clarification on zetCommandListAppendMetricQueryEnd phWaitEvents

[jandres742]

For zetCommandListAppendMetricQueryEnd https://spec.oneapi.io/level-zero/latest/tools/api.html?highlight=zetcommandlistappendmetricqueryend#zetcommandlistappendmetricqueryend we have

phWaitEvents – [in][mbz] must be nullptr

ZE_RESULT_ERROR_INVALID_NULL_POINTER

nullptr == phWaitEvents

so it seems we have contradictory statements. phWaitEvents must be zero, but if so, we return INVALID_NULL_POINTER. We need to choose one of the two.

[bmyates]

This api definition was intentionally changed to mbz in spec 1.1. The error code for nullptr should be removed. This will fix it: #52

[jandres742]

@bmyates : do you remember why it was changed intentionally to mbz in 1.1? if there was a reason at that moment, do we know if that requirement does not longer apply?

[bmyates]

Requirements are unchanged. Original customer report that motivated change:

zetCommandListAppendMetricQueryBegin has no hSignalEvent or phWaitEvents parameters whereas zetCommandListAppendMetricQueryEnd does. If zetCommandListAppendMetricQueryBegin is a barrier, the specification doesn't say so.
MetricQueryEnd should only have a signal and not wait. Doesn't hurt that wait is here but not useful. MetricQueryBegin doesn’t need a signal or wait. Need to be defined as barriers.

[jandres742]

changes merged in #52