Add new transformations to the pass pipeline (#4472)

Adding new transformations to the Intel pass pipeline (already used by
the NVidia pass pipeline)

Signed-off-by: Tiotto, Ettore <[email protected]>

Author: etiotto

third_party/intel/backend/compiler.py

@@ -287,6 +287,12 @@ def make_ttgir(mod, metadata, opt, properties):
             intel.passes.ttgpuir.add_reduce_variable_liveness(pm)
 
         passes.ttgpuir.add_fuse_nested_loops(pm)
+
+        passes.common.add_canonicalizer(pm)
+        passes.ttir.add_triton_licm(pm)
+        passes.common.add_canonicalizer(pm)
+        passes.ttgpuir.add_combine_tensor_select_and_if(pm)
+
         passes.ttgpuir.add_optimize_thread_locality(pm)
         passes.ttgpuir.add_optimize_dot_operands(pm, True)
         passes.common.add_cse(pm)