update gpu pipeline

Author: tkarna

Diff for: src/jit/mlir.cpp

@@ -727,43 +727,46 @@ static const std::string gpu_pipeline =
     "linalg-fuse-elementwise-ops,"
     "arith-expand,"
     "memref-expand,"
-    "arith-bufferize,"
-    "func-bufferize,"
     "func.func(empty-tensor-to-alloc-tensor),"
-    "func.func(scf-bufferize),"
-    "func.func(tensor-bufferize),"
-    "func.func(bufferization-bufferize),"
-    "func.func(linalg-bufferize),"
-    "func.func(linalg-detensorize),"
-    "func.func(tensor-bufferize),"
+    "func.func(tile-for-gpu{tile-sizes=32 in-regions}),"
+    "func.func(tile-for-gpu{tile-sizes=1 in-regions}),"
     "region-bufferize,"
     "canonicalize,"
-    "func.func(finalizing-bufferize),"
+    "one-shot-bufferize,"
+    "cse,"
+    "canonicalize,"
+    "scf-forall-to-parallel,"
+    "cse,"
+    "canonicalize,"
     "imex-remove-temporaries,"
-    "func.func(convert-linalg-to-parallel-loops),"
-    "func.func(scf-parallel-loop-fusion),"
-    // is add-outer-parallel-loop needed?
-    "func.func(imex-add-outer-parallel-loop),"
+    "buffer-deallocation-pipeline,"
+    "func.func(convert-linalg-to-loops),"
     "func.func(gpu-map-parallel-loops),"
-    "func.func(convert-parallel-loops-to-gpu),"
+    "convert-parallel-loops-to-gpu,"
+    "canonicalize,"
+    "cse,"
     "func.func(insert-gpu-allocs{in-regions=1}),"
     "func.func(insert-gpu-copy),"
     "drop-regions,"
     "canonicalize,"
-    "func.func(lower-affine),"
     "gpu-kernel-outlining,"
+    "convert-scf-to-cf,"
+    "convert-cf-to-llvm,"
     "canonicalize,"
     "cse,"
-    "gpu.module(strip-debuginfo,convert-gpu-to-nvvm),"
-    "nvvm-attach-target,"
+    "gpu.module(strip-debuginfo,"
+    "convert-gpu-to-nvvm),"
+    "nvvm-attach-target{chip=sm_80 O=3},"
     "func.func(gpu-async-region),"
     "expand-strided-metadata,"
     "lower-affine,"
     "gpu-to-llvm,"
-    "gpu-module-to-binary{format=fatbin},"
     "convert-func-to-llvm,"
     "convert-math-to-llvm,"
     "finalize-memref-to-llvm,"
+    "canonicalize,"
+    "cse,"
+    "gpu-module-to-binary{format=fatbin},"
     "reconcile-unrealized-casts";
 
 const std::string _passes(get_text_env("SHARPY_PASSES"));