Add -static-openmp driver option

[pirama]

Summary

For Gnu, FreeBSD and NetBSD, this option forces linking with the static
OpenMP host runtime (similar to -static-libgcc and -static-libstdcxx).

Android's NDK will start the shared OpenMP runtime in addition to the static
libomp. In this scenario, the linker will prefer to use the shared library by
default. Add this option to enable linking with the static libomp.

Fixes https://github.com/android-ndk/ndk/issues/1028

[pirama]

Supported this flag for NetBSD and FreeBSD as well.

[danalbert]

Otherwise LGTM

clang/lib/Driver/ToolChains/CommonArgs.cpp

503

Maybe ForceStaticHostRuntime? For configurations where there is only a static runtime available, StaticHostRuntime = false won't actually link the dynamic runtime.

[pirama]

Change parameter name.

[srhines]

Looks really nice. I am sure the NDK developers will be happy to see support for static OpenMP. Do you want to add the public NDK github issue link in the commit message?

[MaskRay]

Instead of more -static-foobar, maybe we need -static=foobar,stdlib,rtlib instead?

Edit: Added -static= in D53238. If that is accepted, you may consider -static=openmp

[pirama]

Mention NDK issue https://github.com/android-ndk/ndk/issues/1028.

[pirama]

In D67200#1660147, @srhines wrote:

Looks really nice. I am sure the NDK developers will be happy to see support for static OpenMP. Do you want to add the public NDK github issue link in the commit message?

Done.

In D67200#1660192, @MaskRay wrote:

Edit: Added -static= in D53238. If that is accepted, you may consider -static=openmp

Thanks Fangrui! That change would be very helpful and I'll move to -static=openmp once it lands. As for this change, I'll wait over the weekend and submit it on Monday to allow everyone a chance to look at it.

[pirama]

I'll update this review addressing @joerg's reply to cfe-commits:

Needs testing for the -static interaction?

Thanks @srhines for pointing me to it - I'd only subscribed to cfe-dev and not cfe-commits so I'd missed it..

[pirama]

Test -static, -static-openmp interaction. Added these only for iomp5 to avoid test-case explosion.

clang/include/clang/Driver/Options.td

 def fopenmp_cuda_blocks_per_sm_EQ : Joined<["-"], "fopenmp-cuda-blocks-per-sm=">, Group<f_Group>,
   Flags<[CC1Option, NoArgumentUnused, HelpHidden]>;
 def fopenmp_cuda_teams_reduction_recs_num_EQ : Joined<["-"], "fopenmp-cuda-teams-reduction-recs-num=">, Group<f_Group>,
   Flags<[CC1Option, NoArgumentUnused, HelpHidden]>;
 def fopenmp_optimistic_collapse : Flag<["-"], "fopenmp-optimistic-collapse">, Group<f_Group>,
   Flags<[CC1Option, NoArgumentUnused, HelpHidden]>;
 def fno_openmp_optimistic_collapse : Flag<["-"], "fno-openmp-optimistic-collapse">, Group<f_Group>,
   Flags<[NoArgumentUnused, HelpHidden]>;
+def static_openmp: Flag<["-"], "static-openmp">,
+  HelpText<"Use the static host OpenMP runtime while linking.">;
 def fno_optimize_sibling_calls : Flag<["-"], "fno-optimize-sibling-calls">, Group<f_Group>;
 def foptimize_sibling_calls : Flag<["-"], "foptimize-sibling-calls">, Group<f_Group>;
 def fno_escaping_block_tail_calls : Flag<["-"], "fno-escaping-block-tail-calls">, Group<f_Group>, Flags<[CC1Option]>;
 def fescaping_block_tail_calls : Flag<["-"], "fescaping-block-tail-calls">, Group<f_Group>;
 def force__cpusubtype__ALL : Flag<["-"], "force_cpusubtype_ALL">;
 def force__flat__namespace : Flag<["-"], "force_flat_namespace">;
 def force__load : Separate<["-"], "force_load">;
 def force_addr : Joined<["-"], "fforce-addr">, Group<clang_ignored_f_Group>;

clang/lib/Driver/ToolChains/CommonArgs.h

void AddAssemblerKPIC(const ToolChain &ToolChain,
                       const llvm::opt::ArgList &Args,
                       llvm::opt::ArgStringList &CmdArgs);
 
 void addArchSpecificRPath(const ToolChain &TC, const llvm::opt::ArgList &Args,
                           llvm::opt::ArgStringList &CmdArgs);
 /// Returns true, if an OpenMP runtime has been added.
 bool addOpenMPRuntime(llvm::opt::ArgStringList &CmdArgs, const ToolChain &TC,
                       const llvm::opt::ArgList &Args,
+                      bool StaticHostRuntime = false,
                       bool IsOffloadingHost = false, bool GompNeedsRT = false);
 
 llvm::opt::Arg *getLastProfileUseArg(const llvm::opt::ArgList &Args);
 llvm::opt::Arg *getLastProfileSampleUseArg(const llvm::opt::ArgList &Args);
 
 bool isObjCAutoRefCount(const llvm::opt::ArgList &Args);
 
 unsigned getLTOParallelism(const llvm::opt::ArgList &Args, const Driver &D);

clang/lib/Driver/ToolChains/CommonArgs.cpp

void tools::addArchSpecificRPath(const ToolChain &TC, const ArgList &Args,
   std::string CandidateRPath = TC.getArchSpecificLibPath();
   if (TC.getVFS().exists(CandidateRPath)) {
     CmdArgs.push_back("-rpath");
     CmdArgs.push_back(Args.MakeArgString(CandidateRPath.c_str()));
   }
 }
 
 bool tools::addOpenMPRuntime(ArgStringList &CmdArgs, const ToolChain &TC,
-                             const ArgList &Args, bool IsOffloadingHost,
-                             bool GompNeedsRT) {
+                             const ArgList &Args, bool StaticHostRuntime,
+                             bool IsOffloadingHost, bool GompNeedsRT) {
   if (!Args.hasFlag(options::OPT_fopenmp, options::OPT_fopenmp_EQ,
                     options::OPT_fno_openmp, false))
     return false;
 
-  switch (TC.getDriver().getOpenMPRuntime(Args)) {
+  Driver::OpenMPRuntimeKind RTKind = TC.getDriver().getOpenMPRuntime(Args);
+
+  if (RTKind == Driver::OMPRT_Unknown)
+    // Already diagnosed.
+    return false;
+
+  if (StaticHostRuntime)
+    CmdArgs.push_back("-Bstatic");
+
+  switch (RTKind) {
   case Driver::OMPRT_OMP:
     CmdArgs.push_back("-lomp");
     break;
   case Driver::OMPRT_GOMP:
     CmdArgs.push_back("-lgomp");
-
-    if (GompNeedsRT)
-      CmdArgs.push_back("-lrt");
     break;
   case Driver::OMPRT_IOMP5:
     CmdArgs.push_back("-liomp5");
     break;
-  case Driver::OMPRT_Unknown:
-    // Already diagnosed.
-    return false;
   }
 
+  if (StaticHostRuntime)
+    CmdArgs.push_back("-Bdynamic");
+
+  if (RTKind == Driver::OMPRT_GOMP && GompNeedsRT)
+      CmdArgs.push_back("-lrt");
+
   if (IsOffloadingHost)
     CmdArgs.push_back("-lomptarget");
 
   addArchSpecificRPath(TC, Args, CmdArgs);
 
   return true;
 }
 

clang/lib/Driver/ToolChains/FreeBSD.cpp

AddGoldPlugin(ToolChain, Args, CmdArgs, Output, Inputs[0],
                   D.getLTOMode() == LTOK_Thin);
   }
 
   bool NeedsSanitizerDeps = addSanitizerRuntimes(ToolChain, Args, CmdArgs);
   bool NeedsXRayDeps = addXRayRuntime(ToolChain, Args, CmdArgs);
   AddLinkerInputs(ToolChain, Inputs, Args, CmdArgs, JA);
 
   if (!Args.hasArg(options::OPT_nostdlib, options::OPT_nodefaultlibs)) {
-    addOpenMPRuntime(CmdArgs, ToolChain, Args);
+    // Use the static OpenMP runtime with -static-openmp
+    bool StaticOpenMP = Args.hasArg(options::OPT_static_openmp) &&
+                        !Args.hasArg(options::OPT_static);
+    addOpenMPRuntime(CmdArgs, ToolChain, Args, StaticOpenMP);
+
     if (D.CCCIsCXX()) {
       if (ToolChain.ShouldLinkCXXStdlib(Args))
         ToolChain.AddCXXStdlibLibArgs(Args, CmdArgs);
       if (Args.hasArg(options::OPT_pg))
         CmdArgs.push_back("-lm_p");
       else
         CmdArgs.push_back("-lm");
     }

clang/lib/Driver/ToolChains/Gnu.cpp

if (!Args.hasArg(options::OPT_nodefaultlibs)) {
         linkSanitizerRuntimeDeps(ToolChain, CmdArgs);
 
       if (NeedsXRayDeps)
         linkXRayRuntimeDeps(ToolChain, CmdArgs);
 
       bool WantPthread = Args.hasArg(options::OPT_pthread) ||
                          Args.hasArg(options::OPT_pthreads);
 
+      // Use the static OpenMP runtime with -static-openmp
+      bool StaticOpenMP = Args.hasArg(options::OPT_static_openmp) &&
+                          !Args.hasArg(options::OPT_static);
+
       // FIXME: Only pass GompNeedsRT = true for platforms with libgomp that
       // require librt. Most modern Linux platforms do, but some may not.
-      if (addOpenMPRuntime(CmdArgs, ToolChain, Args,
+      if (addOpenMPRuntime(CmdArgs, ToolChain, Args, StaticOpenMP,
                            JA.isHostOffloading(Action::OFK_OpenMP),
                            /* GompNeedsRT= */ true))
         // OpenMP runtimes implies pthreads when using the GNU toolchain.
         // FIXME: Does this really make sense for all GNU toolchains?
         WantPthread = true;
 
       AddRunTimeLibs(ToolChain, D, CmdArgs, Args);
 

clang/lib/Driver/ToolChains/NetBSD.cpp

case llvm::Triple::x86_64:
       useLibgcc = false;
       break;
     default:
       break;
     }
   }
 
   if (!Args.hasArg(options::OPT_nostdlib, options::OPT_nodefaultlibs)) {
-    addOpenMPRuntime(CmdArgs, getToolChain(), Args);
+    // Use the static OpenMP runtime with -static-openmp
+    bool StaticOpenMP = Args.hasArg(options::OPT_static_openmp) &&
+                        !Args.hasArg(options::OPT_static);
+    addOpenMPRuntime(CmdArgs, getToolChain(), Args, StaticOpenMP);
+
     if (D.CCCIsCXX()) {
       if (ToolChain.ShouldLinkCXXStdlib(Args))
         ToolChain.AddCXXStdlibLibArgs(Args, CmdArgs);
       CmdArgs.push_back("-lm");
     }
     if (NeedsSanitizerDeps)
       linkSanitizerRuntimeDeps(getToolChain(), CmdArgs);
     if (NeedsXRayDeps)

clang/test/Driver/fopenmp.c

 //
 // CHECK-CC1-NO-OPENMP: "-cc1"
 // CHECK-CC1-NO-OPENMP-NOT: "-fopenmp"
 //
 // RUN: %clang -target x86_64-linux-gnu -fopenmp=libomp %s -o %t -### 2>&1 | FileCheck %s --check-prefix=CHECK-LD-OMP
 // RUN: %clang -target x86_64-linux-gnu -fopenmp=libgomp %s -o %t -### 2>&1 | FileCheck %s --check-prefix=CHECK-LD-GOMP --check-prefix=CHECK-LD-GOMP-RT
 // RUN: %clang -target x86_64-linux-gnu -fopenmp=libiomp5 %s -o %t -### 2>&1 | FileCheck %s --check-prefix=CHECK-LD-IOMP5
 //
+// RUN: %clang -target x86_64-linux-gnu -fopenmp=libomp -static-openmp %s -o %t -### 2>&1 | FileCheck %s --check-prefix=CHECK-LD-STATIC-OMP
+// RUN: %clang -target x86_64-linux-gnu -fopenmp=libgomp -static-openmp %s -o %t -### 2>&1 | FileCheck %s --check-prefix=CHECK-LD-STATIC-GOMP --check-prefix=CHECK-LD-STATIC-GOMP-RT
+// RUN: %clang -target x86_64-linux-gnu -fopenmp=libiomp5 -static-openmp %s -o %t -### 2>&1 | FileCheck %s --check-prefix=CHECK-LD-STATIC-IOMP5
+//
 // RUN: %clang -nostdlib -target x86_64-linux-gnu -fopenmp=libomp %s -o %t -### 2>&1 | FileCheck %s --check-prefix=CHECK-NO-OMP
 // RUN: %clang -nostdlib -target x86_64-linux-gnu -fopenmp=libgomp %s -o %t -### 2>&1 | FileCheck %s --check-prefix=CHECK-NO-GOMP
 // RUN: %clang -nostdlib -target x86_64-linux-gnu -fopenmp=libiomp5 %s -o %t -### 2>&1 | FileCheck %s --check-prefix=CHECK-NO-IOMP5
 //
 // RUN: %clang -target x86_64-darwin -fopenmp=libomp %s -o %t -### 2>&1 | FileCheck %s --check-prefix=CHECK-LD-OMP
 // RUN: %clang -target x86_64-darwin -fopenmp=libgomp %s -o %t -### 2>&1 | FileCheck %s --check-prefix=CHECK-LD-GOMP --check-prefix=CHECK-LD-GOMP-NO-RT
 // RUN: %clang -target x86_64-darwin -fopenmp=libiomp5 %s -o %t -### 2>&1 | FileCheck %s --check-prefix=CHECK-LD-IOMP5
 //
 // RUN: %clang -nostdlib -target x86_64-darwin -fopenmp=libomp %s -o %t -### 2>&1 | FileCheck %s --check-prefix=CHECK-NO-OMP
 // RUN: %clang -nostdlib -target x86_64-darwin -fopenmp=libgomp %s -o %t -### 2>&1 | FileCheck %s --check-prefix=CHECK-NO-GOMP
 // RUN: %clang -nostdlib -target x86_64-darwin -fopenmp=libiomp5 %s -o %t -### 2>&1 | FileCheck %s --check-prefix=CHECK-NO-IOMP5
 //
 // RUN: %clang -target x86_64-freebsd -fopenmp=libomp %s -o %t -### 2>&1 | FileCheck %s --check-prefix=CHECK-LD-OMP
 // RUN: %clang -target x86_64-freebsd -fopenmp=libgomp %s -o %t -### 2>&1 | FileCheck %s --check-prefix=CHECK-LD-GOMP --check-prefix=CHECK-LD-GOMP-NO-RT
 // RUN: %clang -target x86_64-freebsd -fopenmp=libiomp5 %s -o %t -### 2>&1 | FileCheck %s --check-prefix=CHECK-LD-IOMP5
 //
+// RUN: %clang -target x86_64-freebsd -fopenmp=libomp -static-openmp %s -o %t -### 2>&1 | FileCheck %s --check-prefix=CHECK-LD-STATIC-OMP
+// RUN: %clang -target x86_64-freebsd -fopenmp=libgomp -static-openmp %s -o %t -### 2>&1 | FileCheck %s --check-prefix=CHECK-LD-STATIC-GOMP --check-prefix=CHECK-LD-STATIC-GOMP-NO-RT
+// RUN: %clang -target x86_64-freebsd -fopenmp=libiomp5 -static-openmp %s -o %t -### 2>&1 | FileCheck %s --check-prefix=CHECK-LD-STATIC-IOMP5
+//
 // RUN: %clang -nostdlib -target x86_64-freebsd -fopenmp=libomp %s -o %t -### 2>&1 | FileCheck %s --check-prefix=CHECK-NO-OMP
 // RUN: %clang -nostdlib -target x86_64-freebsd -fopenmp=libgomp %s -o %t -### 2>&1 | FileCheck %s --check-prefix=CHECK-NO-GOMP
 // RUN: %clang -nostdlib -target x86_64-freebsd -fopenmp=libiomp5 %s -o %t -### 2>&1 | FileCheck %s --check-prefix=CHECK-NO-IOMP5
 //
 // RUN: %clang -target x86_64-netbsd -fopenmp=libomp %s -o %t -### 2>&1 | FileCheck %s --check-prefix=CHECK-LD-OMP
 // RUN: %clang -target x86_64-netbsd -fopenmp=libgomp %s -o %t -### 2>&1 | FileCheck %s --check-prefix=CHECK-LD-GOMP --check-prefix=CHECK-LD-GOMP-NO-RT
 // RUN: %clang -target x86_64-netbsd -fopenmp=libiomp5 %s -o %t -### 2>&1 | FileCheck %s --check-prefix=CHECK-LD-IOMP5
 //
+// RUN: %clang -target x86_64-netbsd -fopenmp=libomp -static-openmp %s -o %t -### 2>&1 | FileCheck %s --check-prefix=CHECK-LD-STATIC-OMP
+// RUN: %clang -target x86_64-netbsd -fopenmp=libgomp -static-openmp %s -o %t -### 2>&1 | FileCheck %s --check-prefix=CHECK-LD-STATIC-GOMP --check-prefix=CHECK-LD-STATIC-GOMP-NO-RT
+// RUN: %clang -target x86_64-netbsd -fopenmp=libiomp5 -static-openmp %s -o %t -### 2>&1 | FileCheck %s --check-prefix=CHECK-LD-STATIC-IOMP5
+//
 // RUN: %clang -nostdlib -target x86_64-netbsd -fopenmp=libomp %s -o %t -### 2>&1 | FileCheck %s --check-prefix=CHECK-NO-OMP
 // RUN: %clang -nostdlib -target x86_64-netbsd -fopenmp=libgomp %s -o %t -### 2>&1 | FileCheck %s --check-prefix=CHECK-NO-GOMP
 // RUN: %clang -nostdlib -target x86_64-netbsd -fopenmp=libiomp5 %s -o %t -### 2>&1 | FileCheck %s --check-prefix=CHECK-NO-IOMP5
 //
 // RUN: %clang -target x86_64-windows-gnu -fopenmp=libomp %s -o %t -### 2>&1 | FileCheck %s --check-prefix=CHECK-LD-OMP
 // RUN: %clang -target x86_64-windows-gnu -fopenmp=libgomp %s -o %t -### 2>&1 | FileCheck %s --check-prefix=CHECK-LD-GOMP --check-prefix=CHECK-LD-GOMP-NO-RT
 // RUN: %clang -target x86_64-windows-gnu -fopenmp=libiomp5 %s -o %t -### 2>&1 | FileCheck %s --check-prefix=CHECK-LD-IOMP5MD
 //

 // CHECK-NO-GOMP-NOT: "-lgomp"
 //
 // CHECK-NO-IOMP5: "{{.*}}ld{{(.exe)?}}"
 // CHECK-NO-IOMP5-NOT: "-liomp5"
 //
 // CHECK-NO-IOMP5MD: "{{.*}}ld{{(.exe)?}}"
 // CHECK-NO-IOMP5MD-NOT: "-liomp5md"
 //
+// CHECK-LD-STATIC-OMP: "{{.*}}ld{{(.exe)?}}"
+// CHECK-LD-STATIC-OMP: "-Bstatic" "-lomp" "-Bdynamic"
+//
+// CHECK-LD-STATIC-GOMP: "{{.*}}ld{{(.exe)?}}"
+// CHECK-LD-STATIC-GOMP: "-Bstatic" "-lgomp" "-Bdynamic"
+// CHECK-LD-STATIC-GOMP-RT: "-lrt"
+// CHECK-LD-STATIC-NO-GOMP-RT-NOT: "-lrt"
+//
+// CHECK-LD-STATIC-IOMP5: "{{.*}}ld{{(.exe)?}}"
+// CHECK-LD-STATIC-IOMP5: "-Bstatic" "-liomp5" "-Bdynamic"
+//
 // We'd like to check that the default is sane, but until we have the ability
 // to *always* semantically analyze OpenMP without always generating runtime
 // calls (in the event of an unsupported runtime), we don't have a good way to
 // test the CC1 invocation. Instead, just ensure we do eventually link *some*
 // OpenMP runtime.
 //
 // RUN: %clang -target x86_64-linux-gnu -fopenmp %s -o %t -### 2>&1 | FileCheck %s --check-prefix=CHECK-LD-ANY
 // RUN: %clang -target x86_64-darwin -fopenmp %s -o %t -### 2>&1 | FileCheck %s --check-prefix=CHECK-LD-ANY