Merge from develop

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

 #ifndef MIGRAPHX_GUARD_RTGLIB_EXP_HPP
 #define MIGRAPHX_GUARD_RTGLIB_EXP_HPP
 
-#include <migraphx/gpu/lowering.hpp>
 #include <migraphx/gpu/oper.hpp>
-#include <migraphx/manage_ptr.hpp>
-#include <migraphx/instruction.hpp>
-#include <migraphx/operators.hpp>
-#include <migraphx/generate.hpp>
-#include <migraphx/shape_for_each.hpp>
-#include <migraphx/gpu/miopen.hpp>
-#include <migraphx/gpu/hip.hpp>
-#include <migraphx/dfor.hpp>
-#include <migraphx/gpu/device/contiguous.hpp>
 #include <migraphx/gpu/device/exp.hpp>
-#include <migraphx/iterator_for.hpp>
-#include <migraphx/gpu/rocblas.hpp>
-#include <migraphx/gpu/context.hpp>
-#include <migraphx/config.hpp>
-#include <utility>
 
 namespace migraphx {
 inline namespace MIGRAPHX_INLINE_NS {

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

 #ifndef MIGRAPHX_GUARD_RTGLIB_GATHER_HPP
 #define MIGRAPHX_GUARD_RTGLIB_GATHER_HPP
 
-#include <migraphx/gpu/lowering.hpp>
-#include <migraphx/manage_ptr.hpp>
-#include <migraphx/instruction.hpp>
-#include <migraphx/operators.hpp>
-#include <migraphx/generate.hpp>
-#include <migraphx/shape_for_each.hpp>
-#include <migraphx/config.hpp>
+#include <migraphx/shape.hpp>
 #include <migraphx/gpu/miopen.hpp>
-#include <migraphx/gpu/hip.hpp>
-#include <migraphx/dfor.hpp>
-#include <migraphx/gpu/device/contiguous.hpp>
-#include <migraphx/gpu/device/gather.hpp>
-#include <migraphx/iterator_for.hpp>
-#include <migraphx/gpu/rocblas.hpp>
-#include <migraphx/gpu/context.hpp>
-#include <utility>
 
 namespace migraphx {
 inline namespace MIGRAPHX_INLINE_NS {
 namespace gpu {
 
+struct context;
+
 struct hip_gather
 {
     op::gather op;

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

 #ifndef MIGRAPHX_GUARD_RTGLIB_GEMM_HPP
 #define MIGRAPHX_GUARD_RTGLIB_GEMM_HPP
 
-#include <migraphx/gpu/lowering.hpp>
-#include <migraphx/manage_ptr.hpp>
-#include <migraphx/instruction.hpp>
+#include <migraphx/shape.hpp>
 #include <migraphx/operators.hpp>
-#include <migraphx/generate.hpp>
-#include <migraphx/shape_for_each.hpp>
-#include <migraphx/gpu/miopen.hpp>
-#include <migraphx/gpu/hip.hpp>
-#include <migraphx/dfor.hpp>
-#include <migraphx/config.hpp>
-#include <migraphx/gpu/device/contiguous.hpp>
-#include <migraphx/gpu/device/add.hpp>
-#include <migraphx/iterator_for.hpp>
-#include <migraphx/gpu/rocblas.hpp>
-#include <migraphx/gpu/context.hpp>
-#include <utility>
 
 namespace migraphx {
 inline namespace MIGRAPHX_INLINE_NS {
 namespace gpu {
 
+struct context;
+
 struct miopen_gemm
 {
     op::dot op;

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

@@ -23,12 +23,13 @@ void copy_to_gpu(const argument& src, const argument& dst);
 
 struct hip_allocate
 {
+    shape s;
     std::string tag{};
     std::string name() const { return "hip::allocate"; }
     shape compute_shape(const std::vector<shape>& inputs) const
     {
-        check_shapes{inputs}.has(1);
-        return inputs.front();
+        check_shapes{inputs}.has(0);
+        return s;
     }
     argument compute(context&, const shape& output_shape, const std::vector<argument>&) const
     {

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

 #ifndef MIGRAPHX_GUARD_RTGLIB_LEAKY_RELU_HPP
 #define MIGRAPHX_GUARD_RTGLIB_LEAKY_RELU_HPP
 
-#include <migraphx/gpu/lowering.hpp>
-#include <migraphx/manage_ptr.hpp>
-#include <migraphx/instruction.hpp>
-#include <migraphx/operators.hpp>
-#include <migraphx/generate.hpp>
-#include <migraphx/shape_for_each.hpp>
-#include <migraphx/config.hpp>
+#include <migraphx/shape.hpp>
 #include <migraphx/gpu/miopen.hpp>
-#include <migraphx/gpu/hip.hpp>
-#include <migraphx/dfor.hpp>
-#include <migraphx/gpu/device/contiguous.hpp>
-#include <migraphx/gpu/device/add.hpp>
-#include <migraphx/iterator_for.hpp>
-#include <migraphx/gpu/rocblas.hpp>
-#include <migraphx/gpu/context.hpp>
-#include <utility>
 
 namespace migraphx {
 inline namespace MIGRAPHX_INLINE_NS {
 namespace gpu {
 
+struct context;
+
 struct miopen_leaky_relu
 {
     shared<activation_descriptor> ad;

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

 #ifndef MIGRAPHX_GUARD_RTGLIB_LOG_HPP
 #define MIGRAPHX_GUARD_RTGLIB_LOG_HPP
 
-#include <migraphx/gpu/lowering.hpp>
 #include <migraphx/gpu/oper.hpp>
-#include <migraphx/manage_ptr.hpp>
-#include <migraphx/instruction.hpp>
-#include <migraphx/operators.hpp>
-#include <migraphx/generate.hpp>
-#include <migraphx/shape_for_each.hpp>
-#include <migraphx/gpu/miopen.hpp>
-#include <migraphx/gpu/hip.hpp>
-#include <migraphx/dfor.hpp>
-#include <migraphx/gpu/device/contiguous.hpp>
 #include <migraphx/gpu/device/log.hpp>
-#include <migraphx/iterator_for.hpp>
-#include <migraphx/gpu/rocblas.hpp>
-#include <migraphx/gpu/context.hpp>
-#include <migraphx/config.hpp>
-#include <utility>
 
 namespace migraphx {
 inline namespace MIGRAPHX_INLINE_NS {

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

 #ifndef MIGRAPHX_GUARD_RTGLIB_LRN_HPP
 #define MIGRAPHX_GUARD_RTGLIB_LRN_HPP
 
-#include <migraphx/gpu/lowering.hpp>
-#include <migraphx/manage_ptr.hpp>
-#include <migraphx/instruction.hpp>
-#include <migraphx/operators.hpp>
-#include <migraphx/generate.hpp>
-#include <migraphx/shape_for_each.hpp>
-#include <migraphx/config.hpp>
+#include <migraphx/shape.hpp>
 #include <migraphx/gpu/miopen.hpp>
-#include <migraphx/gpu/hip.hpp>
-#include <migraphx/dfor.hpp>
-#include <migraphx/gpu/device/contiguous.hpp>
-#include <migraphx/gpu/device/add.hpp>
-#include <migraphx/iterator_for.hpp>
-#include <migraphx/gpu/rocblas.hpp>
-#include <migraphx/gpu/context.hpp>
-#include <utility>
 
 namespace migraphx {
 inline namespace MIGRAPHX_INLINE_NS {
 namespace gpu {
 
+struct context;
+
 struct miopen_lrn
 {
     shared<lrn_descriptor> ldesc;

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

 #ifndef MIGRAPHX_GUARD_RTGLIB_MAX_HPP
 #define MIGRAPHX_GUARD_RTGLIB_MAX_HPP
 
-#include <migraphx/gpu/lowering.hpp>
 #include <migraphx/gpu/oper.hpp>
-#include <migraphx/manage_ptr.hpp>
-#include <migraphx/instruction.hpp>
-#include <migraphx/operators.hpp>
-#include <migraphx/generate.hpp>
-#include <migraphx/shape_for_each.hpp>
-#include <migraphx/config.hpp>
-#include <migraphx/gpu/miopen.hpp>
-#include <migraphx/gpu/hip.hpp>
-#include <migraphx/dfor.hpp>
-#include <migraphx/gpu/device/contiguous.hpp>
 #include <migraphx/gpu/device/max.hpp>
-#include <migraphx/iterator_for.hpp>
-#include <migraphx/gpu/rocblas.hpp>
-#include <migraphx/gpu/context.hpp>
-#include <utility>
 
 namespace migraphx {
 inline namespace MIGRAPHX_INLINE_NS {

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

 #ifndef MIGRAPHX_GUARD_RTGLIB_MIN_HPP
 #define MIGRAPHX_GUARD_RTGLIB_MIN_HPP
 
-#include <migraphx/gpu/lowering.hpp>
 #include <migraphx/gpu/oper.hpp>
-#include <migraphx/manage_ptr.hpp>
-#include <migraphx/instruction.hpp>
-#include <migraphx/operators.hpp>
-#include <migraphx/generate.hpp>
-#include <migraphx/shape_for_each.hpp>
-#include <migraphx/config.hpp>
-#include <migraphx/gpu/miopen.hpp>
-#include <migraphx/gpu/hip.hpp>
-#include <migraphx/dfor.hpp>
-#include <migraphx/gpu/device/contiguous.hpp>
 #include <migraphx/gpu/device/min.hpp>
-#include <migraphx/iterator_for.hpp>
-#include <migraphx/gpu/rocblas.hpp>
-#include <migraphx/gpu/context.hpp>
-#include <utility>
 
 namespace migraphx {
 inline namespace MIGRAPHX_INLINE_NS {

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

 #ifndef MIGRAPHX_GUARD_RTGLIB_MUL_HPP
 #define MIGRAPHX_GUARD_RTGLIB_MUL_HPP
 
-#include <migraphx/gpu/lowering.hpp>
 #include <migraphx/gpu/oper.hpp>
-#include <migraphx/manage_ptr.hpp>
-#include <migraphx/instruction.hpp>
-#include <migraphx/operators.hpp>
-#include <migraphx/generate.hpp>
-#include <migraphx/shape_for_each.hpp>
-#include <migraphx/config.hpp>
-#include <migraphx/gpu/miopen.hpp>
-#include <migraphx/gpu/hip.hpp>
-#include <migraphx/dfor.hpp>
-#include <migraphx/gpu/device/contiguous.hpp>
 #include <migraphx/gpu/device/mul.hpp>
-#include <migraphx/iterator_for.hpp>
-#include <migraphx/gpu/rocblas.hpp>
-#include <migraphx/gpu/context.hpp>
-#include <utility>
 
 namespace migraphx {
 inline namespace MIGRAPHX_INLINE_NS {

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

 #ifndef MIGRAPHX_GUARD_RTGLIB_UNARY_HPP
 #define MIGRAPHX_GUARD_RTGLIB_UNARY_HPP
 
-#include <migraphx/gpu/lowering.hpp>
-#include <migraphx/manage_ptr.hpp>
-#include <migraphx/instruction.hpp>
-#include <migraphx/operators.hpp>
-#include <migraphx/generate.hpp>
-#include <migraphx/shape_for_each.hpp>
-#include <migraphx/gpu/miopen.hpp>
 #include <migraphx/gpu/hip.hpp>
-#include <migraphx/dfor.hpp>
-#include <migraphx/type_name.hpp>
-#include <migraphx/gpu/device/contiguous.hpp>
-#include <migraphx/iterator_for.hpp>
-#include <migraphx/gpu/rocblas.hpp>
 #include <migraphx/gpu/context.hpp>
+#include <migraphx/shape.hpp>
+#include <migraphx/argument.hpp>
 #include <migraphx/config.hpp>
+#include <migraphx/type_name.hpp>
 #include <utility>
 #include <iostream>
 

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

 #ifndef MIGRAPHX_GUARD_RTGLIB_PAD_HPP
 #define MIGRAPHX_GUARD_RTGLIB_PAD_HPP
 
-#include <migraphx/gpu/lowering.hpp>
-#include <migraphx/manage_ptr.hpp>
-#include <migraphx/instruction.hpp>
+#include <migraphx/shape.hpp>
 #include <migraphx/operators.hpp>
-#include <migraphx/generate.hpp>
-#include <migraphx/shape_for_each.hpp>
-#include <migraphx/config.hpp>
-#include <migraphx/gpu/miopen.hpp>
-#include <migraphx/gpu/hip.hpp>
-#include <migraphx/dfor.hpp>
-#include <migraphx/gpu/device/pad.hpp>
-#include <migraphx/gpu/device/add.hpp>
-#include <migraphx/iterator_for.hpp>
-#include <migraphx/gpu/rocblas.hpp>
-#include <migraphx/gpu/context.hpp>
-#include <utility>
 
 namespace migraphx {
 inline namespace MIGRAPHX_INLINE_NS {
 namespace gpu {
 
+struct context;
+
 struct hip_pad
 {
     op::pad op;

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

 #ifndef MIGRAPHX_GUARD_RTGLIB_POOLING_HPP
 #define MIGRAPHX_GUARD_RTGLIB_POOLING_HPP
 
-#include <migraphx/gpu/lowering.hpp>
-#include <migraphx/manage_ptr.hpp>
-#include <migraphx/instruction.hpp>
+#include <migraphx/shape.hpp>
 #include <migraphx/operators.hpp>
-#include <migraphx/generate.hpp>
-#include <migraphx/shape_for_each.hpp>
-#include <migraphx/config.hpp>
 #include <migraphx/gpu/miopen.hpp>
-#include <migraphx/gpu/hip.hpp>
-#include <migraphx/dfor.hpp>
-#include <migraphx/gpu/device/contiguous.hpp>
-#include <migraphx/gpu/device/add.hpp>
-#include <migraphx/iterator_for.hpp>
-#include <migraphx/gpu/rocblas.hpp>
-#include <migraphx/gpu/context.hpp>
-#include <utility>
 
 namespace migraphx {
 inline namespace MIGRAPHX_INLINE_NS {
 namespace gpu {
 
+struct context;
+
 struct miopen_pooling
 {
     op::pooling op;

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

 #ifndef MIGRAPHX_GUARD_RTGLIB_RELU_HPP
 #define MIGRAPHX_GUARD_RTGLIB_RELU_HPP
 
-#include <migraphx/gpu/lowering.hpp>
-#include <migraphx/manage_ptr.hpp>
-#include <migraphx/instruction.hpp>
-#include <migraphx/operators.hpp>
-#include <migraphx/generate.hpp>
-#include <migraphx/shape_for_each.hpp>
-#include <migraphx/config.hpp>
+#include <migraphx/shape.hpp>
 #include <migraphx/gpu/miopen.hpp>
-#include <migraphx/gpu/hip.hpp>
-#include <migraphx/dfor.hpp>
-#include <migraphx/gpu/device/contiguous.hpp>
-#include <migraphx/gpu/device/add.hpp>
-#include <migraphx/iterator_for.hpp>
-#include <migraphx/gpu/rocblas.hpp>
-#include <migraphx/gpu/context.hpp>
-#include <utility>
 
 namespace migraphx {
 inline namespace MIGRAPHX_INLINE_NS {
 namespace gpu {
 
+struct context;
+
 struct miopen_relu
 {
     shared<activation_descriptor> ad;

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

@@ -2,7 +2,6 @@
 #define MIGRAPHX_GUARD_MIGRAPHLIB_ROCBLAS_HPP
 
 #include <migraphx/manage_ptr.hpp>
-#include <migraphx/operators.hpp>
 #include <migraphx/config.hpp>
 #include <rocblas.h>
 

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

 #ifndef MIGRAPHX_GUARD_RTGLIB_SIGMOID_HPP
 #define MIGRAPHX_GUARD_RTGLIB_SIGMOID_HPP
 
-#include <migraphx/gpu/lowering.hpp>
-#include <migraphx/manage_ptr.hpp>
-#include <migraphx/instruction.hpp>
-#include <migraphx/operators.hpp>
-#include <migraphx/generate.hpp>
-#include <migraphx/shape_for_each.hpp>
-#include <migraphx/config.hpp>
+#include <migraphx/shape.hpp>
 #include <migraphx/gpu/miopen.hpp>
-#include <migraphx/gpu/hip.hpp>
-#include <migraphx/dfor.hpp>
-#include <migraphx/gpu/device/contiguous.hpp>
-#include <migraphx/gpu/device/add.hpp>
-#include <migraphx/iterator_for.hpp>
-#include <migraphx/gpu/rocblas.hpp>
-#include <migraphx/gpu/context.hpp>
-#include <utility>
 
 namespace migraphx {
 inline namespace MIGRAPHX_INLINE_NS {
 namespace gpu {
 
+struct context;
+
 struct miopen_sigmoid
 {
     shared<activation_descriptor> ad;

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

 #ifndef MIGRAPHX_GUARD_RTGLIB_SIN_HPP
 #define MIGRAPHX_GUARD_RTGLIB_SIN_HPP
 
-#include <migraphx/gpu/lowering.hpp>
 #include <migraphx/gpu/oper.hpp>
-#include <migraphx/manage_ptr.hpp>
-#include <migraphx/instruction.hpp>
-#include <migraphx/operators.hpp>
-#include <migraphx/generate.hpp>
-#include <migraphx/shape_for_each.hpp>
-#include <migraphx/gpu/miopen.hpp>
-#include <migraphx/gpu/hip.hpp>
-#include <migraphx/dfor.hpp>
-#include <migraphx/gpu/device/contiguous.hpp>
 #include <migraphx/gpu/device/sin.hpp>
-#include <migraphx/iterator_for.hpp>
-#include <migraphx/gpu/rocblas.hpp>
-#include <migraphx/gpu/context.hpp>
-#include <migraphx/config.hpp>
-#include <utility>
 
 namespace migraphx {
 inline namespace MIGRAPHX_INLINE_NS {

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

 #ifndef MIGRAPHX_GUARD_RTGLIB_SINH_HPP
 #define MIGRAPHX_GUARD_RTGLIB_SINH_HPP
 
-#include <migraphx/gpu/lowering.hpp>
 #include <migraphx/gpu/oper.hpp>
-#include <migraphx/manage_ptr.hpp>
-#include <migraphx/instruction.hpp>
-#include <migraphx/operators.hpp>
-#include <migraphx/generate.hpp>
-#include <migraphx/shape_for_each.hpp>
-#include <migraphx/gpu/miopen.hpp>
-#include <migraphx/gpu/hip.hpp>
-#include <migraphx/dfor.hpp>
-#include <migraphx/gpu/device/contiguous.hpp>
 #include <migraphx/gpu/device/sinh.hpp>
-#include <migraphx/iterator_for.hpp>
-#include <migraphx/gpu/rocblas.hpp>
-#include <migraphx/gpu/context.hpp>
-#include <migraphx/config.hpp>
-#include <utility>
 
 namespace migraphx {
 inline namespace MIGRAPHX_INLINE_NS {

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

 #ifndef MIGRAPHX_GUARD_RTGLIB_SOFTMAX_HPP
 #define MIGRAPHX_GUARD_RTGLIB_SOFTMAX_HPP
 
-#include <migraphx/gpu/lowering.hpp>
-#include <migraphx/manage_ptr.hpp>
-#include <migraphx/instruction.hpp>
+#include <migraphx/shape.hpp>
 #include <migraphx/operators.hpp>
-#include <migraphx/generate.hpp>
-#include <migraphx/shape_for_each.hpp>
-#include <migraphx/config.hpp>
-#include <migraphx/gpu/miopen.hpp>
-#include <migraphx/gpu/hip.hpp>
-#include <migraphx/dfor.hpp>
-#include <migraphx/gpu/device/contiguous.hpp>
-#include <migraphx/gpu/device/add.hpp>
-#include <migraphx/iterator_for.hpp>
-#include <migraphx/gpu/rocblas.hpp>
-#include <migraphx/gpu/context.hpp>
-#include <utility>
 
 namespace migraphx {
 inline namespace MIGRAPHX_INLINE_NS {
 namespace gpu {
 
+struct context;
+
 struct miopen_softmax
 {
     op::softmax op;

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

 #ifndef MIGRAPHX_GUARD_RTGLIB_SUB_HPP
 #define MIGRAPHX_GUARD_RTGLIB_SUB_HPP
 
-#include <migraphx/gpu/lowering.hpp>
 #include <migraphx/gpu/oper.hpp>
-#include <migraphx/manage_ptr.hpp>
-#include <migraphx/instruction.hpp>
-#include <migraphx/operators.hpp>
-#include <migraphx/generate.hpp>
-#include <migraphx/shape_for_each.hpp>
-#include <migraphx/gpu/miopen.hpp>
-#include <migraphx/gpu/hip.hpp>
-#include <migraphx/dfor.hpp>
-#include <migraphx/gpu/device/contiguous.hpp>
 #include <migraphx/gpu/device/sub.hpp>
-#include <migraphx/iterator_for.hpp>
-#include <migraphx/gpu/rocblas.hpp>
-#include <migraphx/gpu/context.hpp>
-#include <migraphx/config.hpp>
-#include <utility>
 
 namespace migraphx {
 inline namespace MIGRAPHX_INLINE_NS {