[Openmp-commits] [openmp] [Libomptarget] Add RPC-based `printf` implementation for OpenMP (PR #85638)

[Joseph Huber via Openmp-commits]

https://github.com/jhuber6 created https://github.com/llvm/llvm-project/pull/85638

Summary:
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 https://github.com/llvm/llvm-project/pull/85331


>From d28a367495bd4a49e6c76c42fd30cba368ef0862 Mon Sep 17 00:00:00 2001
From: Joseph Huber <huberjn at outlook.com>
Date: Mon, 18 Mar 2024 07:57:47 -0500
Subject: [PATCH] [Libomptarget] Add RPC-based `printf` implementation for
 OpenMP

Summary:
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 https://github.com/llvm/llvm-project/pull/85331
---
 openmp/libomptarget/DeviceRTL/CMakeLists.txt |  5 +++++
 openmp/libomptarget/DeviceRTL/src/LibC.cpp   | 13 +++++++++++++
 2 files changed, 18 insertions(+)

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..33fec8135ef039 100644
--- a/openmp/libomptarget/DeviceRTL/src/LibC.cpp
+++ b/openmp/libomptarget/DeviceRTL/src/LibC.cpp
@@ -53,10 +53,23 @@ void memset(void *dst, int C, size_t count) {
     dstc[I] = C;
 }
 
+// If the user built with the GPU C library enabled we will assume that we can
+// call it.
+#ifdef OMPTARGET_HAS_LIBC
+
+// TODO: Remove this handling once we have varargs support.
+extern struct FILE *stdout;
+int32_t rpc_fprintf(FILE *, const char *, void *, uint64_t);
+
+int32_t __llvm_omp_vprintf(const char *Format, void *Arguments, uint32_t Size) {
+  return rpc_fprintf(stdout, Format, Arguments, Size);
+}
+#else
 /// printf() calls are rewritten by CGGPUBuiltin to __llvm_omp_vprintf
 int32_t __llvm_omp_vprintf(const char *Format, void *Arguments, uint32_t Size) {
   return impl::omp_vprintf(Format, Arguments, Size);
 }
+#endif
 }
 
 #pragma omp end declare target