Use 64-bit math in work group size assertion.
authorKenneth Graunke <kenneth.w.graunke@intel.com>
Thu, 13 Jul 2017 01:44:11 +0000 (18:44 -0700)
committerKenneth Graunke <kenneth.w.graunke@intel.com>
Fri, 8 Sep 2017 21:02:14 +0000 (14:02 -0700)
commitfa9244bce9608c01e4dfe49d1e1c108b9d3bdd65
treea83601875dd6826f3e12caaa55acdbe5e8a5d7b2
parent2cf32ffa6ee9daf7f58e34c1b874f4005e4b888f
Use 64-bit math in work group size assertion.

If the maximum workgroup size in each dimension is large enough, this
can overflow, causing the comparison to fail.  For example, on Intel
Haswell, we get:

   assert(2048 * 2048 * 2048 > 2048) -> assert(0 > 2048) -> fail

Casting one value to deInt64 causes the multiplications to occur as
64-bit values, so we don't overflow.

Affects:

dEQP-GLES31.functional.debug.negative_coverage.*.compute.invalid_maximum_work_group_sizes

Components: AOSP

Change-Id: Ie03627a378f4fdc46ce241bbdd3693ae829c217d
modules/gles31/functional/es31fNegativeComputeTests.cpp