add option to specify that objective gradient is constant

- seems that it can speedup in rare cases
- closes coin-or#597

ChangeLog.md

@@ -7,6 +7,11 @@ More detailed information about incremental changes can be found in the
 
 ## 3.14
 
+### 3.14.10 (2022-xx-yy)
+
+- Added option grad_f_constant to specify that objective function is linear.
+  If set, the gradient of the objective will be requested by NLP only once. [#597]
+
 ### 3.14.9 (2022-07-21)
 
 - Fixed mapping of meta data for variable bounds, e.g., variable names,

doc/options.dox

@@ -373,6 +373,14 @@ Possible values: yes, no
 Possible values: yes, no
 </blockquote>
 
+\anchor OPT_grad_f_constant
+<strong>grad_f_constant</strong>: Indicates whether to assume that the objective function is linear
+<blockquote>
+ Activating this option will cause Ipopt to ask for the Gradient of the objective function only once from the NLP and reuse this information later. The default value for this string option is "no".
+
+Possible values: yes, no
+</blockquote>
+
 \anchor OPT_jac_c_constant
 <strong>jac_c_constant</strong>: Indicates whether to assume that all equality constraints are linear
 <blockquote>

src/Algorithm/IpOrigIpoptNLP.cpp

@@ -89,6 +89,12 @@ void OrigIpoptNLP::RegisterOptions(
       "If test is activated and an invalid number is detected, "
       "the matrix is written to output with print_level corresponding to J_MORE_DETAILED; "
       "so beware of large output!");
+   roptions->AddBoolOption(
+      "grad_f_constant",
+      "Indicates whether to assume that the objective function is linear",
+      false,
+      "Activating this option will cause Ipopt to ask for the Gradient of the objective function "
+      "only once from the NLP and reuse this information later.");
    roptions->AddBoolOption(
       "jac_c_constant",
       "Indicates whether to assume that all equality constraints are linear",
@@ -142,6 +148,7 @@ bool OrigIpoptNLP::Initialize(
    options.GetEnumValue("hessian_approximation_space", enum_int, prefix);
    hessian_approximation_space_ = HessianApproximationSpace(enum_int);
 
+   options.GetBoolValue("grad_f_constant", grad_f_constant_, prefix);
    options.GetBoolValue("jac_c_constant", jac_c_constant_, prefix);
    options.GetBoolValue("jac_d_constant", jac_d_constant_, prefix);
    options.GetBoolValue("hessian_constant", hessian_constant_, prefix);
@@ -507,7 +514,12 @@ SmartPtr<const Vector> OrigIpoptNLP::grad_f(
 {
    SmartPtr<Vector> unscaled_grad_f;
    SmartPtr<const Vector> retValue;
-   if( !grad_f_cache_.GetCachedResult1Dep(retValue, &x) )
+   const Vector* dep = NULL;
+   if( !grad_f_constant_ )
+   {
+      dep = &x;
+   }
+   if( !grad_f_cache_.GetCachedResult1Dep(retValue, dep) )
    {
       grad_f_evals_++;
       unscaled_grad_f = x_space_->MakeNew();
@@ -519,7 +531,7 @@ SmartPtr<const Vector> OrigIpoptNLP::grad_f(
       ASSERT_EXCEPTION(success && IsFiniteNumber(unscaled_grad_f->Nrm2()), Eval_Error,
                        "Error evaluating the gradient of the objective function");
       retValue = NLP_scaling()->apply_grad_obj_scaling(ConstPtr(unscaled_grad_f));
-      grad_f_cache_.AddCachedResult1Dep(retValue, &x);
+      grad_f_cache_.AddCachedResult1Dep(retValue, dep);
    }
 
    return retValue;

src/Algorithm/IpOrigIpoptNLP.hpp

@@ -533,6 +533,11 @@ class IPOPTLIB_EXPORT OrigIpoptNLP: public IpoptNLP
     */
    bool check_derivatives_for_naninf_;
 
+   /** Flag indicating if we need to ask for objective
+    *  Gradient only once
+    */
+   bool grad_f_constant_;
+
    /** Flag indicating if we need to ask for equality constraint
     *  Jacobians only once
     */