/* * DriveMovement.cpp * * Created on: 17 Jan 2015 * Author: David */ #include "RepRapFirmware.h" // Prepare this DM for a Cartesian axis move void DriveMovement::PrepareCartesianAxis(const DDA& dda, const PrepParams& params, size_t drive) { const float stepsPerMm = reprap.GetPlatform()->DriveStepsPerUnit(drive) * fabs(dda.directionVector[drive]); mp.cart.twoCsquaredTimesMmPerStepDivA = (uint64_t)(((float)DDA::stepClockRate * (float)DDA::stepClockRate)/(stepsPerMm * dda.acceleration)) * 2; // Acceleration phase parameters mp.cart.accelStopStep = (uint32_t)(dda.accelDistance * stepsPerMm) + 1; startSpeedTimesCdivA = params.startSpeedTimesCdivA; #if CACHE_startSpeedTimesCdivAsquared startSpeedTimesCdivAsquared = isquare64(startSpeedTimesCdivA); #endif // Constant speed phase parameters mp.cart.mmPerStepTimesCdivtopSpeed = (uint32_t)(((float)DDA::stepClockRate * K1)/(stepsPerMm * dda.topSpeed)); accelClocksMinusAccelDistanceTimesCdivTopSpeed = params.accelClocksMinusAccelDistanceTimesCdivTopSpeed; // Deceleration phase parameters // First check whether there is any deceleration at all, otherwise we may get strange results because of rounding errors if (dda.decelDistance * stepsPerMm < 0.5) { mp.cart.decelStartStep = totalSteps + 1; topSpeedTimesCdivAPlusDecelStartClocks = 0; twoDistanceToStopTimesCsquaredDivA = 0; } else { mp.cart.decelStartStep = (uint32_t)(params.decelStartDistance * stepsPerMm) + 1; topSpeedTimesCdivAPlusDecelStartClocks = params.topSpeedTimesCdivAPlusDecelStartClocks; const uint64_t initialDecelSpeedTimesCdivASquared = isquare64(params.topSpeedTimesCdivA); twoDistanceToStopTimesCsquaredDivA = initialDecelSpeedTimesCdivASquared + (uint64_t)((params.decelStartDistance * (DDA::stepClockRateSquared * 2))/dda.acceleration); } // No reverse phase mp.cart.reverseStartStep = totalSteps + 1; mp.cart.fourMaxStepDistanceMinusTwoDistanceToStopTimesCsquaredDivA = 0; } // Prepare this DM for a Delta axis move void DriveMovement::PrepareDeltaAxis(const DDA& dda, const PrepParams& params, size_t drive) { const float stepsPerMm = reprap.GetPlatform()->DriveStepsPerUnit(drive); mp.delta.twoCsquaredTimesMmPerStepDivAK = (uint32_t)((float)DDA::stepClockRateSquared/(stepsPerMm * dda.acceleration * (K2/2))); // Acceleration phase parameters mp.delta.accelStopDsK = (uint32_t)(dda.accelDistance * stepsPerMm * K2); startSpeedTimesCdivA = params.startSpeedTimesCdivA; #if CACHE_startSpeedTimesCdivAsquared startSpeedTimesCdivAsquared = isquare64(startSpeedTimesCdivA); #endif // Constant speed phase parameters mp.delta.mmPerStepTimesCdivtopSpeedK = (uint32_t)(((float)DDA::stepClockRate * K1)/(stepsPerMm * dda.topSpeed)); accelClocksMinusAccelDistanceTimesCdivTopSpeed = params.accelClocksMinusAccelDistanceTimesCdivTopSpeed; // Deceleration phase parameters // First check whether there is any deceleration at all, otherwise we may get strange results because of rounding errors if (dda.decelDistance * stepsPerMm < 0.5) { mp.delta.decelStartDsK = (uint32_t)(dda.totalDistance * stepsPerMm * K2) * 2; // the *2 is to make sure it is definitely high enough topSpeedTimesCdivAPlusDecelStartClocks = 0; twoDistanceToStopTimesCsquaredDivA = 0; } else { mp.delta.decelStartDsK = (uint32_t)(params.decelStartDistance * stepsPerMm * K2); topSpeedTimesCdivAPlusDecelStartClocks = params.topSpeedTimesCdivAPlusDecelStartClocks; const uint64_t initialDecelSpeedTimesCdivASquared = isquare64(params.topSpeedTimesCdivA); twoDistanceToStopTimesCsquaredDivA = initialDecelSpeedTimesCdivASquared + (uint64_t)((params.decelStartDistance * (DDA::stepClockRateSquared * 2))/dda.acceleration); } } // Prepare this DM for an extruder move void DriveMovement::PrepareExtruder(const DDA& dda, const PrepParams& params, size_t drive) { const float stepsPerMm = reprap.GetPlatform()->DriveStepsPerUnit(drive) * fabs(dda.directionVector[drive]); mp.cart.twoCsquaredTimesMmPerStepDivA = (uint64_t)(((float)DDA::stepClockRate * (float)DDA::stepClockRate)/(stepsPerMm * dda.acceleration)) * 2; // Calculate the elasticity compensation parameter (not needed for axis movements, but we do them anyway to keep the code simple) const float compensationTime = reprap.GetPlatform()->GetElasticComp(drive); uint32_t compensationClocks = (uint32_t)(compensationTime * DDA::stepClockRate); const float accelCompensationDistance = compensationTime * (dda.topSpeed - dda.startSpeed); const float accelCompensationSteps = accelCompensationDistance * stepsPerMm; // Calculate the net total step count to allow for compensation (may be negative) // Note that we add totalSteps in floating point mode, to round the number of steps down consistently int32_t netSteps = (int32_t)(((dda.endSpeed - dda.startSpeed) * compensationTime * stepsPerMm) + totalSteps); // Acceleration phase parameters mp.cart.accelStopStep = (uint32_t)((dda.accelDistance * stepsPerMm) + accelCompensationSteps) + 1; startSpeedTimesCdivA = params.startSpeedTimesCdivA + compensationClocks; #if CACHE_startSpeedTimesCdivAsquared startSpeedTimesCdivAsquared = isquare64(startSpeedTimesCdivA); #endif // Constant speed phase parameters mp.cart.mmPerStepTimesCdivtopSpeed = (uint32_t)(((float)DDA::stepClockRate * K1)/(stepsPerMm * dda.topSpeed)); accelClocksMinusAccelDistanceTimesCdivTopSpeed = (int32_t)params.accelClocksMinusAccelDistanceTimesCdivTopSpeed - (int32_t)(compensationClocks * params.compFactor); // Deceleration and reverse phase parameters // First check whether there is any deceleration at all, otherwise we may get strange results because of rounding errors if (dda.decelDistance * stepsPerMm < 0.5) { totalSteps = netSteps; mp.cart.decelStartStep = mp.cart.reverseStartStep = netSteps + 1; topSpeedTimesCdivAPlusDecelStartClocks = 0; mp.cart.fourMaxStepDistanceMinusTwoDistanceToStopTimesCsquaredDivA = 0; twoDistanceToStopTimesCsquaredDivA = 0; } else { mp.cart.decelStartStep = (uint32_t)((params.decelStartDistance * stepsPerMm) + accelCompensationSteps) + 1; const int32_t initialDecelSpeedTimesCdivA = (int32_t)params.topSpeedTimesCdivA - (int32_t)compensationClocks; // signed because it may be negative and we square it const uint64_t initialDecelSpeedTimesCdivASquared = isquare64(initialDecelSpeedTimesCdivA); topSpeedTimesCdivAPlusDecelStartClocks = params.topSpeedTimesCdivAPlusDecelStartClocks - compensationClocks; twoDistanceToStopTimesCsquaredDivA = initialDecelSpeedTimesCdivASquared + (uint64_t)(((params.decelStartDistance + accelCompensationDistance) * (DDA::stepClockRateSquared * 2))/dda.acceleration); const float initialDecelSpeed = dda.topSpeed - dda.acceleration * compensationTime; const float reverseStartDistance = (initialDecelSpeed > 0.0) ? fsquare(initialDecelSpeed)/(2 * dda.acceleration) + params.decelStartDistance : params.decelStartDistance; // Reverse phase parameters if (reverseStartDistance >= dda.totalDistance) { // No reverse phase totalSteps = netSteps; mp.cart.reverseStartStep = netSteps + 1; mp.cart.fourMaxStepDistanceMinusTwoDistanceToStopTimesCsquaredDivA = 0; } else { mp.cart.reverseStartStep = (initialDecelSpeed < 0.0) ? mp.cart.decelStartStep : (twoDistanceToStopTimesCsquaredDivA/mp.cart.twoCsquaredTimesMmPerStepDivA) + 1; // Because the step numbers are rounded down, we may sometimes get a situation in which netSteps = 1 and reverseStartStep = 1. // This would lead to totalSteps = -1, which must be avoided. int32_t overallSteps = (int32_t)(2 * (mp.cart.reverseStartStep - 1)) - netSteps; if (overallSteps > 0) { totalSteps = overallSteps; mp.cart.fourMaxStepDistanceMinusTwoDistanceToStopTimesCsquaredDivA = (int64_t)((2 * (mp.cart.reverseStartStep - 1)) * mp.cart.twoCsquaredTimesMmPerStepDivA) - (int64_t)twoDistanceToStopTimesCsquaredDivA; } else { totalSteps = (uint)max(netSteps, 0); mp.cart.reverseStartStep = totalSteps + 1; mp.cart.fourMaxStepDistanceMinusTwoDistanceToStopTimesCsquaredDivA = 0; } } } } void DriveMovement::DebugPrint(char c, bool isDeltaMovement) const { if (moving || stepError) { debugPrintf("DM%c%s dir=%c steps=%u next=%u sstcda=%u " #if CACHE_startSpeedTimesCdivAsquared "sstcda2=%" PRIu64 " " #endif "acmadtcdts=%d tstcdapdsc=%u tstdca2=%" PRIu64 "\n", c, (stepError) ? " ERR:" : ":", (direction) ? 'F' : 'B', totalSteps, nextStep, startSpeedTimesCdivA, #if CACHE_startSpeedTimesCdivAsquared startSpeedTimesCdivAsquared, #endif accelClocksMinusAccelDistanceTimesCdivTopSpeed, topSpeedTimesCdivAPlusDecelStartClocks, twoDistanceToStopTimesCsquaredDivA); if (isDeltaMovement) { debugPrintf("revss=%d hmz0sK=%d minusAaPlusBbTimesKs=%d dSquaredMinusAsquaredMinusBsquared=%" PRId64 "\n" "2c2mmsdak=%u asdsk=%u dsdsk=%u mmstcdts=%u\n", mp.delta.reverseStartStep, mp.delta.hmz0sK, mp.delta.minusAaPlusBbTimesKs, mp.delta.dSquaredMinusAsquaredMinusBsquaredTimesKsquaredSsquared, mp.delta.twoCsquaredTimesMmPerStepDivAK, mp.delta.accelStopDsK, mp.delta.decelStartDsK, mp.delta.mmPerStepTimesCdivtopSpeedK ); } else { debugPrintf("accelStopStep=%u decelStartStep=%u revStartStep=%u nextStep=%u nextStepTime=%u 2CsqtMmPerStepDivA=%" PRIu64 "\n", mp.cart.accelStopStep, mp.cart.decelStartStep, mp.cart.reverseStartStep, nextStep, nextStepTime, mp.cart.twoCsquaredTimesMmPerStepDivA ); debugPrintf(" mmPerStepTimesCdivtopSpeed=%u fmsdmtstdca2=%" PRId64 "\n", mp.cart.mmPerStepTimesCdivtopSpeed, mp.cart.fourMaxStepDistanceMinusTwoDistanceToStopTimesCsquaredDivA ); } } else { debugPrintf("DM%c: not moving\n", c); } } // The remaining functions are speed-critical, so use full optimisation #pragma GCC optimize ("O3") // Calculate the time since the start of the move when the next step for the specified DriveMovement is due uint32_t DriveMovement::CalcNextStepTimeCartesian(size_t drive) { if (nextStep >= totalSteps) { moving = false; return NoStepTime; } uint32_t lastStepTime = nextStepTime; // pick up the time of the last step ++nextStep; if (nextStep < mp.cart.accelStopStep) { #if CACHE_startSpeedTimesCdivAsquared nextStepTime = isqrt(startSpeedTimesCdivAsquared + (mp.cart.twoCsquaredTimesMmPerStepDivA * nextStep)) - startSpeedTimesCdivA; #else nextStepTime = isqrt(isquare64(startSpeedTimesCdivA) + (mp.cart.twoCsquaredTimesMmPerStepDivA * nextStep)) - startSpeedTimesCdivA; #endif } else if (nextStep < mp.cart.decelStartStep) { nextStepTime = (uint32_t)((int32_t)(((uint64_t)mp.cart.mmPerStepTimesCdivtopSpeed * nextStep)/K1) + accelClocksMinusAccelDistanceTimesCdivTopSpeed); } else if (nextStep < mp.cart.reverseStartStep) { uint64_t temp = mp.cart.twoCsquaredTimesMmPerStepDivA * nextStep; // Allow for possible rounding error when the end speed is zero or very small nextStepTime = (twoDistanceToStopTimesCsquaredDivA > temp) ? topSpeedTimesCdivAPlusDecelStartClocks - isqrt(twoDistanceToStopTimesCsquaredDivA - temp) : topSpeedTimesCdivAPlusDecelStartClocks; } else { if (nextStep == mp.cart.reverseStartStep) { reprap.GetPlatform()->SetDirection(drive, !direction, false); } nextStepTime = topSpeedTimesCdivAPlusDecelStartClocks + isqrt((int64_t)(mp.cart.twoCsquaredTimesMmPerStepDivA * nextStep) - mp.cart.fourMaxStepDistanceMinusTwoDistanceToStopTimesCsquaredDivA); } if ((int32_t)nextStepTime < (int32_t)(lastStepTime + DDA::MinStepTime) && nextStep > 1) { stepError = true; return NoStepTime; } return nextStepTime; } // Calculate the time since the start of the move when the next step for the specified DriveMovement is due uint32_t DriveMovement::CalcNextStepTimeDelta(const DDA &dda, size_t drive) { if (nextStep >= totalSteps) { moving = false; return NoStepTime; } uint32_t lastStepTime = nextStepTime; // pick up the time of the last step ++nextStep; if (nextStep == mp.delta.reverseStartStep) { direction = false; reprap.GetPlatform()->SetDirection(drive, false, false); // going down now } // Calculate d*s*K as an integer, where d = distance the head has travelled, s = steps/mm for this drive, K = a power of 2 to reduce the rounding errors if (direction) { mp.delta.hmz0sK += (int32_t)K2; } else { mp.delta.hmz0sK -= (int32_t)K2; } const int32_t hmz0scK = (int32_t)(((int64_t)mp.delta.hmz0sK * dda.cKc)/Kc); const int32_t t1 = mp.delta.minusAaPlusBbTimesKs + hmz0scK; const int32_t t2 = isqrt(isquare64(t1) + mp.delta.dSquaredMinusAsquaredMinusBsquaredTimesKsquaredSsquared - isquare64(mp.delta.hmz0sK)); const int32_t dsK = (direction) ? t1 - t2 : t1 + t2; // Now feed dsK into a modified version of the step algorithm for Cartesian motion without elasticity compensation if (dsK < 0) { stepError = true; nextStep += 1000000; // so that we can tell what happened in the debug print return NoStepTime; } if ((uint32_t)dsK < mp.delta.accelStopDsK) { #if CACHE_startSpeedTimesCdivAsquared nextStepTime = isqrt(startSpeedTimesCdivAsquared + ((uint64_t)mp.delta.twoCsquaredTimesMmPerStepDivAK * (uint32_t)dsK)) - startSpeedTimesCdivA; #else nextStepTime = isqrt(isquare64(startSpeedTimesCdivA) + ((uint64_t)mp.delta.twoCsquaredTimesMmPerStepDivAK * (uint32_t)dsK)) - startSpeedTimesCdivA; #endif } else if ((uint32_t)dsK < mp.delta.decelStartDsK) { nextStepTime = (uint32_t)((int32_t)(((uint64_t)mp.delta.mmPerStepTimesCdivtopSpeedK * (uint32_t)dsK)/(K1 * K2)) + accelClocksMinusAccelDistanceTimesCdivTopSpeed); } else { uint64_t temp = (uint64_t)mp.delta.twoCsquaredTimesMmPerStepDivAK * (uint32_t)dsK; // Because of possible rounding error when the end speed is zero or very small, we need to check that the square root will work OK nextStepTime = (temp < twoDistanceToStopTimesCsquaredDivA) ? topSpeedTimesCdivAPlusDecelStartClocks - isqrt(twoDistanceToStopTimesCsquaredDivA - temp) : topSpeedTimesCdivAPlusDecelStartClocks; } if ((int32_t)nextStepTime < (int32_t)(lastStepTime + DDA::MinStepTime) && nextStep > 1) { stepError = true; // debugPrintf("%u %u %u %d %d %d %d\n", nextStep, nextStepTime, lastStepTime, dsK, t1, t2, mp.delta.hmz0sK); return NoStepTime; } return nextStepTime; } // Reduce the speed of this movement. Called to reduce the homing speed when we detect we are near the endstop for a drive. void DriveMovement::ReduceSpeed(const DDA& dda, float inverseSpeedFactor) { if (dda.isDeltaMovement) { // Force the linear motion phase mp.delta.accelStopDsK = 0; mp.delta.decelStartDsK = 0xFFFFFFFF; // Adjust the speed mp.delta.mmPerStepTimesCdivtopSpeedK = (uint32_t)(inverseSpeedFactor * mp.delta.mmPerStepTimesCdivtopSpeedK); // Adjust the acceleration clocks to as to maintain continuity of movement const int32_t hmz0scK = (int32_t)(((int64_t)mp.delta.hmz0sK * dda.cKc)/Kc); const int32_t t1 = mp.delta.minusAaPlusBbTimesKs + hmz0scK; const int32_t t2 = isqrt(isquare64(t1) + mp.delta.dSquaredMinusAsquaredMinusBsquaredTimesKsquaredSsquared - isquare64(mp.delta.hmz0sK)); const int32_t dsK = (direction) ? t1 - t2 : t1 + t2; accelClocksMinusAccelDistanceTimesCdivTopSpeed = (int32_t)nextStepTime - (int32_t)(((uint64_t)mp.delta.mmPerStepTimesCdivtopSpeedK * (uint32_t)dsK)/(K1 * K2)); } else { // Force the linear motion phase mp.cart.decelStartStep = totalSteps + 1; mp.cart.accelStopStep = 0; // Adjust the speed mp.cart.mmPerStepTimesCdivtopSpeed = (uint32_t)(inverseSpeedFactor * mp.cart.mmPerStepTimesCdivtopSpeed); // Adjust the acceleration clocks to as to maintain continuity of movement accelClocksMinusAccelDistanceTimesCdivTopSpeed = (int32_t)nextStepTime - (int32_t)(((uint64_t)mp.cart.mmPerStepTimesCdivtopSpeed * nextStep)/K1); } } // Fast 64-bit integer square root function /* static */ uint32_t DriveMovement::isqrt(uint64_t num) { //irqflags_t flags = cpu_irq_save(); //uint32_t t2 = Platform::GetInterruptClocks(); uint32_t numHigh = (uint32_t)(num >> 32); if (numHigh != 0) { uint32_t resHigh = 0; #define iter64a(N) \ { \ uint32_t temp = resHigh + (1 << N); \ if (numHigh >= temp << N) \ { \ numHigh -= temp << N; \ resHigh |= 2 << N; \ } \ } // We need to do 16 iterations iter64a(15); iter64a(14); iter64a(13); iter64a(12); iter64a(11); iter64a(10); iter64a(9); iter64a(8); iter64a(7); iter64a(6); iter64a(5); iter64a(4); iter64a(3); iter64a(2); iter64a(1); iter64a(0); // resHigh is twice the square root of the msw, in the range 0..2^17-1 uint64_t res = (uint64_t)resHigh << 16; uint64_t numAll = ((uint64_t)numHigh << 32) | (uint32_t)num; #define iter64b(N) \ { \ uint64_t temp = res | (1 << N); \ if (numAll >= temp << N) \ { \ numAll -= temp << N; \ res |= 2 << N; \ } \ } // We need to do 16 iterations iter64b(15); iter64b(14); iter64b(13); iter64b(12); iter64b(11); iter64b(10); iter64b(9); iter64b(8); iter64b(7); iter64b(6); iter64b(5); iter64b(4); iter64b(3); iter64b(2); iter64b(1); iter64b(0); uint32_t rslt = (uint32_t)(res >> 1); //uint32_t t3 = Platform::GetInterruptClocks() - t2; if (t3 < minCalcTime) minCalcTime = t3; if (t3 > maxCalcTime) maxCalcTime = t3; //cpu_irq_restore(flags); //uint64_t num3 = (uint64_t)rslt * rslt; if (num3 > num || (num - num3) > 2*rslt) {++sqrtErrors; lastNum = num; lastRes = rslt; } return rslt; } else { // 32-bit square root uint32_t num32 = (uint32_t)num; uint32_t res32 = 0; // Thanks to Wilco Dijksra for this efficient ARM algorithm #define iter32(N) \ { \ uint32_t temp = res32 | (1 << N); \ if (num32 >= temp << N) \ { \ num32 -= temp << N; \ res32 |= 2 << N; \ } \ } // We need to do 16 iterations iter32(15); iter32(14); iter32(13); iter32(12); iter32(11); iter32(10); iter32(9); iter32(8); iter32(7); iter32(6); iter32(5); iter32(4); iter32(3); iter32(2); iter32(1); iter32(0); res32 >>= 1; //uint32_t t3 = Platform::GetInterruptClocks() - t2; if (t3 < minCalcTime) minCalcTime = t3; if (t3 > maxCalcTime) maxCalcTime = t3; //cpu_irq_restore(flags); //uint64_t num3 = (uint64_t)res32 * res32; if (num3 > num || (num - num3) > 2*res32) {++sqrtErrors; lastNum = num; lastRes = res32; } return res32; } } // End