Skip to contentSkip to conversation

Performance improvement in install_sstatepkg

PATCH Applied Mark Hatle 4 participants 6 messages May 10, 2012 – May 11, 2012

lore
Share & tools
lore

Mark Hatle

author
#0 lore raw

cover Performance improvement in install_sstatepkg

Significant performance improvement when large numbers of files are involved with install_sstatepkg.

To reproduce the problem:

bitbake perl
bitbake -c clean perl
time bitbake perl

The following changes since commit 910452727fc277c1caec7612b36c37b58d845350:

eglibc: Add patch to fix /var installation location (2012-05-09 21:42:37 +0100)

are available in the git repository at:
git://git.pokylinux.org/poky-contrib mhatle/sstate
http://git.pokylinux.org/cgit.cgi/poky-contrib/log/?h=mhatle/sstate

Mark Hatle (1):
sstate.bbclass: Improve sstate_installpkg performance

 meta/classes/sstate.bbclass |   28 +++++++++++++++++++++-------
 1 files changed, 21 insertions(+), 7 deletions(-)
Signature · 1 line
1.7.3.4
author
#1 lore raw

PATCH 1/1 sstate.bbclass: Improve sstate_installpkg performance

In a pathological case, lots of files to process, the sstate_installpkg performance was very poor. It interated over each file and ran 3 individual sed commands per file. Changing this to keep iterating but running only a single command took about 1/3 time time.

However, when looking at the corresponding sstate_hardcode_path function, it was clear we could optimize this further.

Using the same encoding logic to specify only the minimumal sed operation necessary, and using xargs to avoid the os.system call the install step was able to be performed in 13% of the original time.

Example timing numbers for perl:

3m7s original code
1m20s single sed, but interating
0m26s using xargs and limited sed

---
 meta/classes/sstate.bbclass |   28 +++++++++++++++++++++-------
 1 files changed, 21 insertions(+), 7 deletions(-)
meta/classes/sstate.bbclass +21 -7
index a8c98e5..ad7d121 100644
@@ -174,18 +174,29 @@ def sstate_installpkg(ss, d):
     bb.build.exec_func('sstate_unpack_package', d)
 
     # Fixup hardcoded paths
+    #
+    # Note: The logic below must match the reverse logic in
+    # sstate_hardcode_path(d)
+
     fixmefn =  sstateinst + "fixmepath"
     if os.path.isfile(fixmefn):
         staging = d.getVar('STAGING_DIR', True)
         staging_target = d.getVar('STAGING_DIR_TARGET', True)
         staging_host = d.getVar('STAGING_DIR_HOST', True)
-        fixmefd = open(fixmefn, "r")
-        fixmefiles = fixmefd.readlines()
-        fixmefd.close()
-        for file in fixmefiles:
-            os.system("sed -i -e s:FIXMESTAGINGDIRTARGET:%s:g %s" % (staging_target, sstateinst + file))
-            os.system("sed -i -e s:FIXMESTAGINGDIRHOST:%s:g %s" % (staging_host, sstateinst + file))
-            os.system("sed -i -e s:FIXMESTAGINGDIR:%s:g %s" % (staging, sstateinst + file))
+
+	if bb.data.inherits_class('native', d) or bb.data.inherits_class('nativesdk', d) or bb.data.inherits_class('crosssdk', d) or bb.data.inherits_class('cross-canadian', d):
+		sstate_sed_cmd = "sed -i -e 's:FIXMESTAGINGDIR:%s:g'" % (staging)
+	elif bb.data.inherits_class('cross', d):
+		sstate_sed_cmd = "sed -i -e 's:FIXMESTAGINGDIRTARGET:%s:g; s:FIXMESTAGINGDIR:%s:g'" % (staging_target, staging)
+	else:
+		sstate_sed_cmd = "sed -i -e 's:FIXMESTAGINGDIRHOST:%s:g'" % (staging_host)
+
+	# Add sstateinst to each filename in fixmepath, use xargs to efficiently call sed
+	sstate_hardcode_cmd = "sed -e 's:^:%s:g' %s | xargs %s" % (sstateinst, fixmefn, sstate_sed_cmd)
+
+	print "Replacing fixme paths in sstate package: %s" % (sstate_hardcode_cmd)
+	os.system(sstate_hardcode_cmd)
+
         # Need to remove this or we'd copy it into the target directory and may 
         # conflict with another writer
         os.remove(fixmefn)
@@ -300,6 +311,9 @@ python sstate_cleanall() {
 def sstate_hardcode_path(d):
 	# Need to remove hardcoded paths and fix these when we install the
 	# staging packages.
+	#
+	# Note: the logic in this function needs to match the reverse logic
+	# in sstate_installpkg(ss, d)
 
 	staging = d.getVar('STAGING_DIR', True)
 	staging_target = d.getVar('STAGING_DIR_TARGET', True)
Signature · 1 line
1.7.3.4
#2 lore raw

Re: [PATCH 1/1] sstate.bbclass: Improve sstate_installpkg performance

Quoted 2 lines · On Thu, 10 May 2012 18:13:38 -0500 Mark Hatle <[email protected]> wrote:
(quoted diff) +1 -0
+	sstate_hardcode_cmd = "sed -e 's:^:%s:g' %s | xargs %s" %

(sstateinst, fixmefn, sstate_sed_cmd)

How confident are we that the file names can never have whitespace in them?

-s

Signature · 1 line
Listen, get this.  Nobody with a good compiler needs to be justified.
#3 lore raw

Re: [PATCH 1/1] sstate.bbclass: Improve sstate_installpkg performance

Quoted 8 lines · Peter Seebach
Quoted 2 lines · On Thu, 10 May 2012 18:13:38 -0500 Mark Hatle <[email protected]> wrote:
(quoted diff) +1 -0
+	sstate_hardcode_cmd = "sed -e 's:^:%s:g' %s | xargs %s" %

(sstateinst, fixmefn, sstate_sed_cmd)

How confident are we that the file names can never have whitespace in them?

Fairly since there are 101 other places this would have broken first.

The day autotools thinks about supporting that, we might start to think about it too. Until then there is little point in worrying about it.

Cheers,

Richard

#4 lore raw

Re: [PATCH 1/1] sstate.bbclass: Improve sstate_installpkg performance

In reply to Mark Hatle
Quoted 65 lines · Mark Hatle

In a pathological case, lots of files to process, the sstate_installpkg performance was very poor. It interated over each file and ran 3 individual sed commands per file. Changing this to keep iterating but running only a single command took about 1/3 time time.

However, when looking at the corresponding sstate_hardcode_path function, it was clear we could optimize this further.

Using the same encoding logic to specify only the minimumal sed operation necessary, and using xargs to avoid the os.system call the install step was able to be performed in 13% of the original time.

Example timing numbers for perl:

3m7s original code
1m20s single sed, but interating
0m26s using xargs and limited sed

---
 meta/classes/sstate.bbclass |   28 +++++++++++++++++++++-------
 1 files changed, 21 insertions(+), 7 deletions(-)
meta/classes/sstate.bbclass +18 -7
index a8c98e5..ad7d121 100644
@@ -174,18 +174,29 @@ def sstate_installpkg(ss, d):
     bb.build.exec_func('sstate_unpack_package', d)
 
     # Fixup hardcoded paths
+    #
+    # Note: The logic below must match the reverse logic in
+    # sstate_hardcode_path(d)
+
     fixmefn =  sstateinst + "fixmepath"
     if os.path.isfile(fixmefn):
         staging = d.getVar('STAGING_DIR', True)
         staging_target = d.getVar('STAGING_DIR_TARGET', True)
         staging_host = d.getVar('STAGING_DIR_HOST', True)
-        fixmefd = open(fixmefn, "r")
-        fixmefiles = fixmefd.readlines()
-        fixmefd.close()
-        for file in fixmefiles:
-            os.system("sed -i -e s:FIXMESTAGINGDIRTARGET:%s:g %s" % (staging_target, sstateinst + file))
-            os.system("sed -i -e s:FIXMESTAGINGDIRHOST:%s:g %s" % (staging_host, sstateinst + file))
-            os.system("sed -i -e s:FIXMESTAGINGDIR:%s:g %s" % (staging, sstateinst + file))
+
+	if bb.data.inherits_class('native', d) or bb.data.inherits_class('nativesdk', d) or bb.data.inherits_class('crosssdk', d) or bb.data.inherits_class('cross-canadian', d):
+		sstate_sed_cmd = "sed -i -e 's:FIXMESTAGINGDIR:%s:g'" % (staging)
+	elif bb.data.inherits_class('cross', d):
+		sstate_sed_cmd = "sed -i -e 's:FIXMESTAGINGDIRTARGET:%s:g; s:FIXMESTAGINGDIR:%s:g'" % (staging_target, staging)
+	else:
+		sstate_sed_cmd = "sed -i -e 's:FIXMESTAGINGDIRHOST:%s:g'" % (staging_host)
+
+	# Add sstateinst to each filename in fixmepath, use xargs to efficiently call sed
+	sstate_hardcode_cmd = "sed -e 's:^:%s:g' %s | xargs %s" % (sstateinst, fixmefn, sstate_sed_cmd)
+
+	print "Replacing fixme paths in sstate package: %s" % (sstate_hardcode_cmd)
+	os.system(sstate_hardcode_cmd)
+
         # Need to remove this or we'd copy it into the target directory and may 
         # conflict with another writer
         os.remove(fixmefn)

meta/classes/sstate.bbclass python sstate_cleanall() { · line 319

Quoted 9 lines
meta/classes/sstate.bbclass +3 -0
@@ -300,6 +311,9 @@ python sstate_cleanall() {
 def sstate_hardcode_path(d):
 	# Need to remove hardcoded paths and fix these when we install the
 	# staging packages.
+	#
+	# Note: the logic in this function needs to match the reverse logic
+	# in sstate_installpkg(ss, d)
 
 	staging = d.getVar('STAGING_DIR', True)
 	staging_target = d.getVar('STAGING_DIR_TARGET', True)

I was thinking this looked familiar, it was the other half of the problem we originally solved:

http://git.yoctoproject.org/cgit.cgi/poky/commit/meta/classes/sstate.bbclass?id=db94ad4cf32d8ce3e97a8287d1c89a58d008a142

:)

Cheers,

Richard

Applied
#5 lore raw
In reply to Mark Hatle
Quoted 17 lines · Mark Hatle

Significant performance improvement when large numbers of files are involved with install_sstatepkg.

To reproduce the problem:

bitbake perl
bitbake -c clean perl
time bitbake perl

The following changes since commit 910452727fc277c1caec7612b36c37b58d845350:

eglibc: Add patch to fix /var installation location (2012-05-09 21:42:37 +0100)

are available in the git repository at:
git://git.pokylinux.org/poky-contrib mhatle/sstate
http://git.pokylinux.org/cgit.cgi/poky-contrib/log/?h=mhatle/sstate

Quoted 6 lines

Mark Hatle (1):
sstate.bbclass: Improve sstate_installpkg performance

  meta/classes/sstate.bbclass |   28 +++++++++++++++++++++-------
  1 files changed, 21 insertions(+), 7 deletions(-)

Nice catch!

Merged into OE-Core

Thanks
Sau!