Unskip more UTs (#3103)

Before:
```
language: passed: 11933, failed: 0, skipped: 104, xfailed: 453, total: 12490, fixme: 0, pass rate (w/o xfailed): 99.14%
tools: passed: 8, failed: 0, skipped: 1, xfailed: 0, total: 9, fixme: 0, pass rate (w/o xfailed): 88.89%
all: passed: 18605, failed: 0, skipped: 126, xfailed: 1215, total: 19946, fixme: 48, pass rate (w/o xfailed): 99.33%
```

After:
```
language: passed: 11934, failed: 0, skipped: 103, xfailed: 453, total: 12490, fixme: 0, pass rate (w/o xfailed): 99.14%
tools: passed: 8, failed: 0, skipped: 0, xfailed: 0, total: 8, fixme: 0, pass rate (w/o xfailed): 100.0%
all: passed: 18606, failed: 0, skipped: 124, xfailed: 1215, total: 19945, fixme: 48, pass rate (w/o xfailed): 99.34%
```

Signed-off-by: Whitney Tsang <[email protected]>

Author: whitneywhtsang

python/test/unit/language/test_core.py

@@ -3779,9 +3779,8 @@ def test_dot3d(B, num_warps, M, N, K, BLOCK_M, BLOCK_N, in_dtype_str, out_dtype_
             pytest.skip("small dots are supported only on HIP at the moment")
 
     if B == 8 and M == 64 and in_dtype_str == "float32" and out_dtype_str == "float32":
-        if not is_interpreter() and torch.cuda.is_available(
-        ) and triton.runtime.driver.active.utils.get_device_properties(
-                torch.cuda.current_device())["max_shared_mem"] < 131072:
+        if not is_interpreter() and triton.runtime.driver.active.utils.get_device_properties(
+                triton.runtime.driver.active.get_current_device())["max_shared_mem"] < 131072:
             pytest.skip(
                 "Skipping tests with B = 8, M = 64, in_type = float32, out_type = float32 due to insufficient shared memory (less than 128 KB per SM) on this GPU."
             )

python/test/unit/language/test_pipeliner.py

@@ -6,7 +6,7 @@
 import triton.language as tl
 import triton.tools.experimental_descriptor
 
-from triton._internal_testing import is_cuda, is_hopper, is_hip_cdna, is_hip_mi200
+from triton._internal_testing import is_cuda, is_hopper, is_hip_cdna, is_hip_mi200, is_xpu
 
 
 def check_capabilities():
@@ -214,8 +214,8 @@ def __exit__(self, exc_type, exc_val, exc_tb):
 @pytest.mark.parametrize("scale", [True, False])
 def test_pipeline_matmul(scale, device):
     check_capabilities()
-    if scale and not (is_cuda() or is_hip_cdna()):
-        pytest.skip("NYI: scale_dot just implemented in CUDA/HIP")
+    if scale and not (is_cuda() or is_hip_cdna() or is_xpu()):
+        pytest.skip("NYI: scale_dot just implemented in CUDA/HIP/XPU")
     M, N, K = 512, 512, 128
     BLOCK_M, BLOCK_N, BLOCK_K = 64, 64, 32
     NUM_STAGES = 4

scripts/skiplist/lts/language.txt

@@ -1585,3 +1585,4 @@ test/unit/language/test_core.py::test_dot[1-128-128-64-4-True-True-none-tf32-flo
 test/unit/language/test_core.py::test_dot[1-128-128-64-4-True-True-none-tf32-float16-float16-1_1]
 test/unit/language/test_core.py::test_dot[1-128-128-64-4-True-False-none-tf32-float16-float16-1_0]
 test/unit/language/test_core.py::test_dot[1-128-128-64-4-True-False-none-tf32-float16-float16-1_1]
+test/unit/language/test_pipeliner.py::test_pipeline_matmul[True]

scripts/test-triton.sh

@@ -202,7 +202,7 @@ run_core_tests() {
     pytest -k "not test_line_info_interpreter" --verbose --device xpu language/test_line_info.py
 
   TRITON_DISABLE_LINE_INFO=1 TRITON_TEST_SUITE=tools \
-    pytest --verbose tools/
+    pytest -k "not test_disam_cubin" --verbose tools
 
   cd $TRITON_PROJ/third_party/intel/python/test
   TRITON_DISABLE_LINE_INFO=1 TRITON_TEST_SUITE=third_party \