[lld-macho] Support -mattr for LTO

[int3]

Summary

There doesn't seem to be an equivalent test for LLD-ELF, so
I've verified our behavior by cribbing a test from
CodeGen/X86/recip-fastmath.ll.

[compnerd]

Does this make sense to have as part of a single commit with D92912 to have a change to populate the LTO configuration options?

[MaskRay]

I think we should only support -mllvm (so all LLVM middle-end/codegen options can be used, including -mllvm -mcpu=xxx, not -mcpu= (I don't think anyone uses ld.lld -mcpu=)

[int3]

Does this make sense to have as part of a single commit with D92912 to have a change to populate the LTO configuration options?

Sure, I can merge them. FWIW, there are a dozen or so other config options that I have yet to add, but I'm doing it incrementally as I figure out how to test each one.

I don't think anyone uses ld.lld -mcpu=

I did this because -mcpu is supported by ld64.

[int3]

merging into D92912

lld/MachO/Driver.cpp

if (!markSubLibrary(searchName))
       error("-sub_library " + searchName + " does not match a supplied dylib");
   }
 
   // Parse LTO options.
   if (auto *arg = args.getLastArg(OPT_mcpu))
     parseClangOption(saver.save("-mcpu=" + StringRef(arg->getValue())),
                      arg->getSpelling());
 
+  for (auto *arg : args.filtered(OPT_mllvm))
+    parseClangOption(arg->getValue(), arg->getSpelling());
+
   initLLVM();
   compileBitcodeFiles();
   replaceCommonSymbols();
 
   StringRef orderFile = args.getLastArgValue(OPT_order_file);
   if (!orderFile.empty())
     parseOrderFile(orderFile);
 

lld/MachO/LTO.cpp

 using namespace lld::macho;
 using namespace llvm;
 using namespace llvm::sys;
 
 static lto::Config createConfig() {
   lto::Config c;
   c.Options = initTargetOptionsFromCodeGenFlags();
   c.CPU = getCPUStr();
+  c.MAttrs = getMAttrs();
   return c;
 }
 
 BitcodeCompiler::BitcodeCompiler() {
   auto backend =
       lto::createInProcessThinBackend(llvm::heavyweight_hardware_concurrency());
   ltoObj = std::make_unique<lto::LTO>(createConfig(), backend);
 }

lld/MachO/Options.td

def page_align_data_atoms : Flag<["-"], "page_align_data_atoms">,
      Flags<[HelpHidden]>,
      Group<grp_rare>;
 def not_for_dyld_shared_cache : Flag<["-"], "not_for_dyld_shared_cache">,
      HelpText<"Prevent system dylibs from being placed into the dylib shared cache">,
      Flags<[HelpHidden]>,
      Group<grp_rare>;
 def mllvm : Separate<["-"], "mllvm">,
      HelpText<"Options to pass to LLVM">,
-     Flags<[HelpHidden]>,
      Group<grp_rare>;
 def mcpu : Separate<["-"], "mcpu">,
      HelpText<"Processor family target for LTO code generation">,
      Group<grp_rare>;
 
 def grp_deprecated : OptionGroup<"deprecated">, HelpText<"DEPRECATED">;
 
 def lazy_framework : Separate<["-"], "lazy_framework">,

lld/test/MachO/mattrs.ll

+; REQUIRES: x86
+; RUN: llvm-as %s -o %t.o
+
+;; Verify that LTO behavior can be tweaked using -mattr.
+
+; RUN: %lld -mcpu haswell -mllvm -mattr=+fma %t.o -o %t.dylib -dylib
+; RUN: llvm-objdump -d --section="__text" --no-leading-addr --no-show-raw-insn %t.dylib | FileCheck %s --check-prefix=FMA
+
+; RUN: %lld -mcpu haswell -mllvm -mattr=-fma %t.o -o %t.dylib -dylib
+; RUN: llvm-objdump -d --section="__text" --no-leading-addr --no-show-raw-insn %t.dylib | FileCheck %s --check-prefix=NO-FMA
+
+; FMA:      <_foo>:
+; FMA-NEXT: vrcpss  %xmm0, %xmm0, %xmm1
+; FMA-NEXT: vfmsub213ss 7(%rip), %xmm1, %xmm0  # 2d4
+; FMA-NEXT: vfnmadd132ss  %xmm1, %xmm1, %xmm0
+; FMA-NEXT: retq
+
+; NO-FMA:      <_foo>:
+; NO-FMA-NEXT: vrcpss %xmm0, %xmm0, %xmm1
+; NO-FMA-NEXT: vmulss %xmm1, %xmm0, %xmm0
+; NO-FMA-NEXT: vmovss 16(%rip), %xmm2  # 2e0
+; NO-FMA-NEXT: vsubss %xmm0, %xmm2, %xmm0
+; NO-FMA-NEXT: vmulss %xmm0, %xmm1, %xmm0
+; NO-FMA-NEXT: vaddss %xmm0, %xmm1, %xmm0
+; NO-FMA-NEXT: retq
+
+target triple = "x86_64-apple-darwin"
+target datalayout = "e-m:o-p270:32:32-p271:32:32-p272:64:64-i64:64-f80:128-n8:16:32:64-S128"
+
+define float @foo(float %x) #0 {
+  %div = fdiv fast float 1.0, %x
+  ret float %div
+}
+
+attributes #0 = { "unsafe-fp-math"="true" "reciprocal-estimates"="divf,vec-divf" }