Revert "REVERTME: modify 'tritongpu_to_llvm.mlir' until issue #2603 is fixed"

This reverts commit 0290b6c.

Author: victor-eds

test/Conversion/tritongpu_to_llvm.mlir

@@ -858,6 +858,8 @@ module attributes {"triton_gpu.num-ctas" = 1 : i32, "triton_gpu.num-warps" = 1 :
 module attributes {"triton_gpu.num-ctas" = 1 : i32, "triton_gpu.num-warps" = 4 : i32} {
   // CHECK-LABEL: convert_layout_slice_mmav2_blocked_reg
   tt.func @convert_layout_slice_mmav2_blocked_reg(%arg0: tensor<1xf16, #slice>) {
+    // CHECK-NOT: st.shared
+    // CHECK-NOT: llvm.load
     %0 = triton_gpu.convert_layout %arg0 : tensor<1xf16, #slice> -> tensor<1xf16, #blocked>
     tt.return
   }
@@ -901,6 +903,8 @@ module attributes {"triton_gpu.num-ctas" = 1 : i32, "triton_gpu.num-warps" = 4 :
 module attributes {"triton_gpu.num-ctas" = 1 : i32, "triton_gpu.num-warps" = 4 : i32} {
   // CHECK-LABEL: convert_layout_mmav3_mmav3_2
   tt.func @convert_layout_mmav3_mmav3_2(%arg0: tensor<16x16xf16, #mma1>) {
+    // CHECK-NOT: st.shared
+    // CHECK-NOT: llvm.load
     %0 = triton_gpu.convert_layout %arg0 : tensor<16x16xf16, #mma1> -> tensor<16x16xf16, #mma0>
     tt.return
   }
@@ -915,6 +919,7 @@ module attributes {"triton_gpu.num-ctas" = 1 : i32, "triton_gpu.num-warps" = 4 :
   // CHECK-LABEL: convert_layout_mmav3_mmav3_3
   tt.func @convert_layout_mmav3_mmav3_3(%arg0: tensor<1x64xf16, #mma1>) {
     // CHECK-NOT: st.shared
+    // CHECK-NOT: llvm.load
     %0 = triton_gpu.convert_layout %arg0 : tensor<1x64xf16, #mma1> -> tensor<1x64xf16, #mma0>
     tt.return
   }