Convert maths macros to inlines

For parity with #10728

Co-Authored-By: ejtagle <ejtagle@hotmail.com>
This commit is contained in:
Scott Lahteine 2018-05-13 10:36:37 -05:00
parent 431a81fc66
commit 7c6dd087c9
2 changed files with 85 additions and 85 deletions

View File

@ -243,59 +243,63 @@ volatile int32_t Stepper::endstops_trigsteps[XYZ];
// intRes = longIn1 * longIn2 >> 24 // intRes = longIn1 * longIn2 >> 24
// uses: // uses:
// r26 to store 0 // A[tmp] to store 0
// r27 to store bits 16-23 of the 48bit result. The top bit is used to round the two byte result. // B[tmp] to store bits 16-23 of the 48bit result. The top bit is used to round the two byte result.
// note that the lower two bytes and the upper byte of the 48bit result are not calculated. // note that the lower two bytes and the upper byte of the 48bit result are not calculated.
// this can cause the result to be out by one as the lower bytes may cause carries into the upper ones. // this can cause the result to be out by one as the lower bytes may cause carries into the upper ones.
// B0 A0 are bits 24-39 and are the returned value // B A are bits 24-39 and are the returned value
// C1 B1 A1 is longIn1 // C B A is longIn1
// D2 C2 B2 A2 is longIn2 // D C B A is longIn2
// //
#define MultiU24X32toH16(intRes, longIn1, longIn2) \ static FORCE_INLINE uint16_t MultiU24X32toH16(uint32_t longIn1, uint32_t longIn2) {
asm volatile ( \ register uint8_t tmp1;
A("clr r26") \ register uint8_t tmp2;
A("mul %A1, %B2") \ register uint16_t intRes;
A("mov r27, r1") \ __asm__ __volatile__(
A("mul %B1, %C2") \ A("clr %[tmp1]")
A("movw %A0, r0") \ A("mul %A[longIn1], %B[longIn2]")
A("mul %C1, %C2") \ A("mov %[tmp2], r1")
A("add %B0, r0") \ A("mul %B[longIn1], %C[longIn2]")
A("mul %C1, %B2") \ A("movw %A[intRes], r0")
A("add %A0, r0") \ A("mul %C[longIn1], %C[longIn2]")
A("adc %B0, r1") \ A("add %B[intRes], r0")
A("mul %A1, %C2") \ A("mul %C[longIn1], %B[longIn2]")
A("add r27, r0") \ A("add %A[intRes], r0")
A("adc %A0, r1") \ A("adc %B[intRes], r1")
A("adc %B0, r26") \ A("mul %A[longIn1], %C[longIn2]")
A("mul %B1, %B2") \ A("add %[tmp2], r0")
A("add r27, r0") \ A("adc %A[intRes], r1")
A("adc %A0, r1") \ A("adc %B[intRes], %[tmp1]")
A("adc %B0, r26") \ A("mul %B[longIn1], %B[longIn2]")
A("mul %C1, %A2") \ A("add %[tmp2], r0")
A("add r27, r0") \ A("adc %A[intRes], r1")
A("adc %A0, r1") \ A("adc %B[intRes], %[tmp1]")
A("adc %B0, r26") \ A("mul %C[longIn1], %A[longIn2]")
A("mul %B1, %A2") \ A("add %[tmp2], r0")
A("add r27, r1") \ A("adc %A[intRes], r1")
A("adc %A0, r26") \ A("adc %B[intRes], %[tmp1]")
A("adc %B0, r26") \ A("mul %B[longIn1], %A[longIn2]")
A("lsr r27") \ A("add %[tmp2], r1")
A("adc %A0, r26") \ A("adc %A[intRes], %[tmp1]")
A("adc %B0, r26") \ A("adc %B[intRes], %[tmp1]")
A("mul %D2, %A1") \ A("lsr %[tmp2]")
A("add %A0, r0") \ A("adc %A[intRes], %[tmp1]")
A("adc %B0, r1") \ A("adc %B[intRes], %[tmp1]")
A("mul %D2, %B1") \ A("mul %D[longIn2], %A[longIn1]")
A("add %B0, r0") \ A("add %A[intRes], r0")
A("clr r1") \ A("adc %B[intRes], r1")
: \ A("mul %D[longIn2], %B[longIn1]")
"=&r" (intRes) \ A("add %B[intRes], r0")
: \ A("clr r1")
"d" (longIn1), \ : [intRes] "=&r" (intRes),
"d" (longIn2) \ [tmp1] "=&r" (tmp1),
: \ [tmp2] "=&r" (tmp2)
"r26" , "r27" \ : [longIn1] "d" (longIn1),
) [longIn2] "d" (longIn2)
: "cc"
);
return intRes;
}
// Some useful constants // Some useful constants
@ -1506,10 +1510,7 @@ void Stepper::isr() {
? _eval_bezier_curve(acceleration_time) ? _eval_bezier_curve(acceleration_time)
: current_block->cruise_rate; : current_block->cruise_rate;
#else #else
MultiU24X32toH16(acc_step_rate, acceleration_time, current_block->acceleration_rate); acc_step_rate = MultiU24X32toH16(acceleration_time, current_block->acceleration_rate) + current_block->initial_rate;
acc_step_rate += current_block->initial_rate;
// upper limit
NOMORE(acc_step_rate, current_block->nominal_rate); NOMORE(acc_step_rate, current_block->nominal_rate);
#endif #endif
@ -1540,7 +1541,6 @@ void Stepper::isr() {
#if ENABLED(BEZIER_JERK_CONTROL) #if ENABLED(BEZIER_JERK_CONTROL)
// If this is the 1st time we process the 2nd half of the trapezoid... // If this is the 1st time we process the 2nd half of the trapezoid...
if (!bezier_2nd_half) { if (!bezier_2nd_half) {
// Initialize the Bézier speed curve // Initialize the Bézier speed curve
_calc_bezier_curve_coeffs(current_block->cruise_rate, current_block->final_rate, current_block->deceleration_time_inverse); _calc_bezier_curve_coeffs(current_block->cruise_rate, current_block->final_rate, current_block->deceleration_time_inverse);
bezier_2nd_half = true; bezier_2nd_half = true;
@ -1553,14 +1553,14 @@ void Stepper::isr() {
#else #else
// Using the old trapezoidal control // Using the old trapezoidal control
MultiU24X32toH16(step_rate, deceleration_time, current_block->acceleration_rate); step_rate = MultiU24X32toH16(deceleration_time, current_block->acceleration_rate);
if (step_rate < acc_step_rate) { // Still decelerating? if (step_rate < acc_step_rate) { // Still decelerating?
step_rate = acc_step_rate - step_rate; step_rate = acc_step_rate - step_rate;
NOLESS(step_rate, current_block->final_rate); NOLESS(step_rate, current_block->final_rate);
} }
else else
step_rate = current_block->final_rate; step_rate = current_block->final_rate;
#endif #endif
// step_rate to timer interval // step_rate to timer interval

View File

@ -61,26 +61,28 @@ extern Stepper stepper;
// uses: // uses:
// r26 to store 0 // r26 to store 0
// r27 to store the byte 1 of the 24 bit result // r27 to store the byte 1 of the 24 bit result
#define MultiU16X8toH16(intRes, charIn1, intIn2) \ static FORCE_INLINE uint16_t MultiU16X8toH16(uint8_t charIn1, uint16_t intIn2) {
asm volatile ( \ register uint8_t tmp;
A("clr r26") \ register uint16_t intRes;
A("mul %A1, %B2") \ __asm__ __volatile__ (
A("movw %A0, r0") \ A("clr %[tmp]")
A("mul %A1, %A2") \ A("mul %[charIn1], %B[intIn2]")
A("add %A0, r1") \ A("movw %A[intRes], r0")
A("adc %B0, r26") \ A("mul %[charIn1], %A[intIn2]")
A("lsr r0") \ A("add %A[intRes], r1")
A("adc %A0, r26") \ A("adc %B[intRes], %[tmp]")
A("adc %B0, r26") \ A("lsr r0")
A("clr r1") \ A("adc %A[intRes], %[tmp]")
: \ A("adc %B[intRes], %[tmp]")
"=&r" (intRes) \ A("clr r1")
: \ : [intRes] "=&r" (intRes),
"d" (charIn1), \ [tmp] "=&r" (tmp)
"d" (intIn2) \ : [charIn1] "d" (charIn1),
: \ [intIn2] "d" (intIn2)
"r26" \ : "cc"
) );
return intRes;
}
class Stepper { class Stepper {
@ -346,17 +348,15 @@ class Stepper {
NOLESS(step_rate, F_CPU / 500000); NOLESS(step_rate, F_CPU / 500000);
step_rate -= F_CPU / 500000; // Correct for minimal speed step_rate -= F_CPU / 500000; // Correct for minimal speed
if (step_rate >= (8 * 256)) { // higher step rate if (step_rate >= (8 * 256)) { // higher step rate
unsigned short table_address = (unsigned short)&speed_lookuptable_fast[(unsigned char)(step_rate >> 8)][0]; uint16_t table_address = (uint16_t)&speed_lookuptable_fast[(uint8_t)(step_rate >> 8)][0],
unsigned char tmp_step_rate = (step_rate & 0x00FF); gain = (uint16_t)pgm_read_word_near(table_address + 2);
unsigned short gain = (unsigned short)pgm_read_word_near(table_address + 2); timer = (uint16_t)pgm_read_word_near(table_address) - MultiU16X8toH16(step_rate & 0x00FF, gain);
MultiU16X8toH16(timer, tmp_step_rate, gain);
timer = (unsigned short)pgm_read_word_near(table_address) - timer;
} }
else { // lower step rates else { // lower step rates
unsigned short table_address = (unsigned short)&speed_lookuptable_slow[0][0]; uint16_t table_address = (uint16_t)&speed_lookuptable_slow[0][0];
table_address += ((step_rate) >> 1) & 0xFFFC; table_address += ((step_rate) >> 1) & 0xFFFC;
timer = (unsigned short)pgm_read_word_near(table_address); timer = (uint16_t)pgm_read_word_near(table_address)
timer -= (((unsigned short)pgm_read_word_near(table_address + 2) * (unsigned char)(step_rate & 0x0007)) >> 3); - (((uint16_t)pgm_read_word_near(table_address + 2) * (uint8_t)(step_rate & 0x0007)) >> 3);
} }
if (timer < 100) { // (20kHz - this should never happen) if (timer < 100) { // (20kHz - this should never happen)
timer = 100; timer = 100;