@@ -596,6 +596,7 @@ static bool mustPreserveGV(const GlobalValue &GV) {
if (const Function *F = dyn_cast(&GV))
return F->isDeclaration() || F->getName().startswith("__asan_") ||
F->getName().startswith("__sanitizer_") ||
+ F->getName() == "__divti3" |
https://github.com/arsenm commented:
128-bit division should already work, we have an IR integer division expansion
for > 64-bit divides. I think moving towards getting the infrastructure to a
place where we can link in compiler-rt binaries is a good thing, but I don't
think we're in a positio
llvmbot wrote:
@llvm/pr-subscribers-clang-driver
Author: Yaxun (Sam) Liu (yxsamliu)
Changes
Currently nvcc supports 128 bit int division in device code. This patch adds
support of 128 bit int division to HIP.
It builds lib functions for 128 bit division in compiler-rt for amdgcn target.