[SYCL][E2E][LIT] Normalize devices not to include "ext_oneapi_"

[aelovikov-intel]

Also see:
#12007 (comment) #12048

[steffenlarsen]

LGTM! 👍

[github-actions]

⚠️ Python code formatter, darker found issues in your code. ⚠️

You can test this locally with the following command:

darker --check --diff -r 66a741bc9d0056e7b350ceed2a01b8531b156af6...e7f9d67d16e9efd0103bacffa25cb1844338eb1c sycl/test-e2e/format.py sycl/test-e2e/lit.cfg.py

View the diff from darker here.

--- format.py	2023-12-18 20:01:36.000000 +0000
+++ format.py	2023-12-18 22:23:20.193457 +0000
@@ -10,15 +10,15 @@
 
 import os
 import re
 
 def get_triple(test, backend):
-    if backend == 'cuda':
-        return 'nvptx64-nvidia-cuda'
-    if backend == 'hip':
-        if test.config.hip_platform == 'NVIDIA':
-            return 'nvptx64-nvidia-cuda'
+    if backend == "cuda":
+        return "nvptx64-nvidia-cuda"
+    if backend == "hip":
+        if test.config.hip_platform == "NVIDIA":
+            return "nvptx64-nvidia-cuda"
         else:
             return 'amdgcn-amd-amdhsa'
     if backend == 'native_cpu':
         return 'native_cpu'
     return 'spir64'
@@ -162,18 +162,22 @@
             # current llvm-lit invocation isn't configured to include it. We
             # don't use ONEAPI_DEVICE_SELECTOR for `%{run-unfiltered-devices}`
             # so that device might still be accessible to some of the tests yet
             # we won't set the environment variable below for such scenario.
             extra_env = []
-            if 'level_zero:gpu' in sycl_devices and litConfig.params.get('ur_l0_debug'):
-                extra_env.append('UR_L0_DEBUG={}'.format(test.config.ur_l0_debug))
-
-            if 'level_zero:gpu' in sycl_devices and litConfig.params.get('ur_l0_leaks_debug'):
-                extra_env.append('UR_L0_LEAKS_DEBUG={}'.format(test.config.ur_l0_leaks_debug))
-
-            if 'cuda:gpu' in sycl_devices:
-                extra_env.append('SYCL_PI_CUDA_ENABLE_IMAGE_SUPPORT=1')
+            if "level_zero:gpu" in sycl_devices and litConfig.params.get("ur_l0_debug"):
+                extra_env.append("UR_L0_DEBUG={}".format(test.config.ur_l0_debug))
+
+            if "level_zero:gpu" in sycl_devices and litConfig.params.get(
+                "ur_l0_leaks_debug"
+            ):
+                extra_env.append(
+                    "UR_L0_LEAKS_DEBUG={}".format(test.config.ur_l0_leaks_debug)
+                )
+
+            if "cuda:gpu" in sycl_devices:
+                extra_env.append("SYCL_PI_CUDA_ENABLE_IMAGE_SUPPORT=1")
 
             return extra_env
 
         extra_env = get_extra_env(devices_for_test)
 
--- lit.cfg.py	2023-12-15 00:20:07.000000 +0000
+++ lit.cfg.py	2023-12-18 22:23:20.412351 +0000
@@ -303,17 +303,19 @@
     config.sycl_devices = list(devices)
 
 if len(config.sycl_devices) > 1:
     lit_config.note('Running on multiple devices, XFAIL-marked tests will be skipped on corresponding devices')
 
-config.sycl_devices = [x.replace('ext_oneapi_', '') for x in config.sycl_devices]
-
-available_devices = {'opencl': ('cpu', 'gpu', 'acc'),
-                     'cuda':('gpu'),
-                     'level_zero':('gpu'),
-                     'hip':('gpu'),
-                     'native_cpu':('cpu')}
+config.sycl_devices = [x.replace("ext_oneapi_", "") for x in config.sycl_devices]
+
+available_devices = {
+    "opencl": ("cpu", "gpu", "acc"),
+    "cuda": ("gpu"),
+    "level_zero": ("gpu"),
+    "hip": ("gpu"),
+    "native_cpu": ("cpu"),
+}
 for d in config.sycl_devices:
      be, dev = d.split(':')
      if be not in available_devices or dev not in available_devices[be]:
           lit_config.error('Unsupported device {}'.format(d))
 
@@ -426,23 +428,23 @@
 
 for sycl_device in config.sycl_devices:
     be, dev = sycl_device.split(':')
     config.available_features.add('any-device-is-' + dev)
     # Use short names for LIT rules.
-    config.available_features.add('any-device-is-' + be)
+    config.available_features.add("any-device-is-" + be)
 
 # That has to be executed last so that all device-independent features have been
 # discovered already.
 config.sycl_dev_features = {}
 
 # Version of the driver for a given device. Empty for non-Intel devices.
 config.intel_driver_ver = {}
 for sycl_device in config.sycl_devices:
     env = copy.copy(llvm_config.config.environment)
-    env['ONEAPI_DEVICE_SELECTOR'] = sycl_device
-    if sycl_device.startswith('cuda:'):
-        env['SYCL_PI_CUDA_ENABLE_IMAGE_SUPPORT'] = '1'
+    env["ONEAPI_DEVICE_SELECTOR"] = sycl_device
+    if sycl_device.startswith("cuda:"):
+        env["SYCL_PI_CUDA_ENABLE_IMAGE_SUPPORT"] = "1"
     # When using the ONEAPI_DEVICE_SELECTOR environment variable, sycl-ls
     # prints warnings that might derail a user thinking something is wrong
     # with their test run. It's just us filtering here, so silence them unless
     # we get an exit status.
     try:

[sommerlukas]

Changes to the kernel fusion test LGTM.

[aelovikov-intel]

2Graph team: I'd suggest to modify your RUN lines to use FileCheck %s --implicit-check-not=LEAK and drop line 6 here and in other tests.

[EwanC]

Thanks for the tip, we can do this as a follow-up PR

[aelovikov-intel]

@intel/dpcpp-tools-reviewers , ping.