[Triton] Add SCCP to the end of make_ttgir (#7069)

Speculative placement of the pass. It needs to run after some IR
simplification but before canonicalizer, typically.

Author: Mogball

third_party/nvidia/backend/compiler.py

@@ -291,6 +291,7 @@ def make_ttgir(mod, metadata, opt, capability):
         if capability // 10 >= 9:
             nvidia.passes.ttnvgpuir.add_tma_lowering(pm)
             nvidia.passes.ttnvgpuir.add_fence_insertion(pm)
+        passes.common.add_sccp(pm)
         passes.common.add_canonicalizer(pm)
         pm.run(mod)
         metadata["cluster_dims"] = (cluster_info.clusterDimX, cluster_info.clusterDimY, cluster_info.clusterDimZ)