From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mga02.intel.com (mga02.intel.com [134.134.136.20]) by dpdk.org (Postfix) with ESMTP id B5F3F5A3E for ; Tue, 12 May 2015 13:03:54 +0200 (CEST) Received: from fmsmga002.fm.intel.com ([10.253.24.26]) by orsmga101.jf.intel.com with ESMTP; 12 May 2015 04:03:47 -0700 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="5.13,414,1427785200"; d="scan'208";a="724591385" Received: from irvmail001.ir.intel.com ([163.33.26.43]) by fmsmga002.fm.intel.com with ESMTP; 12 May 2015 04:03:46 -0700 Received: from sivswdev02.ir.intel.com (sivswdev02.ir.intel.com [10.237.217.46]) by irvmail001.ir.intel.com (8.14.3/8.13.6/MailSET/Hub) with ESMTP id t4CB3jiB025746; Tue, 12 May 2015 12:03:45 +0100 Received: from sivswdev02.ir.intel.com (localhost [127.0.0.1]) by sivswdev02.ir.intel.com with ESMTP id t4CB3f9f029128; Tue, 12 May 2015 12:03:41 +0100 Received: (from pdelarax@localhost) by sivswdev02.ir.intel.com with id t4CB3ewX029066; Tue, 12 May 2015 12:03:40 +0100 From: Pablo de Lara To: dev@dpdk.org Date: Tue, 12 May 2015 12:02:37 +0100 Message-Id: <1431428560-25426-6-git-send-email-pablo.de.lara.guarch@intel.com> X-Mailer: git-send-email 1.7.4.1 In-Reply-To: <1431428560-25426-1-git-send-email-pablo.de.lara.guarch@intel.com> References: <1430837034-21031-1-git-send-email-pablo.de.lara.guarch@intel.com> <1431428560-25426-1-git-send-email-pablo.de.lara.guarch@intel.com> Subject: [dpdk-dev] [PATCH v4 5/6] hash: remove duplicated code X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.15 Precedence: list List-Id: patches and discussions about DPDK List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Tue, 12 May 2015 11:03:55 -0000 rte_jhash is basically like _rte_jhash_2hashes but it returns only 1 hash, instead of 2. In order to remove duplicated code, rte_jhash calls _rte_jhash_2hashes, passing 0 as the second seed and returning just the first hash value. (performance penalty is negligible) The same is done with rte_jhash2. Also, rte_jhash2 is just an specific case where keys are multiple of 32 bits, and where no key alignment check is required. So,to avoid duplicated code, the function calls _rte_jhash_2hashes with check_align = 0 (to use the optimal path) Signed-off-by: Pablo de Lara --- lib/librte_hash/rte_jhash.h | 293 +++++++++---------------------------------- 1 files changed, 60 insertions(+), 233 deletions(-) diff --git a/lib/librte_hash/rte_jhash.h b/lib/librte_hash/rte_jhash.h index f1c0a8a..aa5bb2d 100644 --- a/lib/librte_hash/rte_jhash.h +++ b/lib/librte_hash/rte_jhash.h @@ -102,29 +102,19 @@ extern "C" { #define LOWER16b_MASK rte_le_to_cpu_32(0xffff) #define LOWER24b_MASK rte_le_to_cpu_32(0xffffff) -/** - * The most generic version, hashes an arbitrary sequence - * of bytes. No alignment or length assumptions are made about - * the input key. - * - * @param key - * Key to calculate hash of. - * @param length - * Length of key in bytes. - * @param initval - * Initialising value of hash. - * @return - * Calculated hash value. - */ -static inline uint32_t -rte_jhash(const void *key, uint32_t length, uint32_t initval) +static inline void +__rte_jhash_2hashes(const void *key, uint32_t length, uint32_t *pc, uint32_t *pb, unsigned check_align) { uint32_t a, b, c; /* Set up the internal state */ - a = b = c = RTE_JHASH_GOLDEN_RATIO + ((uint32_t)length) + initval; + a = b = c = RTE_JHASH_GOLDEN_RATIO + ((uint32_t)length) + *pc; + c += *pb; - /* Check key alignment. For x86 architecture, first case is always optimal */ + /* + * Check key alignment. For x86 architecture, first case is always optimal + * If check_align is not set, first case will be used + */ #if defined(RTE_ARCH_X86_64) || defined(RTE_ARCH_I686) || defined(RTE_ARCH_X86_X32) const uint32_t *k = key; const uint32_t s = 0; @@ -132,8 +122,7 @@ rte_jhash(const void *key, uint32_t length, uint32_t initval) const uint32_t *k = (uint32_t *)(uintptr_t)key & (uintptr_t)~3); const uint32_t s = ((uintptr_t)key & 3) * CHAR_BIT; #endif - - if (s == 0) { + if (!check_align || s == 0) { while (length > 12) { a += k[0]; b += k[1]; @@ -172,7 +161,9 @@ rte_jhash(const void *key, uint32_t length, uint32_t initval) a += k[0] & LOWER8b_MASK; break; /* zero length strings require no mixing */ case 0: - return c; + *pc = c; + *pb = b; + return; }; } else { /* all but the last block: affect some 32 bits of (a, b, c) */ @@ -238,63 +229,16 @@ rte_jhash(const void *key, uint32_t length, uint32_t initval) break; /* zero length strings require no mixing */ case 0: - return c; + *pc = c; + *pb = b; + return; } } __rte_jhash_final(a, b, c); - return c; -} - -/** - * A special optimized version that handles 1 or more of uint32_ts. - * The length parameter here is the number of uint32_ts in the key. - * - * @param k - * Key to calculate hash of. - * @param length - * Length of key in units of 4 bytes. - * @param initval - * Initialising value of hash. - * @return - * Calculated hash value. - */ -static inline uint32_t -rte_jhash2(const uint32_t *k, uint32_t length, uint32_t initval) -{ - uint32_t a, b, c; - - /* Set up the internal state */ - a = b = c = RTE_JHASH_GOLDEN_RATIO + (((uint32_t)length) << 2) + initval; - - /* Handle most of the key */ - while (length > 3) { - a += k[0]; - b += k[1]; - c += k[2]; - - __rte_jhash_mix(a, b, c); - - k += 3; - length -= 3; - } - - /* Handle the last 3 uint32_t's */ - switch (length) { - case 3: - c += k[2]; - case 2: - b += k[1]; - case 1: - a += k[0]; - __rte_jhash_final(a, b, c); - /* case 0: nothing left to add */ - case 0: - break; - }; - - return c; + *pc = c; + *pb = b; } /** @@ -315,138 +259,7 @@ rte_jhash2(const uint32_t *k, uint32_t length, uint32_t initval) static inline void rte_jhash_2hashes(const void *key, uint32_t length, uint32_t *pc, uint32_t *pb) { - uint32_t a, b, c; - - /* Set up the internal state */ - a = b = c = RTE_JHASH_GOLDEN_RATIO + ((uint32_t)length) + *pc; - c += *pb; - - /* Check key alignment. For x86 architecture, first case is always optimal */ -#if defined(RTE_ARCH_X86_64) || defined(RTE_ARCH_I686) || defined(RTE_ARCH_X86_X32) - const uint32_t *k = key; - const uint32_t s = 0; -#else - const uint32_t *k = (uint32_t *)(uintptr_t)key & (uintptr_t)~3); - const uint32_t s = ((uintptr_t)key & 3) * CHAR_BIT; -#endif - - if (s == 0) { - while (length > 12) { - a += k[0]; - b += k[1]; - c += k[2]; - - __rte_jhash_mix(a, b, c); - - k += 3; - length -= 12; - } - - switch (length) { - case 12: - c += k[2]; b += k[1]; a += k[0]; break; - case 11: - c += k[2] & LOWER24b_MASK; b += k[1]; a += k[0]; break; - case 10: - c += k[2] & LOWER16b_MASK; b += k[1]; a += k[0]; break; - case 9: - c += k[2] & LOWER8b_MASK; b += k[1]; a += k[0]; break; - case 8: - b += k[1]; a += k[0]; break; - case 7: - b += k[1] & LOWER24b_MASK; a += k[0]; break; - case 6: - b += k[1] & LOWER16b_MASK; a += k[0]; break; - case 5: - b += k[1] & LOWER8b_MASK; a += k[0]; break; - case 4: - a += k[0]; break; - case 3: - a += k[0] & LOWER24b_MASK; break; - case 2: - a += k[0] & LOWER16b_MASK; break; - case 1: - a += k[0] & LOWER8b_MASK; break; - /* zero length strings require no mixing */ - case 0: - *pc = c; - *pb = b; - return; - }; - } else { - /* all but the last block: affect some 32 bits of (a, b, c) */ - while (length > 12) { - a += BIT_SHIFT(k[0], k[1], s); - b += BIT_SHIFT(k[1], k[2], s); - c += BIT_SHIFT(k[2], k[3], s); - __rte_jhash_mix(a, b, c); - - k += 3; - length -= 12; - } - - /* last block: affect all 32 bits of (c) */ - switch (length) { - case 12: - a += BIT_SHIFT(k[0], k[1], s); - b += BIT_SHIFT(k[1], k[2], s); - c += BIT_SHIFT(k[2], k[3], s); - break; - case 11: - a += BIT_SHIFT(k[0], k[1], s); - b += BIT_SHIFT(k[1], k[2], s); - c += BIT_SHIFT(k[2], k[3], s) & LOWER24b_MASK; - break; - case 10: - a += BIT_SHIFT(k[0], k[1], s); - b += BIT_SHIFT(k[1], k[2], s); - c += BIT_SHIFT(k[2], k[3], s) & LOWER16b_MASK; - break; - case 9: - a += BIT_SHIFT(k[0], k[1], s); - b += BIT_SHIFT(k[1], k[2], s); - c += BIT_SHIFT(k[2], k[3], s) & LOWER8b_MASK; - break; - case 8: - a += BIT_SHIFT(k[0], k[1], s); - b += BIT_SHIFT(k[1], k[2], s); - break; - case 7: - a += BIT_SHIFT(k[0], k[1], s); - b += BIT_SHIFT(k[1], k[2], s) & LOWER24b_MASK; - break; - case 6: - a += BIT_SHIFT(k[0], k[1], s); - b += BIT_SHIFT(k[1], k[2], s) & LOWER16b_MASK; - break; - case 5: - a += BIT_SHIFT(k[0], k[1], s); - b += BIT_SHIFT(k[1], k[2], s) & LOWER8b_MASK; - break; - case 4: - a += BIT_SHIFT(k[0], k[1], s); - break; - case 3: - a += BIT_SHIFT(k[0], k[1], s) & LOWER24b_MASK; - break; - case 2: - a += BIT_SHIFT(k[0], k[1], s) & LOWER16b_MASK; - break; - case 1: - a += BIT_SHIFT(k[0], k[1], s) & LOWER8b_MASK; - break; - /* zero length strings require no mixing */ - case 0: - *pc = c; - *pb = b; - return; - } - } - - __rte_jhash_final(a, b, c); - - *pc = c; - *pb = b; + __rte_jhash_2hashes(key, length, pc, pb, 1); } /** @@ -467,40 +280,54 @@ rte_jhash_2hashes(const void *key, uint32_t length, uint32_t *pc, uint32_t *pb) static inline void rte_jhash2_2hashes(const uint32_t *k, uint32_t length, uint32_t *pc, uint32_t *pb) { - uint32_t a, b, c; + __rte_jhash_2hashes((const void *) k, (length << 2), pc, pb, 0); +} - /* Set up the internal state */ - a = b = c = RTE_JHASH_GOLDEN_RATIO + (((uint32_t)length) << 2) + *pc; - c += *pb; +/** + * The most generic version, hashes an arbitrary sequence + * of bytes. No alignment or length assumptions are made about + * the input key. + * + * @param key + * Key to calculate hash of. + * @param length + * Length of key in bytes. + * @param initval + * Initialising value of hash. + * @return + * Calculated hash value. + */ +static inline uint32_t +rte_jhash(const void *key, uint32_t length, uint32_t initval) +{ + uint32_t initval2 = 0; - /* Handle most of the key */ - while (length > 3) { - a += k[0]; - b += k[1]; - c += k[2]; + rte_jhash_2hashes(key, length, &initval, &initval2); - __rte_jhash_mix(a, b, c); + return initval; +} - k += 3; - length -= 3; - } +/** + * A special optimized version that handles 1 or more of uint32_ts. + * The length parameter here is the number of uint32_ts in the key. + * + * @param k + * Key to calculate hash of. + * @param length + * Length of key in units of 4 bytes. + * @param initval + * Initialising value of hash. + * @return + * Calculated hash value. + */ +static inline uint32_t +rte_jhash2(const uint32_t *k, uint32_t length, uint32_t initval) +{ + uint32_t initval2 = 0; - /* Handle the last 3 uint32_t's */ - switch (length) { - case 3: - c += k[2]; - case 2: - b += k[1]; - case 1: - a += k[0]; - __rte_jhash_final(a, b, c); - /* case 0: nothing left to add */ - case 0: - break; - }; + rte_jhash2_2hashes(k, length, &initval, &initval2); - *pc = c; - *pb = b; + return initval; } static inline uint32_t -- 1.7.4.1