[clang] [llvm] [WIP][TargetParser][clang] Parse and store Intel IGCA virtual architecture as an OffloadArch (PR #227468)

[Nikita Kornev via cfe-commits]

================
@@ -92,5 +102,20 @@ INTEL_GPU_COMPAT("xe-bmg", XE_BMG, 40, Render)
 INTEL_GPU_COMPAT("xe-mtl", XE_MTL, 30, Render)
 INTEL_GPU_COMPAT("xe-dg2", XE_DG2, 15, RenderExact)
 
+// IGCA target levels. Only targets defined here are valid.
+#ifndef INTEL_IGCA_TARGET
----------------
KornevNikita wrote:

The new macros don't add anything new, they just duplicate the info from the table above. I believe we should not maintain two sources containing the same info. You could change the original table to contain `c`, `r`, `ca` & `ra` rather than `Compute`, `Render`, etc, I believe that could help to make the logic simpler.

Also I don't see yet why the logic would become complex, could you clarify?

https://github.com/llvm/llvm-project/pull/227468