Isuru Fernando

BinDeps only allows one place to extract the binary. `depsdir(dep)`

> It is assumed that the binaries unpack the libraries into usr/lib. If given a String as its data argument, provides a custom search path for the binaries I should...

Same error on appveyor as well. https://ci.appveyor.com/project/isuruf/symengine-jl/build/1.0.6/job/gbemo86fo8niqjws

Going off topic here. `backend='llvm'` and `backend='lambda'` have wildly different performance numbers in SymEngine. ```python In [10]: eval_expr3 = sm.lambdify((a, b, c, x, y, z), [expr3]) In [9]: %timeit eval_expr3(0.4,...

The remaining failures count the number of times a function is called and the count increased because of ca8b6acc3c73314264137c30b6026ee492ad4b86. I suppose it's okay to update those counters?

Those counters would now depend on whether the function was run with torch.compile or not.

> Are there any other frameworks that support this functionality? Yes, pytorch does and opencv too. > What level of performance benefit it brings to have this feature in oneDNN?...

@vpirogov, Pytorch's `nearest-exact` algorithm and oneDNN's `nearest` are the same. Pytorch's `nearest` is slightly different. oneDNN's nearest (Pytorch's `nearest-exact`) is implemented as ```C++ static inline float linear_map(dim_t y, dim_t y_max,...

The issue is that allocation might happen in C side, for eg: in VkFFT where phase vectors are allocated and there's no way to plug in a custom allocator.

Taking the example of pyvkfft, I was thinking of changing the line https://github.com/vincefn/pyvkfft/blob/5b51a7bcb0d93581c5ea8dcf7edbe77d1b191f2f/src/vkfft_opencl.cpp#L15 from ```C #include "vkFFT.h" ``` to ```C #include #include "vkFFT.h" ``` and that should work without any...