Merge pull request #54 from dkartch-nv/cuda-event
Update EGL_NV_cuda_event
diff --git a/extensions/NV/EGL_NV_cuda_event.txt b/extensions/NV/EGL_NV_cuda_event.txt
index 08154bc..cc54749 100644
--- a/extensions/NV/EGL_NV_cuda_event.txt
+++ b/extensions/NV/EGL_NV_cuda_event.txt
@@ -11,6 +11,7 @@
Debalina Bhattacharjee
Michael Chock
James Jones
+ Daniel Kartch
Contact
@@ -22,7 +23,7 @@
Version
- Version 1, June 20, 2014
+ Version 2, June 28, 2018
Number
@@ -34,11 +35,15 @@
Dependencies
- This extension is written against the language of EGL 1.5.
+ This extension is written against the language of EGL 1.5 and the
+ EGL_EXT_sync_reuse extension.
Either EGL_KHR_fence_sync and the EGLAttrib type or EGL 1.5 are
required.
+ This extension interacts with, but does not require,
+ EGL_EXT_sync_reuse.
+
This extension interacts with EGL_NV_device_cuda.
Overview
@@ -77,31 +82,31 @@
EGL_SYNC_CUDA_EVENT_COMPLETE_NV 0x323D
-Changes to Chapter 3 of the EGL 1.5 Specification (EGL Functions and
-Errors)
+Add to section 3.8.1 (Sync Objects) of the EGL 1.5 specification, after
+the sixth paragraph:
- In section 3.8.1 (Sync Objects) Replace the sixth paragraph:
+ Likewise, a <CUDA event sync object> reflects the status of a
+ corresponding CUDA object. Waiting on this type of sync object is
+ equivalent to waiting for completion of the corresponding linked
+ CUDA event object.
- "A <OpenCL event sync object> reflects the status of a corresponding
- OpenCL object to which the sync object is linked. Likewise, a <CUDA
- event sync object> reflects the status of a corresponding CUDA
- object. These provide another method of sharing images or compute
- results between EGL and the CUDA and OpenCL client APIs (see
- Chapter 9 of the OpenCL Specification and the cl_khr_egl_image
- extension for a second method of synchronization with OpenCL).
- Waiting on either type of sync object is equivalent to waiting for
- completion of the corresponding linked CUDA or OpenCL event object."
+Add a new section following section 3.8.1.2 (Creating and Signaling
+OpenCL Event Sync Objects):
- Add a new section following section 3.8.1.2 (Creating OpenCL Event
- Sync Objects):
-
- "Section 3.8.1.X Creating CUDA Event Sync Objects
+ Section 3.8.1.X Creating and Signaling CUDA Event Sync Objects
If <type> is EGL_SYNC_CUDA_EVENT_NV, a CUDA event sync object is
- created. In this case <attrib_list> must contain the attribute
+ created. The <attrib_list> may contain the attribute
EGL_CUDA_EVENT_HANDLE_NV, set to a pointer to a cudaEvent_t object.
- The object must be properly initialized and recorded by the CUDA API
- (using cudaCreateEvent and cudaEventRecord), and the CUDA device
+ If it does not contain this attribute, the sync object will start in
+ the signaled state, and an event attribute must be provided the
+ first time eglUnsignalSyncEXT is called. Otherwise, a call to
+ eglUnsignalSyncEXT may replace this event attribute or leave it
+ unspecified, causing the previous object to be reused.
+
+ A cudaEvent_t object provided to eglCreateSync or eglUnsignalSyncEXT
+ must be properly initialized and recorded by the CUDA API (using
+ cudaEventCreate and cudaEventRecord), and the CUDA device
used to create the event must correspond to <dpy>[fn1]. Note that
EGL_CUDA_EVENT_HANDLE_NV is not a queryable property of a sync
object.
@@ -119,47 +124,57 @@
EGL_SYNC_STATUS Depends on status of <event>
EGL_SYNC_CONDITION EGL_SYNC_CUDA_EVENT_COMPLETE_NV
- The status of such a sync object depends on the state of <event> at
- the time eglCreateSync was called. If all device work preceding the
- most recent call to cudaEventRecord on the event has not yet
- completed, the status of the linked sync object will be
- EGL_UNSIGNALED. If all such work has completed, the status of the
- linked sync object will be EGL_SIGNALED. Calling cudaEventRecord has
- no effect on a previously created sync object.
+ If an <event> is linked to the sync object, the status of this type
+ of sync object depends on the state of <event> evaluated at the time
+ of the most recent call to eglCreateSync or eglUnsignalSyncEXT. If
+ all device work preceding the most recent call to cudaEventRecord on
+ the event has not yet completed, the status of the linked sync
+ object will be EGL_UNSIGNALED. If all such work has completed, the
+ status of the linked sync object will be EGL_SIGNALED. Calling
+ cudaEventRecord to modify an event has no effect on the sync object
+ while its status is EGL_UNSIGNALED, but will have an effect if the
+ event is reevaluated at a subsequent eglUnsignalSyncEXT call.
The only condition supported for CUDA event sync objects is
EGL_SYNC_CUDA_EVENT_COMPLETE_NV. It is satisfied when all device
work prior to the most recent call to cudaEventRecord at sync
- creation time has completed."
+ unsignaling time has completed.
- Add to the list of errors following 3.8.1.X:
+If EGL_EXT_sync_reuse is not present, then change the second sentence of
+3.8.1.X above to "The <attrib_list> must contain ..." and omit the
+remaining sentences in the paragraph. Omit all references to
+eglUnsignalSyncEXT.
- "If <type> is EGL_SYNC_CUDA_EVENT_NV and EGL_CUDA_EVENT_HANDLE_NV is
- not specified in <attrib_list>, then an EGL_BAD_ATTRIBUTE error is
+In 3.8.1 (Sync Objects), if EGL_EXT_sync_reuse is present, then add the
+following to the error list for eglUnsignalSyncEXT. Otherwise add it to
+the error list for eglCreateSync:
+
+ If <type> is EGL_SYNC_CUDA_EVENT_NV and a EGL_CUDA_EVENT_HANDLE_NV
+ is not linked to the sync object, then an EGL_BAD_ATTRIBUTE error is
generated. If its attribute value is not a valid CUDA event pointer
or has not been initialized as described above, then
EGL_BAD_ATTRIBUTE may be generated, but the results are undefined
- and may include program termination."
+ and may include program termination.
- Modify the third paragraph of section 3.8.1.4 (Querying Sync Object
- Attributes):
+Modify the third paragraph of section 3.8.1.4 (Querying Sync Object
+Attributes):
- "If any eglClientWaitSync or eglWaitSync commands are blocking on
+ If any eglClientWaitSync or eglWaitSync commands are blocking on
<sync> when eglDestroySync is called, <sync> is flagged for deletion
and will be deleted when the associated fence command, OpenCL event
object, or CUDA event object has completed, and <sync> is no longer
blocking any such egl*WaitSync command. Otherwise, the sync object
is destroyed immediately.
- Replace the EGL_SYNC_CONDITION row of table 3.9 with:
+Replace the EGL_SYNC_CONDITION row of table 3.9 with:
- "Attribute Description Supported Sync Objects
- ----------------- ----------------------- ----------------------
- EGL_SYNC_CONDITION Signaling condition EGL_SYNC_FENCE,
- EGL_SYNC_CL_EVENT, or
- EGL_SYNC_CUDA_EVENT_NV
+ Attribute Description Supported Sync Objects
+ ------------------ ----------------------- ----------------------
+ EGL_SYNC_CONDITION Signaling condition EGL_SYNC_FENCE,
+ EGL_SYNC_CL_EVENT, or
+ EGL_SYNC_CUDA_EVENT_NV
- Table 3.9 Attributes Accepted by eglGetSyncAttrib"
+ Table 3.9 Attributes Accepted by eglGetSyncAttrib
Interactions with EGL versions prior to 1.5
@@ -181,5 +196,11 @@
Revision History
+ Version 2, 2018/06/28 (Daniel Kartch)
+ - Rewritten to clearly define interactions with
+ EGL_EXT_sync_reuse, without requiring it
+ - Fixed incorrect CUDA function name
+ - Fixed table spacing
+
Version 1, 2014/06/20 (Michael Chock)
- initial version.
