Morten Ofstad
2004-Oct-19 12:17 UTC
[LLVMdev] Patch for non standard usage of math functions
In lib/Analysis/ScalarEvolution.cpp there is a call to sqrt with an int64_t argument - To make this call in standard C++ a cast to double is required. In lib/CodeGen/LiveIntervalAnalysis.cppn there is a call to pow where the second argument is an unsigned int. The standard C++ call is with 'int' and the compiler does not know if it should promote the unsigned int to float or double so it cannot resolve the overloaded pow function. I solved this by casting to int... m. -------------- next part -------------- An embedded and charset-unspecified text was scrubbed... Name: diff.txt URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20041019/8bd807a5/attachment.txt>
Reasonably Related Threads
- [LLVMdev] Math Library Intrinsics as native intrinsics
- [LLVMdev] LiveIntervals, replace register with representative register?
- [LLVMdev] Math Library Intrinsics as native intrinsics
- [LLVMdev] Math Library Intrinsics as native intrinsics
- Why LLVM cannot optimize this?