Add GPU acceleration

[mborland]

Now that the CUDAfication of math is underway many of the same techniques can be applied here to enable CUDA and SYCL support throughout the library. The main point of difficulty for SYCL will be removing recursion from the constructors of the non-fast types.

[mborland]

spir64-unknown-unknown aka SYCL does not support 128-bit integer types even thought the host compiler (ICPX) does. Need to add additional macro checks.

[mborland]

Deferred for now. Development will exist on cuda-development branch. Maybe for version 3 post-boost review or when we receive a strong demand signal

[mborland]

Once #1352 is in, we will have known support for the backend running on CUDA. Since we use so little of the standard library in the decimal parts of the library I think we should be able to add NVCC rather painlessly.

[mborland]

@xhd1221 Since you've given the thumbs up on this issue, I wanted to let you know that adding CUDA support is underway. Should launch in Boost 1.92 this summer.

[xhd1221]

@xhd1221 Since you've given the thumbs up on this issue, I wanted to let you know that adding CUDA support is underway. Should launch in Boost 1.92 this summer.

Thanks for the update. Great to hear CUDA support is underway — looking forward to trying it in Boost 1.92.