It's known that ATi MB r177 can't work on ATi GPUs with reduced to 128 threads per block number (usual value is 256).
I prepared new version that should solve this issue but for now have no hardware (no such low-end GPU) to test if bugfix was successful or not.
If someone willing to test it, let me know here or (preferably) on Lunatics.

I have 3 HD4650. One old 512MB full size version. One new low profile with 512MB and an other one with 1GB memory low profile also.
Since i always wanted the MB app to run on them, i would like to help getting one done but i havent done beta testing before. If its not to complicated id like to try.

---

Suggestion:
supply the app_info.xml part in the package (.rar, .zip, .7z) (in plain .txt file)
and short instructions (ReadMe_OpenCL.txt) where to put it (e.g. before the closing </app_info> tag in their existing app_info.xml)
+ some links (URL) you find useful for the user to read
so people will not need to search the app_info.xml part on the forums.

---

 


- ALF - "Find out what you don't do well ..... then don't do it!" :)
 

Suggestion:
supply the app_info.xml part in the package (.rar, .zip, .7z) (in plain .txt file)
and short instructions (ReadMe_OpenCL.txt) where to put it (e.g. before the closing </app_info> tag in their existing app_info.xml)
+ some links (URL) you find useful for the user to read
so people will not need to search the app_info.xml part on the forums.

Sounds like a good idea, since 1 typo in app_info.xml file can do a lot of damage.
I now have a i7-2600 +2x EAH5870, WIN 7 64BIT, 8GiG DDR3 RAM.
(And getting really bored running Collatz C. and MW doesn't have much work, a.t.m.)

I did run Astropulse rev.456 on a 5870, which worked OK.
(Quite a difference, between XP and WIN 7 and trying to get used to it)

---

First try completely failed, installed SPP (SDK). Out of SETI work, it's gone....
(I think, BOINC is requesting an executable?)
First I'm going to wait till surgery is advanced enough to replace an eye?!
Just kiddin.............
Is installing the AMD SPP (SDK) necessary? In stalled the 64BIT 2.4 SPP.

---

Ok, give me the dl link. I can test on HD 5450 (1024 mb) and an old 2600 (512 mb) mobility if needed.

Christoph

---

Christoph

You should contact Raistmer, coder-master @LUNATICs website.
And installing the 64BIT version of AMD's APP 2.4, didn't do
much good, all SETI MB's were gone and all MW WU's and GPUz. simply states
Direct Compute 5, works, no OpenCL ?

You should be able to log -in, if you're a member, since their 'policy accepting everybody', has changed.

You could P.M. him, Raistmer.
And try the latest card, 2600 is ba bit light, INHO.
And you can use a HD5 version, better for 5000 series, not 2600.

Original thread article referer.

---

SUMMARY
I have tried and failed to run the r220 application on my 4550. It appears that the GPU never devotes any appreciable resource to the task, and although progress is reported, the job starts over from the beginning after any interruption.

SYSTEM DETAILS
my test host is a Windows 7 Pro system with an E5620 (Westmere) qual CPU with HT enabled. The primary system load is a full set of Einstein tasks, which together consume a little over 2Gbyte of working set, but the system has 12 Gbyte installed.

The GPU card is a "GIGABYTE GV-R455D3-512I Radeon HD 4550 512MB 64-bit DDR3 PCI Express 2.0 x16 HDCP Ready Video Card". GPU-Z reports that I have "atiumdag 8.821.0.0 (Catalyst 11.2)/Win7 64" driver installed.

The app_info segment I used was this:

<app>
		<name>setiathome_enhanced</name>
	</app>
	<file_info>
        	<name>MB_6.10_win_x86_SSE3_OpenCL_ATi_r220.exe</name>
        	<executable/>
    	</file_info>
	<file_info>
	    <name>MultiBeam_Kernels_r220.cl</name>
	    <executable/>
	</file_info>
    	<app_version>
        	<app_name>setiathome_enhanced</app_name>
        	<version_num>610</version_num>
		<avg_ncpus>0.01</avg_ncpus>
		<max_ncpus>0.01</max_ncpus>
                <plan_class>ati13ati</plan_class>
                <cmdline>-period_iterations_num 2 -instances_per_device 1</cmdline>
         	<coproc>
   	         	<type>ATI</type>
		        <count>1</count>
         	</coproc>
	        <flops>15987654321</flops>
	       	<file_ref>
            		<file_name>MB_6.10_win_x86_SSE3_OpenCL_ATi_r220.exe</file_name>
            		<main_program/>                           
        	</file_ref>
		<file_ref>
		    <file_name>MultiBeam_Kernels_r220.cl</file_name>
		    <copy_file/>
		</file_ref>
    	</app_version>

The application does not report error to the messages tab of boincmgr, and the logging of restarts looks like this:

    task 24no10ae.27626.10878.12.10.180_0 suspended by user

This host has been successfully running Raistmer ATI code for Astropulse for many weeks. It has not been running any other SETI work. I copied the non HD5 variant of the three r220 files, and boincmgr downloaded three multibeam results. The one I have tried to run was from this Work Unit. The behavior was that no GPU activity was observed as reported by GPU-Z (load at 0%, temperature, clock and voltage at non-active baseline), while the application reported a substantial fraction of a core of CPU usage. Progress was reported at a rate indicating completion in about 2.5 hours of wall clock time. But on any interruption (I have the computing preference set to suspend GPU work when computer is in use) the progress percentage starts over on task resumption.

I've aborted and reported that first task, so when the SETI site catches up the stderr should be visible, but it is not as yet. I think the relevant stderr was in slot 9 on my host, and after I aborted the task it appears to contain 5 repetitions of this text

Number of period iterations for PulseFind setted to:2
Number of app instances per device setted to:1
Running on device number: 0
Priority of worker thread raised successfully
Priority of process adjusted successfully, below normal priority class used
OpenCL platform detected: Advanced Micro Devices, Inc.
BOINC assigns 0 device, slots 0 to 0 (including) will be checked
Used slot is 0; Info : Building Program (clBuildProgram):main kernels: OK code 0

I don't have much idea whether there is a basic ap problem, or I've made a mistake in attempting to configure for running, or my system has an incompatible configuration. I'm very eager to read any reports of success, especially on an HD4550 GPU.

---

Thanks for full report.
Unfortunately, listed WU has this stderr for your result:
<core_client_version>6.10.58</core_client_version>
<![CDATA[
<message>
aborted by user
</message>
]]>
not much info.

I'll prepare testcase for standalone run and will contact you (via mail probably).

App installed. Currently no AP tasks, Collatz is nwo NNT. Willücome back hen I got an AP task. This host: http://setiathome.berkeley.edu/show_host_detail.php?hostid=5439210

---

Christoph

You should contact Raistmer, coder-master @LUNATICs website.

Current Lunatics member-testers can PM Raistmer @ Lunatics.
Everyone else, should PM him HERE since while non-members can browse the Lunatics board, they can't post or PM on it.

After completing this test cycle for Raistmer, if anyone wishes to regularly participate in additional similar testing, please let Raistmer, any other Lunatics Dev. , or myself know.

I'll prepare testcase for standalone run and will contact you (via mail probably).

Good, I'll be happy to run it.

I hope I'm not just wasting your time with a stupid configuration error on my part--but even so, if the cause can be found, it might help others to succeed.

---

I'll prepare testcase for standalone run and will contact you (via mail probably).

Good, I'll be happy to run it.

I hope I'm not just wasting your time with a stupid configuration error on my part--but even so, if the cause can be found, it might help others to succeed.

Well that can always happen(Murphy's Law :)), understand more testers, knowing what their doing, are needed, so participate..............

Hi, Raistmer, started testing again and am using your Rev516 AP ATI, changed the
ffa_block and ffa_block_fetch 8192 and 4096, unroll 10, count 2, 2 HD5870's (i7-2600). (4x AP WU)
This rig
. Memory, DDR3, 1333MHz is heavily used, so I used SSSE3, not SSE4.1.

A little lag, but also running 8x (SSSE3x) opt. MB, wait and see how this works out,
then also can test the ATI MB app. r.177.
(Set 3TFLOPS, to get some AP and MB work, well I got it, most are already running
High Priority ;) )
Unfortunatly CPU use did go up, also by running 2 on each card,(unroll 12 or above
makes sreen lag unacceptable. GPU use is better too, with this setting.
(Although 1 of the 8 CPU-cores is almost used to load/unload the GPU's, maybe better
to use 1 for GPU loading)
Hope they'll validate.

---

Hi, Raistmer, started testing again and am using your Rev516 AP ATI, changed the
ffa_block and ffa_block_fetch 8192 and 4096, unroll 10, count 2, 2 HD5870's (i7-2600). (4x AP WU)
This rig
.

Hello
Better to report in corresponding release thread so others can use your results.

Hello Raistmer, OK I'll use that particular Forum Thread.

---

PMs sent, please, post stderr.txt content as feedback.

---

SETI apps news
We're not gonna fight them. We're gonna transcend them.

PMs sent, please, post stderr.txt content as feedback.

In my case (system details given in previous post) the observed behavior is this:
1. On launching the supplied test exe, my cursor went to a fresh line immediately. I initially supposed that the Windows application had done a very fast termination, but examination of running processes showed that your test Windows .exe continued to run--showing about 12.3% CPU use--in other words one thread executing nearly all the time.
2. GPU-Z suggests that my ATI 4550 GPU is doing nothing, or very nearly so. GPU load shows as 0%, core clock as 110 MHz, VDDC at 0.9V, and temperature of 60C, all of which are the standard idle behavior, and none of which are like the results when running your Astropulse app.
3. Looking for files in the application directory, I see boinc_lockfile with size 0, stderr.txt with size zero, and result.sah with size 7239 bytes as the new things, and of course the four file unpacked are still there.

As stderr.txt is size zero, there is not yet something to post from there. As of this moment, the application has been running for half an hour, with no change in CPU usage, GPU usage, or files in the directory from which I am running it.

---

One additional observation set: while the CPU consumption, GPU consumption, and result files do not change, and one might imagine nothing going on, Process Explorer shows considerable I/O and context switches by the test Windows application:

at the 53 minutes elapsed point:

1. 8,086,142,000 I/O Read bytes
2. 5,120,204,520 I/O Write bytes
3. 4,694,559 context switches

The rate of I/O read, I/O write, and context switching seems pretty stable.

---

Kill the process and send me stderr.txt file as is, please.

---

SETI apps news
We're not gonna fight them. We're gonna transcend them.

Kill the process and send me stderr.txt file as is, please.

Done.

---