Use half HIP APIs to compute max and min (#1764)

* use half hip functions to compute max and min
* add verify test for min and max

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

@@ -161,8 +161,7 @@ MIGRAPHX_DEVICE_MATH_HALF(fmod, ::fmod)
 // Map math functions to hip half2 functions
 // The half2 type is defined in include/hip/amd_detail/hip_fp16_gcc.h and is 2 16-bit floats
 // packed into a 32-bit number.  See include/hip/amd_detail/hip_fp16_math_fwd.h for the HIP names
-// Most but not all of these math ops have operators of the same names.  Ones not yet implemented
-// at this time are: exp2, exp10, log2, log10, isinf
+// Most but not all of these math ops have operators of the same names.
 MIGRAPHX_DEVICE_MATH_HALF2(abs, ::__habs2)
 MIGRAPHX_DEVICE_MATH_HALF2(ceil, ::h2ceil)
 MIGRAPHX_DEVICE_MATH_HALF2(cos, ::h2cos)
@@ -189,9 +188,8 @@ MIGRAPHX_DEVICE_MATH_BINARY_FOR(float, max, ::max)
 MIGRAPHX_DEVICE_MATH_BINARY_FOR(float, min, ::min)
 MIGRAPHX_DEVICE_MATH_BINARY_FOR(double, max, ::max)
 MIGRAPHX_DEVICE_MATH_BINARY_FOR(double, min, ::min)
-// Add overloads for half that calls the float version
-MIGRAPHX_DEVICE_MATH_BINARY_FOR(migraphx::half, max, ::fmaxf)
-MIGRAPHX_DEVICE_MATH_BINARY_FOR(migraphx::half, min, ::fminf)
+MIGRAPHX_DEVICE_MATH_BINARY_FOR(migraphx::half, max, ::__hmax)
+MIGRAPHX_DEVICE_MATH_BINARY_FOR(migraphx::half, min, ::__hmin)
 
 template <class T, MIGRAPHX_REQUIRES(not is_any_vec<T>())>
 constexpr auto max(const T& a, const T& b)

test/verify/test_min_max.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 "verify_program.hpp"
+#include <migraphx/program.hpp>
+#include <migraphx/generate.hpp>
+#include <migraphx/op/max.hpp>
+#include <migraphx/op/min.hpp>
+
+template <class Op, migraphx::shape::type_t T>
+struct test_min_max : verify_program<test_min_max<Op, T>>
+{
+    migraphx::program create_program() const
+    {
+        migraphx::program p;
+        auto* mm = p.get_main_module();
+        migraphx::shape s{T, {128}};
+        auto x = mm->add_parameter("x", s);
+        auto y = mm->add_parameter("y", s);
+        mm->add_instruction(Op{}, x, y);
+        return p;
+    }
+};
+
+template struct test_min_max<migraphx::op::max, migraphx::shape::float_type>;
+template struct test_min_max<migraphx::op::max, migraphx::shape::half_type>;
+template struct test_min_max<migraphx::op::max, migraphx::shape::double_type>;
+
+template struct test_min_max<migraphx::op::min, migraphx::shape::float_type>;
+template struct test_min_max<migraphx::op::min, migraphx::shape::half_type>;
+template struct test_min_max<migraphx::op::min, migraphx::shape::double_type>;