Siu Chi Chan

branch rebased and resolved conflicts

We are actively looking at function pointer support in hip-clang at the moment. Unfortunately, it's not trivial to implement on the GPU due to several technical challenges that we are...

> The error message for the failing tests is curious. `error while loading shared libraries: libhsa-runtime64.so.1: cannot open shared object file: No such file or directory` That's a known problem...

@jeffdaily do we still need an explicit wait if we clear the asyncops vector?

For the first copy test that failed, I could see that the host is trying to launch a kernel with name: `_ZN6thrust6system4cuda6detail5bulk_6detail15launch_by_valueILj0ENS4_9cuda_taskINS3_14parallel_groupINS3_16concurrent_groupINS3_5agentILm1EEELm0EEELm0EEENS4_7closureINS2_17for_each_n_detail15for_each_kernelENS_5tupleINS4_6cursorILj0EEENS_12zip_iteratorINSG_INS_6detail15normal_iteratorINS_7pointerIiNS2_3tagENS_11use_defaultESO_EEEENSL_INS_10device_ptrIfEEEENS_9null_typeESU_SU_SU_SU_SU_SU_SU_EEEENSK_16wrapped_functionINSK_23unary_transform_functorINS_8identityIiEEEEvEEjSU_SU_SU_SU_SU_SU_EEEEEEEEvT0_` However, only kernel function that is present in...

Currently the fix is only available in HCC's development branch (clang_tot_upgrade), you'll have to build the compiler from source to get it.

I could only reproduce this issue if there's a mismatch between the function declaration and the definition. More specifically, when the declaration is missing the __host__ keyword: `// failing` `extern...

@whchung I'd like to get PR1245 first if possible since it's higher priority. This PR will probably have be updated to adapt to those changes.

By following the steps above, I'm getting compilation errors about `struct __half` and `sobol32_state` below. Also, it's using `g++` to compile, is that expected? scchan@97e6e7dfd750:~/code/mxnet$ make VERBOSE=1 Makefile:270: WARNING: Significant...

I wasn't able to reproduce the memory error you reported. I'm using one GPU, how many GPUs are you using? The error message you are getting suggests that memory accesses...