manual merge

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

+#ifndef MIGRAPHX_GUARD_RTGLIB_DEVICE_COSH_HPP
+#define MIGRAPHX_GUARD_RTGLIB_DEVICE_COSH_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 cosh(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/device/exp.hpp

+#ifndef MIGRAPHX_GUARD_RTGLIB_DEVICE_EXP_HPP
+#define MIGRAPHX_GUARD_RTGLIB_DEVICE_EXP_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 exp(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/device/gather.hpp

+#ifndef MIGRAPHX_GUARD_RTGLIB_DEVICE_GATHER_HPP
+#define MIGRAPHX_GUARD_RTGLIB_DEVICE_GATHER_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 {
+
+argument gather(hipStream_t stream,
+                const migraphx::shape& output_shape,
+                std::vector<migraphx::argument> args,
+                int axis);
+
+} // namespace device
+} // namespace gpu
+} // namespace MIGRAPHX_INLINE_NS
+} // namespace migraphx
+
+#endif

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

+#ifndef MIGRAPHX_GUARD_RTGLIB_DEVICE_LOG_HPP
+#define MIGRAPHX_GUARD_RTGLIB_DEVICE_LOG_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 log(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/device/max.hpp

+
+#ifndef MIGRAPHX_GUARD_RTGLIB_DEVICE_MAX_HPP
+#define MIGRAPHX_GUARD_RTGLIB_DEVICE_MAX_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 max(hipStream_t stream, const argument& result, const argument& arg1, const argument& arg2);
+
+} // namespace device
+} // namespace gpu
+} // namespace MIGRAPHX_INLINE_NS
+} // namespace migraphx
+
+#endif

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

+
+#ifndef MIGRAPHX_GUARD_RTGLIB_DEVICE_MIN_HPP
+#define MIGRAPHX_GUARD_RTGLIB_DEVICE_MIN_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 min(hipStream_t stream, const argument& result, const argument& arg1, const argument& arg2);
+
+} // namespace device
+} // namespace gpu
+} // namespace MIGRAPHX_INLINE_NS
+} // namespace migraphx
+
+#endif

src/targets/gpu/include/migraph/gpu/device/mul.hpp → src/targets/gpu/include/migraphx/gpu/device/mul.hpp

 
-#ifndef MIGRAPH_GUARD_RTGLIB_DEVICE_MUL_HPP
-#define MIGRAPH_GUARD_RTGLIB_DEVICE_MUL_HPP
+#ifndef MIGRAPHX_GUARD_RTGLIB_DEVICE_MUL_HPP
+#define MIGRAPHX_GUARD_RTGLIB_DEVICE_MUL_HPP
 
-#include <migraph/argument.hpp>
-#include <migraph/config.hpp>
+#include <migraphx/argument.hpp>
+#include <migraphx/config.hpp>
 #include <hip/hip_runtime_api.h>
 
-namespace migraph {
-inline namespace MIGRAPH_INLINE_NS {
+namespace migraphx {
+inline namespace MIGRAPHX_INLINE_NS {
 namespace gpu {
 namespace device {
 
@@ -21,7 +21,7 @@ void mul(hipStream_t stream,
 
 } // namespace device
 } // namespace gpu
-} // namespace MIGRAPH_INLINE_NS
-} // namespace migraph
+} // namespace MIGRAPHX_INLINE_NS
+} // namespace migraphx
 
 #endif

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

+
+#ifndef MIGRAPHX_GUARD_RTGLIB_DEVICE_PAD_HPP
+#define MIGRAPHX_GUARD_RTGLIB_DEVICE_PAD_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 {
+
+argument pad(hipStream_t stream,
+             argument result,
+             argument arg1,
+             float value,
+             std::vector<std::int64_t> pads);
+
+} // namespace device
+} // namespace gpu
+} // namespace MIGRAPHX_INLINE_NS
+} // namespace migraphx
+
+#endif

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

+#ifndef MIGRAPHX_GUARD_RTGLIB_DEVICE_SIN_HPP
+#define MIGRAPHX_GUARD_RTGLIB_DEVICE_SIN_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 sin(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/device/sinh.hpp

+#ifndef MIGRAPHX_GUARD_RTGLIB_DEVICE_SINH_HPP
+#define MIGRAPHX_GUARD_RTGLIB_DEVICE_SINH_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 sinh(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/device/tan.hpp

+#ifndef MIGRAPHX_GUARD_RTGLIB_DEVICE_TAN_HPP
+#define MIGRAPHX_GUARD_RTGLIB_DEVICE_TAN_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 tan(hipStream_t stream, const argument& result, const argument& arg);
+
+} // namespace device
+} // namespace gpu
+} // namespace MIGRAPHX_INLINE_NS
+} // namespace migraphx
+
+#endif

src/targets/gpu/include/migraph/gpu/eliminate_workspace.hpp → src/targets/gpu/include/migraphx/gpu/eliminate_workspace.hpp

-#ifndef MIGRAPH_GUARD_RTGLIB_ELIMINATE_WORKSPACE_HPP
-#define MIGRAPH_GUARD_RTGLIB_ELIMINATE_WORKSPACE_HPP
+#ifndef MIGRAPHX_GUARD_RTGLIB_ELIMINATE_WORKSPACE_HPP
+#define MIGRAPHX_GUARD_RTGLIB_ELIMINATE_WORKSPACE_HPP
 
 #include <string>
-#include <migraph/instruction_ref.hpp>
-#include <migraph/config.hpp>
+#include <migraphx/instruction_ref.hpp>
+#include <migraphx/config.hpp>
 
-namespace migraph {
-inline namespace MIGRAPH_INLINE_NS {
+namespace migraphx {
+inline namespace MIGRAPHX_INLINE_NS {
 struct program;
 
 namespace gpu {
@@ -17,7 +17,7 @@ struct eliminate_workspace
     void apply(program& p) const;
 };
 } // namespace gpu
-} // namespace MIGRAPH_INLINE_NS
-} // namespace migraph
+} // namespace MIGRAPHX_INLINE_NS
+} // namespace migraphx
 
 #endif

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/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 miopen_elu
+{
+    shared<activation_descriptor> ad;
+    std::string name() const { return "gpu::elu"; }
+    shape compute_shape(const std::vector<shape>& inputs) const;
+    argument
+    compute(context& ctx, const shape& output_shape, const std::vector<argument>& args) const;
+    int output_alias(const std::vector<shape>& shapes) const { return shapes.size() - 1; }
+};
+
+} // namespace gpu
+} // namespace MIGRAPHX_INLINE_NS
+} // namespace migraphx
+
+#endif

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 {
+namespace gpu {
+
+struct hip_exp : unary_device<hip_exp, device::exp>
+{
+};
+
+} // namespace gpu
+} // namespace MIGRAPHX_INLINE_NS
+} // namespace migraphx
+
+#endif

src/targets/gpu/include/migraph/gpu/fuse_ops.hpp → src/targets/gpu/include/migraphx/gpu/fuse_ops.hpp

-#ifndef MIGRAPH_GUARD_RTGLIB_FUSE_OPS_HPP
-#define MIGRAPH_GUARD_RTGLIB_FUSE_OPS_HPP
+#ifndef MIGRAPHX_GUARD_RTGLIB_FUSE_OPS_HPP
+#define MIGRAPHX_GUARD_RTGLIB_FUSE_OPS_HPP
 
-#include <migraph/program.hpp>
-#include <migraph/config.hpp>
-#include <migraph/gpu/context.hpp>
+#include <migraphx/program.hpp>
+#include <migraphx/config.hpp>
+#include <migraphx/gpu/context.hpp>
 
-namespace migraph {
-inline namespace MIGRAPH_INLINE_NS {
+namespace migraphx {
+inline namespace MIGRAPHX_INLINE_NS {
 
 namespace gpu {
 
@@ -18,7 +18,7 @@ struct fuse_ops
 };
 
 } // namespace gpu
-} // namespace MIGRAPH_INLINE_NS
-} // namespace migraph
+} // namespace MIGRAPHX_INLINE_NS
+} // namespace migraphx
 
 #endif

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/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 hip_gather
+{
+    op::gather op;
+    std::string name() const { return "gpu::gather"; }
+    shape compute_shape(std::vector<shape> inputs) const;
+    argument
+    compute(context& ctx, const shape& output_shape, const std::vector<argument>& args) const;
+    int output_alias(const std::vector<shape>& shapes) const { return shapes.size() - 1; }
+};
+
+} // namespace gpu
+} // namespace MIGRAPHX_INLINE_NS
+} // namespace migraphx
+
+#endif

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

-#ifndef MIGRAPH_GUARD_RTGLIB_GEMM_HPP
-#define MIGRAPH_GUARD_RTGLIB_GEMM_HPP
+#ifndef MIGRAPHX_GUARD_RTGLIB_GEMM_HPP
+#define MIGRAPHX_GUARD_RTGLIB_GEMM_HPP
 
-#include <migraph/gpu/lowering.hpp>
-#include <migraph/manage_ptr.hpp>
-#include <migraph/instruction.hpp>
-#include <migraph/operators.hpp>
-#include <migraph/generate.hpp>
-#include <migraph/shape_for_each.hpp>
-#include <migraph/gpu/miopen.hpp>
-#include <migraph/gpu/hip.hpp>
-#include <migraph/dfor.hpp>
-#include <migraph/config.hpp>
-#include <migraph/gpu/device/contiguous.hpp>
-#include <migraph/gpu/device/add.hpp>
-#include <migraph/iterator_for.hpp>
-#include <migraph/gpu/rocblas.hpp>
-#include <migraph/gpu/context.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/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 migraph {
-inline namespace MIGRAPH_INLINE_NS {
+namespace migraphx {
+inline namespace MIGRAPHX_INLINE_NS {
 namespace gpu {
 
 struct miopen_gemm
@@ -33,7 +33,7 @@ struct miopen_gemm
 };
 
 } // namespace gpu
-} // namespace MIGRAPH_INLINE_NS
-} // namespace migraph
+} // namespace MIGRAPHX_INLINE_NS
+} // namespace migraphx
 
 #endif

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

-#ifndef MIGRAPH_GUARD_MIGRAPHLIB_HIP_HPP
-#define MIGRAPH_GUARD_MIGRAPHLIB_HIP_HPP
+#ifndef MIGRAPHX_GUARD_MIGRAPHLIB_HIP_HPP
+#define MIGRAPHX_GUARD_MIGRAPHLIB_HIP_HPP
 
-#include <migraph/operators.hpp>
-#include <migraph/config.hpp>
+#include <migraphx/operators.hpp>
+#include <migraphx/config.hpp>
 #include <utility>
 
-namespace migraph {
-inline namespace MIGRAPH_INLINE_NS {
+namespace migraphx {
+inline namespace MIGRAPHX_INLINE_NS {
 namespace gpu {
 
-migraph::argument allocate_gpu(const migraph::shape& s, bool host = false);
+argument allocate_gpu(const shape& s, bool host = false);
 
-migraph::argument to_gpu(migraph::argument arg, bool host = false);
+argument to_gpu(const argument& arg, bool host = false);
 
-migraph::argument from_gpu(migraph::argument arg);
+argument from_gpu(const argument& arg);
 
 void set_device(std::size_t id);
 
 void gpu_sync();
 
-void copy_to_gpu(argument src, argument dst);
+void copy_to_gpu(const argument& src, const argument& dst);
 
 struct hip_allocate
 {
@@ -89,7 +89,7 @@ struct hip_copy
 };
 
 } // namespace gpu
-} // namespace MIGRAPH_INLINE_NS
-} // namespace migraph
+} // namespace MIGRAPHX_INLINE_NS
+} // namespace migraphx
 
 #endif

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

-#ifndef MIGRAPH_GUARD_RTGLIB_LEAKY_RELU_HPP
-#define MIGRAPH_GUARD_RTGLIB_LEAKY_RELU_HPP
+#ifndef MIGRAPHX_GUARD_RTGLIB_LEAKY_RELU_HPP
+#define MIGRAPHX_GUARD_RTGLIB_LEAKY_RELU_HPP
 
-#include <migraph/gpu/lowering.hpp>
-#include <migraph/manage_ptr.hpp>
-#include <migraph/instruction.hpp>
-#include <migraph/operators.hpp>
-#include <migraph/generate.hpp>
-#include <migraph/shape_for_each.hpp>
-#include <migraph/config.hpp>
-#include <migraph/gpu/miopen.hpp>
-#include <migraph/gpu/hip.hpp>
-#include <migraph/dfor.hpp>
-#include <migraph/gpu/device/contiguous.hpp>
-#include <migraph/gpu/device/add.hpp>
-#include <migraph/iterator_for.hpp>
-#include <migraph/gpu/rocblas.hpp>
-#include <migraph/gpu/context.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/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 migraph {
-inline namespace MIGRAPH_INLINE_NS {
+namespace migraphx {
+inline namespace MIGRAPHX_INLINE_NS {
 namespace gpu {
 
 struct miopen_leaky_relu
@@ -33,7 +33,7 @@ struct miopen_leaky_relu
 };
 
 } // namespace gpu
-} // namespace MIGRAPH_INLINE_NS
-} // namespace migraph
+} // namespace MIGRAPHX_INLINE_NS
+} // namespace migraphx
 
 #endif

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 {
+namespace gpu {
+
+struct hip_log : unary_device<hip_log, device::log>
+{
+};
+
+} // namespace gpu
+} // namespace MIGRAPHX_INLINE_NS
+} // namespace migraphx
+
+#endif