[clang] [CIR][CUDA] Support built-in CUDA surface type (PR #196079)

Orphic Dreamer via cfe-commits cfe-commits at lists.llvm.org
Mon Jun 15 18:21:44 PDT 2026


Siya-05 wrote:

The surface type lowering itself appears to work, but the test hits existing NYIs in CUDA runtime registration:

- CIRGenNVCUDARuntime::handleVarRegistration (Surface and Texture registration), 
- LoweringPreparePass (Surface registration NYI)

OG CodeGen implements these through registerDeviceSurf / registerDeviceTex, but CIR currently has no corresponding implementation.

Should support for surface registration be added as part of this PR?

https://github.com/llvm/llvm-project/pull/196079


More information about the cfe-commits mailing list