[PATCH] D124669: [flang][driver] Add support for -save-temps

Daniil Dudkin via Phabricator via cfe-commits cfe-commits at lists.llvm.org
Mon May 2 09:41:20 PDT 2022


unterumarmung added inline comments.


================
Comment at: clang/include/clang/Driver/Options.td:4131
 def : Flag<["-"], "no-integrated-as">, Alias<fno_integrated_as>,
-      Flags<[CC1Option, NoXarchOption]>;
+      Flags<[CC1Option,FlangOption,NoXarchOption]>;
 
----------------
rovka wrote:
> awarzynski wrote:
> > unterumarmung wrote:
> > > Why not to add `FC1Option` here and elsewhere like it's done for `CC1Option`?
> > I'm not 100% sure what `-fno-integrated-as` controls in Clang's frontend driver, `clang -cc1`. I'm guessing that it might related to using/not-using LLVM's MCAsmParser. Perhaps for inline assembly?
> > 
> > In Flang, I'm only focusing on `-save-temps` for which I need to make sure that we don't try to call `flang-new -fc1as` or something similar (it does not exist). Instead, `-save-temps` will have to rely on an external assembler.
> > 
> > So, we basically don't require -`fno-integrated-as` in `flang-new -fc1` just yet (that's what `FC1Option` is for - marking an option as available in the frontend driver).
> I'm not 100% sure either (haven't looked at the code), but my understanding of `-fno-integrated-as` is that it forces clang to call the system assembler as opposed to using the LLVM libraries to generate machine code directly. Since flang never uses the system assembler, I would say the integrated assembler is always on in flang. So I'm not convinced it makes sense to add this flag to the driver, unless we're actually adding a 
> `-fc1as`.
@awarzynski @rovka, thank you for the detailed explanation!


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D124669/new/

https://reviews.llvm.org/D124669



More information about the cfe-commits mailing list