Fix client build for TRITON_ENABLE_GPU=OFF (triton-inference-server#1704

)

* Fix client build for TRITON_ENABLE_GPU=OFF

* Update docs

Dockerfile.centos_client

@@ -62,18 +62,18 @@ COPY build build
 COPY src/clients src/clients
 COPY src/core src/core
 
-RUN cd build && \
+RUN mkdir builddir && cd builddir && \
     cmake -DCMAKE_BUILD_TYPE=Release \
           -DCMAKE_INSTALL_PREFIX:PATH=/workspace/install \
           -DTRITON_ENABLE_GRPC=ON \
-          -DTRITON_ENABLE_HTTP=ON && \
+          -DTRITON_ENABLE_HTTP=ON ../build && \
     make -j16 client
 RUN cd install && \
     export VERSION=`cat /workspace/VERSION` && \
     tar zcf /workspace/v$VERSION.clients.tar.gz *
 
 # For CI testing need to install a test script.
-COPY qa/L0_client_tar/test.sh /tmp/test.sh
+COPY qa/L0_client/test.sh /tmp/test.sh
 
 # Install an image needed by the quickstart and other documentation.
 COPY qa/images/mug.jpg images/mug.jpg

Dockerfile.client

@@ -66,18 +66,18 @@ COPY build build
 COPY src/clients src/clients
 COPY src/core src/core
 
-RUN cd build && \
+RUN mkdir builddir && cd builddir && \
     cmake -DCMAKE_BUILD_TYPE=Release \
           -DCMAKE_INSTALL_PREFIX:PATH=/workspace/install \
           -DTRITON_ENABLE_GRPC=ON \
-          -DTRITON_ENABLE_HTTP=ON && \
+          -DTRITON_ENABLE_HTTP=ON ../build && \
     make -j16 client
 RUN cd install && \
     export VERSION=`cat /workspace/VERSION` && \
     tar zcf /workspace/v$VERSION.clients.tar.gz *
 
 # For CI testing need to install a test script.
-COPY qa/L0_client_tar/test.sh /tmp/test.sh
+COPY qa/L0_client/test.sh /tmp/test.sh
 
 # Install an image needed by the quickstart and other documentation.
 COPY qa/images/mug.jpg images/mug.jpg

docs/client_library.rst

@@ -113,13 +113,18 @@ For Ubuntu, the dependencies and how to install them can be found in
 Dockerfile.client. Also note that the dependency name may be different
 depending on the version of the system.
 
-To build on Ubuntu, change to the build/ directory and run the
-following to configure and build::
+To build on Ubuntu, run the following to configure and build::
 
-  $ cd build
-  $ cmake -DCMAKE_BUILD_TYPE=Release
+  $ mkdir builddir && cd builddir
+  $ cmake -DCMAKE_BUILD_TYPE=Release ../build
   $ make -j8 client
 
+If you want to build a version of the client libraries and examples
+that does not include the CUDA shared memory support, use the
+following cmake configuration::
+
+  $ cmake -DTRITON_ENABLE_GPU=OFF -DTRITON_ENABLE_METRICS_GPU=OFF -DCMAKE_BUILD_TYPE=Release ../build
+
 When the build completes the libraries can be found in
 client/install/lib, the corresponding headers in
 client/install/include, and the Python wheel files in
@@ -155,6 +160,12 @@ are using Microsoft Visual Studio, you should do the following::
   > cmake -G"Visual Studio 16 2019" -DCMAKE_BUILD_TYPE=Release
   > MSBuild.exe client.vcxproj -p:Configuration=Release
 
+If you want to build a version of the client libraries and examples
+that does not include the CUDA shared memory support, use the
+following cmake configuration::
+
+  > cmake -G"Visual Studio 16 2019" -DTRITON_ENABLE_GPU=OFF -DTRITON_ENABLE_METRICS_GPU=OFF -DCMAKE_BUILD_TYPE=Release
+
 When the build completes the libraries can be found in
 client\\install\\lib, the corresponding headers in
 client\\install\\include, and the Python wheel files in

qa/L0_client_tar/test.sh → qa/L0_client/test.sh

@@ -25,8 +25,6 @@
 # (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
 # OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 
-CLIENT_LOG=client.log
-
 # Install the tar file
 rm -fr triton_client
 mkdir triton_client
@@ -65,6 +63,24 @@ for l in $WHLS; do
     fi
 done
 
+# This test is running in a non-NVIDIA docker so we can configure the
+# build without GPUs and make sure it works correctly.
+cd /workspace/builddir && rm -fr client ../install/*
+cmake -DTRITON_ENABLE_GPU=OFF -DTRITON_ENABLE_METRICS_GPU=OFF ../build
+make -j16 client
+
+CUDAFILES=`find /workspace/builddir/client/install -name *cuda* | wc -l`
+if [ "$CUDAFILES" != "0" ]; then
+    echo -e "*** unexpected CUDA files in TRITON_ENABLE_GPU=OFF build\n"
+    RET=1
+fi
+
+SHMFILES=`find /workspace/builddir/client/install -name *shm* | wc -l`
+if [ "$SHMFILES" != "7" ]; then
+    echo -e "*** expected 7 SHM files in TRITON_ENABLE_GPU=OFF build, got $SHMFILES\n"
+    RET=1
+fi
+
 if [ $RET -eq 0 ]; then
   echo -e "\n***\n*** Test Passed\n***"
 else

src/clients/c++/examples/CMakeLists.txt

@@ -354,6 +354,7 @@ else()
   target_link_libraries(
     simple_http_shm_client
     PRIVATE TRITON::httpclient_static
+    PRIVATE rt
   )
   install(
     TARGETS simple_http_shm_client

src/clients/c++/library/CMakeLists.txt

@@ -42,7 +42,7 @@ if(${TRITON_ENABLE_GRPC})
 
   set(
       REQUEST_HDRS
-      grpc_client.h common.h
+      grpc_client.h common.h ipc.h
   )
 
   add_library(
@@ -167,7 +167,7 @@ if(${TRITON_ENABLE_HTTP})
 
   set(
       REQUEST_HDRS
-      http_client.h common.h cencode.h
+      http_client.h common.h ipc.h cencode.h
   )
 
   add_library(

src/clients/c++/library/common.h

@@ -27,8 +27,6 @@
 
 /// \file
 
-#include "src/core/constants.h"
-
 #include <algorithm>
 #include <chrono>
 #include <condition_variable>
@@ -41,6 +39,7 @@
 #include <string>
 #include <thread>
 #include <vector>
+#include "src/core/constants.h"
 
 namespace nvidia { namespace inferenceserver { namespace client {
 

src/clients/c++/library/grpc_client.h

@@ -29,18 +29,11 @@
 
 #include <queue>
 #include "src/clients/c++/library/common.h"
+#include "src/clients/c++/library/ipc.h"
 #include "src/core/constants.h"
 #include "src/core/grpc_service.grpc.pb.h"
 #include "src/core/model_config.pb.h"
 
-#ifdef TRITON_ENABLE_GPU
-#include <cuda_runtime_api.h>
-#else
-struct cudaIpcMemHandle_t {
-};
-#endif  // TRITON_ENABLE_GPU
-
-
 namespace nvidia { namespace inferenceserver { namespace client {
 
 /// The key-value map type to be included in the request

src/clients/c++/library/http_client.h

@@ -30,13 +30,7 @@
 #include <map>
 #include <memory>
 #include "src/clients/c++/library/common.h"
-
-#ifdef TRITON_ENABLE_GPU
-#include <cuda_runtime_api.h>
-#else
-struct cudaIpcMemHandle_t {
-};
-#endif  // TRITON_ENABLE_GPU
+#include "src/clients/c++/library/ipc.h"
 
 namespace nvidia { namespace inferenceserver { namespace client {
 

src/clients/c++/library/ipc.h

@@ -0,0 +1,33 @@
+// Copyright (c) 2020, NVIDIA CORPORATION. All rights reserved.
+//
+// Redistribution and use in source and binary forms, with or without
+// modification, are permitted provided that the following conditions
+// are met:
+//  * Redistributions of source code must retain the above copyright
+//    notice, this list of conditions and the following disclaimer.
+//  * Redistributions in binary form must reproduce the above copyright
+//    notice, this list of conditions and the following disclaimer in the
+//    documentation and/or other materials provided with the distribution.
+//  * Neither the name of NVIDIA CORPORATION nor the names of its
+//    contributors may be used to endorse or promote products derived
+//    from this software without specific prior written permission.
+//
+// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS ``AS IS'' AND ANY
+// EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+// IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
+// PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT OWNER OR
+// CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
+// EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
+// PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
+// PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY
+// OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+#pragma once
+
+#ifdef TRITON_ENABLE_GPU
+#include <cuda_runtime_api.h>
+#else
+struct cudaIpcMemHandle_t {
+};
+#endif  // TRITON_ENABLE_GPU

src/clients/python/examples/CMakeLists.txt

@@ -34,12 +34,18 @@ if(${TRITON_ENABLE_HTTP})
       simple_http_string_infer_client.py
       simple_http_async_infer_client.py
       simple_http_sequence_sync_infer_client.py
-      simple_http_cudashm_client.py
       simple_http_shm_client.py
       simple_http_shm_string_client.py
       simple_http_model_control.py
     DESTINATION python
   )
+  if(${TRITON_ENABLE_GPU})
+    install(
+      PROGRAMS
+        simple_http_cudashm_client.py
+      DESTINATION python
+    )
+  endif() # TRITON_ENABLE_GPU
 endif() # TRITON_ENABLE_HTTP
 
 if(${TRITON_ENABLE_GRPC})
@@ -50,7 +56,6 @@ if(${TRITON_ENABLE_GRPC})
       grpc_explicit_int_content_client.py
       grpc_explicit_int8_content_client.py
       grpc_image_client.py
-      simple_grpc_cudashm_client.py
       simple_grpc_health_metadata.py
       simple_grpc_async_infer_client.py
       simple_grpc_infer_client.py
@@ -62,6 +67,13 @@ if(${TRITON_ENABLE_GRPC})
       simple_grpc_model_control.py
     DESTINATION python
   )
+  if(${TRITON_ENABLE_GPU})
+    install(
+      PROGRAMS
+        simple_grpc_cudashm_client.py
+      DESTINATION python
+    )
+  endif() # TRITON_ENABLE_GPU
 endif() # TRITON_ENABLE_GRPC
 
 if(${TRITON_ENABLE_HTTP} OR ${TRITON_ENABLE_GRPC})