Changes in Makefile.autofdo

7594302d9ddd9f198173a658250e7b7debc6ed76 - kbuild: rust: rename flag to `-Zdebuginfo-for-profiling` for Rust >= 1.98

Tue, 02 Jun 2026 17:16:38 +0200

kbuild: rust: rename flag to `-Zdebuginfo-for-profiling` for Rust >= 1.98Starting with Rust 1.98.0 (expected 2026-08-20), the`-Zdebug-info-for-profiling` flag has been renamed to`-Zdebuginfo-for-profiling` (i.e. one less dash, to match `debuginfo`sin other flags) [1].Without this change, one gets in the latest nightlies: error: unknown unstable option: `debug-info-for-profiling`Thus pass the right name.Link: https://github.com/rust-lang/rust/pull/156887 [1]Reviewed-by: Alice Ryhl Acked-by: Nathan Chancellor Link: https://patch.msgid.link/20260602151638.14358-1-ojeda@kernel.orgSigned-off-by: Miguel Ojeda List of files: /linux/scripts/Makefile.autofdo

a960c2cdb19e237f284a3b96eb8a0359abcf0e63 - kbuild: rust: add AutoFDO support

Tue, 31 Mar 2026 12:57:49 +0200

kbuild: rust: add AutoFDO supportThis patch enables AutoFDO build support for Rust code within the Linuxkernel. This allows Rust code to be profiled and optimized based on theprofile.The RUSTFLAGS variable was suffixed with *_AUTOFDO_CLANG to match thenaming of the config option, which is called CONFIG_AUTOFDO_CLANG.This implementation has been verified in Android, first by inspectingthe object files and confirming that they look correct. After that,it was verified as below:1. Running the binderAddInts benchmark [1] with Rust Binder built as rust_binder.ko module, using a Pixel 9 Pro.2. Collecting a profile on a Pixel 10 Pro XL using the app-launch benchmark, which starts different apps many times, on a device with Rust Binder as a built-in kernel module. (C Binder was not present on the device.)3. Using the collected profile, run the binderAddInts benchmark again with Rust Binder built both as a rust_binder.ko module, and as a built-in kernel module.4. In both cases, Rust Binder without AutoFDO was approximately 13% slower than the AutoFDO optimized version. Built-in vs .ko did not make a measurable performance difference.All of the above was verified in conjunction with my helpers inliningseries [2], which confirmed that this worked correctly for helpers tooonce [3] was fixed in the helpers inlining series.Link: https://android.googlesource.com/platform/system/extras/+/920f089/tests/binder/benchmarks/binderAddInts.cpp [1]Link: https://lore.kernel.org/r/20260203-inline-helpers-v2-0-beb8547a03c9@google.com [2]Link: https://lore.kernel.org/r/aasPsbMEsX6iGUl8@google.com [3]Reviewed-by: Rong Xu Reviewed-by: Gary Guo Tested-by: Alice Ryhl Signed-off-by: Alice Ryhl Acked-by: Nicolas Schier Acked-by: Nathan Chancellor Link: https://patch.msgid.link/20260331-autofdo-v2-1-eb5c5964820d@google.com[ Reworded for typos. - Miguel ]Signed-off-by: Miguel Ojeda List of files: /linux/scripts/Makefile.autofdo

2fd65f7afd5a73b685a1651cb651ade120b53e15 - AutoFDO: Enable machine function split optimization for AutoFDO

Sat, 02 Nov 2024 18:51:13 +0100

AutoFDO: Enable machine function split optimization for AutoFDOEnable the machine function split optimization for AutoFDO in Clang.Machine function split (MFS) is a pass in the Clang compiler thatsplits a function into hot and cold parts. The linker groups allcold blocks across functions together. This decreases hot codefragmentation and improves iCache and iTLB utilization.MFS requires a profile so this is enabled only for the AutoFDO builds.Co-developed-by: Han Shen Signed-off-by: Han Shen Signed-off-by: Rong Xu Suggested-by: Sriraman Tallam Suggested-by: Krzysztof Pszeniczny Tested-by: Yonghong Song Tested-by: Yabin Cui Tested-by: Nathan Chancellor Reviewed-by: Kees Cook Signed-off-by: Masahiro Yamada List of files: /linux/scripts/Makefile.autofdo

0847420f5e499a7ab518942fff71482179290163 - AutoFDO: Enable -ffunction-sections for the AutoFDO build

Sat, 02 Nov 2024 18:51:12 +0100

AutoFDO: Enable -ffunction-sections for the AutoFDO buildEnable -ffunction-sections by default for the AutoFDO build.With -ffunction-sections, the compiler places each function in its ownsection named .text.function_name instead of placing all functions inthe .text section. In the AutoFDO build, this allows the linker toutilize profile information to reorganize functions for improvedutilization of iCache and iTLB.Co-developed-by: Han Shen Signed-off-by: Han Shen Signed-off-by: Rong Xu Suggested-by: Sriraman Tallam Tested-by: Yonghong Song Tested-by: Yabin Cui Tested-by: Nathan Chancellor Reviewed-by: Kees Cook Signed-off-by: Masahiro Yamada List of files: /linux/scripts/Makefile.autofdo

315ad8780a129e82e2c5c65ee6e970d91a577acb - kbuild: Add AutoFDO support for Clang build

Sat, 02 Nov 2024 18:51:08 +0100

kbuild: Add AutoFDO support for Clang buildAdd the build support for using Clang's AutoFDO. Building the kernelwith AutoFDO does not reduce the optimization level from thecompiler. AutoFDO uses hardware sampling to gather information aboutthe frequency of execution of different code paths within a binary.This information is then used to guide the compiler's optimizationdecisions, resulting in a more efficient binary. Experimentsshowed that the kernel can improve up to 10% in latency.The support requires a Clang compiler after LLVM 17. This submissionis limited to x86 platforms that support PMU features like LBR onIntel machines and AMD Zen3 BRS. Support for SPE on ARM 1, and BRBE on ARM 1 is part of planned future work.Here is an example workflow for AutoFDO kernel:1) Build the kernel on the host machine with LLVM enabled, for example, $ make menuconfig LLVM=1 Turn on AutoFDO build config: CONFIG_AUTOFDO_CLANG=y With a configuration that has LLVM enabled, use the following command: scripts/config -e AUTOFDO_CLANG After getting the config, build with $ make LLVM=12) Install the kernel on the test machine.3) Run the load tests. The '-c' option in perf specifies the sample event period. We suggest using a suitable prime number, like 500009, for this purpose. For Intel platforms: $ perf record -e BR_INST_RETIRED.NEAR_TAKEN:k -a -N -b -c \ -o -- For AMD platforms: The supported system are: Zen3 with BRS, or Zen4 with amd_lbr_v2 For Zen3: $ cat proc/cpuinfo | grep " brs" For Zen4: $ cat proc/cpuinfo | grep amd_lbr_v2 $ perf record --pfm-events RETIRED_TAKEN_BRANCH_INSTRUCTIONS:k -a \ -N -b -c -o -- 4) (Optional) Download the raw perf file to the host machine.5) To generate an AutoFDO profile, two offline tools are available: create_llvm_prof and llvm_profgen. The create_llvm_prof tool is part of the AutoFDO project and can be found on GitHub (https://github.com/google/autofdo), version v0.30.1 or later. The llvm_profgen tool is included in the LLVM compiler itself. It's important to note that the version of llvm_profgen doesn't need to match the version of Clang. It needs to be the LLVM 19 release or later, or from the LLVM trunk. $ llvm-profgen --kernel --binary= --perfdata= \ -o or $ create_llvm_prof --binary= --profile= \ --format=extbinary --out= Note that multiple AutoFDO profile files can be merged into one via: $ llvm-profdata merge -o ... 6) Rebuild the kernel using the AutoFDO profile file with the same config as step 1, (Note CONFIG_AUTOFDO_CLANG needs to be enabled): $ make LLVM=1 CLANG_AUTOFDO_PROFILE=Co-developed-by: Han Shen Signed-off-by: Han Shen Signed-off-by: Rong Xu Suggested-by: Sriraman Tallam Suggested-by: Krzysztof Pszeniczny Suggested-by: Nick Desaulniers Suggested-by: Stephane Eranian Tested-by: Yonghong Song Tested-by: Yabin Cui Tested-by: Nathan Chancellor Reviewed-by: Kees Cook Tested-by: Peter Jung Signed-off-by: Masahiro Yamada List of files: /linux/scripts/Makefile.autofdo