From nobody Sun May 5 13:52:27 2024 Delivered-To: importer@patchew.org Authentication-Results: mx.zohomail.com; dkim=pass header.i=@intel.com; spf=pass (zohomail.com: domain of gnu.org designates 209.51.188.17 as permitted sender) smtp.mailfrom=qemu-devel-bounces+importer=patchew.org@nongnu.org; dmarc=pass(p=none dis=none) header.from=intel.com ARC-Seal: i=1; a=rsa-sha256; t=1660034246; cv=none; d=zohomail.com; s=zohoarc; b=XlCrUBi5MfIV3ZgrTr6pkaz3EU+8PPtEUAfpog63TCwTY/tq5FzG0G73IDEBZxmOyddjD/g5tckLHp/yg+42zCzdRGjiZNslr+TyDvjMONfjRyldy1XB6BNL8G/UhD/hUVcpcekZUXPZQ1A1oZ3C2XV2Ty/e6qCN/s+7TnwmU/A= ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=zohomail.com; s=zohoarc; t=1660034246; h=Content-Transfer-Encoding:Cc:Date:From:In-Reply-To:List-Subscribe:List-Post:List-Id:List-Archive:List-Help:List-Unsubscribe:MIME-Version:Message-ID:References:Sender:Subject:To; bh=cQ1bSmdOOPlaOP0FTy+lx4ltC0ozitgXZAEbQMGZMIg=; b=J3nV5nZN2uIdZ6ETgr9lkwB/JCbDNWgQpOa6qnBhMMIiBQrrJi/cFhvEvUsasYXXVjCf8kHPlR1ONRv4w0YygN2B5dGCiSFpu+YbJCpk33ocOGkipqN7YgYoaY2aiUiNywvjNALckckITJbV8w89fcfW7ww2MKlaXkUPI7l4OlQ= ARC-Authentication-Results: i=1; mx.zohomail.com; dkim=pass header.i=@intel.com; spf=pass (zohomail.com: domain of gnu.org designates 209.51.188.17 as permitted sender) smtp.mailfrom=qemu-devel-bounces+importer=patchew.org@nongnu.org; dmarc=pass header.from= (p=none dis=none) Return-Path: Received: from lists.gnu.org (lists.gnu.org [209.51.188.17]) by mx.zohomail.com with SMTPS id 1660034246249848.7631718957618; Tue, 9 Aug 2022 01:37:26 -0700 (PDT) Received: from localhost ([::1]:51652 helo=lists1p.gnu.org) by lists.gnu.org with esmtp (Exim 4.90_1) (envelope-from ) id 1oLKjs-0007Sg-2G for importer@patchew.org; Tue, 09 Aug 2022 04:37:24 -0400 Received: from eggs.gnu.org ([2001:470:142:3::10]:56902) by lists.gnu.org with esmtps (TLS1.2:ECDHE_RSA_AES_256_GCM_SHA384:256) (Exim 4.90_1) (envelope-from ) id 1oLKfk-0004bp-2y for qemu-devel@nongnu.org; Tue, 09 Aug 2022 04:33:08 -0400 Received: from mga01.intel.com ([192.55.52.88]:48324) by eggs.gnu.org with esmtps (TLS1.2:ECDHE_RSA_AES_256_GCM_SHA384:256) (Exim 4.90_1) (envelope-from ) id 1oLKfh-0008TS-CT for qemu-devel@nongnu.org; Tue, 09 Aug 2022 04:33:07 -0400 Received: from orsmga008.jf.intel.com ([10.7.209.65]) by fmsmga101.fm.intel.com with ESMTP/TLS/ECDHE-RSA-AES256-GCM-SHA384; 09 Aug 2022 01:33:01 -0700 Received: from xuling-b360m-d3h.sh.intel.com ([10.239.82.110]) by orsmga008.jf.intel.com with ESMTP; 09 Aug 2022 01:32:58 -0700 DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/simple; d=intel.com; i=@intel.com; q=dns/txt; s=Intel; t=1660033985; x=1691569985; h=from:to:cc:subject:date:message-id:in-reply-to: references:mime-version:content-transfer-encoding; bh=S2H7RKS+RzhI5wbMCTgPTGE5hlEfydXncy/X7N6qkbM=; b=AZbLywVCv96e6z0Il9UXu6jxGTcjLHKHYiFvuk9/qgbgFB3EcjNeVWIM 7S9TjYgaFaa5DNIBL5jiaNU4vMQu8EJAPh9OOrybgIWQkQd0+1TaUzGLZ UYM8hVnoA9RFAR3nMJlqrL6yav7VvAmqdN9JR6pdM9coblu+Ns+DsM6Xl e4I+biytAX+UhXI4kNAT8Ee3ytv3bIYxOKJlTS+NkecsN8pmiBvhV1Wj+ nvybAy7r+wdWjUHEbhy8WTR+pusUvhid+KEPIg5lohBz5cEXFdm06jBJK cGbhXrO2kPF6kB/HNBexOsS8AR5MK1LR3h2EN86+Zk36fjz5BG6dP2k02 g==; X-IronPort-AV: E=McAfee;i="6400,9594,10433"; a="316712878" X-IronPort-AV: E=Sophos;i="5.93,224,1654585200"; d="scan'208";a="316712878" X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="5.93,224,1654585200"; d="scan'208";a="633251168" From: ling xu To: qemu-devel@nongnu.org Cc: quintela@redhat.com, dgilbert@redhat.com, ling xu , Zhou Zhao , Jun Jin Subject: [PATCH v4 1/2] Update AVX512 support for xbzrle_encode_buffer function Date: Tue, 9 Aug 2022 16:32:48 +0800 Message-Id: <20220809083249.1646952-2-ling1.xu@intel.com> X-Mailer: git-send-email 2.25.1 In-Reply-To: <20220809083249.1646952-1-ling1.xu@intel.com> References: <20220809083249.1646952-1-ling1.xu@intel.com> MIME-Version: 1.0 Content-Transfer-Encoding: quoted-printable Received-SPF: pass (zohomail.com: domain of gnu.org designates 209.51.188.17 as permitted sender) client-ip=209.51.188.17; envelope-from=qemu-devel-bounces+importer=patchew.org@nongnu.org; helo=lists.gnu.org; Received-SPF: pass client-ip=192.55.52.88; envelope-from=ling1.xu@intel.com; helo=mga01.intel.com X-Spam_score_int: -71 X-Spam_score: -7.2 X-Spam_bar: ------- X-Spam_report: (-7.2 / 5.0 requ) BAYES_00=-1.9, DKIMWL_WL_HIGH=-0.082, DKIM_SIGNED=0.1, DKIM_VALID=-0.1, DKIM_VALID_AU=-0.1, DKIM_VALID_EF=-0.1, RCVD_IN_DNSWL_HI=-5, RCVD_IN_MSPIKE_H3=0.001, RCVD_IN_MSPIKE_WL=0.001, SPF_HELO_NONE=0.001, SPF_PASS=-0.001, T_SCC_BODY_TEXT_LINE=-0.01 autolearn=ham autolearn_force=no X-Spam_action: no action X-BeenThere: qemu-devel@nongnu.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: qemu-devel-bounces+importer=patchew.org@nongnu.org Sender: "Qemu-devel" X-ZohoMail-DKIM: pass (identity @intel.com) X-ZM-MESSAGEID: 1660034247913100001 Content-Type: text/plain; charset="utf-8" This commit update code of avx512 support for xbzrle_encode_buffer function to accelerate xbzrle encoding speed. Compared with C version of xbzrle_encode_buffer function, avx512 version can achieve almost 60%-70% performance improvement on unit test provided by Qemu. In addition, we provide one more unit test called "test_encode_decode_random_avx512", in which dirty data are randomly located in 4K page, and this case can achieve almost 140% performance gain. Signed-off-by: ling xu Co-authored-by: Zhou Zhao Co-authored-by: Jun Jin --- meson.build | 16 ++++ meson_options.txt | 2 + migration/ram.c | 42 ++++++++++- migration/xbzrle.c | 181 +++++++++++++++++++++++++++++++++++++++++++++ migration/xbzrle.h | 4 + 5 files changed, 242 insertions(+), 3 deletions(-) diff --git a/meson.build b/meson.build index 294e9a8f32..4222b77e9f 100644 --- a/meson.build +++ b/meson.build @@ -2262,6 +2262,22 @@ config_host_data.set('CONFIG_AVX512F_OPT', get_optio= n('avx512f') \ int main(int argc, char *argv[]) { return bar(argv[0]); } '''), error_message: 'AVX512F not available').allowed()) =20 +config_host_data.set('CONFIG_AVX512BW_OPT', get_option('avx512bw') \ + .require(have_cpuid_h, error_message: 'cpuid.h not available, cannot ena= ble AVX512BW') \ + .require(cc.links(''' + #pragma GCC push_options + #pragma GCC target("avx512bw") + #include + #include + static int bar(void *a) { + + __m512i x =3D *(__m512i *)a; + __m512i res=3D _mm512_abs_epi8(x); + return res[1]; + } + int main(int argc, char *argv[]) { return bar(argv[0]); } + '''), error_message: 'AVX512BW not available').allowed()) + have_pvrdma =3D get_option('pvrdma') \ .require(rdma.found(), error_message: 'PVRDMA requires OpenFabrics libra= ries') \ .require(cc.compiles(gnu_source_prefix + ''' diff --git a/meson_options.txt b/meson_options.txt index e58e158396..07194bf680 100644 --- a/meson_options.txt +++ b/meson_options.txt @@ -104,6 +104,8 @@ option('avx2', type: 'feature', value: 'auto', description: 'AVX2 optimizations') option('avx512f', type: 'feature', value: 'disabled', description: 'AVX512F optimizations') +option('avx512bw', type: 'feature', value: 'auto', + description: 'AVX512BW optimizations') option('keyring', type: 'feature', value: 'auto', description: 'Linux keyring support') =20 diff --git a/migration/ram.c b/migration/ram.c index dc1de9ddbc..bae7bef236 100644 --- a/migration/ram.c +++ b/migration/ram.c @@ -83,6 +83,35 @@ /* 0x80 is reserved in migration.h start with 0x100 next */ #define RAM_SAVE_FLAG_COMPRESS_PAGE 0x100 =20 +#if defined(CONFIG_AVX512BW_OPT) +static bool is_cpu_support_avx512bw; +#include "qemu/cpuid.h" +static void __attribute__((constructor)) init_cpu_flag(void) +{ + unsigned max =3D __get_cpuid_max(0, NULL); + int a, b, c, d; + is_cpu_support_avx512bw =3D false; + if (max >=3D 1) { + __cpuid(1, a, b, c, d); + /* We must check that AVX is not just available, but usable. */ + if ((c & bit_OSXSAVE) && (c & bit_AVX) && max >=3D 7) { + int bv; + __asm("xgetbv" : "=3Da"(bv), "=3Dd"(d) : "c"(0)); + __cpuid_count(7, 0, a, b, c, d); + /* 0xe6: + * XCR0[7:5] =3D 111b (OPMASK state, upper 256-bit of ZMM0-ZMM= 15 + * and ZMM16-ZMM31 state are enabled by OS) + * XCR0[2:1] =3D 11b (XMM state and YMM state are enabled by O= S) + */ + if ((bv & 0xe6) =3D=3D 0xe6 && (b & bit_AVX512BW)) { + is_cpu_support_avx512bw =3D true; + } + } + } + return ; +} +#endif + XBZRLECacheStats xbzrle_counters; =20 /* struct contains XBZRLE cache and a static page @@ -802,9 +831,16 @@ static int save_xbzrle_page(RAMState *rs, uint8_t **cu= rrent_data, memcpy(XBZRLE.current_buf, *current_data, TARGET_PAGE_SIZE); =20 /* XBZRLE encoding (if there is no overflow) */ - encoded_len =3D xbzrle_encode_buffer(prev_cached_page, XBZRLE.current_= buf, - TARGET_PAGE_SIZE, XBZRLE.encoded_bu= f, - TARGET_PAGE_SIZE); + int (*xbzrle_encode_buffer_func)(uint8_t *, uint8_t *, int, + uint8_t *, int) =3D xbzrle_encode_buffer; + #if defined(CONFIG_AVX512BW_OPT) + if (likely(is_cpu_support_avx512bw)) { + xbzrle_encode_buffer_func =3D xbzrle_encode_buffer_avx512; + } + #endif + encoded_len =3D xbzrle_encode_buffer_func(prev_cached_page, XBZRLE.cur= rent_buf, + TARGET_PAGE_SIZE, XBZRLE.encod= ed_buf, + TARGET_PAGE_SIZE); =20 /* * Update the cache contents, so that it corresponds to the data diff --git a/migration/xbzrle.c b/migration/xbzrle.c index 1ba482ded9..804adc1acb 100644 --- a/migration/xbzrle.c +++ b/migration/xbzrle.c @@ -174,3 +174,184 @@ int xbzrle_decode_buffer(uint8_t *src, int slen, uint= 8_t *dst, int dlen) =20 return d; } + +#if defined(CONFIG_AVX512BW_OPT) +#pragma GCC push_options +#pragma GCC target("avx512bw") + +#include +#include +#define SET_ZERO512(r) r =3D _mm512_set1_epi32(0) +int xbzrle_encode_buffer_avx512(uint8_t *old_buf, uint8_t *new_buf, int sl= en, + uint8_t *dst, int dlen) +{ + uint32_t zrun_len =3D 0, nzrun_len =3D 0; + int d =3D 0, i =3D 0, num =3D 0; + uint8_t *nzrun_start =3D NULL; + int count512s =3D (slen >> 6); + int countResidual =3D slen % 64; + bool never_same =3D true; + while (count512s--) { + if (d + 2 > dlen) { + return -1; + } + __m512i old_data =3D _mm512_mask_loadu_epi8(old_data, + 0xffffffffffffffff, old_buf + i); + __m512i new_data =3D _mm512_mask_loadu_epi8(new_data, + 0xffffffffffffffff, new_b= uf + i); + /* in mask bit 1 for same, 0 for diff */ + __mmask64 comp =3D _mm512_cmpeq_epi8_mask(old_data, new_data); + + int bytesToCheck =3D 64; + bool is_same =3D (comp & 0x1); + while (bytesToCheck) { + if (is_same) { + if (nzrun_len) { + d +=3D uleb128_encode_small(dst + d, nzrun_len); + if (d + nzrun_len > dlen) { + return -1; + } + nzrun_start =3D new_buf + i - nzrun_len; + memcpy(dst + d, nzrun_start, nzrun_len); + d +=3D nzrun_len; + nzrun_len =3D 0; + } + if (comp =3D=3D 0xffffffffffffffff) { + i +=3D 64; + zrun_len +=3D 64; + break; + } + never_same =3D false; + num =3D __builtin_ctzl(~comp); + num =3D (num < bytesToCheck) ? num : bytesToCheck; + zrun_len +=3D num; + bytesToCheck -=3D num; + comp >>=3D num; + i +=3D num; + if (bytesToCheck) { + /* still has different data after same data */ + d +=3D uleb128_encode_small(dst + d, zrun_len); + zrun_len =3D 0; + } else { + break; + } + } + if (never_same || zrun_len) { + /* + * never_same only acts if + * data begins with diff in first count512s + */ + d +=3D uleb128_encode_small(dst + d, zrun_len); + zrun_len =3D 0; + never_same =3D false; + } + /* has diff */ + if ((bytesToCheck =3D=3D 64) && (comp =3D=3D 0x0)) { + i +=3D 64; + nzrun_len +=3D 64; + break; + } + num =3D __builtin_ctzl(comp); + num =3D (num < bytesToCheck) ? num : bytesToCheck; + nzrun_len +=3D num; + bytesToCheck -=3D num; + comp >>=3D num; + i +=3D num; + if (bytesToCheck) { + /* mask like 111000 */ + d +=3D uleb128_encode_small(dst + d, nzrun_len); + /* overflow */ + if (d + nzrun_len > dlen) { + return -1; + } + nzrun_start =3D new_buf + i - nzrun_len; + memcpy(dst + d, nzrun_start, nzrun_len); + d +=3D nzrun_len; + nzrun_len =3D 0; + is_same =3D true; + } + } + } + if (countResidual) { + /* the number of data is less than 64 */ + uint64_t mask =3D pow(2, countResidual); + mask -=3D 1; + __m512i r =3D SET_ZERO512(r); + __m512i old_data =3D _mm512_mask_loadu_epi8(r, mask, old_buf + i); + __m512i new_data =3D _mm512_mask_loadu_epi8(r, mask, new_buf + i); + __mmask64 comp =3D _mm512_cmpeq_epi8_mask(old_data, new_data); + + int bytesToCheck =3D countResidual; + bool is_same =3D (comp & 0x1); + while (bytesToCheck) { + if (is_same) { + if (nzrun_len) { + d +=3D uleb128_encode_small(dst + d, nzrun_len); + if (d + nzrun_len > dlen) { + return -1; + } + nzrun_start =3D new_buf + i - nzrun_len; + memcpy(dst + d, nzrun_start, nzrun_len); + d +=3D nzrun_len; + nzrun_len =3D 0; + } + never_same =3D false; + num =3D __builtin_ctzl(~comp); + num =3D (num < bytesToCheck) ? num : bytesToCheck; + zrun_len +=3D num; + bytesToCheck -=3D num; + comp >>=3D num; + i +=3D num; + if (bytesToCheck) { + /* diff after same */ + d +=3D uleb128_encode_small(dst + d, zrun_len); + zrun_len =3D 0; + } else { + break; + } + } + + if (never_same || zrun_len) { + d +=3D uleb128_encode_small(dst + d, zrun_len); + zrun_len =3D 0; + never_same =3D false; + } + /* has diff */ + num =3D __builtin_ctzl(comp); + num =3D (num < bytesToCheck) ? num : bytesToCheck; + nzrun_len +=3D num; + bytesToCheck -=3D num; + comp >>=3D num; + i +=3D num; + if (bytesToCheck) { + d +=3D uleb128_encode_small(dst + d, nzrun_len); + /* overflow */ + if (d + nzrun_len > dlen) { + return -1; + } + nzrun_start =3D new_buf + i - nzrun_len; + memcpy(dst + d, nzrun_start, nzrun_len); + d +=3D nzrun_len; + nzrun_len =3D 0; + is_same =3D true; + } + } + } + + if (zrun_len) { + return (zrun_len =3D=3D slen) ? 0 : d; + } + if (nzrun_len !=3D 0) { + d +=3D uleb128_encode_small(dst + d, nzrun_len); + /* overflow */ + if (d + nzrun_len > dlen) { + return -1; + } + nzrun_start =3D new_buf + i - nzrun_len; + memcpy(dst + d, nzrun_start, nzrun_len); + d +=3D nzrun_len; + } + return d; +} +#pragma GCC pop_options +#endif diff --git a/migration/xbzrle.h b/migration/xbzrle.h index a0db507b9c..6feb49160a 100644 --- a/migration/xbzrle.h +++ b/migration/xbzrle.h @@ -18,4 +18,8 @@ int xbzrle_encode_buffer(uint8_t *old_buf, uint8_t *new_b= uf, int slen, uint8_t *dst, int dlen); =20 int xbzrle_decode_buffer(uint8_t *src, int slen, uint8_t *dst, int dlen); +#if defined(CONFIG_AVX512BW_OPT) +int xbzrle_encode_buffer_avx512(uint8_t *old_buf, uint8_t *new_buf, int sl= en, + uint8_t *dst, int dlen); +#endif #endif --=20 2.25.1 From nobody Sun May 5 13:52:27 2024 Delivered-To: importer@patchew.org Authentication-Results: mx.zohomail.com; dkim=pass header.i=@intel.com; spf=pass (zohomail.com: domain of gnu.org designates 209.51.188.17 as permitted sender) smtp.mailfrom=qemu-devel-bounces+importer=patchew.org@nongnu.org; dmarc=pass(p=none dis=none) header.from=intel.com ARC-Seal: i=1; a=rsa-sha256; t=1660034414; cv=none; d=zohomail.com; s=zohoarc; b=D+N+/XCE7IpIW1FrJAIp12XsvYfDkZPmtFqvwUBYE6t32xJsEFpOMJMfOyDm/RZwn2uv/I4ic3eSQ9Btcr1vEuVhwqcR8sUnPb8ip+SgAJdV8gFw2Jkoo+JZ1RQ9vNcnJ90ze0vGNdFUs3nig7UfmQG28SLKHnjpXPmzE5ibSyM= ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=zohomail.com; s=zohoarc; t=1660034414; h=Content-Transfer-Encoding:Cc:Date:From:In-Reply-To:List-Subscribe:List-Post:List-Id:List-Archive:List-Help:List-Unsubscribe:MIME-Version:Message-ID:References:Sender:Subject:To; bh=OLJM+rOZjZHvLA1hxeed+KDq1TB38KawrLBRuUvbPeY=; b=I/70rw6WHI82crevqG/eySDMfVzkeD9eZcnFa87dxk///iew1nxls7/lIfieA5JHEe3N6Fz7pkvKGs1ofFbgEBaY81Acia24MwJ/2IUgniXbV7S9r41zhZq302fYOBUzMts7kY8xzzIl7ySSYPNQhczA902/mS4vujjQHFgvi1I= ARC-Authentication-Results: i=1; mx.zohomail.com; dkim=pass header.i=@intel.com; spf=pass (zohomail.com: domain of gnu.org designates 209.51.188.17 as permitted sender) smtp.mailfrom=qemu-devel-bounces+importer=patchew.org@nongnu.org; dmarc=pass header.from= (p=none dis=none) Return-Path: Received: from lists.gnu.org (lists.gnu.org [209.51.188.17]) by mx.zohomail.com with SMTPS id 1660034414815935.2265512601634; Tue, 9 Aug 2022 01:40:14 -0700 (PDT) Received: from localhost ([::1]:56846 helo=lists1p.gnu.org) by lists.gnu.org with esmtp (Exim 4.90_1) (envelope-from ) id 1oLKmb-0002fq-KZ for importer@patchew.org; Tue, 09 Aug 2022 04:40:13 -0400 Received: from eggs.gnu.org ([2001:470:142:3::10]:56912) by lists.gnu.org with esmtps (TLS1.2:ECDHE_RSA_AES_256_GCM_SHA384:256) (Exim 4.90_1) (envelope-from ) id 1oLKfl-0004d2-Nv for qemu-devel@nongnu.org; Tue, 09 Aug 2022 04:33:14 -0400 Received: from mga01.intel.com ([192.55.52.88]:48329) by eggs.gnu.org with esmtps (TLS1.2:ECDHE_RSA_AES_256_GCM_SHA384:256) (Exim 4.90_1) (envelope-from ) id 1oLKfj-0008VR-1E for qemu-devel@nongnu.org; Tue, 09 Aug 2022 04:33:09 -0400 Received: from orsmga008.jf.intel.com ([10.7.209.65]) by fmsmga101.fm.intel.com with ESMTP/TLS/ECDHE-RSA-AES256-GCM-SHA384; 09 Aug 2022 01:33:05 -0700 Received: from xuling-b360m-d3h.sh.intel.com ([10.239.82.110]) by orsmga008.jf.intel.com with ESMTP; 09 Aug 2022 01:33:02 -0700 DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/simple; d=intel.com; i=@intel.com; q=dns/txt; s=Intel; t=1660033987; x=1691569987; h=from:to:cc:subject:date:message-id:in-reply-to: references:mime-version:content-transfer-encoding; bh=F5sOqkDGHzexi1sf9Jyn32uzYDO1PiXfLlG6b2bZZbU=; b=i3OA3O0084Tj+bi+r16TlWKJztw4XTa9QnC6umy7yMIjTYsMNvoIwMqu OgLXIQN8cOw8A2MlU0FQbjn+VIuv5kx1R6xMU3WsLr1DUNut69Tgse1lL r0L9uAF1kTYpGcL7H2Dkh5WqI85u25l3tO4nXiwV0953vWJU5mZRXwiJ6 qA7fhNq24mNeFvnTzUGPUlTUI9VzeG+9VzCRT53Zes8zBjQ9U2C0MaYLM CmXuDDK7rrUVzcbNr30YQGfzvSF8dhaw0LXSoI4m2bMWqQKn8P/9x3F+H MFFKa/biCP5JGE+nlENEQuhdPCcKCEFdE9Y/sgNii4OCFGfFtR6Movotr g==; X-IronPort-AV: E=McAfee;i="6400,9594,10433"; a="316712891" X-IronPort-AV: E=Sophos;i="5.93,224,1654585200"; d="scan'208";a="316712891" X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="5.93,224,1654585200"; d="scan'208";a="633251192" From: ling xu To: qemu-devel@nongnu.org Cc: quintela@redhat.com, dgilbert@redhat.com, ling xu , Zhou Zhao , Jun Jin Subject: [PATCH v4 2/2] Update test code of AVX512 support for xbzrle_encode Date: Tue, 9 Aug 2022 16:32:49 +0800 Message-Id: <20220809083249.1646952-3-ling1.xu@intel.com> X-Mailer: git-send-email 2.25.1 In-Reply-To: <20220809083249.1646952-1-ling1.xu@intel.com> References: <20220809083249.1646952-1-ling1.xu@intel.com> MIME-Version: 1.0 Content-Transfer-Encoding: quoted-printable Received-SPF: pass (zohomail.com: domain of gnu.org designates 209.51.188.17 as permitted sender) client-ip=209.51.188.17; envelope-from=qemu-devel-bounces+importer=patchew.org@nongnu.org; helo=lists.gnu.org; Received-SPF: pass client-ip=192.55.52.88; envelope-from=ling1.xu@intel.com; helo=mga01.intel.com X-Spam_score_int: -71 X-Spam_score: -7.2 X-Spam_bar: ------- X-Spam_report: (-7.2 / 5.0 requ) BAYES_00=-1.9, DKIMWL_WL_HIGH=-0.082, DKIM_SIGNED=0.1, DKIM_VALID=-0.1, DKIM_VALID_AU=-0.1, DKIM_VALID_EF=-0.1, RCVD_IN_DNSWL_HI=-5, RCVD_IN_MSPIKE_H3=0.001, RCVD_IN_MSPIKE_WL=0.001, SPF_HELO_NONE=0.001, SPF_PASS=-0.001, T_SCC_BODY_TEXT_LINE=-0.01 autolearn=ham autolearn_force=no X-Spam_action: no action X-BeenThere: qemu-devel@nongnu.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: qemu-devel-bounces+importer=patchew.org@nongnu.org Sender: "Qemu-devel" X-ZohoMail-DKIM: pass (identity @intel.com) X-ZM-MESSAGEID: 1660034416503100001 Content-Type: text/plain; charset="utf-8" Signed-off-by: ling xu Co-authored-by: Zhou Zhao Co-authored-by: Jun Jin --- tests/unit/test-xbzrle.c | 458 ++++++++++++++++++++++++++++++++++++++- 1 file changed, 457 insertions(+), 1 deletion(-) diff --git a/tests/unit/test-xbzrle.c b/tests/unit/test-xbzrle.c index ef951b6e54..2676123ce3 100644 --- a/tests/unit/test-xbzrle.c +++ b/tests/unit/test-xbzrle.c @@ -16,6 +16,35 @@ =20 #define XBZRLE_PAGE_SIZE 4096 =20 +#if defined(CONFIG_AVX512BW_OPT) +static bool is_cpu_support_avx512bw; +#include "qemu/cpuid.h" +static void __attribute__((constructor)) init_cpu_flag(void) +{ + unsigned max =3D __get_cpuid_max(0, NULL); + int a, b, c, d; + is_cpu_support_avx512bw =3D false; + if (max >=3D 1) { + __cpuid(1, a, b, c, d); + /* We must check that AVX is not just available, but usable. */ + if ((c & bit_OSXSAVE) && (c & bit_AVX) && max >=3D 7) { + int bv; + __asm("xgetbv" : "=3Da"(bv), "=3Dd"(d) : "c"(0)); + __cpuid_count(7, 0, a, b, c, d); + /* 0xe6: + * XCR0[7:5] =3D 111b (OPMASK state, upper 256-bit of ZMM0-ZMM= 15 + * and ZMM16-ZMM31 state are enabled by OS) + * XCR0[2:1] =3D 11b (XMM state and YMM state are enabled by O= S) + */ + if ((bv & 0xe6) =3D=3D 0xe6 && (b & bit_AVX512BW)) { + is_cpu_support_avx512bw =3D true; + } + } + } + return ; +} +#endif + static void test_uleb(void) { uint32_t i, val; @@ -173,11 +202,438 @@ static void test_encode_decode(void) } } =20 +static float *encode_decode_zero(void) +{ + uint8_t *buffer =3D g_malloc0(XBZRLE_PAGE_SIZE); + uint8_t *compressed =3D g_malloc0(XBZRLE_PAGE_SIZE); + uint8_t *buffer512 =3D g_malloc0(XBZRLE_PAGE_SIZE); + uint8_t *compressed512 =3D g_malloc0(XBZRLE_PAGE_SIZE); + int i =3D 0; + int dlen =3D 0, dlen512 =3D 0; + int diff_len =3D g_test_rand_int_range(0, XBZRLE_PAGE_SIZE - 1006); + + for (i =3D diff_len; i > 0; i--) { + buffer[1000 + i] =3D i; + buffer512[1000 + i] =3D i; + } + + buffer[1000 + diff_len + 3] =3D 103; + buffer[1000 + diff_len + 5] =3D 105; + + buffer512[1000 + diff_len + 3] =3D 103; + buffer512[1000 + diff_len + 5] =3D 105; + + /* encode zero page */ + time_t t_start, t_end, t_start512, t_end512; + t_start =3D clock(); + dlen =3D xbzrle_encode_buffer(buffer, buffer, XBZRLE_PAGE_SIZE, compre= ssed, + XBZRLE_PAGE_SIZE); + t_end =3D clock(); + float time_val =3D difftime(t_end, t_start); + g_assert(dlen =3D=3D 0); + + t_start512 =3D clock(); + dlen512 =3D xbzrle_encode_buffer_avx512(buffer512, buffer512, XBZRLE_P= AGE_SIZE, + compressed512, XBZRLE_PAGE_SIZE); + t_end512 =3D clock(); + float time_val512 =3D difftime(t_end512, t_start512); + g_assert(dlen512 =3D=3D 0); + + static float result_zero[2]; + result_zero[0] =3D time_val; + result_zero[1] =3D time_val512; + + g_free(buffer); + g_free(compressed); + g_free(buffer512); + g_free(compressed512); + + return result_zero; +} + +static void test_encode_decode_zero_avx512(void) +{ + int i; + float time_raw =3D 0.0, time_512 =3D 0.0; + float *res; + for (i =3D 0; i < 10000; i++) { + res =3D encode_decode_zero(); + time_raw +=3D res[0]; + time_512 +=3D res[1]; + } + printf("Zero test:\n"); + printf("Raw xbzrle_encode time is %f ms\n", time_raw); + printf("512 xbzrle_encode time is %f ms\n", time_512); +} + +static float *encode_decode_unchanged(void) +{ + uint8_t *compressed =3D g_malloc0(XBZRLE_PAGE_SIZE); + uint8_t *test =3D g_malloc0(XBZRLE_PAGE_SIZE); + uint8_t *compressed512 =3D g_malloc0(XBZRLE_PAGE_SIZE); + uint8_t *test512 =3D g_malloc0(XBZRLE_PAGE_SIZE); + int i =3D 0; + int dlen =3D 0, dlen512 =3D 0; + int diff_len =3D g_test_rand_int_range(0, XBZRLE_PAGE_SIZE - 1006); + + for (i =3D diff_len; i > 0; i--) { + test[1000 + i] =3D i + 4; + test512[1000 + i] =3D i + 4; + } + + test[1000 + diff_len + 3] =3D 107; + test[1000 + diff_len + 5] =3D 109; + + test512[1000 + diff_len + 3] =3D 107; + test512[1000 + diff_len + 5] =3D 109; + + /* test unchanged buffer */ + time_t t_start, t_end, t_start512, t_end512; + t_start =3D clock(); + dlen =3D xbzrle_encode_buffer(test, test, XBZRLE_PAGE_SIZE, compressed, + XBZRLE_PAGE_SIZE); + t_end =3D clock(); + float time_val =3D difftime(t_end, t_start); + g_assert(dlen =3D=3D 0); + + t_start512 =3D clock(); + dlen512 =3D xbzrle_encode_buffer_avx512(test512, test512, XBZRLE_PAGE_= SIZE, + compressed512, XBZRLE_PAGE_SIZE); + t_end512 =3D clock(); + float time_val512 =3D difftime(t_end512, t_start512); + g_assert(dlen512 =3D=3D 0); + + static float result_unchanged[2]; + result_unchanged[0] =3D time_val; + result_unchanged[1] =3D time_val512; + + g_free(test); + g_free(compressed); + g_free(test512); + g_free(compressed512); + + return result_unchanged; +} + +static void test_encode_decode_unchanged_avx512(void) +{ + int i; + float time_raw =3D 0.0, time_512 =3D 0.0; + float *res; + for (i =3D 0; i < 10000; i++) { + res =3D encode_decode_unchanged(); + time_raw +=3D res[0]; + time_512 +=3D res[1]; + } + printf("Unchanged test:\n"); + printf("Raw xbzrle_encode time is %f ms\n", time_raw); + printf("512 xbzrle_encode time is %f ms\n", time_512); +} + +static float *encode_decode_1_byte(void) +{ + uint8_t *buffer =3D g_malloc0(XBZRLE_PAGE_SIZE); + uint8_t *test =3D g_malloc0(XBZRLE_PAGE_SIZE); + uint8_t *compressed =3D g_malloc(XBZRLE_PAGE_SIZE); + uint8_t *buffer512 =3D g_malloc0(XBZRLE_PAGE_SIZE); + uint8_t *test512 =3D g_malloc0(XBZRLE_PAGE_SIZE); + uint8_t *compressed512 =3D g_malloc(XBZRLE_PAGE_SIZE); + int dlen =3D 0, rc =3D 0, dlen512 =3D 0, rc512 =3D 0; + uint8_t buf[2]; + uint8_t buf512[2]; + + test[XBZRLE_PAGE_SIZE - 1] =3D 1; + test512[XBZRLE_PAGE_SIZE - 1] =3D 1; + + time_t t_start, t_end, t_start512, t_end512; + t_start =3D clock(); + dlen =3D xbzrle_encode_buffer(buffer, test, XBZRLE_PAGE_SIZE, compress= ed, + XBZRLE_PAGE_SIZE); + t_end =3D clock(); + float time_val =3D difftime(t_end, t_start); + g_assert(dlen =3D=3D (uleb128_encode_small(&buf[0], 4095) + 2)); + + rc =3D xbzrle_decode_buffer(compressed, dlen, buffer, XBZRLE_PAGE_SIZE= ); + g_assert(rc =3D=3D XBZRLE_PAGE_SIZE); + g_assert(memcmp(test, buffer, XBZRLE_PAGE_SIZE) =3D=3D 0); + + t_start512 =3D clock(); + dlen512 =3D xbzrle_encode_buffer_avx512(buffer512, test512, XBZRLE_PAG= E_SIZE, + compressed512, XBZRLE_PAGE_SIZE); + t_end512 =3D clock(); + float time_val512 =3D difftime(t_end512, t_start512); + g_assert(dlen512 =3D=3D (uleb128_encode_small(&buf512[0], 4095) + 2)); + + rc512 =3D xbzrle_decode_buffer(compressed512, dlen512, buffer512, + XBZRLE_PAGE_SIZE); + g_assert(rc512 =3D=3D XBZRLE_PAGE_SIZE); + g_assert(memcmp(test512, buffer512, XBZRLE_PAGE_SIZE) =3D=3D 0); + + static float result_1_byte[2]; + result_1_byte[0] =3D time_val; + result_1_byte[1] =3D time_val512; + + g_free(buffer); + g_free(compressed); + g_free(test); + g_free(buffer512); + g_free(compressed512); + g_free(test512); + + return result_1_byte; +} + +static void test_encode_decode_1_byte_avx512(void) +{ + int i; + float time_raw =3D 0.0, time_512 =3D 0.0; + float *res; + for (i =3D 0; i < 10000; i++) { + res =3D encode_decode_1_byte(); + time_raw +=3D res[0]; + time_512 +=3D res[1]; + } + printf("1 byte test:\n"); + printf("Raw xbzrle_encode time is %f ms\n", time_raw); + printf("512 xbzrle_encode time is %f ms\n", time_512); +} + +static float *encode_decode_overflow(void) +{ + uint8_t *compressed =3D g_malloc0(XBZRLE_PAGE_SIZE); + uint8_t *test =3D g_malloc0(XBZRLE_PAGE_SIZE); + uint8_t *buffer =3D g_malloc0(XBZRLE_PAGE_SIZE); + uint8_t *compressed512 =3D g_malloc0(XBZRLE_PAGE_SIZE); + uint8_t *test512 =3D g_malloc0(XBZRLE_PAGE_SIZE); + uint8_t *buffer512 =3D g_malloc0(XBZRLE_PAGE_SIZE); + int i =3D 0, rc =3D 0, rc512 =3D 0; + + for (i =3D 0; i < XBZRLE_PAGE_SIZE / 2 - 1; i++) { + test[i * 2] =3D 1; + test512[i * 2] =3D 1; + } + + /* encode overflow */ + time_t t_start, t_end, t_start512, t_end512; + t_start =3D clock(); + rc =3D xbzrle_encode_buffer(buffer, test, XBZRLE_PAGE_SIZE, compressed, + XBZRLE_PAGE_SIZE); + t_end =3D clock(); + float time_val =3D difftime(t_end, t_start); + g_assert(rc =3D=3D -1); + + t_start512 =3D clock(); + rc512 =3D xbzrle_encode_buffer_avx512(buffer512, test512, XBZRLE_PAGE_= SIZE, + compressed512, XBZRLE_PAGE_SIZE); + t_end512 =3D clock(); + float time_val512 =3D difftime(t_end512, t_start512); + g_assert(rc512 =3D=3D -1); + + static float result_overflow[2]; + result_overflow[0] =3D time_val; + result_overflow[1] =3D time_val512; + + g_free(buffer); + g_free(compressed); + g_free(test); + g_free(buffer512); + g_free(compressed512); + g_free(test512); + + return result_overflow; +} + +static void test_encode_decode_overflow_avx512(void) +{ + int i; + float time_raw =3D 0.0, time_512 =3D 0.0; + float *res; + for (i =3D 0; i < 10000; i++) { + res =3D encode_decode_overflow(); + time_raw +=3D res[0]; + time_512 +=3D res[1]; + } + printf("Overflow test:\n"); + printf("Raw xbzrle_encode time is %f ms\n", time_raw); + printf("512 xbzrle_encode time is %f ms\n", time_512); +} + +static float *encode_decode_range_avx512(void) +{ + uint8_t *buffer =3D g_malloc0(XBZRLE_PAGE_SIZE); + uint8_t *compressed =3D g_malloc(XBZRLE_PAGE_SIZE); + uint8_t *test =3D g_malloc0(XBZRLE_PAGE_SIZE); + uint8_t *buffer512 =3D g_malloc0(XBZRLE_PAGE_SIZE); + uint8_t *compressed512 =3D g_malloc(XBZRLE_PAGE_SIZE); + uint8_t *test512 =3D g_malloc0(XBZRLE_PAGE_SIZE); + int i =3D 0, rc =3D 0, rc512 =3D 0; + int dlen =3D 0, dlen512 =3D 0; + + int diff_len =3D g_test_rand_int_range(0, XBZRLE_PAGE_SIZE - 1006); + + for (i =3D diff_len; i > 0; i--) { + buffer[1000 + i] =3D i; + test[1000 + i] =3D i + 4; + buffer512[1000 + i] =3D i; + test512[1000 + i] =3D i + 4; + } + + buffer[1000 + diff_len + 3] =3D 103; + test[1000 + diff_len + 3] =3D 107; + + buffer[1000 + diff_len + 5] =3D 105; + test[1000 + diff_len + 5] =3D 109; + + buffer512[1000 + diff_len + 3] =3D 103; + test512[1000 + diff_len + 3] =3D 107; + + buffer512[1000 + diff_len + 5] =3D 105; + test512[1000 + diff_len + 5] =3D 109; + + /* test encode/decode */ + time_t t_start, t_end, t_start512, t_end512; + t_start =3D clock(); + dlen =3D xbzrle_encode_buffer(test, buffer, XBZRLE_PAGE_SIZE, compress= ed, + XBZRLE_PAGE_SIZE); + t_end =3D clock(); + float time_val =3D difftime(t_end, t_start); + rc =3D xbzrle_decode_buffer(compressed, dlen, test, XBZRLE_PAGE_SIZE); + g_assert(rc < XBZRLE_PAGE_SIZE); + g_assert(memcmp(test, buffer, XBZRLE_PAGE_SIZE) =3D=3D 0); + + t_start512 =3D clock(); + dlen512 =3D xbzrle_encode_buffer_avx512(test512, buffer512, XBZRLE_PAG= E_SIZE, + compressed512, XBZRLE_PAGE_SIZE); + t_end512 =3D clock(); + float time_val512 =3D difftime(t_end512, t_start512); + rc512 =3D xbzrle_decode_buffer(compressed512, dlen512, test512, XBZRLE= _PAGE_SIZE); + g_assert(rc512 < XBZRLE_PAGE_SIZE); + g_assert(memcmp(test512, buffer512, XBZRLE_PAGE_SIZE) =3D=3D 0); + + static float result_range[2]; + result_range[0] =3D time_val; + result_range[1] =3D time_val512; + + g_free(buffer); + g_free(compressed); + g_free(test); + g_free(buffer512); + g_free(compressed512); + g_free(test512); + + return result_range; +} + +static void test_encode_decode_avx512(void) +{ + int i; + float time_raw =3D 0.0, time_512 =3D 0.0; + float *res; + for (i =3D 0; i < 10000; i++) { + res =3D encode_decode_range_avx512(); + time_raw +=3D res[0]; + time_512 +=3D res[1]; + } + printf("Encode decode test:\n"); + printf("Raw xbzrle_encode time is %f ms\n", time_raw); + printf("512 xbzrle_encode time is %f ms\n", time_512); +} + +static float *encode_decode_random(void) +{ + uint8_t *buffer =3D g_malloc0(XBZRLE_PAGE_SIZE); + uint8_t *compressed =3D g_malloc(XBZRLE_PAGE_SIZE); + uint8_t *test =3D g_malloc0(XBZRLE_PAGE_SIZE); + uint8_t *buffer512 =3D g_malloc0(XBZRLE_PAGE_SIZE); + uint8_t *compressed512 =3D g_malloc(XBZRLE_PAGE_SIZE); + uint8_t *test512 =3D g_malloc0(XBZRLE_PAGE_SIZE); + int i =3D 0, rc =3D 0, rc512 =3D 0; + int dlen =3D 0, dlen512 =3D 0; + + int diff_len =3D g_test_rand_int_range(0, XBZRLE_PAGE_SIZE - 1); + /* store the index of diff */ + int dirty_index[diff_len]; + for (int j =3D 0; j < diff_len; j++) { + dirty_index[j] =3D g_test_rand_int_range(0, XBZRLE_PAGE_SIZE - 1); + } + for (i =3D diff_len - 1; i >=3D 0; i--) { + buffer[dirty_index[i]] =3D i; + test[dirty_index[i]] =3D i + 4; + buffer512[dirty_index[i]] =3D i; + test512[dirty_index[i]] =3D i + 4; + } + + time_t t_start, t_end, t_start512, t_end512; + t_start =3D clock(); + dlen =3D xbzrle_encode_buffer(test, buffer, XBZRLE_PAGE_SIZE, compress= ed, + XBZRLE_PAGE_SIZE); + t_end =3D clock(); + float time_val =3D difftime(t_end, t_start); + rc =3D xbzrle_decode_buffer(compressed, dlen, test, XBZRLE_PAGE_SIZE); + g_assert(rc < XBZRLE_PAGE_SIZE); + + t_start512 =3D clock(); + dlen512 =3D xbzrle_encode_buffer_avx512(test512, buffer512, XBZRLE_PAG= E_SIZE, + compressed512, XBZRLE_PAGE_SIZE); + t_end512 =3D clock(); + float time_val512 =3D difftime(t_end512, t_start512); + rc512 =3D xbzrle_decode_buffer(compressed512, dlen512, test512, XBZRLE= _PAGE_SIZE); + g_assert(rc512 < XBZRLE_PAGE_SIZE); + + static float result_random[2]; + result_random[0] =3D time_val; + result_random[1] =3D time_val512; + + g_free(buffer); + g_free(compressed); + g_free(test); + g_free(buffer512); + g_free(compressed512); + g_free(test512); + + return result_random; +} + +static void test_encode_decode_random_avx512(void) +{ + int i; + float time_raw =3D 0.0, time_512 =3D 0.0; + float *res; + for (i =3D 0; i < 10000; i++) { + res =3D encode_decode_random(); + time_raw +=3D res[0]; + time_512 +=3D res[1]; + } + printf("Random test:\n"); + printf("Raw xbzrle_encode time is %f ms\n", time_raw); + printf("512 xbzrle_encode time is %f ms\n", time_512); +} + int main(int argc, char **argv) { g_test_init(&argc, &argv, NULL); g_test_rand_int(); g_test_add_func("/xbzrle/uleb", test_uleb); + #if defined(CONFIG_AVX512BW_OPT) + if (likely(is_cpu_support_avx512bw)) { + g_test_add_func("/xbzrle/encode_decode_zero", test_encode_decode_z= ero_avx512); + g_test_add_func("/xbzrle/encode_decode_unchanged", + test_encode_decode_unchanged_avx512); + g_test_add_func("/xbzrle/encode_decode_1_byte", test_encode_decode= _1_byte_avx512); + g_test_add_func("/xbzrle/encode_decode_overflow", + test_encode_decode_overflow_avx512); + g_test_add_func("/xbzrle/encode_decode", test_encode_decode_avx512= ); + g_test_add_func("/xbzrle/encode_decode_random", test_encode_decode= _random_avx512); + } else { + g_test_add_func("/xbzrle/encode_decode_zero", test_encode_decode_z= ero); + g_test_add_func("/xbzrle/encode_decode_unchanged", + test_encode_decode_unchanged); + g_test_add_func("/xbzrle/encode_decode_1_byte", test_encode_decode= _1_byte); + g_test_add_func("/xbzrle/encode_decode_overflow", + test_encode_decode_overflow); + g_test_add_func("/xbzrle/encode_decode", test_encode_decode); + } + #else g_test_add_func("/xbzrle/encode_decode_zero", test_encode_decode_zero); g_test_add_func("/xbzrle/encode_decode_unchanged", test_encode_decode_unchanged); @@ -185,6 +641,6 @@ int main(int argc, char **argv) g_test_add_func("/xbzrle/encode_decode_overflow", test_encode_decode_overflow); g_test_add_func("/xbzrle/encode_decode", test_encode_decode); - + #endif return g_test_run(); } --=20 2.25.1