int CNAME(BLASLONG m, BLASLONG n, BLASLONG dummy1, FLOAT alpha, FLOAT *a, BLASLONG lda, FLOAT *x, BLASLONG inc_x, FLOAT *y, BLASLONG inc_y, FLOAT *buffer)
{
BLASLONG i;
BLASLONG j;
FLOAT *a_ptr;
FLOAT *x_ptr;
FLOAT *y_ptr;
FLOAT *ap[4];
BLASLONG n1;
BLASLONG m1;
BLASLONG m2;
BLASLONG n2;
FLOAT xbuffer[4],*ybuffer;
if ( m < 1 ) return(0);
if ( n < 1 ) return(0);
ybuffer = buffer;
n1 = n / 4 ;
n2 = n % 4 ;
m1 = m - ( m % 16 );
m2 = (m % NBMAX) - (m % 16) ;
y_ptr = y;
BLASLONG NB = NBMAX;
while ( NB == NBMAX )
{
m1 -= NB;
if ( m1 < 0)
{
if ( m2 == 0 ) break;
NB = m2;
}
a_ptr = a;
x_ptr = x;
zero_y(NB,ybuffer);
for( i = 0; i < n1 ; i++)
{
xbuffer[0] = alpha * x_ptr[0];
x_ptr += inc_x;
xbuffer[1] = alpha * x_ptr[0];
x_ptr += inc_x;
xbuffer[2] = alpha * x_ptr[0];
x_ptr += inc_x;
xbuffer[3] = alpha * x_ptr[0];
x_ptr += inc_x;
ap[0] = a_ptr;
ap[1] = a_ptr + lda;
ap[2] = ap[1] + lda;
ap[3] = ap[2] + lda;
sgemv_kernel_16x4(NB,ap,xbuffer,ybuffer);
a_ptr += 4 * lda;
}
for( i = 0; i < n2 ; i++)
{
xbuffer[0] = alpha * x_ptr[0];
x_ptr += inc_x;
sgemv_kernel_16x1(NB,a_ptr,xbuffer,ybuffer);
a_ptr += 1 * lda;
}
add_y(NB,ybuffer,y_ptr,inc_y);
a += NB;
y_ptr += NB * inc_y;
}
j=0;
while ( j < (m % 16))
{
a_ptr = a;
x_ptr = x;
FLOAT temp = 0.0;
for( i = 0; i < n; i++ )
{
temp += a_ptr[0] * x_ptr[0];
a_ptr += lda;
x_ptr += inc_x;
}
y_ptr[0] += alpha * temp;
y_ptr += inc_y;
a++;
j++;
}
return(0);
}
void process_gcode_command() {
uint32_t backup_f;
// convert relative to absolute
if (next_target.option_relative) {
next_target.target.X += startpoint.X;
next_target.target.Y += startpoint.Y;
next_target.target.Z += startpoint.Z;
#ifdef E_ABSOLUTE
next_target.target.E += startpoint.E;
#endif
}
// E ALWAYS relative, otherwise we overflow our registers after only a few layers
// next_target.target.E += startpoint.E;
// easier way to do this
// startpoint.E = 0;
// moved to dda.c, end of dda_create() and dda_queue.c, next_move()
// implement axis limits
#ifdef X_MIN
if (next_target.target.X < (X_MIN * STEPS_PER_MM_X))
next_target.target.X = X_MIN * STEPS_PER_MM_X;
#endif
#ifdef X_MAX
if (next_target.target.X > (X_MAX * STEPS_PER_MM_X))
next_target.target.X = X_MAX * STEPS_PER_MM_X;
#endif
#ifdef Y_MIN
if (next_target.target.Y < (Y_MIN * STEPS_PER_MM_Y))
next_target.target.Y = Y_MIN * STEPS_PER_MM_Y;
#endif
#ifdef Y_MAX
if (next_target.target.Y > (Y_MAX * STEPS_PER_MM_Y))
next_target.target.Y = Y_MAX * STEPS_PER_MM_Y;
#endif
#ifdef Z_MIN
if (next_target.target.Z < (Z_MIN * STEPS_PER_MM_Z))
next_target.target.Z = Z_MIN * STEPS_PER_MM_Z;
#endif
#ifdef Z_MAX
if (next_target.target.Z > (Z_MAX * STEPS_PER_MM_Z))
next_target.target.Z = Z_MAX * STEPS_PER_MM_Z;
#endif
// The GCode documentation was taken from http://reprap.org/wiki/Gcode .
if (next_target.seen_T) {
//? ==== T: Select Tool ====
//?
//? Example: T1
//?
//? Select extruder number 1 to build with. Extruder numbering starts at 0.
next_tool = next_target.T;
}
if (next_target.seen_G) {
uint8_t axisSelected = 0;
switch (next_target.G) {
// G0 - rapid, unsynchronised motion
// since it would be a major hassle to force the dda to not synchronise, just provide a fast feedrate and hope it's close enough to what host expects
case 0:
//? ==== G0: Rapid move ====
//?
//? Example: G0 X12
//?
//? In this case move rapidly to X = 12 mm. In fact, the RepRap firmware uses exactly the same code for rapid as it uses for controlled moves (see G1 below), as - for the RepRap machine - this is just as efficient as not doing so. (The distinction comes from some old machine tools that used to move faster if the axes were not driven in a straight line. For them G0 allowed any movement in space to get to the destination as fast as possible.)
backup_f = next_target.target.F;
next_target.target.F = MAXIMUM_FEEDRATE_X * 2L;
enqueue(&next_target.target);
next_target.target.F = backup_f;
break;
// G1 - synchronised motion
case 1:
//? ==== G1: Controlled move ====
//?
//? Example: G1 X90.6 Y13.8 E22.4
//?
//? Go in a straight line from the current (X, Y) point to the point (90.6, 13.8), extruding material as the move happens from the current extruded length to a length of 22.4 mm.
enqueue(&next_target.target);
break;
// G2 - Arc Clockwise
// unimplemented
// G3 - Arc Counter-clockwise
// unimplemented
// G4 - Dwell
case 4:
//? ==== G4: Dwell ====
//?
//? Example: G4 P200
//?
//? In this case sit still doing nothing for 200 milliseconds. During delays the state of the machine (for example the temperatures of its extruders) will still be preserved and controlled.
//?
// wait for all moves to complete
queue_wait();
// delay
for (;next_target.P > 0;next_target.P--) {
ifclock(clock_flag_10ms) {
clock_10ms();
}
delay_ms(1);
}
break;
// G20 - inches as units
case 20:
//? ==== G20: Set Units to Inches ====
//?
//? Example: G20
//?
//? Units from now on are in inches.
//?
next_target.option_inches = 1;
break;
// G21 - mm as units
case 21:
//? ==== G21: Set Units to Millimeters ====
//?
//? Example: G21
//?
//? Units from now on are in millimeters. (This is the RepRap default.)
//?
next_target.option_inches = 0;
break;
// G30 - go home via point
case 30:
//? ==== G30: Go home via point ====
//?
//? Undocumented.
enqueue(&next_target.target);
// no break here, G30 is move and then go home
// G28 - go home
case 28:
//? ==== G28: Move to Origin ====
//?
//? Example: G28
//?
//? This causes the RepRap machine to move back to its X, Y and Z zero endstops. It does so
//? accelerating, so as to get there fast. But when it arrives it backs off by 1 mm in each
//? direction slowly, then moves back slowly to the stop. This ensures more accurate positioning.
//?
//? If you add coordinates, then just the axes with coordinates specified will be zeroed. Thus
//?
//? G28 X0 Y72.3
//?
//? will zero the X and Y axes, but not Z. The actual coordinate values are ignored.
//?
queue_wait();
if (next_target.seen_X) {
zero_x();
axisSelected = 1;
}
if (next_target.seen_Y) {
zero_y();
axisSelected = 1;
}
if (next_target.seen_Z) {
zero_z();
axisSelected = 1;
}
// there's no point in moving E, as E has no endstops
if (!axisSelected) {
zero_x();
zero_y();
zero_z();
}
break;
// G90 - absolute positioning
case 90:
//? ==== G90: Set to Absolute Positioning ====
//?
//? Example: G90
//?
//? All coordinates from now on are absolute relative to the origin of the machine. (This is the RepRap default.)
next_target.option_relative = 0;
break;
// G91 - relative positioning
case 91:
//? ==== G91: Set to Relative Positioning ====
//?
//? Example: G91
//?
//? All coordinates from now on are relative to the last position.
next_target.option_relative = 1;
break;
// G92 - set home
case 92:
//? ==== G92: Set Position ====
//?
//? Example: G92 X10 E90
//?
//? Allows programming of absolute zero point, by reseting the current position to the values specified. This would set the machine's X coordinate to 10, and the extrude coordinate to 90. No physical motion will occur.
// wait for queue to empty
queue_wait();
if (next_target.seen_X) {
startpoint.X = current_position.X = next_target.target.X;
axisSelected = 1;
}
if (next_target.seen_Y) {
startpoint.Y = current_position.Y = next_target.target.Y;
axisSelected = 1;
}
if (next_target.seen_Z) {
startpoint.Z = current_position.Z = next_target.target.Z;
axisSelected = 1;
}
if (next_target.seen_E) {
#ifdef E_ABSOLUTE
startpoint.E = current_position.E = next_target.target.E;
#endif
axisSelected = 1;
}
if (axisSelected == 0) {
startpoint.X = current_position.X = next_target.target.X =
startpoint.Y = current_position.Y = next_target.target.Y =
startpoint.Z = current_position.Z = next_target.target.Z = 0;
}
break;
// G161 - Home negative
case 161:
//? ==== G161: Home negative ====
//?
//? Find the minimum limit of the specified axes by searching for the limit switch.
if (next_target.seen_X)
home_x_negative();
if (next_target.seen_Y)
home_y_negative();
if (next_target.seen_Z)
home_z_negative();
break;
// G162 - Home positive
case 162:
//? ==== G162: Home positive ====
//?
//? Find the maximum limit of the specified axes by searching for the limit switch.
if (next_target.seen_X)
home_x_positive();
if (next_target.seen_Y)
home_y_positive();
if (next_target.seen_Z)
home_z_positive();
break;
// unknown gcode: spit an error
default:
sersendf_P(PSTR("E: Bad G-code %d"), next_target.G);
// newline is sent from gcode_parse after we return
return;
}
#ifdef DEBUG
if (DEBUG_POSITION && (debug_flags & DEBUG_POSITION))
print_queue();
#endif
}
void process_gcode_command() {
uint32_t backup_f;
// convert relative to absolute
if (next_target.option_relative) {
next_target.target.X += startpoint.X;
next_target.target.Y += startpoint.Y;
next_target.target.Z += startpoint.Z;
}
// E ALWAYS relative, otherwise we overflow our registers after only a few layers
// next_target.target.E += startpoint.E;
// easier way to do this
// startpoint.E = 0;
// moved to dda.c, end of dda_create() and dda_queue.c, next_move()
// implement axis limits
#ifdef X_MIN
if (next_target.target.X < (X_MIN * STEPS_PER_MM_X))
next_target.target.X = X_MIN * STEPS_PER_MM_X;
#endif
#ifdef X_MAX
if (next_target.target.X > (X_MAX * STEPS_PER_MM_X))
next_target.target.X = X_MAX * STEPS_PER_MM_X;
#endif
#ifdef Y_MIN
if (next_target.target.Y < (Y_MIN * STEPS_PER_MM_Y))
next_target.target.Y = Y_MIN * STEPS_PER_MM_Y;
#endif
#ifdef Y_MAX
if (next_target.target.Y > (Y_MAX * STEPS_PER_MM_Y))
next_target.target.Y = Y_MAX * STEPS_PER_MM_Y;
#endif
#ifdef Z_MIN
if (next_target.target.Z < (Z_MIN * STEPS_PER_MM_Z))
next_target.target.Z = Z_MIN * STEPS_PER_MM_Z;
#endif
#ifdef Z_MAX
if (next_target.target.Z > (Z_MAX * STEPS_PER_MM_Z))
next_target.target.Z = Z_MAX * STEPS_PER_MM_Z;
#endif
if (next_target.seen_T) {
next_tool = next_target.T;
}
if (next_target.seen_G) {
uint8_t axisSelected = 0;
switch (next_target.G) {
// G0 - rapid, unsynchronised motion
// since it would be a major hassle to force the dda to not synchronise, just provide a fast feedrate and hope it's close enough to what host expects
case 0:
backup_f = next_target.target.F;
next_target.target.F = MAXIMUM_FEEDRATE_X * 2L;
enqueue(&next_target.target);
next_target.target.F = backup_f;
break;
// G1 - synchronised motion
case 1:
enqueue(&next_target.target);
break;
// G2 - Arc Clockwise
// unimplemented
// G3 - Arc Counter-clockwise
// unimplemented
// G4 - Dwell
case 4:
// wait for all moves to complete
queue_wait();
// delay
for (;next_target.P > 0;next_target.P--) {
ifclock(CLOCK_FLAG_10MS) {
clock_10ms();
}
delay_ms(1);
}
break;
// G20 - inches as units
case 20:
next_target.option_inches = 1;
break;
// G21 - mm as units
case 21:
next_target.option_inches = 0;
break;
// G30 - go home via point
case 30:
enqueue(&next_target.target);
// no break here, G30 is move and then go home
// G28 - go home
case 28:
queue_wait();
if (next_target.seen_X) {
zero_x();
axisSelected = 1;
}
if (next_target.seen_Y) {
zero_y();
axisSelected = 1;
}
if (next_target.seen_Z) {
zero_z();
axisSelected = 1;
}
if (next_target.seen_E) {
zero_e();
axisSelected = 1;
}
if (!axisSelected) {
zero_x();
zero_y();
zero_z();
zero_e();
}
break;
// G90 - absolute positioning
case 90:
next_target.option_relative = 0;
break;
// G91 - relative positioning
case 91:
next_target.option_relative = 1;
break;
// G92 - set home
case 92:
// wait for queue to empty
queue_wait();
if (next_target.seen_X) {
startpoint.X = current_position.X = next_target.target.X;
axisSelected = 1;
}
if (next_target.seen_Y) {
startpoint.Y = current_position.Y = next_target.target.Y;
axisSelected = 1;
}
if (next_target.seen_Z) {
startpoint.Z = current_position.Z = next_target.target.Z;
axisSelected = 1;
}
if (next_target.seen_E) {
startpoint.E = current_position.E = next_target.target.E;
axisSelected = 1;
}
if (axisSelected == 0) {
startpoint.X = current_position.X =
startpoint.Y = current_position.Y =
startpoint.Z = current_position.Z =
startpoint.E = current_position.E = 0;
}
break;
// G161 - Home negative
case 161:
if (next_target.seen_X)
home_x_negative();
if (next_target.seen_Y)
home_y_negative();
if (next_target.seen_Z)
home_z_negative();
break;
// G162 - Home positive
case 162:
if (next_target.seen_X)
home_x_positive();
if (next_target.seen_Y)
home_y_positive();
if (next_target.seen_Z)
home_z_positive();
break;
// unknown gcode: spit an error
default:
sersendf_P(PSTR("E: Bad G-code %d"), next_target.G);
// newline is sent from gcode_parse after we return
return;
}
#ifdef DEBUG
print_queue();
#endif
}
int CNAME(BLASLONG m, BLASLONG n, BLASLONG dummy1, FLOAT alpha_r,FLOAT alpha_i, FLOAT *a, BLASLONG lda, FLOAT *x, BLASLONG inc_x, FLOAT *y, BLASLONG inc_y, FLOAT *buffer)
{
BLASLONG i;
BLASLONG j;
FLOAT *a_ptr;
FLOAT *x_ptr;
FLOAT *y_ptr;
FLOAT *ap[4];
BLASLONG n1;
BLASLONG m1;
BLASLONG m2;
BLASLONG n2;
FLOAT xbuffer[8],*ybuffer;
#if 0
printf("%s %d %d %.16f %.16f %d %d %d\n","zgemv_n",m,n,alpha_r,alpha_i,lda,inc_x,inc_y);
#endif
if ( m < 1 ) return(0);
if ( n < 1 ) return(0);
ybuffer = buffer;
inc_x *= 2;
inc_y *= 2;
lda *= 2;
n1 = n / 4 ;
n2 = n % 4 ;
m1 = m - ( m % 16 );
m2 = (m % NBMAX) - (m % 16) ;
y_ptr = y;
BLASLONG NB = NBMAX;
while ( NB == NBMAX )
{
m1 -= NB;
if ( m1 < 0)
{
if ( m2 == 0 ) break;
NB = m2;
}
a_ptr = a;
x_ptr = x;
zero_y(NB,ybuffer);
for( i = 0; i < n1 ; i++)
{
xbuffer[0] = x_ptr[0];
xbuffer[1] = x_ptr[1];
x_ptr += inc_x;
xbuffer[2] = x_ptr[0];
xbuffer[3] = x_ptr[1];
x_ptr += inc_x;
xbuffer[4] = x_ptr[0];
xbuffer[5] = x_ptr[1];
x_ptr += inc_x;
xbuffer[6] = x_ptr[0];
xbuffer[7] = x_ptr[1];
x_ptr += inc_x;
ap[0] = a_ptr;
ap[1] = a_ptr + lda;
ap[2] = ap[1] + lda;
ap[3] = ap[2] + lda;
zgemv_kernel_16x4(NB,ap,xbuffer,ybuffer);
a_ptr += 4 * lda;
}
for( i = 0; i < n2 ; i++)
{
xbuffer[0] = x_ptr[0];
xbuffer[1] = x_ptr[1];
x_ptr += inc_x;
zgemv_kernel_16x1(NB,a_ptr,xbuffer,ybuffer);
a_ptr += 1 * lda;
}
add_y(NB,ybuffer,y_ptr,inc_y,alpha_r,alpha_i);
a += 2 * NB;
y_ptr += NB * inc_y;
}
j=0;
while ( j < (m % 16))
{
a_ptr = a;
x_ptr = x;
FLOAT temp_r = 0.0;
FLOAT temp_i = 0.0;
for( i = 0; i < n; i++ )
{
#if ( !defined(CONJ) && !defined(XCONJ) ) || ( defined(CONJ) && defined(XCONJ) )
temp_r += a_ptr[0] * x_ptr[0] - a_ptr[1] * x_ptr[1];
temp_i += a_ptr[0] * x_ptr[1] + a_ptr[1] * x_ptr[0];
#else
temp_r += a_ptr[0] * x_ptr[0] + a_ptr[1] * x_ptr[1];
temp_i += a_ptr[0] * x_ptr[1] - a_ptr[1] * x_ptr[0];
#endif
a_ptr += lda;
x_ptr += inc_x;
}
#if !defined(XCONJ)
y_ptr[0] += alpha_r * temp_r - alpha_i * temp_i;
y_ptr[1] += alpha_r * temp_i + alpha_i * temp_r;
#else
y_ptr[0] += alpha_r * temp_r + alpha_i * temp_i;
y_ptr[1] -= alpha_r * temp_i - alpha_i * temp_r;
#endif
y_ptr += inc_y;
a+=2;
j++;
}
return(0);
}
