change to function call to replace_argument, and remove unnecessary code

src/CMakeLists.txt

@@ -18,7 +18,6 @@ add_library(migraphx
     generate.cpp
     instruction.cpp
     program.cpp
-    quantization.cpp
     quantize_ins.cpp
     shape.cpp
     schedule.cpp

src/include/migraphx/instruction.hpp

@@ -27,7 +27,6 @@ struct instruction
     void replace(const shape& r);
 
     void recompute_shape();
-    void recompute_ins_shape();
 
     void clear_arguments();
 
@@ -70,8 +69,7 @@ struct instruction
 
     static void replace_argument(instruction_ref ins,
                                  instruction_ref old,
-                                 instruction_ref new_ins,
-                                 bool recompute_shape = true);
+                                 instruction_ref new_ins);
 
     static void
     replace(instruction_ref ins, operation o, const shape& r, std::vector<instruction_ref> args);

src/include/migraphx/quantization.hpp

-#ifndef MIGRAPHX_GUARD_MIGRAPHLIB_QUANTIZATION_HPP
-#define MIGRAPHX_GUARD_MIGRAPHLIB_QUANTIZATION_HPP
-
-#include <list>
-#include <unordered_map>
-#include <migraphx/operation.hpp>
-#include <migraphx/literal.hpp>
-#include <migraphx/builtin.hpp>
-#include <migraphx/instruction_ref.hpp>
-#include <migraphx/target.hpp>
-#include <migraphx/tracer.hpp>
-#include <migraphx/env.hpp>
-#include <migraphx/config.hpp>
-#include <algorithm>
-#include <iostream>
-
-namespace migraphx {
-inline namespace MIGRAPHX_INLINE_NS {
-
-void quantize(program& prog);
-
-} // namespace MIGRAPHX_INLINE_NS
-} // namespace migraphx
-
-#endif

src/instruction.cpp

@@ -30,13 +30,6 @@ void instruction::replace(const shape& r)
 
 void instruction::recompute_shape() { replace(compute_shape(op, arguments)); }
 
-void instruction::recompute_ins_shape()
-{
-    auto r = compute_shape(op, arguments);
-    if(r != result)
-        result = r;
-}
-
 void instruction::clear_arguments()
 {
     for(auto&& arg : arguments)
@@ -133,12 +126,10 @@ void instruction::backreference(instruction_ref ref)
 
 void instruction::replace_argument(instruction_ref ins,
                                    instruction_ref old,
-                                   instruction_ref new_ins,
-                                   bool recompute_shape)
+                                   instruction_ref new_ins)
 {
     ins->replace_argument(old, new_ins);
     backreference(ins);
-    if(recompute_shape)
     ins->recompute_shape();
 }
 

src/py/migraphx_py.cpp

@@ -2,7 +2,6 @@
 #include <pybind11/pybind11.h>
 #include <pybind11/stl.h>
 #include <migraphx/program.hpp>
-#include <migraphx/quantization.hpp>
 #include <migraphx/quantize_ins.hpp>
 #include <migraphx/generate.hpp>
 #include <migraphx/cpu/target.hpp>
@@ -183,7 +182,6 @@ PYBIND11_MODULE(migraphx, m)
     });
 
     m.def("generate_argument", &migraphx::generate_argument, py::arg("s"), py::arg("seed") = 0);
-    m.def("quantize", &migraphx::quantize);
     m.def("quantize_ins", &migraphx::quantize_ins);
 
 #ifdef HAVE_GPU

src/quantization.cpp

-#include <migraphx/program.hpp>
-#include <migraphx/quantization.hpp>
-#include <migraphx/stringutils.hpp>
-#include <migraphx/instruction.hpp>
-#include <migraphx/op/fp_conversion.hpp>
-#include <migraphx/target.hpp>
-#include <migraphx/env.hpp>
-#include <migraphx/ranges.hpp>
-#include <migraphx/time.hpp>
-#include <migraphx/iterator_for.hpp>
-#include <iostream>
-#include <sstream>
-#include <algorithm>
-#include <utility>
-
-namespace migraphx {
-inline namespace MIGRAPHX_INLINE_NS {
-
-instruction_ref convert_to_fp16(program& prog, instruction_ref& ins)
-{
-    assert(ins->get_shape().type() == shape::float_type ||
-           ins->get_shape().type() == shape::double_type);
-    assert(contains({"@literal", "@param"}, ins->name()));
-    instruction_ref ins_fp16{};
-    if(ins->name() == "@literal")
-    {
-        shape s = ins->get_shape();
-        auto l  = ins->get_literal();
-        if(s.type() == shape::float_type)
-        {
-            auto tv = l.get<const float>();
-            ins_fp16 =
-                prog.add_literal(literal({shape::half_type, s.lens()}, tv.begin(), tv.end()));
-        }
-        else
-        {
-            auto tv = l.get<const double>();
-            ins_fp16 =
-                prog.add_literal(literal({shape::half_type, s.lens()}, tv.begin(), tv.end()));
-        }
-    }
-    else if(ins->name() == "@param")
-    {
-        if(ins == std::prev(prog.end()))
-        {
-            ins_fp16 = prog.add_instruction(op::fp_conversion{}, ins);
-        }
-        else
-        {
-            ins_fp16 = prog.insert_instruction(std::next(ins), op::fp_conversion{}, ins);
-        }
-    }
-
-    return ins_fp16;
-}
-
-void quantize(program& prog)
-{
-    bool reduced_precision  = false;
-    shape::type_t orig_type = shape::float_type;
-    for(auto ins : iterator_for(prog))
-    {
-        // convert float_type to half_type
-        if(contains({"@literal", "@param"}, ins->name()) &&
-           (ins->get_shape().type() == shape::float_type ||
-            ins->get_shape().type() == shape::double_type))
-        {
-            orig_type     = ins->get_shape().type();
-            auto ins_fp16 = convert_to_fp16(prog, ins);
-            auto outputs  = ins->outputs();
-            for(auto output : outputs)
-            {
-                if(output != ins_fp16)
-                {
-                    instruction::replace_argument(output, ins, ins_fp16, false);
-                }
-            }
-
-            reduced_precision = true;
-        }
-    }
-
-    // add another instruction at last to convert fp16 to fp32
-    if(reduced_precision)
-    {
-        for(auto ins : iterator_for(prog))
-        {
-            if(!contains({"@literal", "@param"}, ins->name()))
-            {
-                ins->recompute_ins_shape();
-            }
-        }
-
-        auto ins = std::prev(prog.end());
-        if(ins->get_shape().type() == shape::half_type)
-        {
-            prog.add_instruction(op::fp_conversion{orig_type}, ins);
-        }
-    }
-}
-
-} // namespace MIGRAPHX_INLINE_NS
-} // namespace migraphx

src/quantize_ins.cpp

@@ -61,6 +61,7 @@ void quantize_ins(program& prog, const std::vector<std::string>& ins_names)
         // process all inputs, if input is a fp32 or fp64, convert it
         // to a fp16 by adding a fp_conversion operator.
         auto inputs = ins->inputs();
+        std::vector<instruction_ref> converted_inputs;
         for(auto input : inputs)
         {
             auto s = input->get_shape();
@@ -77,16 +78,24 @@ void quantize_ins(program& prog, const std::vector<std::string>& ins_names)
                 {
                     input_fp16 = insert_fp16(prog, input, shape::half_type, map_fp16);
                 }
-                instruction::replace_argument(ins, input, input_fp16, false);
+                //instruction::replace_argument(ins, input, input_fp16, false);
+                converted_inputs.push_back(input_fp16);
             }
+            else
+            {
+                converted_inputs.push_back(input);
+            }
+        }
+
+        if (inputs != converted_inputs)
+        {
+            auto op = ins->get_operator();
+            instruction::replace(ins, op, compute_shape(op, converted_inputs), converted_inputs);
         }
-        // recompute the output shape
-        ins->recompute_ins_shape();
 
-        // If output is not the original type, add another instruction
-        // to convert it back to the original type
-        if(ins->get_shape().type() != orig_type)
+        if (ins->get_shape().type() != orig_type)
         {
+            // insert another fp_conversion instruction to convert it back
             if(ins == std::prev(prog.end()))
             {
                 prog.add_instruction(op::fp_conversion{orig_type}, ins);