452 INLINE void transpose_msg_vecs4(const uint8_t *const *inputs,  in transpose_msg_vecs4()  argument
454   out[0] = loadu_128(&inputs[0][block_offset + 0 * sizeof(__m128i)]);  in transpose_msg_vecs4()
455   out[1] = loadu_128(&inputs[1][block_offset + 0 * sizeof(__m128i)]);  in transpose_msg_vecs4()
456   out[2] = loadu_128(&inputs[2][block_offset + 0 * sizeof(__m128i)]);  in transpose_msg_vecs4()
457   out[3] = loadu_128(&inputs[3][block_offset + 0 * sizeof(__m128i)]);  in transpose_msg_vecs4()
458   out[4] = loadu_128(&inputs[0][block_offset + 1 * sizeof(__m128i)]);  in transpose_msg_vecs4()
459   out[5] = loadu_128(&inputs[1][block_offset + 1 * sizeof(__m128i)]);  in transpose_msg_vecs4()
460   out[6] = loadu_128(&inputs[2][block_offset + 1 * sizeof(__m128i)]);  in transpose_msg_vecs4()
461   out[7] = loadu_128(&inputs[3][block_offset + 1 * sizeof(__m128i)]);  in transpose_msg_vecs4()
462   out[8] = loadu_128(&inputs[0][block_offset + 2 * sizeof(__m128i)]);  in transpose_msg_vecs4()
463   out[9] = loadu_128(&inputs[1][block_offset + 2 * sizeof(__m128i)]);  in transpose_msg_vecs4()
464   out[10] = loadu_128(&inputs[2][block_offset + 2 * sizeof(__m128i)]);  in transpose_msg_vecs4()
465   out[11] = loadu_128(&inputs[3][block_offset + 2 * sizeof(__m128i)]);  in transpose_msg_vecs4()
466   out[12] = loadu_128(&inputs[0][block_offset + 3 * sizeof(__m128i)]);  in transpose_msg_vecs4()
467   out[13] = loadu_128(&inputs[1][block_offset + 3 * sizeof(__m128i)]);  in transpose_msg_vecs4()
468   out[14] = loadu_128(&inputs[2][block_offset + 3 * sizeof(__m128i)]);  in transpose_msg_vecs4()
469   out[15] = loadu_128(&inputs[3][block_offset + 3 * sizeof(__m128i)]);  in transpose_msg_vecs4()
471     _mm_prefetch((const void *)&inputs[i][block_offset + 256], _MM_HINT_T0);  in transpose_msg_vecs4()
492 void blake3_hash4_avx512(const uint8_t *const *inputs, size_t blocks,  in blake3_hash4_avx512()  argument
512     transpose_msg_vecs4(inputs, block * BLAKE3_BLOCK_LEN, msg_vecs);  in blake3_hash4_avx512()
709 INLINE void transpose_msg_vecs8(const uint8_t *const *inputs,  in transpose_msg_vecs8()  argument
711   out[0] = loadu_256(&inputs[0][block_offset + 0 * sizeof(__m256i)]);  in transpose_msg_vecs8()
712   out[1] = loadu_256(&inputs[1][block_offset + 0 * sizeof(__m256i)]);  in transpose_msg_vecs8()
713   out[2] = loadu_256(&inputs[2][block_offset + 0 * sizeof(__m256i)]);  in transpose_msg_vecs8()
714   out[3] = loadu_256(&inputs[3][block_offset + 0 * sizeof(__m256i)]);  in transpose_msg_vecs8()
715   out[4] = loadu_256(&inputs[4][block_offset + 0 * sizeof(__m256i)]);  in transpose_msg_vecs8()
716   out[5] = loadu_256(&inputs[5][block_offset + 0 * sizeof(__m256i)]);  in transpose_msg_vecs8()
717   out[6] = loadu_256(&inputs[6][block_offset + 0 * sizeof(__m256i)]);  in transpose_msg_vecs8()
718   out[7] = loadu_256(&inputs[7][block_offset + 0 * sizeof(__m256i)]);  in transpose_msg_vecs8()
719   out[8] = loadu_256(&inputs[0][block_offset + 1 * sizeof(__m256i)]);  in transpose_msg_vecs8()
720   out[9] = loadu_256(&inputs[1][block_offset + 1 * sizeof(__m256i)]);  in transpose_msg_vecs8()
721   out[10] = loadu_256(&inputs[2][block_offset + 1 * sizeof(__m256i)]);  in transpose_msg_vecs8()
722   out[11] = loadu_256(&inputs[3][block_offset + 1 * sizeof(__m256i)]);  in transpose_msg_vecs8()
723   out[12] = loadu_256(&inputs[4][block_offset + 1 * sizeof(__m256i)]);  in transpose_msg_vecs8()
724   out[13] = loadu_256(&inputs[5][block_offset + 1 * sizeof(__m256i)]);  in transpose_msg_vecs8()
725   out[14] = loadu_256(&inputs[6][block_offset + 1 * sizeof(__m256i)]);  in transpose_msg_vecs8()
726   out[15] = loadu_256(&inputs[7][block_offset + 1 * sizeof(__m256i)]);  in transpose_msg_vecs8()
728     _mm_prefetch((const void *)&inputs[i][block_offset + 256], _MM_HINT_T0);  in transpose_msg_vecs8()
747 void blake3_hash8_avx512(const uint8_t *const *inputs, size_t blocks,  in blake3_hash8_avx512()  argument
767     transpose_msg_vecs8(inputs, block * BLAKE3_BLOCK_LEN, msg_vecs);  in blake3_hash8_avx512()
1023 INLINE void transpose_msg_vecs16(const uint8_t *const *inputs,  in transpose_msg_vecs16()  argument
1025   out[0] = loadu_512(&inputs[0][block_offset]);  in transpose_msg_vecs16()
1026   out[1] = loadu_512(&inputs[1][block_offset]);  in transpose_msg_vecs16()
1027   out[2] = loadu_512(&inputs[2][block_offset]);  in transpose_msg_vecs16()
1028   out[3] = loadu_512(&inputs[3][block_offset]);  in transpose_msg_vecs16()
1029   out[4] = loadu_512(&inputs[4][block_offset]);  in transpose_msg_vecs16()
1030   out[5] = loadu_512(&inputs[5][block_offset]);  in transpose_msg_vecs16()
1031   out[6] = loadu_512(&inputs[6][block_offset]);  in transpose_msg_vecs16()
1032   out[7] = loadu_512(&inputs[7][block_offset]);  in transpose_msg_vecs16()
1033   out[8] = loadu_512(&inputs[8][block_offset]);  in transpose_msg_vecs16()
1034   out[9] = loadu_512(&inputs[9][block_offset]);  in transpose_msg_vecs16()
1035   out[10] = loadu_512(&inputs[10][block_offset]);  in transpose_msg_vecs16()
1036   out[11] = loadu_512(&inputs[11][block_offset]);  in transpose_msg_vecs16()
1037   out[12] = loadu_512(&inputs[12][block_offset]);  in transpose_msg_vecs16()
1038   out[13] = loadu_512(&inputs[13][block_offset]);  in transpose_msg_vecs16()
1039   out[14] = loadu_512(&inputs[14][block_offset]);  in transpose_msg_vecs16()
1040   out[15] = loadu_512(&inputs[15][block_offset]);  in transpose_msg_vecs16()
1042     _mm_prefetch((const void *)&inputs[i][block_offset + 256], _MM_HINT_T0);  in transpose_msg_vecs16()
1060 void blake3_hash16_avx512(const uint8_t *const *inputs, size_t blocks,  in blake3_hash16_avx512()  argument
1081     transpose_msg_vecs16(inputs, block * BLAKE3_BLOCK_LEN, msg_vecs);  in blake3_hash16_avx512()
1162 void blake3_hash_many_avx512(const uint8_t *const *inputs, size_t num_inputs,  in blake3_hash_many_avx512()  argument
1168     blake3_hash16_avx512(inputs, blocks, key, counter, increment_counter, flags,  in blake3_hash_many_avx512()
1173     inputs += 16;  in blake3_hash_many_avx512()
1178     blake3_hash8_avx512(inputs, blocks, key, counter, increment_counter, flags,  in blake3_hash_many_avx512()
1183     inputs += 8;  in blake3_hash_many_avx512()
1188     blake3_hash4_avx512(inputs, blocks, key, counter, increment_counter, flags,  in blake3_hash_many_avx512()
1193     inputs += 4;  in blake3_hash_many_avx512()
1198     hash_one_avx512(inputs[0], blocks, key, counter, flags, flags_start,  in blake3_hash_many_avx512()
1203     inputs += 1;  in blake3_hash_many_avx512()