Promote amd-staging to amd-mainline

.azuredevops/rocm-ci.yml

@@ -13,6 +13,7 @@ trigger:
   batch: true
   branches:
     include:
+    - develop
     - amd-staging
     - amd-mainline
   paths:
@@ -29,6 +30,7 @@ pr:
   autoCancel: true
   branches:
     include:
+    - develop
     - amd-staging
     - amd-mainline
   paths:

CHANGELOG.md

@@ -2,6 +2,10 @@
 
 Full documentation for ROCm Compute Profiler is available at [https://rocm.docs.amd.com/projects/rocprofiler-compute/en/latest/](https://rocm.docs.amd.com/projects/rocprofiler-compute/en/latest/).
 
+## Unreleased
+
+* Add Docker files to package the application and dependencies into a single portable and executable standalone binary file  
+
 ## (Unreleased) ROCm Compute Profiler 3.1.0 for ROCm 6.4.0
 
 ### Added

CMakeLists.txt

@@ -372,6 +372,30 @@ add_custom_target(
         "src/${PACKAGE_NAME},cmake/Dockerfile,cmake/rocm_install.sh,docker/docker-entrypoint.sh,src/rocprof_compute_analyze/convertor/mongodb/convert"
     )
 
+# Standalone binary creation
+add_custom_target(
+    standalonebinary
+    # Change working directory to src
+    WORKING_DIRECTORY ${PROJECT_SOURCE_DIR}/src
+    # Check nuitka
+    COMMAND ${Python3_EXECUTABLE} -m pip list | grep -i nuitka > /dev/null 2>&1
+    # Check patchelf
+    COMMAND ${Python3_EXECUTABLE} -m pip list | grep -i patchelf > /dev/null 2>&1
+    # Create VERSION.sha file
+    COMMAND git -C ${PROJECT_SOURCE_DIR} rev-parse HEAD > VERSION.sha
+    # Build standalone binary
+    COMMAND
+        ${Python3_EXECUTABLE} -m nuitka --mode=onefile
+        --include-data-files=${PROJECT_SOURCE_DIR}/VERSION*=./ --enable-plugin=no-qt
+        --include-package-data=dash_svg --include-package=dash_bootstrap_components
+        --include-package=plotly --include-package-data=kaleido
+        --include-package=rocprof_compute_soc --include-package-data=rocprof_compute_soc
+        --include-package-data=utils rocprof-compute
+    # Remove library rpath from executable
+    COMMAND patchelf --remove-rpath rocprof-compute.bin
+    # Move to build directory
+    COMMAND mv rocprof-compute.bin ${CMAKE_BINARY_DIR})
+
 install(
     FILES ${PROJECT_SOURCE_DIR}/LICENSE
     DESTINATION ${CMAKE_INSTALL_DOCDIR}

LICENSE

@@ -19,3 +19,22 @@ AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
 LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
 OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
 SOFTWARE.
+
+This application uses the following dependencies and their usage is governed by their respective licenses
+Python 3 standard library: PSFL
+Nuitka specific runtime code: Apache 2.0 license
+astunparse python library: PSFL
+colorlover python library: MIT
+dash python library: MIT
+matplotlib python library: PSFL
+numpy python library: BSD
+pandas python library: BSD
+pymongo python library: Apache 2.0 license
+pyyaml python library: MIT
+tabulate python library: MIT
+tqdm python library: MIT
+dash-svg python library: MIT
+dash-bootstrap-components python library: MIT
+kaleido python library: MIT
+setuptools python library: MIT
+plotille python library: MIT

README.md

@@ -37,8 +37,8 @@ Users may checkout `amd-staging` to preview upcoming features.
 ## Testing
 
 To quickly get the environment (bash shell) for building and testing, run the following commands:
-* `cd utils/docker_env`
-* `docker compose run app`
+* `cd docker`
+* `docker compose -f docker-compose.test.yml run test`
 
 Inside the docker container, clean, build and install the project with tests enabled:
 ```
@@ -56,6 +56,27 @@ For manual testing, you can find the executable at `install/bin/rocprof-compute`
 
 NOTE: This Dockerfile uses `rocm/dev-ubuntu-22.04` as the base image
 
+## Standalone binary
+
+To create a standalone binary, run the following commands:
+* `cd docker`
+* `docker compose -f docker-compose.standalone.yml run standalone`
+
+You should find the rocprof-compute.bin standalone binary inside the `build` folder in the root directory of the project.
+
+To build the binary we follow these steps:
+* Use RHEL 8 image used to build ROCm as the base image
+* Install python3.8
+* Install dependencies for runtime and for making standalone binary
+* Call the make target which uses Nuitka to build the standalone binary
+
+NOTE: Since RHEL 8 ships with glibc version 2.28, this standalone binary can only be run on environment with glibc version greater than 2.28.
+glibc version can be checked using `ldd --version` command.
+
+NOTE: libnss3.so shared library is required when using --roof-only option which generates roofline data in PDF format 
+
+To test the standalone binary provide the `--call-binary` option to pytest.
+
 ## How to Cite
 
 This software can be cited using a Zenodo

docker/Dockerfile.standalone

@@ -0,0 +1,22 @@
+FROM redhat/ubi8:8.10-1184
+
+WORKDIR /app
+
+RUN yum install -y curl gcc cmake git
+
+# Allows running git commands in /app
+RUN git config --global --add safe.directory /app
+
+RUN yum install -y python38 python38-devel && \
+    yum clean all && \
+    rm -rf /var/cache/yum && \
+    curl -sS https://bootstrap.pypa.io/get-pip.py -o get-pip.py && \
+    python3.8 get-pip.py
+
+CMD ["/bin/bash", "-c", "\
+    python3.8 -m pip install -r requirements.txt \
+    && python3.8 -m pip install nuitka patchelf \
+    && rm -rf build \
+    && cmake -B build -S . \
+    && make -C build standalonebinary \
+"]

utils/docker_env/Dockerfile → docker/Dockerfile.test

@@ -6,10 +6,13 @@ WORKDIR /app
 
 # Update package list and install prerequisites
 RUN apt-get update && apt-get install -y \
-    software-properties-common cmake locales \
+    software-properties-common cmake locales git \
     && add-apt-repository ppa:deadsnakes/ppa \
     && apt-get update
 
+# Allows running git commands in /app
+RUN git config --global --add safe.directory /app
+
 # Generate the desired locale
 RUN locale-gen en_US.UTF-8
 
@@ -19,11 +22,9 @@ RUN apt-get install -y python3.10 python3.10-venv python3.10-dev python3-pip
 # Set Python 3.10 as the default python3
 RUN update-alternatives --install /usr/bin/python3 python3 /usr/bin/python3.10 1
 
-# Copy your application code to the container
-COPY . .
-
 # Install any dependencies specified in requirements.txt
-RUN pip3 install --no-cache-dir -r requirements.txt -r requirements-test.txt
-
-# Command to run your application
-CMD ["/bin/bash"]
+# Run interactive bash shell
+CMD ["/bin/bash", "-c", "\
+    python3.10 -m pip install -r requirements.txt -r requirements-test.txt \
+    && exec /bin/bash \
+"]

docker/docker-compose.standalone.yml

@@ -0,0 +1,12 @@
+services:
+  standalone:
+    build:
+      context: ../
+      dockerfile: docker/Dockerfile.standalone
+    devices:
+      - /dev/kfd
+      - /dev/dri
+    security_opt:
+      - seccomp:unconfined
+    volumes:
+      - ../:/app

docker/docker-compose.test.yml

@@ -0,0 +1,12 @@
+services:
+  test:
+    build:
+      context: ../
+      dockerfile: docker/Dockerfile.test
+    devices:
+      - /dev/kfd
+      - /dev/dri
+    security_opt:
+      - seccomp:unconfined
+    volumes:
+      - ../:/app

src/rocprof_compute_profile/profiler_base.py

@@ -26,6 +26,7 @@
 import logging
 import os
 import re
+import shutil
 import sys
 import time
 from abc import ABC, abstractmethod
@@ -77,6 +78,26 @@ def join_prof(self, out=None):
                 out = self.__args.path + "/pmc_perf.csv"
             files = glob.glob(self.__args.path + "/" + "pmc_perf_*.csv")
             files.extend(glob.glob(self.__args.path + "/" + "SQ_*.csv"))
+
+            if self.get_args().hip_trace:
+                # remove hip api trace ouputs from this list
+                files = [
+                    f
+                    for f in files
+                    if not re.compile(r"^.*_hip_api_trace\.csv$").match(
+                        os.path.basename(f)
+                    )
+                ]
+
+            if self.get_args().kokkos_trace:
+                # remove marker api trace ouputs from this list
+                files = [
+                    f
+                    for f in files
+                    if not re.compile(r"^.*_marker_api_trace\.csv$").match(
+                        os.path.basename(f)
+                    )
+                ]
         elif type(self.__args.path) == list:
             files = self.__args.path
         else:
@@ -266,7 +287,8 @@ def pre_processing(self):
         # verify correct formatting for application binary
         self.__args.remaining = self.__args.remaining[1:]
         if self.__args.remaining:
-            if not Path(self.__args.remaining[0]).is_file():
+            # Ensure that command points to an executable
+            if not shutil.which(self.__args.remaining[0]):
                 console_error(
                     "Your command %s doesn't point to a executable. Please verify."
                     % self.__args.remaining[0]

src/rocprof_compute_soc/analysis_configs/gfx906/0200_system-speed-of-light.yaml

@@ -32,6 +32,12 @@ Panel Config:
             peak: (((($max_sclk * $cu_per_gpu) * 64) * 2) / 1000)
             pop: None # No perf counter
             tips:
+          MFMA FLOPs (F8):
+            value: None # No HW module
+            unit: GFLOP
+            peak: None # No HW module
+            pop: None # No HW module
+            tips:
           MFMA FLOPs (BF16):
             value: None # No perf counter
             unit: GFLOPs

src/rocprof_compute_soc/analysis_configs/gfx906/1000_compute-unit-instruction-mix.yaml

@@ -241,6 +241,12 @@ Panel Config:
             max: None # No HW module
             unit: (instr + $normUnit)
             tips:
+          MFMA-F8:
+            avg: None # No HW module
+            min: None # No HW module
+            max: None # No HW module None # No HW module
+            unit: (instr + $normUnit)
+            tips:
           MFMA-F16:
             avg: None # No HW module
             min: None # No HW module

src/rocprof_compute_soc/analysis_configs/gfx906/1100_compute-unit-compute-pipeline.yaml

@@ -21,16 +21,22 @@ Panel Config:
         metric:
           VALU FLOPs:
             value: None # No perf counter
-            Unit: None
+            unit: None
             peak: None
             pop: None
             tips:
           VALU IOPs:
             value: None # No perf counter
-            Unit: None
+            unit: None
             peak: None
             pop: None
             tips:
+          MFMA FLOPs (F8):
+            value: None # No perf counter
+            unit: GFLOP
+            peak: None # No perf counter
+            pop: None # No perf counter
+            tips:
           MFMA FLOPs (BF16):
             value: None # No perf counter
             Unit: None
@@ -39,25 +45,25 @@ Panel Config:
             tips:
           MFMA FLOPs (F16):
             value: None # No perf counter
-            Unit: None
+            unit: None
             peak: None
             pop: None
             tips:
           MFMA FLOPs (F32):
             value: None # No perf counter
-            Unit: None
+            unit: None
             peak: None
             pop: None
             tips:
           MFMA FLOPs (F64):
             value: None # No perf counter
-            Unit: None
+            unit: None
             peak: None
             pop: None
             tips:
           MFMA IOPs (INT8):
             value: None # No perf counter
-            Unit: None
+            unit: None
             peak: None
             pop: None
             tips:
@@ -174,6 +180,12 @@ Panel Config:
             max: None # No perf counter
             unit: (OPs  + $normUnit)
             tips:
+          F8 OPs:
+            avg: None # No HW module
+            min: None # No HW module
+            max: None # No HW module
+            unit: (OPs  + $normUnit)
+            tips:
           F16 OPs:
             avg: None # No perf counter
             min: None # No perf counter

src/rocprof_compute_soc/analysis_configs/gfx908/0200_system-speed-of-light.yaml

@@ -32,6 +32,12 @@ Panel Config:
             peak: (((($max_sclk * $cu_per_gpu) * 64) * 2) / 1000)
             pop: None # No perf counter
             tips:
+          MFMA FLOPs (F8):
+            value: None # No HW module
+            unit: GFLOP
+            peak: None # No HW module
+            pop: None # No HW module
+            tips:
           MFMA FLOPs (BF16):
             value: None # No perf counter
             unit: GFLOPs

src/rocprof_compute_soc/analysis_configs/gfx908/1000_compute-unit-instruction-mix.yaml

@@ -73,13 +73,13 @@ Panel Config:
           unit: Unit
           tips: Tips
         metric:
-          INT-32:
+          INT32:
             avg: None # No perf counter
             min: None # No perf counter
             max: None # No perf counter
             unit: (instr + $normUnit)
             tips:
-          INT-64:
+          INT64:
             avg: None # No perf counter
             min: None # No perf counter
             max: None # No perf counter
@@ -241,6 +241,12 @@ Panel Config:
             max: None # No HW module
             unit: (instr + $normUnit)
             tips:
+          MFMA-F8:
+            avg: None # No HW module
+            min: None # No HW module
+            max: None # No HW module None # No HW module
+            unit: (instr + $normUnit)
+            tips:
           MFMA-F16:
             avg: None # No HW module
             min: None # No HW module