NVPTX: Add supported CL features

[jvesely]

Summary

Fixes libclc compilation.
Taken from clinfo on one nVidia system.

[jvesely]

add test

[yaxunl]

I have a similar, but yours is more complete. I probably will drop my patch once yours goes through.

test/Misc/nvptx.languageOptsOpenCL.cl

2

-Werror is unnecessary, same as below

6

should be like

#pragma OPENCL EXTENSION cl_xxx : enable

same for below

[jholewinski]

lib/Basic/Targets.cpp

1849

Do you really intend to enable subgroup support?

[jvesely]

remove subgroups support.
better tests.

note: I had to add OPENCL_VERSION defines manually, these can be removed once clang correctly reports the version

[yaxunl]

note: I had to add OPENCL_VERSION defines manually, these can be removed once clang correctly reports the version

clang correctly generates OPENCL_C_VERSION and OPENCL_VERSION_x_y macros. You can use them instead. For CL 1.1/1.2/2.0, you need to use -cl-std=CLx.y to specify OpenCL version.

[jvesely]

even better tests
use existing OPENCL_C_VERSION
use -cl-std=CL for explicit OpenCL1.0

[jvesely]

Report only clinfo exported extensions.

[Anastasia]

test/Misc/nvptx.languageOptsOpenCL.cl

2

Should we check for errors on unsupported extensions for this target too?

[jvesely]

test/Misc/nvptx.languageOptsOpenCL.cl

2

I didn't want to do that. Including unsupported extensions in these tests would imply that every OCL extension needs test in 4 places (nvptx, amdgcn, r600, extension-version.cl), possibly more if more targets are added, which seemed excessive.

[jvesely]

Test all known extensions against expected nvptx outcome (add negative tests)

[jvesely]

add back require registered target

[jvesely]

ping

@jholewinski, is the exported list of features OK with you?

@Anastasia, are the tests OK now?

[jholewinski]

Looks good to me

lib/Basic/Targets.cpp

GPU = llvm::StringSwitch<GPUKind>(Name)
               .Case("sm_21", GK_SM21)
               .Case("sm_30", GK_SM30)
               .Case("sm_35", GK_SM35)
               .Case("sm_37", GK_SM37)
               .Default(GK_NONE);
 
     return GPU != GK_NONE;
   }
+  void setSupportedOpenCLOpts() override {
+    auto &Opts = getSupportedOpenCLOpts();
+    Opts.cl_clang_storage_class_specifiers = 1;
+    Opts.cl_khr_gl_sharing = 1;
+    Opts.cl_khr_gl_event = 1;
+    Opts.cl_khr_d3d10_sharing = 1;
+    Opts.cl_khr_subgroups = 1;
+
+    Opts.cl_khr_fp64 = 1;
+    Opts.cl_khr_byte_addressable_store = 1;
+    Opts.cl_khr_global_int32_base_atomics = 1;
+    Opts.cl_khr_global_int32_extended_atomics = 1;
+    Opts.cl_khr_local_int32_base_atomics = 1;
+    Opts.cl_khr_local_int32_extended_atomics = 1;
+  }
 };
 
 const Builtin::Info NVPTXTargetInfo::BuiltinInfo[] = {
 #define BUILTIN(ID, TYPE, ATTRS)                                               \
   { #ID, TYPE, ATTRS, nullptr, ALL_LANGUAGES, nullptr },
 #define LIBBUILTIN(ID, TYPE, ATTRS, HEADER)                                    \
   { #ID, TYPE, ATTRS, HEADER, ALL_LANGUAGES, nullptr },
 #include "clang/Basic/BuiltinsNVPTX.def"

test/Misc/nvptx.languageOptsOpenCL.cl

+// REQUIRES: nvptx-registered-target
+// RUN: %clang_cc1 -x cl %s -verify -triple nvptx-unknown-unknown -Werror
+// RUN: %clang_cc1 -x cl %s -verify -triple nvptx64-unknown-unknown -Werror
+// expected-no-diagnostics
+
+#pragma EXTENSION cl_clang_storage_class_specifiers: enabled
+#pragma EXTENSION cl_khr_gl_event: enabled
+#pragma EXTENSION cl_khr_gl_sharing: enabled
+#pragma EXTENSION cl_khr_subgroups: enabled
+
+#pragma EXTENSION cl_khr_fp64: enabled
+#pragma EXTENSION cl_khr_byte_addressable_storage: enabled
+#pragma EXTENSION cl_khr_global_int32_base_atomics: enabled
+#pragma EXTENSION cl_khr_global_int32_extended_atomics: enabled
+#pragma EXTENSION cl_khr_local_int32_base_atomics: enabled
+#pragma EXTENSION cl_khr_local_int32_extended_atomics: enabled