Drop GPUCompiler branch pin now that 2.0 is released

Remove the `[sources]` pin to the `tb/compilercaching` branch; the caching
API it relied on shipped in the registered GPUCompiler 2.0 (compat is
already `"2"`).

Adapt `compile_or_lookup` to the released `cached_results` API: it now
returns `nothing` until code exists for the job (rather than always
materializing an empty results struct), so re-fetch after compiling —
which populates Julia's code cache — via `@something`. Matches the final
form merged in JuliaGPU/OpenCL.jl#431.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

Author: vchuravy

Project.toml

@@ -22,9 +22,6 @@ StaticArrays = "90137ffa-7385-5640-81b9-e52037218182"
 UUIDs = "cf7118a7-6976-5b1a-9a39-7adc72f591a4"
 pocl_standalone_jll = "54f56a70-6062-5590-a942-1226658f6c83"
 
-[sources]
-GPUCompiler = {url = "https://github.com/JuliaGPU/GPUCompiler.jl", rev = "tb/compilercaching"}
-
 [weakdeps]
 EnzymeCore = "f151be2c-9106-41f4-ab19-57ee4f262869"
 LinearAlgebra = "37e2e46d-f89d-539d-b4ee-838fcccc9c8e"

src/pocl/compiler/execution.jl

@@ -233,13 +233,16 @@ end
 # by `GPUCompiler.cached_results` (Julia's integrated code cache on 1.11+, which also
 # persists artifacts through precompilation; a session-local store on 1.10).
 #
-# `obj === nothing` identifies an `OpenCLResults` that hasn't been compiled yet. The
-# `compile_hook` check additionally forces the compile path so reflection-style
-# consumers (`@device_code_*`) observe the compilation even on a cache hit.
+# `cached_results` returns `nothing` until code exists for the job; `obj === nothing`
+# then identifies an `OpenCLResults` that hasn't been compiled yet. Compiling populates
+# Julia's code cache, so the post-compile `cached_results` re-fetch is guaranteed to
+# succeed. The `compile_hook` check additionally forces the compile path so
+# reflection-style consumers (`@device_code_*`) observe the compilation even on a hit.
 function compile_or_lookup(@nospecialize(job::CompilerJob))::OpenCLResults
     res = GPUCompiler.cached_results(OpenCLResults, job)
-    if res.obj === nothing || GPUCompiler.compile_hook[] !== nothing
+    if res === nothing || res.obj === nothing || GPUCompiler.compile_hook[] !== nothing
         compiled = compile_to_obj(job)
+        res = @something res GPUCompiler.cached_results(OpenCLResults, job)
         res.obj = compiled.obj
         res.entry = compiled.entry
         res.device_rng = compiled.device_rng