Skip to content

Aligned AMD addrlib structures for 64-bit platforms

Herman Semenov requested to merge GermanAizek/mesa:align-structures-64bit into main

Aligned AMD addrlib structures for 64-bit platforms (move/copy texture addressing and alignment calculator structures optimization)

Diff sizes:

  • _ADDR_REGISTER_VALUE reduced 48 -> 40 bytes
  • _ADDR_CREATE_OUTPUT reduced 32 -> 24 bytes
  • _ADDR_COMPUTE_HTILE_ADDRFROMCOORD_OUTPUT reduced 24 -> 16 bytes
  • _ADDR_COMPUTE_HTILE_COORDFROMADDR_INPUT reduced 64 -> 56 bytes
  • _ADDR_COMPUTE_CMASK_ADDRFROMCOORD_OUTPUT reduced 24 -> 16 bytes
  • _ADDR_COMPUTE_CMASK_COORDFROMADDR_INPUT reduced 56 -> 48 bytes
  • _ADDR_COMPUTE_FMASK_ADDRFROMCOORD_OUTPUT reduced 24 -> 16 bytes
  • _ADDR_COMPUTE_FMASK_COORDFROMADDR_INPUT reduced 64 -> 56 bytes
  • _ADDR2_COMPUTE_CMASK_ADDRFROMCOORD_OUTPUT reduced 24 -> 16 bytes
  • _ADDR2_COMPUTE_CMASK_COORDFROMADDR_INPUT reduced 56 -> 48 bytes
  • _ADDR2_COMPUTE_FMASK_ADDRFROMCOORD_OUTPUT reduced 24 -> 16 bytes
  • _ADDR2_COMPUTE_SUBRESOURCE_OFFSET_FORSWIZZLEPATTERN_INPUT reduced 48 -> 40 bytes
  • _ADDR2_COMPUTE_NONBLOCKCOMPRESSEDVIEW_OUTPUT reduced 40 -> 32 bytes
  • _ADDR3_COMPUTE_SUBRESOURCE_OFFSET_FORSWIZZLEPATTERN_INPUT reduced 48 -> 40 bytes
Edited by Herman Semenov

Merge request reports