[WIP] Enhance MTLN performance

CMakeLists.txt

@@ -57,12 +57,12 @@ if (CMAKE_SYSTEM_NAME MATCHES "Linux")
 	if(CMAKE_Fortran_COMPILER_ID MATCHES "GNU")
 		message(STATUS "Using GNU flags")
 
-		set(CMAKE_CXX_FLAGS "-fopenmp")
-		set(CMAKE_Fortran_FLAGS "-fopenmp -ffree-form -ffree-line-length-none -fdec -fallow-argument-mismatch")
+		set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -fopenmp")
+		set(CMAKE_Fortran_FLAGS "${CMAKE_Fortran_FLAGS} -fopenmp -ffree-form -ffree-line-length-none -fdec -fallow-argument-mismatch")
 
-		set(CMAKE_C_FLAGS_RELEASE "-Ofast")
-		set(CMAKE_CXX_FLAGS_RELEASE "-Ofast")
-		set(CMAKE_Fortran_FLAGS_RELEASE "-Ofast")
+	set(CMAKE_C_FLAGS_RELEASE "${CMAKE_C_FLAGS_RELEASE} -Ofast")
+		set(CMAKE_CXX_FLAGS_RELEASE "${CMAKE_CXX_FLAGS_RELEASE} -Ofast")
+		set(CMAKE_Fortran_FLAGS_RELEASE "${CMAKE_Fortran_FLAGS_RELEASE} -Ofast")
 
 		set(CMAKE_C_FLAGS_DEBUG "-g -O0")
 		set(CMAKE_CXX_FLAGS_DEBUG "-g -O0")

src_main_pub/observation.F90

@@ -743,13 +743,15 @@ subroutine InitObservation(sgg, media, tag_numbers, &
         type(mtln_solver_t), pointer :: mtln_solver
         integer :: i, j
         mtln_solver => GetSolverPtr()
-        do i = 1, ubound(mtln_solver%bundles, 1)
-          if (ubound(mtln_solver%bundles(i)%probes, 1) /= 0) then
-            do j = 1, ubound(mtln_solver%bundles(i)%probes, 1)
-              if (mtln_solver%bundles(i)%probes(j)%in_layer) ThereAreObservation = .true.
-            end do
-          end if
-        end do
+        if (allocated(mtln_solver%bundles)) then 
+          do i = 1, ubound(mtln_solver%bundles, 1)
+            if (ubound(mtln_solver%bundles(i)%probes, 1) /= 0) then
+              do j = 1, ubound(mtln_solver%bundles(i)%probes, 1)
+                if (mtln_solver%bundles(i)%probes(j)%in_layer) ThereAreObservation = .true.
+              end do
+            end if
+          end do
+        end if
       end block
 #endif
       !

src_main_pub/timestepping.F90

@@ -2587,7 +2587,9 @@ subroutine solver_advanceWiresE(this)
       character(len=bufsize) :: buff
 
 #ifdef CompileWithMTLN
+   if (this%thereAre%MTLNbundles) then
       call AdvanceWiresE_mtln(this%sgg,this%Idxh,this%Idyh,this%Idzh,this%eps0,this%mu0)
+   end if
 #else 
 
       if (( (trim(adjustl(this%control%wiresflavor))=='holland') .or. &

src_mtln/CMakeLists.txt

@@ -22,6 +22,8 @@ add_library(mtlnsolver
 	"circuit.F90"
 	"ngspice_interface.F90"
 	"preprocess.F90"
+	"termination_handler.F90"
+	"benchmark.F90"
 )
 
 target_link_libraries(mtlnsolver PRIVATE

src_mtln/benchmark.F90

@@ -0,0 +1,123 @@
+module benchmark_m
+
+    use FDETYPES_m, only: RKIND
+    implicit none
+
+    private
+    public :: benchmark_t
+
+    type benchmark_t
+        real(kind=rkind) :: total_time
+        real(kind=rkind) :: min_time
+        real(kind=rkind) :: max_time
+        real(kind=rkind) :: last_time
+        integer :: count
+        logical :: running
+        real(kind=rkind) :: start_time
+        character(len=128) :: name
+    contains
+        procedure :: benchmark_init
+        procedure :: benchmark_start
+        procedure :: benchmark_stop
+        procedure :: benchmark_get_elapsed
+        procedure :: benchmark_reset
+        procedure :: benchmark_report
+    end type benchmark_t
+
+contains
+
+    subroutine benchmark_init(this, name)
+        class(benchmark_t) :: this
+        character(*), intent(in), optional :: name
+
+        this%total_time = 0.0_rkind
+        this%min_time = huge(1.0_rkind)
+        this%max_time = 0.0_rkind
+        this%last_time = 0.0_rkind
+        this%count = 0
+        this%running = .false.
+        this%start_time = 0.0_rkind
+        if (present(name)) then
+            this%name = trim(name)
+        else
+            this%name = 'benchmark'
+        end if
+    end subroutine benchmark_init
+
+    subroutine benchmark_start(this)
+        class(benchmark_t) :: this
+        integer :: i, n
+
+        call system_clock(count=i, count_rate=n)
+        this%start_time = real(i, kind=rkind) / real(n, kind=rkind)
+        this%running = .true.
+    end subroutine benchmark_start
+
+    subroutine benchmark_stop(this)
+        class(benchmark_t) :: this
+        integer :: i, n
+        real(kind=rkind) :: elapsed
+
+        call system_clock(count=i, count_rate=n)
+        elapsed = real(i, kind=rkind) / real(n, kind=rkind) - this%start_time
+
+        if (elapsed < 0.0_rkind) elapsed = 0.0_rkind
+
+        this%last_time = elapsed
+        this%total_time = this%total_time + elapsed
+        this%count = this%count + 1
+
+        if (elapsed < this%min_time) this%min_time = elapsed
+        if (elapsed > this%max_time) this%max_time = elapsed
+
+        this%running = .false.
+    end subroutine benchmark_stop
+
+    function benchmark_get_elapsed(this) result(elapsed)
+        class(benchmark_t) :: this
+        real(kind=rkind) :: elapsed
+        integer :: i, n
+
+        if (this%running) then
+            call system_clock(count=i, count_rate=n)
+            elapsed = real(i, kind=rkind) / real(n, kind=rkind) - this%start_time
+            if (elapsed < 0.0_rkind) elapsed = 0.0_rkind
+        else
+            elapsed = this%last_time
+        end if
+    end function benchmark_get_elapsed
+
+    subroutine benchmark_reset(this)
+        class(benchmark_t) :: this
+
+        this%total_time = 0.0_rkind
+        this%min_time = huge(1.0_rkind)
+        this%max_time = 0.0_rkind
+        this%last_time = 0.0_rkind
+        this%count = 0
+        this%running = .false.
+        this%start_time = 0.0_rkind
+    end subroutine benchmark_reset
+
+    subroutine benchmark_report(this, unit)
+        class(benchmark_t) :: this
+        integer, intent(in), optional :: unit
+        integer :: u
+
+        u = 6
+        if (present(unit)) u = unit
+
+        if (this%count > 0) then
+            write(u, '(A, A, I5)') 'Benchmark: ', trim(this%name), this%count
+            write(u, '(A, ES12.4)') '  Total time (s): ', this%total_time
+            write(u, '(A, ES12.4)') '  Average time (s): ', this%total_time / this%count
+            write(u, '(A, ES12.4)') '  Min time (s): ', this%min_time
+            write(u, '(A, ES12.4)') '  Max time (s): ', this%max_time
+            write(u, '(A, F10.2, A)') '  Calls per second: ', &
+                 real(this%count, kind=rkind) / this%total_time, ' Hz'
+        else
+            write(u, '(A, A)') 'No measurements for benchmark: ', trim(this%name)
+        end if
+    end subroutine benchmark_report
+
+end module benchmark_m