ogiroux / freestanding Goto Github PK

Hi,
just after reading "https://devblogs.nvidia.com/cuda-turing-new-gpu-compute-possibilities" I became aware of your Cppcon talk and how advanced / improved Volta and later archs. are for C++..
impressed to see a mac.sh..
can share if you already tried on MacOS or script is only for future purposes?
I say as currently there are no drivers for Volta or newer GPUs for MacOS..
or they are? really I found Nvidia Web Driver - 387.10.10.15.15.108 supports Titan V and altough Nvidia removed from web shortly after posting I was able to gather the driver from some good soul on the Internet..
I tested on High Sierra + Xcode 9.4+CUDA 10 and it works OK
I had to do only 2 changes:

1. so I updated build script to point to /Developer/NVIDIA/CUDA-10.0/bin/nvcc
2. I removed --compiler-bindir /usr/local/opt/llvm@5/bin/clang totally..
   seems to work ok without it.. can point why it's needed?

also another question is I see :

<simt/cstdalign> | Except Mac OS X. | <stdalign.h>

can share what's unique in MacOS to not support it?

thanks..
Oscar.

Hi,

I am trying to compile Rapids 0.14 on ppc64le but I am getting the next errors after doing some operations with cuDF

...

--- JIT compile log for ---

../../libcxx/include/limits(408): error: floating constant is out of range

../../libcxx/include/limits(409): error: floating constant is out of range

../../libcxx/include/limits(431): error: floating constant is out of range

I noted Rapids defines limits for x86 while those limits are different on ppc64le, in particular,

x86 limits are

#define __LDBL_MAX__ 1.18973149535723176502e+4932L
#define __LDBL_MIN__ 3.36210314311209350626e-4932L
#define __LDBL_EPSILON__ 1.08420217248550443401e-19L
#define __LDBL_DENORM_MIN__ 3.64519953188247460253e-4951L

while ppc64le GNU's (gcc 7.4) defines

#define __LDBL_MAX__ 1.79769313486231580793728971405301e+308L
#define __LDBL_MIN__ 2.00416836000897277799610805135016e-292L
#define __LDBL_EPSILON__ 4.94065645841246544176568792868221e-324L
#define __LDBL_DENORM_MIN__ 4.94065645841246544176568792868221e-324L

I noted there is support for ppc64le here:

https://github.com/ogiroux/libcxx/blob/36156f0962149ef19ce778110642150b456a182b/include/support/ibm/limits.h

So the question is what flag should be used to enable ppc64le support.

Thanks,

Benjamin

Hi @ogiroux,

thank you for your work, these free-standing functions are useful.

But I have to admit, what we really need and what we are constantly re-implementing are modern C++ primitives with __host__ __device__ annotations such as:

• containers: std::array and std::tuple (and their iterators)
• bonus container: std::vector of which one does actually just need its iterators on device, no resize needed
• type: std::complex

One does not need to take thrust for those, those types should just work out-of-the box with CUDA in a standard-compatible way.