Welcome to MilkyWay@home

milkyway code releases

Message boards : Application Code Discussion : milkyway code releases
Message board moderation

To post messages, you must log in.

1 · 2 · 3 · Next

AuthorMessage
Profile Travis
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist

Send message
Joined: 30 Aug 07
Posts: 2046
Credit: 26,480
RAC: 0
Message 6591 - Posted: 24 Nov 2008, 18:29:49 UTC
Last modified: 26 Nov 2008, 15:19:10 UTC

I've made a code releases directory, so all code releases can now be found here:

http://milkyway.cs.rpi.edu/milkyway/download/code_release/
ID: 6591 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
jedirock
Avatar

Send message
Joined: 8 Nov 08
Posts: 178
Credit: 6,140,854
RAC: 0
Message 6637 - Posted: 24 Nov 2008, 22:42:40 UTC - in response to Message 6591.  

Any idea when some work will be available for testing with this?
ID: 6637 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Travis
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist

Send message
Joined: 30 Aug 07
Posts: 2046
Credit: 26,480
RAC: 0
Message 6638 - Posted: 24 Nov 2008, 22:52:02 UTC - in response to Message 6637.  

Any idea when some work will be available for testing with this?


Probably in a few minutes. I'm fixing the bug that was making the feeder crash, which was caused by the workunits having the wrong appid.
ID: 6638 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Michael Roberts
Avatar

Send message
Joined: 25 Dec 07
Posts: 10
Credit: 217,283
RAC: 0
Message 6641 - Posted: 24 Nov 2008, 23:04:52 UTC - in response to Message 6591.  

Some updates to make.linux for v0.3

  • definition of INC was missing - I've taken that from v0.2
  • please lose the stray tab on the line after the definition of VARIANTFLAGS - make syntax is horrible!
  • definitions of BOINC_API_DIR and BOINC_LIB_DIR are not used
  • an alternative commented-out definition for BOINC_LINK_DIR would be helpful
  • I still suggest a few usage comments!
  • I have prepared a new file Makefile for local definitions which invokes the makefile I need - that means I don't have to make local alterations to make.linux - I show an example below


Here is a patch with the above changes: http://www.pastebin.ca/1266196

Here is the example for Makefile: http://www.pastebin.ca/1266207

ID: 6641 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Travis
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist

Send message
Joined: 30 Aug 07
Posts: 2046
Credit: 26,480
RAC: 0
Message 6643 - Posted: 24 Nov 2008, 23:29:42 UTC - in response to Message 6641.  

Some updates to make.linux for v0.3

  • definition of INC was missing - I've taken that from v0.2
  • please lose the stray tab on the line after the definition of VARIANTFLAGS - make syntax is horrible!
  • definitions of BOINC_API_DIR and BOINC_LIB_DIR are not used
  • an alternative commented-out definition for BOINC_LINK_DIR would be helpful
  • I still suggest a few usage comments!
  • I have prepared a new file Makefile for local definitions which invokes the makefile I need - that means I don't have to make local alterations to make.linux - I show an example below


Here is a patch with the above changes: http://www.pastebin.ca/1266196

Here is the example for Makefile: http://www.pastebin.ca/1266207



Thanks! I'll fix the makefile with these changes and post here when it's updated :)
ID: 6643 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
jedirock
Avatar

Send message
Joined: 8 Nov 08
Posts: 178
Credit: 6,140,854
RAC: 0
Message 6692 - Posted: 25 Nov 2008, 16:50:20 UTC
Last modified: 25 Nov 2008, 16:53:57 UTC

Just some preliminary tests:

  • Stock 0.2 application took about 20 minutes to crunch one WU
  • Slightly optimized 0.2 application with -O2 and SSE enabled too 16.5 minutes to crunch one WU
  • More optimized 0.3 application (-O2, SSE, tree-vectorization, fast math, and unroll loops) is on track for 13.5 minutes for the first WU


Means about 33% speedup using optimized 0.3 application over the 0.2 stock application. I think you'll have to lower your credits again. ;)

Measurements made on a 2GHz Core Duo Mac running Mac OS 10.5.5, 32-bit support only.

EDIT: Yep, first WU crunched at 13.5 minutes. 13m27s to be exact.

ID: 6692 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Travis
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist

Send message
Joined: 30 Aug 07
Posts: 2046
Credit: 26,480
RAC: 0
Message 6694 - Posted: 25 Nov 2008, 16:52:50 UTC - in response to Message 6643.  

I updated the 0.3 code a bit (i think i fixed a bug that was causing some of the windows clients to crash). We'll be releasing the 0.3 app this afternoon.

I also made a couple changes to the make.linux file (to make it emacs friendly).
ID: 6694 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
jedirock
Avatar

Send message
Joined: 8 Nov 08
Posts: 178
Credit: 6,140,854
RAC: 0
Message 6698 - Posted: 25 Nov 2008, 17:01:40 UTC - in response to Message 6692.  

I think you'll have to lower your credits again. ;)

Hmm, maybe not. Just calculating the credit for the first WU I've returned using a self compiled and GCC optimized 0.3 app, and it works out to 44.45 credits per hour, which is less than the average I get on ABC using my 64-bit Linux machine (~60 credits/hour if anyone was wondering). So from first looks, the credit situation should be okay.
ID: 6698 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Travis
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist

Send message
Joined: 30 Aug 07
Posts: 2046
Credit: 26,480
RAC: 0
Message 6710 - Posted: 25 Nov 2008, 18:45:01 UTC - in response to Message 6692.  

Just some preliminary tests:

  • Stock 0.2 application took about 20 minutes to crunch one WU
  • Slightly optimized 0.2 application with -O2 and SSE enabled too 16.5 minutes to crunch one WU
  • More optimized 0.3 application (-O2, SSE, tree-vectorization, fast math, and unroll loops) is on track for 13.5 minutes for the first WU


Means about 33% speedup using optimized 0.3 application over the 0.2 stock application. I think you'll have to lower your credits again. ;)

Measurements made on a 2GHz Core Duo Mac running Mac OS 10.5.5, 32-bit support only.

EDIT: Yep, first WU crunched at 13.5 minutes. 13m27s to be exact.



how does this compare to milksop's version?
ID: 6710 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
jedirock
Avatar

Send message
Joined: 8 Nov 08
Posts: 178
Credit: 6,140,854
RAC: 0
Message 6712 - Posted: 25 Nov 2008, 18:57:39 UTC - in response to Message 6710.  

how does this compare to milksop's version?

If you mean the version that was optimized before, I can't say. I never tried running it. If it's another optimized version here, I haven't tried it yet. All the code I've been running has either been stock or my own compilations.
ID: 6712 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Sabroe_SMC
Avatar

Send message
Joined: 2 Aug 08
Posts: 24
Credit: 374,440,641
RAC: 0
Message 6715 - Posted: 25 Nov 2008, 19:12:40 UTC - in response to Message 6710.  

Milksops Version takes about 5 min 7 sec on a 3 Ghz Quad under Vista 64.
Cheers Sabroe SMC
ID: 6715 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Milksop at try

Send message
Joined: 1 Oct 08
Posts: 106
Credit: 24,162,445
RAC: 0
Message 6725 - Posted: 25 Nov 2008, 19:38:59 UTC - in response to Message 6715.  
Last modified: 25 Nov 2008, 19:46:28 UTC

Milksops Version takes about 5 min 7 sec on a 3 Ghz Quad under Vista 64.
Cheers Sabroe SMC

That is a number for the published 32bit compilation without any SSE. I have used VC98 to slow it down to the performance of the linux build (using gcc3.04) ;)

When compiling it with a newer intel Compiler with SSE2 enabled it would have been roughly a factor of two faster on newer CPUs. But the compatibility would have been worse.
ID: 6725 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Thierry Godefroy

Send message
Joined: 29 Jul 08
Posts: 9
Credit: 2,200,784
RAC: 0
Message 6743 - Posted: 26 Nov 2008, 0:59:01 UTC
Last modified: 26 Nov 2008, 1:00:26 UTC

I managed to compile the code (with BOINC compiled in /usr/src/rpm/BUILD/BOINC) for 32bits Linux. Here is my Makefile:
#-*- Makefile -*-
#
# usage:
#
#  make -f make.linux app_x86_64
#
#  make -f make.linux app_i686
#
#
# change BOINC_DIR locally or define it on the make command line
#
APP_VERSION = 0.3

BOINC_DIR = /usr/src/rpm/BUILD/boinc
BOINC_API_DIR = $(BOINC_DIR)/api
BOINC_LIB_DIR = $(BOINC_DIR)/lib
BOINC_LINK_DIR = /usr/src/rpm/BUILD/boinc
BOINC_LIB_LINK_DIR = $(BOINC_LINK_DIR)/lib
BOINC_API_LINK_DIR = $(BOINC_LINK_DIR)/api

VARIANTFLAGS = -DGMLE_BOINC

CXXFLAGS_ALL = $(VARIANTFLAGS) \
	-O2 -fomit-frame-pointer -frename-registers -fweb -fexpensive-optimizations -fno-strict-aliasing -march=i686 -msse3 -mfpmath=sse -pipe \
	-I$(BOINC_DIR) -I$(BOINC_LIB_DIR) -I$(BOINC_API_DIR)

CXX_i686 = g++
CXXFLAGS_i686 = -m32 $(CXXFLAGS_ALL)
LDFLAGS_i686 = -L/usr/X11R6/lib -L$(BOINC_LIB_LINK_DIR) -L$(BOINC_API_LINK_DIR)

CXX_x86_64 = g++
CXXFLAGS_x86_64 = $(CXXFLAGS_ALL)
LDFLAGS_x86_64 = -L/usr/local/lib
#LDFLAGS_x86_64 = -L/usr/local/lib -L/usr/X11R6/lib
#LDFLAGS_x86_64 = -L$(BOINC_LIB_LINK_DIR) -L$(BOINC_API_LINK_DIR)


APP_DIR = ../astronomy
FGDO_DIR = ..
SCHED_DIR = $(FGDO_DIR)/evaluation
SEARCH_DIR = $(FGDO_DIR)/searches
UTIL_DIR = $(FGDO_DIR)/util

APP_OBJS = \
	$(APP_DIR)/boinc_astronomy.o \
	$(APP_DIR)/atSurveyGeometry.o \
	$(APP_DIR)/numericalIntegration.o \
	$(APP_DIR)/parameters.o \
	$(APP_DIR)/probability.o \
	$(APP_DIR)/stCoords.o \
	$(APP_DIR)/stCnum.o \
	$(APP_DIR)/stMath.o \
	$(APP_DIR)/stVector.o \
	$(APP_DIR)/star_points.o \
	$(APP_DIR)/evaluation_optimized.o \
	$(APP_DIR)/evaluation_state.o
SEARCH_OBJS = \
	$(SEARCH_DIR)/search_parameters.o
UTIL_OBJS = \
	$(UTIL_DIR)/io_util.o \
	$(UTIL_DIR)/settings.o

PROGS = milkyway_$(APP_VERSION)_i686-pc-linux-gnu milkyway_$(APP_VERSION)_x86_64-pc-linux-gnu

all: $(PROGS)

app_i686: OBJ_CXX = $(CXX_i686)
app_i686: OBJ_CXXFLAGS = $(CXXFLAGS_i686)
app_i686: $(APP_OBJS) $(SEARCH_OBJS) $(UTIL_OBJS) $(BOINC_API_LINK_DIR)/libboinc_api.a $(BOINC_LIB_LINK_DIR)/libboinc.a
	$(CXX_i686) $(LDFLAGS_i686) $(CXXFLAGS_i686) -Wl --export_dynamic -o milkyway_$(APP_VERSION)_i686-pc-linux-gnu $(APP_OBJS) $(SEARCH_OBJS) $(UTIL_OBJS) -lm -lboinc_api -lboinc -pthread

app_x86_64: OBJ_CXX = $(CXX_x86_64)
app_x86_64: OBJ_CXXFLAGS = $(CXXFLAGS_x86_64)
app_x86_64: $(APP_OBJS) $(SEARCH_OBJS) $(UTIL_OBJS) $(BOINC_API_LINK_DIR)/libboinc_api.a $(BOINC_LIB_LINK_DIR)/libboinc.a
	$(CXX_x86_64) $(LDFLAGS_x86_64) $(CXXFLAGS_x86_64) -Wl --export_dynamic -o milkyway_$(APP_VERSION)_x86_64-pc-linux-gnu $(APP_OBJS) $(SEARCH_OBJS) $(UTIL_OBJS) -lm -lboinc_api -lboinc -pthread

.C.o:
	$(OBJ_CXX) $(OBJ_CXXFLAGS) $(INC) -Wall -x c++ -c $< -o $@
.c.o:
	$(OBJ_CXX) $(OBJ_CXXFLAGS) $(INC) -Wall -x c++ -c $< -o $@

clean:
	rm -f $(APP_OBJS) $(UTIL_OBJS) $(SEARCH_OBJS);

clean_all:
	rm -f $(PROGS) $(APP_OBJS) $(UTIL_OBJS) $(SEARCH_OBJS);


However, it looks like I can't get BOINC to download WUs for the new app, it reports:
mer. 26 nov. 2008 01:53:35 CET|Milkyway@home|Sending scheduler request: To fetch work. Requesting 227032 seconds of work, reporting 0 completed tasks
mer. 26 nov. 2008 01:53:40 CET|Milkyway@home|Scheduler request succeeded: got 0 new tasks
mer. 26 nov. 2008 01:53:40 CET|Milkyway@home|Message from server: No work sent

while there is work waiting on the servers (the "Message from server: No work sent" is usually an indication that there is no WU matching the application version).

I guess I'm having something wrong in my app_info.xml file, but I can't find out hat exactly. Here it is:
<app_info>
  <app>
    <name>milkyway</name>
  </app>
<file_info>
  <name>milkyway_0.3_i686-pc-linux-gnu</name>
  <executable/>
</file_info>
<app_version>
  <app_name>milkyway</app_name>
  <version_num>0.3</version_num>
  <file_ref>
    <file_name>milkyway_0.3_i686-pc-linux-gnu</file_name>
    <main_program/>
  </file_ref>
</app_version>
<app_info>

I tried to change name & app_name for "astronomy" and version_num for 2, 3, 0.2, 0.3 & 1.22, but all to no avail...

What am I doing wrong, please ?
ID: 6743 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile banditwolf
Avatar

Send message
Joined: 12 Nov 07
Posts: 2425
Credit: 524,164
RAC: 0
Message 6744 - Posted: 26 Nov 2008, 1:17:17 UTC - in response to Message 6743.  

Could just be that there isn't much work at the moment.
Doesn't expecting the unexpected make the unexpected the expected?
If it makes sense, DON'T do it.
ID: 6744 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Thierry Godefroy

Send message
Joined: 29 Jul 08
Posts: 9
Credit: 2,200,784
RAC: 0
Message 6746 - Posted: 26 Nov 2008, 1:25:44 UTC - in response to Message 6744.  

Could just be that there isn't much work at the moment.


There are over 1200 WUs available for download following the server status page, and they won't download either when I set back Milkyway to the old app (same "No work sent" report indicating there are WUs but that they don't match the app version). I'd rather say that there are no old app WUs left and that I didn't get the version and/or app name right in my app_info.xml file...
ID: 6746 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
John Clark

Send message
Joined: 4 Oct 08
Posts: 1734
Credit: 64,228,409
RAC: 0
Message 6749 - Posted: 26 Nov 2008, 1:47:52 UTC

This, I think, is common to many

I am using thw Milksop client so I would expect conventional WUs to D/L. But none come, and others running the test clients seem to be in the samee boat.

As Bandit says, no work?
ID: 6749 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Travis
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist

Send message
Joined: 30 Aug 07
Posts: 2046
Credit: 26,480
RAC: 0
Message 6751 - Posted: 26 Nov 2008, 2:22:56 UTC - in response to Message 6749.  

This, I think, is common to many

I am using thw Milksop client so I would expect conventional WUs to D/L. But none come, and others running the test clients seem to be in the samee boat.

As Bandit says, no work?


There's work available, and the assimilator is generating work... something is slowing up the process and we're not quite sure what yet.
ID: 6751 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
jedirock
Avatar

Send message
Joined: 8 Nov 08
Posts: 178
Credit: 6,140,854
RAC: 0
Message 6849 - Posted: 27 Nov 2008, 3:18:39 UTC - in response to Message 6743.  

However, it looks like I can't get BOINC to download WUs for the new app, it reports:
mer. 26 nov. 2008 01:53:35 CET|Milkyway@home|Sending scheduler request: To fetch work. Requesting 227032 seconds of work, reporting 0 completed tasks
mer. 26 nov. 2008 01:53:40 CET|Milkyway@home|Scheduler request succeeded: got 0 new tasks
mer. 26 nov. 2008 01:53:40 CET|Milkyway@home|Message from server: No work sent

while there is work waiting on the servers (the "Message from server: No work sent" is usually an indication that there is no WU matching the application version).

Getting the same thing here now, 1500WUs reported ready to sent at the time of this writing. Just trying to use the new 0.4 app, although I was getting the same error with my 0.3 app.
ID: 6849 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Travis
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist

Send message
Joined: 30 Aug 07
Posts: 2046
Credit: 26,480
RAC: 0
Message 6850 - Posted: 27 Nov 2008, 3:24:47 UTC - in response to Message 6849.  

However, it looks like I can't get BOINC to download WUs for the new app, it reports:
mer. 26 nov. 2008 01:53:35 CET|Milkyway@home|Sending scheduler request: To fetch work. Requesting 227032 seconds of work, reporting 0 completed tasks
mer. 26 nov. 2008 01:53:40 CET|Milkyway@home|Scheduler request succeeded: got 0 new tasks
mer. 26 nov. 2008 01:53:40 CET|Milkyway@home|Message from server: No work sent

while there is work waiting on the servers (the "Message from server: No work sent" is usually an indication that there is no WU matching the application version).

Getting the same thing here now, 1500WUs reported ready to sent at the time of this writing. Just trying to use the new 0.4 app, although I was getting the same error with my 0.3 app.


I've been trying to debug the assimilator, it's been crashing from some random seg fault that i haven't been able to track down. thats why the work has been scarce for the new app.
ID: 6850 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Travis
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist

Send message
Joined: 30 Aug 07
Posts: 2046
Credit: 26,480
RAC: 0
Message 6851 - Posted: 27 Nov 2008, 3:24:47 UTC - in response to Message 6849.  

However, it looks like I can't get BOINC to download WUs for the new app, it reports:
mer. 26 nov. 2008 01:53:35 CET|Milkyway@home|Sending scheduler request: To fetch work. Requesting 227032 seconds of work, reporting 0 completed tasks
mer. 26 nov. 2008 01:53:40 CET|Milkyway@home|Scheduler request succeeded: got 0 new tasks
mer. 26 nov. 2008 01:53:40 CET|Milkyway@home|Message from server: No work sent

while there is work waiting on the servers (the "Message from server: No work sent" is usually an indication that there is no WU matching the application version).

Getting the same thing here now, 1500WUs reported ready to sent at the time of this writing. Just trying to use the new 0.4 app, although I was getting the same error with my 0.3 app.


I've been trying to debug the assimilator, it's been crashing from some random seg fault that i haven't been able to track down. thats why the work has been scarce for the new app.
ID: 6851 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
1 · 2 · 3 · Next

Message boards : Application Code Discussion : milkyway code releases

©2024 Astroinformatics Group