Apply comments to the paper (#327)

* Clarify phrasing

* change figure format

* add list of GPU solvers currently supported

Author: tmigot

paper/2025-09-06_ipopt_lbfgs_trunk_tron_cutest_Float64_0_291_time_pp.pdf

@@ -46,6 +46,7 @@ for solver ∈ keys(stats)
   println(io, "Solver: $solver")
   pretty_stats(io, stats[solver][!, cols], hdr_override = header)
 end
+close(io)
 
 # Path because of Ipopt on windows issue:
 stats[:ipopt].elapsed_time .= stats[:ipopt].real_time
@@ -72,7 +73,7 @@ for solver ∈ keys(stats)
   push!(tim, solver => tim_solver)
 end
 tim_best = ones(nproblems) * Inf # time of the best 
-for i=1:nproblems
+for i = 1:nproblems
   for solver ∈ keys(stats)
     tim_best[i] = min(tim[solver][i], tim_best[i])
   end
@@ -85,11 +86,19 @@ end
 using Plots
 gr()
 
+# Save individual performance profiles
 for i = 1:length(costs)
   title_name = "" # "Unconstrained solvers on CUTEst w.r.t. $(costnames[i])"
-  performance_profile(stats, costs[i], title=title_name)
-  png("$name" * "_" * costnames[i] * "_pp.png")
+  performance_profile(stats, costs[i], title = title_name)
+
+  base = string(name, "_", costnames[i], "_pp")
+  # Save as SVG and PDF (add "eps" here if you also want EPS)
+  savefig("$base.svg")
+  savefig("$base.pdf")
 end
 
+# Combined profile_solvers plot
 profile_solvers(stats, costs, costnames)
-png("$name")
+# Save the combined figure as SVG and PDF
+savefig("$name.svg")
+savefig("$name.pdf")

paper/2025-09-06_ipopt_lbfgs_trunk_tron_cutest_Float64_0_291_time_pp.svg

@@ -56,9 +56,9 @@ A nonlinear least-squares problem is a special case of \eqref{eq:nlp}, where $f(
 TRON and TRUNK have specialized implementations leveraging the structure of residual models to improve performance and scalability.
 
 A key strength of `JSOSolvers.jl` lies in its efficiency and flexibility.
-The solvers support fully in-place execution, allowing repeated solves with zero memory allocation, which is particularly beneficial in high-performance and GPU computing environments where memory management is critical.
+The solvers support fully in-place execution, allowing repeated solves without additional memory allocation, which is particularly beneficial in high-performance and GPU computing environments where memory management is critical.
 The solvers support any floating-point type, including extended and multi-precision types such as BigFloat, DoubleFloats or QuadMath.
-In addition, several solvers support GPU arrays, broadening the range of hardware where the package can be effectively deployed, for instance when used together with `ExaModels.jl` [@shin2024accelerating].
+Moreover, TRUNK, TRUNK-NLS, and FOMO support GPU arrays, broadening the range of hardware where the package can be effectively deployed, for instance when used together with `ExaModels.jl` [@shin2024accelerating].
 The package documentation and \url{https://jso.dev/tutorials} provide examples illustrating the use of different floating-point systems.
 Furthermore, the solvers expose in-place function variants, allowing multiple optimization problems with identical dimensions and data types to be solved efficiently without reallocations.
 
@@ -107,10 +107,11 @@ include("benchmark.jl") # run the benchmark and store the result in a JLD2 file
 include("analyze_benchmark.jl") # make the figure
 ```
 -->
-![Unconstrained solvers on CUTEst with respect to the elapsed time.](2025-09-06_ipopt_lbfgs_trunk_tron_cutest_Float64_0_291_time_pp.png){ width=100% }
+![Unconstrained solvers on CUTEst with respect to the elapsed time.](2025-09-06_ipopt_lbfgs_trunk_tron_cutest_Float64_0_291_time_pp.pdf){ width=100% }
 
 # Acknowledgements
 
 Dominique Orban is partially supported by an NSERC Discovery Grant.
 
 # References
+