[Libomptarget] Emit a special warning when no images are found

[jhuber6]

Summary

When offloading is mandatory we can emit a more helpful message if we
did not find any compatible images with the user's system.

Fixes #60221

[jhuber6]

Cleanup unrelated lines.

[ye-luo]

How does the above message behaves when there is no GPU on the node.

[jhuber6]

In D142369#4073769, @ye-luo wrote:

How does the above message behaves when there is no GPU on the node.

This will fire anytime we fail with a mandatory offloading but there were no images registered to a plugin. Maybe this could happen if the plugin failed to load the x86 plugin. That plugin uses the architecture just like any other, but it's usually empty. I should maybe add a special string if it's empty.

[jhuber6]

Add special handling for if the Info is null or contains and empty string.

[ye-luo]

When libomptarget failed to load the amd plugin due to missing hsa runtime so files.
I got

yeluo@epyc-server:~$ clang++ -fopenmp --offload-arch=gfx90a main.cpp
yeluo@epyc-server:~$ OMP_TARGET_OFFLOAD=mandatory ./a.out 
Libomptarget error: Consult https://openmp.llvm.org/design/Runtimes.html for debugging options.
Libomptarget error: None of the images found are compatible: gfx90a
Libomptarget error: Source location information not present. Compile with -g or -gline-tables-only.
Libomptarget fatal error 1: failure of target construct while offloading is mandatory

None of the images found are compatible: gfx90a this line is a bit hard to digest. Could you print a list of loaded plugins and a list of images.

[jhuber6]

In D142369#4074092, @ye-luo wrote:

When libomptarget failed to load the amd plugin due to missing hsa runtime so files.
I got

yeluo@epyc-server:~$ clang++ -fopenmp --offload-arch=gfx90a main.cpp
yeluo@epyc-server:~$ OMP_TARGET_OFFLOAD=mandatory ./a.out 
Libomptarget error: Consult https://openmp.llvm.org/design/Runtimes.html for debugging options.
Libomptarget error: None of the images found are compatible: gfx90a
Libomptarget error: Source location information not present. Compile with -g or -gline-tables-only.
Libomptarget fatal error 1: failure of target construct while offloading is mandatory

None of the images found are compatible: gfx90a this line is a bit hard to digest. Could you print a list of loaded plugins and a list of images.

The following is the list of images found. I could try to make that more explicit. And by default we should have every plugin loaded.

[ye-luo]

In D142369#4074095, @jhuber6 wrote:

In D142369#4074092, @ye-luo wrote:

When libomptarget failed to load the amd plugin due to missing hsa runtime so files.
I got

yeluo@epyc-server:~$ clang++ -fopenmp --offload-arch=gfx90a main.cpp
yeluo@epyc-server:~$ OMP_TARGET_OFFLOAD=mandatory ./a.out 
Libomptarget error: Consult https://openmp.llvm.org/design/Runtimes.html for debugging options.
Libomptarget error: None of the images found are compatible: gfx90a
Libomptarget error: Source location information not present. Compile with -g or -gline-tables-only.
Libomptarget fatal error 1: failure of target construct while offloading is mandatory

None of the images found are compatible: gfx90a this line is a bit hard to digest. Could you print a list of loaded plugins and a list of images.

The following is the list of images found. I could try to make that more explicit. And by default we should have every plugin loaded.

By list all the successfully loaded plugins and images. I can figure out if the issue was caused by missing plugin or missing image.

[jhuber6]

In D142369#4074913, @ye-luo wrote:

By list all the successfully loaded plugins and images. I can figure out if the issue was caused by missing plugin or missing image.

It's tough to tell which image was intended for which plugin. The best we could do is print out every plugin, I guess this can happen if the library paths aren't set up correctly. Printing this would require making the RTLName present in every build. Maybe I can just print a special message if we didn't find any plugins. Since I think that's more likely to be the cause of failure vs. a single one.

[jhuber6]

Adding message if we didn't find any plugins. Usually happens if the paths are wrong.

[ye-luo]

It doesn't print well

$ OMP_TARGET_OFFLOAD=mandatory ./a.out 
Libomptarget error: Consult https://openmp.llvm.org/design/Runtimes.html for debugging options.
Libomptarget error: None of the images ( gfx90a 
Libomptarget error: Source location information not present. Compile with -g or -gline-tables-only.
Libomptarget fatal error 1: failure of target construct while offloading is mandatory

[jdoerfert]

openmp/libomptarget/src/omptarget.cpp

284

None of the images, what?

[jhuber6]

Updating, error message now looks like this.

Libomptarget error: Consult https://openmp.llvm.org/design/Runtimes.html for debugging options.
Libomptarget error: No images with a supported architecture found (sm_70)
Libomptarget error: Source location information not present. Compile with -g or -gline-tables-only.
Libomptarget fatal error 1: failure of target construct while offloading is mandatory
[1]    159574 IOT instruction (core dumped)  OMP_TARGET_OFFLOAD=mandatory ./a.out

[ye-luo]

Still prefer a more clear message.

openmp/libomptarget/src/omptarget.cpp

284

FAILURE_MESSAGE("No images found compatible with the hardware. Found ");
fprintf(stderr, "(%s) images\n", llvm::join(Archs, ",").c_str());

[jhuber6]

Changing message

openmp/libomptarget/src/omptarget.cpp

case tgt_mandatory:
     if (!Success) {
       if (getInfoLevel() & OMP_INFOTYPE_DUMP_TABLE)
         for (auto &Device : PM->Devices)
           dumpTargetPointerMappings(Loc, *Device);
       else
         FAILURE_MESSAGE("Consult https://openmp.llvm.org/design/Runtimes.html "
                         "for debugging options.\n");
 
+      if (PM->RTLs.UsedRTLs.empty()) {
+        llvm::SmallVector<llvm::StringRef> Archs;
+        llvm::transform(PM->Images, std::back_inserter(Archs),
+                        [](const auto &x) {
+                          return !x.second.Arch ? "empty" : x.second.Arch;
+                        });
+        FAILURE_MESSAGE(
+            "No images found compatible with the installed hardware. ");
+        fprintf(stderr, "Found (%s)\n", llvm::join(Archs, ",").c_str());
+      }
+
       SourceInfo Info(Loc);
       if (Info.isAvailible())
         fprintf(stderr, "%s:%d:%d: ", Info.getFilename(), Info.getLine(),
                 Info.getColumn());
       else
         FAILURE_MESSAGE("Source location information not present. Compile with "
                         "-g or -gline-tables-only.\n");
       FATAL_MESSAGE0(