blob: 59ab0fd64fe5c0fdf0fcc36323b5f55c7aac9337 [file] [log] [blame]
@ Function: WT_Interpolate
@ Processor: ARM-E
@ Description: the main synthesis function when fetching
@ wavetable samples.
@ C-callable.
@ Usage:
@ void WT_Interpolate(
@ S_WT_VOICE *pWTVoice,
@ S_WT_FRAME *pWTFrame);
@ Copyright Sonic Network Inc. 2004
@ Revision Control:
@ $Revision: 496 $
@ $Date: 2006-12-11 14:33:26 -0800 (Mon, 11 Dec 2006) $
@ where:
@ S_WT_VOICE *pWTVoice
@ S_WT_FRAME *pWTFrame;
#include ""
.global WT_Interpolate
@ Register usage
@ --------------
pWTVoice .req r0
pWTFrame .req r1
numSamples .req r2
phaseIncrement .req r3
pOutputBuffer .req r4
tmp0 .req r1 @reuse register
tmp1 .req r5
tmp2 .req r6
pLoopEnd .req r7
pLoopStart .req r8
pPhaseAccum .req r9
phaseFrac .req r10
phaseFracMask .req r11
@SaveRegs RLIST {r4-r11,lr}
@RestoreRegs RLIST {r4-r11,pc}
STMFD sp!,{r4-r11,lr}
@ Fetch parameters from structures
LDR pOutputBuffer, [pWTFrame, #m_pAudioBuffer]
LDR numSamples, [pWTFrame, #m_numSamples]
LDR phaseIncrement, [pWTFrame, #m_phaseIncrement]
LDR pPhaseAccum, [pWTVoice, #m_pPhaseAccum]
LDR phaseFrac, [pWTVoice, #m_phaseFrac]
LDR pLoopStart, [pWTVoice, #m_pLoopStart]
LDR pLoopEnd, [pWTVoice, #m_pLoopEnd]
ADD pLoopEnd, pLoopEnd, #1 @ need loop end to equal last sample + 1
SUBS tmp0, pPhaseAccum, pLoopEnd @ check for loop end
ADDGE pPhaseAccum, pLoopStart, tmp0 @ loop back to start
#ifdef SAMPLES_8_BIT
LDRSB tmp0, [pPhaseAccum] @ tmp0 = x0
LDRSB tmp1, [pPhaseAccum, #1] @ tmp1 = x1
#elif SAMPLES_16_BIT
LDRSH tmp0, [pPhaseAccum] @ tmp0 = x0
LDRSH tmp1, [pPhaseAccum, #2] @ tmp1 = x1
#error Must define one of SAMPLES_8_BIT or SAMPLES_16_BIT.
ADD tmp2, phaseIncrement, phaseFrac @ increment pointer here to avoid pipeline stall
SUB tmp1, tmp1, tmp0 @ tmp1 = x1 - x0
SMULBB tmp1, phaseFrac, tmp1 @ tmp1 = phaseFrac * tmp2
@ This section performs a gain adjustment of -12dB for 16-bit samples
@ or +36dB for 8-bit samples. For a high quality synthesizer, the output
@ can be set to full scale, however if the filter is used, it can overflow
@ with certain coefficients and signal sources. In this case, either a
@ saturation operation should take in the filter before scaling back to
@ 16 bits or the signal path should be increased to 18 bits or more.
#ifdef SAMPLES_8_BIT
MOV tmp0, tmp0, LSL #6 @ boost 8-bit signal by 36dB
#elif SAMPLES_16_BIT
MOV tmp0, tmp0, ASR #2 @ reduce 16-bit signal by 12dB
#error Must define one of SAMPLES_8_BIT or SAMPLES_16_BIT.
ADD tmp1, tmp0, tmp1, ASR #(NUM_EG1_FRAC_BITS-6) @ tmp1 = tmp0 + (tmp1 >> (15-6))
@ = x0 + f * (x1 - x0) == interpolated result
STRH tmp1, [pOutputBuffer], #NEXT_OUTPUT_PCM @ *pOutputBuffer++ = interpolated result
@ carry overflow from fraction to integer portion
AND phaseFrac, tmp2, phaseFracMask @ nphaseFrac = frac part
SUBS numSamples, numSamples, #1
BGT InterpolationLoop
@ update and store phase
STR pPhaseAccum, [pWTVoice, #m_pPhaseAccum]
STR phaseFrac, [pWTVoice, #m_phaseFrac]
LDMFD sp!,{r4-r11,lr}
BX lr