Welcome to the V3.0 OpenVMS Alpha October 2000 port of the SETI@home software! For more information about the SETI@home program, please see their WWW page at: http://setiathome.ssl.berkeley.edu/ Bug Fixes Between V2.4 and V3.0 -------------------------------- The SETI@home project has made many bug fixes and changes in the base code since V2.4. The biggest change is an improved processing algorithm which allows new kinds of analyses to be done in the same time. Even with the new analyses, this new algorithm appears to improve the per-workunit performance running under the OpenVMS Operating System using EV6 and EV67 processors. The following bug fixes and changes are specific to the OpenVMS SETI@home client: * The OpenVMS SETI@home client V3.0 no longer creates a new version of USER_INFO.SAH for each new work unit. If you care for some reason, the first time you run it, it will also purge all of your old USER_INFO.SAH files. * The OpenVMS SETI@home client V3.0 no longer raises its priority to 3 when it runs in a process or batch queue with a lower priority. The new behavior is that it will always change priority to what is specified by the -nice parameter (assuming sufficient privilege). If -nice is not specified it will lower its priority to 3 if necessary, but it will not raise its prority. * The OpenVMS SETI@home client V3.0 is a bit more clever about changing its process name when you use the -procname switch. It now modifies the '@' to different characters until it finds a unique process name. This means that the name should reflect the job's done percentage more accurately. * The SETI.COM command file has a new feature. If you specify "TIME" in parameter 5 (along with LOG if you wish, separated by a comma), the procedure will create and update a file named TIMELOG.TXT. Each time a work unit finishes, the time, cputime, and done percent will be recorded in TIMELOG.TXT. If the done percent is not close to 100, this data point is probably bogus. It means the SETI client exited without error, but also with finishing a work unit. Bug Fixes Between V1.3a and V2.4 -------------------------------- The SETI@home project has made many bug fixes and changes since V1.3. The following should be of particular interest to those running the OpenVMS Seti client: o The client no longer gives an HPARITH divide-by-zero trap on certain work units. o The V2.4 client is able to work through a Microsoft firewall. o The V2.4 client does not print out as much information while it is running as the V1.3 client did. You can re-enable the information with the -verbose switch. o The V2.4 client does not save state as often. This should reduce the I/O overhead, particularly on fast processors. The following changes were VMS-specific: o The OpenVMS V2.4 client no longer produces multiple versions of files in the working directory. o The OpenVMS V2.4 client now comes as a single image which runs on all OpenVMS versions later than V6.2. o The SETI.COM parameter have changed to be more consistent in the different modes, and to allow more options. Installation and Startup: ------------------------- The OpenVMS SETI@home client should run on any version of OpenVMS Alpha after V6.2, (for example V6.2, V7.1, V7.1-1H1, V7.1-1H2, V7.1-2, V7.2, V7.2-1, V7.2-1H1), although we have not tested it on every version. We will NOT be providing a port for OpenVMS VAX. While the port is not difficult, we found that it ran too slow to be practical. It seems likely that this performance problem is because of the need to use IEEE floating point (which is not available in hardware on the VAX), although we did not spend much time analyzing it. To install SETI@home on an OpenVMS Alpha system, follow the directions below. We encourage you to run as many simultaneous jobs as you have CPUs. To run multiple copies, however, you must create multiple directories. First, unpack and restore the saveset. $ ! Device is your choice. $ create/dir device:[seti] ! Only if it does not exist already $ RUN setiathome_3_0_alpha_openvms_bck.exe $ BACKUP setiathome_3_0_alpha_openvms.bck/save device:[seti] $ SET DEFAULT device:[seti] The following paragraphs are meant as general guidelines only. The SETI.COM procedure is a template and should be modified to fit your system's needs. To set up the subdirectories and to fetch the first work unit, you can run SETI interactively. If you are going to run multiple copies, you must perform the following command once for each copy you wish to run. $ @DEVICE:[SETI]SETI.COM INTERACTIVE CPU1 Once it is running, you can Control-Y. The state information is saved and the next time you start it in that directory, it will restart where you left off. For the second CPU's work unit, use cpu2, and so on. NOTE: When you run SETI interactively, your process priority will be lowered (but not raised) to 3. Alternatively, you can run the program interactively only once and copy the USER_INFO.SAH file that the program created into your other directories. We recommend that you set up a batch queue, with a lower than normal priority, just for the SETI jobs. The job limit should again be the number of CPUs that you have (or that you want to dedicate to SETI) $ INIT/QUE/START/BASE_PRIORITY=3/JOB_LIMIT=3 SETI$BATCH/BATCH You can now submit the jobs to the queue. You can use the included command procedure as a template. Be sure to edit it if you are using other devices, directories or queue names. $ @SYS$SYSDEVICE:[SETI]SETI.COM SUBMIT 3 Release Notes: -------------- - Corporate firewalls may prevent the system from properly connecting with the SETIATHOME system. You may specify a proxy on the command line by using the '-proxy' statement. Please check with your local network administrator for the correct proxy to use. If you use SETI.COM, you must specify your proxy as parameter 3. For example: $ @SETI SUBMIT 1 "-proxy bear.company.com:8080" - The Seti client no longer prints out lots of status information by default (you can enable this info with -verbose). However, SETI.COM still submits the job with /NOLOG by default. You can specify "LOG" as parameter 5 to SETI.COM to turn on logging if you need to debug a problem. Note that there is a switch, -procname, to make the Seti client change its process name to reflect current progress. Specify this, and any other option in parameter 3 of SETI.COM. For example: $ @SETI SUBMIT 1 "-proxy bear.company.com:8080 -procname" - The "-email" option does not work correctly. - The OpenVMS Seti client has been compiled to optimize the code sequence for clients running on "EV6" (21264) and later versions of the Alpha CPU chip. We did this using the switches /OPTIMIZE=TUNE=EV6/ARCHITECTURE=GENERIC. Notice that with these switches, the compiler generates some hardware instructions (such as square root) which only exist on newer processors. However, it will generate code with tests such that these instructions are only executed on processors that support them. This means that the Seti client will still run acceptably on older Alpha chips. - The Graphic version of the Seti client is not available for OpenVMS yet. We will try to get to it as spare time permits. (I know...I said that before!) - Version 3.0 of the Seti client consists of a single executable image which is built to run on OpenVMS V6.2 and above. It is nonetheless built with the latest compilers and run time libraries, so there is little penalty on the latest OpenVMS versions. Problem Reporting ----------------- The most common problem is network access and proxies; you need to resolve that within your own company. Remember also that occasionally the Seti server at Berkeley is down or swamped. Remember that sometimes the client waits for many minutes before retrying after a failure. Almost invariably, network or server problems are the cause when you see the client staying in the LEF or HIB state. Allow the client to keep trying to return results of fetch a new work unit for a day or two and check the setiathome website before becoming concerned. This port of the SETIATHOME software has been verified against the baseline results. Very few changes were made during the port, so we suggest that all problems be reported directly to SETIATHOME. However, if you have problems that are clearly OpenVMS-specific, or suggestions, we would be happy to hear from you and we will try to help as our work schedules permit. Thanks to Compaq Computer Corporation for the use of their engineering resources during this port. Jeff Friedrichs ( jeff.friedrichs@compaq.com ) Burns Fisher ( burns.fisher@compaq.com ) 12-Oct-2000 ===============================================================