From: Randy Dunlap <rdunlap@xenotime.net>
To: "Radosław Smogura" <mail@smogura.eu>
Cc: linux-mm@kvack.org, Yongqiang Yang <xiaoqiangnk@gmail.com>,
linux-ext4@vger.kernel.org
Subject: Re: [PATCH 01/18] Added hacking menu for override optimization by GCC.
Date: Thu, 16 Feb 2012 07:44:54 -0800 [thread overview]
Message-ID: <4F3D2476.4030409@xenotime.net> (raw)
In-Reply-To: <1329402705-25454-1-git-send-email-mail@smogura.eu>
On 02/16/2012 06:31 AM, Radosław Smogura wrote:
> From: mail@smogura.eu <mail@smogura.eu>
>
> This patch gives ability for add some "-fno-..." options for GCC
> and to force -O1 optimization. Supporting files, like Kconfig, Makefile
> are auto-generated due to large amount of available options.
>
> Patch helps to debug kernel.
Note: I only see patches 1-10.
I fix a few typos below (comments in the generated files only).
> ---
> Makefile | 11 ++++
> lib/Kconfig.debug | 2 +
> lib/Kconfig.debug.optim | 102 ++++++++++++++++++++++++++++++++++++
> scripts/Makefile.optim.inc | 23 ++++++++
> scripts/debug/make_config_optim.sh | 88 +++++++++++++++++++++++++++++++
> 5 files changed, 226 insertions(+), 0 deletions(-)
> create mode 100644 lib/Kconfig.debug.optim
> create mode 100644 scripts/Makefile.optim.inc
> create mode 100644 scripts/debug/make_config_optim.sh
>
> diff --git a/Makefile b/Makefile
> index 7c44b67..bc9a961 100644
> --- a/Makefile
> +++ b/Makefile
> @@ -558,12 +558,23 @@ endif # $(dot-config)
> # Defaults to vmlinux, but the arch makefile usually adds further targets
> all: vmlinux
>
> +ifdef CONFIG_HACK_OPTIM_FORCE_O1_LEVEL
> +KBUILD_CFLAGS += -O1
> +else
> +
> ifdef CONFIG_CC_OPTIMIZE_FOR_SIZE
> KBUILD_CFLAGS += -Os
> else
> KBUILD_CFLAGS += -O2
> endif
>
> +endif
> +
> +# Include makefile for optimization override
> +ifdef CONFIG_HACK_OPTIM
> +include $(srctree)/scripts/Makefile.optim.inc
> +endif
> +
> include $(srctree)/arch/$(SRCARCH)/Makefile
>
> ifneq ($(CONFIG_FRAME_WARN),0)
> diff --git a/lib/Kconfig.debug b/lib/Kconfig.debug
> index 8745ac7..928265e 100644
> --- a/lib/Kconfig.debug
> +++ b/lib/Kconfig.debug
> @@ -1274,5 +1274,7 @@ source "lib/Kconfig.kgdb"
>
> source "lib/Kconfig.kmemcheck"
>
> +source "lib/Kconfig.debug.optim"
> +
> config TEST_KSTRTOX
> tristate "Test kstrto*() family of functions at runtime"
> diff --git a/lib/Kconfig.debug.optim b/lib/Kconfig.debug.optim
> new file mode 100644
> index 0000000..09b1012
> --- /dev/null
> +++ b/lib/Kconfig.debug.optim
> @@ -0,0 +1,102 @@
> +# This file was auto generated. It's utility configuration
^^^^^^^^^^^^^^^^^^^^^^^^^^
That sentence(?) (fragment?) is not helpful. I would drop it.
> +# Distributed under GPL v2 License
> +
> +menuconfig HACK_OPTIM
> + bool "Allows to override GCC optimization"
Allows the user to override GCC optimization
> + depends on DEBUG_KERNEL && EXPERIMENTAL
> + help
> + If you say Y here you will be able to override
> + how GCC optimize kernel code. This will create
optimizes
> + more debug friendly, but with not guarentee
friendly code, but will not guarantee
> + about same runi, like production, kernel.
the same running, like a production kernel.
> +
> + If you say Y here probably You will want say
you will want to say Y
> + for all suboptions
for all suboptions.
> +
> +if HACK_OPTIM
> +
> +config HACK_OPTIM_FORCE_O1_LEVEL
> + bool "Forces -O1 optimization level"
> + ---help---
> + This will change how GCC optimize code. Code
optimizes
> + may be slower and larger but will be more debug
> + "friendly".
> +
> + In some cases there is low chance that kernel
> + will run different then normal, reporting or not
will run differently than normal,
> + some bugs or errors. Refere to GCC manual for
Refer
> + more details.
> +
> + You SHOULD say N here.
> +
> +config HACK_OPTIM__fno_inline_functions_called_once
> + bool "Adds -fno-inline-functions-called-once parameter to gcc invoke line."
> + ---help---
> + This will change how GCC optimize code. Code
optimizes
> + may be slower and larger but will be more debug
> + "friendly".
> +
> + In some cases there is low chance that kernel
> + will run different then normal, reporting or not
will run differently than normal,
> + some bugs or errors. Refere to GCC manual for
Refer
> + more details.
> +
> + You SHOULD say N here.
> +
> +config HACK_OPTIM__fno_combine_stack_adjustments
> + bool "Adds -fno-combine-stack-adjustments parameter to gcc invoke line."
> + ---help---
> + This will change how GCC optimize code. Code
optimizes
> + may be slower and larger but will be more debug
> + "friendly".
> +
> + In some cases there is low chance that kernel
> + will run different then normal, reporting or not
will run differently than normal,
> + some bugs or errors. Refere to GCC manual for
Refer
> + more details.
> +
> + You SHOULD say N here.
> +
> +config HACK_OPTIM__fno_tree_dce
> + bool "Adds -fno-tree-dce parameter to gcc invoke line."
> + ---help---
> + This will change how GCC optimize code. Code
optimizes
> + may be slower and larger but will be more debug
> + "friendly".
> +
> + In some cases there is low chance that kernel
> + will run different then normal, reporting or not
will run differently than normal,
> + some bugs or errors. Refere to GCC manual for
Refer
> + more details.
> +
> + You SHOULD say N here.
> +
> +config HACK_OPTIM__fno_tree_dominator_opts
> + bool "Adds -fno-tree-dominator-opts parameter to gcc invoke line."
> + ---help---
> + This will change how GCC optimize code. Code
same ...
> + may be slower and larger but will be more debug
> + "friendly".
> +
> + In some cases there is low chance that kernel
> + will run different then normal, reporting or not
> + some bugs or errors. Refere to GCC manual for
> + more details.
> +
> + You SHOULD say N here.
> +
> +config HACK_OPTIM__fno_dse
> + bool "Adds -fno-dse parameter to gcc invoke line."
> + ---help---
> + This will change how GCC optimize code. Code
same ...
> + may be slower and larger but will be more debug
> + "friendly".
> +
> + In some cases there is low chance that kernel
> + will run different then normal, reporting or not
> + some bugs or errors. Refere to GCC manual for
> + more details.
> +
> + You SHOULD say N here.
> +
> +endif #HACK_OPTIM
> diff --git a/scripts/Makefile.optim.inc b/scripts/Makefile.optim.inc
> new file mode 100644
> index 0000000..e78cc92
> --- /dev/null
> +++ b/scripts/Makefile.optim.inc
> @@ -0,0 +1,23 @@
> +# This file was auto generated. It's utility configuration
> +# Distributed under GPL v2 License
> +
> +ifdef CONFIG_HACK_OPTIM__fno_inline_functions_called_once
> + KBUILD_CFLAGS += -fno-inline-functions-called-once
> +endif
> +
> +ifdef CONFIG_HACK_OPTIM__fno_combine_stack_adjustments
> + KBUILD_CFLAGS += -fno-combine-stack-adjustments
> +endif
> +
> +ifdef CONFIG_HACK_OPTIM__fno_tree_dce
> + KBUILD_CFLAGS += -fno-tree-dce
> +endif
> +
> +ifdef CONFIG_HACK_OPTIM__fno_tree_dominator_opts
> + KBUILD_CFLAGS += -fno-tree-dominator-opts
> +endif
> +
> +ifdef CONFIG_HACK_OPTIM__fno_dse
> + KBUILD_CFLAGS += -fno-dse
> +endif
> +
> diff --git a/scripts/debug/make_config_optim.sh b/scripts/debug/make_config_optim.sh
> new file mode 100644
> index 0000000..26865923
> --- /dev/null
> +++ b/scripts/debug/make_config_optim.sh
> @@ -0,0 +1,88 @@
> +#!/bin/sh
> +
> +## Utility script for generating optimization override options
> +## for kernel compilation.
> +##
> +## Distributed under GPL v2 license
> +## (c) Radosław Smogura, 2011
> +
> +# Prefix added for variable
> +CFG_PREFIX="HACK_OPTIM"
> +
> +KCFG="Kconfig.debug.optim"
> +MKFI="Makefile.optim.inc"
> +
> +OPTIMIZATIONS_PARAMS="-fno-inline-functions-called-once \
> + -fno-combine-stack-adjustments \
> + -fno-tree-dce \
> + -fno-tree-dominator-opts \
> + -fno-dse "
> +
> +echo "# This file was auto generated. It's utility configuration" > $KCFG
> +echo "# Distributed under GPL v2 License" >> $KCFG
> +echo >> $KCFG
> +echo "menuconfig ${CFG_PREFIX}" >> $KCFG
> +echo -e "\tbool \"Allows to override GCC optimization\"" >> $KCFG
> +echo -e "\tdepends on DEBUG_KERNEL && EXPERIMENTAL" >> $KCFG
> +echo -e "\thelp" >> $KCFG
> +echo -e "\t If you say Y here you will be able to override" >> $KCFG
> +echo -e "\t how GCC optimize kernel code. This will create" >> $KCFG
> +echo -e "\t more debug friendly, but with not guarentee" >> $KCFG
> +echo -e "\t about same runi, like production, kernel." >> $KCFG
> +echo >> $KCFG
> +echo -e "\t If you say Y here probably You will want say" >> $KCFG
> +echo -e "\t for all suboptions" >> $KCFG
> +echo >> $KCFG
> +echo "if ${CFG_PREFIX}" >> $KCFG
> +echo >> $KCFG
> +
> +echo "# This file was auto generated. It's utility configuration" > $MKFI
> +echo "# Distributed under GPL v2 License" >> $MKFI
> +echo >> $MKFI
> +
> +# Insert standard override optimization level
> +# This is exception, and this value will not be included
> +# in auto generated makefile. Support for this value
> +# is hard coded in main Makefile.
> +echo -e "config ${CFG_PREFIX}_FORCE_O1_LEVEL" >> $KCFG
> +echo -e "\tbool \"Forces -O1 optimization level\"" >> $KCFG
> +echo -e "\t---help---" >> $KCFG
> +echo -e "\t This will change how GCC optimize code. Code" >> $KCFG
> +echo -e "\t may be slower and larger but will be more debug" >> $KCFG
> +echo -e "\t \"friendly\"." >> $KCFG
> +echo >> $KCFG
> +echo -e "\t In some cases there is low chance that kernel" >> $KCFG
> +echo -e "\t will run different then normal, reporting or not" >> $KCFG
> +echo -e "\t some bugs or errors. Refere to GCC manual for" >> $KCFG
> +echo -e "\t more details." >> $KCFG
> +echo >> $KCFG
> +echo -e "\t You SHOULD say N here." >> $KCFG
> +echo >> $KCFG
> +
> +for o in $OPTIMIZATIONS_PARAMS ; do
> + cfg_o="${CFG_PREFIX}_${o//-/_}";
> + echo "Processing param ${o} config variable will be $cfg_o";
> +
> + # Generate kconfig entry
> + echo -e "config ${cfg_o}" >> $KCFG
> + echo -e "\tbool \"Adds $o parameter to gcc invoke line.\"" >> $KCFG
> + echo -e "\t---help---" >> $KCFG
> + echo -e "\t This will change how GCC optimize code. Code" >> $KCFG
> + echo -e "\t may be slower and larger but will be more debug" >> $KCFG
> + echo -e "\t \"friendly\"." >> $KCFG
> + echo >> $KCFG
> + echo -e "\t In some cases there is low chance that kernel" >> $KCFG
> + echo -e "\t will run different then normal, reporting or not" >> $KCFG
> + echo -e "\t some bugs or errors. Refere to GCC manual for" >> $KCFG
> + echo -e "\t more details." >> $KCFG
> + echo >> $KCFG
> + echo -e "\t You SHOULD say N here." >> $KCFG
> + echo >> $KCFG
> +
> + #Generate Make for include
> + echo "ifdef CONFIG_${cfg_o}" >> $MKFI
> + echo -e "\tKBUILD_CFLAGS += $o" >> $MKFI
> + echo "endif" >> $MKFI
> + echo >> $MKFI
> +done;
> +echo "endif #${CFG_PREFIX}" >> $KCFG
--
~Randy
*** Remember to use Documentation/SubmitChecklist when testing your code ***
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>
next prev parent reply other threads:[~2012-02-16 15:44 UTC|newest]
Thread overview: 26+ messages / expand[flat|nested] mbox.gz Atom feed top
2012-02-16 14:31 [PATCH 01/18] Added hacking menu for override optimization by GCC Radosław Smogura
2012-02-16 14:31 ` [PATCH 02/18] Change of refcounting method for compound pages and atomic heads Radosław Smogura
2012-02-16 14:31 ` [PATCH 03/18] Compound read / write locking aka get / put Radosław Smogura
2012-02-16 14:31 ` [PATCH 04/18] Page flag for tracking compound deque Radosław Smogura
2012-02-16 14:31 ` [PATCH 05/18] Various VM_BUG_ON for securing tail pages usage Radosław Smogura
2012-02-16 14:31 ` [PATCH 06/18] Make migrate pages fucntion more flexible Radosław Smogura
2012-02-16 14:31 ` [PATCH 07/18] Configuration menu for Huge Page Cache Radosław Smogura
2012-02-16 14:31 ` [PATCH 08/18] Generic routines for defragmenting pagecache Radosław Smogura
2012-02-16 14:31 ` [PATCH 09/18] Allowing usage of MAP_HUGETLB in mmap Radosław Smogura
2012-02-16 14:31 ` [PATCH 10/18] Support for huge page faulting Radosław Smogura
2012-02-16 14:31 ` [PATCH 11/18] Basic support (faulting) for huge pages for shmfs Radosław Smogura
2012-02-16 14:31 ` [PATCH 12/18] Additional macros for pmd operations Radosław Smogura
2012-02-16 14:31 ` [PATCH 13/18] Zapping and freeing huge mappings Radosław Smogura
2012-02-16 14:31 ` [PATCH 14/18] Fixes for proc memory Radosław Smogura
2012-02-16 14:31 ` [PATCH 15/18] Splitting and truncating Radosław Smogura
2012-02-16 14:31 ` [PATCH 16/18] SHM: Support for splitting on truncation Radosław Smogura
2012-02-16 14:31 ` [PATCH 17/18] [Experimental] Support for huge pages in EXT 4 Radosław Smogura
2012-02-16 14:31 ` [PATCH 18/18] [WIP] Dummy patch for details Radosław Smogura
2012-02-16 15:44 ` Randy Dunlap [this message]
2012-02-16 16:11 ` [PATCH 01/18] Added hacking menu for override optimization by GCC Radosław Smogura
2012-02-16 19:09 ` Michal Nazarewicz
2012-02-16 20:26 ` Radosław Smogura
2012-02-16 21:59 ` Michal Nazarewicz
2012-02-16 22:40 ` Radosław Smogura
2012-02-16 23:11 ` Michal Nazarewicz
2012-02-17 14:33 ` [PATCH] " Radosław Smogura
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=4F3D2476.4030409@xenotime.net \
--to=rdunlap@xenotime.net \
--cc=linux-ext4@vger.kernel.org \
--cc=linux-mm@kvack.org \
--cc=mail@smogura.eu \
--cc=xiaoqiangnk@gmail.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).