atrosinenko added a subscriber: scanon. atrosinenko added a comment. On one hand, this `clzdi()` implementation should be completely harmless when fallback is implemented via `clzsi()`. On the other hand, it may possibly be completely removed. This code snippet mimics one already existing for `clzsi` in int_types.h and for currently supported targets it can probably be even as simple as this (provided we can use C99):
#if ULLONG_MAX == 0xFFFFFFFFFFFFFFFF #define clzdi __builtin_clzll #define ctzdi __builtin_ctzll #else #error could not determine appropriate clzdi macro for this system #endif ================ Comment at: compiler-rt/lib/builtins/int_lib.h:149 +static int __inline clzdi(di_int a) { +#if defined __LP64__ ---------------- @scanon This function was copied from fp_lib.h almost unchanged, so it can be used from fp_extend.h (the version from that file is almost identical except it does not use the properly-sized `clzsi()` macro). But is this manual fallback implementation useful nowadays? As I can see, the implementation in fp_lib.h remained almost untouched since 2010 when this file was uploaded via {rGb1fdde1d01}. Now, `__builtin_clzll` is used in many places of compiler-rt/builtins library for seemingly the same purpose. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D86221/new/ https://reviews.llvm.org/D86221 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits