Cory Fields

> > Or are we adding it to the linker elsewhere? > > We are passing the min version as part of `-platform_version` in `darwin_LDFLAGS`. Note that we also check...

@TheCharlatan Mind adding a commit to fixup the includes [as per this conversation](https://github.com/bitcoin/bitcoin/pull/28051#pullrequestreview-2041704403)?

Here's a godbolt test that shows how various compilers perform: https://gcc.godbolt.org/z/nTadqEP83 To test with/without builtins, comment/un-comment the `DISABLE_BUILTIN_BSWAPS` define at line 14. From my tests: - clang: all c++20 supporting...

Added a quick note about `std::byteswap` and c++23.

@aureleoules Thanks! These numbers really don't make much sense to me and don't match what I've seen locally, but I'll work on trying to reproduce.

From some more local tests, it looks like it's the clz changes slowing things down for whatever reason. Still investigating.

I've dropped the clz changes as a test and kept only the endian/byteswap change. Locally on my machine the benchmarks look the same before and after. If the corecheck benchmarks...

Aha, I finally tracked down the culprit! No wonder the benchmarks weren't making sense! The problem was the removal of this from `crypto/common.h`: ```c++ #if defined(HAVE_CONFIG_H) #include #endif ``` Turns...

Converted to a draft while I'm still messing with this. I added a new commit to add the `bitcoin-config.h` include where it's necessary. I'm not done investigating that yet, but...

I've pushed a commit which temporarily puts the includes back in the low-level headers for the sake of addressing the benchmarks first, setting aside the possibility of missing defines. @aureleoules's...