Don't update NoTrappingFPMath and FPDenormalMode in resetTargetOptions

[ostannard]

Summary

We'd like to remove this whole function, because these are properties of
functions, not the target as a whole. These two are easy to remove
because they are only used for emitting ARM build attributes, which
expects them to represent the defaults for the whole module, not just
the last function generated.

This is needed to get correct build attributes when using IPRA on ARM,
because IPRA causes resetTargetOptions to get called before
ARMAsmPrinter::emitAttributes.

[SjoerdMeijer]

I agree with this:

We'd like to remove this whole function, because these are properties of functions

that's why this looks good to me. Perhaps just a minor advance warning about this:

These two are easy to remove because they are only used for emitting ARM build attributes

because build attributes and downstream math library selection issues was the whole reason to add this.
If we don't set these build attributes, I am almost certain that without there will be downstream fall out because of this, so we might want to set these build attributes somewhere else (and indeed do it the right way).

llvm/trunk/lib/Target/TargetMachine.cpp

do {                                                                         \
     else                                                                       \
       Options.X = DefaultOptions.X;                                            \
   } while (0)
 
   RESET_OPTION(UnsafeFPMath, "unsafe-fp-math");
   RESET_OPTION(NoInfsFPMath, "no-infs-fp-math");
   RESET_OPTION(NoNaNsFPMath, "no-nans-fp-math");
   RESET_OPTION(NoSignedZerosFPMath, "no-signed-zeros-fp-math");
-  RESET_OPTION(NoTrappingFPMath, "no-trapping-math");
-
-  StringRef Denormal =
-    F.getFnAttribute("denormal-fp-math").getValueAsString();
-  if (Denormal == "ieee")
-    Options.FPDenormalMode = FPDenormal::IEEE;
-  else if (Denormal == "preserve-sign")
-    Options.FPDenormalMode = FPDenormal::PreserveSign;
-  else if (Denormal == "positive-zero")
-    Options.FPDenormalMode = FPDenormal::PositiveZero;
-  else
-    Options.FPDenormalMode = DefaultOptions.FPDenormalMode;
 }
 
 /// Returns the code generation relocation model. The choices are static, PIC,
 /// and dynamic-no-pic.
 Reloc::Model TargetMachine::getRelocationModel() const { return RM; }
 
 /// Returns the code model. The choices are small, kernel, medium, large, and
 /// target default.

llvm/trunk/test/CodeGen/ARM/build-attributes-fn-attr6.ll

 ; Check FP options -fno-trapping-math and -fdenormal-fp-math. They are passed
 ; as function attributes, which map on to build attributes ABI_FP_exceptions
 ; ABI_FP_denormal. In the backend we therefore have a check to see if all
 ; functions have consistent function attributes values. Here we check two
 ; functions have inconsistent values, and that a default is returned. 
 
 ; RUN: llc < %s -mtriple=armv7-linux-gnueabi -mcpu=cortex-a15  | FileCheck %s
+; RUN: llc < %s -mtriple=armv7-linux-gnueabi -mcpu=cortex-a15 -enable-ipra | FileCheck %s
 
 ; CHECK: .eabi_attribute 20, 1
 
 define i32 @foo1() local_unnamed_addr #0 {
 entry:
   ret i32 42
 }
 
 define i32 @foo2() local_unnamed_addr #1 {
 entry:
   ret i32 42
 }
 
 attributes #0 = { minsize norecurse nounwind optsize readnone "denormal-fp-math"="preserve-sign"}
 attributes #1 = { minsize norecurse nounwind optsize readnone "denormal-fp-math"="positive-zero"}