Add gpuCI support (dask-contrib#240)

* Add dask-cudf tests with GPU marker

* Change --gpu option to --rungpu

* Add build scripts for gpuCI

* xfail GPU table creation from memory

* Activate correct environment

* Force reinstall openjdk to resolve Maven issues

* Try to install Java side with debug switches

* Pass options to correct commnand

* Set JAVA_HOME to gpuCI workspace

* Specifiy MAVEN_OPTS

* Move pytest marker to pytest.ini

* Add GPU tests for Context

* Bump RAPIDS_VER to 21.12

* xfail GPU tests that aren't currently working

* Rerun tests

* Use gpu param for table creation in test_input_types

Author: charlesbluca

conftest.py

@@ -1,3 +1,12 @@
 import pytest
 
 pytest_plugins = ["distributed.utils_test", "tests.integration.fixtures"]
+
+
+def pytest_addoption(parser):
+    parser.addoption("--rungpu", action="store_true", help="run tests meant for GPU")
+
+
+def pytest_runtest_setup(item):
+    if "gpu" in item.keywords and not item.config.getoption("--rungpu"):
+        pytest.skip("need --rungpu option to run")

continuous_integration/gpuci/axis.yaml

@@ -0,0 +1,13 @@
+PYTHON_VER:
+- "3.8"
+
+CUDA_VER:
+- "11.2"
+
+LINUX_VER:
+- ubuntu18.04
+
+RAPIDS_VER:
+- "21.12"
+
+excludes:

continuous_integration/gpuci/build.sh

@@ -0,0 +1,62 @@
+##################################################
+# dask-sql GPU build and test script for CI      #
+##################################################
+set -e
+NUMARGS=$#
+ARGS=$*
+
+# Arg parsing function
+function hasArg {
+    (( ${NUMARGS} != 0 )) && (echo " ${ARGS} " | grep -q " $1 ")
+}
+
+# Set path and build parallel level
+export PATH=/opt/conda/bin:/usr/local/cuda/bin:$PATH
+export PARALLEL_LEVEL=${PARALLEL_LEVEL:-4}
+
+# Set home to the job's workspace
+export HOME="$WORKSPACE"
+
+# specify maven options
+export MAVEN_OPTS="-Dmaven.repo.local=${WORKSPACE}/.m2/repository"
+
+# Switch to project root; also root of repo checkout
+cd "$WORKSPACE"
+
+# Determine CUDA release version
+export CUDA_REL=${CUDA_VERSION%.*}
+
+################################################################################
+# SETUP - Check environment
+################################################################################
+
+gpuci_logger "Check environment variables"
+env
+
+gpuci_logger "Check GPU usage"
+nvidia-smi
+
+gpuci_logger "Activate conda env"
+. /opt/conda/etc/profile.d/conda.sh
+conda activate dask_sql
+
+gpuci_logger "Install dask"
+python -m pip install git+https://github.com/dask/dask
+
+gpuci_logger "Install distributed"
+python -m pip install git+https://github.com/dask/distributed
+
+gpuci_logger "Install dask-sql"
+pip install -e ".[dev]"
+python setup.py java
+
+gpuci_logger "Check Python version"
+python --version
+
+gpuci_logger "Check conda environment"
+conda info
+conda config --show-sources
+conda list --show-channel-urls
+
+gpuci_logger "Python py.test for dask-sql"
+py.test $WORKSPACE -n 4 -v -m gpu --rungpu --junitxml="$WORKSPACE/junit-dask-sql.xml" --cov-config="$WORKSPACE/.coveragerc" --cov=dask_sql --cov-report=xml:"$WORKSPACE/dask-sql-coverage.xml" --cov-report term

pytest.ini

@@ -5,3 +5,5 @@ addopts =
     --cov-report=term-missing
 testpaths =
     tests
+markers =
+    gpu: marks tests that require GPUs (skipped by default, run with '--rungpu')

tests/integration/test_create.py

@@ -8,7 +8,8 @@
 
 
 @skip_if_external_scheduler
-def test_create_from_csv(c, df, temporary_data_file):
+@pytest.mark.parametrize("gpu", [False, pytest.param(True, marks=pytest.mark.gpu)])
+def test_create_from_csv(c, df, temporary_data_file, gpu):
     df.to_csv(temporary_data_file, index=False)
 
     c.sql(
@@ -17,7 +18,8 @@ def test_create_from_csv(c, df, temporary_data_file):
             new_table
         WITH (
             location = '{temporary_data_file}',
-            format = 'csv'
+            format = 'csv',
+            gpu = {gpu}
         )
     """
     )
@@ -28,10 +30,28 @@ def test_create_from_csv(c, df, temporary_data_file):
     """
     ).compute()
 
+    if gpu:
+        result_df = result_df.to_pandas()
+
     assert_frame_equal(result_df, df)
 
 
-def test_cluster_memory(client, c, df):
+@pytest.mark.parametrize(
+    "gpu",
+    [
+        False,
+        pytest.param(
+            True,
+            marks=[
+                pytest.mark.gpu,
+                pytest.mark.xfail(
+                    reason="dataframes on memory currently aren't being converted to dask-cudf"
+                ),
+            ],
+        ),
+    ],
+)
+def test_cluster_memory(client, c, df, gpu):
     client.publish_dataset(df=dd.from_pandas(df, npartitions=1))
 
     c.sql(
@@ -40,7 +60,8 @@ def test_cluster_memory(client, c, df):
             new_table
         WITH (
             location = 'df',
-            format = 'memory'
+            format = 'memory',
+            gpu = {gpu}
         )
     """
     )
@@ -51,11 +72,15 @@ def test_cluster_memory(client, c, df):
     """
     ).compute()
 
+    if gpu:
+        return_df = return_df.to_pandas()
+
     assert_frame_equal(df, return_df)
 
 
 @skip_if_external_scheduler
-def test_create_from_csv_persist(c, df, temporary_data_file):
+@pytest.mark.parametrize("gpu", [False, pytest.param(True, marks=pytest.mark.gpu)])
+def test_create_from_csv_persist(c, df, temporary_data_file, gpu):
     df.to_csv(temporary_data_file, index=False)
 
     c.sql(
@@ -65,7 +90,8 @@ def test_create_from_csv_persist(c, df, temporary_data_file):
         WITH (
             location = '{temporary_data_file}',
             format = 'csv',
-            persist = True
+            persist = True,
+            gpu = {gpu}
         )
     """
     )
@@ -76,6 +102,9 @@ def test_create_from_csv_persist(c, df, temporary_data_file):
     """
     ).compute()
 
+    if gpu:
+        return_df = return_df.to_pandas()
+
     assert_frame_equal(df, return_df)
 
 
@@ -143,15 +172,29 @@ def test_create_from_query(c, df):
 
 
 @skip_if_external_scheduler
-def test_view_table_persist(c, temporary_data_file, df):
+@pytest.mark.parametrize(
+    "gpu",
+    [
+        False,
+        pytest.param(
+            True,
+            marks=(
+                pytest.mark.gpu,
+                pytest.mark.xfail(reason="to_pandas() changes int precision"),
+            ),
+        ),
+    ],
+)
+def test_view_table_persist(c, temporary_data_file, df, gpu):
     df.to_csv(temporary_data_file, index=False)
     c.sql(
         f"""
         CREATE TABLE
             new_table
         WITH (
             location = '{temporary_data_file}',
-            format = 'csv'
+            format = 'csv',
+            gpu = {gpu}
         )
     """
     )
@@ -177,21 +220,27 @@ def test_view_table_persist(c, temporary_data_file, df):
     """
     )
 
-    assert_frame_equal(
-        c.sql("SELECT c FROM count_view").compute(), pd.DataFrame({"c": [700]})
-    )
-    assert_frame_equal(
-        c.sql("SELECT c FROM count_table").compute(), pd.DataFrame({"c": [700]})
-    )
+    from_view = c.sql("SELECT c FROM count_view").compute()
+    from_table = c.sql("SELECT c FROM count_table").compute()
+
+    if gpu:
+        from_view = from_view.to_pandas()
+        from_table = from_table.to_pandas()
+
+    assert_frame_equal(from_view, pd.DataFrame({"c": [700]}))
+    assert_frame_equal(from_table, pd.DataFrame({"c": [700]}))
 
     df.iloc[:10].to_csv(temporary_data_file, index=False)
 
-    assert_frame_equal(
-        c.sql("SELECT c FROM count_view").compute(), pd.DataFrame({"c": [10]})
-    )
-    assert_frame_equal(
-        c.sql("SELECT c FROM count_table").compute(), pd.DataFrame({"c": [700]})
-    )
+    from_view = c.sql("SELECT c FROM count_view").compute()
+    from_table = c.sql("SELECT c FROM count_table").compute()
+
+    if gpu:
+        from_view = from_view.to_pandas()
+        from_table = from_table.to_pandas()
+
+    assert_frame_equal(from_view, pd.DataFrame({"c": [10]}))
+    assert_frame_equal(from_table, pd.DataFrame({"c": [700]}))
 
 
 def test_replace_and_error(c, temporary_data_file, df):