[PATCH] D72841: [RFC] Add support for pragma float_control, to control precision and exception behavior at the source level

[Melanie Blower via Phabricator via cfe-commits]

mibintc marked 3 inline comments as done.
mibintc added a comment.

@sepavloff Thanks a lot for your comments. I added a few replies and I have one question, added inline.



================
Comment at: clang/lib/Parse/ParsePragma.cpp:2522
+  IdentifierInfo *II = Tok.getIdentifierInfo();
+  PragmaFloatControlDecl::FloatControlKind Kind =
+    llvm::StringSwitch<PragmaFloatControlDecl::FloatControlKind>(
----------------
sepavloff wrote:
> Does such treatment allow a pragma like:
> 
>     #pragma #pragma float_control(except, on), push
> 
> The comment to `PragmaFloatControlHandler::HandlePragma` says it is valid.
Yes, #pragma float_control(except, on, push) is allowed. That's inherited from the Microsoft pragma of the same name. I need to change the .rst documentation about this. #pragma float_control(push) or #pragma float_control(pop) is also supported. Here's a link to the Microsoft doc, https://docs.microsoft.com/en-us/cpp/preprocessor/float-control?view=vs-2019 


================
Comment at: clang/lib/Sema/SemaAttr.cpp:430
+    break;
+  case PragmaFloatControlDecl::FC_Push:
+  case PragmaFloatControlDecl::FC_Pop:
----------------
sepavloff wrote:
> `push` cannot be combined with `precise`?
Yes it can be combined: #pragma float_control(precise, push); the "push" is coded into the Action, the action can be either "set" or "push". I'm using pre-existing code in clang which provides support for microsoft-style pragma push/pop/set.  For example, clang supports the Microsoft pragma code_seg which supports a stack of code segment names. 


================
Comment at: clang/lib/Sema/SemaExpr.cpp:13129
     if (FunctionDecl *F = dyn_cast<FunctionDecl>(CurContext)) {
+      // If the expression occurs inside an internal global_var_init_function
+      // then the FunctionDecl is not availble
----------------
sepavloff wrote:
> The standard says that static initializers execute in default FP mode.
> The standard says ...
Are you sure about this one?  Can you please provide the standards reference so I can study it?


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D72841/new/

https://reviews.llvm.org/D72841