€•{      Œsphinx.addnodes”Œdocument”“”)”}”(Œ	rawsource”Œ ”Œchildren”]”(Œtranslations”ŒLanguagesNode”“”)”}”(hhh]”(h Œpending_xref”“”)”}”(hhh]”Œdocutils.nodes”ŒText”“”ŒChinese (Simplified)”…””}”Œparent”hsbaŒ
attributes”}”(Œids”]”Œclasses”]”Œnames”]”Œdupnames”]”Œbackrefs”]”Œ	refdomain”Œstd”Œreftype”Œdoc”Œ	reftarget”Œ'/translations/zh_CN/dev-tools/propeller”Œmodname”NŒ	classname”NŒrefexplicit”ˆuŒtagname”hhhubh)”}”(hhh]”hŒChinese (Traditional)”…””}”hh2sbah}”(h]”h ]”h"]”h$]”h&]”Œ	refdomain”h)Œreftype”h+Œ	reftarget”Œ'/translations/zh_TW/dev-tools/propeller”Œmodname”NŒ	classname”NŒrefexplicit”ˆuh1hhhubh)”}”(hhh]”hŒItalian”…””}”hhFsbah}”(h]”h ]”h"]”h$]”h&]”Œ	refdomain”h)Œreftype”h+Œ	reftarget”Œ'/translations/it_IT/dev-tools/propeller”Œmodname”NŒ	classname”NŒrefexplicit”ˆuh1hhhubh)”}”(hhh]”hŒJapanese”…””}”hhZsbah}”(h]”h ]”h"]”h$]”h&]”Œ	refdomain”h)Œreftype”h+Œ	reftarget”Œ'/translations/ja_JP/dev-tools/propeller”Œmodname”NŒ	classname”NŒrefexplicit”ˆuh1hhhubh)”}”(hhh]”hŒKorean”…””}”hhnsbah}”(h]”h ]”h"]”h$]”h&]”Œ	refdomain”h)Œreftype”h+Œ	reftarget”Œ'/translations/ko_KR/dev-tools/propeller”Œmodname”NŒ	classname”NŒrefexplicit”ˆuh1hhhubh)”}”(hhh]”hŒPortuguese (Brazilian)”…””}”hh‚sbah}”(h]”h ]”h"]”h$]”h&]”Œ	refdomain”h)Œreftype”h+Œ	reftarget”Œ'/translations/pt_BR/dev-tools/propeller”Œmodname”NŒ	classname”NŒrefexplicit”ˆuh1hhhubh)”}”(hhh]”hŒSpanish”…””}”hh–sbah}”(h]”h ]”h"]”h$]”h&]”Œ	refdomain”h)Œreftype”h+Œ	reftarget”Œ'/translations/sp_SP/dev-tools/propeller”Œmodname”NŒ	classname”NŒrefexplicit”ˆuh1hhhubeh}”(h]”h ]”h"]”h$]”h&]”Œcurrent_language”ŒEnglish”uh1h
hhŒ	_document”hŒsource”NŒline”NubhŒcomment”“”)”}”(hŒ SPDX-License-Identifier: GPL-2.0”h]”hŒ SPDX-License-Identifier: GPL-2.0”…””}”hh·sbah}”(h]”h ]”h"]”h$]”h&]”Œ	xml:space”Œpreserve”uh1hµhhh²hh³ŒA/var/lib/git/docbuild/linux/Documentation/dev-tools/propeller.rst”h´KubhŒsection”“”)”}”(hhh]”(hŒtitle”“”)”}”(hŒ%Using Propeller with the Linux kernel”h]”hŒ%Using Propeller with the Linux kernel”…””}”(hhÏh²hh³Nh´Nubah}”(h]”h ]”h"]”h$]”h&]”uh1hÍhhÊh²hh³hÇh´KubhŒ	paragraph”“”)”}”(hXÅ  This enables Propeller build support for the kernel when using Clang
compiler. Propeller is a profile-guided optimization (PGO) method used
to optimize binary executables. Like AutoFDO, it utilizes hardware
sampling to gather information about the frequency of execution of
different code paths within a binary. Unlike AutoFDO, this information
is then used right before linking phase to optimize (among others)
block layout within and across functions.”h]”hXÅ  This enables Propeller build support for the kernel when using Clang
compiler. Propeller is a profile-guided optimization (PGO) method used
to optimize binary executables. Like AutoFDO, it utilizes hardware
sampling to gather information about the frequency of execution of
different code paths within a binary. Unlike AutoFDO, this information
is then used right before linking phase to optimize (among others)
block layout within and across functions.”…””}”(hhßh²hh³Nh´Nubah}”(h]”h ]”h"]”h$]”h&]”uh1hÝh³hÇh´KhhÊh²hubhÞ)”}”(hŒ<A few important notes about adopting Propeller optimization:”h]”hŒ<A few important notes about adopting Propeller optimization:”…””}”(hhíh²hh³Nh´Nubah}”(h]”h ]”h"]”h$]”h&]”uh1hÝh³hÇh´KhhÊh²hubhŒenumerated_list”“”)”}”(hhh]”(hŒ	list_item”“”)”}”(hŒÐAlthough it can be used as a standalone optimization step, it is
strongly recommended to apply Propeller on top of AutoFDO,
AutoFDO+ThinLTO or Instrument FDO. The rest of this document
assumes this paradigm.
”h]”hÞ)”}”(hŒÏAlthough it can be used as a standalone optimization step, it is
strongly recommended to apply Propeller on top of AutoFDO,
AutoFDO+ThinLTO or Instrument FDO. The rest of this document
assumes this paradigm.”h]”hŒÏAlthough it can be used as a standalone optimization step, it is
strongly recommended to apply Propeller on top of AutoFDO,
AutoFDO+ThinLTO or Instrument FDO. The rest of this document
assumes this paradigm.”…””}”(hj  h²hh³Nh´Nubah}”(h]”h ]”h"]”h$]”h&]”uh1hÝh³hÇh´Khj  ubah}”(h]”h ]”h"]”h$]”h&]”uh1j   hhýh²hh³hÇh´Nubj  )”}”(hŒÄPropeller uses another round of profiling on top of
AutoFDO/AutoFDO+ThinLTO/iFDO. The whole build process involves
"build-afdo - train-afdo - build-propeller - train-propeller -
build-optimized".
”h]”hÞ)”}”(hŒÃPropeller uses another round of profiling on top of
AutoFDO/AutoFDO+ThinLTO/iFDO. The whole build process involves
"build-afdo - train-afdo - build-propeller - train-propeller -
build-optimized".”h]”hŒÇPropeller uses another round of profiling on top of
AutoFDO/AutoFDO+ThinLTO/iFDO. The whole build process involves
â€œbuild-afdo - train-afdo - build-propeller - train-propeller -
build-optimizedâ€.”…””}”(hj  h²hh³Nh´Nubah}”(h]”h ]”h"]”h$]”h&]”uh1hÝh³hÇh´Khj  ubah}”(h]”h ]”h"]”h$]”h&]”uh1j   hhýh²hh³hÇh´Nubj  )”}”(hŒVPropeller requires LLVM 19 release or later for Clang/Clang++
and the linker(ld.lld).
”h]”hÞ)”}”(hŒUPropeller requires LLVM 19 release or later for Clang/Clang++
and the linker(ld.lld).”h]”hŒUPropeller requires LLVM 19 release or later for Clang/Clang++
and the linker(ld.lld).”…””}”(hj6  h²hh³Nh´Nubah}”(h]”h ]”h"]”h$]”h&]”uh1hÝh³hÇh´Khj2  ubah}”(h]”h ]”h"]”h$]”h&]”uh1j   hhýh²hh³hÇh´Nubj  )”}”(hŒyIn addition to LLVM toolchain, Propeller requires a profiling
conversion tool: https://github.com/google/llvm-propeller.
”h]”hÞ)”}”(hŒxIn addition to LLVM toolchain, Propeller requires a profiling
conversion tool: https://github.com/google/llvm-propeller.”h]”(hŒOIn addition to LLVM toolchain, Propeller requires a profiling
conversion tool: ”…””}”(hjN  h²hh³Nh´NubhŒ	reference”“”)”}”(hŒ(https://github.com/google/llvm-propeller”h]”hŒ(https://github.com/google/llvm-propeller”…””}”(hjX  h²hh³Nh´Nubah}”(h]”h ]”h"]”h$]”h&]”Œrefuri”jZ  uh1jV  hjN  ubhŒ.”…””}”(hjN  h²hh³Nh´Nubeh}”(h]”h ]”h"]”h$]”h&]”uh1hÝh³hÇh´KhjJ  ubah}”(h]”h ]”h"]”h$]”h&]”uh1j   hhýh²hh³hÇh´Nubeh}”(h]”h ]”h"]”h$]”h&]”Œenumtype”Œarabic”Œprefix”hŒsuffix”Œ.”uh1hûhhÊh²hh³hÇh´KubhÞ)”}”(hŒRCurrent supported architectures include x86/X86_64 (via LBR),
and arm64 (via SPE).”h]”hŒRCurrent supported architectures include x86/X86_64 (via LBR),
and arm64 (via SPE).”…””}”(hj‚  h²hh³Nh´Nubah}”(h]”h ]”h"]”h$]”h&]”uh1hÝh³hÇh´K!hhÊh²hubhÞ)”}”(hŒ@The Propeller optimization process involves the following steps:”h]”hŒ@The Propeller optimization process involves the following steps:”…””}”(hj  h²hh³Nh´Nubah}”(h]”h ]”h"]”h$]”h&]”uh1hÝh³hÇh´K$hhÊh²hubhü)”}”(hhh]”(j  )”}”(hXn  Initial building: Build the AutoFDO or AutoFDO+ThinLTO binary as
you would normally do, but with a set of compile-time / link-time
flags, so that a special metadata section is created within the
kernel binary. The special section is only intend to be used by the
profiling tool, it is not part of the runtime image, nor does it
change kernel run time text sections.
”h]”hÞ)”}”(hXm  Initial building: Build the AutoFDO or AutoFDO+ThinLTO binary as
you would normally do, but with a set of compile-time / link-time
flags, so that a special metadata section is created within the
kernel binary. The special section is only intend to be used by the
profiling tool, it is not part of the runtime image, nor does it
change kernel run time text sections.”h]”hXm  Initial building: Build the AutoFDO or AutoFDO+ThinLTO binary as
you would normally do, but with a set of compile-time / link-time
flags, so that a special metadata section is created within the
kernel binary. The special section is only intend to be used by the
profiling tool, it is not part of the runtime image, nor does it
change kernel run time text sections.”…””}”(hj¥  h²hh³Nh´Nubah}”(h]”h ]”h"]”h$]”h&]”uh1hÝh³hÇh´K&hj¡  ubah}”(h]”h ]”h"]”h$]”h&]”uh1j   hjž  h²hh³hÇh´Nubj  )”}”(hXs  Profiling: The above kernel is then run with a representative
workload to gather execution frequency data. This data is collected
using hardware sampling, via perf. Propeller is most effective on
platforms supporting advanced PMU features like LBR on Intel
machines. This step is the same as profiling the kernel for AutoFDO
(the exact perf parameters can be different).
”h]”hÞ)”}”(hXr  Profiling: The above kernel is then run with a representative
workload to gather execution frequency data. This data is collected
using hardware sampling, via perf. Propeller is most effective on
platforms supporting advanced PMU features like LBR on Intel
machines. This step is the same as profiling the kernel for AutoFDO
(the exact perf parameters can be different).”h]”hXr  Profiling: The above kernel is then run with a representative
workload to gather execution frequency data. This data is collected
using hardware sampling, via perf. Propeller is most effective on
platforms supporting advanced PMU features like LBR on Intel
machines. This step is the same as profiling the kernel for AutoFDO
(the exact perf parameters can be different).”…””}”(hj½  h²hh³Nh´Nubah}”(h]”h ]”h"]”h$]”h&]”uh1hÝh³hÇh´K-hj¹  ubah}”(h]”h ]”h"]”h$]”h&]”uh1j   hjž  h²hh³hÇh´Nubj  )”}”(hŒqPropeller profile generation: Perf output file is converted to a
pair of Propeller profiles via an offline tool.
”h]”hÞ)”}”(hŒpPropeller profile generation: Perf output file is converted to a
pair of Propeller profiles via an offline tool.”h]”hŒpPropeller profile generation: Perf output file is converted to a
pair of Propeller profiles via an offline tool.”…””}”(hjÕ  h²hh³Nh´Nubah}”(h]”h ]”h"]”h$]”h&]”uh1hÝh³hÇh´K4hjÑ  ubah}”(h]”h ]”h"]”h$]”h&]”uh1j   hjž  h²hh³hÇh´Nubj  )”}”(hXG  Optimized build: Build the AutoFDO or AutoFDO+ThinLTO optimized
binary as you would normally do, but with a compile-time /
link-time flag to pick up the Propeller compile time and link time
profiles. This build step uses 3 profiles - the AutoFDO profile,
the Propeller compile-time profile and the Propeller link-time
profile.
”h]”hÞ)”}”(hXF  Optimized build: Build the AutoFDO or AutoFDO+ThinLTO optimized
binary as you would normally do, but with a compile-time /
link-time flag to pick up the Propeller compile time and link time
profiles. This build step uses 3 profiles - the AutoFDO profile,
the Propeller compile-time profile and the Propeller link-time
profile.”h]”hXF  Optimized build: Build the AutoFDO or AutoFDO+ThinLTO optimized
binary as you would normally do, but with a compile-time /
link-time flag to pick up the Propeller compile time and link time
profiles. This build step uses 3 profiles - the AutoFDO profile,
the Propeller compile-time profile and the Propeller link-time
profile.”…””}”(hjí  h²hh³Nh´Nubah}”(h]”h ]”h"]”h$]”h&]”uh1hÝh³hÇh´K7hjé  ubah}”(h]”h ]”h"]”h$]”h&]”uh1j   hjž  h²hh³hÇh´Nubj  )”}”(hŒDeployment: The optimized kernel binary is deployed and used
in production environments, providing improved performance
and reduced latency.
”h]”hÞ)”}”(hŒŒDeployment: The optimized kernel binary is deployed and used
in production environments, providing improved performance
and reduced latency.”h]”hŒŒDeployment: The optimized kernel binary is deployed and used
in production environments, providing improved performance
and reduced latency.”…””}”(hj  h²hh³Nh´Nubah}”(h]”h ]”h"]”h$]”h&]”uh1hÝh³hÇh´K>hj  ubah}”(h]”h ]”h"]”h$]”h&]”uh1j   hjž  h²hh³hÇh´Nubeh}”(h]”h ]”h"]”h$]”h&]”j}  j~  j  hj€  j  uh1hûhhÊh²hh³hÇh´K&ubhÉ)”}”(hhh]”(hÎ)”}”(hŒPreparation”h]”hŒPreparation”…””}”(hj"  h²hh³Nh´Nubah}”(h]”h ]”h"]”h$]”h&]”uh1hÍhj  h²hh³hÇh´KCubhÞ)”}”(hŒConfigure the kernel with::”h]”hŒConfigure the kernel with:”…””}”(hj0  h²hh³Nh´Nubah}”(h]”h ]”h"]”h$]”h&]”uh1hÝh³hÇh´KEhj  h²hubhŒliteral_block”“”)”}”(hŒ/CONFIG_AUTOFDO_CLANG=y
CONFIG_PROPELLER_CLANG=y”h]”hŒ/CONFIG_AUTOFDO_CLANG=y
CONFIG_PROPELLER_CLANG=y”…””}”hj@  sbah}”(h]”h ]”h"]”h$]”h&]”hÅhÆuh1j>  h³hÇh´KGhj  h²hubeh}”(h]”Œpreparation”ah ]”h"]”Œpreparation”ah$]”h&]”uh1hÈhhÊh²hh³hÇh´KCubhÉ)”}”(hhh]”(hÎ)”}”(hŒCustomization”h]”hŒCustomization”…””}”(hjY  h²hh³Nh´Nubah}”(h]”h ]”h"]”h$]”h&]”uh1hÍhjV  h²hh³hÇh´KKubhÞ)”}”(hX  The default CONFIG_PROPELLER_CLANG setting covers kernel space objects
for Propeller builds. One can, however, enable or disable Propeller build
for individual files and directories by adding a line similar to the
following to the respective kernel Makefile:”h]”hX  The default CONFIG_PROPELLER_CLANG setting covers kernel space objects
for Propeller builds. One can, however, enable or disable Propeller build
for individual files and directories by adding a line similar to the
following to the respective kernel Makefile:”…””}”(hjg  h²hh³Nh´Nubah}”(h]”h ]”h"]”h$]”h&]”uh1hÝh³hÇh´KMhjV  h²hubhŒbullet_list”“”)”}”(hhh]”(j  )”}”(hŒIFor enabling a single file (e.g. foo.o)::

 PROPELLER_PROFILE_foo.o := y
”h]”(hÞ)”}”(hŒ)For enabling a single file (e.g. foo.o)::”h]”hŒ(For enabling a single file (e.g. foo.o):”…””}”(hj~  h²hh³Nh´Nubah}”(h]”h ]”h"]”h$]”h&]”uh1hÝh³hÇh´KRhjz  ubj?  )”}”(hŒPROPELLER_PROFILE_foo.o := y”h]”hŒPROPELLER_PROFILE_foo.o := y”…””}”hjŒ  sbah}”(h]”h ]”h"]”h$]”h&]”hÅhÆuh1j>  h³hÇh´KThjz  ubeh}”(h]”h ]”h"]”h$]”h&]”uh1j   hjw  h²hh³hÇh´Nubj  )”}”(hŒCFor enabling all files in one directory::

 PROPELLER_PROFILE := y
”h]”(hÞ)”}”(hŒ)For enabling all files in one directory::”h]”hŒ(For enabling all files in one directory:”…””}”(hj¤  h²hh³Nh´Nubah}”(h]”h ]”h"]”h$]”h&]”uh1hÝh³hÇh´KVhj   ubj?  )”}”(hŒPROPELLER_PROFILE := y”h]”hŒPROPELLER_PROFILE := y”…””}”hj²  sbah}”(h]”h ]”h"]”h$]”h&]”hÅhÆuh1j>  h³hÇh´KXhj   ubeh}”(h]”h ]”h"]”h$]”h&]”uh1j   hjw  h²hh³hÇh´Nubj  )”}”(hŒ8For disabling one file::

 PROPELLER_PROFILE_foo.o := n
”h]”(hÞ)”}”(hŒFor disabling one file::”h]”hŒFor disabling one file:”…””}”(hjÊ  h²hh³Nh´Nubah}”(h]”h ]”h"]”h$]”h&]”uh1hÝh³hÇh´KZhjÆ  ubj?  )”}”(hŒPROPELLER_PROFILE_foo.o := n”h]”hŒPROPELLER_PROFILE_foo.o := n”…””}”hjØ  sbah}”(h]”h ]”h"]”h$]”h&]”hÅhÆuh1j>  h³hÇh´K\hjÆ  ubeh}”(h]”h ]”h"]”h$]”h&]”uh1j   hjw  h²hh³hÇh´Nubj  )”}”(hŒFFor disabling all files in one directory::

 PROPELLER__PROFILE := n

”h]”(hÞ)”}”(hŒ*For disabling all files in one directory::”h]”hŒ)For disabling all files in one directory:”…””}”(hjð  h²hh³Nh´Nubah}”(h]”h ]”h"]”h$]”h&]”uh1hÝh³hÇh´K^hjì  ubj?  )”}”(hŒPROPELLER__PROFILE := n”h]”hŒPROPELLER__PROFILE := n”…””}”hjþ  sbah}”(h]”h ]”h"]”h$]”h&]”hÅhÆuh1j>  h³hÇh´K`hjì  ubeh}”(h]”h ]”h"]”h$]”h&]”uh1j   hjw  h²hh³hÇh´Nubeh}”(h]”h ]”h"]”h$]”h&]”Œbullet”Œ-”uh1ju  h³hÇh´KRhjV  h²hubeh}”(h]”Œcustomization”ah ]”h"]”Œcustomization”ah$]”h&]”uh1hÈhhÊh²hh³hÇh´KKubhÉ)”}”(hhh]”(hÎ)”}”(hŒWorkflow”h]”hŒWorkflow”…””}”(hj%  h²hh³Nh´Nubah}”(h]”h ]”h"]”h$]”h&]”uh1hÍhj"  h²hh³hÇh´KdubhÞ)”}”(hŒEHere is an example workflow for building an AutoFDO+Propeller kernel:”h]”hŒEHere is an example workflow for building an AutoFDO+Propeller kernel:”…””}”(hj3  h²hh³Nh´Nubah}”(h]”h ]”h"]”h$]”h&]”uh1hÝh³hÇh´Kfhj"  h²hubhü)”}”(hhh]”(j  )”}”(hX1  Assuming an AutoFDO profile is already collected following
instructions in the AutoFDO document, build the kernel on the host
machine, with AutoFDO and Propeller build configs ::

   CONFIG_AUTOFDO_CLANG=y
   CONFIG_PROPELLER_CLANG=y

and ::

   $ make LLVM=1 CLANG_AUTOFDO_PROFILE=<autofdo-profile-name>
”h]”(hÞ)”}”(hŒ²Assuming an AutoFDO profile is already collected following
instructions in the AutoFDO document, build the kernel on the host
machine, with AutoFDO and Propeller build configs ::”h]”hŒ¯Assuming an AutoFDO profile is already collected following
instructions in the AutoFDO document, build the kernel on the host
machine, with AutoFDO and Propeller build configs”…””}”(hjH  h²hh³Nh´Nubah}”(h]”h ]”h"]”h$]”h&]”uh1hÝh³hÇh´KhhjD  ubj?  )”}”(hŒ/CONFIG_AUTOFDO_CLANG=y
CONFIG_PROPELLER_CLANG=y”h]”hŒ/CONFIG_AUTOFDO_CLANG=y
CONFIG_PROPELLER_CLANG=y”…””}”hjV  sbah}”(h]”h ]”h"]”h$]”h&]”hÅhÆuh1j>  h³hÇh´KlhjD  ubhÞ)”}”(hŒand ::”h]”hŒand”…””}”(hjd  h²hh³Nh´Nubah}”(h]”h ]”h"]”h$]”h&]”uh1hÝh³hÇh´KohjD  ubj?  )”}”(hŒ:$ make LLVM=1 CLANG_AUTOFDO_PROFILE=<autofdo-profile-name>”h]”hŒ:$ make LLVM=1 CLANG_AUTOFDO_PROFILE=<autofdo-profile-name>”…””}”hjr  sbah}”(h]”h ]”h"]”h$]”h&]”hÅhÆuh1j>  h³hÇh´KqhjD  ubeh}”(h]”h ]”h"]”h$]”h&]”uh1j   hjA  h²hh³hÇh´Nubj  )”}”(hŒ(Install the kernel on the test machine.
”h]”hÞ)”}”(hŒ'Install the kernel on the test machine.”h]”hŒ'Install the kernel on the test machine.”…””}”(hjŠ  h²hh³Nh´Nubah}”(h]”h ]”h"]”h$]”h&]”uh1hÝh³hÇh´Kshj†  ubah}”(h]”h ]”h"]”h$]”h&]”uh1j   hjA  h²hh³hÇh´Nubj  )”}”(hX™  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 <count> -o <perf_file> -- <loadtest>

- For AMD platforms::

   $ perf record --pfm-event RETIRED_TAKEN_BRANCH_INSTRUCTIONS:k -a -N -b -c <count> -o <perf_file> -- <loadtest>

- For arm64 with SPE::
  There are a few kernel features that must be enabled to collect SPE profiles on Arm.
  Below is a list of the required features:

   - CONFIG_ARM_SPE_PMU=y
   - CONFIG_PID_IN_CONTEXTIDR=y
   - kpti=off

  Use the following command to generate SPE perf data file::

   $ perf record -e 'arm_spe_0/branch_filter=1,load_filter=0,store_filter=0/' -a -N -c <count> --no-switch-events -o <perf_file> -- <loadtest>

  Note you can repeat the above steps to collect multiple <perf_file>s.
”h]”(hÞ)”}”(hŒ—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.”h]”hŒ›Run the load tests. The â€˜-câ€™ option in perf specifies the sample
event period. We suggest using a suitable prime number, like 500009,
for this purpose.”…””}”(hj¢  h²hh³Nh´Nubah}”(h]”h ]”h"]”h$]”h&]”uh1hÝh³hÇh´Kuhjž  ubjv  )”}”(hhh]”(j  )”}”(hŒwFor Intel platforms::

 $ perf record -e BR_INST_RETIRED.NEAR_TAKEN:k -a -N -b -c <count> -o <perf_file> -- <loadtest>
”h]”(hÞ)”}”(hŒFor Intel platforms::”h]”hŒFor Intel platforms:”…””}”(hj·  h²hh³Nh´Nubah}”(h]”h ]”h"]”h$]”h&]”uh1hÝh³hÇh´Kyhj³  ubj?  )”}”(hŒ^$ perf record -e BR_INST_RETIRED.NEAR_TAKEN:k -a -N -b -c <count> -o <perf_file> -- <loadtest>”h]”hŒ^$ perf record -e BR_INST_RETIRED.NEAR_TAKEN:k -a -N -b -c <count> -o <perf_file> -- <loadtest>”…””}”hjÅ  sbah}”(h]”h ]”h"]”h$]”h&]”hÅhÆuh1j>  h³hÇh´K{hj³  ubeh}”(h]”h ]”h"]”h$]”h&]”uh1j   hj°  ubj  )”}”(hŒ…For AMD platforms::

 $ perf record --pfm-event RETIRED_TAKEN_BRANCH_INSTRUCTIONS:k -a -N -b -c <count> -o <perf_file> -- <loadtest>
”h]”(hÞ)”}”(hŒFor AMD platforms::”h]”hŒFor AMD platforms:”…””}”(hjÝ  h²hh³Nh´Nubah}”(h]”h ]”h"]”h$]”h&]”uh1hÝh³hÇh´K}hjÙ  ubj?  )”}”(hŒn$ perf record --pfm-event RETIRED_TAKEN_BRANCH_INSTRUCTIONS:k -a -N -b -c <count> -o <perf_file> -- <loadtest>”h]”hŒn$ perf record --pfm-event RETIRED_TAKEN_BRANCH_INSTRUCTIONS:k -a -N -b -c <count> -o <perf_file> -- <loadtest>”…””}”hjë  sbah}”(h]”h ]”h"]”h$]”h&]”hÅhÆuh1j>  h³hÇh´KhjÙ  ubeh}”(h]”h ]”h"]”h$]”h&]”uh1j   hj°  ubj  )”}”(hXè  For arm64 with SPE::
There are a few kernel features that must be enabled to collect SPE profiles on Arm.
Below is a list of the required features:

 - CONFIG_ARM_SPE_PMU=y
 - CONFIG_PID_IN_CONTEXTIDR=y
 - kpti=off

Use the following command to generate SPE perf data file::

 $ perf record -e 'arm_spe_0/branch_filter=1,load_filter=0,store_filter=0/' -a -N -c <count> --no-switch-events -o <perf_file> -- <loadtest>

Note you can repeat the above steps to collect multiple <perf_file>s.
”h]”(hÞ)”}”(hŒ“For arm64 with SPE::
There are a few kernel features that must be enabled to collect SPE profiles on Arm.
Below is a list of the required features:”h]”hŒ“For arm64 with SPE::
There are a few kernel features that must be enabled to collect SPE profiles on Arm.
Below is a list of the required features:”…””}”(hj  h²hh³Nh´Nubah}”(h]”h ]”h"]”h$]”h&]”uh1hÝh³hÇh´Khjÿ  ubhŒblock_quote”“”)”}”(hŒ?- CONFIG_ARM_SPE_PMU=y
- CONFIG_PID_IN_CONTEXTIDR=y
- kpti=off
”h]”jv  )”}”(hhh]”(j  )”}”(hŒCONFIG_ARM_SPE_PMU=y”h]”hÞ)”}”(hj  h]”hŒCONFIG_ARM_SPE_PMU=y”…””}”(hj  h²hh³Nh´Nubah}”(h]”h ]”h"]”h$]”h&]”uh1hÝh³hÇh´K…hj  ubah}”(h]”h ]”h"]”h$]”h&]”uh1j   hj  ubj  )”}”(hŒCONFIG_PID_IN_CONTEXTIDR=y”h]”hÞ)”}”(hj3  h]”hŒCONFIG_PID_IN_CONTEXTIDR=y”…””}”(hj5  h²hh³Nh´Nubah}”(h]”h ]”h"]”h$]”h&]”uh1hÝh³hÇh´K†hj1  ubah}”(h]”h ]”h"]”h$]”h&]”uh1j   hj  ubj  )”}”(hŒ	kpti=off
”h]”hÞ)”}”(hŒkpti=off”h]”hŒkpti=off”…””}”(hjL  h²hh³Nh´Nubah}”(h]”h ]”h"]”h$]”h&]”uh1hÝh³hÇh´K‡hjH  ubah}”(h]”h ]”h"]”h$]”h&]”uh1j   hj  ubeh}”(h]”h ]”h"]”h$]”h&]”j  j  uh1ju  h³hÇh´K…hj  ubah}”(h]”h ]”h"]”h$]”h&]”uh1j  h³hÇh´K…hjÿ  ubhÞ)”}”(hŒ:Use the following command to generate SPE perf data file::”h]”hŒ9Use the following command to generate SPE perf data file:”…””}”(hjl  h²hh³Nh´Nubah}”(h]”h ]”h"]”h$]”h&]”uh1hÝh³hÇh´K‰hjÿ  ubj?  )”}”(hŒ‹$ perf record -e 'arm_spe_0/branch_filter=1,load_filter=0,store_filter=0/' -a -N -c <count> --no-switch-events -o <perf_file> -- <loadtest>”h]”hŒ‹$ perf record -e 'arm_spe_0/branch_filter=1,load_filter=0,store_filter=0/' -a -N -c <count> --no-switch-events -o <perf_file> -- <loadtest>”…””}”hjz  sbah}”(h]”h ]”h"]”h$]”h&]”hÅhÆuh1j>  h³hÇh´K‹hjÿ  ubhÞ)”}”(hŒENote you can repeat the above steps to collect multiple <perf_file>s.”h]”hŒENote you can repeat the above steps to collect multiple <perf_file>s.”…””}”(hjˆ  h²hh³Nh´Nubah}”(h]”h ]”h"]”h$]”h&]”uh1hÝh³hÇh´Khjÿ  ubeh}”(h]”h ]”h"]”h$]”h&]”uh1j   hj°  ubeh}”(h]”h ]”h"]”h$]”h&]”j  j  uh1ju  h³hÇh´Kyhjž  ubeh}”(h]”h ]”h"]”h$]”h&]”uh1j   hjA  h²hh³Nh´Nubj  )”}”(hŒ>(Optional) Download the raw perf file(s) to the host machine.
”h]”hÞ)”}”(hŒ=(Optional) Download the raw perf file(s) to the host machine.”h]”hŒ=(Optional) Download the raw perf file(s) to the host machine.”…””}”(hj¬  h²hh³Nh´Nubah}”(h]”h ]”h"]”h$]”h&]”uh1hÝh³hÇh´Khj¨  ubah}”(h]”h ]”h"]”h$]”h&]”uh1j   hjA  h²hh³hÇh´Nubj  )”}”(hXÀ  Use the generate_propeller_profiles tool (https://github.com/google/llvm-propeller) to
generate Propeller profile. ::

   $ generate_propeller_profiles \
          --binary=<vmlinux> --profile=<perf_file> \
          --format=propeller --propeller_output_module_name \
          --out=<propeller_profile_prefix>_cc_profile.txt \
          --propeller_symorder=<propeller_profile_prefix>_ld_profile.txt

"<propeller_profile_prefix>" can be something like "/home/user/dir/any_string".

This command generates a pair of Propeller profiles:
"<propeller_profile_prefix>_cc_profile.txt" and
"<propeller_profile_prefix>_ld_profile.txt".

If there are more than 1 perf_file collected in the previous step,
you can create a temp list file "<perf_file_list>" with each line
containing one perf file name and run::

   $ generate_propeller_profiles \
          --binary=<vmlinux> --profile=@<perf_file_list> \
          --format=propeller --propeller_output_module_name \
          --out=<propeller_profile_prefix>_cc_profile.txt \
          --propeller_symorder=<propeller_profile_prefix>_ld_profile.txt

For arm64 SPE, add the option '--profiler=perf_spe', like::

   $ generate_propeller_profiles  \
          --binary=<vmlinux> --profile=<perf_file> \
          --profiler=perf_spe \
          --format=propeller --propeller_output_module_name \
          --out=<propeller_profile_prefix>_cc_profile.txt \
          --propeller_symorder=<propeller_profile_prefix>_ld_profile.txt
”h]”(hÞ)”}”(hŒuUse the generate_propeller_profiles tool (https://github.com/google/llvm-propeller) to
generate Propeller profile. ::”h]”(hŒ*Use the generate_propeller_profiles tool (”…””}”(hjÄ  h²hh³Nh´NubjW  )”}”(hŒ(https://github.com/google/llvm-propeller”h]”hŒ(https://github.com/google/llvm-propeller”…””}”(hjÌ  h²hh³Nh´Nubah}”(h]”h ]”h"]”h$]”h&]”Œrefuri”jÎ  uh1jV  hjÄ  ubhŒ ) to
generate Propeller profile.”…””}”(hjÄ  h²hh³Nh´Nubeh}”(h]”h ]”h"]”h$]”h&]”uh1hÝh³hÇh´K‘hjÀ  ubj?  )”}”(hX  $ generate_propeller_profiles \
       --binary=<vmlinux> --profile=<perf_file> \
       --format=propeller --propeller_output_module_name \
       --out=<propeller_profile_prefix>_cc_profile.txt \
       --propeller_symorder=<propeller_profile_prefix>_ld_profile.txt”h]”hX  $ generate_propeller_profiles \
       --binary=<vmlinux> --profile=<perf_file> \
       --format=propeller --propeller_output_module_name \
       --out=<propeller_profile_prefix>_cc_profile.txt \
       --propeller_symorder=<propeller_profile_prefix>_ld_profile.txt”…””}”hjå  sbah}”(h]”h ]”h"]”h$]”h&]”hÅhÆuh1j>  h³hÇh´K”hjÀ  ubhÞ)”}”(hŒO"<propeller_profile_prefix>" can be something like "/home/user/dir/any_string".”h]”hŒWâ€œ<propeller_profile_prefix>â€ can be something like â€œ/home/user/dir/any_stringâ€.”…””}”(hjó  h²hh³Nh´Nubah}”(h]”h ]”h"]”h$]”h&]”uh1hÝh³hÇh´KšhjÀ  ubhÞ)”}”(hŒ‘This command generates a pair of Propeller profiles:
"<propeller_profile_prefix>_cc_profile.txt" and
"<propeller_profile_prefix>_ld_profile.txt".”h]”hŒ™This command generates a pair of Propeller profiles:
â€œ<propeller_profile_prefix>_cc_profile.txtâ€ and
â€œ<propeller_profile_prefix>_ld_profile.txtâ€.”…””}”(hj  h²hh³Nh´Nubah}”(h]”h ]”h"]”h$]”h&]”uh1hÝh³hÇh´KœhjÀ  ubhÞ)”}”(hŒ¬If there are more than 1 perf_file collected in the previous step,
you can create a temp list file "<perf_file_list>" with each line
containing one perf file name and run::”h]”hŒ¯If there are more than 1 perf_file collected in the previous step,
you can create a temp list file â€œ<perf_file_list>â€ with each line
containing one perf file name and run:”…””}”(hj  h²hh³Nh´Nubah}”(h]”h ]”h"]”h$]”h&]”uh1hÝh³hÇh´K hjÀ  ubj?  )”}”(hX  $ generate_propeller_profiles \
       --binary=<vmlinux> --profile=@<perf_file_list> \
       --format=propeller --propeller_output_module_name \
       --out=<propeller_profile_prefix>_cc_profile.txt \
       --propeller_symorder=<propeller_profile_prefix>_ld_profile.txt”h]”hX  $ generate_propeller_profiles \
       --binary=<vmlinux> --profile=@<perf_file_list> \
       --format=propeller --propeller_output_module_name \
       --out=<propeller_profile_prefix>_cc_profile.txt \
       --propeller_symorder=<propeller_profile_prefix>_ld_profile.txt”…””}”hj  sbah}”(h]”h ]”h"]”h$]”h&]”hÅhÆuh1j>  h³hÇh´K¤hjÀ  ubhÞ)”}”(hŒ;For arm64 SPE, add the option '--profiler=perf_spe', like::”h]”hŒ>For arm64 SPE, add the option â€˜--profiler=perf_speâ€™, like:”…””}”(hj+  h²hh³Nh´Nubah}”(h]”h ]”h"]”h$]”h&]”uh1hÝh³hÇh´KªhjÀ  ubj?  )”}”(hX)  $ generate_propeller_profiles  \
       --binary=<vmlinux> --profile=<perf_file> \
       --profiler=perf_spe \
       --format=propeller --propeller_output_module_name \
       --out=<propeller_profile_prefix>_cc_profile.txt \
       --propeller_symorder=<propeller_profile_prefix>_ld_profile.txt”h]”hX)  $ generate_propeller_profiles  \
       --binary=<vmlinux> --profile=<perf_file> \
       --profiler=perf_spe \
       --format=propeller --propeller_output_module_name \
       --out=<propeller_profile_prefix>_cc_profile.txt \
       --propeller_symorder=<propeller_profile_prefix>_ld_profile.txt”…””}”hj9  sbah}”(h]”h ]”h"]”h$]”h&]”hÅhÆuh1j>  h³hÇh´K¬hjÀ  ubeh}”(h]”h ]”h"]”h$]”h&]”uh1j   hjA  h²hh³hÇh´Nubj  )”}”(hŒïRebuild the kernel using the AutoFDO and Propeller
profiles. ::

   CONFIG_AUTOFDO_CLANG=y
   CONFIG_PROPELLER_CLANG=y

and ::

   $ make LLVM=1 CLANG_AUTOFDO_PROFILE=<profile_file> CLANG_PROPELLER_PROFILE_PREFIX=<propeller_profile_prefix>”h]”(hÞ)”}”(hŒ?Rebuild the kernel using the AutoFDO and Propeller
profiles. ::”h]”hŒ<Rebuild the kernel using the AutoFDO and Propeller
profiles.”…””}”(hjQ  h²hh³Nh´Nubah}”(h]”h ]”h"]”h$]”h&]”uh1hÝh³hÇh´K³hjM  ubj?  )”}”(hŒ/CONFIG_AUTOFDO_CLANG=y
CONFIG_PROPELLER_CLANG=y”h]”hŒ/CONFIG_AUTOFDO_CLANG=y
CONFIG_PROPELLER_CLANG=y”…””}”hj_  sbah}”(h]”h ]”h"]”h$]”h&]”hÅhÆuh1j>  h³hÇh´K¶hjM  ubhÞ)”}”(hŒand ::”h]”hŒand”…””}”(hjm  h²hh³Nh´Nubah}”(h]”h ]”h"]”h$]”h&]”uh1hÝh³hÇh´K¹hjM  ubj?  )”}”(hŒl$ make LLVM=1 CLANG_AUTOFDO_PROFILE=<profile_file> CLANG_PROPELLER_PROFILE_PREFIX=<propeller_profile_prefix>”h]”hŒl$ make LLVM=1 CLANG_AUTOFDO_PROFILE=<profile_file> CLANG_PROPELLER_PROFILE_PREFIX=<propeller_profile_prefix>”…””}”hj{  sbah}”(h]”h ]”h"]”h$]”h&]”hÅhÆuh1j>  h³hÇh´K»hjM  ubeh}”(h]”h ]”h"]”h$]”h&]”uh1j   hjA  h²hh³hÇh´Nubeh}”(h]”h ]”h"]”h$]”h&]”j}  j~  j  hj€  Œ)”uh1hûhj"  h²hh³hÇh´Khubeh}”(h]”Œworkflow”ah ]”h"]”Œworkflow”ah$]”h&]”uh1hÈhhÊh²hh³hÇh´Kdubeh}”(h]”Œ%using-propeller-with-the-linux-kernel”ah ]”h"]”Œ%using propeller with the linux kernel”ah$]”h&]”uh1hÈhhh²hh³hÇh´Kubeh}”(h]”h ]”h"]”h$]”h&]”Œsource”hÇuh1hŒcurrent_source”NŒcurrent_line”NŒsettings”Œdocutils.frontend”ŒValues”“”)”}”(hÍNŒ	generator”NŒ	datestamp”NŒsource_link”NŒ
source_url”NŒtoc_backlinks”Œentry”Œfootnote_backlinks”KŒsectnum_xform”KŒstrip_comments”NŒstrip_elements_with_classes”NŒstrip_classes”NŒreport_level”KŒ
halt_level”KŒexit_status_level”KŒdebug”NŒwarning_stream”NŒ	traceback”ˆŒinput_encoding”Œ	utf-8-sig”Œinput_encoding_error_handler”Œstrict”Œoutput_encoding”Œutf-8”Œoutput_encoding_error_handler”jÉ  Œerror_encoding”Œutf-8”Œerror_encoding_error_handler”Œbackslashreplace”Œlanguage_code”Œen”Œrecord_dependencies”NŒconfig”NŒ	id_prefix”hŒauto_id_prefix”Œid”Œdump_settings”NŒdump_internals”NŒdump_transforms”NŒdump_pseudo_xml”NŒexpose_internals”NŒstrict_visitor”NŒ_disable_config”NŒ_source”hÇŒ_destination”NŒ_config_files”]”Œ7/var/lib/git/docbuild/linux/Documentation/docutils.conf”aŒfile_insertion_enabled”ˆŒraw_enabled”KŒline_length_limit”M'Œpep_references”NŒpep_base_url”Œhttps://peps.python.org/”Œpep_file_url_template”Œpep-%04d”Œrfc_references”NŒrfc_base_url”Œ&https://datatracker.ietf.org/doc/html/”Œ	tab_width”KŒtrim_footnote_reference_space”‰Œsyntax_highlight”Œlong”Œsmart_quotes”ˆŒsmartquotes_locales”]”Œcharacter_level_inline_markup”‰Œdoctitle_xform”‰Œdocinfo_xform”KŒsectsubtitle_xform”‰Œimage_loading”Œlink”Œembed_stylesheet”‰Œcloak_email_addresses”ˆŒsection_self_link”‰Œenv”NubŒreporter”NŒindirect_targets”]”Œsubstitution_defs”}”Œsubstitution_names”}”Œrefnames”}”Œrefids”}”Œnameids”}”(j£  j   jS  jP  j  j  j›  j˜  uŒ	nametypes”}”(j£  ‰jS  ‰j  ‰j›  ‰uh}”(j   hÊjP  j  j  jV  j˜  j"  uŒfootnote_refs”}”Œcitation_refs”}”Œautofootnotes”]”Œautofootnote_refs”]”Œsymbol_footnotes”]”Œsymbol_footnote_refs”]”Œ	footnotes”]”Œ	citations”]”Œautofootnote_start”KŒsymbol_footnote_start”K Œ
id_counter”Œcollections”ŒCounter”“”}”…”R”Œparse_messages”]”Œtransform_messages”]”Œtransformer”NŒinclude_log”]”Œ
decoration”Nh²hub.