Sean Baxter

 There is no _attribute-specifier-seq_ in that position, by the standard. I do parse attributes after the ( ), (accessible through the _lambda-specifiers_ in the grammar). The...

I have given a good deal of thought to this before. My problem is potentially circular definitions, and cases where you have to query the noexcept state of a function...

thanks, i'm taking a look.

Surprising bug. Circle sees this extern template declaration and therefore doesn't do implicit instantiation of the destructor: fs_dir.h: ```cpp // Use explicit instantiations of these types. Any inconsistency in the...

Is this what you have on your system? ```cpp #ifndef __CUDA_ARCH__ #define __DEF_IF_HOST { } #else /* !__CUDA_ARCH__ */ #define __DEF_IF_HOST ; #endif /* __CUDA_ARCH__ */ ``` Not sure what...

The definition of `atomicAdd` that you want is in `device_atomic_functions.hpp`. (Yes there are definitions in different files.) But including that breaks on undeclared identifiers `__any` and `__all`. I remember removing...

Like circle or nvc++. The current headers is attaching a bogus { } definition to __host__ atomicAdd, and the correct definition to __device__ atomicAdd, and using CUDA_ARCH to separate the...

Ya, it seems reasonable to make that work. I will put it on the list.

I started language builtin f-strings. Reserved f"" for that. But I haven't settled on a grammar for formatting, or really compelling way to plug in with user-defined types. Do you...

Thanks for the report. I'll get to work on it this week.