Re-enable CUDA tests. (#8237)
* Remove additional bazel file and add cuda config for turing.
* Add CUDA target and driver option to turing builds.
* Remove flags to enable cuda build pending VM updates.
* Disable cuda in bazel tests
diff --git a/build_tools/bazel/iree_check_test.bzl b/build_tools/bazel/iree_check_test.bzl
index b34ef02..a72fe41 100644
--- a/build_tools/bazel/iree_check_test.bzl
+++ b/build_tools/bazel/iree_check_test.bzl
@@ -127,6 +127,9 @@
tests = []
for src in srcs:
+ # CUDA backend/driver not supported by Bazel build.
+ if target_backend == "cuda" or driver == "cuda":
+ continue
test_name = "_".join([name, src])
iree_check_test(
name = test_name,
@@ -198,6 +201,9 @@
# could just create a test suite. The latter seems simpler and more readable.
tests = []
for backend, driver in target_backends_and_drivers:
+ # CUDA backend/driver not supported by Bazel build.
+ if backend == "cuda" or driver == "cuda":
+ continue
suite_name = "_".join([name, backend, driver])
iree_check_single_backend_test_suite(
name = suite_name,
diff --git a/build_tools/cmake/clean_build.sh b/build_tools/cmake/clean_build.sh
index 543412b..0af2e4e 100755
--- a/build_tools/cmake/clean_build.sh
+++ b/build_tools/cmake/clean_build.sh
@@ -17,4 +17,4 @@
cd ${ROOT_DIR?}
rm -rf build/
-./build_tools/cmake/rebuild.sh
+./build_tools/cmake/rebuild.sh "$@"
diff --git a/build_tools/cmake/rebuild.sh b/build_tools/cmake/rebuild.sh
index 68e9420..4151aa5 100755
--- a/build_tools/cmake/rebuild.sh
+++ b/build_tools/cmake/rebuild.sh
@@ -50,5 +50,5 @@
"-DIREE_ENABLE_ASSERTIONS=ON"
)
-"$CMAKE_BIN" "${CMAKE_ARGS[@]?}" ..
+"$CMAKE_BIN" "${CMAKE_ARGS[@]?}" "$@" ..
"$CMAKE_BIN" --build .
diff --git a/build_tools/kokoro/gcp_ubuntu/cmake-bazel/linux/x86-turing/build.sh b/build_tools/kokoro/gcp_ubuntu/cmake-bazel/linux/x86-turing/build.sh
index a1cff76..0baddcd 100755
--- a/build_tools/kokoro/gcp_ubuntu/cmake-bazel/linux/x86-turing/build.sh
+++ b/build_tools/kokoro/gcp_ubuntu/cmake-bazel/linux/x86-turing/build.sh
@@ -57,6 +57,8 @@
# TODO(gcmn): It would be nice to be able to build and test as much as possible,
# so a build failure only prevents building/testing things that depend on it and
# we can still run the other tests.
+# TODO: Add "-DIREE_TARGET_BACKEND_CUDA=ON -DIREE_HAL_DRIVER_CUDA=ON" once the
+# VMs have been updated with the correct CUDA SDK.
echo "Configuring CMake"
"${CMAKE_BIN}" -B "${CMAKE_BUILD_DIR?}" -G Ninja \
-DIREE_TF_TOOLS_ROOT="${BAZEL_BINDIR?}/iree_tf_compiler/" \
diff --git a/build_tools/kokoro/gcp_ubuntu/cmake/linux/x86-turing/build.sh b/build_tools/kokoro/gcp_ubuntu/cmake/linux/x86-turing/build.sh
index 8363787..28751fe 100755
--- a/build_tools/kokoro/gcp_ubuntu/cmake/linux/x86-turing/build.sh
+++ b/build_tools/kokoro/gcp_ubuntu/cmake/linux/x86-turing/build.sh
@@ -33,6 +33,8 @@
# TODO(gcmn): It would be nice to be able to build and test as much as possible,
# so a build failure only prevents building/testing things that depend on it and
# we can still run the other tests.
+# TODO: Add "-DIREE_TARGET_BACKEND_CUDA=ON -DIREE_HAL_DRIVER_CUDA=ON" once the
+# VMs have been updated with the correct CUDA SDK.
echo "Building with cmake"
./build_tools/cmake/clean_build.sh
diff --git a/iree/compiler/Dialect/HAL/Target/CUDA/test/BUILD b/iree/compiler/Dialect/HAL/Target/CUDA/test/BUILD
deleted file mode 100644
index ac1f0e0..0000000
--- a/iree/compiler/Dialect/HAL/Target/CUDA/test/BUILD
+++ /dev/null
@@ -1,26 +0,0 @@
-# Copyright 2021 The IREE Authors
-#
-# Licensed under the Apache License v2.0 with LLVM Exceptions.
-# See https://llvm.org/LICENSE.txt for license information.
-# SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
-
-# load("//build_tools/bazel:iree_lit_test.bzl", "iree_lit_test_suite")
-# load("//build_tools/bazel:enforce_glob.bzl", "enforce_glob")
-
-package(
- default_visibility = ["//visibility:public"],
- features = ["layering_check"],
- licenses = ["notice"], # Apache 2.0
-)
-
-# iree_lit_test_suite(
-# name = "lit",
-# srcs = enforce_glob(
-# ["smoketest.mlir"],
-# include = ["*.mlir"],
-# ),
-# tools = [
-# "//iree/tools:iree-opt",
-# "@llvm-project//llvm:FileCheck",
-# ],
-# )
diff --git a/iree/compiler/Dialect/HAL/Target/CUDA/test/CMakeLists.txt b/iree/compiler/Dialect/HAL/Target/CUDA/test/CMakeLists.txt
index 0b6559b..31c1f25 100644
--- a/iree/compiler/Dialect/HAL/Target/CUDA/test/CMakeLists.txt
+++ b/iree/compiler/Dialect/HAL/Target/CUDA/test/CMakeLists.txt
@@ -1,13 +1,11 @@
-################################################################################
-# Autogenerated by build_tools/bazel_to_cmake/bazel_to_cmake.py from #
-# iree/compiler/Dialect/HAL/Target/CUDA/test/BUILD #
-# #
-# Use iree_cmake_extra_content from iree/build_defs.oss.bzl to add arbitrary #
-# CMake-only content. #
-# #
-# To disable autogeneration for this file entirely, delete this header. #
-################################################################################
-
iree_add_all_subdirs()
-### BAZEL_TO_CMAKE_PRESERVES_ALL_CONTENT_BELOW_THIS_LINE ###
+iree_lit_test_suite(
+ NAME
+ lit
+ SRCS
+ "smoketest.mlir"
+ TOOLS
+ FileCheck
+ iree::tools::iree-opt
+)
diff --git a/iree/hal/cuda/cts/CMakeLists.txt b/iree/hal/cuda/cts/CMakeLists.txt
index 72c2269..e2d6e72 100644
--- a/iree/hal/cuda/cts/CMakeLists.txt
+++ b/iree/hal/cuda/cts/CMakeLists.txt
@@ -4,27 +4,26 @@
# See https://llvm.org/LICENSE.txt for license information.
# SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
-# Temporarily disabled pending build system changes.
-# iree_hal_cts_test_suite(
-# DRIVER_NAME
-# cuda
-# DRIVER_REGISTRATION_HDR
-# "iree/hal/cuda/registration/driver_module.h"
-# DRIVER_REGISTRATION_FN
-# "iree_hal_cuda_driver_module_register"
-# COMPILER_TARGET_BACKEND
-# "cuda"
-# EXECUTABLE_FORMAT
-# "\"PTXE\""
-# DEPS
-# iree::hal::cuda::registration
-# EXCLUDED_TESTS
-# # This test depends on iree_hal_cuda_stream_command_buffer_update_buffer
-# # via iree_hal_buffer_view_allocate_buffer, which is not implemented yet.
-# "command_buffer_dispatch"
-# # Non-push descriptor sets are not implemented in the CUDA backend yet.
-# "descriptor_set"
-# # Semaphores are not implemented in the CUDA backend yet.
-# "semaphore_submission"
-# "semaphore"
-# )
+iree_hal_cts_test_suite(
+ DRIVER_NAME
+ cuda
+ DRIVER_REGISTRATION_HDR
+ "iree/hal/cuda/registration/driver_module.h"
+ DRIVER_REGISTRATION_FN
+ "iree_hal_cuda_driver_module_register"
+ COMPILER_TARGET_BACKEND
+ "cuda"
+ EXECUTABLE_FORMAT
+ "\"PTXE\""
+ DEPS
+ iree::hal::cuda::registration
+ EXCLUDED_TESTS
+ # This test depends on iree_hal_cuda_stream_command_buffer_update_buffer
+ # via iree_hal_buffer_view_allocate_buffer, which is not implemented yet.
+ "command_buffer_dispatch"
+ # Non-push descriptor sets are not implemented in the CUDA backend yet.
+ "descriptor_set"
+ # Semaphores are not implemented in the CUDA backend yet.
+ "semaphore_submission"
+ "semaphore"
+)
diff --git a/iree/samples/simple_embedding/BUILD b/iree/samples/simple_embedding/BUILD
index 4914245..21911ee 100644
--- a/iree/samples/simple_embedding/BUILD
+++ b/iree/samples/simple_embedding/BUILD
@@ -273,7 +273,8 @@
inline = True,
)
-# Temporarily disabled pending build system changes.
+# Disabled because CUDA is not universally available and Bazel does not
+# support configurability in a reasonable or useful way.
# iree_cmake_extra_content(
# content = """
# if(${IREE_HAL_DRIVER_CUDA} AND (${IREE_TARGET_BACKEND_CUDA} OR DEFINED IREE_HOST_BINARY_ROOT))
@@ -310,14 +311,13 @@
# ],
# )
-# Simple embedding is failing in the CI.
-# native_test(
-# name = "simple_embedding_cuda_test",
-# tags = [
-# "driver=cuda",
-# ],
-# src = ":simple_embedding_cuda",
-# )
+# # native_test(
+# # name = "simple_embedding_cuda_test",
+# # src = ":simple_embedding_cuda",
+# # tags = [
+# # "driver=cuda",
+# # ],
+# # )
# iree_cmake_extra_content(
# content = """
diff --git a/iree/test/e2e/cuda_specific/BUILD b/iree/test/e2e/cuda_specific/BUILD
deleted file mode 100644
index a067545..0000000
--- a/iree/test/e2e/cuda_specific/BUILD
+++ /dev/null
@@ -1,35 +0,0 @@
-# Copyright 2021 The IREE Authors
-#
-# Licensed under the Apache License v2.0 with LLVM Exceptions.
-# See https://llvm.org/LICENSE.txt for license information.
-# SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
-
-# Tests for end-to-end IREE support specific to the CUDA backend to be able to
-# incrementally enable features.
-
-# load("//build_tools/bazel:iree_check_test.bzl", "iree_check_single_backend_test_suite")
-
-package(
- default_visibility = ["//visibility:public"],
- features = ["layering_check"],
- licenses = ["notice"], # Apache 2.0
-)
-
-# Temporarily disabled pending build system changes.
-# iree_check_single_backend_test_suite(
-# name = "check_cuda",
-# srcs = [
-# "dot.mlir",
-# ],
-# compiler_flags = ["-iree-input-type=mhlo"],
-# driver = "cuda",
-# tags = [
-# # CUDA cuInit fails with sanitizer on.
-# "noasan",
-# "nomsan",
-# "notsan",
-# "noubsan",
-# "requires-gpu-nvidia",
-# ],
-# target_backend = "cuda",
-# )
diff --git a/iree/test/e2e/cuda_specific/CMakeLists.txt b/iree/test/e2e/cuda_specific/CMakeLists.txt
index ab173dc..4730d22 100644
--- a/iree/test/e2e/cuda_specific/CMakeLists.txt
+++ b/iree/test/e2e/cuda_specific/CMakeLists.txt
@@ -1,13 +1,29 @@
-################################################################################
-# Autogenerated by build_tools/bazel_to_cmake/bazel_to_cmake.py from #
-# iree/test/e2e/cuda_specific/BUILD #
-# #
-# Use iree_cmake_extra_content from iree/build_defs.oss.bzl to add arbitrary #
-# CMake-only content. #
-# #
-# To disable autogeneration for this file entirely, delete this header. #
-################################################################################
+# Copyright 2022 The IREE Authors
+#
+# Licensed under the Apache License v2.0 with LLVM Exceptions.
+# See https://llvm.org/LICENSE.txt for license information.
+# SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+
+# Tests for end-to-end IREE support specific to the CUDA backend to be able to
+# incrementally enable features.
iree_add_all_subdirs()
-### BAZEL_TO_CMAKE_PRESERVES_ALL_CONTENT_BELOW_THIS_LINE ###
+iree_check_single_backend_test_suite(
+ NAME
+ check_cuda
+ SRCS
+ "dot.mlir"
+ TARGET_BACKEND
+ "cuda"
+ DRIVER
+ "cuda"
+ COMPILER_FLAGS
+ "-iree-input-type=mhlo"
+ LABELS
+ "noasan"
+ "nomsan"
+ "notsan"
+ "noubsan"
+ "requires-gpu-nvidia"
+)
diff --git a/iree/test/e2e/linalg_ext_ops/BUILD b/iree/test/e2e/linalg_ext_ops/BUILD
index 343c8b1..824d951 100644
--- a/iree/test/e2e/linalg_ext_ops/BUILD
+++ b/iree/test/e2e/linalg_ext_ops/BUILD
@@ -13,30 +13,29 @@
licenses = ["notice"], # Apache 2.0
)
-# Temporarily disabled pending build system changes.
-# iree_check_single_backend_test_suite(
-# name = "check_cuda",
-# srcs = enforce_glob(
-# # keep sorted
-# [
-# "reverse.mlir",
-# "scan.mlir",
-# ],
-# include = ["*.mlir"],
-# exclude = [
-# ],
-# ),
-# driver = "cuda",
-# tags = [
-# # CUDA cuInit fails with sanitizer on.
-# "noasan",
-# "nomsan",
-# "notsan",
-# "noubsan",
-# "requires-gpu-nvidia",
-# ],
-# target_backend = "cuda",
-# )
+iree_check_single_backend_test_suite(
+ name = "check_cuda",
+ srcs = enforce_glob(
+ # keep sorted
+ [
+ "reverse.mlir",
+ "scan.mlir",
+ ],
+ include = ["*.mlir"],
+ exclude = [
+ ],
+ ),
+ driver = "cuda",
+ tags = [
+ # CUDA cuInit fails with sanitizer on.
+ "noasan",
+ "nomsan",
+ "notsan",
+ "noubsan",
+ "requires-gpu-nvidia",
+ ],
+ target_backend = "cuda",
+)
iree_check_single_backend_test_suite(
name = "check_dylib-llvm-aot_dylib",
diff --git a/iree/test/e2e/linalg_ext_ops/CMakeLists.txt b/iree/test/e2e/linalg_ext_ops/CMakeLists.txt
index 7aed204..3d9d628 100644
--- a/iree/test/e2e/linalg_ext_ops/CMakeLists.txt
+++ b/iree/test/e2e/linalg_ext_ops/CMakeLists.txt
@@ -12,6 +12,24 @@
iree_check_single_backend_test_suite(
NAME
+ check_cuda
+ SRCS
+ "reverse.mlir"
+ "scan.mlir"
+ TARGET_BACKEND
+ "cuda"
+ DRIVER
+ "cuda"
+ LABELS
+ "noasan"
+ "nomsan"
+ "notsan"
+ "noubsan"
+ "requires-gpu-nvidia"
+)
+
+iree_check_single_backend_test_suite(
+ NAME
check_dylib-llvm-aot_dylib
SRCS
"reverse.mlir"
diff --git a/iree/test/e2e/models/BUILD b/iree/test/e2e/models/BUILD
index d0b1107..ae84e50 100644
--- a/iree/test/e2e/models/BUILD
+++ b/iree/test/e2e/models/BUILD
@@ -74,20 +74,19 @@
target_backend = "vulkan-spirv",
)
-# Temporarily disabled pending build system changes.
-# iree_check_single_backend_test_suite(
-# name = "check_cuda_cuda",
-# timeout = "long",
-# srcs = CHECK_FRAMEWORK_TESTS,
-# compiler_flags = ["-iree-input-type=mhlo"],
-# driver = "cuda",
-# tags = [
-# # CUDA cuInit fails with sanitizer on.
-# "noasan",
-# "nomsan",
-# "notsan",
-# "noubsan",
-# "requires-gpu-nvidia",
-# ],
-# target_backend = "cuda",
-# )
+iree_check_single_backend_test_suite(
+ name = "check_cuda_cuda",
+ timeout = "long",
+ srcs = CHECK_FRAMEWORK_TESTS,
+ compiler_flags = ["-iree-input-type=mhlo"],
+ driver = "cuda",
+ tags = [
+ # CUDA cuInit fails with sanitizer on.
+ "noasan",
+ "nomsan",
+ "notsan",
+ "noubsan",
+ "requires-gpu-nvidia",
+ ],
+ target_backend = "cuda",
+)
diff --git a/iree/test/e2e/models/CMakeLists.txt b/iree/test/e2e/models/CMakeLists.txt
index a668981..a15eb09 100644
--- a/iree/test/e2e/models/CMakeLists.txt
+++ b/iree/test/e2e/models/CMakeLists.txt
@@ -58,4 +58,24 @@
"-iree-input-type=mhlo"
)
+iree_check_single_backend_test_suite(
+ NAME
+ check_cuda_cuda
+ SRCS
+ "bert_encoder_unrolled_fake_weights.mlir"
+ "mobilenetv3_fake_weights.mlir"
+ TARGET_BACKEND
+ "cuda"
+ DRIVER
+ "cuda"
+ COMPILER_FLAGS
+ "-iree-input-type=mhlo"
+ LABELS
+ "noasan"
+ "nomsan"
+ "notsan"
+ "noubsan"
+ "requires-gpu-nvidia"
+)
+
### BAZEL_TO_CMAKE_PRESERVES_ALL_CONTENT_BELOW_THIS_LINE ###
diff --git a/iree/test/e2e/regression/BUILD b/iree/test/e2e/regression/BUILD
index cc78f00..bc0e05b 100644
--- a/iree/test/e2e/regression/BUILD
+++ b/iree/test/e2e/regression/BUILD
@@ -91,22 +91,21 @@
target_backend = "vulkan-spirv",
)
-# Temporarily disabled pending build system changes.
-# iree_check_single_backend_test_suite(
-# name = "check_regression_cuda",
-# srcs = BACKEND_TESTS,
-# compiler_flags = ["-iree-input-type=mhlo"],
-# driver = "cuda",
-# tags = [
-# # CUDA cuInit fails with sanitizer on.
-# "noasan",
-# "nomsan",
-# "notsan",
-# "noubsan",
-# "requires-gpu-nvidia",
-# ],
-# target_backend = "cuda",
-# )
+iree_check_single_backend_test_suite(
+ name = "check_regression_cuda",
+ srcs = BACKEND_TESTS,
+ compiler_flags = ["-iree-input-type=mhlo"],
+ driver = "cuda",
+ tags = [
+ # CUDA cuInit fails with sanitizer on.
+ "noasan",
+ "nomsan",
+ "notsan",
+ "noubsan",
+ "requires-gpu-nvidia",
+ ],
+ target_backend = "cuda",
+)
py_binary(
name = "generate_e2e_matmul_tests",
diff --git a/iree/test/e2e/regression/CMakeLists.txt b/iree/test/e2e/regression/CMakeLists.txt
index 9942708..31cc3df 100644
--- a/iree/test/e2e/regression/CMakeLists.txt
+++ b/iree/test/e2e/regression/CMakeLists.txt
@@ -96,6 +96,34 @@
"-iree-input-type=mhlo"
)
+iree_check_single_backend_test_suite(
+ NAME
+ check_regression_cuda
+ SRCS
+ "dynamic_abs.mlir"
+ "dynamic_add.mlir"
+ "dynamic_dot.mlir"
+ "dynamic_reduce_min.mlir"
+ "dynamic_torch_index_select_high_rank.mlir"
+ "dynamic_torch_index_select_negative.mlir"
+ "dynamic_torch_index_select_scalar.mlir"
+ "dynamic_torch_index_select_vector.mlir"
+ "linalg_ext_ops.mlir"
+ "linalg_ops.mlir"
+ TARGET_BACKEND
+ "cuda"
+ DRIVER
+ "cuda"
+ COMPILER_FLAGS
+ "-iree-input-type=mhlo"
+ LABELS
+ "noasan"
+ "nomsan"
+ "notsan"
+ "noubsan"
+ "requires-gpu-nvidia"
+)
+
iree_generated_trace_runner_test(
NAME
e2e_matmul_direct_i8_small
diff --git a/iree/test/e2e/tensor_ops/BUILD b/iree/test/e2e/tensor_ops/BUILD
index ba43404..1c694b2 100644
--- a/iree/test/e2e/tensor_ops/BUILD
+++ b/iree/test/e2e/tensor_ops/BUILD
@@ -51,30 +51,29 @@
target_backend = "dylib-llvm-aot",
)
-# Temporarily disabled pending build system changes.
-# iree_check_single_backend_test_suite(
-# name = "check_cuda",
-# srcs = enforce_glob(
-# # keep sorted
-# [
-# "extract_slice.mlir",
-# "tensor_insert_slice.mlir",
-# ],
-# include = ["*.mlir"],
-# exclude = [
-# "tensor_cast.mlir",
-# ],
-# ),
-# driver = "cuda",
-# tags = [
-# "noasan",
-# "nomsan",
-# "notsan",
-# "noubsan",
-# "requires-gpu-nvidia",
-# ],
-# target_backend = "cuda",
-# )
+iree_check_single_backend_test_suite(
+ name = "check_cuda",
+ srcs = enforce_glob(
+ # keep sorted
+ [
+ "extract_slice.mlir",
+ "tensor_insert_slice.mlir",
+ ],
+ include = ["*.mlir"],
+ exclude = [
+ "tensor_cast.mlir",
+ ],
+ ),
+ driver = "cuda",
+ tags = [
+ "noasan",
+ "nomsan",
+ "notsan",
+ "noubsan",
+ "requires-gpu-nvidia",
+ ],
+ target_backend = "cuda",
+)
iree_check_single_backend_test_suite(
name = "check_vulkan-spirv_vulkan",
diff --git a/iree/test/e2e/tensor_ops/CMakeLists.txt b/iree/test/e2e/tensor_ops/CMakeLists.txt
index 279a5c4..2acc931 100644
--- a/iree/test/e2e/tensor_ops/CMakeLists.txt
+++ b/iree/test/e2e/tensor_ops/CMakeLists.txt
@@ -38,6 +38,24 @@
iree_check_single_backend_test_suite(
NAME
+ check_cuda
+ SRCS
+ "extract_slice.mlir"
+ "tensor_insert_slice.mlir"
+ TARGET_BACKEND
+ "cuda"
+ DRIVER
+ "cuda"
+ LABELS
+ "noasan"
+ "nomsan"
+ "notsan"
+ "noubsan"
+ "requires-gpu-nvidia"
+)
+
+iree_check_single_backend_test_suite(
+ NAME
check_vulkan-spirv_vulkan
SRCS
"extract_slice.mlir"
diff --git a/iree/test/e2e/xla_ops/BUILD b/iree/test/e2e/xla_ops/BUILD
index 50f0354..20e3b73 100644
--- a/iree/test/e2e/xla_ops/BUILD
+++ b/iree/test/e2e/xla_ops/BUILD
@@ -19,167 +19,166 @@
licenses = ["notice"], # Apache 2.0
)
-# Temporarily disabled pending build system changes.
-# iree_check_single_backend_test_suite(
-# name = "check_cuda_graph",
-# srcs = enforce_glob(
-# # keep sorted
-# [
-# "abs.mlir",
-# "add.mlir",
-# "batch_norm_inference.mlir",
-# "bitcast_convert.mlir",
-# "broadcast.mlir",
-# "broadcast_add.mlir",
-# "broadcast_in_dim.mlir",
-# "clamp.mlir",
-# "compare.mlir",
-# "concatenate.mlir",
-# "constant.mlir",
-# "convert.mlir",
-# "convolution.mlir",
-# "cosine.mlir",
-# "divide.mlir",
-# "dot.mlir",
-# "dot_general.mlir",
-# "dynamic_slice.mlir",
-# "dynamic_update_slice.mlir",
-# "exponential.mlir",
-# "exponential_fp16.mlir",
-# "exponential_minus_one.mlir",
-# "fft.mlir",
-# "finite.mlir",
-# "floor.mlir",
-# "gather.mlir",
-# "iota.mlir",
-# "log.mlir",
-# "log_plus_one.mlir",
-# "maximum.mlir",
-# "minimum.mlir",
-# "multiply.mlir",
-# "negate.mlir",
-# "pad.mlir",
-# "pow.mlir",
-# "reduce.mlir",
-# "reduce_window.mlir",
-# "remainder.mlir",
-# "reshape.mlir",
-# "reverse.mlir",
-# "rng_normal.mlir",
-# "rng_uniform.mlir",
-# "rsqrt.mlir",
-# "scatter.mlir",
-# "scatter_dynamic.mlir",
-# "select.mlir",
-# "sine.mlir",
-# "slice.mlir",
-# "sort.mlir",
-# "sqrt.mlir",
-# "subtract.mlir",
-# "tanh.mlir",
-# "torch_index_select.mlir",
-# "transpose.mlir",
-# "while.mlir",
-# ],
-# include = ["*.mlir"],
-# exclude = [
-# "round.mlir",
-# ],
-# ),
-# compiler_flags = ["-iree-input-type=mhlo"],
-# driver = "cuda",
-# runner_args = ["--cuda_use_streams=false"],
-# tags = [
-# # CUDA cuInit fails with sanitizer on.
-# "noasan",
-# "nomsan",
-# "notsan",
-# "noubsan",
-# "requires-gpu-nvidia",
-# ],
-# target_backend = "cuda",
-# )
+iree_check_single_backend_test_suite(
+ name = "check_cuda_graph",
+ srcs = enforce_glob(
+ # keep sorted
+ [
+ "abs.mlir",
+ "add.mlir",
+ "batch_norm_inference.mlir",
+ "bitcast_convert.mlir",
+ "broadcast.mlir",
+ "broadcast_add.mlir",
+ "broadcast_in_dim.mlir",
+ "clamp.mlir",
+ "compare.mlir",
+ "concatenate.mlir",
+ "constant.mlir",
+ "convert.mlir",
+ "convolution.mlir",
+ "cosine.mlir",
+ "divide.mlir",
+ "dot.mlir",
+ "dot_general.mlir",
+ "dynamic_slice.mlir",
+ "dynamic_update_slice.mlir",
+ "exponential.mlir",
+ "exponential_fp16.mlir",
+ "exponential_minus_one.mlir",
+ "fft.mlir",
+ "finite.mlir",
+ "floor.mlir",
+ "gather.mlir",
+ "iota.mlir",
+ "log.mlir",
+ "log_plus_one.mlir",
+ "maximum.mlir",
+ "minimum.mlir",
+ "multiply.mlir",
+ "negate.mlir",
+ "pad.mlir",
+ "pow.mlir",
+ "reduce.mlir",
+ "reduce_window.mlir",
+ "remainder.mlir",
+ "reshape.mlir",
+ "reverse.mlir",
+ "rng_normal.mlir",
+ "rng_uniform.mlir",
+ "rsqrt.mlir",
+ "scatter.mlir",
+ "scatter_dynamic.mlir",
+ "select.mlir",
+ "sine.mlir",
+ "slice.mlir",
+ "sort.mlir",
+ "sqrt.mlir",
+ "subtract.mlir",
+ "tanh.mlir",
+ "torch_index_select.mlir",
+ "transpose.mlir",
+ "while.mlir",
+ ],
+ include = ["*.mlir"],
+ exclude = [
+ "round.mlir",
+ ],
+ ),
+ compiler_flags = ["-iree-input-type=mhlo"],
+ driver = "cuda",
+ runner_args = ["--cuda_use_streams=false"],
+ tags = [
+ # CUDA cuInit fails with sanitizer on.
+ "noasan",
+ "nomsan",
+ "notsan",
+ "noubsan",
+ "requires-gpu-nvidia",
+ ],
+ target_backend = "cuda",
+)
# # Run cuda tests using stream command buffer
-# iree_check_single_backend_test_suite(
-# name = "check_cuda_streams",
-# srcs = enforce_glob(
-# # keep sorted
-# [
-# "abs.mlir",
-# "add.mlir",
-# "batch_norm_inference.mlir",
-# "bitcast_convert.mlir",
-# "broadcast.mlir",
-# "broadcast_add.mlir",
-# "broadcast_in_dim.mlir",
-# "clamp.mlir",
-# "compare.mlir",
-# "concatenate.mlir",
-# "constant.mlir",
-# "convert.mlir",
-# "convolution.mlir",
-# "cosine.mlir",
-# "divide.mlir",
-# "dot.mlir",
-# "dot_general.mlir",
-# "dynamic_slice.mlir",
-# "dynamic_update_slice.mlir",
-# "exponential.mlir",
-# "exponential_fp16.mlir",
-# "exponential_minus_one.mlir",
-# "fft.mlir",
-# "finite.mlir",
-# "floor.mlir",
-# "gather.mlir",
-# "iota.mlir",
-# "log.mlir",
-# "log_plus_one.mlir",
-# "maximum.mlir",
-# "minimum.mlir",
-# "multiply.mlir",
-# "negate.mlir",
-# "pad.mlir",
-# "pow.mlir",
-# "reduce.mlir",
-# "reduce_window.mlir",
-# "remainder.mlir",
-# "reshape.mlir",
-# "reverse.mlir",
-# "rng_normal.mlir",
-# "rng_uniform.mlir",
-# "rsqrt.mlir",
-# "scatter.mlir",
-# "scatter_dynamic.mlir",
-# "select.mlir",
-# "sine.mlir",
-# "slice.mlir",
-# "sort.mlir",
-# "sqrt.mlir",
-# "subtract.mlir",
-# "tanh.mlir",
-# "torch_index_select.mlir",
-# "transpose.mlir",
-# "while.mlir",
-# ],
-# include = ["*.mlir"],
-# exclude = [
-# "round.mlir",
-# ],
-# ),
-# compiler_flags = ["-iree-input-type=mhlo"],
-# driver = "cuda",
-# runner_args = ["--cuda_use_streams=true"],
-# tags = [
-# # CUDA cuInit fails with sanitizer on.
-# "noasan",
-# "nomsan",
-# "notsan",
-# "noubsan",
-# "requires-gpu-nvidia",
-# ],
-# target_backend = "cuda",
-# )
+iree_check_single_backend_test_suite(
+ name = "check_cuda_streams",
+ srcs = enforce_glob(
+ # keep sorted
+ [
+ "abs.mlir",
+ "add.mlir",
+ "batch_norm_inference.mlir",
+ "bitcast_convert.mlir",
+ "broadcast.mlir",
+ "broadcast_add.mlir",
+ "broadcast_in_dim.mlir",
+ "clamp.mlir",
+ "compare.mlir",
+ "concatenate.mlir",
+ "constant.mlir",
+ "convert.mlir",
+ "convolution.mlir",
+ "cosine.mlir",
+ "divide.mlir",
+ "dot.mlir",
+ "dot_general.mlir",
+ "dynamic_slice.mlir",
+ "dynamic_update_slice.mlir",
+ "exponential.mlir",
+ "exponential_fp16.mlir",
+ "exponential_minus_one.mlir",
+ "fft.mlir",
+ "finite.mlir",
+ "floor.mlir",
+ "gather.mlir",
+ "iota.mlir",
+ "log.mlir",
+ "log_plus_one.mlir",
+ "maximum.mlir",
+ "minimum.mlir",
+ "multiply.mlir",
+ "negate.mlir",
+ "pad.mlir",
+ "pow.mlir",
+ "reduce.mlir",
+ "reduce_window.mlir",
+ "remainder.mlir",
+ "reshape.mlir",
+ "reverse.mlir",
+ "rng_normal.mlir",
+ "rng_uniform.mlir",
+ "rsqrt.mlir",
+ "scatter.mlir",
+ "scatter_dynamic.mlir",
+ "select.mlir",
+ "sine.mlir",
+ "slice.mlir",
+ "sort.mlir",
+ "sqrt.mlir",
+ "subtract.mlir",
+ "tanh.mlir",
+ "torch_index_select.mlir",
+ "transpose.mlir",
+ "while.mlir",
+ ],
+ include = ["*.mlir"],
+ exclude = [
+ "round.mlir",
+ ],
+ ),
+ compiler_flags = ["-iree-input-type=mhlo"],
+ driver = "cuda",
+ runner_args = ["--cuda_use_streams=true"],
+ tags = [
+ # CUDA cuInit fails with sanitizer on.
+ "noasan",
+ "nomsan",
+ "notsan",
+ "noubsan",
+ "requires-gpu-nvidia",
+ ],
+ target_backend = "cuda",
+)
iree_check_single_backend_test_suite(
name = "check_dylib-llvm-aot_dylib",
diff --git a/iree/test/e2e/xla_ops/CMakeLists.txt b/iree/test/e2e/xla_ops/CMakeLists.txt
index 0b0086e..05236c9 100644
--- a/iree/test/e2e/xla_ops/CMakeLists.txt
+++ b/iree/test/e2e/xla_ops/CMakeLists.txt
@@ -12,6 +12,156 @@
iree_check_single_backend_test_suite(
NAME
+ check_cuda_graph
+ SRCS
+ "abs.mlir"
+ "add.mlir"
+ "batch_norm_inference.mlir"
+ "bitcast_convert.mlir"
+ "broadcast.mlir"
+ "broadcast_add.mlir"
+ "broadcast_in_dim.mlir"
+ "clamp.mlir"
+ "compare.mlir"
+ "concatenate.mlir"
+ "constant.mlir"
+ "convert.mlir"
+ "convolution.mlir"
+ "cosine.mlir"
+ "divide.mlir"
+ "dot.mlir"
+ "dot_general.mlir"
+ "dynamic_slice.mlir"
+ "dynamic_update_slice.mlir"
+ "exponential.mlir"
+ "exponential_fp16.mlir"
+ "exponential_minus_one.mlir"
+ "fft.mlir"
+ "finite.mlir"
+ "floor.mlir"
+ "gather.mlir"
+ "iota.mlir"
+ "log.mlir"
+ "log_plus_one.mlir"
+ "maximum.mlir"
+ "minimum.mlir"
+ "multiply.mlir"
+ "negate.mlir"
+ "pad.mlir"
+ "pow.mlir"
+ "reduce.mlir"
+ "reduce_window.mlir"
+ "remainder.mlir"
+ "reshape.mlir"
+ "reverse.mlir"
+ "rng_normal.mlir"
+ "rng_uniform.mlir"
+ "rsqrt.mlir"
+ "scatter.mlir"
+ "scatter_dynamic.mlir"
+ "select.mlir"
+ "sine.mlir"
+ "slice.mlir"
+ "sort.mlir"
+ "sqrt.mlir"
+ "subtract.mlir"
+ "tanh.mlir"
+ "torch_index_select.mlir"
+ "transpose.mlir"
+ "while.mlir"
+ TARGET_BACKEND
+ "cuda"
+ DRIVER
+ "cuda"
+ COMPILER_FLAGS
+ "-iree-input-type=mhlo"
+ RUNNER_ARGS
+ "--cuda_use_streams=false"
+ LABELS
+ "noasan"
+ "nomsan"
+ "notsan"
+ "noubsan"
+ "requires-gpu-nvidia"
+)
+
+iree_check_single_backend_test_suite(
+ NAME
+ check_cuda_streams
+ SRCS
+ "abs.mlir"
+ "add.mlir"
+ "batch_norm_inference.mlir"
+ "bitcast_convert.mlir"
+ "broadcast.mlir"
+ "broadcast_add.mlir"
+ "broadcast_in_dim.mlir"
+ "clamp.mlir"
+ "compare.mlir"
+ "concatenate.mlir"
+ "constant.mlir"
+ "convert.mlir"
+ "convolution.mlir"
+ "cosine.mlir"
+ "divide.mlir"
+ "dot.mlir"
+ "dot_general.mlir"
+ "dynamic_slice.mlir"
+ "dynamic_update_slice.mlir"
+ "exponential.mlir"
+ "exponential_fp16.mlir"
+ "exponential_minus_one.mlir"
+ "fft.mlir"
+ "finite.mlir"
+ "floor.mlir"
+ "gather.mlir"
+ "iota.mlir"
+ "log.mlir"
+ "log_plus_one.mlir"
+ "maximum.mlir"
+ "minimum.mlir"
+ "multiply.mlir"
+ "negate.mlir"
+ "pad.mlir"
+ "pow.mlir"
+ "reduce.mlir"
+ "reduce_window.mlir"
+ "remainder.mlir"
+ "reshape.mlir"
+ "reverse.mlir"
+ "rng_normal.mlir"
+ "rng_uniform.mlir"
+ "rsqrt.mlir"
+ "scatter.mlir"
+ "scatter_dynamic.mlir"
+ "select.mlir"
+ "sine.mlir"
+ "slice.mlir"
+ "sort.mlir"
+ "sqrt.mlir"
+ "subtract.mlir"
+ "tanh.mlir"
+ "torch_index_select.mlir"
+ "transpose.mlir"
+ "while.mlir"
+ TARGET_BACKEND
+ "cuda"
+ DRIVER
+ "cuda"
+ COMPILER_FLAGS
+ "-iree-input-type=mhlo"
+ RUNNER_ARGS
+ "--cuda_use_streams=true"
+ LABELS
+ "noasan"
+ "nomsan"
+ "notsan"
+ "noubsan"
+ "requires-gpu-nvidia"
+)
+
+iree_check_single_backend_test_suite(
+ NAME
check_dylib-llvm-aot_dylib
SRCS
"abs.mlir"
diff --git a/iree/tools/CMakeLists.txt b/iree/tools/CMakeLists.txt
index 31046df..833970d 100644
--- a/iree/tools/CMakeLists.txt
+++ b/iree/tools/CMakeLists.txt
@@ -19,31 +19,31 @@
# Enable compiler targets based on options.
set(IREE_COMPILER_TARGETS "")
set(IREE_COMPILER_TARGET_COPTS "")
-if("${IREE_TARGET_BACKEND_DYLIB_LLVM_AOT}" OR "${IREE_TARGET_BACKEND_WASM_LLVM_AOT}")
+if(IREE_TARGET_BACKEND_DYLIB_LLVM_AOT OR IREE_TARGET_BACKEND_WASM_LLVM_AOT)
list(APPEND IREE_COMPILER_TARGETS iree::compiler::Dialect::HAL::Target::LLVM)
list(APPEND IREE_COMPILER_TARGET_COPTS "-DIREE_HAVE_LLVMAOT_TARGET")
endif()
-if("${IREE_TARGET_BACKEND_METAL_SPIRV}")
+if(IREE_TARGET_BACKEND_METAL_SPIRV)
list(APPEND IREE_COMPILER_TARGETS iree::compiler::Dialect::HAL::Target::MetalSPIRV)
list(APPEND IREE_COMPILER_TARGET_COPTS "-DIREE_HAVE_METALSPIRV_TARGET")
endif()
-if("${IREE_TARGET_BACKEND_VMVX}")
+if(IREE_TARGET_BACKEND_VMVX)
list(APPEND IREE_COMPILER_TARGETS iree::compiler::Dialect::HAL::Target::VMVX)
list(APPEND IREE_COMPILER_TARGET_COPTS "-DIREE_HAVE_VMVX_TARGET")
endif()
-if("${IREE_TARGET_BACKEND_VULKAN_SPIRV}")
+if(IREE_TARGET_BACKEND_VULKAN_SPIRV)
list(APPEND IREE_COMPILER_TARGETS iree::compiler::Dialect::HAL::Target::VulkanSPIRV)
list(APPEND IREE_COMPILER_TARGET_COPTS "-DIREE_HAVE_VULKANSPIRV_TARGET")
endif()
-if("${IREE_TARGET_BACKEND_WEBGPU}")
+if(IREE_TARGET_BACKEND_WEBGPU)
list(APPEND IREE_COMPILER_TARGETS iree::compiler::Dialect::HAL::Target::WebGPU)
list(APPEND IREE_COMPILER_TARGET_COPTS "-DIREE_HAVE_WEBGPU_TARGET")
endif()
-# if("${IREE_TARGET_BACKEND_CUDA}")
-# list(APPEND IREE_COMPILER_TARGETS iree::compiler::Dialect::HAL::Target::CUDA)
-# list(APPEND IREE_COMPILER_TARGET_COPTS "-DIREE_HAVE_CUDA_TARGET")
-# endif()
-if("${IREE_TARGET_BACKEND_ROCM}")
+if(IREE_TARGET_BACKEND_CUDA)
+ list(APPEND IREE_COMPILER_TARGETS iree::compiler::Dialect::HAL::Target::CUDA)
+ list(APPEND IREE_COMPILER_TARGET_COPTS "-DIREE_HAVE_CUDA_TARGET")
+endif()
+if(IREE_TARGET_BACKEND_ROCM)
list(APPEND IREE_COMPILER_TARGETS iree::compiler::Dialect::HAL::Target::ROCM)
list(APPEND IREE_COMPILER_TARGET_COPTS "-DIREE_HAVE_ROCM_TARGET")
endif()
