[Openmp-commits] [openmp] f81879c - [Libomptarget] Add RPC-based printf implementation for OpenMP #85638

[Joseph Huber via Openmp-commits]

Author: Joseph Huber
Date: 2024-04-10T13:36:25-05:00
New Revision: f81879c0f70ee5a1cf1d5b716dfd49d1a271cc2d

URL: https://github.com/llvm/llvm-project/commit/f81879c0f70ee5a1cf1d5b716dfd49d1a271cc2d
DIFF: https://github.com/llvm/llvm-project/commit/f81879c0f70ee5a1cf1d5b716dfd49d1a271cc2d.diff

LOG: [Libomptarget] Add RPC-based printf implementation for OpenMP #85638

Summary:
Relanding after reverting, only applies to AMDGPU for now.

This patch adds an implementation of printf that's provided by the GPU
C library runtime. This pritnf currently implemented using the same
wrapper handling that OpenMP sets up. This will be removed once we have
proper varargs support.

This printf differs from the one CUDA offers in that it is synchronous
and uses a finite size. Additionally we support pretty much every
format specifier except the %n option.

Depends on #85331

Added: 
    

Modified: 
    openmp/libomptarget/DeviceRTL/CMakeLists.txt
    openmp/libomptarget/DeviceRTL/src/LibC.cpp

Removed: 
    


################################################################################
diff  --git a/openmp/libomptarget/DeviceRTL/CMakeLists.txt b/openmp/libomptarget/DeviceRTL/CMakeLists.txt
index 2509f1276ccee6..2e7f28df24d649 100644
--- a/openmp/libomptarget/DeviceRTL/CMakeLists.txt
+++ b/openmp/libomptarget/DeviceRTL/CMakeLists.txt
@@ -122,6 +122,11 @@ set(clang_opt_flags -O3 -mllvm -openmp-opt-disable -DSHARED_SCRATCHPAD_SIZE=512
 set(link_opt_flags  -O3        -openmp-opt-disable -attributor-enable=module -vectorize-slp=false )
 set(link_export_flag -passes=internalize -internalize-public-api-file=${source_directory}/exports)
 
+# If the user built with the GPU C library enabled we will use that instead.
+if(${LIBOMPTARGET_GPU_LIBC_SUPPORT})
+  list(APPEND clang_opt_flags -DOMPTARGET_HAS_LIBC)
+endif()
+
 # Prepend -I to each list element
 set (LIBOMPTARGET_LLVM_INCLUDE_DIRS_DEVICERTL "${LIBOMPTARGET_LLVM_INCLUDE_DIRS}")
 list(TRANSFORM LIBOMPTARGET_LLVM_INCLUDE_DIRS_DEVICERTL PREPEND "-I")

diff  --git a/openmp/libomptarget/DeviceRTL/src/LibC.cpp b/openmp/libomptarget/DeviceRTL/src/LibC.cpp
index af675b97256f6c..e587c3057f5ba5 100644
--- a/openmp/libomptarget/DeviceRTL/src/LibC.cpp
+++ b/openmp/libomptarget/DeviceRTL/src/LibC.cpp
@@ -25,13 +25,26 @@ int32_t omp_vprintf(const char *Format, void *Arguments, uint32_t) {
 } // namespace impl
 #pragma omp end declare variant
 
-// We do not have a vprintf implementation for AMD GPU yet so we use a stub.
 #pragma omp begin declare variant match(device = {arch(amdgcn)})
+
+#ifdef OMPTARGET_HAS_LIBC
+// TODO: Remove this handling once we have varargs support.
+extern "C" struct FILE *stdout;
+extern "C" int32_t rpc_fprintf(FILE *, const char *, void *, uint64_t);
+
+namespace impl {
+int32_t omp_vprintf(const char *Format, void *Arguments, uint32_t Size) {
+  return rpc_fprintf(stdout, Format, Arguments, Size);
+}
+} // namespace impl
+#else
+// We do not have a vprintf implementation for AMD GPU so we use a stub.
 namespace impl {
 int32_t omp_vprintf(const char *Format, void *Arguments, uint32_t) {
   return -1;
 }
 } // namespace impl
+#endif
 #pragma omp end declare variant
 
 extern "C" {