[PATCH 1/5] Use BASIC_FLAGS for profile feedback

2014-07-04 Thread Andi Kleen
From: Andi Kleen Use BASIC_CFLAGS instead of CFLAGS to set up the profile feedback option in the Makefile. This allows still overriding CFLAGS on the make command line without disabling profile feedback. Signed-off-by: Andi Kleen --- Makefile | 4 ++-- 1 file changed, 2 insertions(+), 2

[PATCH 2/5] Don't define away __attribute__ on gcc

2014-07-04 Thread Andi Kleen
From: Andi Kleen Profile feedback sets -DNO_NORETURN, which causes the compat header file to go into a default #else block. That #else block defines away __attribute__(). Doing so causes all kinds of problems with the Linux and gcc system headers: in particular it makes the xmmintrin.h headers

[PATCH 3/5] Run the perf test suite for profile feedback too

2014-07-04 Thread Andi Kleen
From: Andi Kleen Open: If the perf test suite is representative enough it may be reasonable to only run that and skip the much longer full test suite. Thoughts? Signed-off-by: Andi Kleen --- Makefile | 1 + 1 file changed, 1 insertion(+) diff --git a/Makefile b/Makefile index a9770ac

Fix the bitrotted profile feedback build

2014-07-04 Thread Andi Kleen
The profile feedback build support had bitrotted. This patchkit fixes it and improves it slightly by adding a new profile-fast target. -Andi -- To unsubscribe from this list: send the line "unsubscribe git" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.

[PATCH 4/5] Fix profile feedback with -jN and add profile-fast

2014-07-04 Thread Andi Kleen
From: Andi Kleen Profile feedback always failed for me with -jN. The problem was that there was no implicit ordering between the profile generate stage and the profile use stage. So some objects in the later stage would be linked with profile generate objects, and fail due to the missing -lgcov

[PATCH 5/5] Add a little script to compare two make perf runs

2014-07-04 Thread Andi Kleen
From: Andi Kleen Signed-off-by: Andi Kleen --- diff-res | 26 ++ 1 file changed, 26 insertions(+) create mode 100755 diff-res diff --git a/diff-res b/diff-res new file mode 100755 index 000..90d57be --- /dev/null +++ b/diff-res @@ -0,0 +1,26 @@ +#!/usr/bin/python

Re: [PATCH 5/5] Add a little script to compare two make perf runs

2014-07-06 Thread Andi Kleen
> a justification why the geometric mean is used here would increase my > confident significantly. It's just a standard way to summarize sets of benchmarks. For example SPEC uses the same approach. Anyways the script is not essential to the rest of the profile feedback feature. Just ignore it i

Re: [PATCH 3/5] Run the perf test suite for profile feedback too

2014-07-07 Thread Andi Kleen
On Mon, Jul 07, 2014 at 02:06:57PM -0700, Junio C Hamano wrote: > Andi Kleen writes: > > > From: Andi Kleen > > > > Open: If the perf test suite is representative enough it may > > be reasonable to only run that and skip the much longer full > > test sui

Profile feedback patchkit v2

2014-07-07 Thread Andi Kleen
Fix the bitrotted profile feedback support. Changes to v1: - Remove obsolete comment - Remove controversal diff script. -- To unsubscribe from this list: send the line "unsubscribe git" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-

[PATCH 1/4] Use BASIC_FLAGS for profile feedback

2014-07-07 Thread Andi Kleen
From: Andi Kleen Use BASIC_CFLAGS instead of CFLAGS to set up the profile feedback option in the Makefile. This allows still overriding CFLAGS on the make command line without disabling profile feedback. Signed-off-by: Andi Kleen --- Makefile | 4 ++-- 1 file changed, 2 insertions(+), 2

[PATCH 4/4] Fix profile feedback with -jN and add profile-fast

2014-07-07 Thread Andi Kleen
From: Andi Kleen Profile feedback always failed for me with -jN. The problem was that there was no implicit ordering between the profile generate stage and the profile use stage. So some objects in the later stage would be linked with profile generate objects, and fail due to the missing -lgcov

[PATCH 3/4] Run the perf test suite for profile feedback too

2014-07-07 Thread Andi Kleen
From: Andi Kleen Signed-off-by: Andi Kleen --- Makefile | 1 + 1 file changed, 1 insertion(+) diff --git a/Makefile b/Makefile index a9770ac..ba64be9 100644 --- a/Makefile +++ b/Makefile @@ -1647,6 +1647,7 @@ ifeq ($(filter all,$(MAKECMDGOALS)),all) all:: profile-clean $(MAKE

[PATCH 2/4] Don't define away __attribute__ on gcc

2014-07-07 Thread Andi Kleen
From: Andi Kleen Profile feedback sets -DNO_NORETURN, which causes the compat header file to go into a default #else block. That #else block defines away __attribute__(). Doing so causes all kinds of problems with the Linux and gcc system headers: in particular it makes the xmmintrin.h headers

Re: make profile issue on Git 2.1.0

2014-08-19 Thread Andi Kleen
> Maybe we should do this: Looks good to me. > As a side note, while testing this I noticed that the "make perf" run > goes a lot faster if you set GIT_PERF_REPEAT_COUNT=1. This is bad for > actually measuring things, but probably fine for profile feedback. I > don't use the profile builds myself

Re: 2.6.12 hangs on boot

2005-07-18 Thread Andi Kleen
Can you please test if this patch fixes it? -Andi Don't compare linux processor index with APICID Fixes boot up lockups on some machines where CPU apic ids don't start with 0 Signed-off-by: Andi Kleen <[EMAIL PROTECTED]> Index: linux/arch/x86_64/