Initial attempt at kernel for nonmaxsuppression (WORK IN PROGRESS)

Just grabbing what I did for gather. This will probably be something more
complicated with a few different kernels due to all the comparissions
and things that need to be vectorized

src/targets/gpu/jit/nonmaxsuppression.cpp

+/*
+ * The MIT License (MIT)
+ *
+ * Copyright (c) 2015-2022 Advanced Micro Devices, Inc. All rights reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to deal
+ * in the Software without restriction, including without limitation the rights
+ * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+ * copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
+ * THE SOFTWARE.
+ */
+#include <migraphx/gpu/compiler.hpp>
+#include <migraphx/make_op.hpp>
+#include <migraphx/gpu/context.hpp>
+
+#include <migraphx/gpu/compile_hip_code_object.hpp>
+#include <migraphx/gpu/compile_hip.hpp>
+
+namespace migraphx {
+inline namespace MIGRAPHX_INLINE_NS {
+namespace gpu {
+
+// NOLINTNEXTLINE
+static const char* const nonmaxsupression_kernel = R"__migraphx__(
+#include <migraphx/kernels/gather.hpp>
+#include <migraphx/kernels/ops.hpp>
+#include <migraphx/kernels/integral_constant.hpp>
+#include <migraphx/kernels/generic_constant.hpp>
+#include <args.hpp>
+
+namespace migraphx {
+
+extern "C" {
+
+__global__ void nonmaxsuppression_kernel(void* in_data, void* in_indices, void* output) 
+{
+    make_tensors()(in_data, in_indices, output)([](auto&&... xs) { 
+        gather<${axis}>(xs...); 
+    });
+}
+
+}
+
+} // namespace migraphx
+
+)__migraphx__";
+
+struct nonmaxsuppresion_compiler : compiler<nonmaxsuppression_compiler>
+{
+    std::vector<std::string> names() const { return {"nonmaxsuppression"}; }
+
+    operation compile_op(context& ctx, const std::vector<shape>& inputs, const value& v) const
+    {
+        hip_compile_options options;
+        const auto& out_s = inputs.back();
+        options.set_launch_params(v, compute_global_for(ctx, out_s.elements()));
+        options.inputs         = inputs;
+        options.output         = out_s;
+        options.kernel_name    = "nonmaxsuppression_kernel";
+        options.virtual_inputs = inputs;
+
+        auto axis = v.at("axis").to<std::string>();
+
+        auto src = interpolate_string(nonmaxsuppression_kernel, {{"axis", axis}});
+
+        return compile_hip_code_object(src, options);
+    }
+
+    compiler_replace compile(context& ctx, instruction_ref ins, const operation& op) const
+    {
+        return replace(compile_op(ctx, to_shapes(ins->inputs()), op.to_value()));
+    }
+};
+
+} // namespace gpu
+} // namespace MIGRAPHX_INLINE_NS
+} // namespace migraphx

src/targets/gpu/kernels/include/migraphx/kernels/nonmaxsuppression.hpp

+/*
+ * The MIT License (MIT)
+ *
+ * Copyright (c) 2015-2022 Advanced Micro Devices, Inc. All rights reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to deal
+ * in the Software without restriction, including without limitation the rights
+ * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+ * copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
+ * THE SOFTWARE.
+ */
+#ifndef MIGRAPHX_GUARD_KERNELS_NONMAXSUPPRESSION_HPP
+#define MIGRAPHX_GUARD_KERNELS_NONMAXSUPPRESSION_HPP
+
+#include <migraphx/kernels/index.hpp>
+#include <migraphx/kernels/shape.hpp>
+#include <migraphx/kernels/algorithm.hpp>
+#include <migraphx/kernels/tensor_view.hpp>
+
+namespace migraphx {
+
+template <int Axis, class Input, class Indices>
+constexpr auto gather_shape(Input input, Indices indices)
+{
+    auto lengths = input.lens;
+
+    lengths[Axis] = indices.elements();
+    return make_shape(lengths, input.strides);
+}
+
+template <int Axis, class Input, class Indices, class Output>
+__device__ void nonmaxsuppression(Input input, Indices indices, Output output)
+{
+    auto ind           = make_index();
+    auto axis_dim_size = input.get_shape().lens[Axis];
+
+    constexpr auto out_comp = nonmaxsuppression_shape<Axis>(get_shape_c<Input>{}, get_shape_c<Indices>{});
+
+    ind.global_stride(output.get_shape().elements(), [&](auto i) {
+        auto idx      = out_comp.multi(i);
+        auto in_index = indices[idx[Axis]];
+
+        auto new_in_index = (in_index < 0) ? in_index + axis_dim_size : in_index;
+
+        idx[Axis] = new_in_index;
+
+        output[i] = input[idx];
+    });
+}
+
+} // namespace migraphx
+#endif