Adding code for release v0.4.1 (commit 92a8fcd)

Author: smatysik-nv

.github/ISSUE_TEMPLATE/bug_report_form.yml

@@ -28,7 +28,7 @@ body:
     attributes:
       label: Version
       description: What version of nvImageCodec are you running?
-      placeholder: "example: 0.4.0"
+      placeholder: "example: 0.4.1"
     validations:
       required: true
 

CMakeLists.txt

@@ -18,7 +18,7 @@ cmake_minimum_required(VERSION 3.18 FATAL_ERROR)
 include(CMakeDependentOption)
 
 set(NVIMGCODEC_PROJECT_NAME "nvimgcodec")
-set(NVIMGCODEC_VERSION "0.4.0")
+set(NVIMGCODEC_VERSION "0.4.1")
 
 if (NVIMGCODEC_FLAVOR)
     set(NVIMGCODEC_FLAVOR "${NVIMGCODEC_FLAVOR}")  # could be nightly, weekly, etc if necessary

CONTRIBUTING.md

@@ -0,0 +1,29 @@
+## nvImageCodec OSS Contribution Rules
+
+The nvImageCodec project is develop internally at NVIDIA. The source code of nvImageCodec are made available to the public regularly with each release. External contributions are welcome with the reservation that you must agree to the Developer Certificate of Origin presented below and certify that contribution is under the Apache License Version 2.0.
+
+All enhancement, bugfix, or change requests must begin with the creation of a [nvImageCodec Issue Request](https://github.com/nvidia/nvImageCodec/issues).
+
+### Developer Certificate of Origin
+
+```
+  Developer Certificate of Origin
+  Version 1.1
+
+  Copyright (C) 2004, 2006 The Linux Foundation and its contributors.
+
+  Everyone is permitted to copy and distribute verbatim copies of this license document, but changing it is not allowed.
+
+
+  Developer's Certificate of Origin 1.1
+
+  By making a contribution to this project, I certify that:
+
+  (a) The contribution was created in whole or in part by me and I have the right to submit it under the open source license indicated in the file; or
+
+  (b) The contribution is based upon previous work that, to the best of my knowledge, is covered under an appropriate open source license and I have the right under that license to submit that work with modifications, whether created in whole or in part by me, under the same open source license (unless I am permitted to submit under a different license), as indicated in the file; or
+
+  (c) The contribution was provided directly to me by some other person who certified (a), (b) or (c) and I have not modified it.
+
+  (d) I understand and agree that this project and the contribution are public and that a record of the contribution (including all personal information I submit with it, including my sign-off) is maintained indefinitely and may be redistributed consistent with this project or the open source license(s) involved.
+```

README.md

@@ -1,6 +1,6 @@
 # nvImageCodec
 
-![Version](https://img.shields.io/badge/Version-v0.4.0--beta-blue)
+![Version](https://img.shields.io/badge/Version-v0.4.1--beta-blue)
 [![License](https://img.shields.io/badge/License-Apache_2.0-yellogreen.svg)](https://opensource.org/licenses/Apache-2.0)
 
 ![Platform](https://img.shields.io/badge/Platform-linux--x86__64_%7C_linux--aarch64_%7C_windows--64_wsl2_%7C_windows--64-blue)
@@ -114,6 +114,12 @@ pip install nvidia-nvimgcodec-cu11
 pip install nvidia-nvimgcodec-cu12
 ```
 
+### nvImageCodec Python for Tegra (CUDA 12.x only)
+
+```
+pip install nvidia-nvimgcodec-tegra-cu12
+```
+
 ### Optional installation of nvJPEG library
 
 If you do not have CUDA Toolkit installed, or you would like install nvJPEG library independently, you can use the instructions described below.
@@ -246,17 +252,17 @@ This will generate in build directory *.zip or *tar.xz files
 #### Tar file installation
 
 ```
-tar -xvf nvimgcodec-0.4.0.0-cuda12-x86_64-linux-lib.tar.gz -C /opt/nvidia/
+tar -xvf nvimgcodec-0.4.1.0-cuda12-x86_64-linux-lib.tar.gz -C /opt/nvidia/
 ```
 
 #### DEB File Installation
 ```
-sudo apt-get install -y ./nvimgcodec-0.4.0.0-cuda12-x86_64-linux-lib.deb
+sudo apt-get install -y ./nvimgcodec-0.4.1.0-cuda12-x86_64-linux-lib.deb
 ```
 #### Python WHL File Installation
 
 ```
-pip install nvidia_nvimgcodec_cu12-0.4.0-py3-none-manylinux2014_x86_64.whl
+pip install nvidia_nvimgcodec_cu12-0.4.1-py3-none-manylinux2014_x86_64.whl
 ```
 
 ### Installation from sources
@@ -316,7 +322,7 @@ Run Python API tests
 First install python wheel. You would also need to have installed all Python tests dependencies (see Dockerfiles). 
 
 ```
-pip install nvidia_nvimgcodec_cu12-0.4.0.x-py3-none-manylinux2014_x86_64.whl
+pip install nvidia_nvimgcodec_cu12-0.4.1.x-py3-none-manylinux2014_x86_64.whl
 ```
 
 Run tests

example/nvimtrans/main.cpp

@@ -143,7 +143,7 @@ void fill_encode_params(const CommandLineParams& params, fs::path output_path, n
 }
 
 void generate_backends_without_hw_gpu(std::vector<nvimgcodecBackend_t>& nvimgcds_backends) {
-    nvimgcds_backends.resize(3);
+    nvimgcds_backends.resize(3, nvimgcodecBackend_t{NVIMGCODEC_STRUCTURE_TYPE_BACKEND, sizeof(nvimgcodecBackend_t), nullptr});
     nvimgcds_backends[0].kind = NVIMGCODEC_BACKEND_KIND_CPU_ONLY;
     nvimgcds_backends[0].params = {NVIMGCODEC_STRUCTURE_TYPE_BACKEND_PARAMS, sizeof(nvimgcodecBackendParams_t), nullptr, 1.0f, NVIMGCODEC_LOAD_HINT_POLICY_FIXED};
     nvimgcds_backends[1].kind = NVIMGCODEC_BACKEND_KIND_GPU_ONLY;

extensions/nvtiff/cuda_decoder.cpp

@@ -527,7 +527,12 @@ nvimgcodecStatus_t Decoder::decode(const nvimgcodecImageDesc_t* image, const nvi
         }
 
         auto& nvtiff_decoder = resources.decoder(image_info.cuda_stream);
-        XM_CHECK_NVTIFF(nvtiffDecodeCheckSupported(resources.nvtiff_stream_, nvtiff_decoder, resources.decode_params_, 0));
+        bool decode_check_ret = nvtiffDecodeCheckSupported(resources.nvtiff_stream_, nvtiff_decoder, resources.decode_params_, 0);
+        if (decode_check_ret != NVTIFF_STATUS_SUCCESS) {
+            NVIMGCODEC_LOG_INFO(framework_, plugin_id_, "nvtiffDecodeCheckSupported returned error code: " << decode_check_ret);
+            image->imageReady(image->instance, NVIMGCODEC_PROCESSING_STATUS_CODESTREAM_UNSUPPORTED);
+            return NVIMGCODEC_STATUS_CODESTREAM_UNSUPPORTED;
+        }
 
         bool need_conversion =
             (dec_image_info.sample_format != image_info.sample_format && image_info.sample_format != NVIMGCODEC_SAMPLEFORMAT_I_UNCHANGED) ||
@@ -560,6 +565,9 @@ nvimgcodecStatus_t Decoder::decode(const nvimgcodecImageDesc_t* image, const nvi
             NVIMGCODEC_LOG_DEBUG(framework_, plugin_id_, "LaunchConvertNormKernel");
             nvimgcodec::LaunchConvertNormKernel(image_info, dec_image_info, image_info.cuda_stream);
 
+            // Record event on user stream to synchronize on next iteration
+            XM_CHECK_CUDA(cudaEventRecord(resources.event_, image_info.cuda_stream));
+
             image->imageReady(image->instance, NVIMGCODEC_PROCESSING_STATUS_SUCCESS);
             return NVIMGCODEC_STATUS_SUCCESS;
         } else {
@@ -569,6 +577,9 @@ nvimgcodecStatus_t Decoder::decode(const nvimgcodecImageDesc_t* image, const nvi
             XM_CHECK_NVTIFF(nvtiffDecodeImage(
                 resources.nvtiff_stream_, nvtiff_decoder, resources.decode_params_, 0, output_ptr, image_info.cuda_stream));
 
+            // Record event on user stream to synchronize on next iteration
+            XM_CHECK_CUDA(cudaEventRecord(resources.event_, image_info.cuda_stream));
+
             image->imageReady(image->instance, NVIMGCODEC_PROCESSING_STATUS_SUCCESS);
             return NVIMGCODEC_STATUS_SUCCESS;
         }

python/bundle-wheel.sh

@@ -209,8 +209,8 @@ echo "Fixed hashed names"
 patch_rpath() {
     local FILE=$1
     UPDIRS=$(dirname $(echo "$FILE" | sed "s|$PKGNAME_PATH||") | sed 's/[^\/][^\/]*/../g')
-    echo "Setting rpath of $FILE to '\$ORIGIN:\$ORIGIN$UPDIRS:\$ORIGIN$UPDIRS/.libs:\$ORIGIN/../../nvjpeg/lib:\$ORIGIN/../../nvjpeg2k/lib:/usr/local/cuda/lib64'"
-    patchelf --set-rpath "\$ORIGIN:\$ORIGIN$UPDIRS:\$ORIGIN$UPDIRS/.libs:\$ORIGIN/../../nvjpeg/lib:\$ORIGIN/../../nvjpeg2k/lib:/usr/local/cuda/lib64" $FILE
+    echo "Setting rpath of $FILE to '\$ORIGIN:\$ORIGIN$UPDIRS:\$ORIGIN$UPDIRS/.libs:\$ORIGIN/../../nvjpeg/lib:\$ORIGIN/../../nvjpeg2k/lib:\$ORIGIN/../../nvtiff/lib:\$ORIGIN/../../nvcomp:/usr/local/cuda/lib64'"
+    patchelf --set-rpath "\$ORIGIN:\$ORIGIN$UPDIRS:\$ORIGIN$UPDIRS/.libs:\$ORIGIN/../../nvjpeg/lib:\$ORIGIN/../../nvjpeg2k/lib:\$ORIGIN/../../nvtiff/lib:\$ORIGIN/../../nvcomp:/usr/local/cuda/lib64" $FILE
     patchelf --print-rpath $FILE
 }
 echo "Fixing rpath of main files..."

src/image_generic_codec.h

@@ -146,10 +146,15 @@ struct SampleEntry : public IImage
 
 struct PerThread
 {
-    explicit PerThread()
+    explicit PerThread(int device_id)
     {
-        CHECK_CUDA(cudaStreamCreateWithFlags(&stream, cudaStreamNonBlocking));
-        CHECK_CUDA(cudaEventCreate(&event));
+        if (device_id == NVIMGCODEC_DEVICE_CPU_ONLY) {
+            stream = nullptr;
+            event = nullptr;
+        } else {
+            CHECK_CUDA(cudaStreamCreateWithFlags(&stream, cudaStreamNonBlocking));
+            CHECK_CUDA(cudaEventCreate(&event));
+        }
     }
     PerThread(const PerThread& oth) = delete;
     PerThread& operator=(const PerThread& oth) = delete;
@@ -245,7 +250,10 @@ class ImageGenericCodec
                 }
             }
         }
-        per_thread_.resize(num_threads_ + 1);
+        per_thread_.reserve(num_threads_ + 1);
+        for (size_t i = 0; i < num_threads_ + 1; i++) {
+            per_thread_.emplace_back(exec_params_.device_id);
+        }
 
         size_t total_num_processors = 0;
         for (size_t codec_idx = 0; codec_idx < codec_registry_->getCodecsCount(); codec_idx++) {
@@ -384,6 +392,8 @@ class ImageGenericCodec
 
     void preSync(SampleEntry<ProcessorEntry>& sample, int tid)
     {
+        if (exec_params_.device_id == NVIMGCODEC_DEVICE_CPU_ONLY)
+            return;
         auto& t = per_thread_[tid];
         auto user_cuda_stream = sample.orig_image_info.cuda_stream;
         if (t.user_streams.find(user_cuda_stream) == t.user_streams.end()) {
@@ -400,6 +410,8 @@ class ImageGenericCodec
 
     void postSync(int tid)
     {
+        if (exec_params_.device_id == NVIMGCODEC_DEVICE_CPU_ONLY)
+            return;
         auto& t = per_thread_[tid];
         if (!t.user_streams.empty()) {
             nvtx3::scoped_range marker{"sync"};