Optimize Transposed Convolution for CL backend (FP32/16)

This patch optimizes transposed convolution for CL backend by rewriting it in a single kernel instead of three (flip_kernel + upsample + conv). The new kernel skips the upsampling step which reduces the input space of convolution by stride_x * stride_y, resulting in significant performance improvement. It also skips the kernel flipping by traversing the weights accordingly, thus reduces the memory footprint.

Resolves: COMPMID-5676
Signed-off-by: Gunes Bayir <gunes.bayir@arm.com>
Change-Id: I8a333212dc7c5f7f0597aa58b0d56d44814baa14
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/8588
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Gian Marco Iodice <gianmarco.iodice@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Benchmark: Arm Jenkins <bsgcomp@arm.com>

1 files changed, 15 insertions, 1 deletions

diff --git a/tests/datasets/ShapeDatasets.h b/tests/datasets/ShapeDatasets.h
index f0ad9fa693..e4277a981e 100644
--- a/tests/datasets/ShapeDatasets.h
+++ b/tests/datasets/ShapeDatasets.h
@@ -549,7 +549,7 @@ public:
     {
         TensorShape{ 5U, 5U, 7U, 4U, 3U },
                      TensorShape{ 5U, 5U, 4U, 13U, 2U },
-                     TensorShape{ 5U, 5U, 3U, 5U , 2U},
+                     TensorShape{ 5U, 5U, 3U, 5U, 2U },
     })
     {
     }
@@ -718,6 +718,7 @@ public:
     SmallDeconvolutionShapes()
         : ShapeDataset("InputShape",
     {
+        // Multiple Vector Loops for FP32
         TensorShape{ 5U, 4U, 3U, 2U },
                      TensorShape{ 5U, 5U, 3U },
                      TensorShape{ 11U, 13U, 4U, 3U }
@@ -726,6 +727,19 @@ public:
     }
 };
 
+class SmallDeconvolutionShapesWithLargerChannels final : public ShapeDataset
+{
+public:
+    SmallDeconvolutionShapesWithLargerChannels()
+        : ShapeDataset("InputShape",
+    {
+        // Multiple Vector Loops for all data types
+        TensorShape{ 5U, 5U, 35U }
+    })
+    {
+    }
+};
+
 /** Data set containing tiny tensor shapes for direct convolution. */
 class TinyDirectConvolutionShapes final : public ShapeDataset
 {