1bbbf1280Sopenharmony_ci/* 2bbbf1280Sopenharmony_ci * strncmp - compare two strings with limit 3bbbf1280Sopenharmony_ci * 4bbbf1280Sopenharmony_ci * Copyright (c) 2018-2021, Arm Limited. 5bbbf1280Sopenharmony_ci * SPDX-License-Identifier: MIT 6bbbf1280Sopenharmony_ci */ 7bbbf1280Sopenharmony_ci 8bbbf1280Sopenharmony_ci#include "../asmdefs.h" 9bbbf1280Sopenharmony_ci 10bbbf1280Sopenharmony_ci#if __ARM_FEATURE_SVE 11bbbf1280Sopenharmony_ci/* Assumptions: 12bbbf1280Sopenharmony_ci * 13bbbf1280Sopenharmony_ci * ARMv8-a, AArch64 14bbbf1280Sopenharmony_ci * SVE Available. 15bbbf1280Sopenharmony_ci */ 16bbbf1280Sopenharmony_ci 17bbbf1280Sopenharmony_ciENTRY (__strncmp_aarch64_sve) 18bbbf1280Sopenharmony_ci PTR_ARG (0) 19bbbf1280Sopenharmony_ci PTR_ARG (1) 20bbbf1280Sopenharmony_ci SIZE_ARG (2) 21bbbf1280Sopenharmony_ci setffr /* initialize FFR */ 22bbbf1280Sopenharmony_ci mov x3, 0 /* initialize off */ 23bbbf1280Sopenharmony_ci 24bbbf1280Sopenharmony_ci0: whilelo p0.b, x3, x2 /* while off < max */ 25bbbf1280Sopenharmony_ci b.none 9f 26bbbf1280Sopenharmony_ci 27bbbf1280Sopenharmony_ci ldff1b z0.b, p0/z, [x0, x3] 28bbbf1280Sopenharmony_ci ldff1b z1.b, p0/z, [x1, x3] 29bbbf1280Sopenharmony_ci rdffrs p1.b, p0/z 30bbbf1280Sopenharmony_ci b.nlast 2f 31bbbf1280Sopenharmony_ci 32bbbf1280Sopenharmony_ci /* First fault did not fail: the vector up to max is valid. 33bbbf1280Sopenharmony_ci Avoid depending on the contents of FFR beyond the branch. 34bbbf1280Sopenharmony_ci Increment for a whole vector, even if we've only read a partial. 35bbbf1280Sopenharmony_ci This is significantly cheaper than INCP, and since OFF is not 36bbbf1280Sopenharmony_ci used after the loop it is ok to increment OFF past MAX. */ 37bbbf1280Sopenharmony_ci incb x3 38bbbf1280Sopenharmony_ci cmpeq p1.b, p0/z, z0.b, z1.b /* compare strings */ 39bbbf1280Sopenharmony_ci cmpne p2.b, p0/z, z0.b, 0 /* search for ~zero */ 40bbbf1280Sopenharmony_ci nands p2.b, p0/z, p1.b, p2.b /* ~(eq & ~zero) -> ne | zero */ 41bbbf1280Sopenharmony_ci b.none 0b 42bbbf1280Sopenharmony_ci 43bbbf1280Sopenharmony_ci /* Found end-of-string or inequality. */ 44bbbf1280Sopenharmony_ci1: brkb p2.b, p0/z, p2.b /* find first such */ 45bbbf1280Sopenharmony_ci lasta w0, p2, z0.b /* extract each char */ 46bbbf1280Sopenharmony_ci lasta w1, p2, z1.b 47bbbf1280Sopenharmony_ci sub x0, x0, x1 /* return comparison */ 48bbbf1280Sopenharmony_ci ret 49bbbf1280Sopenharmony_ci 50bbbf1280Sopenharmony_ci /* First fault failed: only some of the vector is valid. 51bbbf1280Sopenharmony_ci Perform the comparison only on the valid bytes. */ 52bbbf1280Sopenharmony_ci2: cmpeq p2.b, p1/z, z0.b, z1.b /* compare strings, as above */ 53bbbf1280Sopenharmony_ci cmpne p3.b, p1/z, z0.b, 0 54bbbf1280Sopenharmony_ci nands p2.b, p1/z, p2.b, p3.b 55bbbf1280Sopenharmony_ci b.any 1b 56bbbf1280Sopenharmony_ci 57bbbf1280Sopenharmony_ci /* No inequality or zero found. Re-init FFR, incr and loop. */ 58bbbf1280Sopenharmony_ci setffr 59bbbf1280Sopenharmony_ci incp x3, p1.b 60bbbf1280Sopenharmony_ci b 0b 61bbbf1280Sopenharmony_ci 62bbbf1280Sopenharmony_ci /* Found end-of-count. */ 63bbbf1280Sopenharmony_ci9: mov x0, 0 /* return equal */ 64bbbf1280Sopenharmony_ci ret 65bbbf1280Sopenharmony_ci 66bbbf1280Sopenharmony_ciEND (__strncmp_aarch64_sve) 67bbbf1280Sopenharmony_ci 68bbbf1280Sopenharmony_ci#endif 69bbbf1280Sopenharmony_ci 70