Merge pull request #4150 from silx-kit/fix_opencl_convolution_cpu

pierrepaleo · web-flow · commit 415b020023cb · 2024-08-02T16:43:42.000+02:00
`silx.opencl.convolution`: Fixed separable convolution on CPU that crashed in some cases
diff --git a/src/silx/opencl/convolution.py b/src/silx/opencl/convolution.py
@@ -385,6 +385,8 @@ def _batched_convolution(self, axis, input_ref=None, output_ref=None):
             self.kernel_args, input_ref, output_ref
         )
         ev = opencl_kernel(*opencl_kernel_args)
+        if self.queue.device.type == cl.device_type.CPU:
+            ev.wait()
         if self.profile:
             self.events.append(EventDescription("batched convolution", ev))
 

Original file line number	Diff line number	Diff line change
`@@ -385,6 +385,8 @@ def _batched_convolution(self, axis, input_ref=None, output_ref=None):`
`385`	`385`	`self.kernel_args, input_ref, output_ref`
`386`	`386`	`)`
`387`	`387`	`ev = opencl_kernel(*opencl_kernel_args)`
	`388`	`+ if self.queue.device.type == cl.device_type.CPU:`
	`389`	`+ ev.wait()`
`388`	`390`	`if self.profile:`
`389`	`391`	`self.events.append(EventDescription("batched convolution", ev))`
`390`	`392`