[PATCH] Change the fast-isel-abort option from bool to int to enable "levels"

Eric Christopher echristo at gmail.com
Fri Feb 27 10:16:38 PST 2015


One reply inline and one small cosmetic comment. Otherwise LGTM.

-eric


================
Comment at: lib/CodeGen/SelectionDAG/SelectionDAGISel.cpp:418
@@ -418,3 +417,3 @@
   assert((!EnableFastISelAbort || TM.Options.EnableFastISel) &&
-         "-fast-isel-abort requires -fast-isel");
+         "-fast-isel-abort>0 requires -fast-isel");
 
----------------
Probably want some nice whitespace, but that's cosmetic.

================
Comment at: lib/CodeGen/SelectionDAG/SelectionDAGISel.cpp:171
@@ -170,9 +170,3 @@
                    "instruction selector"));
-static cl::opt<bool>
-EnableFastISelAbort("fast-isel-abort", cl::Hidden,
-          cl::desc("Enable abort calls when \"fast\" instruction selection "
-                   "fails to lower an instruction"));
-static cl::opt<bool>
-EnableFastISelAbortArgs("fast-isel-abort-args", cl::Hidden,
-          cl::desc("Enable abort calls when \"fast\" instruction selection "
-                   "fails to lower a formal argument"));
+static cl::opt<int> EnableFastISelAbort(
+    "fast-isel-abort", cl::Hidden,
----------------
joker.eph wrote:
> echristo wrote:
> > Can we make this default to the current default behavior? It'll mean that you don't need to update testcases too :)
> The existing default value is "off". I believe it is gonna be the same here.
> I had to update the tests because previously you would be able to just add -fast-isel-abort while now you need -fast-isel-abort=1
> I don't know if there is a way to say that -fast-isel-abort with no argument will be like =1. 
> Even if it is possible, I'm not even sure it is desirable as reading a command line with -fast-isel-abort alone makes me expect to always abort.
Right. Was just trying to keep the command line from changing, but you've convinced me. :)

http://reviews.llvm.org/D7941

EMAIL PREFERENCES
  http://reviews.llvm.org/settings/panel/emailpreferences/






More information about the llvm-commits mailing list