Commit Graph

2 Commits

Author SHA1 Message Date
MrVallentin e20fdd83c4 Fixed repeated words 2018-11-01 17:51:07 +01:00
Jon Leech 894211de5f (The previous commit didn't actually include internal gitlab changes
since 1.1.89; fixed now).

Change log for October 28, 2018 Vulkan 1.1.90 spec update:

  * Update release number to 90.

Public Issues:

  * Tag flink:vkQueueWaitIdle as `externsync` in `vk.xml` (public pull
    request 815).
  * Update README (public pull request 834).
  * `VK_NV_framebuffer_mixed_samples` and `VK_AMD_mixed_attachment_samples`
    had confusing and contradictory valid usage statements when read in the
    all-extensions spec build. Change them to explicitly mention which
    extension each is for (public issue Vulkan-ValidationLayers/issues/353).

Internal Issues:

  * Update `COPYING.md` to clarify how externally generated Vulkan
    Specifications (for translations, annotations, or other reasons) must be
    copyrighted, and acknowledge the Exception Clause on the `vk.xml`
    license (internal issue 1079).
  * Specify that flink:vkGetPhysicalDeviceImageFormatProperties may: return
    pname:maxMipLevels 1 if the format is ycbcr (internal issue 1361).
  * Clarify previously underspecified language for
    flink:vkCmdPushConstants::pname:pStageFlags regarding use of push
    constants across multiple pipelines (internal issue 1403).
  * Fix typo in XML/headers for
    ename:VK_STRUCTURE_TYPE_IMAGE_DRM_FORMAT_MODIFIER_EXPLICIT_CREATE_INFO_EXT,
    which was previously
    etext:VK_STRUCTURE_TYPE_IMAGE_EXCPLICIT_DRM_FORMAT_MODIFIER_CREATE_INFO_EXT
    (internal issue 1428).
  * Fix markup of equations that were sporadically breaking the
    `optimize-pdf` step of PDF generation, due (apparently) to inconsistent
    treatment of unwrapped multicharacter terms by different LaTeX parsers
    (internal issue 1435).
  * For the <<memory-model-synchronizes-with synchronizes-with>> memory
    model relation cases involving a release barrier plus relaxed atomic
    write, treat the atomic as if it were a release atomic and allow the
    acquire side to read from its hypothetical release sequence. This is
    more consistent with how C++ defines synchronization for release fences
    (internal issue cross-api/memory-model#72).
  * Minor editorial changes to the <<memory-model, memory model>> appendix
    based on external feedback.
2018-10-28 23:53:18 -07:00