Mercurial > hg > beaglert

annotate projects/d-box/audio_routines.S @ 0:8a575ba3ab52

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Initial commit.
author andrewm
date Fri, 31 Oct 2014 19:10:17 +0100
parents
children
rev   line source
andrewm@0 1 @
andrewm@0 2 @ audio_routines.S
andrewm@0 3 @
andrewm@0 4 @ NEON-based functions for time-critical audio processing
andrewm@0 5 @
andrewm@0 6 @ Andrew McPherson 2014
andrewm@0 7 @ Queen Mary University of London
andrewm@0 8 @
andrewm@0 9
andrewm@0 10 .syntax unified
andrewm@0 11 .arch armv7-a
andrewm@0 12 .fpu neon
andrewm@0 13
andrewm@0 14 @ void oscillator_bank_neon(int numAudioFrames, float *audioOut,
andrewm@0 15 @ int activePartialNum, int lookupTableSize,
andrewm@0 16 @ float *phases, float *frequencies, float *amplitudes,
andrewm@0 17 @ float *freqDerivatives, float *ampDerivatives,
andrewm@0 18 @ float *lookupTable);
andrewm@0 19
andrewm@0 20 @ Registers:
andrewm@0 21 @ r0: numAudioFrames How many frames to render
andrewm@0 22 @ r1: audioOut Buffer for audio output samples [stereo]
andrewm@0 23 @ r2: activePartialNum How many active partials to render
andrewm@0 24 @ r3: lookupTableSize Size of lookup table
andrewm@0 25 @ ---- other arguments start on the stack and are moved: -----
andrewm@0 26 @ r4: phases Phase of each oscillator (pointer)
andrewm@0 27 @ r5: frequencies Normalised frequency of each oscillator (pointer)
andrewm@0 28 @ r6: amplitudes Normalised amplitude of each oscillator (pointer)
andrewm@0 29 @ r7: freqDerivatives Derivative of frequency for each oscillator (pointer)
andrewm@0 30 @ r8: ampDerivatives Derivative of amplitude for each oscillator (pointer)
andrewm@0 31 @ r9: lookupTable Lookup table containing one oscillation
andrewm@0 32 @
andrewm@0 33 @ Alignment requirements:
andrewm@0 34 @ audioOut: 8-byte boundary
andrewm@0 35 @ phases: 16-byte boundary
andrewm@0 36 @ frequencies: 16-byte boundary
andrewm@0 37 @ amplitudes: 16-byte boundary
andrewm@0 38 @ freqDerivatives: 16-byte bounary
andrewm@0 39 @ ampDerivatives: 16-byte boundary
andrewm@0 40 @ lookupTable: 4-byte boundary (TODO: check this)
andrewm@0 41
andrewm@0 42 .align 2
andrewm@0 43 .global oscillator_bank_neon
andrewm@0 44 .thumb
andrewm@0 45 .thumb_func
andrewm@0 46 .type oscillator_bank_neon, %function
andrewm@0 47 oscillator_bank_neon:
andrewm@0 48
andrewm@0 49
andrewm@0 50 dSample .dn D6.F32
andrewm@0 51 qPhases .qn Q8.F32
andrewm@0 52 dPhases_0 .dn D16.F32
andrewm@0 53 dPhases_1 .dn D17.F32
andrewm@0 54 qFreqs .qn Q9.F32
andrewm@0 55 dFreqs_0 .dn D18.F32
andrewm@0 56 dFreqs_1 .dn D19.F32
andrewm@0 57 qAmps .qn Q10.F32
andrewm@0 58 dAmps_0 .dn D20.F32
andrewm@0 59 dAmps_1 .dn D21.F32
andrewm@0 60 qFreqDs .qn Q11.F32
andrewm@0 61 dFreqDs_0 .dn D22.F32
andrewm@0 62 dFreqDs_1 .dn D23.F32
andrewm@0 63 qAmpDs .qn Q12.F32
andrewm@0 64 dAmpDs_0 .dn D24.F32
andrewm@0 65 dAmpDs_1 .dn D25.F32
andrewm@0 66
andrewm@0 67 qBaseInts .qn Q13.U32 @ Base indexes: unsigned ints x4
andrewm@0 68 dBaseInts_0 .dn D26.U32
andrewm@0 69 dBaseInts_1 .dn D27.U32
andrewm@0 70 qFractions .qn Q14.F32 @ Fraction indexes: floats x4
andrewm@0 71 qTableBase .qn Q15.U32 @ Base of lookup table
andrewm@0 72
andrewm@0 73 cmp r0, #0 @ Check for trivial case 1: zero frames
andrewm@0 74 it eq
andrewm@0 75 bxeq lr @ Return if that's the case (otherwise might have odd behaviour)
andrewm@0 76 cmp r2, #4 @ Check for trivial case 2: zero oscillators
andrewm@0 77 it lt
andrewm@0 78 bxlt lr @ Return if that's the case
andrewm@0 79
andrewm@0 80 push {r4-r11} @ Now arguments start 32 bytes above SP
andrewm@0 81 add r11, sp, #32 @ Pointer to 32 bytes into the stack
andrewm@0 82 ldm r11, {r4-r9} @ Load 6 arguments into registers
andrewm@0 83
andrewm@0 84 vdup qTableBase, r9 @ Move lookup table base index into 4 ints
andrewm@0 85
andrewm@0 86 @ Outer loop: iterate over the number of oscillators, choosing 4 at a
andrewm@0 87 @ time to work with.
andrewm@0 88 oscbank_oscillator_loop:
andrewm@0 89 vld1 {dPhases_0, dPhases_1}, [r4] @ no increment; will store at end of sample loop
andrewm@0 90 vld1 {dFreqs_0, dFreqs_1}, [r5]
andrewm@0 91 vld1 {dAmps_0, dAmps_1}, [r6]
andrewm@0 92 vld1 {dFreqDs_0, dFreqDs_1}, [r7]! @ increment; won't update at end of sample loop
andrewm@0 93 vld1 {dAmpDs_0, dAmpDs_1}, [r8]!
andrewm@0 94
andrewm@0 95 push {r0-r1,r4-r8}
andrewm@0 96 @ --- inner loop: iterate over the number of samples ---
andrewm@0 97 oscbank_sample_loop:
andrewm@0 98 vcvt qBaseInts, qPhases @ Take floor(phases)
andrewm@0 99 vmov q2.f32, #1.0 @ Load 1.0 into every slot of q2
andrewm@0 100 vshl q0.U32, qBaseInts, #2 @ Shift the indexes left 2 (*4 for float addressing)
andrewm@0 101 vcvt qFractions, qBaseInts @ int back to float
andrewm@0 102 vadd q0.U32, q0.U32, qTableBase @ Find memory addresses
andrewm@0 103
andrewm@0 104 vmov r4, r5, d0 @ Move two indexes to ARM registers
andrewm@0 105 vmov r6, r7, d1 @ Move two more indexes to ARM registers
andrewm@0 106 vsub qFractions, qPhases, qFractions @ fraction = phase - floor(phase)
andrewm@0 107
andrewm@0 108 vldr.64 d0, [r4] @ Load two consecutive floats at each location
andrewm@0 109 vldr.64 d1, [r5] @ These hold the previous and following samples in the table
andrewm@0 110 vldr.64 d2, [r6] @ TODO: check whether these work at 4-byte alignment
andrewm@0 111 vldr.64 d3, [r7]
andrewm@0 112
andrewm@0 113 @ Format at this point:
andrewm@0 114 @ Osc0(before) Osc0(after) Osc1(before) Osc1(after) Osc2(before) Osc2(after) Osc3(before) Osc3(after)
andrewm@0 115 @ We want:
andrewm@0 116 @ Osc0(before) Osc1(before) Osc2(before) Osc3(before) Osc0(after) Osc1(after) Osc2(after) Osc3(after)
andrewm@0 117
andrewm@0 118 vuzp.32 q0, q1 @ Now q0 contains before, q1 contains after
andrewm@0 119 vsub q2.f32, q2.f32, qFractions @ q2 = 1.0 - fraction
andrewm@0 120 vmul q1.f32, q1.f32, qFractions @ q1 = fraction * after
andrewm@0 121 vmul q0.f32, q0.f32, q2.f32 @ q0 = (1.0 - fraction) * before
andrewm@0 122
andrewm@0 123 vadd qPhases, qPhases, qFreqs @ Update phases
andrewm@0 124 vadd qFreqs, qFreqs, qFreqDs @ Update frequencies
andrewm@0 125
andrewm@0 126 vadd q0.f32, q0.f32, q1.f32 @ Add two interpolated components to get the final sample
andrewm@0 127 vdup q2.u32, r3 @ Put lookup table size into each element of q2
andrewm@0 128 vcvt qBaseInts, qPhases @ Take floor of new phases
andrewm@0 129 vmul q0.f32, q0.f32, qAmps @ Multiply samples by current amplitude
andrewm@0 130
andrewm@0 131 vld1 dSample, [r1] @ Load the current stereo samples
andrewm@0 132 vpadd d2.f32, d0.f32, d1.f32 @ Pairwise accumulate q0 (output sample) into d2
andrewm@0 133
andrewm@0 134 vand q2, q2, qBaseInts @ Logical AND of new phase int leaves 1 bit set only if phase >= table size
andrewm@0 135 vpadd d3.f32, d2.f32, d2.f32 @ Pairwise accumulate d2 into d0 --> d0[0] and d0[1] both hold total of 4 oscillators
andrewm@0 136 vadd qAmps, qAmps, qAmpDs @ Update amplitudes
andrewm@0 137 vcvt q0.f32, q2.u32 @ Convert int back to float after AND operation
andrewm@0 138
andrewm@0 139 vadd dSample, dSample, d3.f32 @ Add oscillator outputs to each channel
andrewm@0 140
andrewm@0 141 subs r0, r0, #1 @ numFrames--
andrewm@0 142 vsub qPhases, qPhases, q0.f32 @ Keep phases in table range
andrewm@0 143 vst1 dSample, [r1]! @ Store back in buffer and increment by 8
andrewm@0 144
andrewm@0 145 it gt
andrewm@0 146 bgt oscbank_sample_loop @ Loop if numFrames > 0
andrewm@0 147
andrewm@0 148 @ --- end inner loop ---
andrewm@0 149 pop {r0-r1,r4-r8} @ Restore registers: restores audioOut and numFrames, among others
andrewm@0 150
andrewm@0 151 vst1 {dPhases_0, dPhases_1}, [r4]! @ Store phases back to array
andrewm@0 152 vst1 {dFreqs_0, dFreqs_1}, [r5]! @ Store frequencies back to array
andrewm@0 153 vst1 {dAmps_0, dAmps_1}, [r6]! @ Store amplitudes back to array
andrewm@0 154 @ No need to update r7, r8
andrewm@0 155
andrewm@0 156 subs r2, r2, #4 @ numPartials -= 4
andrewm@0 157 it gt
andrewm@0 158 bgt oscbank_oscillator_loop @ Loop if numPartials > 0
andrewm@0 159
andrewm@0 160 pop {r4-r11}
andrewm@0 161 bx lr
andrewm@0 162
andrewm@0 163
andrewm@0 164 @ void wavetable_interpolate_neon(int numSamplesIn, int numSamplesOut,
andrewm@0 165 @ float *tableIn, float *tableOut);
andrewm@0 166
andrewm@0 167 @ Registers:
andrewm@0 168 @ r0: numSamplesIn Size of the input table
andrewm@0 169 @ r1: numSamplesOut Size of the output table
andrewm@0 170 @ r2: tableIn Pointer to input table
andrewm@0 171 @ r3: tableOut Pointer to output table
andrewm@0 172
andrewm@0 173 @ Alignment requirements:
andrewm@0 174 @ tableIn: 8-byte boundary
andrewm@0 175 @ tableOut: 8-byte boundary
andrewm@0 176
andrewm@0 177 .align 2
andrewm@0 178 .global wavetable_interpolate_neon
andrewm@0 179 .thumb
andrewm@0 180 .thumb_func
andrewm@0 181 .type wavetable_interpolate_neon, %function
andrewm@0 182 wavetable_interpolate_neon:
andrewm@0 183 @ TODO
andrewm@0 184
andrewm@0 185 bx lr