|
| 1 | +--- |
| 2 | +title: 2026-03-12 - HLSL Working Group Minutes |
| 3 | +--- |
| 4 | + |
| 5 | +* Discussion topics |
| 6 | + * @s-perron mentioned that [llvm/offload-test-suite#756] has been merged, |
| 7 | + adding depth texture support for the test suite for SPIR-V. |
| 8 | + * @icohedron mentioned [llvm/wg-hlsl#394], a proposal to move HLSL library |
| 9 | + function definitions into tablegen, rather than implementing the header |
| 10 | + entirely by hand. |
| 11 | + * @farzonl pointed out that we're getting quite a few PRs from folks outside |
| 12 | + of the meeting, and reminded people to watch for these so that they get |
| 13 | + good and prompt reviews. |
| 14 | + * @bogner pointed out that the vector elementwise-alignment datalayout |
| 15 | + specifier is now available as of [llvm/llvm-project#180617]. He's |
| 16 | + implementing this for DirectX but would like help for the SPIR-V side. |
| 17 | + * @s-perron about the status around [llvm/llvm-project#183367]. @bogner |
| 18 | + pointed out that we'd like to be consistent between DirectX and SPIR-V here |
| 19 | + if possible, but if not we should follow @s-perron's suggestion to convert |
| 20 | + the "unbounded" value when processing the intrinsic itself for SPIR-V. |
| 21 | + * @s-perron reminded everyone that when writing Vulkan specific tests, the |
| 22 | + flags to `spirv-val` need to be set appropriately so that it doesn't run |
| 23 | + its validation under the wrong environment. |
| 24 | + |
| 25 | +[llvm/offload-test-suite#756]: |
| 26 | + https://github.com/llvm/offload-test-suite/pull/756 |
| 27 | +[llvm/wg-hlsl#394]: |
| 28 | + https://github.com/llvm/wg-hlsl/pull/394 |
| 29 | +[llvm/llvm-project#180617]: |
| 30 | + https://github.com/llvm/llvm-project/pull/180617 |
| 31 | +[llvm/llvm-project#183367]: |
| 32 | + https://github.com/llvm/llvm-project/issues/183367 |
0 commit comments