Skip to content

Commit

Permalink
Simplify SSSE3 nibble2base function
Browse files Browse the repository at this point in the history
  • Loading branch information
rhpvorderman committed Jul 8, 2024
1 parent b8145e6 commit fb7dd52
Showing 1 changed file with 16 additions and 39 deletions.
55 changes: 16 additions & 39 deletions simd.c
Original file line number Diff line number Diff line change
Expand Up @@ -123,48 +123,25 @@ static void nibble2base_ssse3(uint8_t *nib, char *seq, int len) {
char *seq_cursor = seq;
uint8_t *nibble_cursor = nib;
const char *seq_vec_end_ptr = seq_end_ptr - (2 * sizeof(__m128i) - 1);
__m128i first_upper_shuffle = _mm_setr_epi8(
0, -1, 1, -1, 2, -1, 3, -1, 4, -1, 5, -1, 6, -1, 7, -1);
__m128i first_lower_shuffle = _mm_setr_epi8(
-1, 0, -1, 1, -1, 2, -1, 3, -1, 4, -1, 5, -1, 6, -1, 7);
__m128i second_upper_shuffle = _mm_setr_epi8(
8, -1, 9, -1, 10, -1, 11, -1, 12, -1, 13, -1, 14, -1, 15, -1);
__m128i second_lower_shuffle = _mm_setr_epi8(
-1, 8, -1, 9, -1, 10, -1, 11, -1, 12, -1, 13, -1, 14, -1, 15);
__m128i nuc_lookup_vec = _mm_lddqu_si128((__m128i *)seq_nt16_str);
/* Work on 16 encoded characters at the time resulting in 32 decoded characters
Examples are given for 8 encoded characters A until H to keep it readable.
Encoded stored as |AB|CD|EF|GH|
Shuffle into |AB|00|CD|00|EF|00|GH|00| and
|00|AB|00|CD|00|EF|00|GH|
Shift upper to the right resulting into
|0A|B0|0C|D0|0E|F0|0G|H0| and
|00|AB|00|CD|00|EF|00|GH|
Merge with or resulting into (X stands for garbage)
|0A|XB|0C|XD|0E|XF|0G|XH|
Bitwise and with 0b1111 leads to:
|0A|0B|0C|0D|0E|0F|0G|0H|
We can use the resulting 4-bit integers as indexes for the shuffle of
the nucleotide lookup. */
/* Nucleotides are encoded 4-bits per nucleotide and stored in 8-bit bytes
as follows: |AB|CD|EF|GH|. The 4-bit codes (going from 0-15) can be used
together with the pshufb instruction as a lookup table. The most efficient
way is to use bitwise AND and shift to create two vectors. One with all
the upper codes (|A|C|E|G|) and one with the lower codes (|B|D|F|H|).
The lookup can then be performed and the resulting vectors can be
interleaved again using the unpack instructions. */
while (seq_cursor < seq_vec_end_ptr) {
__m128i encoded = _mm_lddqu_si128((__m128i *)nibble_cursor);

__m128i first_upper = _mm_shuffle_epi8(encoded, first_upper_shuffle);
__m128i first_lower = _mm_shuffle_epi8(encoded, first_lower_shuffle);
__m128i shifted_first_upper = _mm_srli_epi64(first_upper, 4);
__m128i first_merged = _mm_or_si128(shifted_first_upper, first_lower);
__m128i first_indexes = _mm_and_si128(first_merged, _mm_set1_epi8(15));
__m128i first_nucleotides = _mm_shuffle_epi8(nuc_lookup_vec, first_indexes);
__m128i encoded_upper = _mm_srli_epi64(encoded, 4);
encoded_upper = _mm_and_si128(encoded_upper, _mm_set1_epi8(15));
__m128i encoded_lower = _mm_and_si128(encoded, _mm_set1_epi8(15));
__m128i nucs_upper = _mm_shuffle_epi8(nuc_lookup_vec, encoded_upper);
__m128i nucs_lower = _mm_shuffle_epi8(nuc_lookup_vec, encoded_lower);
__m128i first_nucleotides = _mm_unpacklo_epi8(nucs_upper, nucs_lower);
__m128i second_nucleotides = _mm_unpackhi_epi8(nucs_upper, nucs_lower);
_mm_storeu_si128((__m128i *)seq_cursor, first_nucleotides);

__m128i second_upper = _mm_shuffle_epi8(encoded, second_upper_shuffle);
__m128i second_lower = _mm_shuffle_epi8(encoded, second_lower_shuffle);
__m128i shifted_second_upper = _mm_srli_epi64(second_upper, 4);
__m128i second_merged = _mm_or_si128(shifted_second_upper, second_lower);
__m128i second_indexes = _mm_and_si128(second_merged, _mm_set1_epi8(15));
__m128i second_nucleotides = _mm_shuffle_epi8(nuc_lookup_vec, second_indexes);
_mm_storeu_si128((__m128i *)(seq_cursor + 16), second_nucleotides);

_mm_storeu_si128((__m128i *)(seq_cursor + (sizeof(__m128i))),
second_nucleotides);
nibble_cursor += sizeof(__m128i);
seq_cursor += 2 * sizeof(__m128i);
}
Expand Down

0 comments on commit fb7dd52

Please sign in to comment.