bits = cmp_mask_against_input(input_lo, input_hi, _mm256_set1_epi8('"'));
quote_bits = quote_bits & ~odd_ends;
dumpbits(quote_bits, "quote_bits");
// pdep pattern is alternating 0 and 1 bits, starting with 0 or 1 depending on whether
// we're in a quote-pair from the previous iteration
u64 pdep_pattern = even_bits ^ prev_iter_inside_quote;
u64 starting_quotes = _pdep_u64(pdep_pattern, quote_bits);
// now starting quotes is the opening quote in each pair
dumpbits(pdep_pattern, "pdep_pattern");
dumpbits(starting_quotes, "starting quotes");
// record whether we borrow from out of bit 63 (whether we're in a quote at iteration end)
// we also need to subtract 1 if we need to; this reflects
// previous iteration status. This needs to be effectively subtracted from our starting
// quotes as we need to get a carry even if there aren't any quote bits in order to preserve
// our 'in a quote' status inside a long string
// aside from that it is simple: just clear starting quotes and subtract them, leaving us
// with one bits where we are inside quotes (a half-open range including the start quote
// but not the end quote).
u64 subtracted_starts;
bool iter_ends_quote = __builtin_usubll_overflow(
(quote_bits ^ starting_quotes),
starting_quotes - prev_iter_inside_quote,
prev_iter_inside_quote = iter_ends_quote ? 0xffffffffffffffffULL : 0x0ULL;
dumpbits(subtracted_starts, "subtracted_starts");
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment