From 2a33e6a1427ec3e6d7b359f93414d781a6c7d21f Mon Sep 17 00:00:00 2001 From: Ruben Vorderman Date: Mon, 8 Jul 2024 07:51:30 +0200 Subject: [PATCH] Simplify SSSE3 nibble2base function --- simd.c | 54 ++++++++++++++++-------------------------------------- 1 file changed, 16 insertions(+), 38 deletions(-) diff --git a/simd.c b/simd.c index ba3edea7b..c0038fa82 100644 --- a/simd.c +++ b/simd.c @@ -123,48 +123,26 @@ 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); }