Add GatherND operator (#1089)

Add ref and gpu implementations for ONNX op GatherND

Resolves #1032

Author: turneram

src/CMakeLists.txt

@@ -109,6 +109,7 @@ register_migraphx_ops(
     flatten
     floor
     gather
+    gathernd
     get_tuple_elem
     greater
     gru

src/include/migraphx/op/gathernd.hpp

@@ -0,0 +1,131 @@
+#ifndef MIGRAPHX_GUARD_OPERATORS_GATHERND_HPP
+#define MIGRAPHX_GUARD_OPERATORS_GATHERND_HPP
+
+#include <migraphx/check_shapes.hpp>
+#include <migraphx/shape_for_each.hpp>
+#include <migraphx/par_for.hpp>
+
+namespace migraphx {
+inline namespace MIGRAPHX_INLINE_NS {
+namespace op {
+
+struct gathernd
+{
+    int batch_dims = 0;
+
+    template <class Self, class F>
+    static auto reflect(Self& self, F f)
+    {
+        return pack(f(self.batch_dims, "batch_dims"));
+    }
+
+    std::string name() const { return "gathernd"; }
+
+    shape compute_shape(std::vector<shape> inputs) const
+    {
+        check_shapes{inputs, *this}.has(2);
+        auto r = inputs.front().lens().size();
+        auto q = inputs.back().lens().size();
+        auto k = inputs.back().lens().back();
+        if(k > r - batch_dims)
+        {
+            MIGRAPHX_THROW("GATHERND: Indices of length " + std::to_string(k) +
+                           " cannot be used to access data of rank " +
+                           std::to_string(r - batch_dims));
+        }
+        auto indices_lens_iter = inputs.back().lens().begin();
+        auto output_lens_size  = q + r - k - batch_dims - 1;
+        std::vector<std::size_t> output_lens(output_lens_size);
+        std::copy(indices_lens_iter, indices_lens_iter + (q - 1), output_lens.begin());
+        if(k < r - batch_dims)
+        {
+            auto data_lens = inputs.front().lens();
+            std::copy(
+                data_lens.begin() + batch_dims + k, data_lens.end(), output_lens.begin() + q - 1);
+        }
+        shape output_shape{inputs.front().type(), output_lens};
+        return output_shape;
+    }
+
+    argument compute(const shape& output_shape, std::vector<argument> args) const
+    {
+        argument result{output_shape};
+        visit_all(result, args[0])([&](auto output, auto data) {
+            args[1].visit([&](auto indices) {
+                auto indices_shape        = indices.get_shape();
+                auto indices_shape_lens   = indices_shape.lens();
+                auto data_shape           = data.get_shape();
+                auto data_shape_lens      = data_shape.lens();
+                auto k                    = indices_shape.lens().back();
+                const auto num_slice_dims = k;
+                std::size_t num_slices    = std::accumulate(indices_shape_lens.begin(),
+                                                         indices_shape_lens.end() - 1,
+                                                         1,
+                                                         std::multiplies<std::size_t>());
+                std::size_t slice_size  = std::accumulate(data_shape_lens.begin() + k + batch_dims,
+                                                         data_shape_lens.end(),
+                                                         1,
+                                                         std::multiplies<std::size_t>());
+                std::size_t num_batches = std::accumulate(data_shape_lens.begin(),
+                                                          data_shape_lens.begin() + batch_dims,
+                                                          1,
+                                                          std::multiplies<std::size_t>());
+                std::size_t data_batch_stride =
+                    std::accumulate(data_shape_lens.begin() + batch_dims,
+                                    data_shape_lens.end(),
+                                    1,
+                                    std::multiplies<std::size_t>());
+                auto num_slices_per_batch = num_slices / num_batches;
+
+                std::vector<std::size_t> sizes_from_slice_dims(num_slice_dims);
+                {
+                    auto running_product = slice_size;
+                    for(std::size_t i = 0; i < num_slice_dims; ++i)
+                    {
+                        sizes_from_slice_dims[num_slice_dims - 1 - i] = running_product;
+                        running_product *= data_shape_lens[batch_dims + num_slice_dims - 1 - i];
+                    }
+                }
+
+                std::vector<std::size_t> input_slice_offsets(num_slices);
+                par_for(num_slices, [&](const auto i) {
+                    std::size_t batch_idx = i / num_slices_per_batch;
+
+                    auto slice_indices                = indices.begin() + (i * num_slice_dims);
+                    std::size_t relative_slice_offset = 0;
+                    for(size_t dim_idx = 0; dim_idx < num_slice_dims; ++dim_idx)
+                    {
+                        int64_t index                   = *(slice_indices + dim_idx);
+                        const std::size_t input_dim_idx = batch_dims + dim_idx;
+                        const auto input_dim            = data_shape_lens[input_dim_idx];
+                        if(index < -static_cast<int64_t>(input_dim) or
+                           index >= static_cast<int64_t>(input_dim))
+                            MIGRAPHX_THROW("GatherND: index " + std::to_string(index) +
+                                           " is out of bounds for dim of len " +
+                                           std::to_string(input_dim));
+                        if(index < 0)
+                            index += input_dim;
+
+                        relative_slice_offset += index * sizes_from_slice_dims[dim_idx];
+                    }
+
+                    input_slice_offsets[i] =
+                        (batch_idx * data_batch_stride) + relative_slice_offset;
+                });
+
+                par_for(num_slices * slice_size, [&](const auto i) {
+                    auto slice_offset = input_slice_offsets[i / slice_size];
+                    output[i]         = data[slice_offset + i % slice_size];
+                });
+            });
+        });
+
+        return result;
+    }
+};
+
+} // namespace op
+} // namespace MIGRAPHX_INLINE_NS
+} // namespace migraphx
+
+#endif

src/include/migraphx/operators.hpp

@@ -35,6 +35,7 @@
 #include <migraphx/op/flatten.hpp>
 #include <migraphx/op/floor.hpp>
 #include <migraphx/op/gather.hpp>
+#include <migraphx/op/gathernd.hpp>
 #include <migraphx/op/get_tuple_elem.hpp>
 #include <migraphx/op/greater.hpp>
 #include <migraphx/op/gru.hpp>

src/onnx/parse_generic_op.cpp

@@ -28,6 +28,7 @@ struct parse_generic_op : op_parser<parse_generic_op>
                 {"Flatten", "flatten"},
                 {"Floor", "floor"},
                 {"Gather", "gather"},
+                {"GatherND", "gathernd"},
                 {"Identity", "identity"},
                 {"IsNaN", "isnan"},
                 {"LeakyRelu", "leaky_relu"},

src/targets/gpu/jit/gathernd.cpp

@@ -0,0 +1,75 @@
+#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>
+#include <migraphx/ranges.hpp>
+#include <migraphx/reduce_dims.hpp>
+#include <migraphx/stringutils.hpp>
+#include <migraphx/dead_code_elimination.hpp>
+#include <migraphx/eliminate_common_subexpression.hpp>
+#include <migraphx/module.hpp>
+#include <migraphx/pass_manager.hpp>
+
+namespace migraphx {
+inline namespace MIGRAPHX_INLINE_NS {
+namespace gpu {
+
+// NOLINTNEXTLINE
+static const char* const gathernd_kernel = R"__migraphx__(
+#include <migraphx/kernels/gathernd.hpp>
+#include <migraphx/kernels/basic_ops.hpp>
+#include <migraphx/kernels/integral_constant.hpp>
+#include <migraphx/kernels/generic_constant.hpp>
+#include <args.hpp>
+
+namespace migraphx {
+
+extern "C" {
+
+__global__ void gathernd_kernel(void* in_data, void* in_indices, void* output) 
+{
+    make_tensors()(in_data, in_indices, output)([](auto&&... xs) { 
+        auto settings = make_gathernd_settings(MIGRAPHX_MAKE_CONSTANT(int64_t{BATCH_DIMS}));
+        gathernd(xs..., settings); 
+    });
+}
+
+}
+
+} // namespace migraphx
+
+)__migraphx__";
+
+struct gathernd_compiler : compiler<gathernd_compiler>
+{
+    std::vector<std::string> names() const { return {"gathernd"}; }
+
+    operation compile_op(context& ctx, const std::vector<shape>& inputs, const value& v) const
+    {
+        hip_compile_options options;
+        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    = "gathernd_kernel";
+        options.virtual_inputs = inputs;
+
+        // batch_dims
+        assert(v.contains("batch_dims"));
+        auto batch_dims = v.at("batch_dims").to<int64_t>();
+        options.params += " -DBATCH_DIMS=" + std::to_string(batch_dims);
+
+        return compile_hip_code_object(gathernd_kernel, 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/algorithm.hpp

@@ -21,6 +21,16 @@ struct greater
     }
 };
 
+template <class InputIt, class T, class BinaryOperation>
+constexpr T accumulate(InputIt first, InputIt last, T init, BinaryOperation op)
+{
+    for(; first != last; ++first)
+    {
+        init = op(std::move(init), *first);
+    }
+    return init;
+}
+
 template <class InputIt, class OutputIt>
 constexpr OutputIt copy(InputIt first, InputIt last, OutputIt d_first)
 {

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

@@ -0,0 +1,81 @@
+#ifndef MIGRAPHX_GUARD_KERNELS_GATHERND_HPP
+#define MIGRAPHX_GUARD_KERNELS_GATHERND_HPP
+
+#include <migraphx/kernels/index.hpp>
+#include <migraphx/kernels/algorithm.hpp>
+
+namespace migraphx {
+
+template <class T>
+struct gathernd_settings
+{
+    T batch_dims{};
+};
+
+template <class... Ts>
+constexpr gathernd_settings<Ts...> make_gathernd_settings(Ts... xs)
+{
+    return {xs...};
+}
+
+template <class T, class U, class V, class Settings>
+__device__ void gathernd(const T& data_t, const U& indices_t, const V& output_t, Settings s)
+{
+    auto ind           = make_index();
+    auto batch_dims    = s.batch_dims;
+    auto output_shape  = output_t.get_shape();
+    auto indices_shape = indices_t.get_shape();
+    auto data_shape    = data_t.get_shape();
+
+    auto indices_shape_lens = indices_shape.lens;
+    auto data_shape_lens    = data_shape.lens;
+    auto num_slice_dims     = indices_shape_lens.back();
+    std::size_t num_slices  = accumulate(indices_shape_lens.begin(),
+                                        indices_shape_lens.end() - 1,
+                                        1,
+                                        std::multiplies<std::size_t>());
+    std::size_t slice_size  = accumulate(data_shape_lens.begin() + num_slice_dims + batch_dims,
+                                        data_shape_lens.end(),
+                                        1,
+                                        std::multiplies<std::size_t>());
+    const std::size_t num_batches       = accumulate(data_shape_lens.begin(),
+                                               data_shape_lens.begin() + batch_dims,
+                                               1,
+                                               std::multiplies<std::size_t>());
+    const std::size_t data_batch_stride = accumulate(data_shape_lens.begin() + batch_dims,
+                                                     data_shape_lens.end(),
+                                                     1,
+                                                     std::multiplies<std::size_t>());
+    const auto num_slices_per_batch     = num_slices / num_batches;
+
+    ind.global_stride(output_shape.elements(), [&](auto i) {
+        const auto* indices_ptr     = indices_t.data();
+        const std::size_t j         = i / slice_size;
+        const std::size_t batch_idx = j / num_slices_per_batch;
+
+        auto* slice_indices               = indices_ptr + (j * num_slice_dims);
+        std::size_t relative_slice_offset = 0;
+        for(std::size_t idx = 0; idx < num_slice_dims; ++idx)
+        {
+            int64_t index                   = slice_indices[idx];
+            const std::size_t input_dim_idx = batch_dims + idx;
+            const auto input_dim            = data_shape_lens[input_dim_idx];
+            assert(index >= -static_cast<int64_t>(input_dim) and
+                   index < static_cast<int64_t>(input_dim));
+            if(index < 0)
+                index += input_dim;
+            std::size_t size_from_slice_dims =
+                accumulate(data_shape_lens.begin() + batch_dims + idx + 1,
+                           data_shape_lens.begin() + batch_dims + num_slice_dims,
+                           slice_size,
+                           std::multiplies<std::size_t>());
+            relative_slice_offset += index * size_from_slice_dims;
+        }
+
+        auto slice_offset = (batch_idx * data_batch_stride) + relative_slice_offset;
+        output_t[i]       = data_t[slice_offset + i % slice_size];
+    });
+}
+
+} // namespace migraphx
+#endif

test/onnx/gathernd_batch_dims_test.onnx

@@ -0,0 +1,19 @@
+gathernd_batch_dims_test:�

+/
+data
+indices
y"GatherND*
+
+batch_dims
�
gathernd_batch_dims_testZ
+data
+

+
+
+Z
+indices
+
+
+
b
+
y
+

+
+B

test/onnx/gathernd_test.onnx

@@ -0,0 +1,16 @@
+gathernd_test:q
+
+data
+indices
y"GatherNDgathernd_testZ
+data
+

+
+Z
+indices
+
+
+b
+
y
+
+

+B