[CI][E2E] CPU AOT tests fail to build on windows CI machines due to having no listed CPU devices

[ayylol]

Describe the bug

The following tests are failing due to not having a listed cpu device on windows build ci machine:

SYCL :: AOT/cpu.cpp
SYCL :: NewOffloadDriver/cpu.cpp
SYCL :: SpecConstants/2020/non_native/multiple-targets.cpp
SYCL :: OptionalKernelFeatures/is_compatible/is_compatible_spir64_x86_64.cpp
SYCL :: AOT/multiple-devices.cpp
SYCL :: AOT/reqd-sg-size.cpp
SYCL :: ProgramManager/uneven_kernel_split.cpp

With errors like the following:

********************
FAIL: SYCL :: AOT/cpu.cpp (104 of 2303)
******************** TEST 'SYCL :: AOT/cpu.cpp' FAILED ********************
Exit Code: 2147482647

Command Output (stdout):
--
# RUN: at line 16
D:/github/actions-runner/_work/llvm/llvm/install/bin/clang++.exe  -Werror -fsycl -fsycl-targets=spir64_x86_64  D:\github\actions-runner\_work\llvm\llvm\llvm\sycl\test-e2e\AOT/Inputs/aot.cpp -Xsycl-target-backend "--bo=-g" -Xsycl-target-backend "--bo=-cl-opt-disable" -o D:\github\actions-runner\_work\llvm\llvm\build-e2e\AOT\Output\cpu.cpp.tmp2.out
# executed command: D:/github/actions-runner/_work/llvm/llvm/install/bin/clang++.exe -Werror -fsycl -fsycl-targets=spir64_x86_64 'D:\github\actions-runner\_work\llvm\llvm\llvm\sycl\test-e2e\AOT/Inputs/aot.cpp' -Xsycl-target-backend --bo=-g -Xsycl-target-backend --bo=-cl-opt-disable -o 'D:\github\actions-runner\_work\llvm\llvm\build-e2e\AOT\Output\cpu.cpp.tmp2.out'
# .---command stderr------------
# | Failed to retrieve OpenCL platform count: -1001 (Unknown error code)
# | llvm-foreach: 
# | clang++: error: x86_64 compiler command failed with exit code 2147482647 (use -v to see invocation)
# `-----------------------------
# error: command failed with exit status: 0x7ffffc17

https://github.com/intel/llvm/actions/runs/13929259216/job/38985533716?pr=17335
https://github.com/intel/llvm/actions/runs/13950218284/job/39050385825?pr=17335

To reproduce

1. Include a code snippet that is as short as possible
2. Specify the command which should be used to compile the program
3. Specify the command which should be used to launch the program
4. Indicate what is wrong and what was expected

Environment

• OS: [e.g Windows/Linux]
• Target device and vendor: [e.g. Intel GPU]
• DPC++ version: [e.g. commit hash or output of clang++ --version]
• Dependencies version: [e.g. the output of sycl-ls --verbose]

Additional context

No response

[KornevNikita]

also, what about this:
clang++: error: cannot find libdevice for sm_50; provide path to different CUDA installation via '--cuda-path', or pass '-nocudalib' to build without linking with libdevice
It passes on linux with no nvidia-gpu, can this also be adjusted on win?

same with compiling for amd:
clang++: error: cannot find ROCm device library; provide its path via '--rocm-path' or '--rocm-device-lib-path', or pass '-nogpulib' to build without ROCm device library

[sarnex]

im gonna install rocm, opencl cpu and cuda on all windows runners

[sarnex]

fyi its prolly gonna take me a while, i'm hitting numerous bugs not even in our software :)

[sarnex]

@KornevNikita @ayylol Ok I believe it's finally done, every Windows runner should have the OpenCL CPU runtime installed, as well as the CUDA and HIP SDKs, and the latter can be checked with the cuda_dev_kit and hip_dev_kit LIT features. You'll need the latest HEAD commit though because I fixed the lit detection literally right before posting this comment.

Can you guys test it and see if it's working as expected?

[KornevNikita]

If the pre-commit passes then I have no concerns - #17809

[ayylol]

@KornevNikita @ayylol Ok I believe it's finally done, every Windows runner should have the OpenCL CPU runtime installed, as well as the CUDA and HIP SDKs, and the latter can be checked with the cuda_dev_kit and hip_dev_kit LIT features. You'll need the latest HEAD commit though because I fixed the lit detection literally right before posting this comment.

Can you guys test it and see if it's working as expected?

@sarnex Looks like the CPU AOT tests are working now, but I see the tests that required opencl_icd are failing to build with this message:

FAIL: SYCL :: Adapters/interop-opencl-make-kernel-bundle.cpp (6 of 2328)
******************** TEST 'SYCL :: Adapters/interop-opencl-make-kernel-bundle.cpp' FAILED ********************
Exit Code: 1

Command Output (stdout):
--
# RUN: at line 3
D:/github/_work/llvm/llvm/install/bin/clang++.exe  -Werror -fsycl -fsycl-targets=spir64  D:\github\_work\llvm\llvm\llvm\sycl\test-e2e\Adapters\interop-opencl-make-kernel-bundle.cpp -o D:\github\_work\llvm\llvm\build-e2e\Adapters\Output\interop-opencl-make-kernel-bundle.cpp.tmp.out -LC:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v12.8/lib/x64 -lOpenCL
# executed command: D:/github/_work/llvm/llvm/install/bin/clang++.exe -Werror -fsycl -fsycl-targets=spir64 'D:\github\_work\llvm\llvm\llvm\sycl\test-e2e\Adapters\interop-opencl-make-kernel-bundle.cpp' -o 'D:\github\_work\llvm\llvm\build-e2e\Adapters\Output\interop-opencl-make-kernel-bundle.cpp.tmp.out' -LC:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v12.8/lib/x64 -lOpenCL
# .---command stderr------------
# | clang++: error: no such file or directory: 'Files/NVIDIA'
# | clang++: error: no such file or directory: 'GPU'
# | clang++: error: no such file or directory: 'Computing'
# | clang++: error: no such file or directory: 'Toolkit/CUDA/v12.8/lib/x64'
# `-----------------------------
# error: command failed with exit status: 1

--

From the message im guessing its related?
https://github.com/intel/llvm/actions/runs/14222943143/job/39857723374?pr=17335

[sarnex]

thanks let me investigate, need to add quotes somewhere

[sarnex]

@ayylol can you try this patch

diff --git a/sycl/test-e2e/lit.cfg.py b/sycl/test-e2e/lit.cfg.py
index 2dab110b3877..6f98a84be791 100644
--- a/sycl/test-e2e/lit.cfg.py
+++ b/sycl/test-e2e/lit.cfg.py
@@ -712,6 +712,9 @@ if "cuda:gpu" in config.sycl_devices:
         config.cuda_libs_dir = os.path.join(os.environ["CUDA_PATH"], r"lib64")
         config.cuda_include = os.path.join(os.environ["CUDA_PATH"], "include")
 
+config.cuda_libs_dir = quote_path(config.cuda_libs_dir)
+config.cuda_include = quote_path(config.cuda_include)
+
 config.substitutions.append(("%threads_lib", config.sycl_threads_lib))
 
 if lit_config.params.get("ze_debug"):

thx

[ayylol]

@ayylol can you try this patch

diff --git a/sycl/test-e2e/lit.cfg.py b/sycl/test-e2e/lit.cfg.py
index 2dab110b3877..6f98a84be791 100644
--- a/sycl/test-e2e/lit.cfg.py
+++ b/sycl/test-e2e/lit.cfg.py
@@ -712,6 +712,9 @@ if "cuda:gpu" in config.sycl_devices:
         config.cuda_libs_dir = os.path.join(os.environ["CUDA_PATH"], r"lib64")
         config.cuda_include = os.path.join(os.environ["CUDA_PATH"], "include")
 
+config.cuda_libs_dir = quote_path(config.cuda_libs_dir)
+config.cuda_include = quote_path(config.cuda_include)
+
 config.substitutions.append(("%threads_lib", config.sycl_threads_lib))
 
 if lit_config.params.get("ze_debug"):

thx

Ok, added that as a commit on top of #17335 https://github.com/intel/llvm/actions/runs/14224919991

[sarnex]

didnt work, ill investigate more

[sarnex]

@ayylol ok can you revert the previous one and try this, it might just actually work

diff --git a/sycl/test-e2e/lit.cfg.py b/sycl/test-e2e/lit.cfg.py
index 2dab110b3877..222ae37208b5 100644
--- a/sycl/test-e2e/lit.cfg.py
+++ b/sycl/test-e2e/lit.cfg.py
@@ -511,6 +511,8 @@ with test_env():
 
 # Check for OpenCL ICD
 if config.opencl_libs_dir:
+    config.opencl_libs_dir = quote_path(config.opencl_libs_dir)
+    config.opencl_include_dir = quote_path(config.opencl_include_dir)
     if cl_options:
         config.substitutions.append(
             ("%opencl_lib", " " + config.opencl_libs_dir + "/OpenCL.lib")

[ayylol]

@ayylol ok can you revert the previous one and try this, it might just actually work

diff --git a/sycl/test-e2e/lit.cfg.py b/sycl/test-e2e/lit.cfg.py
index 2dab110b3877..222ae37208b5 100644
--- a/sycl/test-e2e/lit.cfg.py
+++ b/sycl/test-e2e/lit.cfg.py
@@ -511,6 +511,8 @@ with test_env():
 
 # Check for OpenCL ICD
 if config.opencl_libs_dir:
+    config.opencl_libs_dir = quote_path(config.opencl_libs_dir)
+    config.opencl_include_dir = quote_path(config.opencl_include_dir)
     if cl_options:
         config.substitutions.append(
             ("%opencl_lib", " " + config.opencl_libs_dir + "/OpenCL.lib")

Ok, its running here https://github.com/intel/llvm/actions/runs/14227019977