Issues with MCX and 'pattern' srctype

Nicholas Dana

unread,

Apr 18, 2016, 2:58:53 PM4/18/16

to mcx-users

I've been using MCX to estimate fluences for some photoacoustic and diffuse reflectance applications, but I'm getting some consistent crashes when using the 'pattern' srctype.

My current implementation uses a previous flux output (at a shorter wavelength) to generate a slice (i.e. pattern image) that corresponds to what fluorescent sources would be in the tissue at that slice. I then scale this image to [0-1] and define cfg.srcpos, cfg.srcparam1 and cfg.srcparam2 based on where I generated the slice from. I also modify the cfg.prop to reflect the optical properties at a different wavelength and keep the cfg.nphoton low for testing purposes. All other cfg parameters stay the same. When I attempt to run it thus, I invariably get a MATLAB crash and a lengthy error detail.

When I modify the script slightly to use a 'gaussian' source type to mimic the image generated from above and define cfg.srcpos & cfg.srcparam1 appropriately, it executes fine and I get a flux output in line with expectations.

I've attached my code with the 'pattern' srctype section commented out, and the 'gaussian' section filled in. I've also posted the error I received when I use the 'pattern' srctype below, . I've checked and rechecked the cfg.srcpattern parameter to make sure it is sized and scaled properly. Any experience a similar problem and solve it? It also seems to crash at the same time after calling mcxlab, which is right before it displays 'mcx.srcparam2'. Any help is greatly appreciated.

------------------------------------------------------------------------

Access violation detected at Mon Apr 18 13:56:00 2016

------------------------------------------------------------------------

Configuration:

Crash Decoding : Disabled

Default Encoding : windows-1252

Graphics card 1 : DemoForge ( 0x0 ) Mirage Driver Version 2.0.105.0

Graphics card 2 : NVIDIA ( 0x10de ) NVIDIA Quadro K600 Version 10.18.13.5390

MATLAB Architecture: win64

MATLAB Root : C:\Program Files\MATLAB\R2014b

MATLAB Version : 8.4.0.150421 (R2014b)

Operating System : Microsoft Windows 7 Enterprise

Processor ID : x86 Family 6 Model 60 Stepping 3, GenuineIntel

Software OpenGL : 0

Virtual Machine : Java 1.7.0_11-b21 with Oracle Corporation Java HotSpot(TM) 64-Bit Server VM mixed mode

Window System : Version 6.1 (Build 7601: Service Pack 1)

Fault Count: 1

Abnormal termination:

Access violation

Register State (from fault):

RAX = 000000000003d090 RBX = 0000000004026df0

RCX = 000000000007afb0 RDX = 000000000000000a

RSP = 0000000004026b70 RBP = 000000000e8b6000

RSI = 000000000001ebec RDI = 000000003a3d7360

R8 = 00000000100e0080 R9 = 00000000000002a0

R10 = 0000000000010000 R11 = 00000000000001d0

R12 = 000000003a3d7090 R13 = 0000000000000001

R14 = 0000000000000000 R15 = 0000000000000001

RIP = 000000000fc021b0 EFL = 00010287

CS = 0033 FS = 0053 GS = 002b

Stack Trace (from fault):

[ 0] 0x000000000fc021b0 C:\Users\nd5583\Box Sync\Data\mcx\mcxlab\mcxlab.mexw64+00008624

[ 1] 0x000000000fc02875 C:\Users\nd5583\Box Sync\Data\mcx\mcxlab\mcxlab.mexw64+00010357 mexFunction+00000357

[ 2] 0x00000000fc5f3701 C:\Program Files\MATLAB\R2014b\bin\win64\libmex.dll+00079617 mexRunMexFile+00000129

[ 3] 0x00000000fc5f2762 C:\Program Files\MATLAB\R2014b\bin\win64\libmex.dll+00075618 inSwapMexfileReader+00000690

[ 4] 0x00000000fc5f2248 C:\Program Files\MATLAB\R2014b\bin\win64\libmex.dll+00074312 mexUnlock+00006008

[ 5] 0x0000000004cf05e3 C:\Program Files\MATLAB\R2014b\bin\win64\m_dispatcher.dll+00067043 Mfh_file::dispatch_fh+00000659

[ 6] 0x0000000004cf0aee C:\Program Files\MATLAB\R2014b\bin\win64\m_dispatcher.dll+00068334 Mfunction_handle::dispatch+00000766

[ 7] 0x0000000004e02b2f C:\Program Files\MATLAB\R2014b\bin\win64\m_interpreter.dll+00273199 inGetNameInSymbolTable+00017071

[ 8] 0x0000000004eeb6f7 C:\Program Files\MATLAB\R2014b\bin\win64\m_interpreter.dll+01226487 inFullEvalFcn+00011159

[ 9] 0x0000000004e05592 C:\Program Files\MATLAB\R2014b\bin\win64\m_interpreter.dll+00284050 inGetNameInSymbolTable+00027922

[ 10] 0x0000000004e01b4f C:\Program Files\MATLAB\R2014b\bin\win64\m_interpreter.dll+00269135 inGetNameInSymbolTable+00013007

[ 11] 0x0000000004e01ab1 C:\Program Files\MATLAB\R2014b\bin\win64\m_interpreter.dll+00268977 inGetNameInSymbolTable+00012849

[ 12] 0x0000000004e2406f C:\Program Files\MATLAB\R2014b\bin\win64\m_interpreter.dll+00409711 inPathNotification::function_delete_notification+00098591

[ 13] 0x0000000004e227af C:\Program Files\MATLAB\R2014b\bin\win64\m_interpreter.dll+00403375 inPathNotification::function_delete_notification+00092255

[ 14] 0x0000000004e21359 C:\Program Files\MATLAB\R2014b\bin\win64\m_interpreter.dll+00398169 inPathNotification::function_delete_notification+00087049

[ 15] 0x0000000004cf07a4 C:\Program Files\MATLAB\R2014b\bin\win64\m_dispatcher.dll+00067492 Mfh_file::dispatch_fh+00001108

[ 16] 0x0000000004cf0aee C:\Program Files\MATLAB\R2014b\bin\win64\m_dispatcher.dll+00068334 Mfunction_handle::dispatch+00000766

[ 17] 0x0000000004e02b2f C:\Program Files\MATLAB\R2014b\bin\win64\m_interpreter.dll+00273199 inGetNameInSymbolTable+00017071

[ 18] 0x0000000004eeb6f7 C:\Program Files\MATLAB\R2014b\bin\win64\m_interpreter.dll+01226487 inFullEvalFcn+00011159

[ 19] 0x0000000004e05592 C:\Program Files\MATLAB\R2014b\bin\win64\m_interpreter.dll+00284050 inGetNameInSymbolTable+00027922

[ 20] 0x0000000004e01b4f C:\Program Files\MATLAB\R2014b\bin\win64\m_interpreter.dll+00269135 inGetNameInSymbolTable+00013007

[ 21] 0x0000000004e01ab1 C:\Program Files\MATLAB\R2014b\bin\win64\m_interpreter.dll+00268977 inGetNameInSymbolTable+00012849

[ 22] 0x0000000004e2406f C:\Program Files\MATLAB\R2014b\bin\win64\m_interpreter.dll+00409711 inPathNotification::function_delete_notification+00098591

[ 23] 0x0000000004e227af C:\Program Files\MATLAB\R2014b\bin\win64\m_interpreter.dll+00403375 inPathNotification::function_delete_notification+00092255

[ 24] 0x0000000004e21359 C:\Program Files\MATLAB\R2014b\bin\win64\m_interpreter.dll+00398169 inPathNotification::function_delete_notification+00087049

[ 25] 0x0000000004cf07a4 C:\Program Files\MATLAB\R2014b\bin\win64\m_dispatcher.dll+00067492 Mfh_file::dispatch_fh+00001108

[ 26] 0x0000000004cf0aee C:\Program Files\MATLAB\R2014b\bin\win64\m_dispatcher.dll+00068334 Mfunction_handle::dispatch+00000766

[ 27] 0x0000000004e54a88 C:\Program Files\MATLAB\R2014b\bin\win64\m_interpreter.dll+00608904 inPathNotification::function_delete_notification+00297784

[ 28] 0x0000000004e54a04 C:\Program Files\MATLAB\R2014b\bin\win64\m_interpreter.dll+00608772 inPathNotification::function_delete_notification+00297652

[ 29] 0x0000000004e549c5 C:\Program Files\MATLAB\R2014b\bin\win64\m_interpreter.dll+00608709 inPathNotification::function_delete_notification+00297589

[ 30] 0x0000000004e54992 C:\Program Files\MATLAB\R2014b\bin\win64\m_interpreter.dll+00608658 inPathNotification::function_delete_notification+00297538

[ 31] 0x0000000004e54947 C:\Program Files\MATLAB\R2014b\bin\win64\m_interpreter.dll+00608583 inPathNotification::function_delete_notification+00297463

[ 32] 0x0000000004dcd69d C:\Program Files\MATLAB\R2014b\bin\win64\m_interpreter.dll+00054941 inEvalExp+00001309

[ 33] 0x0000000004ef525d C:\Program Files\MATLAB\R2014b\bin\win64\m_interpreter.dll+01266269 inEvalCmdWithLocalReturn+00000285

[ 34] 0x0000000004ef5181 C:\Program Files\MATLAB\R2014b\bin\win64\m_interpreter.dll+01266049 inEvalCmdWithLocalReturn+00000065

[ 35] 0x000000000636e31d C:\Program Files\MATLAB\R2014b\bin\win64\libmwbridge.dll+00058141 mnGetPrompt+00001517

[ 36] 0x000000000636ed09 C:\Program Files\MATLAB\R2014b\bin\win64\libmwbridge.dll+00060681 mnParser+00000745

[ 37] 0x00000000fd0dd834 C:\Program Files\MATLAB\R2014b\bin\win64\mcr.dll+00383028 mcrInstance::mnParser_on_interpreter_thread+00000036

[ 38] 0x00000000fd0a68e7 C:\Program Files\MATLAB\R2014b\bin\win64\mcr.dll+00157927 mcr::runtime::setInterpreterThreadToCurrent+00019751

[ 39] 0x00000000fd0a6923 C:\Program Files\MATLAB\R2014b\bin\win64\mcr.dll+00157987 mcr::runtime::setInterpreterThreadToCurrent+00019811

[ 40] 0x00000000fd0a7121 C:\Program Files\MATLAB\R2014b\bin\win64\mcr.dll+00160033 mcr::runtime::setInterpreterThreadToCurrent+00021857

[ 41] 0x00000000fab5d3a6 C:\Program Files\MATLAB\R2014b\bin\win64\uiw.dll+00512934 UIW_AttachThreadInput+00001270

[ 42] 0x00000000fab5cc35 C:\Program Files\MATLAB\R2014b\bin\win64\uiw.dll+00511029 wsd_to_MSW+00004373

[ 43] 0x00000000fab5ccb9 C:\Program Files\MATLAB\R2014b\bin\win64\uiw.dll+00511161 wsd_to_MSW+00004505

[ 44] 0x00000000775d8802 C:\WINDOWS\system32\USER32.dll+00165890 GetMenuBarInfo+00000626

[ 45] 0x00000000775bf5e7 C:\WINDOWS\system32\USER32.dll+00062951 SystemParametersInfoW+00000247

[ 46] 0x00000000775c4869 C:\WINDOWS\system32\USER32.dll+00084073 IsProcessDPIAware+00000453

[ 47] 0x000000007781ba75 C:\WINDOWS\SYSTEM32\ntdll.dll+00309877 KiUserCallbackDispatcher+00000031

[ 48] 0x00000000775c90ca C:\WINDOWS\system32\USER32.dll+00102602 PeekMessageW+00000186

[ 49] 0x00000000775c50de C:\WINDOWS\system32\USER32.dll+00086238 GetKeyState+00000238

[ 50] 0x00000000775c3a47 C:\WINDOWS\system32\USER32.dll+00080455 PeekMessageA+00000087

[ 51] 0x00000000fab0a26b C:\Program Files\MATLAB\R2014b\bin\win64\uiw.dll+00172651 UIW_ShowMenuItem+00003995

[ 52] 0x00000000fab5dce2 C:\Program Files\MATLAB\R2014b\bin\win64\uiw.dll+00515298 UIW_SuspendAttachThreadInput+00000466

[ 53] 0x0000000100118223 C:\Program Files\MATLAB\R2014b\bin\win64\libmwservices.dll+01212963 services::system_events::PpeDispatchHook::dispatchOne+00032291

[ 54] 0x0000000100123c95 C:\Program Files\MATLAB\R2014b\bin\win64\libmwservices.dll+01260693 sysq::addProcessPendingEventsUnitTestHook+00005813

[ 55] 0x0000000100123d50 C:\Program Files\MATLAB\R2014b\bin\win64\libmwservices.dll+01260880 sysq::addProcessPendingEventsUnitTestHook+00006000

[ 56] 0x0000000100125b75 C:\Program Files\MATLAB\R2014b\bin\win64\libmwservices.dll+01268597 sysq::getCondition+00004629

[ 57] 0x00000001001274ba C:\Program Files\MATLAB\R2014b\bin\win64\libmwservices.dll+01275066 svWS_ProcessPendingEvents+00000138

[ 58] 0x00000000fd0a783d C:\Program Files\MATLAB\R2014b\bin\win64\mcr.dll+00161853 mcr::runtime::setInterpreterThreadToCurrent+00023677

[ 59] 0x00000000fd0a84aa C:\Program Files\MATLAB\R2014b\bin\win64\mcr.dll+00165034 mcr::runtime::setInterpreterThreadToCurrent+00026858

[ 60] 0x00000000fd0a0135 C:\Program Files\MATLAB\R2014b\bin\win64\mcr.dll+00131381 mcr_process_events+00007477

[ 61] 0x00000000fd0a1a12 C:\Program Files\MATLAB\R2014b\bin\win64\mcr.dll+00137746 mcr_process_events+00013842

[ 62] 0x0000000008128f2d C:\Program Files\MATLAB\R2014b\bin\win64\MVMLocal.dll+00167725 mvm_server::inproc::LocalFactory::terminate+00048285

[ 63] 0x00000000fa8ef1ea C:\Program Files\MATLAB\R2014b\bin\win64\mvm.dll+00782826 mvm::detail::initLocalMvmHack+00000490

[ 64] 0x00000000fa8ef7a8 C:\Program Files\MATLAB\R2014b\bin\win64\mvm.dll+00784296 mvm::detail::SessionImpl::privateSession+00000376

[ 65] 0x00000000fa8ef9f8 C:\Program Files\MATLAB\R2014b\bin\win64\mvm.dll+00784888 mvm::detail::SessionImpl::privateSession+00000968

[ 66] 0x0000000140005ddd C:\Program Files\MATLAB\R2014b\bin\win64\matlab.exe+00024029

[ 67] 0x0000000140007870 C:\Program Files\MATLAB\R2014b\bin\win64\matlab.exe+00030832

[ 68] 0x00000000776c59bd C:\WINDOWS\system32\kernel32.dll+00088509 BaseThreadInitThunk+00000013

[ 69] 0x00000000777fa2e1 C:\WINDOWS\SYSTEM32\ntdll.dll+00172769 RtlUserThreadStart+00000033

This error was detected while a MEX-file was running. If the MEX-file

is not an official MathWorks function, please examine its source code

for errors. Please consult the External Interfaces Guide for information

on debugging MEX-files.

If this problem is reproducible, please submit a Service Request via:

http://www.mathworks.com/support/contact_us/

A technical support engineer might contact you with further information.

Thank you for your help.** This crash report has been saved to disk as C:\Users\nd5583\AppData\Local\Temp\matlab_crash_dump.4692-1 **

mcTest.m

Nicholas Dana

unread,

Apr 18, 2016, 2:59:49 PM4/18/16

to mcx-users

Typo in above, the 'gaussian' srctype is commented out in the code attached.

Qianqian Fang

unread,

Apr 19, 2016, 12:50:49 AM4/19/16

to mcx-...@googlegroups.com

On 04/18/2016 02:58 PM, Nicholas Dana wrote:

I've been using MCX to estimate fluences for some photoacoustic and diffuse reflectance applications, but I'm getting some consistent crashes when using the 'pattern' srctype.

My current implementation uses a previous flux output (at a shorter wavelength) to generate a slice (i.e. pattern image) that corresponds to what fluorescent sources would be in the tissue at that slice. I then scale this image to [0-1] and define cfg.srcpos, cfg.srcparam1 and cfg.srcparam2 based on where I generated the slice from. I also modify the cfg.prop to reflect the optical properties at a different wavelength and keep the cfg.nphoton low for testing purposes. All other cfg parameters stay the same. When I attempt to run it thus, I invariably get a MATLAB crash and a lengthy error detail.

hi Nicholas

a quick comparison between your script and the built-in 'pattern'
source testing script in mcxlab/examples/demo_mcxlab_srctype.m:

https://github.com/fangq/mcx/blob/master/mcxlab/examples/demo_mcxlab_srctype.m#L116-L152

I found your srcparam1 and srcparam2 have reverted
dimensions of your input pattern array (pat). I am not sure if
this will cause a crash, but perhaps it is worth swapping
srcparam1 and srcparam2 definitions and see what happens.

Also, if you know how to recompile mcx, I strongly recommend
upgrading to the latest git version. There have been many
updates since the last release, and I am working on making
a new release, hopefully make it ready in the next few days.

I am happy to debug this problem, if you can send me
your mcxOut.mat file and your cfg.prop values.

Qianqian

--
You received this message because you are subscribed to the Google Groups "mcx-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to mcx-users+...@googlegroups.com.
To post to this group, send email to mcx-...@googlegroups.com.
Visit this group at https://groups.google.com/group/mcx-users.
For more options, visit https://groups.google.com/d/optout.

Nicholas Dana

unread,

Apr 19, 2016, 2:00:03 PM4/19/16

to mcx-users

Dr. Fang,

Thank you for your quick reply. Based on your suggestion, I swapped srcparam1 and srcparam2, as well as modifying which corners they point to, as so:

cfg.srcpos = [0 0 ii];

cfg.srcparam1 = [0 size(cfg.vol,2) 0 size(pat,2)];

cfg.srcparam2 = [size(cfg.vol,1) size(cfg.vol,2) 0 size(pat,1)];

The crash still occurred in the same fashion for all cases, as before. Below are my cfg.prop values

cfg.prop = [

0.0000 0.0010 1.0000 1.0000

0.0000 0.0100 1.0000 1.0000

0.1585 0.4358 0.9100 1.4000

1.0000 0.1000 0 1.5000];

I should say that I'm using a very small voxel size (10 um) as I'm simulating short wavelength light. Originally I had had some issues with defining the cfg.unitinmm = 0.01. A work around I found was to let the model interpret the voxel size as default, but scale the absorption & scattering to reflect grid units, as opposed to millimeters. This is how I've come up with the numbers I have. As such, my vol is large (500x500x500) and mcxOut.mat is large also. I've included a link to it on dropbox for you to download if you will. I can, of course, modify the solution space to use a smaller volume for testing purposes, but this is the same output as was generated previously and which worked when using the 'gaussian' srctype.

https://www.dropbox.com/s/o34t59ajewryr65/mcxOut.mat?dl=0

As far as version, I'm using mcxlab 0.9.7-2, from 2014/09/26. This is the most recent version I'm seeing on sourceforge. I see that http://mcx.space/#mmc has a link button that suggests there is v2016.4 of MCX, but I don't see such a version at sourceforge (am I missing something?). If there is a newer version, I can try recompiling.

Thanks in advance!

Nicholas Dana

unread,

Apr 20, 2016, 6:33:58 PM4/20/16

to mcx-users

So I tried recreating the issue by running mcxlab on a 2nd machine, still Windows 7 with CUDA 4.0. The script either caused Matlab to crash outright, or appeared to execute, but flux.data a 3D array full of nothing but NaN. It seemed to only execute the script without crashing IF I had previously run another mcxlab instance in matlab prior running mcTest.m, though that was intermittent. I tested on previous machine and it exhibited same behavior, sometimes executing but outputting NaN, but usually crashing. The on-screen output of the script looked normal, save that values reported for normalization factor, absorbed energy, etc were garbage due to the NaN values (e.g. normalization factor: #NAN0000x1)

Hope this helps

Qianqian Fang

unread,

Apr 20, 2016, 7:28:25 PM4/20/16

to mcx-...@googlegroups.com

On 04/20/2016 06:33 PM, Nicholas Dana wrote:

So I tried recreating the issue by running mcxlab on a 2nd machine, still Windows 7 with CUDA 4.0. The script either caused Matlab to crash outright, or appeared to execute, but flux.data a 3D array full of nothing but NaN. It seemed to only execute the script without crashing IF I had previously run another mcxlab instance in matlab prior running mcTest.m, though that was intermittent. I tested on previous machine and it exhibited same behavior, sometimes executing but outputting NaN, but usually crashing. The on-screen output of the script looked normal, save that values reported for normalization factor, absorbed energy, etc were garbage due to the NaN values (e.g. normalization factor: #NAN0000x1)

hi Nicholas

as I mentioned earlier, I am working on making a new release (v2016.4)
both Linux and Mac are almost ready, now I am testing on windows.
Once this is done (hopefully by tomorrow), I will upload the precompiled
binaries, as well as the source code (github always has the latest).

I did run your script yesterday. I noticed you used a very large
vol array. This is not efficient for debugging purposes. You also
defined some extra fields that are not supported by mcxlab.
Especially, you defined cfg.vox_size, which is not a valid mcxlab
option. You should use cfg.unitinmm.

after fixing most of these issues, I was able to run your simulation,
I did not get a crash. The results look reasonable (screenshots attached).

Here are my scripts, first for the Gaussian source:

load mcxOut
oldvol=cfg.vol;
cfg.vol=oldvol(1:5:end,1:5:end,1:5:end); % downsample by 5x
cfg.unitinmm=0.05;    % set length unit accordingly
cfg.srcpos=[50 50 1];
cfg.srcparam1=[90/5 0 0 0];
cfg.srcparam2=[0 0 0 0];
cfg.detpos=[50 10 1 40];
cfg.respin=1; % I have a dedicated GPU

fl=mcxlab(cfg);
hs=slice(log10(double(fl.data)),50, 50, 10); set(hs,'linestyle','none')

Launching MCXLAB - Monte Carlo eXtreme for MATLAB & GNU Octave ...
Running simulations for configuration #1 ...
WARNING: redundant field 'WL'
mcx.dim=[100 100 100];
mcx.seed=1732221598;
mcx.nphoton=5e+08;
mcx.srcpos=[50 50 1];
mcx.srcdir=[0 0 1];
mcx.srctype='gaussian';
mcx.srcparam1=[18 0 0 0];
mcx.medianum=4;
mcx.tstart=0;
mcx.tend=5e-09;
mcx.tstep=5e-09;
mcx.autopilot=1;
mcx.gpuid=1;
mcx.respin=1;
mcx.isreflect=0;
mcx.isrefint=0;
WARNING: redundant field 'issavedet'
mcx.detnum=1;
mcx.maxdetphoton=1e+07;
WARNING: redundant field 'vox_size'
mcx.unitinmm=0.05;
mcx.srcparam2=[0 0 0 0];
###############################################################################
#                      Monte Carlo eXtreme (MCX) -- CUDA                      #
#          Copyright (c) 2009-2016 Qianqian Fang <q.fang at neu.edu>          #
#                             http://mcx.space/                               #
#                                                                             #
#         Computational Imaging Laboratory (CIL) [http://fanglab.org]         #
#            Department of Bioengineering, Northeastern University            #
###############################################################################
#    The MCX Project is funded by the NIH/NIGMS under grant R01-GM114365      #
###############################################################################
$Rev::c52b47 $ Last $Date::2016-04-06 01:50:15 -04$ by $Author::Qianqian Fang $
###############################################################################
- variant name: [Fermi] compiled for GPU Capability [100] with CUDA [7000]
- compiled with: RNG [xorshift128+] with Seed Length [4]
- this version CAN save photons at the detectors

GPU=1 (GeForce GTX 980 Ti) threadph=11097 extra=13600 np=500000032 nthread=45056 maxgate=1 repetition=1
initializing streams ...    init complete : 0 ms
requesting 3072 bytes of shared memory
lauching MCX simulation for time window [0.00e+00ns 5.00e+00ns] ...
simulation run# 1 ...     kernel complete:     16481 ms
retrieving fields ...     transfer complete:    16483 ms
data normalization complete : 16484 ms
normalizing raw data ...    normalization factor alpha=159.999985
simulated 500000000 photons (500000000) with 45056 threads (repeat x1)
MCX simulation speed: 30354.54 photon/ms
total simulated energy: 500000032.00    absorbed: 18.48011%
(loss due to initial specular reflection is excluded in the total)

Then for the pattern source:

cfg.srctype='pattern';ii=80/5;pat = flux.data(1:5:end,1:5:end,ii*5).*(cfg.vol(:,:,ii)==2);patmax = max(pat(:));cfg.srcpattern= pat./patmax; cfg.srcpos=[1 1 ii]; cfg.srcparam1=[size(cfg.vol,1),0 0 size(pat,1)];cfg.srcparam2=[0 size(cfg.vol,1) 0 size(pat,2)];fl2=mcxlab(cfg);hs=slice(log10(double(fl2.data)),50, 50, 10); set(hs,'linestyle','none') Launching MCXLAB - Monte Carlo eXtreme for MATLAB & GNU Octave ... Running simulations for configuration #1 ... WARNING: redundant field 'WL' mcx.dim=[100 100 100]; mcx.seed=1732221598; mcx.nphoton=5e+08; mcx.srcpos=[1 1 16]; mcx.srcdir=[0 0 1]; mcx.srctype='pattern'; mcx.srcparam1=[100 0 0 100]; mcx.medianum=4; mcx.tstart=0; mcx.tend=5e-09; mcx.tstep=5e-09; mcx.autopilot=1; mcx.gpuid=1; mcx.respin=1; mcx.isreflect=0; mcx.isrefint=0; WARNING: redundant field 'issavedet' mcx.detnum=1; mcx.maxdetphoton=1e+07; WARNING: redundant field 'vox_size' mcx.unitinmm=0.05; mcx.srcparam2=[0 100 0 100]; mcx.srcpattern=[100 100]; ############################################################################### # Monte Carlo eXtreme (MCX) -- CUDA # # Copyright (c) 2009-2016 Qianqian Fang <q.fang at neu.edu> # # http://mcx.space/ # # # # Computational Imaging Laboratory (CIL) [http://fanglab.org] # # Department of Bioengineering, Northeastern University # ############################################################################### # The MCX Project is funded by the NIH/NIGMS under grant R01-GM114365 # ############################################################################### $Rev::c52b47 $ Last $Date::2016-04-06 01:50:15 -04$ by $Author::Qianqian Fang $ ############################################################################### - variant name: [Fermi] compiled for GPU Capability [100] with CUDA [7000] - compiled with: RNG [xorshift128+] with Seed Length [4] - this version CAN save photons at the detectors GPU=1 (GeForce GTX 980 Ti) threadph=11097 extra=13600 np=500000032 nthread=45056 maxgate=1 repetition=1 initializing streams ... init complete : 1 ms requesting 3072 bytes of shared memory lauching MCX simulation for time window [0.00e+00ns 5.00e+00ns] ... simulation run# 1 ... kernel complete: 48022 ms retrieving fields ... transfer complete: 48024 ms data normalization complete : 48024 ms normalizing raw data ... normalization factor alpha=5879.846680 simulated 500000000 photons (500000000) with 45056 threads (repeat x1) MCX simulation speed: 10413.85 photon/ms total simulated energy: 13605796.54 absorbed: 14.32143% (loss due to initial specular reflection is excluded in the total)

by the way, try to set up a dedicated GPU if you want to take full
advantage of MCX. You can see from my log, if I use a dedicated GPU setting
(a Maxwell, 980Ti), I can get a speed of 30354 photon/ms (for Gaussian src),
but if I use your setting, cfg.respin=3334, I was only able to get 1021 photon/ms.
You got a 30x slow-down using a non-dedicated GPU!

Qianqian

gaussian_profile.png

fluxdata_profile.png

Nicholas Dana

unread,

Apr 20, 2016, 9:22:06 PM4/20/16

to mcx-users

Thanks again for your fast response Dr. Fang. I copied your code exactly and experienced the same issues I had had.

clear; close all;clc

load mcxOut
oldvol=cfg.vol;


cfg.vol=oldvol(1:5:end,1:5:end,1:5:end); % downsample by 5x
cfg.unitinmm=0.05;    % set length unit accordingly
cfg.srcpos=[50 50 1];
cfg.srcparam1=[90/5 0 0 0];
cfg.srcparam2=[0 0 0 0];  
cfg.detpos=[50 10 1 40];
cfg.respin=1;  %

 I have a dedicated GPU
cfg.nphoton = 1e6;



fl=mcxlab(cfg);
hs=slice(log10(double(fl.data)),50, 50, 10); set(hs,'linestyle','none')

Launching MCXLAB - Monte Carlo eXtreme for MATLAB & GNU Octave ...
Running simulations for configuration #1 ...

mcx.dim=[100 100 100];
mcx.seed=1.73222e+09;
mcx.nphoton=1e+06;

mcx.srcpos=[50 50 1];
mcx.srcdir=[0 0 1];
mcx.srctype='gaussian';
mcx.srcparam1=[18 0 0 0];
mcx.medianum=4;

mcx.unitinmm=0.05;

mcx.tstart=0;
mcx.tend=5e-09;
mcx.tstep=5e-09;
mcx.autopilot=1;
mcx.gpuid=1;
mcx.respin=1;
mcx.isreflect=0;
mcx.isrefint=0;

mcx.issavedet=0;
mcx.detnum=1;
mcx.maxdetphoton=1e+07;

mcx.srcparam2=[0 0 0 0];

autopilot mode: setting thread number to 12288, block size to 64 and time gates to 1

###############################################################################
# Monte Carlo eXtreme (MCX) -- CUDA #

# Copyright (c) 2009-2013 Qianqian Fang <fangq at nmr.mgh.harvard.edu> #
# #
# Martinos Center for Biomedical Imaging, Massachusetts General Hospital #
###############################################################################
$MCX $Rev:: 325 $ Last Commit $Date:: 2014-09-23 13:34:57#$ by $Author:: fangq$
###############################################################################
- variant name: [Fermi MCX] compiled for GPU Capability [100] with CUDA [4000]
- compiled with: RNG [Logistic-Lattice] with Seed Length [5]

- this version CAN save photons at the detectors

threadph=81 oddphotons=4672 np=1000000 nthread=12288 repetition=1

initializing streams ... init complete : 1 ms

requesting 0 bytes of shared memory

lauching MCX simulation for time window [0.00e+00ns 5.00e+00ns] ...

simulation run# 1 ... kernel complete: 1235 ms
retrieving fields ... transfer complete: 1250 ms
normalizing raw data ... normalization factor alpha=91940.070313
data normalization complete : 1265 ms
simulated 1000000 photons (1000000) with 12288 threads (repeat x1)
MCX simulation speed: 822.37 photon/ms
exit energy: 8.91699609e+04 + absorbed energy: 9.10830063e+05 = total: 1.00000000e+06

And the output image looks similar to yours, albeit much noisier. Then, with the second portion:

cfg.srctype='pattern';
ii=80/5;
pat = flux.data(1:5:end,1:5:end,ii*5).*(cfg.vol(:,:,ii)==2);
patmax = max(pat(:));
cfg.srcpattern= pat./patmax; 
cfg.srcpos=[1 1 ii]; 
cfg.srcparam1=[size(cfg.vol,1),0 0 size(pat,1)];
cfg.srcparam2=[0 size(cfg.vol,1) 0 size(pat,2)];
fl2=mcxlab(cfg);
hs=slice(log10(double(fl2.data)),50, 50, 10); set(hs,'linestyle','none')

Launching MCXLAB - Monte Carlo eXtreme for MATLAB & GNU Octave ... Running simulations for configuration #1 ...

mcx.dim=[100 100 100]; mcx.seed=1.73222e+09; mcx.nphoton=1e+06;

mcx.srcpos=[1 1 16]; mcx.srcdir=[0 0 1]; mcx.srctype='pattern'; mcx.srcparam1=[100 0 0 100]; mcx.medianum=4;

mcx.unitinmm=0.05;

mcx.tstart=0; mcx.tend=5e-09; mcx.tstep=5e-09; mcx.autopilot=1; mcx.gpuid=1; mcx.respin=1; mcx.isreflect=0; mcx.isrefint=0;

mcx.issavedet=0; mcx.detnum=1; mcx.maxdetphoton=1e+07;

mcx.srcparam2=[0 100 0 100]; mcx.srcpattern=[100 100];

autopilot mode: setting thread number to 12288, block size to 64 and time gates to 1

############################################################################### # Monte Carlo eXtreme (MCX) -- CUDA #

# Copyright (c) 2009-2013 Qianqian Fang <fangq at nmr.mgh.harvard.edu> # # # # Martinos Center for Biomedical Imaging, Massachusetts General Hospital # ############################################################################### $MCX $Rev:: 325 $ Last Commit $Date:: 2014-09-23 13:34:57#$ by $Author:: fangq$ ############################################################################### - variant name: [Fermi MCX] compiled for GPU Capability [100] with CUDA [4000] - compiled with: RNG [Logistic-Lattice] with Seed Length [5]

- this version CAN save photons at the detectors

threadph=81 oddphotons=4672 np=1000000 nthread=12288 repetition=1

initializing streams ... init complete : 0 ms

requesting 0 bytes of shared memory

lauching MCX simulation for time window [0.00e+00ns 5.00e+00ns] ...

simulation run# 1 ... kernel complete: 1453 ms retrieving fields ... transfer complete: 1456 ms normalizing raw data ... normalization factor alpha=1.#QNAN0 data normalization complete : 1470 ms simulated 1000000 photons (1000000) with 12288 threads (repeat x1) MCX simulation speed: 695.89 photon/ms exit energy: 1.#QNAN000e+00 + absorbed energy: 1.#QNAN000e+00 = total: 1.#QNAN000e+00

And then Matlab throws an error trying to execute slice. Similar to previously, fl2 is full of NaN. The only difference I see is you're using Cuda 7.0, presumably on Linux. I have the ability to install Ubuntu alongside Windows 7 on this machine, which perhaps I'll look into if the updated versions don't resolve this issue.

Per your recommendation for a dedicated graphics card, this machine has 2 graphics cards, with the secondary card being an Nvidia GT 730. While I can certainly use mcxlab on either card (by specifying gpuid), I still get a kernel launch timeout error, regardless of which card is being used. I've looked in the previous posts edited TdrDelay, which helps, but I still get a timeout even with a dedicated GPU. Perhaps there's another setting I missed?

Thanks again for your help!

...

gaussian_profile_ND.png

Reply all

Reply to author

Forward