Tim Blechmann

> A data point: I've tried this build on macOS 10.14 (Intel) and 11.6 (M1). On the former I seemed to be able to run it with thread pinning (i.e....

> I think we would have to find out which logical cpus belong to the same physical cores, put them on a vector and sort them so that adjacent elements...

> I guess there is no way around checking the actual HW topology? ah, quite possibly > With 2 DSP threads (= default) performance is great (almost 2x), but with...

it probably doesn't use the macro to generate the entry point functions (wild guess, didn't have a look into the code)

as a temporary workaround ``` set(ENV{SDKROOT} "${CMAKE_OSX_SYSROOT}") ``` seems to work. maybe this would also be a strategy for #226?

related to specializing `std::is_arithmetic`, this code gives me a compile error with catch2 v3 ``` TEST_CASE( "testme", "" ) { using test_type = strong::type< std::int8_t, struct tag, strong::arithmetic, strong::default_constructible, strong::equality...

> The failure with Windows C++20 is issue #18, so never mind that. I need to think a bit if having both a full set of comparison functions `=` in...

> This will need some fixes to the path normalization code in approval tests. do you have any pointers for the entry points to do these fixes?

it seems to be related to fmt's `ostream` support and `strong::ostreamable`: ``` #include #include #include using type = strong:: type< uint64_t, struct tag, strong::ostreamable, strong::default_constructible >; void foo() { fmt::print("{}",...

~~the `STONG_CONSTEXPR` caused problems for me:~~ ``` namespace fmt { template struct formatter : formatter { template // STRONG_CONSTEXPR decltype(auto) format(const Type& t, FormatContext& fc) const noexcept(noexcept(std::declval().format(std::declval(), fc))) { return...