From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from mail-qt1-f171.google.com (mail-qt1-f171.google.com [209.85.160.171]) (using TLSv1.2 with cipher ECDHE-RSA-AES128-GCM-SHA256 (128/128 bits)) (No client certificate requested) by smtp.subspace.kernel.org (Postfix) with ESMTPS id 9D94321731D for ; Thu, 7 Nov 2024 18:44:34 +0000 (UTC) Authentication-Results: smtp.subspace.kernel.org; arc=none smtp.client-ip=209.85.160.171 ARC-Seal:i=1; a=rsa-sha256; d=subspace.kernel.org; s=arc-20240116; t=1731005076; cv=none; b=gERDqvrlUuAOp2NVf5mJMOIfrXy+tuvIIgtR/4UaKfHeipc+hsOjI85T09kBJ7jecl/2Aq5PYfgvwhtiAJzfdw67QxuAhUQZgWpcCEH4VqMplU0ervdMlxPMkac5hlIPEBznjNzpzfeG6gldCV7Xoc9mUuG50O+XzPbjqe/TDB8= ARC-Message-Signature:i=1; a=rsa-sha256; d=subspace.kernel.org; s=arc-20240116; t=1731005076; c=relaxed/simple; bh=Tam3C3FSWBIBa879uPr6lOw08wkgu4jeV8XEi7er5NA=; h=MIME-Version:References:In-Reply-To:From:Date:Message-ID:Subject: To:Cc:Content-Type; b=u8tejYDj4Y1bepTFNftsv5e/Jo2P85h7vIPKf/4JwaiSyV+7LyV0D7dkhWnQF4QlTk8ykaR9UA24r93K9Es1naYXwHFv/2aWisLfFDdEaqZN0TMiPGBV/o07Nzqf3Nh3/rxhQlitZOaflR8GMzC4uuJyyQHmgOL7IGABlsyeXpI= ARC-Authentication-Results:i=1; smtp.subspace.kernel.org; dmarc=pass (p=reject dis=none) header.from=google.com; spf=pass smtp.mailfrom=google.com; dkim=pass (2048-bit key) header.d=google.com header.i=@google.com header.b=i0yziqJR; arc=none smtp.client-ip=209.85.160.171 Authentication-Results: smtp.subspace.kernel.org; dmarc=pass (p=reject dis=none) header.from=google.com Authentication-Results: smtp.subspace.kernel.org; spf=pass smtp.mailfrom=google.com Authentication-Results: smtp.subspace.kernel.org; dkim=pass (2048-bit key) header.d=google.com header.i=@google.com header.b="i0yziqJR" Received: by mail-qt1-f171.google.com with SMTP id d75a77b69052e-460a8d1a9b7so21111cf.1 for ; Thu, 07 Nov 2024 10:44:34 -0800 (PST) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=google.com; s=20230601; t=1731005073; x=1731609873; darn=vger.kernel.org; h=content-transfer-encoding:cc:to:subject:message-id:date:from :in-reply-to:references:mime-version:from:to:cc:subject:date :message-id:reply-to; bh=J4nPpYXP0vVd6qbOkqNtUFVqt4H36QKYv9JFjcka9R4=; b=i0yziqJRI0ILnBAjYvAjt4vLzPR+UFA6Ww7FdJYeIE3gPS0Vtn+dT3CmfvFdlnaAC+ 2fNSDAkA9MwKUYhyh6CTqP1zDvl2DQq44OdVRpm+pmayYzTrJTztaO5fZ30Jgfw0qg1B fSLEkaeZuLLCLyUUCgpdFzSj196jVKWEzMvFfEfviYnc2RnxBMaD6JenbvwqMAzGE7Y6 1tGMldevX4HH0VAspk48woA+B7QI59I7OeewGTsJcEqrM1sGv61VEkTIgYZWcOCLv2gN v4VSh53JVXdPiKb/W9BOfBZgBO/ZF2cGg0DtvEo1TaYTe6EoRFsSwJlxLEeKPsV7VOEf Ec9Q== X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20230601; t=1731005073; x=1731609873; h=content-transfer-encoding:cc:to:subject:message-id:date:from :in-reply-to:references:mime-version:x-gm-message-state:from:to:cc :subject:date:message-id:reply-to; bh=J4nPpYXP0vVd6qbOkqNtUFVqt4H36QKYv9JFjcka9R4=; b=hP0pQeKtIn750d6onzhIJcVY7cYJumPYk1nmm+Dw7M/rXT+0zSYLBz6Zi77W4P3HT/ 5Rn/bBZKLLvXOels2Z7/l7iy5PfZj9rrPxcIPcEIKM2rMy+FYlitIk/AygeE1xG6rVqa VTB+R56PZ1bszhgQgKLLZ5gfT5AsBn1vVN0JfnEGI4ssQ2wsLRuLv66DsqG4xtsC1qRz TuQk/2hSGtPCLdymaO1ZLT6D2Pz7PIth446f98ZxndWoOpWX98zoy8oTHUVDKn5qpmXV NwpjwIGOpc8whrYM+SugU5ExTwRZdLDOloc/JyXBExf0/1hvtDx9k7DsDhturLpUF7oi z31w== X-Forwarded-Encrypted: i=1; AJvYcCUvQfn8omOzVnoX8hB6eh/W/emgMUexP0ZF+lxbX8vK6lpMYmsyVezN+U/xhZoD8NRskyyGMQ+RCu4=@vger.kernel.org X-Gm-Message-State: AOJu0YwtUEfYDCOsowFa1Fa3ih9E73VOqUhiWWLq1mecno3qpZtI65a3 f7eP3owhUdFj7bpqhMOKa517bKOy7aXag+JxHo/06UsaOhFM1OvygWjUFiwhxVvt99J1r8GAHYR qKElQgRUyVU0iuTg72rdsaWu//1Ty/jFkNXVr X-Gm-Gg: ASbGncu+PtFSG5W3d2LKPuwGMyjkBZht4nGYLyLOh1XOLJvvVkugi0DF71YZ/xiLfNy Ga3yO8Cz9FlTjrmIxLIUHiSBJt+M+mjHkHgji03h/mwUR3La+rRg9A0b+0D1L X-Google-Smtp-Source: AGHT+IHrq73eKO0bY8+FfuWfYvfEG6yvw63fkpfi5hna41D9nyt3oGP9CfN+ewePinNTBwT8SH0tAPXgnDQz/bQ4nhs= X-Received: by 2002:a05:622a:4b0a:b0:462:c158:9f5b with SMTP id d75a77b69052e-462fa610ffbmr4822701cf.19.1731005073310; Thu, 07 Nov 2024 10:44:33 -0800 (PST) Precedence: bulk X-Mailing-List: workflows@vger.kernel.org List-Id: List-Subscribe: List-Unsubscribe: MIME-Version: 1.0 References: <20241102175115.1769468-1-xur@google.com> In-Reply-To: From: Rong Xu Date: Thu, 7 Nov 2024 10:44:21 -0800 Message-ID: Subject: Re: [PATCH v7 0/7] Add AutoFDO and Propeller support for Clang build To: Masahiro Yamada Cc: Alice Ryhl , Andrew Morton , Arnd Bergmann , Bill Wendling , Borislav Petkov , Breno Leitao , Brian Gerst , Dave Hansen , David Li , Han Shen , Heiko Carstens , "H. Peter Anvin" , Ingo Molnar , Jann Horn , Jonathan Corbet , Josh Poimboeuf , Juergen Gross , Justin Stitt , Kees Cook , "Mike Rapoport (IBM)" , Nathan Chancellor , Nick Desaulniers , Nicolas Schier , "Paul E. McKenney" , Peter Zijlstra , Sami Tolvanen , Thomas Gleixner , Wei Yang , workflows@vger.kernel.org, Miguel Ojeda , Maksim Panchenko , "David S. Miller" , Andreas Larsson , Yonghong Song , Yabin Cui , Krzysztof Pszeniczny , Sriraman Tallam , Stephane Eranian , x86@kernel.org, linux-arch@vger.kernel.org, sparclinux@vger.kernel.org, linux-doc@vger.kernel.org, linux-kbuild@vger.kernel.org, linux-kernel@vger.kernel.org, llvm@lists.linux.dev Content-Type: text/plain; charset="UTF-8" Content-Transfer-Encoding: quoted-printable Thanks for the explanation. On Thu, Nov 7, 2024 at 6:58=E2=80=AFAM Masahiro Yamada wrote: > > On Thu, Nov 7, 2024 at 4:00=E2=80=AFAM Rong Xu wrote: > > > > On Wed, Nov 6, 2024 at 8:09=E2=80=AFAM Masahiro Yamada wrote: > > > > > > On Sun, Nov 3, 2024 at 2:51=E2=80=AFAM Rong Xu wrote= : > > > > > > > > Hi, > > > > > > > > This patch series is to integrate AutoFDO and Propeller support int= o > > > > the Linux kernel. AutoFDO is a profile-guided optimization techniqu= e > > > > that leverages hardware sampling to enhance binary performance. > > > > Unlike Instrumentation-based FDO (iFDO), AutoFDO offers a user-frie= ndly > > > > and straightforward application process. While iFDO generally yield= s > > > > superior profile quality and performance, our findings reveal that > > > > AutoFDO achieves remarkable effectiveness, bringing performance clo= se > > > > to iFDO for benchmark applications. > > > > > > > > Propeller is a profile-guided, post-link optimizer that improves > > > > the performance of large-scale applications compiled with LLVM. It > > > > operates by relinking the binary based on an additional round of ru= ntime > > > > profiles, enabling precise optimizations that are not possible at > > > > compile time. Similar to AutoFDO, Propeller too utilizes hardware > > > > sampling to collect profiles and apply post-link optimizations to i= mprove > > > > the benchmark=E2=80=99s performance over and above AutoFDO. > > > > > > > > Our empirical data demonstrates significant performance improvement= s > > > > with AutoFDO and Propeller, up to 10% on microbenchmarks and up to = 5% > > > > on large warehouse-scale benchmarks. This makes a strong case for t= heir > > > > inclusion as supported features in the upstream kernel. > > > > > > > > Background > > > > > > > > A significant fraction of fleet processing cycles (excluding idle t= ime) > > > > from data center workloads are attributable to the kernel. Ware-hou= se > > > > scale workloads maximize performance by optimizing the production k= ernel > > > > using iFDO (a.k.a instrumented PGO, Profile Guided Optimization). > > > > > > > > iFDO can significantly enhance application performance but its use > > > > within the kernel has raised concerns. AutoFDO is a variant of FDO = that > > > > uses the hardware=E2=80=99s Performance Monitoring Unit (PMU) to co= llect > > > > profiling data. While AutoFDO typically yields smaller performance > > > > gains than iFDO, it presents unique benefits for optimizing kernels= . > > > > > > > > AutoFDO eliminates the need for instrumented kernels, allowing a si= ngle > > > > optimized kernel to serve both execution and profile collection. It= also > > > > minimizes slowdown during profile collection, potentially yielding > > > > higher-fidelity profiling, especially for time-sensitive code, comp= ared > > > > to iFDO. Additionally, AutoFDO profiles can be obtained from produc= tion > > > > environments via the hardware=E2=80=99s PMU whereas iFDO profiles r= equire > > > > carefully curated load tests that are representative of real-world > > > > traffic. > > > > > > > > AutoFDO facilitates profile collection across diverse targets. > > > > Preliminary studies indicate significant variation in kernel hot sp= ots > > > > within Google=E2=80=99s infrastructure, suggesting potential perfor= mance gains > > > > through target-specific kernel customization. > > > > > > > > Furthermore, other advanced compiler optimization techniques, inclu= ding > > > > ThinLTO and Propeller can be stacked on top of AutoFDO, similar to = iFDO. > > > > ThinLTO achieves better runtime performance through whole-program > > > > analysis and cross module optimizations. The main difference betwee= n > > > > traditional LTO and ThinLTO is that the latter is scalable in time = and > > > > memory. > > > > > > > > This patch series adds AutoFDO and Propeller support to the kernel.= The > > > > actual solution comes in six parts: > > > > > > > > [P 1] Add the build support for using AutoFDO in Clang > > > > > > > > Add the basic support for AutoFDO build and provide the > > > > instructions for using AutoFDO. > > > > > > > > [P 2] Fix objtool for bogus warnings when -ffunction-sections is en= abled > > > > > > > > [P 3] Adjust symbol ordering in text output sections > > > > > > > > [P 4] Add markers for text_unlikely and text_hot sections > > > > > > > > [P 5] Enable =E2=80=93ffunction-sections for the AutoFDO build > > > > > > > > [P 6] Enable Machine Function Split (MFS) optimization for AutoFDO > > > > > > > > [P 7] Add Propeller configuration to the kernel build > > > > > > > > Patch 1 provides basic AutoFDO build support. Patches 2 to 6 furthe= r > > > > enhance the performance of AutoFDO builds and are functionally depe= ndent > > > > on Patch 1. Patch 7 enables support for Propeller and is dependent = on > > > > patch 2 to patch 4. > > > > > > > > Caveats > > > > > > > > AutoFDO is compatible with both GCC and Clang, but the patches in t= his > > > > series are exclusively applicable to LLVM 17 or newer for AutoFDO a= nd > > > > LLVM 19 or newer for Propeller. For profile conversion, two differe= nt > > > > tools could be used, llvm_profgen or create_llvm_prof. llvm_profgen > > > > needs to be the LLVM 19 or newer, or just the LLVM trunk. Alternati= vely, > > > > create_llvm_prof v0.30.1 or newer can be used instead of llvm-profg= en. > > > > > > > > Additionally, the build is only supported on x86 platforms equipped > > > > with PMU capabilities, such as LBR on Intel machines. More > > > > specifically: > > > > * Intel platforms: works on every platform that supports LBR; > > > > we have tested on Skylake. > > > > * AMD platforms: tested on AMD Zen3 with the BRS feature. The kern= el > > > > needs to be configured with =E2=80=9CCONFIG_PERF_EVENTS_AMD_BRS= =3Dy", To > > > > check, use > > > > $ cat /proc/cpuinfo | grep =E2=80=9C brs=E2=80=9D > > > > For the AMD Zen4, AMD LBRV2 is supported, but we suspect a bug w= ith > > > > AMD LBRv2 implementation in Genoa which blocks the usage. > > > > > > > > For ARM, we plan to send patches for SPE-based Propeller when > > > > AutoFDO for Arm is ready. > > > > > > > > Experiments and Results > > > > > > > > Experiments were conducted to compare the performance of AutoFDO-op= timized > > > > kernel images (version 6.9.x) against default builds.. The evaluati= on > > > > encompassed both open source microbenchmarks and real-world product= ion > > > > services from Google and Meta. The selected microbenchmarks include= d Neper, > > > > a network subsystem benchmark, and UnixBench which is a comprehensi= ve suite > > > > for assessing various kernel operations. > > > > > > > > For Neper, AutoFDO optimization resulted in a 6.1% increase in thro= ughput > > > > and a 10.6% reduction in latency. UnixBench saw a 2.2% improvement = in its > > > > index score under low system load and a 2.6% improvement under high= system > > > > load. > > > > > > > > For further details on the improvements observed in Google and Meta= 's > > > > production services, please refer to the LLVM discourse post: > > > > https://discourse.llvm.org/t/optimizing-the-linux-kernel-with-autof= do-including-thinlto-and-propeller/79108 > > > > > > > > Thanks, > > > > > > > > Rong Xu and Han Shen > > > > > > > > > I applied this series to linux-kbuild. > > > > > > > Thanks for taking the patch! > > > > > As I mentioned before, I do not like #ifdef because > > > it hides (not fixes) issues only for default cases. > > > > We followed the suggestion and removed most of the #if (or #ifdef) in > > the linker script. > > I just checked: there are two #ifdef remaining: > > (1) in the propeller patch for .llvm_bb_addr_map > > (2) in linker script patch for arch/sparc/kernel/vmlinux.lds.S. > > > > I think it's likely safe to remove the checks for head_64.o in > > non-SPARC64 builds and .llvm_bb_addr_map symbols in non-propeller build= s. > > > > SPARC64 builds should always produce head_64.o, and non-SPARC64 > > builds shouldn't. > > > > Propeller builds always generate .llvm_bb_addr_map symbols, and the > > linker will omit the section if it's empty in non-propeller builds. > > > > Keeping the checks is harmless and might slightly reduce linker > > workload for matching. > > But If you'd prefer to remove them, I'm happy to provide a patch. > > > I am talking about the #ifdef in include/asm-generic/vmlinux.lds.h > > > Yeah, it is me who (reluctantly) accepted cb87481ee89d. > > Now, the #ifdef has become a little more complicated. > The default case is safe, but there are hidden issues. > > Some issues are easy to fix, so I sent some patches. > https://lore.kernel.org/linux-kbuild/20241106161445.189399-1-masahiroy@ke= rnel.org/T/#t > https://lore.kernel.org/linux-kbuild/20241106161445.189399-1-masahiroy@ke= rnel.org/T/#m4e4fa70386696e903b68d3fe1d7277e9a63fbefe > https://lore.kernel.org/linux-kbuild/20241107111519.GA15424@willie-the-tr= uck/T/#mccf6d49ddd11c90dcc583d7a68934bb3311da880 I did notice the issues for .data.* -- that is one of the reasons we separated text from data in our patch. > > For example, see e41f501d3912. > > When CONFIG_LD_DEAD_CODE_DATA_ELIMINATION=3Dy or > CONFIG_LTO_CLANG=3Dy or CONFIG_AUTOFDO_CLANG=3Dy or > CONFIG_PROPELLER_CLANG=3Dy, the .text.startup sections > will go to TEXT_MAIN instead of INIT_TEXT. > This is not a fatal issue, but we cannot reuse memory for .text.startup > sections. > > Removing the #ifdef (i.e. reverting cb87481ee89d) is more difficult > because we need to take a closer look at potential impacts for all > architectures. I'm not sure if there is a naming convention for section names in the kerne= l. For special sections, we should avoid using .text.* or .data.*, instead, using "..', or use other prefixes. The compiler can generate sections names like .text.hot.*", ".text.unknown.= *", ".text.unlikely.*", ".text.split.*", ".text.startup." or ".text.exit. It seems we've addressed most of them except .text.startup and .text.exit. For text.startup and .text.exit, have you considered renaming the sections within the linker script -- they are fixed strings and should be able to be rename= d. > > I understood you did not want to take a risk to break random architecture= s, > so I decided to postpone the #ifdef issue and accept your patch set. Thanks for the understanding! > > -- > Best Regards > Masahiro Yamada