From 87021f7c0ac4620eafd185ff11905ee643f72b6c Mon Sep 17 00:00:00 2001 From: Michael Sevakis Date: Sat, 18 May 2013 01:45:03 -0400 Subject: SPC Codec: Refactor for CPU and clean up some things. CPU optimization gets its own files in which to fill-in optimizable routines. Some pointless #if 0's for profiling need removal. Those macros are empty if not profiling. Force some functions that are undesirable to be force-inlined by the compiler to be not inlined. Change-Id: Ia7b7e45380d7efb20c9b1a4d52e05db3ef6bbaab --- lib/rbcodec/codecs/libspc/cpu/spc_dsp_armv6.c | 244 ++++++++++++++++++++++++++ 1 file changed, 244 insertions(+) create mode 100644 lib/rbcodec/codecs/libspc/cpu/spc_dsp_armv6.c (limited to 'lib/rbcodec/codecs/libspc/cpu/spc_dsp_armv6.c') diff --git a/lib/rbcodec/codecs/libspc/cpu/spc_dsp_armv6.c b/lib/rbcodec/codecs/libspc/cpu/spc_dsp_armv6.c new file mode 100644 index 0000000000..2e3de87613 --- /dev/null +++ b/lib/rbcodec/codecs/libspc/cpu/spc_dsp_armv6.c @@ -0,0 +1,244 @@ +/*************************************************************************** + * __________ __ ___. + * Open \______ \ ____ ____ | | _\_ |__ _______ ___ + * Source | _// _ \_/ ___\| |/ /| __ \ / _ \ \/ / + * Jukebox | | ( <_> ) \___| < | \_\ ( <_> > < < + * Firmware |____|_ /\____/ \___ >__|_ \|___ /\____/__/\_ \ + * \/ \/ \/ \/ \/ + * $Id$ + * + * Copyright (C) 2010 Michael Sevakis (jhMikeS) + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation; either version 2 + * of the License, or (at your option) any later version. + * + * This software is distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY + * KIND, either express or implied. + * + ****************************************************************************/ +#if !SPC_NOINTERP + +#define SPC_GAUSSIAN_FAST_INTERP +static inline int gaussian_fast_interp( int16_t const* samples, + int32_t position, + int16_t const* fwd, + int16_t const* rev ) +{ + int output; + int t0, t1, t2, t3; + + asm volatile ( + /* NOTE: often-unaligned accesses */ + "ldr %[t0], [%[samp]] \n" /* t0=i0i1 */ + "ldr %[t2], [%[fwd]] \n" /* t2=f0f1 */ + "ldr %[t1], [%[samp], #4] \n" /* t1=i2i3 */ + "ldr %[t3], [%[rev]] \n" /* t3=r0r1 */ + "smuad %[out], %[t0], %[t2] \n" /* out=f0*i0+f1*i1 */ + "smladx %[out], %[t1], %[t3], %[out] \n" /* out+=r1*i2+r0*i3 */ + : [out]"=r"(output), + [t0]"=&r"(t0), [t1]"=&r"(t1), [t2]"=&r"(t2), [t3]"=r"(t3) + : [fwd]"r"(fwd), [rev]"r"(rev), + [samp]"r"(samples + (position >> 12))); + + return output; +} + +#define SPC_GAUSSIAN_FAST_AMP +static inline int gaussian_fast_amp( struct voice_t* voice, int output, + int* amp_0, int* amp_1 ) +{ + int t0; + + asm volatile ( + "mov %[t0], %[out], asr #(11-5) \n" /* To do >> 16 below */ + "mul %[out], %[t0], %[envx] \n" + : [out]"+r"(output), [t0]"=&r"(t0) + : [envx]"r"((int) voice->envx)); + + asm volatile ( + "smulwb %[a0], %[out], %[v0] \n" /* amp * vol >> 16 */ + "smulwb %[a1], %[out], %[v1] \n" + : [a0]"=&r"(*amp_0), [a1]"=r"(*amp_1) + : [out]"r"(output), + [v0]"r"(voice->volume [0]), + [v1]"r"(voice->volume [1])); + + return output >> 5; /* 'output' still 5 bits too big */ +} + +#define SPC_GAUSSIAN_SLOW_INTERP +static inline int gaussian_slow_interp( int16_t const* samples, + int32_t position, + int16_t const* fwd, + int16_t const* rev ) +{ + int output; + int t0, t1, t2, t3; + + asm volatile ( + /* NOTE: often-unaligned accesses */ + "ldr %[t0], [%[samp]] \n" /* t0=i0i1 */ + "ldr %[t2], [%[fwd]] \n" /* t2=f0f1 */ + "ldr %[t1], [%[samp], #4] \n" /* t1=i2i3 */ + "ldr %[t3], [%[rev]] \n" /* t3=f2f3 */ + "smulbb %[out], %[t0], %[t2] \n" /* out=f0*i0 */ + "smultt %[t0], %[t0], %[t2] \n" /* t0=f1*i1 */ + "smulbt %[t2], %[t1], %[t3] \n" /* t2=r1*i2 */ + "smultb %[t3], %[t1], %[t3] \n" /* t3=r0*i3 */ + : [out]"=r"(output), + [t0]"=&r"(t0), [t1]"=&r"(t1), [t2]"=&r"(t2), [t3]"=r"(t3) + : [fwd]"r"(fwd), [rev]"r"(rev), + [samp]"r"(samples + (position >> 12))); + + asm volatile ( + "mov %[out], %[out], asr #12 \n" + "add %[t0], %[out], %[t0], asr #12 \n" + "add %[t2], %[t0], %[t2], asr #12 \n" + "pkhbt %[t0], %[t2], %[t3], asl #4 \n" /* t3[31:16], t2[15:0] */ + "sadd16 %[t0], %[t0], %[t0] \n" /* t3[31:16]*2, t2[15:0]*2 */ + "qsubaddx %[out], %[t0], %[t0] \n" /* out[15:0]= + * sat16(t3[31:16]+t2[15:0]) */ + : [out]"+r"(output), + [t0]"+r"(t0), [t2]"+r"(t2), [t3]"+r"(t3)); + + /* output will be sign-extended in next step */ + return output; +} + +#define SPC_GAUSSIAN_SLOW_AMP +static inline int gaussian_slow_amp( struct voice_t* voice, int output, + int* amp_0, int* amp_1 ) +{ + asm volatile ( + "smulbb %[out], %[out], %[envx]" + : [out]"+r"(output) + : [envx]"r"(voice->envx)); + + asm volatile ( + "mov %[out], %[out], asr #11 \n" + "bic %[out], %[out], #0x1 \n" + "smulbb %[amp_0], %[out], %[v0] \n" + "smulbb %[amp_1], %[out], %[v1] \n" + : [out]"+r"(output), + [amp_0]"=&r"(*amp_0), [amp_1]"=r"(*amp_1) + : [v0]"r"(voice->volume[0]), [v1]"r"(voice->volume[1])); + + return output; +} + +#endif /* !SPC_NOINTERP */ + +#if !SPC_NOECHO + +#define SPC_DSP_ECHO_APPLY + +/* Echo filter history */ +static int32_t fir_buf[FIR_BUF_CNT] IBSS_ATTR_SPC + __attribute__(( aligned(FIR_BUF_ALIGN*1) )); + +static inline void echo_init( struct Spc_Dsp* this ) +{ + this->fir.ptr = fir_buf; + ci->memset( fir_buf, 0, sizeof fir_buf ); +} + +static inline void echo_apply(struct Spc_Dsp* this, + uint8_t* const echo_ptr, int* out_0, int* out_1) +{ + /* Keep last 8 samples */ + int32_t* fir_ptr; + int t0; + asm volatile ( + "ldr %[t0], [%[ep]] \n" + "add %[p], %[t_p], #4 \n" + "bic %[t_p], %[p], %[mask] \n" + "str %[t0], [%[p], #-4] \n" + /* duplicate at +8 eliminates wrap checking below */ + "str %[t0], [%[p], #28] \n" + : [p]"=&r"(fir_ptr), [t_p]"+r"(this->fir.ptr), + [t0]"=&r"(t0) + : [ep]"r"(echo_ptr), [mask]"i"(~FIR_BUF_MASK)); + + int32_t* fir_coeff = (int32_t *)this->fir.coeff; + + asm volatile ( /* L0R0 = acc0 */ + "ldmia %[p]!, { r2-r5 } \n" /* L1R1-L4R4 = r2-r5 */ + "ldmia %[c]!, { r0-r1 } \n" /* C0C1-C2C3 = r0-r1 */ + "pkhbt %[acc0], %[t0], r2, asl #16 \n" /* L0R0,L1R1->L0L1,R0R1 */ + "pkhtb r2, r2, %[t0], asr #16 \n" + "smuad %[acc0], %[acc0], r0 \n" /* acc0=L0*C0+L1*C1 */ + "smuad %[acc1], r2, r0 \n" /* acc1=R0*C0+R1*C1 */ + "pkhbt %[t0], r3, r4, asl #16 \n" /* L2R2,L3R3->L2L3,R2R3 */ + "pkhtb r4, r4, r3, asr #16 \n" + "smlad %[acc0], %[t0], r1, %[acc0] \n" /* acc0+=L2*C2+L3*C3 */ + "smlad %[acc1], r4, r1, %[acc1] \n" /* acc1+=R2*C2+R3*C3 */ + "ldmia %[p], { r2-r4 } \n" /* L5R5-L7R7 = r2-r4 */ + "ldmia %[c], { r0-r1 } \n" /* C4C5-C6C7 = r0-r1 */ + "pkhbt %[t0], r5, r2, asl #16 \n" /* L4R4,L5R5->L4L5,R4R5 */ + "pkhtb r2, r2, r5, asr #16 \n" + "smlad %[acc0], %[t0], r0, %[acc0] \n" /* acc0+=L4*C4+L5*C5 */ + "smlad %[acc1], r2, r0, %[acc1] \n" /* acc1+=R4*C4+R5*C5 */ + "pkhbt %[t0], r3, r4, asl #16 \n" /* L6R6,L7R7->L6L7,R6R7 */ + "pkhtb r4, r4, r3, asr #16 \n" + "smlad %[acc0], %[t0], r1, %[acc0] \n" /* acc0+=L6*C6+L7*C7 */ + "smlad %[acc1], r4, r1, %[acc1] \n" /* acc1+=R6*C6+R7*C7 */ + : [t0]"+r"(t0), [acc0]"=&r"(*out_0), [acc1]"=&r"(*out_1), + [p]"+r"(fir_ptr), [c]"+r"(fir_coeff) + : + : "r0", "r1", "r2", "r3", "r4", "r5"); +} + +#define SPC_DSP_ECHO_FEEDBACK +static inline void echo_feedback(struct Spc_Dsp* this, uint8_t* echo_ptr, + int echo_0, int echo_1, int fb_0, int fb_1) +{ + int e0, e1; + asm volatile ( + "mov %[e0], %[ei0], asl #7 \n" + "mov %[e1], %[ei1], asl #7 \n" + "mla %[e0], %[fb0], %[ef], %[e0] \n" + "mla %[e1], %[fb1], %[ef], %[e1] \n" + : [e0]"=&r"(e0), [e1]"=&r"(e1) + : [ei0]"r"(echo_0), [ei1]"r"(echo_1), + [fb0]"r"(fb_0), [fb1]"r"(fb_1), + [ef]"r"((int)this->r.g.echo_feedback)); + asm volatile ( + "ssat %[e0], #16, %[e0], asr #14 \n" + "ssat %[e1], #16, %[e1], asr #14 \n" + "pkhbt %[e0], %[e0], %[e1], lsl #16 \n" + "str %[e0], [%[ep]] \n" + : [e0]"+r"(e0), [e1]"+r"(e1) + : [ep]"r"((int32_t *)echo_ptr)); +} + +#define SPC_DSP_GENERATE_OUTPUT +static inline void echo_output( struct Spc_Dsp* this, int global_muting, + int global_vol_0, int global_vol_1, int chans_0, int chans_1, + int fb_0, int fb_1, int* out_0, int* out_1 ) +{ + int t0, t1; + + asm volatile ( + "mul %[t0], %[gv0], %[ch0] \n" + "mul %[t1], %[gv1], %[ch1] \n" + : [t0]"=&r"(t0), [t1]"=r"(t1) + : [gv0]"r"(global_vol_0), [gv1]"r"(global_vol_1), + [ch0]"r"(chans_0), [ch1]"r"(chans_1)); + asm volatile ( + "mla %[t0], %[i0], %[ev0], %[t0] \n" + "mla %[t1], %[i1], %[ev1], %[t1] \n" + : [t0]"+r"(t0), [t1]"+r"(t1) + : [i0]"r"(fb_0), [i1]"r"(fb_1), + [ev0]"r"((int)this->r.g.echo_volume_0), + [ev1]"r"((int)this->r.g.echo_volume_1)); + asm volatile ( + "mov %[o0], %[t0], asr %[gm] \n" + "mov %[o1], %[t1], asr %[gm] \n" + : [o0]"=&r"(*out_0), [o1]"=r"(*out_1) + : [t0]"r"(t0), [t1]"r"(t1), + [gm]"r"(global_muting)); +} + +#endif /* SPC_NOECHO */ -- cgit v1.2.3