slangpy - npm Package Compare versions

Comparing version 0.34.0 to 0.35.0

slangpy/tests/slangpy_tests/test_raytracing.py

+# SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+
+import pytest
+import numpy as np
+
+import slangpy as spy
+from slangpy import DeviceType, Module
+from slangpy.types.tensor import Tensor
+from slangpy.testing import helpers
+
+
+def build_blas(
+    device: spy.Device, vertices: np.ndarray, indices: np.ndarray
+) -> spy.AccelerationStructure:
+    vertex_buffer = device.create_buffer(
+        usage=spy.BufferUsage.shader_resource,
+        label="vertex_buffer",
+        data=vertices,
+    )
+
+    index_buffer = device.create_buffer(
+        usage=spy.BufferUsage.shader_resource,
+        label="index_buffer",
+        data=indices,
+    )
+
+    blas_input_triangles = spy.AccelerationStructureBuildInputTriangles(
+        {
+            "vertex_buffers": [vertex_buffer],
+            "vertex_format": spy.Format.rgb32_float,
+            "vertex_count": vertices.size // 3,
+            "vertex_stride": vertices.itemsize * 3,
+            "index_buffer": index_buffer,
+            "index_format": spy.IndexFormat.uint32,
+            "index_count": indices.size,
+            "flags": spy.AccelerationStructureGeometryFlags.opaque,
+        }
+    )
+
+    blas_build_desc = spy.AccelerationStructureBuildDesc(
+        {
+            "inputs": [blas_input_triangles],
+        }
+    )
+
+    blas_sizes = device.get_acceleration_structure_sizes(blas_build_desc)
+
+    blas_scratch_buffer = device.create_buffer(
+        size=blas_sizes.scratch_size,
+        usage=spy.BufferUsage.unordered_access,
+        label="blas_scratch_buffer",
+    )
+
+    blas = device.create_acceleration_structure(
+        size=blas_sizes.acceleration_structure_size,
+        label="blas",
+    )
+
+    command_encoder = device.create_command_encoder()
+    command_encoder.build_acceleration_structure(
+        desc=blas_build_desc, dst=blas, src=None, scratch_buffer=blas_scratch_buffer
+    )
+    device.submit_command_buffer(command_encoder.finish())
+
+    return blas
+
+
+def build_tlas(
+    device: spy.Device, instance_list: spy.AccelerationStructureInstanceList
+) -> spy.AccelerationStructure:
+    tlas_build_desc = spy.AccelerationStructureBuildDesc(
+        {
+            "inputs": [instance_list.build_input_instances()],
+        }
+    )
+
+    tlas_sizes = device.get_acceleration_structure_sizes(tlas_build_desc)
+
+    tlas_scratch_buffer = device.create_buffer(
+        size=tlas_sizes.scratch_size,
+        usage=spy.BufferUsage.unordered_access,
+        label="tlas_scratch_buffer",
+    )
+
+    tlas = device.create_acceleration_structure(
+        size=tlas_sizes.acceleration_structure_size,
+        label="tlas",
+    )
+
+    command_encoder = device.create_command_encoder()
+    command_encoder.build_acceleration_structure(
+        desc=tlas_build_desc, dst=tlas, src=None, scratch_buffer=tlas_scratch_buffer
+    )
+    device.submit_command_buffer(command_encoder.finish())
+
+    return tlas
+
+
[e-mail]("device_type", helpers.DEFAULT_DEVICE_TYPES)
+def test_raytracing(device_type: DeviceType):
+    device = helpers.get_device(device_type)
+
+    if not device.has_feature(spy.Feature.acceleration_structure):
+        pytest.skip("Acceleration structures not supported on this device")
+    if not device.has_feature(spy.Feature.ray_tracing):
+        pytest.skip("Ray tracing not supported on this device")
+
+    vertices = np.array([-1, -1, 0, 1, -1, 0, -1, 1, 0], dtype=np.float32)
+    indices = np.array([0, 1, 2], dtype=np.uint32)
+    blas = build_blas(device, vertices, indices)
+
+    instance_list = device.create_acceleration_structure_instance_list(1)
+    instance_list.write(
+        0,
+        {
+            "transform": spy.float3x4.identity(),
+            "instance_id": 0,
+            "instance_mask": 0xFF,
+            "instance_contribution_to_hit_group_index": 0,
+            "flags": spy.AccelerationStructureInstanceFlags.none,
+            "acceleration_structure": blas.handle,
+        },
+    )
+    tlas = build_tlas(device, instance_list)
+
+    tensor = Tensor.zeros(device, (64, 64, 3), dtype=float)
+    module = Module(device.load_module("test_raytracing.slang"))
+
+    module.trace.ray_tracing(
+        hit_groups=[{"hit_group_name": "hit_group", "closest_hit_entry_point": "closest_hit"}],
+        miss_entry_points=["miss"],
+        max_recursion=1,
+        max_ray_payload_size=12,
+    )(tid=spy.call_id(), tlas=tlas, _result=tensor)
+
+    data = tensor.to_numpy()
+
+    # spy.tev.show(spy.Bitmap(data))
+
+    assert np.allclose(data[0, 0, :], [0, 0, 0], atol=0.01)
+    assert np.allclose(data[0, 63, :], [1, 0, 0], atol=0.01)
+    assert np.allclose(data[63, 0, :], [0, 1, 0], atol=0.01)
+    assert np.allclose(data[63, 63, :], [1, 0, 1], atol=0.01)
+
+
+if __name__ == "__main__":
+    pytest.main([__file__, "-v", "-s"])

METADATA

 Metadata-Version: 2.4
 Name: slangpy
-Version: 0.34.0
+Version: 0.35.0
 Summary: Easily call Slang functions and integrate with PyTorch auto diff directly from Python.
@@ -103,5 +103,5 @@ Author-email: Simon Kallweit <[e-mail]>, Chris Cummings <[e-mail]>, Benedikt Bitterli <[e-mail]>, Sai Bangaru <[e-mail]>, Yong He <[e-mail]>
     note = {https://github.com/shader-slang/slangpy},
-    version = {0.34.0},
+    version = {0.35.0},
     year = 2025
 }
 ```

RECORD

 slangpy/__init__.py,sha256=PsbtOD5NRRgDcN2C7SbPnePNpAzce326xQM9tZhBT9M,1948
-slangpy/__init__.pyi,sha256=L-SxD_v3BYdqsi4I3Uq6qtDXIThLa1pe68fVQfnuOBE,192810
-slangpy/libsgl.dylib,sha256=Bi8bqjreB8AgnP9PHk62KxeyydQiLf76e8Wq4jgt6EY,8336864
-slangpy/libslang-glsl-module.dylib,sha256=Hqm_WyMZpzM8m3-ARtacg5QFtECzV_ysBHhrx0ADOTE,1292880
-slangpy/libslang-glslang.dylib,sha256=9mH1jRPBHOU-RevJHtFJUyXCIYJkKZCFfSwYb1St_9o,7876544
-slangpy/libslang-rhi.a,sha256=bturAvxqevJ7j3wpuQ7mlXwAhGp1WS0mSUzkoHH1RXE,74103624
-slangpy/libslang-rt.dylib,sha256=9BKh0LwHqoOOPSYKittTf1PbtlObgoiIKTDjP5pmnaw,249056
-slangpy/libslang.dylib,sha256=xnn0H5N-CT0C-M3-hQS7Sgr5jX9TdfQm4SNVQSlc3yY,19208720
-slangpy/slangpy_ext.cpython-310-darwin.so,sha256=HIo0pZthekBN5_vCo83eDQYsIV_v9hhFw09mqT7uZQg,6608624
+slangpy/__init__.pyi,sha256=uitnhvr1sP-WmPkWxGCbQfZZhXPFNVqP2S1sHqIhJbE,193026
+slangpy/libsgl.dylib,sha256=fe5P5-act_QWXSUr83y1b4TnSlIFkltDGx-j1r__L2g,8336896
+slangpy/libslang-glsl-module.dylib,sha256=M-suEgdNUFMlBuSlG5-wUm1hsUcu0r85QcYAhYD2Vf8,1292880
+slangpy/libslang-glslang.dylib,sha256=0941UZI3GrFNd-JilmnaEsMnglDmndBgvOZFkAnZ948,8033520
+slangpy/libslang-rhi.a,sha256=FG1-ksx-jWgFvpR3ScrEU4wBU-d_el91yfocrLG8PT0,74108576
+slangpy/libslang-rt.dylib,sha256=C7DXKhnoycvr4OiHKSSfWXgqtML4mnyg50Lm_gsZQaQ,249968
+slangpy/libslang.dylib,sha256=3wuOwjodITu-7CPh7XtRX5S7q7PIcm1hGgz8HBO3d8Q,19373712
+slangpy/slangpy_ext.cpython-310-darwin.so,sha256=ghED3AR43maqxD68k0YbXI5-OwCGUL1CAvx0emmB6ZI,6630720
 slangpy/benchmarks/conftest.py,sha256=rAYyblUkfrklcEe_TbJ0UuSAzQ8b0pmAbrWG2_s1nrg,438
@@ -34,11 +34,11 @@ slangpy/benchmarks/test_benchmark_interop.py,sha256=7xkivr7F1sD7_VRUVivVy125whirpZkbuttauURqgDw,15467
 slangpy/core/__init__.py,sha256=TJYDzVSv8gf6xRO-8P6Fg8BpYxp3rjdxiUgt18yd_R4,94
-slangpy/core/calldata.py,sha256=z_suSAbkfptrWhIB0tBQgoN_AOfXBjSPg_HQG5GOfd8,18003
-slangpy/core/callsignature.py,sha256=UqZgXTkXT7ObkB7SOi2iDlDd3FQUkSu_dEtjO79TOY8,25905
-slangpy/core/dispatchdata.py,sha256=J0yHRgwYJno6n9WStq3mbdJLJuSQiZmcrdsO52cBMJ0,10739
+slangpy/core/calldata.py,sha256=oVK6IexBHGWsp7Qffx1wS75QymE8V57HCRCKC-bmxUM,21458
+slangpy/core/callsignature.py,sha256=_ljcrJe1Z8E6ckaWaUaDkWuauP3ZGzBl3EEMlGPT6XU,26970
+slangpy/core/dispatchdata.py,sha256=S176OEjLyzVovpkomRllpxVproTy4KttSRnU38ie5Z0,10955
 slangpy/core/enums.py,sha256=lLLVsfRqLWNMDOE7eaxOyy6t6KNzbXjv2JsvmepltU0,211
-slangpy/core/function.py,sha256=dnor5p_ANTHnpGS7tUYRT8n6OzVi6abEx5V__ODW0iw,20499
+slangpy/core/function.py,sha256=25UwWJ2-Hjesiyk0GextPawvmnHLV5vgsMEr6QN6pDA,23663
 slangpy/core/instance.py,sha256=Q-Jjt1rU9p1-j_aFQA01DdWCqKaPGfn3C3xgZ_QF8eU,4535
 slangpy/core/jupyter.py,sha256=iRB4SgmlI3qCk0itlUUGYjB--BwvloKpvB8ZT7UOzVY,9676
 slangpy/core/logging.py,sha256=LW5fu_5neXYMwnsF8dYXalmQGZJl6j3xWTfKgJ_xZ0k,8688
-slangpy/core/module.py,sha256=wt2uQYR0Y-Yl1tyfDzsd6sMEAOeNJMWwXis3tftjjRI,8134
+slangpy/core/module.py,sha256=4q9evVKFmNpB81mEUiiyOoV8rGLrJkId1tGQd6K4LXc,8215
 slangpy/core/native.py,sha256=SpEiYl6WrtB8c1n7CBUVcdsAmnmqpN37DS3JbtPMEqU,151
@@ -103,3 +103,3 @@ slangpy/core/packedarg.py,sha256=8yD7ZHZ3I7J2LHGVKdMDpBL7UK-11XrJ5bU6UCsScYY,2060
 slangpy/experimental/gridarg.py,sha256=ZRTZCmN3dQLS0oTZ7JPf9CK8GpEpC489XkaQG4wUyj4,4207
-slangpy/include/slang-deprecated.h,sha256=Y3lFxhYrnatDzVFybqurn_r2k3hB_NjkM3zU4gQczDU,69513
+slangpy/include/slang-deprecated.h,sha256=4HgvR46cTImMBvQI3u2Weukaz9hykr0P_xLVG1ZZgSA,69655
 slangpy/include/slang-image-format-defs.h,sha256=Na2baM-oiRV7K_Sk7zkmO0rV8Pdg_-5mrKA0OOg2O80,2421
@@ -109,4 +109,4 @@ slangpy/include/slang-rhi-config.h,sha256=Em9yoqMRi3H21r1rWhJgTwSbWUk88IEQzrQ-ot2Ppf0,403
 slangpy/include/slang-user-config.h,sha256=-DTfKSDXKnCWR_NU6n96RYbQyYmM5ms2IFQH8UWNn7Y,151
-slangpy/include/slang.h,sha256=x0nFtGeAM_HmiReKpSjzPIPdyOTnWRAXkNt-KCR6paE,178544
-slangpy/include/sgl/sgl.h,sha256=R3uIVJZJM5Q-CUtCv9KDCvvpEjyLerKONwt5jpT7aTU,610
+slangpy/include/slang.h,sha256=sQbZxKt6WhP8F8wvU8EnyMzltucfmIWgQ_7tdZnhRvM,179036
+slangpy/include/sgl/sgl.h,sha256=XEOQ9UxGba9xzNvoyrUDI7QNRAmxa5SOJUF8sWUWtKM,610
 slangpy/include/sgl/sgl_pch.h,sha256=SD1ALBpwwkcMlCnU_V8nHdSvS6-NZIXZOv5Rd6_r5z8,131
@@ -121,3 +121,3 @@ slangpy/include/sgl/app/app.h,sha256=tyvkNXNrsvsYWL98uSxVKcYesdiJ2ZRqdwJpfw9dbwA,2809
 slangpy/include/sgl/core/error.h,sha256=o-Mi9Fo-j3sDX1BZnIljNdLQtRylKiJuPYr3AQvCCS0,6727
-slangpy/include/sgl/core/file_stream.h,sha256=KvaJHbPP-WSReT9k7DeMHtiVbRNkyzuvevkqAcdVlmI,1787
+slangpy/include/sgl/core/file_stream.h,sha256=7JXL14o9sdLC4DJmKoyJ5jWyN6wIBdK1ALudklAXYQs,1509
 slangpy/include/sgl/core/file_system_watcher.h,sha256=1D_gaJh-GtPUNEiBySx46VwPZw4rvjeh2e_qE3tUlyo,3968
@@ -140,3 +140,3 @@ slangpy/include/sgl/core/format.h,sha256=WP8Qy7Y8sfnJFiHqodsmzPsi8W5VRlQM5-8gpi0WNqU,903
 slangpy/include/sgl/core/static_vector.h,sha256=Giyz8df4Cz0apEfaAOyPHx51pd8X-TO_AiDeq4fdP14,3070
-slangpy/include/sgl/core/stream.h,sha256=RNwFNcYmq6lPC9-zZprqSySyV9UabevNuuDQgWHZ5Fs,1507
+slangpy/include/sgl/core/stream.h,sha256=mdrcEPnkxZSdrzI9pjoqCvEsGFkOX0KxJsRV5UlvIJs,1786
 slangpy/include/sgl/core/string.h,sha256=azYJO32kfk9ThHoLsEjQMiTiYUPt-2Uadve1mEvA0v8,8223
@@ -188,3 +188,3 @@ slangpy/include/sgl/core/thread.h,sha256=A9W4s8uQ2_BAxL-yUZlGZaIrWSngO9bMUZiGzyK7qpE,670
 slangpy/include/sgl/math/matrix.h,sha256=ipSvS3MPZqHHKbahaG4OSowi4WxUeU2WqV45sDCQQQo,143
-slangpy/include/sgl/math/matrix_math.h,sha256=uYo82bL_D2IKntAQ-c5TipUQ7YFOkR9NO7t3LR-07zo,24114
+slangpy/include/sgl/math/matrix_math.h,sha256=jazXRhe9NoPrwrXQnYG_t8yFmGvYlMrRsNt28CUE7EQ,24382
 slangpy/include/sgl/math/matrix_types.h,sha256=n-Yg6I3EmZIsWoKxM6xLFW9i0Kgv4axTDdy7tO5NlSI,5645
@@ -216,3 +216,3 @@ slangpy/include/sgl/math/quaternion.h,sha256=SjZfbLOa0GI7eGIyyi0exmcyNIG6L9T1J3A92nYLCek,151
 slangpy/math/__init__.py,sha256=iasSej-enqU0bZTocJyJYiXTfADJ9sd89NwHtF1PDI0,58
-slangpy/math/__init__.pyi,sha256=x1kH1ceRfEIZJ5fNipmAI-UuilqP3wRA133pAg9Og-4,113876
+slangpy/math/__init__.pyi,sha256=Qp57jIhdbzO3YpMu6wyo1CfVDi8HJRdu8xJL1Ysx7Xs,113931
 slangpy/platform/__init__.py,sha256=iasSej-enqU0bZTocJyJYiXTfADJ9sd89NwHtF1PDI0,58
@@ -243,3 +243,3 @@ slangpy/platform/__init__.pyi,sha256=Xew0oHwYHVdPnScU_wCGmf1dVnfjPhXDfdlzycqjL9U,2599
 slangpy/slangpy/__init__.py,sha256=iasSej-enqU0bZTocJyJYiXTfADJ9sd89NwHtF1PDI0,58
-slangpy/slangpy/__init__.pyi,sha256=8bJzuPxdgETTHbvRSpvVldDk37T88E7LDjK43LLJILU,22329
+slangpy/slangpy/__init__.pyi,sha256=ETgClZZDNoMZqtHrORVjG7_J34Vp8EzyUpjBwOdpIEM,22491
 slangpy/testing/__init__.py,sha256=iasSej-enqU0bZTocJyJYiXTfADJ9sd89NwHtF1PDI0,58
@@ -323,3 +323,3 @@ slangpy/testing/helpers.py,sha256=aQBRM5wYX_MW11cqWzd_HBnjekS28KsHTEaOVOoaXQM,15087
 slangpy/tests/slangpy_tests/test_buffers.py,sha256=f0bNcEB_E20QS4ag505SzbKSk_HtXVR6oDAlcUV6d_M,1265
-slangpy/tests/slangpy_tests/test_caching.py,sha256=lw5uauQyGuc7Mk1ggy_Ps96AJQ3ttrCTeMbK-F8DfpM,2045
+slangpy/tests/slangpy_tests/test_caching.py,sha256=O3jBBSeuMBkYNkRZdpFlISnLA8C-nxo1t0TOIpmRkDM,2029
 slangpy/tests/slangpy_tests/test_call_group_integrations.py,sha256=Zid7TVEjFK7XkbFsLSNDLFPp1tt5pQ3FAxX29XudA5E,32480
@@ -349,2 +349,4 @@ slangpy/tests/slangpy_tests/test_call_groups.py,sha256=_1BD8Upp-qeCbUsI--NaZ2zLKZJvjxZAjH3kBEd25RY,26207
 slangpy/tests/slangpy_tests/test_raw_dispatch.py,sha256=pj6ut45b9I1mRrBTntlyp8eCbaruY3JWVGurFoJ1Q_g,6140
+slangpy/tests/slangpy_tests/test_raytracing.py,sha256=Dww27b51bZIBxD6hvzoi5Sxrzlqcxk8E0dW_isrVOTs,4720
+slangpy/tests/slangpy_tests/test_raytracing.slang,sha256=-q5YFKPv144lQxMc2j1DCTJFq1rk0c1QJxtjCF61Cxw,903
 slangpy/tests/slangpy_tests/test_reflection2.py,sha256=qO6EIWdnCYyHa-TQIU5m31WiU15Ga2oLp-41qjBTEUc,12989
@@ -388,3 +390,3 @@ slangpy/tests/slangpy_tests/test_return_types.py,sha256=MwnQy8791jYrelOFt8-9ImEL6H8zUUUfJRJ_Z2kXXz4,3731
 slangpy/types/__init__.py,sha256=fmSHX1CNHOdyiFQ6Q2QVsTHghJiADss_Wf9Y23SruNU,731
-slangpy/types/buffer.py,sha256=jpoIB5oL3jJE5MhtpkiMAdpF_AMR8e3bYEEPVjN8duM,16323
+slangpy/types/buffer.py,sha256=3cJBBREBwzLtiAIHaXn6woL3JCY-YAseInt3h2Npiu4,16584
 slangpy/types/callidarg.py,sha256=hIPIfFw17TbzL5rpmBcpZN5lsQK4ge66ZAVwO9wc-yA,2623
@@ -400,6 +402,6 @@ slangpy/types/diffpair.py,sha256=K9C4kplAdFJbjeDoe6f7JYbM43Cs3S60ygjGonrrVqg,1627
 slangpy/ui/__init__.pyi,sha256=6oKnW1qupqQCPSe82znQzssxR5mw6zAuM4mGr3CaSvQ,30512
-slangpy-0.34.0.dist-info/licenses/LICENSE,sha256=tzjZi3clQUtTTBGaxmEmh2uMvUCqiuSCENsZBxS4FnA,1490
-slangpy-0.34.0.dist-info/METADATA,sha256=-30QRMsOk7W4oCUsmiO_USKf5NCLRt0aFy6SGUYfPnU,4323
-slangpy-0.34.0.dist-info/WHEEL,sha256=QbwVjH-TlueSdsOKSic424jsJD30r52gknYt4wXVwF0,109
-slangpy-0.34.0.dist-info/top_level.txt,sha256=Wo7_Eny8d-MI3ZHT-XHGGMEKKIK38FlAmUCz8AffxA0,8
-slangpy-0.34.0.dist-info/RECORD,,
+slangpy-0.35.0.dist-info/licenses/LICENSE,sha256=tzjZi3clQUtTTBGaxmEmh2uMvUCqiuSCENsZBxS4FnA,1490
+slangpy-0.35.0.dist-info/METADATA,sha256=Ue9H520aYJVwxkzlynefSeDPFT0wbm1O4nvFtvwhM94,4323
+slangpy-0.35.0.dist-info/WHEEL,sha256=QbwVjH-TlueSdsOKSic424jsJD30r52gknYt4wXVwF0,109
+slangpy-0.35.0.dist-info/top_level.txt,sha256=Wo7_Eny8d-MI3ZHT-XHGGMEKKIK38FlAmUCz8AffxA0,8
+slangpy-0.35.0.dist-info/RECORD,,

slangpy/core/calldata.py

@@ -135,4 +135,7 @@ # SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 
-            # Set call data mode based on device type
-            if build_info.module.device.info.type == DeviceType.cuda:
+            # Set call data mode based on device and pipeline type
+            if (
+                build_info.module.device.info.type == DeviceType.cuda
+                and build_info.pipeline_type == PipelineType.compute
+            ):
                 self.call_data_mode = CallDataMode.entry_point
@@ -312,5 +315,8 @@ else:
             # Check if we've already built this module.
-            if hash in build_info.module.compute_pipeline_cache:
+            if hash in build_info.module.pipeline_cache:
                 # Get pipeline from cache if we have
-                self.compute_pipeline = build_info.module.compute_pipeline_cache[hash]
+                self.pipeline = build_info.module.pipeline_cache[hash]
+                # Get shader table from cache if the pipeline is a raytracing pipeline
+                if build_info.pipeline_type == PipelineType.ray_tracing:
+                    self.shader_table = build_info.module.shader_table_cache[hash]
                 self.device = build_info.module.device
@@ -325,15 +331,69 @@ self.log_debug(f"  Found cached pipeline with hash {hash}")
                 module = session.load_module_from_source(hash, code)
-                ep = module.entry_point(f"compute_main", type_conformances)
                 opts = SlangLinkOptions()
                 opts.dump_intermediates = _DUMP_SLANG_INTERMEDIATES
                 opts.dump_intermediates_prefix = sanitized
-                program = session.link_program(
-                    [module, build_info.module.device_module] + build_info.module.link,
-                    [ep],
-                    opts,
-                )
-                self.compute_pipeline = device.create_compute_pipeline(
-                    program, defer_target_compilation=True
-                )
-                build_info.module.compute_pipeline_cache[hash] = self.compute_pipeline
+                if build_info.pipeline_type == PipelineType.compute:
+                    # Create compute pipeline
+                    ep = module.entry_point(f"compute_main", type_conformances)
+                    program = session.link_program(
+                        [module, build_info.module.device_module] + build_info.module.link,
+                        [ep],
+                        opts,
+                    )
+                    self.pipeline = device.create_compute_pipeline(
+                        program, defer_target_compilation=True
+                    )
+                    build_info.module.pipeline_cache[hash] = self.pipeline
+                elif build_info.pipeline_type == PipelineType.ray_tracing:
+                    # Create ray tracing pipeline
+                    eps = [module.entry_point(f"raygen_main", type_conformances)]
+                    hit_group_names: list[str] = []
+                    for hit_group in build_info.ray_tracing_hit_groups:
+                        hit_group_names.append(hit_group.hit_group_name)
+                        if hit_group.closest_hit_entry_point != "":
+                            eps.append(
+                                build_info.module.device_module.entry_point(
+                                    hit_group.closest_hit_entry_point
+                                )
+                            )
+                        if hit_group.any_hit_entry_point != "":
+                            eps.append(
+                                build_info.module.device_module.entry_point(
+                                    hit_group.any_hit_entry_point
+                                )
+                            )
+                        if hit_group.intersection_entry_point != "":
+                            eps.append(
+                                build_info.module.device_module.entry_point(
+                                    hit_group.intersection_entry_point
+                                )
+                            )
+                    for miss_entry_point in build_info.ray_tracing_miss_entry_points:
+                        eps.append(build_info.module.device_module.entry_point(miss_entry_point))
+
+                    program = session.link_program(
+                        [module, build_info.module.device_module] + build_info.module.link,
+                        eps,
+                        opts,
+                    )
+                    self.pipeline = device.create_ray_tracing_pipeline(
+                        program,
+                        hit_groups=build_info.ray_tracing_hit_groups,
+                        max_recursion=build_info.ray_tracing_max_recursion,
+                        max_ray_payload_size=build_info.ray_tracing_max_ray_payload_size,
+                        max_attribute_size=build_info.ray_tracing_max_attribute_size,
+                        flags=build_info.ray_tracing_flags,
+                        defer_target_compilation=True,
+                    )
+                    build_info.module.pipeline_cache[hash] = self.pipeline
+                    self.shader_table = device.create_shader_table(
+                        program,
+                        ray_gen_entry_points=["raygen_main"],
+                        miss_entry_points=build_info.ray_tracing_miss_entry_points,
+                        hit_group_names=hit_group_names,
+                        callable_entry_points=build_info.ray_tracing_callable_entry_points,
+                    )
+                    build_info.module.shader_table_cache[hash] = self.shader_table
+                else:
+                    raise RuntimeError("Unknown pipeline type")
                 self.device = device
@@ -340,0 +400,0 @@ self.log_debug(f"  Build succesful")

slangpy/core/callsignature.py

@@ -5,6 +5,7 @@ # SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 from slangpy.core.native import AccessType, CallMode, CallDataMode, NativeMarshall
+from slangpy.core.function import PipelineType
 
 import slangpy.bindings.typeregistry as tr
 import slangpy.reflection as slr
-from slangpy import ModifierID, TypeReflection, DeviceType
+from slangpy import ModifierID, TypeReflection
 from slangpy.bindings.marshall import Marshall, BindContext, ReturnContext
@@ -601,20 +602,32 @@ from slangpy.bindings.boundvariable import (
     # Generate the main function
-    cg.kernel.append_line('[shader("compute")]')
-    if call_group_size != 1:
-        cg.kernel.append_line(f"[numthreads({call_group_size}, 1, 1)]")
+    if build_info.pipeline_type == PipelineType.compute:
+        cg.kernel.append_line('[shader("compute")]')
+        if call_group_size != 1:
+            cg.kernel.append_line(f"[numthreads({call_group_size}, 1, 1)]")
+        else:
+            cg.kernel.append_line("[numthreads(32, 1, 1)]")
+        # Note: While flat_call_thread_id is 3-dimensional, we consider it "flat" and 1-dimensional because of the
+        #       true call group shape of [x, 1, 1] and only use the first dimension for the call thread id.
+        if is_entry_point:
+            cg.kernel.append_line(
+                "void compute_main(int3 flat_call_thread_id: SV_DispatchThreadID, int3 flat_call_group_id: SV_GroupID, int flat_call_group_thread_id: SV_GroupIndex, uniform CallData call_data)"
+            )
+        else:
+            cg.kernel.append_line(
+                "void compute_main(int3 flat_call_thread_id: SV_DispatchThreadID, int3 flat_call_group_id: SV_GroupID, int flat_call_group_thread_id: SV_GroupIndex)"
+            )
+    elif build_info.pipeline_type == PipelineType.ray_tracing:
+        cg.kernel.append_line('[shader("raygen")]')
+        if is_entry_point:
+            cg.kernel.append_line("void raygen_main(uniform CallData call_data)")
+        else:
+            cg.kernel.append_line("void raygen_main()")
     else:
-        cg.kernel.append_line("[numthreads(32, 1, 1)]")
+        raise RuntimeError(f"Unknown pipeline type: {build_info.pipeline_type}")
 
-    # Note: While flat_call_thread_id is 3-dimensional, we consider it "flat" and 1-dimensional because of the
-    #       true call group shape of [x, 1, 1] and only use the first dimension for the call thread id.
-    if is_entry_point:
-        cg.kernel.append_line(
-            "void compute_main(int3 flat_call_thread_id: SV_DispatchThreadID, int3 flat_call_group_id: SV_GroupID, int flat_call_group_thread_id: SV_GroupIndex, CallData call_data)"
-        )
-    else:
-        cg.kernel.append_line(
-            "void compute_main(int3 flat_call_thread_id: SV_DispatchThreadID, int3 flat_call_group_id: SV_GroupID, int flat_call_group_thread_id: SV_GroupIndex)"
-        )
     cg.kernel.begin_block()
 
+    if build_info.pipeline_type == PipelineType.ray_tracing:
+        cg.kernel.append_statement("int3 flat_call_thread_id = DispatchRaysIndex();")
+
     cg.kernel.append_statement("if (any(flat_call_thread_id >= call_data._thread_count)) return")
@@ -628,10 +641,18 @@
     if call_data_len > 0:
-        cg.kernel.append_line(
-            f"""
-        if (!init_thread_local_call_shape_info(flat_call_group_thread_id,
-            flat_call_group_id, flat_call_thread_id, call_data._grid_stride,
-            call_data._grid_dim, call_data._call_dim))
-            return;"""
-        )
-
+        if build_info.pipeline_type == PipelineType.compute:
+            cg.kernel.append_line(
+                f"""
+    if (!init_thread_local_call_shape_info(flat_call_group_thread_id,
+        flat_call_group_id, flat_call_thread_id, call_data._grid_stride,
+        call_data._grid_dim, call_data._call_dim))
+        return;"""
+            )
+        elif build_info.pipeline_type == PipelineType.ray_tracing:
+            cg.kernel.append_line(
+                f"""
+    if (!init_thread_local_call_shape_info(0,
+        uint3(0), flat_call_thread_id, call_data._grid_stride,
+        call_data._grid_dim, call_data._call_dim))
+        return;"""
+            )
         context_args += ", CallShapeInfo::get_call_id().shape"
@@ -638,0 +659,0 @@

slangpy/core/dispatchdata.py

@@ -12,3 +12,10 @@ # SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 
-from slangpy import CommandEncoder, ShaderCursor, SlangLinkOptions, uint3, DeviceType
+from slangpy import (
+    CommandEncoder,
+    ComputePipeline,
+    ShaderCursor,
+    SlangLinkOptions,
+    uint3,
+    DeviceType,
+)
 from slangpy.core.native import NativeCallRuntimeOptions
@@ -178,5 +185,8 @@ from slangpy.bindings.marshall import BindContext
             # Check if we've already built this module.
-            if hash in build_info.module.compute_pipeline_cache:
+            if hash in build_info.module.pipeline_cache:
                 # Get pipeline from cache if we have
-                self.compute_pipeline = build_info.module.compute_pipeline_cache[hash]
+                pipeline = build_info.module.pipeline_cache[hash]
+                if not isinstance(pipeline, ComputePipeline):
+                    raise RuntimeError("Pipeline cache entry is not a ComputePipeline")
+                self.compute_pipeline = pipeline
                 self.device = build_info.module.device
@@ -183,0 +193,0 @@ else:

slangpy/core/function.py

 # SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
-from typing import TYPE_CHECKING, Any, Callable, Optional, Protocol, Union, cast
+from typing import TYPE_CHECKING, Any, Callable, Optional, Protocol, Union, cast, Sequence
+from enum import Enum
 
@@ -13,3 +14,12 @@ from slangpy.core.native import (
 from slangpy.reflection import SlangFunction, SlangType
-from slangpy import CommandEncoder, TypeConformance, uint3, Logger, NativeHandle, NativeHandleType
+from slangpy import (
+    CommandEncoder,
+    TypeConformance,
+    uint3,
+    Logger,
+    NativeHandle,
+    NativeHandleType,
+    RayTracingPipelineFlags,
+    HitGroupDesc,
+)
 from slangpy.slangpy import Shape
@@ -22,2 +32,3 @@ from slangpy.bindings.typeregistry import PYTHON_SIGNATURES
     from slangpy.core.struct import Struct
+    from slangpy import HitGroupDescParam
 
@@ -44,2 +55,7 @@ ENABLE_CALLDATA_CACHE = True
 
+class PipelineType(Enum):
+    compute = 0
+    ray_tracing = 1
+
+
 class FunctionBuildInfo:
@@ -66,2 +82,10 @@ def __init__(self) -> None:
         self.call_group_shape: Optional[Shape] = None
+        self.pipeline_type: PipelineType = PipelineType.compute
+        self.ray_tracing_hit_groups: list[HitGroupDesc] = []
+        self.ray_tracing_miss_entry_points: list[str] = []
+        self.ray_tracing_callable_entry_points: list[str] = []
+        self.ray_tracing_max_recursion: int = 0
+        self.ray_tracing_max_ray_payload_size: int = 0
+        self.ray_tracing_max_attribute_size: int = 8
+        self.ray_tracing_flags: RayTracingPipelineFlags = RayTracingPipelineFlags.none
 
@@ -160,2 +184,26 @@
 
+    def ray_tracing(
+        self,
+        hit_groups: Sequence["HitGroupDescParam"],
+        miss_entry_points: Sequence[str] = [],
+        callable_entry_points: Sequence[str] = [],
+        max_recursion: int = 1,
+        max_ray_payload_size: int = 32,
+        max_attribute_size: int = 8,
+        flags: RayTracingPipelineFlags = RayTracingPipelineFlags.none,
+    ):
+        """
+        Specify the ray tracing pipeline configuration.
+        """
+        return FunctionNodeRayTracing(
+            self,
+            hit_groups,
+            miss_entry_points,
+            callable_entry_points,
+            max_recursion,
+            max_ray_payload_size,
+            max_attribute_size,
+            flags,
+        )
+
     @property
@@ -463,2 +511,41 @@ def bwds(self):
 
+class FunctionNodeRayTracing(FunctionNode):
+    def __init__(
+        self,
+        parent: NativeFunctionNode,
+        hit_groups: Sequence["HitGroupDescParam"],
+        miss_entry_points: Sequence[str],
+        callable_entry_points: Sequence[str],
+        max_recursion: int,
+        max_ray_payload_size: int,
+        max_attribute_size: int,
+        flags: RayTracingPipelineFlags,
+    ):
+        super().__init__(
+            parent,
+            FunctionNodeType.ray_tracing,
+            {
+                "hit_groups": [HitGroupDesc(hit_group) for hit_group in hit_groups],  # type: ignore
+                "miss_entry_points": list(miss_entry_points),
+                "callable_entry_points": list(callable_entry_points),
+                "max_recursion": max_recursion,
+                "max_ray_payload_size": max_ray_payload_size,
+                "max_attribute_size": max_attribute_size,
+                "flags": flags,
+            },
+        )
+        self.slangpy_signature = f"({hit_groups}, {miss_entry_points}, {callable_entry_points}, {max_recursion}, {max_ray_payload_size}, {max_attribute_size}, {flags})"
+
+    def _populate_build_info(self, info: FunctionBuildInfo):
+        d = cast(dict[str, Any], self._native_data)
+        info.pipeline_type = PipelineType.ray_tracing
+        info.ray_tracing_hit_groups = d["hit_groups"]
+        info.ray_tracing_miss_entry_points = d["miss_entry_points"]
+        info.ray_tracing_callable_entry_points = d["callable_entry_points"]
+        info.ray_tracing_max_recursion = d["max_recursion"]
+        info.ray_tracing_max_ray_payload_size = d["max_ray_payload_size"]
+        info.ray_tracing_max_attribute_size = d["max_attribute_size"]
+        info.ray_tracing_flags = d["flags"]
+
+
 class FunctionNodeBwds(FunctionNode):
@@ -465,0 +552,0 @@ def __init__(self, parent: NativeFunctionNode) -> None:

slangpy/core/module.py

@@ -7,3 +7,3 @@ # SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 
-from slangpy import ComputePipeline, SlangModule, Device, Logger
+from slangpy import Pipeline, ShaderTable, SlangModule, Device, Logger
 from slangpy.core.native import NativeCallDataCache
@@ -77,3 +77,4 @@ from slangpy.reflection import SlangProgramLayout
         self.dispatch_data_cache: dict[str, "DispatchData"] = {}
-        self.compute_pipeline_cache: dict[str, ComputePipeline] = {}
+        self.pipeline_cache: dict[str, Pipeline] = {}
+        self.shader_table_cache: dict[str, ShaderTable] = {}
         self.logger: Optional[Logger] = None
@@ -215,3 +216,4 @@
         self.dispatch_data_cache = {}
-        self.compute_pipeline_cache = {}
+        self.pipeline_cache = {}
+        self.shader_table_cache = {}
         self._attr_cache = {}
@@ -218,0 +220,0 @@

slangpy/include/sgl/core/file_stream.h

@@ -15,16 +15,3 @@ // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 
-class SGL_API EOFException : public std::runtime_error {
-public:
-    EOFException(const std::string& what, size_t gcount)
-        : std::runtime_error(what)
-        , m_gcount(gcount)
-    {
-    }
 
-    size_t gcount() const { return m_gcount; }
-
-private:
-    size_t m_gcount;
-};
-
 class SGL_API FileStream : public Stream {
@@ -31,0 +18,0 @@ SGL_OBJECT(FileStream)

slangpy/include/sgl/core/stream.h

@@ -10,2 +10,16 @@ // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 
+class SGL_API EOFException : public std::runtime_error {
+public:
+    EOFException(const std::string& what, size_t gcount)
+        : std::runtime_error(what)
+        , m_gcount(gcount)
+    {
+    }
+
+    size_t gcount() const { return m_gcount; }
+
+private:
+    size_t m_gcount;
+};
+
 /**
@@ -12,0 +26,0 @@ * \brief Base class for all stream objects.

slangpy/include/sgl/math/matrix_math.h

@@ -714,2 +714,13 @@ // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 
+template<floating_point T>
+[[nodiscard]] inline matrix<T, 4, 4> matrix_4x4_from_3x4(const matrix<T, 3, 4>& m)
+{
+    matrix<T, 4, 4> result;
+    for (int r = 0; r < 3; ++r) {
+        result.set_row(r, m.get_row(r));
+    }
+    result[3][3] = T(1);
+    return result;
+}
+
 template<typename T, int R, int C>
@@ -716,0 +727,0 @@ [[nodiscard]] std::string to_string(const matrix<T, R, C>& m)

slangpy/include/sgl/sgl.h

@@ -8,3 +8,3 @@ // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 #define SGL_VERSION_MAJOR 0
-#define SGL_VERSION_MINOR 34
+#define SGL_VERSION_MINOR 35
 #define SGL_VERSION_PATCH 0
@@ -11,0 +11,0 @@

slangpy/slangpy/__init__.pyi

@@ -336,9 +336,16 @@ from collections.abc import Mapping, Sequence
     @property
-    def compute_pipeline(self) -> slangpy.ComputePipeline:
+    def pipeline(self) -> slangpy.Pipeline:
         """N/A"""
 
-    @compute_pipeline.setter
-    def compute_pipeline(self, arg: slangpy.ComputePipeline, /) -> None: ...
+    @pipeline.setter
+    def pipeline(self, arg: slangpy.Pipeline, /) -> None: ...
 
     @property
+    def shader_table(self) -> slangpy.ShaderTable:
+        """N/A"""
+
+    @shader_table.setter
+    def shader_table(self, arg: slangpy.ShaderTable, /) -> None: ...
+
+    @property
     def call_dimensionality(self) -> int:
@@ -712,2 +719,4 @@ """N/A"""
 
+    ray_tracing = 5
+
 class NativeFunctionNode(NativeObject):
@@ -714,0 +723,0 @@ def __init__(self, parent: NativeFunctionNode | None, type: FunctionNodeType, data: object | None) -> None:

slangpy/tests/slangpy_tests/test_caching.py

@@ -58,3 +58,3 @@ # SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
     assert float_float_cd != mapped_float_float_cd
-    assert float_float_cd.compute_pipeline == mapped_float_float_cd.compute_pipeline
+    assert float_float_cd.pipeline == mapped_float_float_cd.pipeline
 
@@ -61,0 +61,0 @@

slangpy/types/buffer.py

@@ -73,2 +73,8 @@ # SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 
+def _hot_reload_lookup_module(device: Device):
+    if device in global_lookup_modules:
+        dummy_module = device.load_module_from_source("slangpy_layout", 'import "slangpy";')
+        global_lookup_modules[device].on_hot_reload(dummy_module.layout)
+
+
 def get_lookup_module(device: Device) -> SlangProgramLayout:
@@ -78,3 +84,3 @@ if device not in global_lookup_modules:
         device.register_device_close_callback(_on_device_close)
-        device.register_shader_hot_reload_callback(lambda _: _load_lookup_module(device))
+        device.register_shader_hot_reload_callback(lambda _: _hot_reload_lookup_module(device))
 
@@ -81,0 +87,0 @@ return global_lookup_modules[device]

Alert delta unavailable

Currently unable to show alert delta for PyPI packages.

No metric changes