diff --git a/.gitignore b/.gitignore
index 51f3eee22..af22edb51 100644
--- a/.gitignore
+++ b/.gitignore
@@ -82,6 +82,8 @@ tuning_list
 etc/tune
 2kprime.1
 drprimes.txt
+etc/multiplying*
+etc/squaring*
 
 # ignore stuff generated by "make manual" and "make poster"
 *.aux
@@ -134,3 +136,20 @@ build*/
 # kdevelop section
 .kdev4/
 *.kdev4
+
+# ignore cmake files
+CMakeFiles
+Makefile
+cmake_install.cmake
+
+
+
+
+
+
+
+
+
+
+
+
diff --git a/CMakeLists.txt b/CMakeLists.txt
index 014fb1883..2e4e649e2 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -32,7 +32,8 @@ include(sources.cmake)
 # Options
 #-----------------------------------------------------------------------------
 option(BUILD_SHARED_LIBS "Build shared library and only the shared library if \"ON\", default is static" OFF)
-
+option(BUILD_TUNING "Run a tuning program for the fast multiplication/squaring algorithms if \"ON\"" OFF)
+option(BUILD_GRAPHS "Run a benchmark of the fast multiplication/squaring algorithms and make graphics if \"ON\"" OFF)
 #-----------------------------------------------------------------------------
 # Compose CFLAGS
 #-----------------------------------------------------------------------------
@@ -137,6 +138,14 @@ if(BUILD_TESTING)
     add_subdirectory(demo)
 endif()
 
+#-----------------------------------------------------------------------------
+# tuning and benchmark targets
+#-----------------------------------------------------------------------------
+
+if(BUILD_TUNING OR BUILD_GRAPHS)
+    add_subdirectory(etc ${CMAKE_CURRENT_SOURCE_DIR}/etc)
+endif()
+
 #-----------------------------------------------------------------------------
 # Install/export targets and files
 #-----------------------------------------------------------------------------
diff --git a/demo/timing.c b/demo/timing.c
index a66ce8505..ce33d3a23 100644
--- a/demo/timing.c
+++ b/demo/timing.c
@@ -55,42 +55,35 @@ static unsigned int lbit(void)
    }
 }
 
-/* RDTSC from Scott Duplichan */
-static uint64_t TIMFUNC(void)
-{
-#if defined __GNUC__
-#if defined(__i386__) || defined(__x86_64__)
-   /* version from http://www.mcs.anl.gov/~kazutomo/rdtsc.html
-    * the old code always got a warning issued by gcc, clang did not complain...
-    */
-   unsigned hi, lo;
-   __asm__ __volatile__("rdtsc" : "=a"(lo), "=d"(hi));
-   return ((uint64_t)lo)|(((uint64_t)hi)<<32);
-#else /* gcc-IA64 version */
-   unsigned long result;
-   __asm__ __volatile__("mov %0=ar.itc":"=r"(result)::"memory");
-
-   while (__builtin_expect((int) result == -1, 0))
-      __asm__ __volatile__("mov %0=ar.itc":"=r"(result)::"memory");
-
-   return result;
-#endif
 
-   /* Microsoft and Intel Windows compilers */
-#elif defined _M_IX86
-   __asm rdtsc
-#elif defined _M_AMD64
-   return __rdtsc();
-#elif defined _M_IA64
-#if defined __INTEL_COMPILER
-#include <ia64intrin.h>
+#if defined(_WIN32)
+#  include <windows.h>
 #endif
-   return __getReg(3116);
+
+static uint64_t TIMFUNC(void)
+{
+#if _POSIX_C_SOURCE >= 199309L
+#define LTM_BILLION 1000000000
+   struct timespec ts;
+
+   /* TODO: Sets errno in case of error. Use? */
+   clock_gettime(CLOCK_MONOTONIC, &ts);
+   return (((uint64_t)ts.tv_sec) * LTM_BILLION + (uint64_t)ts.tv_nsec);
+#elif defined(_WIN32)
+   LARGE_INTEGER ticks;
+   QueryPerformanceCounter(&ticks);
+   return (uint64_t)ticks.QuadPart;
 #else
-#error need rdtsc function for this build
+   clock_t t;
+   t = clock();
+   if (t < (clock_t)(0)) {
+      return (uint64_t)(0);
+   }
+   return (uint64_t)(t);
 #endif
 }
 
+
 #define DO2(x) do { mp_err err = x; err = x; (void)err; }while(0)
 #define DO4(x) DO2(x); DO2(x)
 #define DO8(x) DO4(x); DO4(x)
@@ -141,6 +134,12 @@ int main(int argc, char **argv)
    int n, cnt, ix, old_kara_m, old_kara_s, old_toom_m, old_toom_s;
    unsigned rr;
 
+#ifdef _WIN32
+   LARGE_INTEGER Frequency;
+#else
+   struct timespec ts;
+#endif
+
    CHECK_OK(mp_init(&a));
    CHECK_OK(mp_init(&b));
    CHECK_OK(mp_init(&c));
@@ -150,10 +149,21 @@ int main(int argc, char **argv)
 
    srand(LTM_TIMING_RAND_SEED);
 
-
+#ifdef _WIN32
+   QueryPerformanceFrequency(&Frequency);
+   CLK_PER_SEC = (uint64) Frequency;
+#elif _POSIX_C_SOURCE >= 199309L
+   /* returns -1 for an error and 0 for okay, sets errno (not used here) */
+   if (clock_getres(CLOCK_MONOTONIC, &ts)) {
+      fprintf(stderr, "%d, clock_getres failed\n", __LINE__);
+      exit(EXIT_FAILURE);
+   }
+   CLK_PER_SEC = LTM_BILLION / ts.tv_nsec;
+#else
    CLK_PER_SEC = TIMFUNC();
    sleep(1);
    CLK_PER_SEC = TIMFUNC() - CLK_PER_SEC;
+#endif
 
    printf("CLK_PER_SEC == %" PRIu64 "\n", CLK_PER_SEC);
 
diff --git a/doc/bn.tex b/doc/bn.tex
index 63e71633b..dfedc8295 100644
--- a/doc/bn.tex
+++ b/doc/bn.tex
@@ -102,6 +102,7 @@ \section{License}
 
 \section{Building LibTomMath}
 
+
 LibTomMath is meant to be very ``GCC friendly'' as it comes with a makefile well suited for GCC.
 However, the library will also build in MSVC, Borland C out of the box.  For any other ISO C
 compiler a makefile will have to be made by the end
@@ -270,6 +271,53 @@ \subsection{Testing}
 test was invoked.  If an error is detected the program will exit with a dump of the relevant
 numbers it was working with.
 
+\subsection{CMake}
+Some of the options above are also available with CMake.
+
+\subsubsection{Shared Library}
+The default is a static library. To produce a shared library use the CMake option
+\begin{alltt}
+-DBUILD_SHARED_LIBS=ON
+\end{alltt}
+
+\subsubsection{Testing}
+To run the testsuite use option
+\begin{alltt}
+-DBUILD_TESTING=ON
+\end{alltt}
+
+\subsubsection{Tuning}
+To run the tuning itself use option
+\begin{alltt}
+-DBUILD_TUNING=ON
+\end{alltt}
+
+To run a benchmark with the tuned library and print plots of the benchmark tables use option
+\begin{alltt}
+-DBUILD_GRAPHS=ON
+\end{alltt}
+
+To compile with LTO (Link Time Optimization) use option
+\begin{alltt}
+-DCOMPILE_LTO=ON
+\end{alltt}
+
+There are several build types available:
+\begin{description}
+\item[Debug] Build a library with debugging symbols (\texttt{-g3}) and no extra optimization
+\item[Release] Build the normal release version (\texttt{-O3 -funroll-loops -fomit-frame-pointer}) (default)
+\item[RelWithDebInfo] Build a library with debugging symbols (\texttt{-g3 -O2}) and a bit of optimization
+\item[MinSizeRel] Build a small sized library (\texttt{-Os})
+\end{description}
+The build types are case-sensitive!
+
+Choose one with:
+\begin{alltt}
+-DCMAKE_BUILD_TYPE=buildtype
+\end{alltt}
+
+
+
 \section{Build Configuration}
 LibTomMath can configured at build time in two phases we shall call ``depends'' and
 ``trims''. Each phase changes how the library is built and they are applied one after another
@@ -1600,13 +1648,35 @@ \section{Tuning Polynomial Basis Routines}
 make tune
 \end{alltt}
 
-This will run a benchmark, computes the medians, rewrites \texttt{bncore.c}, and recompiles
-\texttt{bncore.c} and relinks the library.
+With CMake
+\begin{alltt}
+cmake --build /path/to/build/dir -DBUILD_TUNING=ON
+\end{alltt}
+
+
+This will run a benchmark, computes the medians, rewrites \texttt{tommath\_cutoffs.h}, recompiles
+\texttt{cutoffs.c}, and relinks the library.
 
 The benchmark itself can be fine--tuned in the file \texttt{etc/tune\_it.sh}.
 
 The program \texttt{etc/tune} is also able to print a list of values for printing curves with e.g.:
-\texttt{gnuplot}. type \texttt{./etc/tune -h} to get a list of all available options.
+\texttt{gnuplot}. Type \texttt{./etc/tune -h} to get a list of all the available options. There
+are a lot.
+
+To get some nice plots in \texttt{etc} try
+
+\begin{alltt}
+make graphs
+\end{alltt}
+
+With CMake
+\begin{alltt}
+cmake --build /path/to/build/dir -DBUILD_GRAPHS=ON
+\end{alltt}
+
+This will run a benchmark, computes the medians, rewrites \texttt{tommath\_cutoffs.h}, recompiles
+\texttt{cutoffs.c}, relinks the library and runs gnuplot to print plots in the PNG format. The size
+of the images is fixed in the file \texttt{etc/plot\_graphs.gp} and has to be changed manually.
 
 \chapter{Modular Reduction}
 
diff --git a/etc/CMakeLists.txt b/etc/CMakeLists.txt
new file mode 100644
index 000000000..0d6bc1c2c
--- /dev/null
+++ b/etc/CMakeLists.txt
@@ -0,0 +1,51 @@
+# SPDX-License-Identifier: Unlicense
+#
+# LibTomMath, a free open source portable number theoretic multiple-precision
+# integer (MPI) library written entirely in C.
+#
+
+cmake_minimum_required(VERSION 3.10)
+
+set(LTM_TUNE tune-ltm)
+
+# This file can be included from the top level or used stand-alone
+if(PROJECT_NAME)
+    set(LIBRARY_NAME ${PROJECT_NAME})
+else()
+    # Define an independent project and all the necessary stuff around
+    project(${LTM_TUNE}
+        LANGUAGES C)
+    set(LIBRARY_NAME libtommath)
+    find_package(${LIBRARY_NAME})
+    if(NOT CMAKE_BUILD_TYPE AND NOT CMAKE_CONFIGURATION_TYPES)
+        set(CMAKE_BUILD_TYPE "Release")
+    endif()
+endif()
+
+add_executable(tune
+    ${CMAKE_CURRENT_SOURCE_DIR}/tune.c
+)
+
+target_include_directories(tune PRIVATE
+    ${CMAKE_CURRENT_SOURCE_DIR}
+    ${CMAKE_CURRENT_SOURCE_DIR}/..
+)
+
+target_link_libraries(tune PRIVATE
+    ${LIBRARY_NAME}
+)
+
+target_compile_options(tune PRIVATE
+    ${LTM_C_FLAGS}
+)
+target_link_options(tune BEFORE PUBLIC
+    ${LTM_LD_FLAGS}
+)
+
+if(BUILD_GRAPHS)
+   # used in tune_it.sh
+   find_program(GNUPLOT gnuplot)
+   add_custom_command(TARGET tune POST_BUILD COMMAND  ${CMAKE_CURRENT_SOURCE_DIR}/tune_it.sh 1000 WORKING_DIRECTORY ${CMAKE_CURRENT_SOURCE_DIR} VERBATIM)
+else()
+   add_custom_command(TARGET tune POST_BUILD COMMAND  ${CMAKE_CURRENT_SOURCE_DIR}/tune_it.sh WORKING_DIRECTORY ${CMAKE_CURRENT_SOURCE_DIR} VERBATIM)
+endif()
diff --git a/etc/makefile b/etc/makefile
index 0e178396f..ecdf9b1f7 100644
--- a/etc/makefile
+++ b/etc/makefile
@@ -6,7 +6,7 @@ LTM_TUNE_CFLAGS = $(CFLAGS) $(LTM_CFLAGS) -Wall -W -Wextra -Wshadow -O3 -I../
 # libname when you can't install the lib with install
 LIBNAME=../libtommath.a
 
-all: pprime tune test_standalone mersenne drprime 2kprime mont
+all: pprime tune test_standalone mersenne drprime 2kprime mont getlimbsize graph
 
 #provable primes
 pprime: pprime.o
@@ -36,10 +36,15 @@ drprime: drprime.o
 mont: mont.o
 	$(CC) $(LTM_TUNE_CFLAGS) mont.o $(LIBNAME) -o mont
 
+# Make pretty pictures (1000 is the maximum number of limbs to print for mul/sqr)
+# "tune" runs twice because it runs automatically when build.
+graphs: tune
+	./tune_it.sh 1000
 
 clean:
 	rm -f *.log *.o *.obj *.exe pprime tune mersenne drprime mont 2kprime pprime.dat \
-        tuning_list multiplying squaring test *.da *.dyn *.dpi *~
+        tuning_list get_limbsize out *.da *.dyn *.dpi *~ cmake_install.cmake Makefile
 	rm -rf .libs
+	rm -rf CMakeFiles
 
 .PHONY: tune
diff --git a/etc/makefile.icc b/etc/makefile.icc
index 9217f7b1d..2a7272a1e 100644
--- a/etc/makefile.icc
+++ b/etc/makefile.icc
@@ -32,20 +32,10 @@ tune: tune.o
 	$(CC) $(CFLAGS) tune.o $(LIBNAME) -o tune
 	./tune_it.sh
 
-# same app but using RDTSC for higher precision [requires 80586+], coff based gcc installs [e.g. ming, cygwin, djgpp]
-tune86: tune.c
-	nasm -f coff timer.asm
-	$(CC) -DX86_TIMER $(CFLAGS) tune.c timer.o  $(LIBNAME) -o tune86
-
-# for cygwin
-tune86c: tune.c
-	nasm -f gnuwin32 timer.asm
-	$(CC) -DX86_TIMER $(CFLAGS) tune.c timer.o  $(LIBNAME) -o tune86
-
-#make tune86 for linux or any ELF format
-tune86l: tune.c
-	nasm -f elf -DUSE_ELF timer.asm
-	$(CC) -DX86_TIMER $(CFLAGS) tune.c timer.o $(LIBNAME) -o tune86l
+# Make pretty pictures (1000 is the maximum number of limbs to print for mul/sqr)
+# "tune" runs twice because it runs automatically when build.
+graphs: tune
+	./tune_it.sh 1000
 
 # spits out mersenne primes
 mersenne: mersenne.o
@@ -64,4 +54,4 @@ mont: mont.o
 
 
 clean:
-	rm -f *.log *.o *.obj *.exe pprime tune mersenne drprime tune86 tune86l mont 2kprime pprime.dat *.il tuning_list
+	rm -f *.log *.o *.obj *.exe pprime tune mersenne drprime mont 2kprime pprime.dat get_limbsize *.il tuning_list
diff --git a/etc/plot_graphs.gp b/etc/plot_graphs.gp
new file mode 100644
index 000000000..c78511d1c
--- /dev/null
+++ b/etc/plot_graphs.gp
@@ -0,0 +1,19 @@
+set term pngcairo size 720,540
+# Good for most colorblinds
+set colorsequence podo
+
+set key top left;
+
+set ylabel "Time"
+set xlabel "Operand size (limbs)"
+
+set output "multiplying".ARG1.".png";
+set title "Comparing fast and slow multiplying [".ARG1." bits limbsize]";
+plot "multiplying".ARG1."" using 1:2 w lines t "slow", "multiplying".ARG1."" using 1:3 w lines t "fast"
+
+set output "squaring".ARG1.".png";
+set title "Comparing fast and slow squaring [".ARG1." bits limbsize]";
+plot "squaring".ARG1."" using 1:2 w lines t "slow", "squaring".ARG1."" using 1:3 w lines t "fast"
+
+
+
diff --git a/etc/timer.asm b/etc/timer.asm
deleted file mode 100644
index 35890d985..000000000
--- a/etc/timer.asm
+++ /dev/null
@@ -1,37 +0,0 @@
-; x86 timer in NASM
-;
-; Tom St Denis, tomstdenis@iahu.ca
-[bits 32]
-[section .data]
-time dd 0, 0
-
-[section .text]
-
-%ifdef USE_ELF
-[global t_start]
-t_start:
-%else
-[global _t_start]
-_t_start:
-%endif
-   push edx
-   push eax
-   rdtsc
-   mov [time+0],edx
-   mov [time+4],eax
-   pop eax
-   pop edx
-   ret
-   
-%ifdef USE_ELF
-[global t_read]
-t_read:
-%else
-[global _t_read]
-_t_read:
-%endif
-   rdtsc
-   sub eax,[time+4]
-   sbb edx,[time+0]
-   ret
-   
\ No newline at end of file
diff --git a/etc/tune.c b/etc/tune.c
index 8ad202890..2f78da9cc 100644
--- a/etc/tune.c
+++ b/etc/tune.c
@@ -24,6 +24,9 @@ static uint64_t s_time_mul(int size);
 static uint64_t s_time_sqr(int size);
 static void s_usage(char *s);
 
+#if defined(_WIN32)
+#  include <windows.h>
+#endif
 static uint64_t s_timer_function(void)
 {
 #if _POSIX_C_SOURCE >= 199309L
@@ -33,6 +36,10 @@ static uint64_t s_timer_function(void)
    /* TODO: Sets errno in case of error. Use? */
    clock_gettime(CLOCK_MONOTONIC, &ts);
    return (((uint64_t)ts.tv_sec) * LTM_BILLION + (uint64_t)ts.tv_nsec);
+#elif defined(_WIN32)
+   LARGE_INTEGER ticks;
+   QueryPerformanceCounter(&ticks);
+   return (uint64_t)ticks.QuadPart;
 #else
    clock_t t;
    t = clock();
@@ -42,7 +49,6 @@ static uint64_t s_timer_function(void)
    return (uint64_t)(t);
 #endif
 }
-
 /* generic ISO C timer */
 static uint64_t s_timer_tmp;
 static void s_timer_start(void)
@@ -66,7 +72,7 @@ static uint64_t s_time_mul(int size)
    int x;
    mp_err  e;
    mp_int  a, b, c, d;
-   uint64_t t1;
+   uint64_t t1 = 0u;
 
    if ((e = mp_init_multi(&a, &b, &c, &d, NULL)) != MP_OKAY) {
       t1 = UINT64_MAX;
@@ -82,12 +88,14 @@ static uint64_t s_time_mul(int size)
       goto LBL_ERR;
    }
 
-   s_timer_start();
+
    for (x = 0; x < s_number_of_test_loops; x++) {
+      s_timer_start();
       if ((e = mp_mul(&a,&b,&c)) != MP_OKAY) {
          t1 = UINT64_MAX;
          goto LBL_ERR;
       }
+      t1 += s_timer_stop();
       if (s_check_result == 1) {
          if ((e = s_mp_mul_full(&a,&b,&d)) != MP_OKAY) {
             t1 = UINT64_MAX;
@@ -101,7 +109,7 @@ static uint64_t s_time_mul(int size)
       }
    }
 
-   t1 = s_timer_stop();
+
 LBL_ERR:
    mp_clear_multi(&a, &b, &c, &d, NULL);
    return t1;
@@ -112,7 +120,7 @@ static uint64_t s_time_sqr(int size)
    int x;
    mp_err  e;
    mp_int  a, b, c;
-   uint64_t t1;
+   uint64_t t1 = 0u;
 
    if ((e = mp_init_multi(&a, &b, &c, NULL)) != MP_OKAY) {
       t1 = UINT64_MAX;
@@ -124,12 +132,14 @@ static uint64_t s_time_sqr(int size)
       goto LBL_ERR;
    }
 
-   s_timer_start();
+
    for (x = 0; x < s_number_of_test_loops; x++) {
+      s_timer_start();
       if ((e = mp_sqr(&a,&b)) != MP_OKAY) {
          t1 = UINT64_MAX;
          goto LBL_ERR;
       }
+      t1 += s_timer_stop();
       if (s_check_result == 1) {
          if ((e = s_mp_sqr(&a,&c)) != MP_OKAY) {
             t1 = UINT64_MAX;
@@ -142,7 +152,7 @@ static uint64_t s_time_sqr(int size)
       }
    }
 
-   t1 = s_timer_stop();
+
 LBL_ERR:
    mp_clear_multi(&a, &b, &c, NULL);
    return t1;
@@ -288,7 +298,7 @@ int main(int argc, char **argv)
    int opt;
    struct cutoffs orig, updated;
 
-   FILE *squaring, *multiplying;
+   FILE *squaring, *multiplying, *out;
    char mullog[256] = "multiplying";
    char sqrlog[256] = "squaring";
    s_number_of_test_loops = 64;
@@ -309,6 +319,14 @@ int main(int argc, char **argv)
    if (argc != 1) {
       for (opt = 1; (opt < argc) && (argv[opt][0] == '-'); opt++) {
          switch (argv[opt][1]) {
+         case 'Z':
+            out = fopen("out", "w");
+            if (out == NULL) {
+               fprintf(stderr, "Opening file \"%s\" failed\n", "out");
+               exit(EXIT_FAILURE);
+            }
+            fprintf(out,"%d",MP_DIGIT_BIT);
+            exit(EXIT_SUCCESS);
          case 'T':
             args.testmode = 1;
             s_check_result = 1;
@@ -433,6 +451,7 @@ int main(int argc, char **argv)
       }
    }
 
+
    /*
      mp_rand uses the cryptographically secure
      source of the OS by default. That is too expensive, too slow and
@@ -469,7 +488,6 @@ int main(int argc, char **argv)
          if (test[n].fn != NULL) {
             s_run(test[n].name, test[n].fn, test[n].cutoff);
             *test[n].update = *test[n].cutoff;
-            *test[n].cutoff = INT_MAX;
          }
       }
    }
@@ -479,14 +497,10 @@ int main(int argc, char **argv)
              updated.SQR_KARATSUBA,
              updated.MUL_TOOM,
              updated.SQR_TOOM);
-   } else {
-      printf("MUL_KARATSUBA_CUTOFF = %d\n", updated.MUL_KARATSUBA);
-      printf("SQR_KARATSUBA_CUTOFF = %d\n", updated.SQR_KARATSUBA);
-      printf("MUL_TOOM_CUTOFF = %d\n", updated.MUL_TOOM);
-      printf("SQR_TOOM_CUTOFF = %d\n", updated.SQR_TOOM);
    }
 
    if (args.print == 1) {
+
       printf("Printing data for graphing to \"%s\" and \"%s\"\n",mullog, sqrlog);
 
       multiplying = fopen(mullog, "w+");
diff --git a/etc/tune_it.sh b/etc/tune_it.sh
index dba5b6968..2666080bd 100755
--- a/etc/tune_it.sh
+++ b/etc/tune_it.sh
@@ -61,7 +61,7 @@ i=1
 while [ $i -le $LIMIT ]; do
    RNUM=$(LCG)
    printf "\r%d" $i
-   "$MPWD"/tune -t -r $RLOOPS -L $LAG -S "$RNUM" -o $OFFSET >> $FILE_NAME || die "tune" $?
+   "$MPWD"/tune -t  -r $RLOOPS -L $LAG -S "$RNUM" -o $OFFSET >> $FILE_NAME || die "tune" $?
    i=$((i + 1))
 done
 
@@ -92,15 +92,29 @@ END_OF_INPUT
 i=$(tail -n +2 $FILE_NAME | wc -l)
 # our median point will be at $i entries
 i=$(( (i / 2) + 1 ))
-TMP=$(median $FILE_NAME 1 $i)
-echo "#define MP_DEFAULT_MUL_KARATSUBA_CUTOFF $TMP"
-echo "#define MP_DEFAULT_MUL_KARATSUBA_CUTOFF $TMP" >> $TOMMATH_CUTOFFS_H || die "(km) Appending to $TOMMATH_CUTOFFS_H" $?
-TMP=$(median $FILE_NAME 2 $i)
-echo "#define MP_DEFAULT_SQR_KARATSUBA_CUTOFF $TMP"
-echo "#define MP_DEFAULT_SQR_KARATSUBA_CUTOFF $TMP" >> $TOMMATH_CUTOFFS_H || die "(ks) Appending to $TOMMATH_CUTOFFS_H" $?
-TMP=$(median $FILE_NAME 3 $i)
-echo "#define MP_DEFAULT_MUL_TOOM_CUTOFF      $TMP"
-echo "#define MP_DEFAULT_MUL_TOOM_CUTOFF      $TMP" >> $TOMMATH_CUTOFFS_H || die "(tc3m) Appending to $TOMMATH_CUTOFFS_H" $?
-TMP=$(median $FILE_NAME 4 $i)
-echo "#define MP_DEFAULT_SQR_TOOM_CUTOFF      $TMP"
-echo "#define MP_DEFAULT_SQR_TOOM_CUTOFF      $TMP" >> $TOMMATH_CUTOFFS_H || die "(tc3s) Appending to $TOMMATH_CUTOFFS_H" $?
+TMP1=$(median $FILE_NAME 1 $i)
+echo "#define MP_DEFAULT_MUL_KARATSUBA_CUTOFF $TMP1"
+echo "#define MP_DEFAULT_MUL_KARATSUBA_CUTOFF $TMP1" >> $TOMMATH_CUTOFFS_H || die "(km) Appending to $TOMMATH_CUTOFFS_H" $?
+TMP2=$(median $FILE_NAME 2 $i)
+echo "#define MP_DEFAULT_SQR_KARATSUBA_CUTOFF $TMP2"
+echo "#define MP_DEFAULT_SQR_KARATSUBA_CUTOFF $TMP2" >> $TOMMATH_CUTOFFS_H || die "(ks) Appending to $TOMMATH_CUTOFFS_H" $?
+TMP3=$(median $FILE_NAME 3 $i)
+echo "#define MP_DEFAULT_MUL_TOOM_CUTOFF      $TMP3"
+echo "#define MP_DEFAULT_MUL_TOOM_CUTOFF      $TMP3" >> $TOMMATH_CUTOFFS_H || die "(tc3m) Appending to $TOMMATH_CUTOFFS_H" $?
+TMP4=$(median $FILE_NAME 4 $i)
+echo "#define MP_DEFAULT_SQR_TOOM_CUTOFF      $TMP4"
+echo "#define MP_DEFAULT_SQR_TOOM_CUTOFF      $TMP4" >> $TOMMATH_CUTOFFS_H || die "(tc3s) Appending to $TOMMATH_CUTOFFS_H" $?
+
+# Print the tables for the graphs. Please do not change.
+if [ $# -eq 1 ]; then
+   OPTION0=$1
+   # Do not forget to raise if there are more fast algorithms with higher cutoffs.
+   if [ $OPTION0 -lt 500 ]; then
+      OPTION0=500
+   fi
+   "$MPWD"/tune -Z
+   "$MPWD"/tune -p -r $RLOOPS -L $LAG -S "$RNUM" -o $OFFSET -M $1 -G $(cat out) -s $TMP1,$TMP2,$TMP3,$TMP4
+   gnuplot -c  plot_graphs.gp $(cat out)
+fi
+
+
diff --git a/logs/Makefile b/logs/Makefile
index 59987e519..0d73db987 100644
--- a/logs/Makefile
+++ b/logs/Makefile
@@ -6,4 +6,4 @@ cmp:
 	gnuplot before_after.dem
 
 clean:
-	rm -f *-*.log *.png graphs-*.dem
+	rm -f *.log *.png graphs-*.dem
diff --git a/logs/before_after.dem b/logs/before_after.dem
index 33721ad20..3d7b9edee 100644
--- a/logs/before_after.dem
+++ b/logs/before_after.dem
@@ -1,26 +1,34 @@
 set terminal png
+# Good for most colorblinds
+set colorsequence podo
+set key top left;
+
 set ylabel "Cycles per Operation"
 set xlabel "Operand size (bits)"
 
-set output "addsub-ba.png"
+set output "addsub-ba-".ARG1.".png"
+set title "Addition/subtraction before and after profiling. Version: ".ARG1
 plot 'add-before.log' smooth bezier title "Addition (before)", \
       'add-after.log' smooth bezier title "Addition (after)", \
       'sub-before.log' smooth bezier title "Subtraction (before)", \
       'sub-after.log' smooth bezier title "Subtraction (after)"
 
-set output "mult-ba.png"
+set output "mult-ba-".ARG1.".png"
+set title "Multiplication before and after profiling. Version: ".ARG1
 plot 'mult-before.log' smooth bezier title "Multiplication (without Karatsuba) (before)", \
       'mult-after.log' smooth bezier title "Multiplication (without Karatsuba) (after)", \
       'mult_kara-before.log' smooth bezier title "Multiplication (Karatsuba) (before)", \
       'mult_kara-after.log' smooth bezier title "Multiplication (Karatsuba) (after)"
 
-set output "sqr-ba.png"
+set output "sqr-ba-".ARG1.".png"
+set title "Squaring before and after profiling. Version: ".ARG1
 plot 'sqr-before.log' smooth bezier title "Squaring (without Karatsuba) (before)", \
       'sqr-after.log' smooth bezier title "Squaring (without Karatsuba) (after)", \
       'sqr_kara-before.log' smooth bezier title "Squaring (Karatsuba) (before)", \
       'sqr_kara-after.log' smooth bezier title "Squaring (Karatsuba) (after)"
 
-set output "expt-ba.png"
+set output "expt-ba-".ARG1.".png"
+set title "Exponentiation algorithms before and after profiling. Version: ".ARG1
 plot 'expt-before.log' smooth bezier title "Exptmod (Montgomery) (before)", \
       'expt-after.log' smooth bezier title "Exptmod (Montgomery) (after)", \
       'expt_dr-before.log' smooth bezier title "Exptmod (Diminished Radix) (before)", \
@@ -30,7 +38,8 @@ plot 'expt-before.log' smooth bezier title "Exptmod (Montgomery) (before)", \
       'expt_2kl-before.log' smooth bezier title "Exptmod (2k-l Reduction) (before)", \
       'expt_2kl-after.log' smooth bezier title "Exptmod (2k-l Reduction) (after)"
 
-set output "invmod-ba.png"
+set output "invmod-ba-".ARG1.".png"
+set title "Invmod algorithms before and after profiling. Version: ".ARG1
 plot 'invmod-before.log' smooth bezier title "Modular Inverse (before)", \
       'invmod-after.log' smooth bezier title "Modular Inverse (after)"
 
diff --git a/makefile b/makefile
index 8f211f5f2..e112fb2ff 100644
--- a/makefile
+++ b/makefile
@@ -69,17 +69,62 @@ profiled:
 	rm -f *.a *.o timing
 	make CFLAGS="$(CFLAGS) -fbranch-probabilities"
 
+# run tune first, than optimize branching
+profiled_tuned: tune
+	make CFLAGS="$(CFLAGS) -fprofile-arcs" timing
+	./timing
+	rm -f *.a *.o timing
+	make CFLAGS="$(CFLAGS) -fbranch-probabilities"
+
+# run tune first, optimize branching, run tune again
+# (running it in a loop until the timings stabilize is indeed tempting)
+profiled_tuned_tuned: tune
+	make CFLAGS="$(CFLAGS) -fprofile-arcs" timing
+	./timing
+	rm -f *.a *.o timing
+	make CFLAGS="$(CFLAGS) -fbranch-probabilities"
+	./etc/tune_it.sh
+
 #make a single object profiled library
 amalgamated_timing: pre_gen
 	$(CC) $(LTM_CFLAGS) -fprofile-arcs -c pre_gen/tommath_amalgam.c -o tommath_amalgam.o
 	$(CC) $(LTM_CFLAGS) -DMP_VERSION=\"before\" demo/timing.c tommath_amalgam.o -lgcov -o timing
 
+
+amalgamated_timing_tuned: tune pre_gen
+	$(CC) $(LTM_CFLAGS) -fprofile-arcs -c pre_gen/tommath_amalgam.c -o tommath_amalgam.o
+	$(CC) $(LTM_CFLAGS) -DMP_VERSION=\"before\" demo/timing.c tommath_amalgam.o -lgcov -o timing
+
 profiled_single: amalgamated_timing
 	./timing
 	rm -f *.o timing
 	$(CC) $(LTM_CFLAGS) -fbranch-probabilities -c pre_gen/tommath_amalgam.c -o tommath_amalgam.o
 	$(AR) $(ARFLAGS) $(LIBNAME) tommath_amalgam.o
 
+# run tune first, than optimize branching
+profiled_single_tuned: amalgamated_timing_tuned
+	./timing
+	rm -f *.a *.o timing
+	$(CC) $(LTM_CFLAGS) -fbranch-probabilities -c pre_gen/tommath_amalgam.c -o tommath_amalgam.o
+	$(AR) $(ARFLAGS) $(LIBNAME) tommath_amalgam.o
+
+# run tune first, optimize branching, run tune again
+# (running it in a loop until the timings stabilize is indeed tempting)
+profiled_single_tuned_tuned: amalgamated_timing_tuned
+	./timing
+	rm -f *.o timing
+	$(CC) $(LTM_CFLAGS) -fbranch-probabilities -c pre_gen/tommath_amalgam.c -o tommath_amalgam.o
+	$(AR) $(ARFLAGS) $(LIBNAME) tommath_amalgam.o
+	rm -f etc/tune.o
+	$(CC) $(LTM_CFLAGS) -c etc/tune.c -o etc/tune.o
+	$(CC) $(LTM_CFLAGS) tommath_amalgam.o  etc/tune.o -o etc/tune
+	./etc/tune_it.sh
+	rm -f *.a *.o
+	$(CC) $(LTM_CFLAGS) -fprofile-arcs -c pre_gen/tommath_amalgam.c -o tommath_amalgam.o
+	$(CC) $(LTM_CFLAGS) -fbranch-probabilities -c pre_gen/tommath_amalgam.c -o tommath_amalgam.o
+	$(AR) $(ARFLAGS) $(LIBNAME) tommath_amalgam.o
+	
+
 install: $(LIBNAME) .install_common
 	install -m 644 $(LIBNAME) $(DESTDIR)$(LIBPATH)
 	install -m 644 $(HEADERS_PUB) $(DESTDIR)$(INCPATH)
@@ -108,6 +153,10 @@ tune: $(LIBNAME)
 	$(MAKE) -C etc tune CFLAGS="$(LTM_CFLAGS) -I../"
 	$(MAKE)
 
+graphs: $(LIBNAME)
+	$(MAKE) -C etc graphs CFLAGS="$(LTM_CFLAGS) -I../"
+	$(MAKE)
+
 etc-all: $(LIBNAME)
 	$(MAKE) -C etc all CFLAGS="$(LTM_CFLAGS) -I../"
 	$(MAKE)
@@ -127,6 +176,16 @@ cmp: profiled_single
 	./timing
 	$(MAKE) -C logs/ cmp
 
+cmp_tuned: profiled_single_tuned
+	$(CC) $(LTM_CFLAGS) -DMP_VERSION=\"after\" demo/timing.c $(LIBNAME) -lgcov -o timing
+	./timing
+	$(MAKE) -C logs/ cmp
+
+cmp_tuned_tuned: profiled_single_tuned_tuned
+	$(CC) $(LTM_CFLAGS) -DMP_VERSION=\"after\" demo/timing.c $(LIBNAME) -lgcov -o timing
+	./timing
+	$(MAKE) -C logs/ cmp
+
 zipup:
 	$(MAKE) clean
 	$(MAKE) .zipup
diff --git a/makefile.mingw b/makefile.mingw
index e2445e8a0..80e66652c 100644
--- a/makefile.mingw
+++ b/makefile.mingw
@@ -86,6 +86,11 @@ tune: $(LIBNAME_S)
 	$(MAKE) -C etc tune
 	$(MAKE)
 
+graphs: $(LIBNAME)
+	$(MAKE) -C etc graphs
+	$(MAKE)
+
+
 clean:
 	@-cmd /c del /Q /S *.o *.a *.exe *.dll 2>nul
 
diff --git a/makefile.msvc b/makefile.msvc
index 8feb425c4..4fa88517e 100644
--- a/makefile.msvc
+++ b/makefile.msvc
@@ -86,6 +86,10 @@ tune: $(LIBMAIN_S)
 	$(MAKE) -C etc tune
 	$(MAKE)
 
+graphs: $(LIBNAME)
+	$(MAKE) -C etc graphs
+	$(MAKE)
+
 clean-obj:
 	@-cmd /c del /Q /S *.OBJ 2>nul
 
diff --git a/makefile.shared b/makefile.shared
index 50c335269..7bf0e1aae 100644
--- a/makefile.shared
+++ b/makefile.shared
@@ -92,8 +92,19 @@ $(foreach demo, $(strip $(DEMOS)), $(eval $(call DEMO_template,$(demo))))
 mtest:
 	cd mtest ; $(CC) $(LTM_CFLAGS) -O0 mtest.c $(LTM_LDFLAGS) -o mtest
 
-tune: $(LIBNAME)
-	$(LTCOMPILE) $(LTM_CFLAGS) -c etc/tune.c -o etc/tune.o
-	$(LTLINK) $(LTM_LDFLAGS) -o etc/tune etc/tune.o $(LIBNAME)
-	cd etc/; /bin/sh tune_it.sh; cd ..
+# etc/tune.c uses hidden symbols, so we assume that only difference between the shared and static
+# libraries is just that: they are shared or static and simply build a static library to run
+# etc/tune
+tune:
+	$(MAKE) -f makefile tune
+	$(MAKE) -f makefile clean
 	$(MAKE) -f makefile.shared
+
+graphs:
+	$(MAKE) -f makefile graphs
+	$(MAKE) -f makefile clean
+	$(MAKE) -f makefile.shared
+
+
+
+
diff --git a/makefile.unix b/makefile.unix
index 58642098d..653b6a5ec 100644
--- a/makefile.unix
+++ b/makefile.unix
@@ -85,6 +85,10 @@ tune: $(LIBMAIN_S)
 	$(MAKE) -C etc tune
 	$(MAKE)
 
+graphs: $(LIBNAME)
+	$(MAKE) -C etc graphs
+	$(MAKE)
+
 #NOTE: this makefile works also on cygwin, thus we need to delete *.exe
 clean:
 	-@rm -f $(OBJECTS) $(LIBMAIN_S)
diff --git a/makefile_include.mk b/makefile_include.mk
index d47ea2ba2..4aadc770f 100644
--- a/makefile_include.mk
+++ b/makefile_include.mk
@@ -168,7 +168,6 @@ check: test
 	./test
 
 #make the code coverage of the library
-#
 coverage: LTM_CFLAGS += -fprofile-arcs -ftest-coverage -DTIMING_NO_LOGS
 coverage: LTM_LFLAGS += -lgcov
 coverage: LTM_LDFLAGS += -lgcov
@@ -192,8 +191,9 @@ cleancov: cleancov-clean clean
 clean:
 	rm -f *.gcda *.gcno *.gcov *.bat *.o *.a *.obj *.lib *.exe *.dll etclib/*.o \
 				demo/*.o test timing mtest_opponent mtest/mtest mtest/mtest.exe tuning_list \
-				*.s tommath_amalgam.c pre_gen/tommath_amalgam.c *.da *.dyn *.dpi tommath.tex \
+				*.s tommath_amalgam.c *.da *.dyn *.dpi tommath.tex \
+                                cmake_install.cmake Makefile \
 				`find . -type f | grep [~] | xargs` *.lo *.la
-	rm -rf .libs/ demo/.libs
+	rm -rf .libs/ demo/.libs CMakeFiles pre_gen
 	${MAKE} -C etc/ clean MAKE=${MAKE}
 	${MAKE} -C doc/ clean MAKE=${MAKE}