Skip to content
GitLab
Projects Groups Snippets
  • /
  • Help
    • Help
    • Support
    • Community forum
    • Submit feedback
    • Contribute to GitLab
  • Sign in / Register
  • mesa mesa
  • Project information
    • Project information
    • Activity
    • Labels
    • Members
  • Repository
    • Repository
    • Files
    • Commits
    • Branches
    • Tags
    • Contributors
    • Graph
    • Compare
  • Issues 3,055
    • Issues 3,055
    • List
    • Boards
    • Service Desk
    • Milestones
  • Merge requests 1,003
    • Merge requests 1,003
  • CI/CD
    • CI/CD
    • Pipelines
    • Jobs
    • Schedules
  • Deployments
    • Deployments
    • Releases
  • Packages and registries
    • Packages and registries
    • Container Registry
  • Analytics
    • Analytics
    • Value stream
    • CI/CD
    • Repository
  • Activity
  • Graph
  • Create a new issue
  • Jobs
  • Commits
  • Issue Boards
Collapse sidebar
  • MesaMesa
  • mesamesa
  • Issues
  • #3208
Closed
Open
Issue created Jun 30, 2020 by Emma Anholt@anholtOwner

ir3_cf breaks nir_op_fquantize2f16

ir3_cf's "don't optimize out f2f32(f2f16(x))" thing doesn't work because cf will first push the f2f16 into whatever produced x, so we then we have (for example) cov.f16f32(add.f hr, z, w)) and ir3_cf can't detect it any more. I could disable optimization out of cov.f16f32, but I'm dubious of this pass even needing to exist: I went looking to see what work cf is doing for us on my shader-db, and it seems to all be "convert our bary instructions to half". This feels like work that should be done at the GLSL and NIR levels.

Testcase: dEQP-VK.spirv_assembly.instruction.graphics.opquantize.too_small_vert

To upload designs, you'll need to enable LFS and have an admin enable hashed storage. More information
Assignee
Assign to
Time tracking