int Test_3x3getRot(void)
{
// Init an array flanked by guard pages
btMatrix3x3 in1[ARRAY_SIZE];
btQuaternion out[ARRAY_SIZE];
btQuaternion out2[ARRAY_SIZE];
// Init the data
size_t i, j;
for (i = 0; i < ARRAY_SIZE; i++)
{
in1[i] = btMatrix3x3(rand_f4(), rand_f4(), rand_f4());
out[i] = btQuaternion(qtNAN_f4());
out2[i] = btQuaternion(qtNAN_f4());
M3x3getRot_ref(in1[i], out[i]);
in1[i].getRotation(out2[i]);
if (out[i] != out2[i])
{
vlog("Error - M3x3getRot result error! ");
vlog("failure @ %ld\n", i);
vlog(
"\ncorrect = (%10.7f, %10.7f, %10.7f, %10.7f) "
"\ntested = (%10.7f, %10.7f, %10.7f, %10.7f) \n",
out[i].x(), out[i].y(), out[i].z(), out[i].w(),
out2[i].x(), out2[i].y(), out2[i].z(), out2[i].w());
return -1;
}
}
uint64_t scalarTime, vectorTime;
uint64_t startTime, bestTime, currentTime;
bestTime = ~(bestTime & 0); //-1ULL;
scalarTime = 0;
for (j = 0; j < LOOPCOUNT; j++)
{
startTime = ReadTicks();
for (i = 0; i < ARRAY_SIZE; i++)
M3x3getRot_ref(in1[i], out[i]);
currentTime = ReadTicks() - startTime;
scalarTime += currentTime;
if (currentTime < bestTime)
bestTime = currentTime;
}
if (0 == gReportAverageTimes)
scalarTime = bestTime;
else
scalarTime /= LOOPCOUNT;
bestTime = ~(bestTime & 0); //-1ULL;
vectorTime = 0;
for (j = 0; j < LOOPCOUNT; j++)
{
startTime = ReadTicks();
for (i = 0; i < ARRAY_SIZE; i++)
{
in1[i].getRotation(out2[i]);
}
currentTime = ReadTicks() - startTime;
vectorTime += currentTime;
if (currentTime < bestTime)
bestTime = currentTime;
}
if (0 == gReportAverageTimes)
vectorTime = bestTime;
else
vectorTime /= LOOPCOUNT;
vlog("Timing:\n");
vlog("\t scalar\t vector\n");
vlog("\t%10.2f\t%10.2f\n", TicksToCycles(scalarTime) / ARRAY_SIZE, TicksToCycles(vectorTime) / ARRAY_SIZE);
return 0;
}
int Test_3x3mulM1M2(void)
{
// Init an array flanked by guard pages
btMatrix3x3 in1[ARRAY_SIZE];
btMatrix3x3 in2[ARRAY_SIZE];
btMatrix3x3 out[ARRAY_SIZE];
btMatrix3x3 out2[ARRAY_SIZE];
// Init the data
size_t i, j;
for( i = 0; i < ARRAY_SIZE; i++ )
{
in1[i] = btMatrix3x3(rand_f4(), rand_f4(), rand_f4() );
in2[i] = btMatrix3x3(rand_f4(), rand_f4(), rand_f4() );
out[i] = M3x3mulM1M2_ref(in1[i], in2[i]);
out2[i] = (in1[i] * in2[i]);
if( out[i] != out2[i] )
{
vlog( "Error - M3x3mulM1M2 result error! ");
vlog( "failure @ %ld\n", i);
btVector3 m0, m1, m2;
m0 = out[i].getRow(0);
m1 = out[i].getRow(1);
m2 = out[i].getRow(2);
vlog( "\ncorrect = (%10.4f, %10.4f, %10.4f, %10.4f) "
"\n (%10.4f, %10.4f, %10.4f, %10.4f) "
"\n (%10.4f, %10.4f, %10.4f, %10.4f) \n",
m0.m_floats[0], m0.m_floats[1], m0.m_floats[2], m0.m_floats[3],
m1.m_floats[0], m1.m_floats[1], m1.m_floats[2], m1.m_floats[3],
m2.m_floats[0], m2.m_floats[1], m2.m_floats[2], m2.m_floats[3]);
m0 = out2[i].getRow(0);
m1 = out2[i].getRow(1);
m2 = out2[i].getRow(2);
vlog( "\ntested = (%10.4f, %10.4f, %10.4f, %10.4f) "
"\n (%10.4f, %10.4f, %10.4f, %10.4f) "
"\n (%10.4f, %10.4f, %10.4f, %10.4f) \n",
m0.m_floats[0], m0.m_floats[1], m0.m_floats[2], m0.m_floats[3],
m1.m_floats[0], m1.m_floats[1], m1.m_floats[2], m1.m_floats[3],
m2.m_floats[0], m2.m_floats[1], m2.m_floats[2], m2.m_floats[3]);
return -1;
}
}
uint64_t scalarTime, vectorTime;
uint64_t startTime, bestTime, currentTime;
bestTime = -1LL;
scalarTime = 0;
for (j = 0; j < LOOPCOUNT; j++)
{
startTime = ReadTicks();
for( i = 0; i < ARRAY_SIZE; i++ )
out[i] = M3x3mulM1M2_ref(in1[i], in2[i]);
currentTime = ReadTicks() - startTime;
scalarTime += currentTime;
if( currentTime < bestTime )
bestTime = currentTime;
}
if( 0 == gReportAverageTimes )
scalarTime = bestTime;
else
scalarTime /= LOOPCOUNT;
bestTime = -1LL;
vectorTime = 0;
for (j = 0; j < LOOPCOUNT; j++)
{
startTime = ReadTicks();
for( i = 0; i < ARRAY_SIZE; i++ )
out2[i] = (in1[i] * in2[i]);
currentTime = ReadTicks() - startTime;
vectorTime += currentTime;
if( currentTime < bestTime )
bestTime = currentTime;
}
if( 0 == gReportAverageTimes )
vectorTime = bestTime;
else
vectorTime /= LOOPCOUNT;
vlog( "Timing:\n" );
vlog( "\t scalar\t vector\n" );
vlog( "\t%10.2f\t%10.2f\n", TicksToCycles( scalarTime ) / ARRAY_SIZE, TicksToCycles( vectorTime ) / ARRAY_SIZE );
return 0;
}
int Test_3x3timesTranspose(void)
{
// Init an array flanked by guard pages
btMatrix3x3 in1[ARRAY_SIZE];
btMatrix3x3 in2[ARRAY_SIZE];
btMatrix3x3 out[ARRAY_SIZE];
btMatrix3x3 out2[ARRAY_SIZE];
// Init the data
size_t i, j;
for( i = 0; i < ARRAY_SIZE; i++ )
{
in1[i] = btMatrix3x3(rand_f4(), rand_f4(), rand_f4() );
in2[i] = btMatrix3x3(rand_f4(), rand_f4(), rand_f4() );
out[i] = timesTranspose(in1[i], in2[i]);
out2[i] = in1[i].timesTranspose(in2[i]);
if( out[i] != out2[i] )
{
printf( "failure @ %ld\n", i);
return -1;
}
}
uint64_t scalarTime, vectorTime;
uint64_t startTime, bestTime, currentTime;
bestTime = -1LL;
scalarTime = 0;
for (j = 0; j < LOOPCOUNT; j++) {
startTime = ReadTicks();
for( i = 0; i < ARRAY_SIZE; i++ )
out[i] = timesTranspose(in1[i], in2[i]);
currentTime = ReadTicks() - startTime;
scalarTime += currentTime;
if( currentTime < bestTime )
bestTime = currentTime;
}
if( 0 == gReportAverageTimes )
scalarTime = bestTime;
else
scalarTime /= LOOPCOUNT;
bestTime = -1LL;
vectorTime = 0;
for (j = 0; j < LOOPCOUNT; j++) {
startTime = ReadTicks();
for( i = 0; i < ARRAY_SIZE; i++ )
out[i] = in1[i].timesTranspose(in2[i]);
currentTime = ReadTicks() - startTime;
vectorTime += currentTime;
if( currentTime < bestTime )
bestTime = currentTime;
}
if( 0 == gReportAverageTimes )
vectorTime = bestTime;
else
vectorTime /= LOOPCOUNT;
vlog( "Timing:\n" );
vlog( "\t scalar\t vector\n" );
vlog( "\t%10.2f\t%10.2f\n", TicksToCycles( scalarTime ) / ARRAY_SIZE, TicksToCycles( vectorTime ) / ARRAY_SIZE );
return 0;
}
