Hi,

We just released v1.20 which should fix an important bug which prevented proper validation, in particular on AMD 4xxx GPUs. Let's try and collect your feedback to this specific app version (and this one only) in this thread.

FYI, please use the latest Catalyst driver (>=12.1) and BOINC client (>=7.0.12).

Thanks,
Oliver

Hi, I have some tasks that all gave errors.

http://albert.phys.uwm.edu/results.php?hostid=1353&offset=0&show_names=0&state=5&appid=

I run on an ATI 4850 with 512MB RAM and Catalyst 11.12
BM 7.0.12

---

I run on an ATI 4850 with 512MB RAM and Catalyst 11.12

As discussed earlier, we (still) need ~490 MB free GPU memory. This doesn't seem to be the case here. We'd be able to reduce that by 120 MB but that would mean to exclude all OpenCL 1.0 GPUs, including your 4xxx series card.

FYI, we think about doing this anyway since the OpenCL 1.0 devices don't offer a significant performance gain over the CPU version so it seems worthwhile dropping their support in favor of more relaxed memory requirements.


Cheers,
Oliver

I run on an ATI 4850 with 512MB RAM and Catalyst 11.12

As discussed earlier, we (still) need ~490 MB free GPU memory. This does't seem to be the case here. We'd be able to reduce that by 120 MB but that would mean to exclude all OpenCL 1.0 GPUs, including your 4xxx series card.

FYI, we think about doing this anyway since the OpenCL 1.0 devices don't offer a significant performance gain over the CPU version so it seems worthwhile dropping their support in favor of more relaxed memory requirements.


Cheers,
Oliver

Well, if the mem need was 450Mb you would have a few more 4xxx cards....
I think that my card has about 480Mb free so if the demand is 470MB one could run the app on 48xx cards.

---

As I said, we can reduce it but only by dropping OpenCL 1.0 (i.e. AMD 4xxx) support.

Oliver

As I said, we can reduce it but only by dropping OpenCL 1.0 (i.e. AMD 4xxx) support.

Oliver

Ok

---

Have tried several Catalyst driver/AMD-APP-SDK/Boinc combinations.
OS = Ubuntu/oneiric, 2 x AMD 5770 (Juniper) Cards

ATI GPU 0: ATI Radeon HD 5700 series (Juniper) (CAL version 1.4.1664, 1024MB, 817MB available, 2720 GFLOPS peak)
ATI GPU 1: ATI Radeon HD 5700 series (Juniper) (CAL version 1.4.1664, 1024MB, 983MB available, 2720 GFLOPS peak)
02-Feb-2012 16:45:01 [---] OpenCL: ATI GPU 0: Juniper (driver version CAL 1.4.1664, device version OpenCL 1.1 AMD-APP (851.4), 512MB, 817MB available)
02-Feb-2012 16:45:01 [---] OpenCL: ATI GPU 1: Juniper (driver version CAL 1.4.1664, device version OpenCL 1.1 AMD-APP (851.4), 512MB, 983MB available)

Last combination is Catalyst 12.1, AMD-APP-SDK-v2.6, (without OpenCL v1.2 support) boinc 7.0.12. Results are like=>

http://albert.phys.uwm.edu/result.php?resultid=114016

clinfo reports:clinfo

Number of platforms:                             1
  Platform Profile:                              FULL_PROFILE
  Platform Version:                              OpenCL 1.1 AMD-APP (851.4)
  Platform Name:                                 AMD Accelerated Parallel Processing
  Platform Vendor:                               Advanced Micro Devices, Inc.
  Platform Extensions:                           cl_khr_icd cl_amd_event_callback cl_amd_offline_devices


  Platform Name:                                 AMD Accelerated Parallel Processing
Number of devices:                               3
  Device Type:                                   CL_DEVICE_TYPE_GPU
  Device ID:                                     4098
  Board name:                                    ATI Radeon HD 5700 Series
  Device Topology:                               PCI[ B#2, D#0, F#0 ]
  Max compute units:                             10
  Max work items dimensions:                     3
    Max work items[0]:                           256
    Max work items[1]:                           256
    Max work items[2]:                           256
  Max work group size:                           256
  Preferred vector width char:                   16
  Preferred vector width short:                  8
  Preferred vector width int:                    4
  Preferred vector width long:                   2
  Preferred vector width float:                  4
  Preferred vector width double:                 0
  Native vector width char:                      16
  Native vector width short:                     8
  Native vector width int:                       4
  Native vector width long:                      2
  Native vector width float:                     4
  Native vector width double:                    0
  Max clock frequency:                           0Mhz
  Address bits:                                  32
  Max memory allocation:                         134217728
  Image support:                                 Yes
  Max number of images read arguments:           128
  Max number of images write arguments:          8
  Max image 2D width:                            8192
  Max image 2D height:                           8192
  Max image 3D width:                            2048
  Max image 3D height:                           2048
  Max image 3D depth:                            2048
  Max samplers within kernel:                    16
  Max size of kernel argument:                   1024
  Alignment (bits) of base address:              2048
  Minimum alignment (bytes) for any datatype:    128
  Single precision floating point capability
    Denorms:                                     No
    Quiet NaNs:                                  Yes
    Round to nearest even:                       Yes
    Round to zero:                               Yes
    Round to +ve and infinity:                   Yes
    IEEE754-2008 fused multiply-add:             Yes
  Cache type:                                    None
  Cache line size:                               0
  Cache size:                                    0
  Global memory size:                            536870912
  Constant buffer size:                          65536
  Max number of constant args:                   8
  Local memory type:                             Scratchpad
  Local memory size:                             32768
  Kernel Preferred work group size multiple:     64
  Error correction support:                      0
  Unified memory for Host and Device:            0
  Profiling timer resolution:                    1
  Device endianess:                              Little
  Available:                                     Yes
  Compiler available:                            Yes
  Execution capabilities:                                
    Execute OpenCL kernels:                      Yes
    Execute native function:                     No
  Queue properties:                              
    Out-of-Order:                                No
    Profiling :                                  Yes
  Platform ID:                                   0x7fedcd03c100
  Name:                                          Juniper
  Vendor:                                        Advanced Micro Devices, Inc.
  Device OpenCL C version:                       OpenCL C 1.1 
  Driver version:                                CAL 1.4.1664
  Profile:                                       FULL_PROFILE
  Version:                                       OpenCL 1.1 AMD-APP (851.4)
  Extensions:                                    cl_khr_global_int32_base_atomics cl_khr_global_int32_extended_atomics cl_khr_local_int32_base_atomics cl_khr_local_int32_extended_atomics cl_khr_3d_image_writes cl_khr_byte_addressable_store cl_khr_gl_sharing cl_ext_atomic_counters_32 cl_amd_device_attribute_query cl_amd_vec3 cl_amd_printf cl_amd_media_ops cl_amd_popcnt 

Great, initial estimated time to completion, 286 hours. No wonder it went into HP. But that DCF of 11 is way out of whack.
i3-530, Win 7 - 64bit, 8GB RAM, ATI HD6850 2GB, Catalysts 11.12, BOINC 7.0.12

---

Jord.

BOINC FAQ Service

They say most of your brain shuts down in cryo-sleep. All but the primitive side, the animal side. No wonder I'm still awake.

Last combination is Catalyst 12.1, AMD-APP-SDK-v2.6, (without OpenCL v1.2 support) boinc 7.0.12. Results are like=>

http://albert.phys.uwm.edu/result.php?resultid=114016

clinfo reports:clinfo

  Global memory size:                            536870912

Your card has 512 MB video memory. That's right on the edge of what our app requires (see above). In your case it's not sufficient - error 5 means "out of resources" (memory).

Sorry. You would be one who would benefit if we drop OpenCL 1.0 support in favor of lower memory requirements...

Cheers,
Oliver

PS: BOINC's memory information still seems to be off. I'll forward this...

Great, initial estimated time to completion, 286 hours. No wonder it went into HP. But that DCF of 11 is way out of whack.
i3-530, Win 7 - 64bit, 8GB RAM, ATI HD6850 2GB, Catalysts 11.12, BOINC 7.0.12

I'll forward this to Bernd.


Thanks,
Oliver

Oops you're right, video memory is reported as 1024MB for both cards. With 817MB, 983MB available, but global memory for OpenCL is reported from AMD-APP-SDK as 512MB , but (strange) boinc says 817MB, 983MB available). Is there a tweak in OpenCL configuration? amdccle (Catalyst Control Center) says 1024MB Video Memory for both cards. Hmm...?

Oops you're right, video memory is reported as 1024MB for both cards. With 817MB, 983MB available, but global memory for OpenCL is reported from AMD-APP-SDK as 512MB , but (strange) boinc says 817MB, 983MB available). Is there a tweak in OpenCL configuration? amdccle (Catalyst Control Center) says 1024MB Video Memory for both cards. Hmm...?

This is a (sort of) known issue with AMD GPUs/drivers. Do you happen to know the actual/correct value of your card? Look at the retail box if it helps :-)


Oliver

According to apple specs it's a MacPro with the standard graphics card ATI Radeon 5770 1GB video mem (one auxilary power cable). Maybe AMD-APP-SDK limits to 50%?

According to apple specs it's a MacPro with the standard graphics card ATI Radeon 5770 1GB video mem (one auxilary power cable). Maybe AMD-APP-SDK limits to 50%?

Hang on, do you run Linux on your MacPro?

Anyway, you're witnessing the mess we've to deal with here, AMD's driver and CAL/OpenCL runtimes don't reliably report RAM sizes on some GPUs. They don't even agree with each other. I wouldn't wonder if the driver/runtime tells our app that there's no more memory available (hence the error you see) even when there is... This is valuable hint indeed!

Oliver

Yes, i' am running Linux on my MacPro. (3.1)
So here we are:
BOINC 1GB
FGLRX 1GB
AMD-SDK 512MB
Card label, Apple 1GB

I installed the driver AFTER the AMD-SDK., so the openCL runtime libs maybe replaced by the driver installation.

And now linux kernel:

id:
display
description: VGA compatible controller
product: Juniper [Radeon HD 5700 Series]
vendor: ATI Technologies Inc
physical id:
0
bus info:
pci@0000:01:00.0
version: 00
width: 64 bits
clock: 33MHz
capabilities: pm pciexpress msi vga_controller bus_master cap_list rom
configuration:
driver = fglrx_pci
latency = 0
resources:
irq : 60
memory : 80000000-9fffffff
memory : c0b00000-c0b1ffff
ioport : 3000(size=256)
memory : c0b20000-c0b3ffff

Someone with power of 2 capability may calculate ram from here

I dont know what is problem in linux clean install ubuntu 11.10 x64 latest 12.1 catalyst. all in clean install. Still <core_client_version>7.0.12</core_client_version>
<![CDATA[
<message>
process exited with code 255 (0xff, -1)
</message>
<stderr_txt>
[12:12:23][2765][INFO ] Application startup - thank you for supporting Einstein@Home!
[12:12:23][2765][INFO ] Starting data processing...
[12:12:23][2765][ERROR] Failed to get OpenCL platform/device info from BOINC (error: -1)!

http://albert.phys.uwm.edu/result.php?resultid=109647

http://albert.phys.uwm.edu/result.php?resultid=109647
[12:12:23][2765][ERROR] Demodulation failed (error: -1)!
12:12:23 (2765): called boinc_finish

</stderr_txt>
]]>

3 gentoo 64bit boxes all with the same problem and they are crunching openCL wus for poem and primegrid - exit with code 255 - unable to get openCL info - error -1

---

Idont know but other openCL pojects run fine but not this. Poem@home fine, moowrapper CAL fine. I not tested MW.

Looks like it's more on linux pc, my win7 runs fine
http://albert.phys.uwm.edu/show_host_detail.php?hostid=1894
the last known failing wu is from 28.Jan.

---

I think its restricted to linux 64 bit machines

---