examples/BuddyConvolution/:

linalg-pooling-nhwc-max.mlir: Linalg Dialect handwritten mlir;
conv2d-nhwc-fhwc-vec.mlir: Vector Dialect handwritten mlir file;
makefile update related commands.

midend/lib/Conversion/ConvVectorizationn/: PoolingNhwcMaxVectorization.cpp implements vectorisation.

tests/Conversion/:
pooling-nhwc-max-vectorisation.mlir: a test file.

Before vectorisation optimization: 1.71661e-05, after vectorization optimization: 1.09673e-05 (vectorization size is 32), speedup ratio is 0.454. The amount of data in the example is small and the optimization effect is not obvious.

examples/BuddyNext/pooling-nhwc-max-vec.mlir

@@ -40,6 +40,7 @@ module {
     %dim_4_upbound_tmp = arith.subi %dim_4, %vl_step : index
     %dim_4_upbound = arith.addi %dim_4_upbound_tmp, %c1 : index
 
+    %t_start = call @rtclock() : () -> f64
     affine.for %arg3 = #map(%c0) to #map(%dim_1) {
       affine.for %arg4 = #map(%c0) to #map(%dim_2) {
         affine.for %arg5 = #map(%c0) to #map(%dim_3) {
@@ -83,6 +84,14 @@ module {
         }
       }
     }
+    %t_end = call @rtclock() : () -> f64
+    %time = arith.subf %t_end, %t_start : f64
+    %printed_output = memref.cast %arg2 : memref<?x?x?x?xf32> to memref<*xf32>
+    call @printMemrefF32(%printed_output) : (memref<*xf32>) -> ()
+
+    // Print timings.
+    vector.print %time : f64
+
     return
   }
 
@@ -105,20 +114,14 @@ module {
     linalg.fill ins(%cf1_32 : f32) outs(%b : memref<?x?xf32>)
     linalg.fill ins(%cf1_32 : f32) outs(%c : memref<?x?x?x?xf32>)
 
-    %t0 = call @rtclock() : () -> f64
-    call @pooling_nhwc_max(%a, %b, %c) : (memref<?x?x?x?xf32>, memref<?x?xf32>, memref<?x?x?x?xf32>) -> ()
-    %t1 = call @rtclock() : () -> f64
     // All the elements of the MemRef are the same,
     // only check the first line to verify the correctness.
     // CHECK: Unranked Memref
     // CHECK: [
     // CHECK: [
     // CHECK: [
     // CHECK: [1{{(, 1)*}}],
-    %print_C = memref.cast %c : memref<?x?x?x?xf32> to memref<*xf32>
-    call @printMemrefF32(%print_C) : (memref<*xf32>) -> ()
-    %time = arith.subf %t1, %t0 : f64
-    vector.print %time : f64
+    call @pooling_nhwc_max(%a, %b, %c) : (memref<?x?x?x?xf32>, memref<?x?xf32>, memref<?x?x?x?xf32>) -> ()
 
     memref.dealloc %c : memref<?x?x?x?xf32>
     memref.dealloc %b : memref<?x?xf32>

examples/MLIRLinalg/linalg-pooling-nhwc-max.mlir

@@ -18,9 +18,20 @@ module{
   func.func private @printMemrefF32(memref<*xf32>)
 
   func.func @pooling_nhwc_max(%a : memref<?x?x?x?xf32>, %b : memref<?x?xf32>, %c : memref<?x?x?x?xf32>) {
+    %t_start = call @rtclock() : () -> f64
+
     linalg.pooling_nhwc_max {dilations = dense<1> : vector<2xi64>, strides = dense<2> : vector<2xi64>} 
       ins(%a, %b : memref<?x?x?x?xf32>, memref<?x?xf32>) 
       outs(%c : memref<?x?x?x?xf32>)
+
+    %t_end = call @rtclock() : () -> f64
+    %time = arith.subf %t_end, %t_start : f64
+    %printed_output = memref.cast %c : memref<?x?x?x?xf32> to memref<*xf32>
+    call @printMemrefF32(%printed_output) : (memref<*xf32>) -> ()
+
+    // Print timings.
+    vector.print %time : f64
+
     return
   }
 
@@ -68,21 +79,15 @@ module{
     %v1 = call @alloc2_f32(%c2, %c2, %f0) : (index, index, f32) -> memref<?x?xf32>
     %v2 = call @alloc_f32(%c1, %c12, %c12, %c6, %f0) : (index, index, index, index, f32) -> memref<?x?x?x?xf32>
 
-    %t0 = call @rtclock() : () -> f64
-    call @pooling_nhwc_max(%v0, %v1, %v2) : (memref<?x?x?x?xf32>, memref<?x?xf32>, memref<?x?x?x?xf32>) -> ()
-    %t1 = call @rtclock() : () -> f64
     // All the elements of the MemRef are the same,
     // only check the first line to verify the correctness.
     // CHECK: Unranked Memref
     // CHECK: [
     // CHECK: [
     // CHECK: [
     // CHECK: [1{{(, 1)*}}],
-    %print_C = memref.cast %v2 : memref<?x?x?x?xf32> to memref<*xf32>
-    call @printMemrefF32(%print_C) : (memref<*xf32>) -> ()
-    %time = arith.subf %t1, %t0 : f64
-    vector.print %time : f64
-
+    call @pooling_nhwc_max(%v0, %v1, %v2) : (memref<?x?x?x?xf32>, memref<?x?xf32>, memref<?x?x?x?xf32>) -> ()
+
     memref.dealloc %v0 : memref<?x?x?x?xf32>
     memref.dealloc %v1 : memref<?x?xf32>
     memref.dealloc %v2 : memref<?x?x?x?xf32>