Merge branch 'master' into fp16

src/targets/cpu/gemm.cpp

@@ -4,6 +4,7 @@
 #include <blaze/math/CustomMatrix.h>
 
 namespace migraph {
+inline namespace MIGRAPH_INLINE_NS {
 namespace cpu {
 
 template <class T>
@@ -93,5 +94,5 @@ void migemm(
 }
 
 } // namespace cpu
-
+} // namespace MIGRAPH_INLINE_NS
 } // namespace migraph

src/targets/cpu/include/migraph/cpu/context.hpp

 #ifndef MIGRAPH_GUARD_RTGLIB_CONTEXT_HPP
 #define MIGRAPH_GUARD_RTGLIB_CONTEXT_HPP
 
+#include <migraph/config.hpp>
+
 namespace migraph {
+inline namespace MIGRAPH_INLINE_NS {
 namespace cpu {
 
 struct context
@@ -10,6 +13,7 @@ struct context
 };
 
 } // namespace cpu
+} // namespace MIGRAPH_INLINE_NS
 } // namespace migraph
 
 #endif

src/targets/cpu/include/migraph/cpu/gemm.hpp

@@ -2,15 +2,17 @@
 #define MIGRAPH_GUARD_RTGLIB_CPU_GEMM_HPP
 
 #include <migraph/argument.hpp>
+#include <migraph/config.hpp>
 
 namespace migraph {
+inline namespace MIGRAPH_INLINE_NS {
 namespace cpu {
 
 void migemm(
     const argument& c_arg, const argument& a_arg, const argument& b_arg, float alpha, float beta);
 
 } // namespace cpu
-
+} // namespace MIGRAPH_INLINE_NS
 } // namespace migraph
 
 #endif

src/targets/cpu/include/migraph/cpu/lowering.hpp

@@ -2,8 +2,10 @@
 #define MIGRAPH_GUARD_RTGLIB_CPU_LOWERING_HPP
 
 #include <migraph/program.hpp>
+#include <migraph/config.hpp>
 
 namespace migraph {
+inline namespace MIGRAPH_INLINE_NS {
 namespace cpu {
 
 struct lowering
@@ -13,7 +15,7 @@ struct lowering
 };
 
 } // namespace cpu
-
+} // namespace MIGRAPH_INLINE_NS
 } // namespace migraph
 
 #endif

src/targets/cpu/include/migraph/cpu/target.hpp

@@ -3,8 +3,10 @@
 
 #include <migraph/program.hpp>
 #include <migraph/cpu/context.hpp>
+#include <migraph/config.hpp>
 
 namespace migraph {
+inline namespace MIGRAPH_INLINE_NS {
 namespace cpu {
 
 struct target
@@ -15,7 +17,7 @@ struct target
 };
 
 } // namespace cpu
-
+} // namespace MIGRAPH_INLINE_NS
 } // namespace migraph
 
 #endif

src/targets/cpu/lowering.cpp

@@ -10,6 +10,7 @@
 #include <utility>
 
 namespace migraph {
+inline namespace MIGRAPH_INLINE_NS {
 namespace cpu {
 
 template <typename T>
@@ -606,6 +607,7 @@ struct cpu_apply
         apply_map["sin"]        = simple_op<cpu_unary<sin_op>>();
         apply_map["cos"]        = simple_op<cpu_unary<cos_op>>();
         apply_map["tan"]        = simple_op<cpu_unary<tan_op>>();
+        apply_map["relu"]       = simple_op<cpu_unary<relu_op>>();
         apply_map["add"]        = simple_op<cpu_binary<add_op>>();
         apply_map["sub"]        = simple_op<cpu_binary<sub_op>>();
         apply_map["mul"]        = simple_op<cpu_binary<mul_op>>();
@@ -619,11 +621,7 @@ struct cpu_apply
         init();
         for(auto it : iterator_for(*prog))
         {
-            if(it->name() == "activation")
-            {
-                apply_activation(it);
-            }
-            else if(it->name() == "pooling")
+            if(it->name() == "pooling")
             {
                 apply_pooling(it);
             }
@@ -647,13 +645,6 @@ struct cpu_apply
         prog->replace_instruction(ins, T{op}, ins->inputs());
     }
 
-    void apply_activation(instruction_ref ins)
-    {
-        auto&& op = any_cast<op::activation>(ins->get_operator());
-        if(op.mode == "relu")
-            prog->replace_instruction(ins, cpu_unary<relu_op>{}, ins->inputs());
-    }
-
     void apply_pooling(instruction_ref ins)
     {
         auto&& op = any_cast<op::pooling>(ins->get_operator());
@@ -667,5 +658,5 @@ struct cpu_apply
 void lowering::apply(program& p) const { cpu_apply{&p}.apply(); }
 
 } // namespace cpu
-
+} // namespace MIGRAPH_INLINE_NS
 } // namespace migraph

src/targets/cpu/target.cpp

@@ -4,6 +4,7 @@
 #include <migraph/auto_contiguous.hpp>
 
 namespace migraph {
+inline namespace MIGRAPH_INLINE_NS {
 namespace cpu {
 
 std::string target::name() const { return "cpu"; }
@@ -14,5 +15,5 @@ std::vector<pass> target::get_passes(migraph::context&) const
 }
 
 } // namespace cpu
-
+} // namespace MIGRAPH_INLINE_NS
 } // namespace migraph

src/targets/gpu/add.cpp

 #include <migraph/gpu/add.hpp>
 #include <migraph/operators.hpp>
 #include <migraph/manage_ptr.hpp>
+#include <migraph/config.hpp>
 #include <migraph/gpu/miopen.hpp>
 #include <utility>
 
 namespace migraph {
+inline namespace MIGRAPH_INLINE_NS {
 namespace gpu {
 
 shape hip_add::compute_shape(const std::vector<shape>& inputs) const
@@ -49,5 +51,5 @@ argument miopen_add::compute(context& ctx,
 }
 
 } // namespace gpu
-
+} // namespace MIGRAPH_INLINE_NS
 } // namespace migraph

src/targets/gpu/batchnorm.cpp

@@ -5,6 +5,7 @@
 #include <utility>
 
 namespace migraph {
+inline namespace MIGRAPH_INLINE_NS {
 namespace gpu {
 
 shape miopen_batch_norm_inference::compute_shape(const std::vector<shape>& inputs) const
@@ -42,5 +43,5 @@ argument miopen_batch_norm_inference::compute(context& ctx,
 }
 
 } // namespace gpu
-
+} // namespace MIGRAPH_INLINE_NS
 } // namespace migraph

src/targets/gpu/concat.cpp

@@ -6,6 +6,7 @@
 #include <utility>
 
 namespace migraph {
+inline namespace MIGRAPH_INLINE_NS {
 namespace gpu {
 
 shape hip_concat::compute_shape(std::vector<shape> inputs) const
@@ -23,5 +24,5 @@ argument hip_concat::compute(context& ctx,
 }
 
 } // namespace gpu
-
+} // namespace MIGRAPH_INLINE_NS
 } // namespace migraph

src/targets/gpu/contiguous.cpp

@@ -5,6 +5,7 @@
 #include <utility>
 
 namespace migraph {
+inline namespace MIGRAPH_INLINE_NS {
 namespace gpu {
 
 shape miopen_contiguous::compute_shape(const std::vector<shape>& inputs) const
@@ -24,5 +25,5 @@ argument miopen_contiguous::compute(context& ctx,
 }
 
 } // namespace gpu
-
+} // namespace MIGRAPH_INLINE_NS
 } // namespace migraph

src/targets/gpu/convolution.cpp

@@ -5,6 +5,7 @@
 #include <utility>
 
 namespace migraph {
+inline namespace MIGRAPH_INLINE_NS {
 namespace gpu {
 
 shape miopen_convolution::compute_shape(const std::vector<shape>& inputs) const
@@ -81,5 +82,5 @@ shape miopen_convolution::compile(context& ctx,
 }
 
 } // namespace gpu
-
+} // namespace MIGRAPH_INLINE_NS
 } // namespace migraph

src/targets/gpu/device/add.cpp

@@ -2,6 +2,7 @@
 #include <migraph/gpu/device/nary.hpp>
 
 namespace migraph {
+inline namespace MIGRAPH_INLINE_NS {
 namespace gpu {
 namespace device {
 
@@ -21,4 +22,5 @@ void add(hipStream_t stream,
 
 } // namespace device
 } // namespace gpu
+} // namespace MIGRAPH_INLINE_NS
 } // namespace migraph

src/targets/gpu/device/add_relu.cpp

@@ -2,6 +2,7 @@
 #include <migraph/gpu/device/nary.hpp>
 
 namespace migraph {
+inline namespace MIGRAPH_INLINE_NS {
 namespace gpu {
 namespace device {
 
@@ -26,4 +27,5 @@ void add_relu(hipStream_t stream,
 
 } // namespace device
 } // namespace gpu
+} // namespace MIGRAPH_INLINE_NS
 } // namespace migraph

src/targets/gpu/device/concat.cpp

@@ -5,6 +5,7 @@
 #include <migraph/gpu/device/launch.hpp>
 
 namespace migraph {
+inline namespace MIGRAPH_INLINE_NS {
 namespace gpu {
 namespace device {
 
@@ -33,4 +34,5 @@ argument concat(hipStream_t stream,
 
 } // namespace device
 } // namespace gpu
+} // namespace MIGRAPH_INLINE_NS
 } // namespace migraph

src/targets/gpu/device/contiguous.cpp

@@ -3,6 +3,7 @@
 #include <migraph/gpu/device/nary.hpp>
 
 namespace migraph {
+inline namespace MIGRAPH_INLINE_NS {
 namespace gpu {
 namespace device {
 
@@ -13,4 +14,5 @@ void contiguous(hipStream_t stream, argument result, argument arg)
 
 } // namespace device
 } // namespace gpu
+} // namespace MIGRAPH_INLINE_NS
 } // namespace migraph

src/targets/gpu/device/include/migraph/gpu/device/launch.hpp

@@ -2,8 +2,10 @@
 #define MIGRAPH_GUARD_RTGLIB_DEVICE_LAUNCH_HPP
 
 #include <hip/hip_runtime.h>
+#include <migraph/config.hpp>
 
 namespace migraph {
+inline namespace MIGRAPH_INLINE_NS {
 namespace gpu {
 namespace device {
 
@@ -58,6 +60,7 @@ inline auto gs_launch(hipStream_t stream, std::size_t n, std::size_t local = 102
 
 } // namespace device
 } // namespace gpu
+} // namespace MIGRAPH_INLINE_NS
 } // namespace migraph
 
 #endif

src/targets/gpu/device/include/migraph/gpu/device/nary.hpp

@@ -6,8 +6,10 @@
 #include <migraph/gpu/device/types.hpp>
 #include <migraph/functional.hpp>
 #include <migraph/ranges.hpp>
+#include <migraph/config.hpp>
 
 namespace migraph {
+inline namespace MIGRAPH_INLINE_NS {
 namespace gpu {
 namespace device {
 
@@ -394,6 +396,7 @@ inline auto nary(hipStream_t stream,
 
 } // namespace device
 } // namespace gpu
+} // namespace MIGRAPH_INLINE_NS
 } // namespace migraph
 
 #endif

src/targets/gpu/device/include/migraph/gpu/device/tensor.hpp

@@ -3,8 +3,10 @@
 
 #include <hip/hip_runtime.h>
 #include <migraph/functional.hpp>
+#include <migraph/config.hpp>
 
 namespace migraph {
+inline namespace MIGRAPH_INLINE_NS {
 namespace gpu {
 namespace device {
 
@@ -85,6 +87,7 @@ struct hip_tensor_descriptor
 
 } // namespace device
 } // namespace gpu
+} // namespace MIGRAPH_INLINE_NS
 } // namespace migraph
 
 #endif

src/targets/gpu/device/include/migraph/gpu/device/types.hpp

@@ -9,8 +9,10 @@
 #define MIGRAPH_GUARD_RTGLIB_GPU_DEVICE_TYPES_HPP
 
 #include <migraph/half.hpp>
+#include <migraph/config.hpp>
 
 namespace migraph {
+inline namespace MIGRAPH_INLINE_NS {
 namespace gpu {
 namespace device {
 
@@ -75,6 +77,7 @@ device_type<T>* device_cast(T* x)
 
 } // namespace device
 } // namespace gpu
+} // namespace MIGRAPH_INLINE_NS
 } // namespace migraph
 
 #endif