================
@@ -2931,7 +2931,8 @@ class CodeGenFunction : public CodeGenTypeCache {
                            RawAddress *Alloca = nullptr);
   RawAddress CreateMemTemp(QualType T, CharUnits Align,
                            const Twine &Name = "tmp",
-                           RawAddress *Alloca = nullptr);
+                           RawAddress *Alloca = nullptr,
+                           LangAS OverrideSemaForOpenCL = LangAS::Default);
----------------
vtjnash wrote:

Yeah, this was supposed to be a bit gross to keep anyone from using it. But how 
do you feel about this branch to kill it once these 2 PRs land and then 
https://github.com/vtjnash/llvm-project/pull/new/jn/opencl-param-addrspace (WIP 
for commit list, but passes tests). There's still awkwardness with Implicit 
parameters not being added by Sema, but it at least moves to the construction 
site, instead of the usage site. I got confused before by the number of 
assertion and test failures and thought that meant OpenCL Sema couldn't be 
fixed to remove it's special case here, but it turns out I seem to have been 
wrong about that, and I had just not refactored some of the code enough before.

After that branch, only OpenMP would be a special case in this function (maybe 
unfixable, since some of its VarDecls are secretly not allocas but are rather 
managed malloc calls from CGOpenMPRuntime)

https://github.com/llvm/llvm-project/pull/181256
_______________________________________________
cfe-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

Reply via email to