hw brickbench: Benchmark test for NXT and EV3
Re: hw brickbench: Benchmark test for NXT and EV3
Display output tests are impossible right now, since the functionality has not been implemented yet.
I implemented a native shellsort opcode for char arrays and it showed a 1000-fold speed up, when compared to running the same algorithm in ROBOTC. I haven't had time to make it work for other types yet.
= Xander
I implemented a native shellsort opcode for char arrays and it showed a 1000-fold speed up, when compared to running the same algorithm in ROBOTC. I haven't had time to make it work for other types yet.
= Xander
| My Blog: I'd Rather Be Building Robots (http://botbench.com)
| RobotC 3rd Party Driver Suite: (http://rdpartyrobotcdr.sourceforge.net)
| Some people, when confronted with a problem, think, "I know, I'll use threads,"
| and then two they hav erpoblesms. (@nedbat)
| RobotC 3rd Party Driver Suite: (http://rdpartyrobotcdr.sourceforge.net)
| Some people, when confronted with a problem, think, "I know, I'll use threads,"
| and then two they hav erpoblesms. (@nedbat)
Re: hw brickbench: Benchmark test for NXT and EV3
Xander,
thank you for this information,
but then please just post your code as-it-is, in the version by which you generated your published results!
The RobotC benchmarks can be updated at any time when your code will be available in an updated version, too.
thank you for this information,
but then please just post your code as-it-is, in the version by which you generated your published results!
The RobotC benchmarks can be updated at any time when your code will be available in an updated version, too.
Re: hw brickbench: Benchmark test for NXT and EV3
Helmut,
I've added it to the samples folder that comes with ROBOTC 4.x. I am not able to access that stuff right now, but I will when I get home.
= Xander
Edit: see attached.
I've added it to the samples folder that comes with ROBOTC 4.x. I am not able to access that stuff right now, but I will when I get home.
= Xander
Edit: see attached.
- Attachments
-
- benchmark.c
- (9.56 KiB) Downloaded 1075 times
| My Blog: I'd Rather Be Building Robots (http://botbench.com)
| RobotC 3rd Party Driver Suite: (http://rdpartyrobotcdr.sourceforge.net)
| Some people, when confronted with a problem, think, "I know, I'll use threads,"
| and then two they hav erpoblesms. (@nedbat)
| RobotC 3rd Party Driver Suite: (http://rdpartyrobotcdr.sourceforge.net)
| Some people, when confronted with a problem, think, "I know, I'll use threads,"
| and then two they hav erpoblesms. (@nedbat)
Re: hw brickbench: Benchmark test for NXT and EV3
thx for the code, Xander.
at the top it's still announced to be the 3.6 code - is that true?
or is it the 4.0 code?
at the top it's still announced to be the 3.6 code - is that true?
or is it the 4.0 code?
Re: hw brickbench: Benchmark test for NXT and EV3
Oh, I didn't change that part, this is what's shipped with 4.x. I'll fix the header in the benchmark when the next batch of functionality gets added to ROBOTC.
= Xander
= Xander
| My Blog: I'd Rather Be Building Robots (http://botbench.com)
| RobotC 3rd Party Driver Suite: (http://rdpartyrobotcdr.sourceforge.net)
| Some people, when confronted with a problem, think, "I know, I'll use threads,"
| and then two they hav erpoblesms. (@nedbat)
| RobotC 3rd Party Driver Suite: (http://rdpartyrobotcdr.sourceforge.net)
| Some people, when confronted with a problem, think, "I know, I'll use threads,"
| and then two they hav erpoblesms. (@nedbat)
-
- Posts: 323
- Joined: 29 Sep 2010, 05:03
Re: hw brickbench: Benchmark test for NXT and EV3
Doc the source code for the leJOS version for the EV3 is here:
https://drive.google.com/file/d/0BwTg7x ... sp=sharing
and the NXT here:
https://drive.google.com/file/d/0BwTg7x ... sp=sharing
For some reason you do not seem to be showing the results for test 0-3 for the EV3 version of leJOS.
https://drive.google.com/file/d/0BwTg7x ... sp=sharing
and the NXT here:
https://drive.google.com/file/d/0BwTg7x ... sp=sharing
For some reason you do not seem to be showing the results for test 0-3 for the EV3 version of leJOS.
Re: hw brickbench: Benchmark test for NXT and EV3
thank you very much, Andy!
I'll post it here in pure text for better reading:
hw benchmark for leJOS/Java for NXT:
########################################################################################################
hw benchmark for leJOS/Java for EV3:
########################################################################################################
BTW: Is meanwhile everything finished for the EV3, too, e.g. display benchmarks (or will be still anything changed here), and has the float vs. double thing been fixed intermediately?
What about a java-fw-included sort algorithm?
ps
for the NXT code you wrote " leJOS/EV3 0.8.0-alpha" - but it's probably 0.9.anything, pls CMIIW...?
pps
another question re. the Java JIT compiler:
have the benchs been run once or several times before you took the time?
I'll post it here in pure text for better reading:
hw benchmark for leJOS/Java for NXT:
Code: Select all
import javax.microedition.lcdui.Graphics;
import lejos.nxt.Button;
import lejos.nxt.LCD;
//hw brickbench
//benchmark test for NXT/EV3 and similar Micro Controllers
//PL: leJOS/NXT 0.9.1beta
//Autor: (C) Helmut Wunder 2013
// Ported to leJOS by Andy Shaw.
//freie Verwendung für private Zwecke
//für kommerzielle Zwecke nur nach Genehmigung durch den Autor.
//protected under the friendly Creative Commons Attribution-NonCommercial-ShareAlike 3.0 Unported License
//http://creativecommons.org/licenses/by-nc-sa/3.0/
//version 1.08
// PLEASE NOTE: This is not a good example of a Java program. It is a simple port of a C program. No
// attempt has been made to make use of Java features.
public class MBBenchmark
{
static long[] runtime = new long[8];
static int[] a = new int[500], b = new int[500], c = new int[500],
t = new int[500];
// --------------------------------------------
// Mersenne Twister
// --------------------------------------------
static int[] y = new int[25];
static int index = 25 + 1;
static final int M = 7;
static final int[] A = { 0, 0x8ebfd028 };
static int randM()
{
if (index >= 25)
{
int k;
if (index > 25)
{
int r = 9, s = 3402;
for (k = 0; k < 25; ++k)
{
r = 509845221 * r + 3;
s *= s + 1;
y[k] = s + (r >> 10);
}
}
for (k = 0; k < 25 - M; ++k)
y[k] = y[k + M] ^ (y[k] >> 1) ^ A[y[k] & 1];
for (; k < 25; ++k)
y[k] = y[k + (M - 25)] ^ (y[k] >> 1) ^ A[y[k] & 1];
index = 0;
}
int e = y[index++];
e ^= (e << 7) & 0x2b5b2500;
e ^= (e << 15) & 0xdb8b0000;
e ^= (e >> 16);
return e;
}
// --------------------------------------------
// Matrix Algebra
// --------------------------------------------
// matrix * matrix multiplication (matrix product)
static void MatrixMatrixMult(int N, int M, int K, double A[][],
double B[][], double C[][])
{
int i, j, s; // matrix A: N x M // B: M x K // C: N x K
for (i = 0; i < N; ++i)
{
for (j = 0; j < K; ++j)
{
C[i][j] = 0;
for (s = 0; s < M; ++s)
{
C[i][j] = C[i][j] + A[i][s] * B[s][j];
}
}
}
}
// matrix determinant
static double MatrixDet(int N, double A[][])
{
int i, j, i_count, j_count, count = 0;
double[][] Asub = new double[N - 1][N - 1];
double det = 0;
if (N == 1)
return A[0][0];
if (N == 2)
return (A[0][0] * A[1][1] - A[0][1] * A[1][0]);
for (count = 0; count < N; count++)
{
i_count = 0;
for (i = 1; i < N; i++)
{
j_count = 0;
for (j = 0; j < N; j++)
{
if (j == count)
continue;
Asub[i_count][j_count] = A[i][j];
j_count++;
}
i_count++;
}
det += Math.pow(-1, count) * A[0][count] * MatrixDet(N - 1, Asub);
}
return det;
}
// --------------------------------------------
// shell sort
// --------------------------------------------
static void shellsort(int size, int[] A)
{
int i, j, increment;
int temp;
increment = size / 2;
while (increment > 0)
{
for (i = increment; i < size; i++)
{
j = i;
temp = A[i];
while ((j >= increment) && (A[j - increment] > temp))
{
A[j] = A[j - increment];
j = j - increment;
}
A[j] = temp;
}
if (increment == 2)
increment = 1;
else
increment = (int) (increment / 2.2);
}
}
// --------------------------------------------
// benchmark test procedures
// --------------------------------------------
static int test_Int_Add()
{
int i = 1, j = 11, k = 112, l = 1111, m = 11111, n = -1, o = -11, p = -111, q = -1112, r = -11111;
int x;
int s = 0;
for (x = 0; x < 20000; ++x)
{
s += i;
s += j;
s += k;
s += l;
s += m;
s += n;
s += o;
s += p;
s += q;
s += r;
}
return s;
}
static int test_Int_Mult()
{
int x, y;
int s = 0;
for (y = 0; y < 2000; ++y)
{
s = 1;
for (x = 1; x <= 13; ++x)
{
s *= x;
}
for (x = 13; x > 0; --x)
{
s /= x;
}
}
return s;
}
static float test_float_math()
{
float s = (float) Math.PI;
int y;
for (y = 0; y < 5000; ++y)
{
s *= Math.sqrt(s);
s = (float) Math.sin(s);
s *= Math.cos(10.5 * s);
s = (float) Math.sqrt(s);
s = (float) Math.exp(s);
}
return s;
}
static int test_rand_MT()
{
int s = 0;
int y;
for (y = 0; y < 5000; ++y)
{
s = randM() % 10001;
}
return s;
}
static float test_matrix_math()
{
int x;
double[][] A = new double[2][2], B = new double[2][2], C = new double[2][2];
double[][] O = new double[3][3], T = new double[3][3];
int s;
for (x = 0; x < 250; ++x)
{
A[0][0] = 1;
A[0][1] = 3;
A[1][0] = 2;
A[1][1] = 4;
B[0][0] = 10;
B[0][1] = 30;
B[1][0] = 20;
B[1][1] = 40;
MatrixMatrixMult(2, 2, 2, A, B, C);
A[0][0] = 1;
A[0][1] = 3;
A[1][0] = 2;
A[1][1] = 4;
MatrixDet(2, A);
O[0][0] = 1;
O[0][1] = 4;
O[0][2] = 7;
O[1][0] = 2;
O[1][1] = 5;
O[1][2] = 8;
O[2][0] = 3;
O[2][1] = 6;
O[2][2] = 9;
MatrixDet(3, O);
}
s = (int) (O[0][0] * O[1][1] * O[2][2]);
return s;
}
static int test_Sort()
{
int y;
int[] t = new int[500];
for (y = 0; y < 50; ++y)
{
System.arraycopy(a, 0, t, 0, t.length);
shellsort(500, t);
System.arraycopy(b, 0, t, 0, t.length);
shellsort(500, t);
System.arraycopy(c, 0, t, 0, t.length);
shellsort(500, t);
}
return y;
}
static long test_TextOut()
{
int y;
for (y = 0; y < 20; ++y)
{
LCD.clear();
LCD.drawString("" + 0 + " " + 1000 + " int_Add", 0, 0);
LCD.drawString("" + 0 + " " + 1010 + " int_Mult", 0, 1);
LCD.drawString("" + 0 + " " + 1020 + " float_op", 0, 2);
LCD.drawString("" + 0 + " " + 1030 + " randomize", 0, 3);
LCD.drawString("" + 0 + " " + 1040 + " matrix_algb", 0, 4);
LCD.drawString("" + 0 + " " + 1050 + " arr_sort", 0, 5);
LCD.drawString("" + 0 + " " + 1060 + " display_text", 0, 6);
LCD.drawString("" + 0 + " " + 1070 + " testing...", 0, 7);
}
return 99;
}
static int test_graphics()
{
int x = 1, y;
Graphics g = new Graphics();;
for (y = 0; y < 100; ++y)
{
g.clear();
g.drawArc(50, 40, 10, 10, 0, 360);
g.fillArc(30, 24, 10, 10, 0, 360);
g.drawLine(10, 10, 60, 60);
g.drawLine(50, 20, 90, 70);
g.drawRect(20, 20, 40, 40);
g.fillRect(65, 25, 20, 30);
g.drawArc(100, 30, 15, 20, 0, 360);
}
return x;
}
static void displayValue(int testNo, long testTime, String testName)
{
LCD.drawInt(testNo, 2, 0, testNo);
LCD.drawInt((int) testTime, 6, 3, testNo);
LCD.drawString(testName, 10, testNo);
}
static void displayValues()
{
displayValue(0, runtime[0], "int_Add");
displayValue(1, runtime[1], "int_Mult");
displayValue(2, runtime[2], "float_op");
displayValue(3, runtime[3], "randomize");
displayValue(4, runtime[4], "matrix_algb");
displayValue(5, runtime[5], "arr_sort");
displayValue(6, runtime[6], "displ_txt");
displayValue(7, runtime[7], "graphics");
}
static public void main(String[] args)
{
long time0;
int x, y;
float s;
int i;
LCD.clear();
LCD.drawString("hw brickbench", 0, 1);
LCD.drawString("(C)H.Wunder 2013", 0, 2);
LCD.drawString("initializing...", 0, 4);
for (y = 0; y < 500; ++y)
{
a[y] = randM() % 30000;
b[y] = randM() % 30000;
c[y] = randM() % 30000;
}
LCD.clear();
time0 = System.currentTimeMillis();
s = test_Int_Add();
runtime[0] = System.currentTimeMillis() - time0;
displayValue(0, runtime[0], "int_Add ");
System.out.println("int_Add " + runtime[0]);
time0 = System.currentTimeMillis();
s = test_Int_Mult();
runtime[1] = System.currentTimeMillis() - time0;
displayValue(1, runtime[1], "int_Mult");
System.out.println("int_Mult " + runtime[1]);
time0 = System.currentTimeMillis();
s = test_float_math();
runtime[2] = System.currentTimeMillis() - time0;
displayValue(2, runtime[2], "float_op");
System.out.println("float_op " + runtime[2]);
time0 = System.currentTimeMillis();
s = test_rand_MT();
runtime[3] = System.currentTimeMillis() - time0;
displayValue(3, runtime[3], "randomize");
System.out.println("randomize " + runtime[3]);
time0 = System.currentTimeMillis();
s = test_matrix_math();
runtime[4] = System.currentTimeMillis() - time0;
displayValue(4, runtime[4], "matrix_algb");
System.out.println("matrx_algb " + runtime[4]);
time0 = System.currentTimeMillis();
s = test_Sort();
runtime[5] = System.currentTimeMillis() - time0;
displayValue(5, runtime[5], "arr_sort");
System.out.println("arr_sort " + runtime[5]);
time0 = System.currentTimeMillis();
s = test_TextOut();
runtime[6] = System.currentTimeMillis() - time0;
System.out.println("text " + runtime[6]);
LCD.clear();
displayValues();
time0 = System.currentTimeMillis();
s = test_graphics();
runtime[7] = System.currentTimeMillis() - time0;
System.out.println("graphics " + runtime[7]);
LCD.clear();
displayValues();
Button.waitForAnyPress();
LCD.clear();
y = 0;
for (x = 0; x < 8; ++x)
{
y += runtime[x];
}
LCD.drawString("Total ms: " + y, 0, 0);
LCD.drawString("Benchmark: " + (50000000 / y), 0, 1);
System.out.println("total ms: " + y);
System.out.println("benchmark: " + 50000000 / y);
Button.waitForAnyPress();
LCD.clear();
}
}
hw benchmark for leJOS/Java for EV3:
Code: Select all
import lejos.hardware.Button;
import lejos.hardware.ev3.LocalEV3;
import lejos.hardware.lcd.GraphicsLCD;
import lejos.hardware.lcd.LCD;
//hw brickbench
//benchmark test for NXT/EV3 and similar Micro Controllers
//PL: leJOS/EV3 0.8.0-alpha
//Autor: (C) Helmut Wunder 2013
// Ported to leJOS by Andy Shaw.
//freie Verwendung für private Zwecke
//für kommerzielle Zwecke nur nach Genehmigung durch den Autor.
//protected under the friendly Creative Commons Attribution-NonCommercial-ShareAlike 3.0 Unported License
//http://creativecommons.org/licenses/by-nc-sa/3.0/
//version 1.08
// PLEASE NOTE: This is not a good example of a Java program. It is a simple port of a C program. No
// attempt has been made to make use of Java features.
public class MBBenchmark
{
static long[] runtime = new long[8];
static int[] a = new int[500], b = new int[500], c = new int[500],
t = new int[500];
// --------------------------------------------
// Mersenne Twister
// --------------------------------------------
static int[] y = new int[25];
static int index = 25 + 1;
static final int M = 7;
static final int[] A = { 0, 0x8ebfd028 };
static int randM()
{
if (index >= 25)
{
int k;
if (index > 25)
{
int r = 9, s = 3402;
for (k = 0; k < 25; ++k)
{
r = 509845221 * r + 3;
s *= s + 1;
y[k] = s + (r >> 10);
}
}
for (k = 0; k < 25 - M; ++k)
y[k] = y[k + M] ^ (y[k] >> 1) ^ A[y[k] & 1];
for (; k < 25; ++k)
y[k] = y[k + (M - 25)] ^ (y[k] >> 1) ^ A[y[k] & 1];
index = 0;
}
int e = y[index++];
e ^= (e << 7) & 0x2b5b2500;
e ^= (e << 15) & 0xdb8b0000;
e ^= (e >> 16);
return e;
}
// --------------------------------------------
// Matrix Algebra
// --------------------------------------------
// matrix * matrix multiplication (matrix product)
static void MatrixMatrixMult(int N, int M, int K, double A[][],
double B[][], double C[][])
{
int i, j, s; // matrix A: N x M // B: M x K // C: N x K
for (i = 0; i < N; ++i)
{
for (j = 0; j < K; ++j)
{
C[i][j] = 0;
for (s = 0; s < M; ++s)
{
C[i][j] = C[i][j] + A[i][s] * B[s][j];
}
}
}
}
// matrix determinant
static double MatrixDet(int N, double A[][])
{
int i, j, i_count, j_count, count = 0;
double[][] Asub = new double[N - 1][N - 1];
double det = 0;
if (N == 1)
return A[0][0];
if (N == 2)
return (A[0][0] * A[1][1] - A[0][1] * A[1][0]);
for (count = 0; count < N; count++)
{
i_count = 0;
for (i = 1; i < N; i++)
{
j_count = 0;
for (j = 0; j < N; j++)
{
if (j == count)
continue;
Asub[i_count][j_count] = A[i][j];
j_count++;
}
i_count++;
}
det += Math.pow(-1, count) * A[0][count] * MatrixDet(N - 1, Asub);
}
return det;
}
// --------------------------------------------
// shell sort
// --------------------------------------------
static void shellsort(int size, int[] A)
{
int i, j, increment;
int temp;
increment = size / 2;
while (increment > 0)
{
for (i = increment; i < size; i++)
{
j = i;
temp = A[i];
while ((j >= increment) && (A[j - increment] > temp))
{
A[j] = A[j - increment];
j = j - increment;
}
A[j] = temp;
}
if (increment == 2)
increment = 1;
else
increment = (int) (increment / 2.2);
}
}
// --------------------------------------------
// benchmark test procedures
// --------------------------------------------
static int test_Int_Add()
{
int i = 1, j = 11, k = 112, l = 1111, m = 11111, n = -1, o = -11, p = -111, q = -1112, r = -11111;
int x;
int s = 0;
for (x = 0; x < 20000; ++x)
{
s += i;
s += j;
s += k;
s += l;
s += m;
s += n;
s += o;
s += p;
s += q;
s += r;
}
return s;
}
static int test_Int_Mult()
{
int x, y;
int s = 0;
for (y = 0; y < 2000; ++y)
{
s = 1;
for (x = 1; x <= 13; ++x)
{
s *= x;
}
for (x = 13; x > 0; --x)
{
s /= x;
}
}
return s;
}
static float test_float_math()
{
float s = (float) Math.PI;
int y;
for (y = 0; y < 5000; ++y)
{
s *= Math.sqrt(s);
s = (float) Math.sin(s);
s *= Math.cos(10.5 * s);
s = (float) Math.sqrt(s);
s = (float) Math.exp(s);
}
return s;
}
static int test_rand_MT()
{
int s = 0;
int y;
for (y = 0; y < 5000; ++y)
{
s = randM() % 10001;
}
return s;
}
static float test_matrix_math()
{
int x;
double[][] A = new double[2][2], B = new double[2][2], C = new double[2][2];
double[][] O = new double[3][3], T = new double[3][3];
int s;
for (x = 0; x < 250; ++x)
{
A[0][0] = 1;
A[0][1] = 3;
A[1][0] = 2;
A[1][1] = 4;
B[0][0] = 10;
B[0][1] = 30;
B[1][0] = 20;
B[1][1] = 40;
MatrixMatrixMult(2, 2, 2, A, B, C);
A[0][0] = 1;
A[0][1] = 3;
A[1][0] = 2;
A[1][1] = 4;
MatrixDet(2, A);
O[0][0] = 1;
O[0][1] = 4;
O[0][2] = 7;
O[1][0] = 2;
O[1][1] = 5;
O[1][2] = 8;
O[2][0] = 3;
O[2][1] = 6;
O[2][2] = 9;
MatrixDet(3, O);
}
s = (int) (O[0][0] * O[1][1] * O[2][2]);
return s;
}
static int test_Sort()
{
int y;
int[] t = new int[500];
for (y = 0; y < 50; ++y)
{
System.arraycopy(a, 0, t, 0, t.length);
shellsort(500, t);
System.arraycopy(b, 0, t, 0, t.length);
shellsort(500, t);
System.arraycopy(c, 0, t, 0, t.length);
shellsort(500, t);
}
return y;
}
static long test_TextOut()
{
int y;
for (y = 0; y < 20; ++y)
{
LCD.clear();
LCD.drawString("" + 0 + " " + 1000 + " int_Add", 0, 0);
LCD.drawString("" + 0 + " " + 1010 + " int_Mult", 0, 1);
LCD.drawString("" + 0 + " " + 1020 + " float_op", 0, 2);
LCD.drawString("" + 0 + " " + 1030 + " randomize", 0, 3);
LCD.drawString("" + 0 + " " + 1040 + " matrix_algb", 0, 4);
LCD.drawString("" + 0 + " " + 1050 + " arr_sort", 0, 5);
LCD.drawString("" + 0 + " " + 1060 + " display_text", 0, 6);
LCD.drawString("" + 0 + " " + 1070 + " testing...", 0, 7);
}
return 99;
}
static int test_graphics()
{
int x = 1, y;
GraphicsLCD g = LocalEV3.get().getGraphicsLCD();
for (y = 0; y < 100; ++y)
{
g.clear();
g.drawArc(50, 40, 10, 10, 0, 360);
g.fillArc(30, 24, 10, 10, 0, 360);
g.drawLine(10, 10, 60, 60);
g.drawLine(50, 20, 90, 70);
g.drawRect(20, 20, 40, 40);
g.fillRect(65, 25, 20, 30);
g.drawArc(100, 30, 15, 20, 0, 360);
}
return x;
}
static void displayValue(int testNo, long testTime, String testName)
{
LCD.drawInt(testNo, 2, 0, testNo);
LCD.drawInt((int) testTime, 6, 3, testNo);
LCD.drawString(testName, 10, testNo);
}
static void displayValues()
{
displayValue(0, runtime[0], "int_Add");
displayValue(1, runtime[1], "int_Mult");
displayValue(2, runtime[2], "float_op");
displayValue(3, runtime[3], "randomize");
displayValue(4, runtime[4], "matrix_algb");
displayValue(5, runtime[5], "arr_sort");
displayValue(6, runtime[6], "displ_txt");
displayValue(7, runtime[7], "graphics");
}
static public void main(String[] args)
{
long time0;
int x, y;
float s;
int i;
LCD.clear();
LCD.drawString("hw brickbench", 0, 1);
LCD.drawString("(C)H.Wunder 2013", 0, 2);
LCD.drawString("initializing...", 0, 4);
// Java on the EV3 uses a JIT compiler, run the test multiple times to
// allow the results to be stable
for (int ii = 0; ii < 20; ii++)
{
for (y = 0; y < 500; ++y)
{
a[y] = randM() % 30000;
b[y] = randM() % 30000;
c[y] = randM() % 30000;
}
LCD.clear();
time0 = System.currentTimeMillis();
s = test_Int_Add();
runtime[0] = System.currentTimeMillis() - time0;
displayValue(0, runtime[0], "int_Add ");
System.out.println("int_Add " + runtime[0]);
time0 = System.currentTimeMillis();
s = test_Int_Mult();
runtime[1] = System.currentTimeMillis() - time0;
displayValue(1, runtime[1], "int_Mult");
System.out.println("int_Mult " + runtime[1]);
time0 = System.currentTimeMillis();
s = test_float_math();
runtime[2] = System.currentTimeMillis() - time0;
displayValue(2, runtime[2], "float_op");
System.out.println("float_op " + runtime[2]);
time0 = System.currentTimeMillis();
s = test_rand_MT();
runtime[3] = System.currentTimeMillis() - time0;
displayValue(3, runtime[3], "randomize");
System.out.println("randomize " + runtime[3]);
time0 = System.currentTimeMillis();
s = test_matrix_math();
runtime[4] = System.currentTimeMillis() - time0;
displayValue(4, runtime[4], "matrix_algb");
System.out.println("matrx_algb " + runtime[4]);
time0 = System.currentTimeMillis();
s = test_Sort();
runtime[5] = System.currentTimeMillis() - time0;
displayValue(5, runtime[5], "arr_sort");
System.out.println("arr_sort " + runtime[5]);
time0 = System.currentTimeMillis();
s = test_TextOut();
runtime[6] = System.currentTimeMillis() - time0;
System.out.println("text " + runtime[6]);
LCD.clear();
displayValues();
time0 = System.currentTimeMillis();
s = test_graphics();
runtime[7] = System.currentTimeMillis() - time0;
System.out.println("graphics " + runtime[7]);
LCD.clear();
displayValues();
}
Button.waitForAnyPress();
LCD.clear();
y = 0;
for (x = 0; x < 8; ++x)
{
y += runtime[x];
}
LCD.drawString("Total ms: " + y, 0, 0);
LCD.drawString("Benchmark: " + (50000000 / y), 0, 1);
System.out.println("total ms: " + y);
System.out.println("benchmark: " + 50000000 / y);
Button.waitForAnyPress();
LCD.clear();
}
}
BTW: Is meanwhile everything finished for the EV3, too, e.g. display benchmarks (or will be still anything changed here), and has the float vs. double thing been fixed intermediately?
What about a java-fw-included sort algorithm?
ps
for the NXT code you wrote " leJOS/EV3 0.8.0-alpha" - but it's probably 0.9.anything, pls CMIIW...?
pps
another question re. the Java JIT compiler:
have the benchs been run once or several times before you took the time?
Last edited by HaWe on 13 May 2014, 14:14, edited 1 time in total.
-
- Posts: 323
- Joined: 29 Sep 2010, 05:03
Re: hw brickbench: Benchmark test for NXT and EV3
All of the results are as posted. Things may improve over time, but that is how they are at the moment.
The display results could be improved but it is not a priority for us at the moment. The double/float issue is just a difference between Java and the other systems. Java uses double for most floating point operations (in particular the sin/cos/sqrt/pow etc. functions return doubles), the current benchmark uses floats. This means that the results for the floating point tests are comparing Java using doubles against most of the other systems using floats. On most modern systems this is not a big deal, on the EV3 it may be. On the NXT it certainly is (you can see this in the results). But that is how things are if you use standard Java. I've not made any attempt to use Java features for things like array sorting etc.
Ah yes the version number for the NXT should be 0.9.1beta.
Is there a reason for the question marks against the NXT results?
The display results could be improved but it is not a priority for us at the moment. The double/float issue is just a difference between Java and the other systems. Java uses double for most floating point operations (in particular the sin/cos/sqrt/pow etc. functions return doubles), the current benchmark uses floats. This means that the results for the floating point tests are comparing Java using doubles against most of the other systems using floats. On most modern systems this is not a big deal, on the EV3 it may be. On the NXT it certainly is (you can see this in the results). But that is how things are if you use standard Java. I've not made any attempt to use Java features for things like array sorting etc.
Ah yes the version number for the NXT should be 0.9.1beta.
Is there a reason for the question marks against the NXT results?
Re: hw brickbench: Benchmark test for NXT and EV3
version thing fixed,
"?" removed (I simply missed it).
Thanks again!
Ergebnisse / Results (Version 1.17)
link to the most updated charts:
http://www.mindstormsforum.de/viewtopic ... 095#p64772
"?" removed (I simply missed it).
Thanks again!
Ergebnisse / Results (Version 1.17)
Code: Select all
HaWe brickbench © test (time in ms) NXT NXT NXT NXT EV3 EV3 EV3 EV3 Arduino Arduino
progr.language, API NXC RobotC leJOS Toppers C/C++ BCC/gpp C leJOS RobotC C#/Mono Mega2560 Due
Firmware / vers. JH's EFW 3.6 0.9.1 nxtOSEK BIOS LegoFW**) loop(?) 4.0.x loop#3(-#1) Sketch 1.5 Sketch 1.5
===================================================================================================================================================
0 integer add/subtr 3353 1561 727 29 3 4(-xxx)? 336 2(-15) 129 8
1 integer multiply/division 6441 1150 463 182 13 11(-xxx)? 589 26(-35) 1219 7
2 float operations 1716 413 2543 320 50 45(-xxx)? 226 155(-212) 887 83
3 Mersenne Tw. PRNG (&,^,>>,<<) 5372 1838 73 55 4 7(-xxx)? 1204 13(-19) 315 8
---------------------------------------------------------------------------------------------------------------------------------------------------
4 matrix algebra (prod/determ.) 3356 585 68 39 10 60(-xxx)? 325 43(-229) 243 23
5 int array sort [500] 332 95879 18756 491 113 139(-xxx)? 20932 172(-299) 1366 171
---------------------------------------------------------------------------------------------------------------------------------------------------
6 display text (N/A: penalty=1000) *) 781 32 26 17 28 218(-xxx)? +1000+? 268(-395) 92016 35347
7 display graphics (N/A: penalty=1000) *) 384 ? 95? 11480 182 51 371(-xxx)? +1000+? 185(-475)? 224137 138600
===================================================================================================================================================
execution time 21,735 101,553 35,117 1,362 272 855(-xxx)? <23600+? 864 316153 174247
===================================================================================================================================================
10 runt. code mem. (score= ln^3) 120k(110) 112k(105) ? 224k(158) <64M(1340) ? ? 42M(1210) 248k(167) 512k(240)
11 variable mem. (score= ln^3) 32k(42) 15k(20) ? 64k(72) <64M(1340) ? 15k(20) 42M(1210) 8k (8) 96k (95)
12 BT raw master+slave (score 10x) 4 (40) 2 (20) ? 2 (20) 0 ? ? 0 0 0
13 RS485/i2c raw master+slv.(score 10x) 32(320) 2 (20) ? 32(320) 0 ? ? 0 110(1100) 220(2200)
14 BT, +MC I/O rem.protocol (score 50x) 0 0 ? 0 0 ? ? 0 0 0
15 RS485, +MC I/O rem.prot. (score 50x) 0 0 ? 0 0 ? ? 0 0 0
16 USB MC Daisy-Chain (score 50x) 0 0 ? 0 0 ? ? 0 0 0
17 WiFi MC I/O rem.protocol (score 50x) 0 0 ? 0 0 ? ? 0 0 0
18 feat. double=64bit float (score 100) 0 0 100 0 100 100 0 100 0 100
19 feat.>=4D array/matrix op. (score 100) 100 0 ? 0 100 ? 0 100 0 0
20 Recursions (score 100) 0 0 ? 100 100 ? 0 100 100 100
21 Multitthreading (score 100) 100 100 100 100 (100) 100 100 100 (100) 100
22 API complete (sensor+motor+screen) + + + (+) --- + - (-) + +
23 small IDE made-to-measure (score 500) 500 500 0 0 500 0 500 0 500 500
===================================================================================================================================================
Brick platform NXT NXT NXT NXT EV3 EV3 EV3 EV3 Ardu.2560 Ardu.Due
progr.language, Firmware NXC RobotC leJOS nxtOSEK BCC/gpp C leJOS RobotC C#/Mono C/C++ C/C++
===================================================================================================================================================
speed bench score (50,000,000/exec.time) 2,300 0,485 1,423 36,711 183,823 58,479 >2,100? 57,870 0,158 0,287
environnment bench score 1,312 0,672 ?0,200 0,790 [---3,080] ?0,200 0,627? 2,820 2,119 2,487
---------------------------------------------------------------------------------------------------------------------------------------------------
brickbench system benchmark 3,612 1,157? 1,623 ? 37.501 [---186,883] 58,579? >2,700? 60,690 3,359 2,774
===================================================================================================================================================
http://www.mindstormsforum.de/viewtopic ... 095#p64772
Last edited by HaWe on 03 Jan 2015, 14:08, edited 2 times in total.
Re: hw brickbench: Benchmark test for NXT and EV3
still many Lejos and RobotC issues and data missing, but Monobrick has been finished yet by adding display classes (thanks to Vlad Ruzov!).
latest most updated benchmark overview table can be found here:
http://www.mindstormsforum.de/viewtopic ... 095#p64772
latest most updated benchmark overview table can be found here:
http://www.mindstormsforum.de/viewtopic ... 095#p64772
Last edited by HaWe on 04 Jan 2015, 13:27, edited 1 time in total.
Who is online
Users browsing this forum: No registered users and 1 guest