commit 6044784ca3bb538503f604ff3643447feaf5f5c5
author Slawomir Cygan <slawomir.cygan@intel.com> Mon Nov 14 17:57:31 2016 +0100
committer Pyry Haulos <phaulos@google.com> Thu Nov 17 13:26:51 2016 -0500
tree cfe45e6be49ba4ca75419efc5873afdfd583ad11
parent 096528efbd7a10c2d88b237a4995837bd4fc2f26

Fix race condition between query pool reset and accessing query results

Affects tests:
dEQP-VK.query_pool.occlusion_query.get_results*(wait_query|wait_none)*

These tests access query pool results on host, without any synchronization
with command buffer which performs the query - they either use
VK_QUERY_RESULT_WAIT_BIT or allow for incomplete results to be captured.

This may have a race condition, mentioned by spec:

"""
Applications must take care to ensure that use of the
VK_QUERY_RESULT_WAIT_BIT bit has the desired effect. For example, if
a query has been used previously and a command buffer records the
commands vkCmdResetQueryPool, vkCmdBeginQuery, and vkCmdEndQuery for
that query, then the query will remain in the available state until the
 vkCmdResetQueryPool command executes on a queue.
"""

In affected tests, this change moves vkCmdResetQueryPool to separate
command buffer that is sumitted and completed before query is used on
device & host.

Additionally code handling RESULTS_MODE_COPY mode is refactored to use
new haveSeparateCopyCmdBuf() function instead of tons of flags and conditions.

Change-Id: Id4a538b912030f2a86b09593fecaae37f1819a71