Merge branch 'extend_gemm_op' into seq2seq_example

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

 #ifndef MIGRAPHX_GUARD_RTGLIB_ATAN_HPP
 #define MIGRAPHX_GUARD_RTGLIB_ATAN_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/atan.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/batchnorm.hpp

 #ifndef MIGRAPHX_GUARD_RTGLIB_BATCHNORM_HPP
 #define MIGRAPHX_GUARD_RTGLIB_BATCHNORM_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/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 <migraphx/config.hpp>
-#include <utility>
 
 namespace migraphx {
 inline namespace MIGRAPHX_INLINE_NS {
 namespace gpu {
 
+struct context;
+
 struct miopen_batch_norm_inference
 {
     op::batch_norm_inference op;

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

 #ifndef MIGRAPHX_GUARD_RTGLIB_CONCAT_HPP
 #define MIGRAPHX_GUARD_RTGLIB_CONCAT_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/concat.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_concat
 {
     op::concat op;

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

 #ifndef MIGRAPHX_GUARD_RTGLIB_CONTIGUOUS_HPP
 #define MIGRAPHX_GUARD_RTGLIB_CONTIGUOUS_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_contiguous
 {
     op::contiguous op;

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

 #ifndef MIGRAPHX_GUARD_RTGLIB_CONVOLUTION_HPP
 #define MIGRAPHX_GUARD_RTGLIB_CONVOLUTION_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_convolution
 {
     op::convolution op;

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

 #ifndef MIGRAPHX_GUARD_RTGLIB_COS_HPP
 #define MIGRAPHX_GUARD_RTGLIB_COS_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/cos.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/cosh.hpp

 #ifndef MIGRAPHX_GUARD_RTGLIB_COSH_HPP
 #define MIGRAPHX_GUARD_RTGLIB_COSH_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/cosh.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/elu.hpp

 #ifndef MIGRAPHX_GUARD_RTGLIB_ELU_HPP
 #define MIGRAPHX_GUARD_RTGLIB_ELU_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_elu
 {
     shared<activation_descriptor> ad;

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/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;