Merge update_cuda_build into update_amd_build

Author: vlad-perevezentsev

CMakeLists.txt

@@ -28,7 +28,9 @@ option(DPCTL_GENERATE_COVERAGE_FOR_PYBIND11_EXTENSIONS
 set(DPCTL_TARGET_CUDA
     ""
     CACHE STRING
-    "Build DPCTL to target CUDA devices. Set to ON to use default architecture (sm_50), or to a specific architecture like sm_80."
+    "Build DPCTL to target CUDA device. "
+    "Set to a truthy value (e.g., ON, TRUE) to use default architecture (sm_50), "
+    "or to a specific architecture like sm_80."
 )
 set(DPCTL_TARGET_HIP
     ""
@@ -55,14 +57,14 @@ set(_dpctl_sycl_targets)
 set(_dpctl_cuda_arch)
 if ("x${DPCTL_SYCL_TARGETS}" STREQUAL "x")
    if (NOT "x${DPCTL_TARGET_CUDA}" STREQUAL "x")
-      if (DPCTL_TARGET_CUDA STREQUAL "ON")
-         set(_dpctl_cuda_arch "sm_50")
-      elseif(DPCTL_TARGET_CUDA MATCHES "^sm_")
+      if(DPCTL_TARGET_CUDA MATCHES "^sm_")
          set(_dpctl_cuda_arch ${DPCTL_TARGET_CUDA})
+      elseif(DPCTL_TARGET_CUDA MATCHES "^(ON|TRUE|YES|Y|1)$")
+         set(_dpctl_cuda_arch "sm_50")
       else()
          message(FATAL_ERROR
             "Invalid value for DPCTL_TARGET_CUDA: \"${DPCTL_TARGET_CUDA}\". "
-            "Expected 'ON' or an architecture like 'sm_80'."
+            "Expected 'ON', 'TRUE', 'YES', 'Y', '1', or a CUDA architecture like 'sm_80'."
          )
       endif()
       set(_dpctl_sycl_targets "nvidia_gpu_${_dpctl_cuda_arch},spir64-unknown-unknown")

docs/doc_sources/beginners_guides/installation.rst

@@ -166,7 +166,8 @@ which accepts a specific compute architecture string:
 
     python scripts/build_locally.py --verbose --cmake-opts="-DDPCTL_TARGET_CUDA=sm_80"
 
-To use the default architecture (``sm_50``), use:
+To use the default architecture (``sm_50``),
+set ``DPCTL_TARGET_CUDA`` to a value such as ``ON``, ``TRUE``, ``YES``, ``Y``, or ``1``:
 
 .. code-block:: bash
 
@@ -175,10 +176,11 @@ To use the default architecture (``sm_50``), use:
 Note that kernels are built for ``sm_50`` by default, allowing them to work on a wider
 range of architectures, but limiting the usage of more recent CUDA features.
 
-For reference, compute architecture strings like ``sm_80`` are based on
-CUDA Compute Capability. A complete mapping between NVIDIA GPU models and their
-respective ``sm_XX`` values can be found in the official
-`CUDA GPU Compute Capability <https://developer.nvidia.com/cuda-gpus>`_.
+For reference, compute architecture strings like ``sm_80`` correspond to specific
+CUDA Compute Capabilities (e.g., Compute Capability 8.0 corresponds to ``sm_80``).
+A complete mapping between NVIDIA GPU models and their respective
+Compute Capabilities can be found in the official
+`CUDA GPU Compute Capability <https://developer.nvidia.com/cuda-gpus>`_ documentation.
 
 A full list of available SYCL alias targets is available in the
 `DPC++ Compiler User Manual <https://intel.github.io/llvm/UsersManual.html>`_.

dpctl/tensor/_ctors.py

@@ -111,8 +111,6 @@ def _asarray_from_usm_ndarray(
         raise TypeError(
             f"Expected dpctl.tensor.usm_ndarray, got {type(usm_ndary)}"
         )
-    if dtype is None:
-        dtype = usm_ndary.dtype
     if usm_type is None:
         usm_type = usm_ndary.usm_type
     if sycl_queue is not None:
@@ -122,6 +120,8 @@ def _asarray_from_usm_ndarray(
         copy_q = normalize_queue_device(sycl_queue=sycl_queue, device=exec_q)
     else:
         copy_q = usm_ndary.sycl_queue
+    if dtype is None:
+        dtype = _map_to_device_dtype(usm_ndary.dtype, copy_q)
     # Conditions for zero copy:
     can_zero_copy = copy is not True
     #    dtype is unchanged

dpctl/tests/test_tensor_asarray.py

@@ -623,3 +623,27 @@ def test_asarray_support_for_usm_ndarray_protocol(usm_type):
     assert x.dtype == y3.dtype
     assert y3.usm_data.reference_obj is None
     assert dpt.all(x[dpt.newaxis, :] == y3)
+
+
+@pytest.mark.parametrize("dt", [dpt.float16, dpt.float64, dpt.complex128])
+def test_asarray_to_device_with_unsupported_dtype(dt):
+    aspect = "fp16" if dt == dpt.float16 else "fp64"
+    try:
+        d0 = dpctl.select_device_with_aspects(aspect)
+    except dpctl.SyclDeviceCreationError:
+        pytest.skip("No device with aspect for test")
+    d1 = None
+    for d in dpctl.get_devices():
+        if d.default_selector_score < 0:
+            pass
+        try:
+            d1 = dpctl.select_device_with_aspects(
+                d.device_type.name, excluded_aspects=[aspect]
+            )
+        except dpctl.SyclDeviceCreationError:
+            pass
+    if d1 is None:
+        pytest.skip("No device with missing aspect for test")
+    x = dpt.ones(10, dtype=dt, device=d0)
+    y = dpt.asarray(x, device=d1)
+    assert y.sycl_device == d1