tkfx

 0x00007fd71508b1ec: mov    0x1b0(%r15),%r10   ;*invokestatic currentThread
                                                ; - java.lang.ThreadLocal::get@0 (line 143)
                                                ; - Opt::handleRequest@3 (line 70)
 0x00007fd71508b1f3: mov    0x50(%r10),%r11d   ;*getfield threadLocals
                                                ; - java.lang.ThreadLocal::getMap@1 (line 213)
                                                ; - java.lang.ThreadLocal::get@6 (line 144)
                                                ; - Opt::handleRequest@3 (line 70)

tkfx

private static final ThreadLocal<Integer> counter = new ThreadLocal<Integer>();

private static void handleRequest() {
	if (counter.get() == null) {
		counter.set(0);
	}

	counter.set(counter.get() + 1);

tkfx

private static void calcLine(int a, int b, int from, int to) {
	Line l = new Line(a, b);
	for (int x = from; x <= to; x++) {
		int y = (l.a * x + l.b);
		System.err.println("(" + x + ", " + y + ")");
	}
}

tkfx

private static void calcLine(int a, int b, int from, int to) {
	Line l = new Line(a, b);
	for (int x = from; x <= to; x++) {
		int y = l.getY(x);
		System.err.println("(" + x + ", " + y + ")");
	}
}

static class Line {
	public final int a;

tkfx

private static double[] loopUnrolling(double[][] matrix1, double[] vector1) {
	double[] result = new double[vector1.length];

	for (int i = 0; i < matrix1.length; i++) {
	for (int j = 0; j < vector1.length; j++) {
			result[i] += matrix1[i][j] * vector1[j];
		}
	}

	return result;

tkfx

private static double[] loopUnrolling2(double[][] matrix1, double[] vector1) {
	double[] result = new double[vector1.length];

	for (int i = 0; i < matrix1.length; i++) {
		result[i] += matrix1[i][0] * vector1[0];
		result[i] += matrix1[i][1] * vector1[1];
		result[i] += matrix1[i][2] * vector1[2];
		// and maybe it will expand even further - e.g. 4 iterations, thus
		// adding code to fix the indexing
		// which we would waste more time doing correctly and efficiently

tkfx

....
0x00007fd715060743: vmovsd 0x10(%r8,%rcx,8),%xmm0  ;*daload
                                                   ; - Opt::loopUnrolling@26 (line 179)
0x00007fd71506074a: vmovsd 0x10(%rbp),%xmm1        ;*daload
                                                   ; - Opt::loopUnrolling@36 (line 179)
0x00007fd71506074f: vmulsd 0x10(%r12,%r9,8),%xmm1,%xmm1
0x00007fd715060756: vaddsd %xmm0,%xmm1,%xmm0       ;*dadd
                                                   ; - Opt::loopUnrolling@38 (line 179)
0x00007fd71506075a: vmovsd %xmm0,0x10(%r8,%rcx,8)  ;*dastore
                                                   ; - Opt::loopUnrolling@39 (line 179)

tkfx

0x00007fd715062d0c: cmp    %edx,%esi
0x00007fd715062d0e: jge    0x00007fd715062d24  ;*if_icmplt
                                               ; - Opt::isOpt@4 (line 117)

tkfx

private static int isOpt(int x, int y) {
	int veryHardCalculation = 0;

	if (x < y) {
		// this would not require a jump
		veryHardCalculation = y * 1000 + x;
		return veryHardCalculation;
	} else {
		veryHardCalculation = x * 1000 + y;
		return veryHardCalculation;

tkfx

private static int isOpt(int x, int y) {
	int veryHardCalculation = 0;

	if (x >= y) {
		veryHardCalculation = x * 1000 + y;
	} else {
		veryHardCalculation = y * 1000 + x;
	}
	return veryHardCalculation;
}