Exception : LoadWorld : File does not start with HPCEHeatWorldV0

[luisi-at]

System: macOS 10.14 GPU: AMD Radeon Pro 460

Running the v5 task code, I keep getting an issue when trying to produce an output, be it binary, text or .bmp. The code compiles successfully but upon running gets the following exception:

Has anyone else suffered from the same issue and how did you rectify it?

Thanks.

[JBacchelli]

It works fine on my local machine (macOS 14.10.1, Intel Iris Graphics 650), but I get the same error for tests 20 to 25.

For example, in jjd06_logs_2019-02-08_15-51-10/log/test_20.log :

Loaded world with w=64, h=64
Stepping by dt=0.1 for n=1
Loaded world with w=64, h=64
Stepping by dt=0.1 for n=1
Found 1 platforms
    Platform 0 : NVIDIA Corporation
Choosing platform 0
Found 1 devices
    Device 0 : GeForce GTX 970
Choosing device 0
Exception : LoadWorld : File does not start with HPCEHeatWorldV0.

[Odog1996]

I had this issue on V5. The specification is very vague on the creation of the packed array but this was the cause of the error for me:

kernel.setArg(1, packed)

This is incorrect and should be set to:

kernel.setArg(1, buffProperties)

with packed only passed in as part of the write buffer:

queue.enqueueWriteBuffer(buffProperties, CL_TRUE, 0, cbBuffer, &packed[0]);

as the array you pass the OpenCL kernel needs to be declared as an OpenCL Buffer. Obviously your argument number is likely to differ from mine above.

[luisi-at]

This also occurs when there is no output piped to the bin/render_world, likely due to a segfault happening somewhere in the code. In my case this was found to be due to a poorly bracketed (!((world.properties[index] & hpce::Cell_Fixed) || (world.properties[index] & hpce::Cell_Insulator))) Check that the array during encoding is being indexed correctly, or that the kernel is being indexed correctly.

[jjd06]

Ah, the old buffer vs what's being buffered problem. It's good that this kind of confusion is being addressed now so that these things become second nature for CW5 & 6!