Lines Matching +full:v1 +full:- +full:v6
2 // Accelerated CRC-T10DIF using arm64 NEON and Crypto Extensions instructions
14 // Implement fast CRC-T10DIF computation with SSE and PCLMULQDQ instructions
62 // /white-papers/fast-crc-computation-generic-polynomials-pclmulqdq-paper.pdf
69 .arch armv8-a+crypto
276 CPU_LE( rev64 v1.16b, v1.16b )
281 CPU_LE( rev64 v6.16b, v6.16b )
284 CPU_LE( ext v1.16b, v1.16b, v1.16b, #8 )
289 CPU_LE( ext v6.16b, v6.16b, v6.16b, #8 )
305 // While >= 128 data bytes remain (not counting v0-v7), fold the 128
306 // bytes v0-v7 into them, storing the result back into v0-v7.
308 fold_32_bytes \p, v0, v1
311 fold_32_bytes \p, v6, v7
316 // Now fold the 112 bytes in v0-v6 into the 16 bytes in v7.
323 fold_16_bytes \p, v1, v5
324 fold_16_bytes \p, v2, v6
327 fold_16_bytes \p, v4, v6
330 fold_16_bytes \p, v6, v7
336 adds len, len, #(128-16)
367 ldr q0, [buf, #-16]
371 // v1 = high order part of second chunk: v7 left-shifted by 'len' bytes.
375 tbl v1.16b, {v7.16b}, v2.16b
377 // v3 = first chunk: v7 right-shifted by '16-len' bytes.
382 // Convert to 8-bit masks: 'len' 0x00 bytes, then '16-len' 0xff bytes.
385 // v2 = second chunk: 'len' bytes from v0 (low-order bytes),
386 // then '16-len' bytes from v1 (high-order bytes).
387 bsl v2.16b, v1.16b, v0.16b
396 // Reduce the 128-bit value M(x), stored in v7, to the final 16-bit CRC.
405 // x^64. This produces a 128-bit value congruent to x^64 * M(x) and
411 // Fold the high 32 bits into the low 96 bits. This produces a 96-bit
413 ext v1.16b, v0.16b, v2.16b, #12 // extract high 32 bits
415 __pmull_\p v1, v1, fold_consts // high 32 bits * x^48 * (x^48 mod G(x))
416 eor v0.16b, v0.16b, v1.16b // + low bits
423 __pmull_\p v1, v0, fold_consts, 2 // high 32 bits * floor(x^48 / G(x))
424 ushr v1.2d, v1.2d, #32 // /= x^32
425 __pmull_\p v1, v1, fold_consts // *= G(x)
427 eor v0.16b, v0.16b, v1.16b // + low 16 nonzero bits
451 // Load the fold-across-16-bytes constants.
469 stp x29, x30, [sp, #-16]!
508 // For 1 <= len <= 15, the 16-byte vector beginning at &byteshift_table[16 -
510 // ..., 0x80} XOR the index vector to shift right by '16 - len' bytes.