One curious question -- unless I've misread things horribly, it looks like the D benchmarks for Lomuto branch-free are consistently slower than for C++. Any idea why that is? I would expect gcc/gdc and clang/ldc to produce effectively identical results for code like this.
Wrt. the LDC results, LDC v1.17 was shipped with LLVM 8.0.1, while the used clang is v10 based on LLVM 10, so that might account for some slight diffs.
