Add recip op for gpu (#488)

* add recip gpu and tests

* formatting

* remove to_hip_type
Co-authored-by: mvermeulen <5479696+mvermeulen@users.noreply.github.com>

src/onnx/onnx.cpp

@@ -60,6 +60,7 @@ struct onnx_parser
         add_generic_op("Log", op::log{});
         add_generic_op("Floor", op::floor{});
         add_generic_op("Identity", op::identity{});
+        add_generic_op("Reciprocal", op::recip{});
         add_generic_op("Relu", op::relu{});
         add_generic_op("Round", op::round{});
         add_generic_op("Sigmoid", op::sigmoid{});

src/targets/gpu/CMakeLists.txt

@@ -47,6 +47,7 @@ add_library(migraphx_device
     device/pad.cpp
     device/pow.cpp
     device/prelu.cpp
+    device/recip.cpp
     device/reduce_max.cpp
     device/reduce_mean.cpp
     device/reduce_min.cpp

src/targets/gpu/device/recip.cpp

+#include <migraphx/gpu/device/recip.hpp>
+#include <migraphx/gpu/device/nary.hpp>
+#include <migraphx/gpu/device/types.hpp>
+
+namespace migraphx {
+inline namespace MIGRAPHX_INLINE_NS {
+namespace gpu {
+namespace device {
+
+void recip(hipStream_t stream, const argument& result, const argument& arg)
+{
+    nary(stream, result, arg)([](auto x) __device__ { return 1 / x; });
+}
+
+} // namespace device
+} // namespace gpu
+} // namespace MIGRAPHX_INLINE_NS
+} // namespace migraphx

src/targets/gpu/include/migraphx/gpu/device/recip.hpp

+#ifndef MIGRAPHX_GUARD_RTGLIB_DEVICE_RECIP_HPP
+#define MIGRAPHX_GUARD_RTGLIB_DEVICE_RECIP_HPP
+
+#include <migraphx/argument.hpp>
+#include <migraphx/config.hpp>
+#include <hip/hip_runtime_api.h>
+
+namespace migraphx {
+inline namespace MIGRAPHX_INLINE_NS {
+namespace gpu {
+namespace device {
+
+void recip(hipStream_t stream, const argument& result, const argument& arg);
+
+} // namespace device
+} // namespace gpu
+} // namespace MIGRAPHX_INLINE_NS
+} // namespace migraphx
+
+#endif

src/targets/gpu/include/migraphx/gpu/recip.hpp

+#ifndef MIGRAPHX_GUARD_RTGLIB_RECIP_HPP
+#define MIGRAPHX_GUARD_RTGLIB_RECIP_HPP
+
+#include <migraphx/gpu/oper.hpp>
+#include <migraphx/gpu/device/recip.hpp>
+
+namespace migraphx {
+inline namespace MIGRAPHX_INLINE_NS {
+namespace gpu {
+
+struct hip_recip : unary_device<hip_recip, device::recip>
+{
+};
+
+} // namespace gpu
+} // namespace MIGRAPHX_INLINE_NS
+} // namespace migraphx
+
+#endif

src/targets/gpu/lowering.cpp

@@ -71,6 +71,7 @@
 #include <migraphx/gpu/sqdiff.hpp>
 #include <migraphx/gpu/int8_conv_pack.hpp>
 #include <migraphx/gpu/prelu.hpp>
+#include <migraphx/gpu/recip.hpp>
 #include <utility>
 #include <functional>
 #include <algorithm>
@@ -166,6 +167,7 @@ struct miopen_apply
         add_generic_op<hip_sigmoid>("sigmoid");
         add_generic_op<hip_ceil>("ceil");
         add_generic_op<hip_floor>("floor");
+        add_generic_op<hip_recip>("recip");
 
         add_extend_op<miopen_contiguous, op::contiguous>("contiguous");
         add_extend_op<hip_concat, op::concat>("concat");

test/cpu_ops_test.cpp

@@ -2419,4 +2419,19 @@ TEST_CASE(op_capture)
     EXPECT(migraphx::verify_range(vec, cap_vec));
 }
 
+TEST_CASE(recip_test)
+{
+    migraphx::program p;
+    migraphx::shape s{migraphx::shape::double_type, {3}};
+    std::vector<float> data{-0.5f, 0.1f, 0.5f};
+    auto l = p.add_literal(migraphx::literal{s, data});
+    p.add_instruction(migraphx::op::recip{}, l);
+    p.compile(migraphx::cpu::target{});
+    auto result = p.eval({}).back();
+    std::vector<float> results_vector(3);
+    result.visit([&](auto output) { results_vector.assign(output.begin(), output.end()); });
+    std::vector<float> gold = {-2.0f, 10.0f, 2.0f};
+    EXPECT(migraphx::verify_range(results_vector, gold));
+}
+
 int main(int argc, const char* argv[]) { test::run(argc, argv); }

test/gpu/ops_test.cpp

@@ -4473,4 +4473,16 @@ struct test_convert : verify_program<test_convert>
     };
 };
 
+struct test_recip : verify_program<test_recip>
+{
+    migraphx::program create_program() const
+    {
+        migraphx::program p;
+        migraphx::shape s{migraphx::shape::double_type, {3}};
+        auto x = p.add_parameter("x", s);
+        p.add_instruction(migraphx::op::recip{}, x);
+        return p;
+    }
+};
+
 int main(int argc, const char* argv[]) { test::run(argc, argv); }

test/onnx/gen_onnx.py

@@ -1455,6 +1455,20 @@ def prelu_brcst_test():
     return ([node], [arg0, arg1], [arg_out])
 
 
+@onnx_test
+def recip_test():
+    x = helper.make_tensor_value_info('x', TensorProto.FLOAT, [3])
+    y = helper.make_tensor_value_info('y', TensorProto.FLOAT, [3])
+
+    node = onnx.helper.make_node(
+        'Reciprocal',
+        inputs=['x'],
+        outputs=['y'],
+    )
+
+    return ([node], [x], [y])
+
+
 @onnx_test
 def reducel1_test():
     x = helper.make_tensor_value_info('x', TensorProto.FLOAT, [3, 4, 5, 6])

test/onnx/onnx_test.cpp

@@ -1133,6 +1133,17 @@ TEST_CASE(prelu_brcst_test)
     EXPECT(p == prog);
 }
 
+TEST_CASE(recip_test)
+{
+    migraphx::program p;
+    auto input = p.add_parameter("x", migraphx::shape{migraphx::shape::float_type, {3}});
+    p.add_instruction(migraphx::op::recip{}, input);
+
+    auto prog = optimize_onnx("recip_test.onnx");
+
+    EXPECT(p == prog);
+}
+
 TEST_CASE(reducel1_test)
 {
     migraphx::program p;

test/onnx/recip_test.onnx

+
+recip_test:B
+
+
x
y"
+Reciprocal
+recip_testZ
+
x
+
+

+b
+
y
+
+

+B
\ No newline at end of file