I've had several WUs that would run for 13 seconds, pause and start over again in an endless loop of 13 second activity. The Intel OpenCL seems poorly supported too.

I've had several WUs that would run for 13 seconds, pause and start over again in an endless loop of 13 second activity. The Intel OpenCL seems poorly supported too.

That url is not available for us to view, It is for your use only, with your computers hidden we cannot see what the problem is,

But it's probably this:

WUs with very unusual Autocorr parameters

Claggy

Got the same behavior since 2 weeks on some WUs on all my PC with GPU:
Start computing then "Postponed: Cuda runtime, memory related failure, threadsafe temporary Exit", then waiting for reschedule and restart computing from 0...

Stderr.txt example:

setiathome_CUDA: Found 1 CUDA device(s):
Device 1: Quadro FX 880M, 1024 MiB, regsPerBlock 16384
computeCap 1.2, multiProcs 6
clockRate = 1210 MHz
In cudaAcc_initializeDevice(): Boinc passed DevPref 1
setiathome_CUDA: CUDA Device 1 specified, checking...
Device 1: Quadro FX 880M is okay
SETI@home using CUDA accelerated device Quadro FX 880M
pulsefind: blocks per SM 1 (Pre-Fermi default)
pulsefind: periods per launch 100 (default)
Priority of process set to BELOW_NORMAL (default) successfully
Priority of worker thread set successfully

setiathome enhanced x41zc, Cuda 2.30

Detected setiathome_enhanced_v7 task. Autocorrelations enabled, size 0k elements.
Work Unit Info:
...............
WU true angle range is : 0.545577
re-using dev_GaussFitResults array for dev_AutoCorrIn, 256 bytes
re-using dev_GaussFitResults+32x8 array for dev_AutoCorrOut, 256 bytes
Error on launch (ac_reducePartial<<<grid3, block3, blksize*sizeof(float3)>>>( (float *)dev_AutoCorrIn, dev_ac_partials )), file c:/[Projects]/__Sources/sah_v7_opt/Xbranch/client/cuda/cudaAcc_autocorr.cu, line 200: invalid configuration argument
Exiting
cudaAcc_free() called...
cudaAcc_free() running...
cudaAcc_free() PulseFind freed...
cudaAcc_free() Gaussfit freed...
cudaAcc_free() AutoCorrelation freed...
cudaAcc_free() DONE.
Cuda sync'd & freed.
Preemptively acknowledging a safe temporary exit->
Exit Status: 0
boinc_exit(): requesting safe worker shutdown ->
boinc_exit(): received safe worker shutdown acknowledge ->
Cuda threadsafe ExitProcess() initiated, rval 0
...

Maybe an incorrect argument when calling the functions:
int cudaAcc_FindAutoCorrelation(float *AutoCorrelation, int ac_fftlen, int fft_num )

and the CUDA kernel with the following dimGrid/dimBlock parameters:
int len = ac_fftlen/2;
int blksize = 256;
dim3 block3(blksize,1,1);
dim3 grid3(len/blksize,1,1);
CUDA_ACC_SAFE_LAUNCH( (ac_reducePartial<<<grid3, block3, blksize*sizeof(float3)>>>( (float *)dev_AutoCorrIn, dev_ac_partials )),true);

---

WU's on first