From 6c95c2dd574ebc3217c949a17016eb071935bc3b Mon Sep 17 00:00:00 2001
From: Georgios Pinitas <georgios.pinitas@arm.com>
Date: Mon, 20 Aug 2018 16:06:58 +0100
Subject: COMPMID-1188: Static tuning of CLScale

Change-Id: Icf1cc00d9861fdb8766d0b8fd33ca90833863927
Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/144830
Reviewed-by: Anthony Barbier <anthony.barbier@arm.com>
Tested-by: Jenkins <bsgcomp@arm.com>
---
 src/runtime/CL/tuners/BifrostTuner.cpp | 32 ++++++++++++++++++++++++++++++++
 1 file changed, 32 insertions(+)

(limited to 'src/runtime/CL/tuners/BifrostTuner.cpp')

diff --git a/src/runtime/CL/tuners/BifrostTuner.cpp b/src/runtime/CL/tuners/BifrostTuner.cpp
index fa67710cc8..2d52f3392e 100644
--- a/src/runtime/CL/tuners/BifrostTuner.cpp
+++ b/src/runtime/CL/tuners/BifrostTuner.cpp
@@ -249,6 +249,34 @@ void tune_pooling_kernel(CLPoolingLayerKernel &k)
 
     k.set_lws_hint(lws_hint);
 }
+
+void tune_scale_kernel(CLScaleKernel &k)
+{
+    cl::NDRange               lws_hint      = k.lws_hint();
+    const GPUTarget           gpu_target    = k.get_target();
+    const DataType            dt            = k.input()->info()->data_type();
+    const InterpolationPolicy interpolation = k._interpolationPolicy;
+
+    // Configure the local work size for Bifrost, interpolation (bilinear) and datatype F32.
+    // The value are obtained via exhaustive autotuning.
+    if(gpu_target_is_in(gpu_target, GPUTarget::G71, GPUTarget::G72) && (dt == DataType::F32) && (interpolation == InterpolationPolicy::BILINEAR))
+    {
+        auto dim_0 = k.output()->info()->dimension(0);
+        if(dim_0 == 480)
+        {
+            lws_hint = cl::NDRange(2, 1);
+        }
+        else if(dim_0 == 3120)
+        {
+            lws_hint = cl::NDRange(2, 8);
+        }
+        else if(dim_0 == 4160)
+        {
+            lws_hint = cl::NDRange(4, 8);
+        }
+        k.set_lws_hint(lws_hint);
+    }
+}
 } // namespace
 
 void BifrostTuner::tune_kernel_static(ICLKernel &kernel)
@@ -281,6 +309,10 @@ void BifrostTuner::tune_kernel_static(ICLKernel &kernel)
     {
         tune_pooling_kernel(*utils::cast::polymorphic_downcast<CLPoolingLayerKernel *>(&kernel));
     }
+    else if(dynamic_cast<CLScaleKernel *>(&kernel) != nullptr)
+    {
+        tune_scale_kernel(*utils::cast::polymorphic_downcast<CLScaleKernel *>(&kernel));
+    }
 }
 
 void BifrostTuner::tune_kernel_dynamic(ICLKernel &kernel)
-- 
cgit v1.2.1