doe300

Unless you explicitly disabled `REGISTER_POKE_KERNEL`, the "Execution: failed" error means that the kernel execution timed out. This could be due to one of two problems: - The execution of a...

When `VC4CL` is built in emulation mode (`MOCK_HAL` enabled), it is built without ICD loader support. Thus, to run `clinfo` with the emulation version, you need to make sure, `clinfo`...

I tried it out myself too yesterday and it took a while but finished. Which is bad in this case, since it doesn't help to find the problem. So from...

Sorry, didn't have time to look into that. Since the error is not reproducible on the emulator, I will have to take a look at the generated source code, but...

So it looks like for a work-group size of 12 (the maximum), the execution hangs, probably somewhere in `barrier(...)`. For a work-group-size of less than 12, it at least passes,...

So, turns out I did semaphore access wrong, that is why the execution timed out. Now the execution passes, but the result seems to still be wrong.

The difference between `6579.5039` and `6579.5034` is 1 ULP, same for `6708.3955` and `6708.3950`. The problem here is probably, how the constant `3.14` gets stored (`3.140000` vs. `3.1399998`). Can you...

Thanks for investing time to better analyse this error. > See the differences when using FACTOR=3.1415 or FACTOR=2.0. The difference here is that `2.0` can be exactly represented as `float`,...

So I can partially reproduce the behavior you are seeing: Up to 10240, everything is okay, for 102400 I get a lot of errors. When I get back to

So for the float multiplication, I did some testing and it looks like the `fmul` instruction does different rounding than the host CPU. I will have to try to find...