summaryrefslogtreecommitdiff
path: root/thirdparty/opus/celt
diff options
context:
space:
mode:
Diffstat (limited to 'thirdparty/opus/celt')
-rw-r--r--thirdparty/opus/celt/arch.h13
-rw-r--r--thirdparty/opus/celt/arm/arm_celt_map.c24
-rw-r--r--thirdparty/opus/celt/arm/armcpu.c49
-rw-r--r--thirdparty/opus/celt/arm/armcpu.h6
-rw-r--r--thirdparty/opus/celt/arm/celt_neon_intr.c61
-rw-r--r--thirdparty/opus/celt/arm/pitch_arm.h68
-rw-r--r--thirdparty/opus/celt/bands.c4
-rw-r--r--thirdparty/opus/celt/celt.h2
-rw-r--r--thirdparty/opus/celt/celt_decoder.c10
-rw-r--r--thirdparty/opus/celt/celt_encoder.c13
-rw-r--r--thirdparty/opus/celt/celt_lpc.c3
-rw-r--r--thirdparty/opus/celt/cwrs.c2
-rw-r--r--thirdparty/opus/celt/fixed_generic.h16
-rw-r--r--thirdparty/opus/celt/kiss_fft.c2
-rw-r--r--thirdparty/opus/celt/mathops.c2
-rw-r--r--thirdparty/opus/celt/pitch.c67
-rw-r--r--thirdparty/opus/celt/pitch.h20
-rw-r--r--thirdparty/opus/celt/rate.c2
-rw-r--r--thirdparty/opus/celt/vq.c2
-rw-r--r--thirdparty/opus/celt/x86/pitch_sse.h34
-rw-r--r--thirdparty/opus/celt/x86/x86_celt_map.c2
-rw-r--r--thirdparty/opus/celt/x86/x86cpu.c2
22 files changed, 293 insertions, 111 deletions
diff --git a/thirdparty/opus/celt/arch.h b/thirdparty/opus/celt/arch.h
index 9f74ddd267..8ceab5fe10 100644
--- a/thirdparty/opus/celt/arch.h
+++ b/thirdparty/opus/celt/arch.h
@@ -78,6 +78,15 @@ static OPUS_INLINE void _celt_fatal(const char *str, const char *file, int line)
#define UADD32(a,b) ((a)+(b))
#define USUB32(a,b) ((a)-(b))
+/* Set this if opus_int64 is a native type of the CPU. */
+/* Assume that all LP64 architectures have fast 64-bit types; also x86_64
+ (which can be ILP32 for x32) and Win64 (which is LLP64). */
+#if defined(__x86_64__) || defined(__LP64__) || defined(_WIN64)
+#define OPUS_FAST_INT64 1
+#else
+#define OPUS_FAST_INT64 0
+#endif
+
#define PRINT_MIPS(file)
#ifdef FIXED_POINT
@@ -118,7 +127,9 @@ static OPUS_INLINE opus_int16 SAT16(opus_int32 x) {
#include "fixed_generic.h"
-#ifdef OPUS_ARM_INLINE_EDSP
+#ifdef OPUS_ARM_PRESUME_AARCH64_NEON_INTR
+#include "arm/fixed_arm64.h"
+#elif OPUS_ARM_INLINE_EDSP
#include "arm/fixed_armv5e.h"
#elif defined (OPUS_ARM_INLINE_ASM)
#include "arm/fixed_armv4.h"
diff --git a/thirdparty/opus/celt/arm/arm_celt_map.c b/thirdparty/opus/celt/arm/arm_celt_map.c
index ee6c244786..4d4d069a86 100644
--- a/thirdparty/opus/celt/arm/arm_celt_map.c
+++ b/thirdparty/opus/celt/arm/arm_celt_map.c
@@ -36,6 +36,9 @@
#if defined(OPUS_HAVE_RTCD)
# if defined(FIXED_POINT)
+# if ((defined(OPUS_ARM_MAY_HAVE_NEON) && !defined(OPUS_ARM_PRESUME_NEON)) || \
+ (defined(OPUS_ARM_MAY_HAVE_MEDIA) && !defined(OPUS_ARM_PRESUME_MEDIA)) || \
+ (defined(OPUS_ARM_MAY_HAVE_EDSP) && !defined(OPUS_ARM_PRESUME_EDSP)))
opus_val32 (*const CELT_PITCH_XCORR_IMPL[OPUS_ARCHMASK+1])(const opus_val16 *,
const opus_val16 *, opus_val32 *, int , int) = {
celt_pitch_xcorr_c, /* ARMv4 */
@@ -43,8 +46,10 @@ opus_val32 (*const CELT_PITCH_XCORR_IMPL[OPUS_ARCHMASK+1])(const opus_val16 *,
MAY_HAVE_MEDIA(celt_pitch_xcorr), /* Media */
MAY_HAVE_NEON(celt_pitch_xcorr) /* NEON */
};
+
+# endif
# else /* !FIXED_POINT */
-# if defined(OPUS_ARM_MAY_HAVE_NEON_INTR)
+# if defined(OPUS_ARM_MAY_HAVE_NEON_INTR) && !defined(OPUS_ARM_PRESUME_NEON_INTR)
void (*const CELT_PITCH_XCORR_IMPL[OPUS_ARCHMASK+1])(const opus_val16 *,
const opus_val16 *, opus_val32 *, int, int) = {
celt_pitch_xcorr_c, /* ARMv4 */
@@ -55,6 +60,23 @@ void (*const CELT_PITCH_XCORR_IMPL[OPUS_ARCHMASK+1])(const opus_val16 *,
# endif
# endif /* FIXED_POINT */
+#if defined(FIXED_POINT) && defined(OPUS_HAVE_RTCD) && \
+ defined(OPUS_ARM_MAY_HAVE_NEON_INTR) && !defined(OPUS_ARM_PRESUME_NEON_INTR)
+
+void (*const XCORR_KERNEL_IMPL[OPUS_ARCHMASK + 1])(
+ const opus_val16 *x,
+ const opus_val16 *y,
+ opus_val32 sum[4],
+ int len
+) = {
+ xcorr_kernel_c, /* ARMv4 */
+ xcorr_kernel_c, /* EDSP */
+ xcorr_kernel_c, /* Media */
+ xcorr_kernel_neon_fixed, /* Neon */
+};
+
+#endif
+
# if defined(OPUS_ARM_MAY_HAVE_NEON_INTR)
# if defined(HAVE_ARM_NE10)
# if defined(CUSTOM_MODES)
diff --git a/thirdparty/opus/celt/arm/armcpu.c b/thirdparty/opus/celt/arm/armcpu.c
index 5e5d10c344..694a63b78e 100644
--- a/thirdparty/opus/celt/arm/armcpu.c
+++ b/thirdparty/opus/celt/arm/armcpu.c
@@ -37,11 +37,12 @@
#include "cpu_support.h"
#include "os_support.h"
#include "opus_types.h"
+#include "arch.h"
-#define OPUS_CPU_ARM_V4 (1)
-#define OPUS_CPU_ARM_EDSP (1<<1)
-#define OPUS_CPU_ARM_MEDIA (1<<2)
-#define OPUS_CPU_ARM_NEON (1<<3)
+#define OPUS_CPU_ARM_V4_FLAG (1<<OPUS_ARCH_ARM_V4)
+#define OPUS_CPU_ARM_EDSP_FLAG (1<<OPUS_ARCH_ARM_EDSP)
+#define OPUS_CPU_ARM_MEDIA_FLAG (1<<OPUS_ARCH_ARM_MEDIA)
+#define OPUS_CPU_ARM_NEON_FLAG (1<<OPUS_ARCH_ARM_NEON)
#if defined(_MSC_VER)
/*For GetExceptionCode() and EXCEPTION_ILLEGAL_INSTRUCTION.*/
@@ -55,20 +56,22 @@ static OPUS_INLINE opus_uint32 opus_cpu_capabilities(void){
/* MSVC has no OPUS_INLINE __asm support for ARM, but it does let you __emit
* instructions via their assembled hex code.
* All of these instructions should be essentially nops. */
-# if defined(OPUS_ARM_MAY_HAVE_EDSP)
+# if defined(OPUS_ARM_MAY_HAVE_EDSP) || defined(OPUS_ARM_MAY_HAVE_MEDIA) \
+ || defined(OPUS_ARM_MAY_HAVE_NEON) || defined(OPUS_ARM_MAY_HAVE_NEON_INTR)
__try{
/*PLD [r13]*/
__emit(0xF5DDF000);
- flags|=OPUS_CPU_ARM_EDSP;
+ flags|=OPUS_CPU_ARM_EDSP_FLAG;
}
__except(GetExceptionCode()==EXCEPTION_ILLEGAL_INSTRUCTION){
/*Ignore exception.*/
}
-# if defined(OPUS_ARM_MAY_HAVE_MEDIA)
+# if defined(OPUS_ARM_MAY_HAVE_MEDIA) \
+ || defined(OPUS_ARM_MAY_HAVE_NEON) || defined(OPUS_ARM_MAY_HAVE_NEON_INTR)
__try{
/*SHADD8 r3,r3,r3*/
__emit(0xE6333F93);
- flags|=OPUS_CPU_ARM_MEDIA;
+ flags|=OPUS_CPU_ARM_MEDIA_FLAG;
}
__except(GetExceptionCode()==EXCEPTION_ILLEGAL_INSTRUCTION){
/*Ignore exception.*/
@@ -77,7 +80,7 @@ static OPUS_INLINE opus_uint32 opus_cpu_capabilities(void){
__try{
/*VORR q0,q0,q0*/
__emit(0xF2200150);
- flags|=OPUS_CPU_ARM_NEON;
+ flags|=OPUS_CPU_ARM_NEON_FLAG;
}
__except(GetExceptionCode()==EXCEPTION_ILLEGAL_INSTRUCTION){
/*Ignore exception.*/
@@ -107,26 +110,26 @@ opus_uint32 opus_cpu_capabilities(void)
while(fgets(buf, 512, cpuinfo) != NULL)
{
-# if defined(OPUS_ARM_MAY_HAVE_EDSP) || defined(OPUS_ARM_MAY_HAVE_NEON) || defined(OPUS_ARM_MAY_HAVE_NEON_INTR)
+# if defined(OPUS_ARM_MAY_HAVE_EDSP) || defined(OPUS_ARM_MAY_HAVE_MEDIA) \
+ || defined(OPUS_ARM_MAY_HAVE_NEON) || defined(OPUS_ARM_MAY_HAVE_NEON_INTR)
/* Search for edsp and neon flag */
if(memcmp(buf, "Features", 8) == 0)
{
char *p;
-# if defined(OPUS_ARM_MAY_HAVE_EDSP)
p = strstr(buf, " edsp");
if(p != NULL && (p[5] == ' ' || p[5] == '\n'))
- flags |= OPUS_CPU_ARM_EDSP;
-# endif
+ flags |= OPUS_CPU_ARM_EDSP_FLAG;
# if defined(OPUS_ARM_MAY_HAVE_NEON) || defined(OPUS_ARM_MAY_HAVE_NEON_INTR)
p = strstr(buf, " neon");
if(p != NULL && (p[5] == ' ' || p[5] == '\n'))
- flags |= OPUS_CPU_ARM_NEON;
+ flags |= OPUS_CPU_ARM_NEON_FLAG;
# endif
}
# endif
-# if defined(OPUS_ARM_MAY_HAVE_MEDIA)
+# if defined(OPUS_ARM_MAY_HAVE_MEDIA) \
+ || defined(OPUS_ARM_MAY_HAVE_NEON) || defined(OPUS_ARM_MAY_HAVE_NEON_INTR)
/* Search for media capabilities (>= ARMv6) */
if(memcmp(buf, "CPU architecture:", 17) == 0)
{
@@ -134,7 +137,7 @@ opus_uint32 opus_cpu_capabilities(void)
version = atoi(buf+17);
if(version >= 6)
- flags |= OPUS_CPU_ARM_MEDIA;
+ flags |= OPUS_CPU_ARM_MEDIA_FLAG;
}
# endif
}
@@ -156,18 +159,26 @@ int opus_select_arch(void)
opus_uint32 flags = opus_cpu_capabilities();
int arch = 0;
- if(!(flags & OPUS_CPU_ARM_EDSP))
+ if(!(flags & OPUS_CPU_ARM_EDSP_FLAG)) {
+ /* Asserts ensure arch values are sequential */
+ celt_assert(arch == OPUS_ARCH_ARM_V4);
return arch;
+ }
arch++;
- if(!(flags & OPUS_CPU_ARM_MEDIA))
+ if(!(flags & OPUS_CPU_ARM_MEDIA_FLAG)) {
+ celt_assert(arch == OPUS_ARCH_ARM_EDSP);
return arch;
+ }
arch++;
- if(!(flags & OPUS_CPU_ARM_NEON))
+ if(!(flags & OPUS_CPU_ARM_NEON_FLAG)) {
+ celt_assert(arch == OPUS_ARCH_ARM_MEDIA);
return arch;
+ }
arch++;
+ celt_assert(arch == OPUS_ARCH_ARM_NEON);
return arch;
}
diff --git a/thirdparty/opus/celt/arm/armcpu.h b/thirdparty/opus/celt/arm/armcpu.h
index ac5744606e..820262ff5f 100644
--- a/thirdparty/opus/celt/arm/armcpu.h
+++ b/thirdparty/opus/celt/arm/armcpu.h
@@ -66,6 +66,12 @@
# if defined(OPUS_HAVE_RTCD)
int opus_select_arch(void);
+
+#define OPUS_ARCH_ARM_V4 (0)
+#define OPUS_ARCH_ARM_EDSP (1)
+#define OPUS_ARCH_ARM_MEDIA (2)
+#define OPUS_ARCH_ARM_NEON (3)
+
# endif
#endif
diff --git a/thirdparty/opus/celt/arm/celt_neon_intr.c b/thirdparty/opus/celt/arm/celt_neon_intr.c
index 47dce15ba5..47bbe3dc22 100644
--- a/thirdparty/opus/celt/arm/celt_neon_intr.c
+++ b/thirdparty/opus/celt/arm/celt_neon_intr.c
@@ -37,7 +37,66 @@
#include <arm_neon.h>
#include "../pitch.h"
-#if !defined(FIXED_POINT)
+#if defined(FIXED_POINT)
+void xcorr_kernel_neon_fixed(const opus_val16 * x, const opus_val16 * y, opus_val32 sum[4], int len)
+{
+ int j;
+ int32x4_t a = vld1q_s32(sum);
+ /* Load y[0...3] */
+ /* This requires len>0 to always be valid (which we assert in the C code). */
+ int16x4_t y0 = vld1_s16(y);
+ y += 4;
+
+ for (j = 0; j + 8 <= len; j += 8)
+ {
+ /* Load x[0...7] */
+ int16x8_t xx = vld1q_s16(x);
+ int16x4_t x0 = vget_low_s16(xx);
+ int16x4_t x4 = vget_high_s16(xx);
+ /* Load y[4...11] */
+ int16x8_t yy = vld1q_s16(y);
+ int16x4_t y4 = vget_low_s16(yy);
+ int16x4_t y8 = vget_high_s16(yy);
+ int32x4_t a0 = vmlal_lane_s16(a, y0, x0, 0);
+ int32x4_t a1 = vmlal_lane_s16(a0, y4, x4, 0);
+
+ int16x4_t y1 = vext_s16(y0, y4, 1);
+ int16x4_t y5 = vext_s16(y4, y8, 1);
+ int32x4_t a2 = vmlal_lane_s16(a1, y1, x0, 1);
+ int32x4_t a3 = vmlal_lane_s16(a2, y5, x4, 1);
+
+ int16x4_t y2 = vext_s16(y0, y4, 2);
+ int16x4_t y6 = vext_s16(y4, y8, 2);
+ int32x4_t a4 = vmlal_lane_s16(a3, y2, x0, 2);
+ int32x4_t a5 = vmlal_lane_s16(a4, y6, x4, 2);
+
+ int16x4_t y3 = vext_s16(y0, y4, 3);
+ int16x4_t y7 = vext_s16(y4, y8, 3);
+ int32x4_t a6 = vmlal_lane_s16(a5, y3, x0, 3);
+ int32x4_t a7 = vmlal_lane_s16(a6, y7, x4, 3);
+
+ y0 = y8;
+ a = a7;
+ x += 8;
+ y += 8;
+ }
+
+ for (; j < len; j++)
+ {
+ int16x4_t x0 = vld1_dup_s16(x); /* load next x */
+ int32x4_t a0 = vmlal_s16(a, y0, x0);
+
+ int16x4_t y4 = vld1_dup_s16(y); /* load next y */
+ y0 = vext_s16(y0, y4, 1);
+ a = a0;
+ x++;
+ y++;
+ }
+
+ vst1q_s32(sum, a);
+}
+
+#else
/*
* Function: xcorr_kernel_neon_float
* ---------------------------------
diff --git a/thirdparty/opus/celt/arm/pitch_arm.h b/thirdparty/opus/celt/arm/pitch_arm.h
index 8626ed75b9..14331169ee 100644
--- a/thirdparty/opus/celt/arm/pitch_arm.h
+++ b/thirdparty/opus/celt/arm/pitch_arm.h
@@ -46,10 +46,53 @@ opus_val32 celt_pitch_xcorr_edsp(const opus_val16 *_x, const opus_val16 *_y,
opus_val32 *xcorr, int len, int max_pitch);
# endif
-# if !defined(OPUS_HAVE_RTCD)
+# if defined(OPUS_HAVE_RTCD) && \
+ ((defined(OPUS_ARM_MAY_HAVE_NEON) && !defined(OPUS_ARM_PRESUME_NEON)) || \
+ (defined(OPUS_ARM_MAY_HAVE_MEDIA) && !defined(OPUS_ARM_PRESUME_MEDIA)) || \
+ (defined(OPUS_ARM_MAY_HAVE_EDSP) && !defined(OPUS_ARM_PRESUME_EDSP)))
+extern opus_val32
+(*const CELT_PITCH_XCORR_IMPL[OPUS_ARCHMASK+1])(const opus_val16 *,
+ const opus_val16 *, opus_val32 *, int, int);
+# define OVERRIDE_PITCH_XCORR (1)
+# define celt_pitch_xcorr(_x, _y, xcorr, len, max_pitch, arch) \
+ ((*CELT_PITCH_XCORR_IMPL[(arch)&OPUS_ARCHMASK])(_x, _y, \
+ xcorr, len, max_pitch))
+
+# elif defined(OPUS_ARM_PRESUME_EDSP) || \
+ defined(OPUS_ARM_PRESUME_MEDIA) || \
+ defined(OPUS_ARM_PRESUME_NEON)
# define OVERRIDE_PITCH_XCORR (1)
# define celt_pitch_xcorr(_x, _y, xcorr, len, max_pitch, arch) \
((void)(arch),PRESUME_NEON(celt_pitch_xcorr)(_x, _y, xcorr, len, max_pitch))
+
+# endif
+
+# if defined(OPUS_ARM_MAY_HAVE_NEON_INTR)
+void xcorr_kernel_neon_fixed(
+ const opus_val16 *x,
+ const opus_val16 *y,
+ opus_val32 sum[4],
+ int len);
+# endif
+
+# if defined(OPUS_HAVE_RTCD) && \
+ (defined(OPUS_ARM_MAY_HAVE_NEON_INTR) && !defined(OPUS_ARM_PRESUME_NEON_INTR))
+
+extern void (*const XCORR_KERNEL_IMPL[OPUS_ARCHMASK + 1])(
+ const opus_val16 *x,
+ const opus_val16 *y,
+ opus_val32 sum[4],
+ int len);
+
+# define OVERRIDE_XCORR_KERNEL (1)
+# define xcorr_kernel(x, y, sum, len, arch) \
+ ((*XCORR_KERNEL_IMPL[(arch) & OPUS_ARCHMASK])(x, y, sum, len))
+
+# elif defined(OPUS_ARM_PRESUME_NEON_INTR)
+# define OVERRIDE_XCORR_KERNEL (1)
+# define xcorr_kernel(x, y, sum, len, arch) \
+ ((void)arch, xcorr_kernel_neon_fixed(x, y, sum, len))
+
# endif
#else /* Start !FIXED_POINT */
@@ -57,12 +100,27 @@ opus_val32 celt_pitch_xcorr_edsp(const opus_val16 *_x, const opus_val16 *_y,
#if defined(OPUS_ARM_MAY_HAVE_NEON_INTR)
void celt_pitch_xcorr_float_neon(const opus_val16 *_x, const opus_val16 *_y,
opus_val32 *xcorr, int len, int max_pitch);
-#if !defined(OPUS_HAVE_RTCD) || defined(OPUS_ARM_PRESUME_NEON_INTR)
-#define OVERRIDE_PITCH_XCORR (1)
+#endif
+
+# if defined(OPUS_HAVE_RTCD) && \
+ (defined(OPUS_ARM_MAY_HAVE_NEON_INTR) && !defined(OPUS_ARM_PRESUME_NEON_INTR))
+extern void
+(*const CELT_PITCH_XCORR_IMPL[OPUS_ARCHMASK+1])(const opus_val16 *,
+ const opus_val16 *, opus_val32 *, int, int);
+
+# define OVERRIDE_PITCH_XCORR (1)
+# define celt_pitch_xcorr(_x, _y, xcorr, len, max_pitch, arch) \
+ ((*CELT_PITCH_XCORR_IMPL[(arch)&OPUS_ARCHMASK])(_x, _y, \
+ xcorr, len, max_pitch))
+
+# elif defined(OPUS_ARM_PRESUME_NEON_INTR)
+
+# define OVERRIDE_PITCH_XCORR (1)
# define celt_pitch_xcorr(_x, _y, xcorr, len, max_pitch, arch) \
((void)(arch),celt_pitch_xcorr_float_neon(_x, _y, xcorr, len, max_pitch))
-#endif
-#endif
+
+# endif
#endif /* end !FIXED_POINT */
+
#endif
diff --git a/thirdparty/opus/celt/bands.c b/thirdparty/opus/celt/bands.c
index 25f229e267..87eaa6c031 100644
--- a/thirdparty/opus/celt/bands.c
+++ b/thirdparty/opus/celt/bands.c
@@ -414,7 +414,7 @@ static void stereo_merge(celt_norm * OPUS_RESTRICT X, celt_norm * OPUS_RESTRICT
/* Compensating for the mid normalization */
xp = MULT16_32_Q15(mid, xp);
/* mid and side are in Q15, not Q14 like X and Y */
- mid2 = SHR32(mid, 1);
+ mid2 = SHR16(mid, 1);
El = MULT16_16(mid2, mid2) + side - 2*xp;
Er = MULT16_16(mid2, mid2) + side + 2*xp;
if (Er < QCONST32(6e-4f, 28) || El < QCONST32(6e-4f, 28))
@@ -714,7 +714,7 @@ static void compute_theta(struct band_ctx *ctx, struct split_ctx *sctx,
if (qn!=1)
{
if (encode)
- itheta = (itheta*qn+8192)>>14;
+ itheta = (itheta*(opus_int32)qn+8192)>>14;
/* Entropy coding of the angle. We use a uniform pdf for the
time split, a step for stereo, and a triangular one for the rest. */
diff --git a/thirdparty/opus/celt/celt.h b/thirdparty/opus/celt/celt.h
index a423b95046..d1f7eb690d 100644
--- a/thirdparty/opus/celt/celt.h
+++ b/thirdparty/opus/celt/celt.h
@@ -209,7 +209,7 @@ void comb_filter_const_c(opus_val32 *y, opus_val32 *x, int T, int N,
#endif
#ifndef OVERRIDE_COMB_FILTER_CONST
-# define comb_filter_const(y, x, T, N, g10, g11, g12, arch) \
+# define comb_filter_const(y, x, T, N, g10, g11, g12, arch) \
((void)(arch),comb_filter_const_c(y, x, T, N, g10, g11, g12))
#endif
diff --git a/thirdparty/opus/celt/celt_decoder.c b/thirdparty/opus/celt/celt_decoder.c
index b688f2a4e3..b978bb34d1 100644
--- a/thirdparty/opus/celt/celt_decoder.c
+++ b/thirdparty/opus/celt/celt_decoder.c
@@ -82,6 +82,7 @@ struct OpusCustomDecoder {
int error;
int last_pitch_index;
int loss_count;
+ int skip_plc;
int postfilter_period;
int postfilter_period_old;
opus_val16 postfilter_gain;
@@ -164,8 +165,6 @@ OPUS_CUSTOM_NOSTATIC int opus_custom_decoder_init(CELTDecoder *st, const CELTMod
st->signalling = 1;
st->arch = opus_select_arch();
- st->loss_count = 0;
-
opus_custom_decoder_ctl(st, OPUS_RESET_STATE);
return OPUS_OK;
@@ -447,7 +446,7 @@ static void celt_decode_lost(CELTDecoder * OPUS_RESTRICT st, int N, int LM)
loss_count = st->loss_count;
start = st->start;
- noise_based = loss_count >= 5 || start != 0;
+ noise_based = loss_count >= 5 || start != 0 || st->skip_plc;
if (noise_based)
{
/* Noise-based PLC/CNG */
@@ -832,6 +831,10 @@ int celt_decode_with_ec(CELTDecoder * OPUS_RESTRICT st, const unsigned char *dat
return frame_size/st->downsample;
}
+ /* Check if there are at least two packets received consecutively before
+ * turning on the pitch-based PLC */
+ st->skip_plc = st->loss_count != 0;
+
if (dec == NULL)
{
ec_dec_init(&_dec,(unsigned char*)data,len);
@@ -1198,6 +1201,7 @@ int opus_custom_decoder_ctl(CELTDecoder * OPUS_RESTRICT st, int request, ...)
((char*)&st->DECODER_RESET_START - (char*)st));
for (i=0;i<2*st->mode->nbEBands;i++)
oldLogE[i]=oldLogE2[i]=-QCONST16(28.f,DB_SHIFT);
+ st->skip_plc = 1;
}
break;
case OPUS_GET_PITCH_REQUEST:
diff --git a/thirdparty/opus/celt/celt_encoder.c b/thirdparty/opus/celt/celt_encoder.c
index 41fbfd49c8..3ee7a4d3f7 100644
--- a/thirdparty/opus/celt/celt_encoder.c
+++ b/thirdparty/opus/celt/celt_encoder.c
@@ -1175,10 +1175,10 @@ static int run_prefilter(CELTEncoder *st, celt_sig *in, celt_sig *prefilter_mem,
if (N>COMBFILTER_MAXPERIOD)
{
- OPUS_MOVE(prefilter_mem+c*COMBFILTER_MAXPERIOD, pre[c]+N, COMBFILTER_MAXPERIOD);
+ OPUS_COPY(prefilter_mem+c*COMBFILTER_MAXPERIOD, pre[c]+N, COMBFILTER_MAXPERIOD);
} else {
OPUS_MOVE(prefilter_mem+c*COMBFILTER_MAXPERIOD, prefilter_mem+c*COMBFILTER_MAXPERIOD+N, COMBFILTER_MAXPERIOD-N);
- OPUS_MOVE(prefilter_mem+c*COMBFILTER_MAXPERIOD+COMBFILTER_MAXPERIOD-N, pre[c]+COMBFILTER_MAXPERIOD, N);
+ OPUS_COPY(prefilter_mem+c*COMBFILTER_MAXPERIOD+COMBFILTER_MAXPERIOD-N, pre[c]+COMBFILTER_MAXPERIOD, N);
}
} while (++c<CC);
@@ -1281,12 +1281,15 @@ static int compute_vbr(const CELTMode *mode, AnalysisInfo *analysis, opus_int32
if ((!has_surround_mask||lfe) && (constrained_vbr || bitrate<64000))
{
- opus_val16 rate_factor;
+ opus_val16 rate_factor = Q15ONE;
+ if (bitrate < 64000)
+ {
#ifdef FIXED_POINT
- rate_factor = MAX16(0,(bitrate-32000));
+ rate_factor = MAX16(0,(bitrate-32000));
#else
- rate_factor = MAX16(0,(1.f/32768)*(bitrate-32000));
+ rate_factor = MAX16(0,(1.f/32768)*(bitrate-32000));
#endif
+ }
if (constrained_vbr)
rate_factor = MIN16(rate_factor, QCONST16(0.67f, 15));
target = base_target + (opus_int32)MULT16_32_Q15(rate_factor, target-base_target);
diff --git a/thirdparty/opus/celt/celt_lpc.c b/thirdparty/opus/celt/celt_lpc.c
index f02145af0d..b410a21c5f 100644
--- a/thirdparty/opus/celt/celt_lpc.c
+++ b/thirdparty/opus/celt/celt_lpc.c
@@ -49,8 +49,7 @@ int p
float *lpc = _lpc;
#endif
- for (i = 0; i < p; i++)
- lpc[i] = 0;
+ OPUS_CLEAR(lpc, p);
if (ac[0] != 0)
{
for (i = 0; i < p; i++) {
diff --git a/thirdparty/opus/celt/cwrs.c b/thirdparty/opus/celt/cwrs.c
index 2fa9f89cd6..9722f0ac86 100644
--- a/thirdparty/opus/celt/cwrs.c
+++ b/thirdparty/opus/celt/cwrs.c
@@ -74,7 +74,7 @@ int log2_frac(opus_uint32 val, int frac)
/*Although derived separately, the pulse vector coding scheme is equivalent to
a Pyramid Vector Quantizer \cite{Fis86}.
Some additional notes about an early version appear at
- http://people.xiph.org/~tterribe/notes/cwrs.html, but the codebook ordering
+ https://people.xiph.org/~tterribe/notes/cwrs.html, but the codebook ordering
and the definitions of some terms have evolved since that was written.
The conversion from a pulse vector to an integer index (encoding) and back
diff --git a/thirdparty/opus/celt/fixed_generic.h b/thirdparty/opus/celt/fixed_generic.h
index ac67d37ce8..1cfd6d6989 100644
--- a/thirdparty/opus/celt/fixed_generic.h
+++ b/thirdparty/opus/celt/fixed_generic.h
@@ -37,16 +37,32 @@
#define MULT16_16SU(a,b) ((opus_val32)(opus_val16)(a)*(opus_val32)(opus_uint16)(b))
/** 16x32 multiplication, followed by a 16-bit shift right. Results fits in 32 bits */
+#if OPUS_FAST_INT64
+#define MULT16_32_Q16(a,b) ((opus_val32)SHR((opus_int64)((opus_val16)(a))*(b),16))
+#else
#define MULT16_32_Q16(a,b) ADD32(MULT16_16((a),SHR((b),16)), SHR(MULT16_16SU((a),((b)&0x0000ffff)),16))
+#endif
/** 16x32 multiplication, followed by a 16-bit shift right (round-to-nearest). Results fits in 32 bits */
+#if OPUS_FAST_INT64
+#define MULT16_32_P16(a,b) ((opus_val32)PSHR((opus_int64)((opus_val16)(a))*(b),16))
+#else
#define MULT16_32_P16(a,b) ADD32(MULT16_16((a),SHR((b),16)), PSHR(MULT16_16SU((a),((b)&0x0000ffff)),16))
+#endif
/** 16x32 multiplication, followed by a 15-bit shift right. Results fits in 32 bits */
+#if OPUS_FAST_INT64
+#define MULT16_32_Q15(a,b) ((opus_val32)SHR((opus_int64)((opus_val16)(a))*(b),15))
+#else
#define MULT16_32_Q15(a,b) ADD32(SHL(MULT16_16((a),SHR((b),16)),1), SHR(MULT16_16SU((a),((b)&0x0000ffff)),15))
+#endif
/** 32x32 multiplication, followed by a 31-bit shift right. Results fits in 32 bits */
+#if OPUS_FAST_INT64
+#define MULT32_32_Q31(a,b) ((opus_val32)SHR((opus_int64)(a)*(opus_int64)(b),31))
+#else
#define MULT32_32_Q31(a,b) ADD32(ADD32(SHL(MULT16_16(SHR((a),16),SHR((b),16)),1), SHR(MULT16_16SU(SHR((a),16),((b)&0x0000ffff)),15)), SHR(MULT16_16SU(SHR((b),16),((a)&0x0000ffff)),15))
+#endif
/** Compile-time conversion of float constant to 16-bit value */
#define QCONST16(x,bits) ((opus_val16)(.5+(x)*(((opus_val32)1)<<(bits))))
diff --git a/thirdparty/opus/celt/kiss_fft.c b/thirdparty/opus/celt/kiss_fft.c
index 4ed37d2bb7..1f8fd05321 100644
--- a/thirdparty/opus/celt/kiss_fft.c
+++ b/thirdparty/opus/celt/kiss_fft.c
@@ -191,7 +191,7 @@ static void kf_bfly3(
kiss_fft_cpx * Fout_beg = Fout;
#ifdef FIXED_POINT
- epi3.r = -16384;
+ /*epi3.r = -16384;*/ /* Unused */
epi3.i = -28378;
#else
epi3 = st->twiddles[fstride*m];
diff --git a/thirdparty/opus/celt/mathops.c b/thirdparty/opus/celt/mathops.c
index 3f8c5dcc0e..21a01f52e4 100644
--- a/thirdparty/opus/celt/mathops.c
+++ b/thirdparty/opus/celt/mathops.c
@@ -164,7 +164,7 @@ opus_val16 celt_cos_norm(opus_val32 x)
{
return _celt_cos_pi_2(EXTRACT16(x));
} else {
- return NEG32(_celt_cos_pi_2(EXTRACT16(65536-x)));
+ return NEG16(_celt_cos_pi_2(EXTRACT16(65536-x)));
}
} else {
if (x&0x0000ffff)
diff --git a/thirdparty/opus/celt/pitch.c b/thirdparty/opus/celt/pitch.c
index 1d89cb0342..bf46e7d562 100644
--- a/thirdparty/opus/celt/pitch.c
+++ b/thirdparty/opus/celt/pitch.c
@@ -412,6 +412,41 @@ void pitch_search(const opus_val16 * OPUS_RESTRICT x_lp, opus_val16 * OPUS_RESTR
RESTORE_STACK;
}
+#ifdef FIXED_POINT
+static opus_val16 compute_pitch_gain(opus_val32 xy, opus_val32 xx, opus_val32 yy)
+{
+ opus_val32 x2y2;
+ int sx, sy, shift;
+ opus_val32 g;
+ opus_val16 den;
+ if (xy == 0 || xx == 0 || yy == 0)
+ return 0;
+ sx = celt_ilog2(xx)-14;
+ sy = celt_ilog2(yy)-14;
+ shift = sx + sy;
+ x2y2 = MULT16_16_Q14(VSHR32(xx, sx), VSHR32(yy, sy));
+ if (shift & 1) {
+ if (x2y2 < 32768)
+ {
+ x2y2 <<= 1;
+ shift--;
+ } else {
+ x2y2 >>= 1;
+ shift++;
+ }
+ }
+ den = celt_rsqrt_norm(x2y2);
+ g = MULT16_32_Q15(den, xy);
+ g = VSHR32(g, (shift>>1)-1);
+ return EXTRACT16(MIN32(g, Q15ONE));
+}
+#else
+static opus_val16 compute_pitch_gain(opus_val32 xy, opus_val32 xx, opus_val32 yy)
+{
+ return xy/celt_sqrt(1+xx*yy);
+}
+#endif
+
static const int second_check[16] = {0, 0, 3, 2, 3, 2, 5, 2, 3, 2, 3, 2, 5, 2, 3, 2};
opus_val16 remove_doubling(opus_val16 *x, int maxperiod, int minperiod,
int N, int *T0_, int prev_period, opus_val16 prev_gain, int arch)
@@ -450,18 +485,7 @@ opus_val16 remove_doubling(opus_val16 *x, int maxperiod, int minperiod,
yy = yy_lookup[T0];
best_xy = xy;
best_yy = yy;
-#ifdef FIXED_POINT
- {
- opus_val32 x2y2;
- int sh, t;
- x2y2 = 1+HALF32(MULT32_32_Q31(xx,yy));
- sh = celt_ilog2(x2y2)>>1;
- t = VSHR32(x2y2, 2*(sh-7));
- g = g0 = VSHR32(MULT16_32_Q15(celt_rsqrt_norm(t), xy),sh+1);
- }
-#else
- g = g0 = xy/celt_sqrt(1+xx*yy);
-#endif
+ g = g0 = compute_pitch_gain(xy, xx, yy);
/* Look for any pitch at T/k */
for (k=2;k<=15;k++)
{
@@ -484,24 +508,13 @@ opus_val16 remove_doubling(opus_val16 *x, int maxperiod, int minperiod,
T1b = celt_udiv(2*second_check[k]*T0+k, 2*k);
}
dual_inner_prod(x, &x[-T1], &x[-T1b], N, &xy, &xy2, arch);
- xy += xy2;
- yy = yy_lookup[T1] + yy_lookup[T1b];
-#ifdef FIXED_POINT
- {
- opus_val32 x2y2;
- int sh, t;
- x2y2 = 1+MULT32_32_Q31(xx,yy);
- sh = celt_ilog2(x2y2)>>1;
- t = VSHR32(x2y2, 2*(sh-7));
- g1 = VSHR32(MULT16_32_Q15(celt_rsqrt_norm(t), xy),sh+1);
- }
-#else
- g1 = xy/celt_sqrt(1+2.f*xx*1.f*yy);
-#endif
+ xy = HALF32(xy + xy2);
+ yy = HALF32(yy_lookup[T1] + yy_lookup[T1b]);
+ g1 = compute_pitch_gain(xy, xx, yy);
if (abs(T1-prev_period)<=1)
cont = prev_gain;
else if (abs(T1-prev_period)<=2 && 5*k*k < T0)
- cont = HALF32(prev_gain);
+ cont = HALF16(prev_gain);
else
cont = 0;
thresh = MAX16(QCONST16(.3f,15), MULT16_16_Q15(QCONST16(.7f,15),g0)-cont);
diff --git a/thirdparty/opus/celt/pitch.h b/thirdparty/opus/celt/pitch.h
index 65a77a6ecc..d3503532a0 100644
--- a/thirdparty/opus/celt/pitch.h
+++ b/thirdparty/opus/celt/pitch.h
@@ -187,25 +187,6 @@ celt_pitch_xcorr_c(const opus_val16 *_x, const opus_val16 *_y,
opus_val32 *xcorr, int len, int max_pitch);
#if !defined(OVERRIDE_PITCH_XCORR)
-/*Is run-time CPU detection enabled on this platform?*/
-# if defined(OPUS_HAVE_RTCD) && (defined(OPUS_ARM_ASM) \
- || (defined(OPUS_ARM_MAY_HAVE_NEON_INTR) \
- && !defined(OPUS_ARM_PRESUME_NEON_INTR)))
-extern
-# if defined(FIXED_POINT)
-opus_val32
-# else
-void
-# endif
-(*const CELT_PITCH_XCORR_IMPL[OPUS_ARCHMASK+1])(const opus_val16 *,
- const opus_val16 *, opus_val32 *, int, int);
-
-# define OVERRIDE_PITCH_XCORR
-# define celt_pitch_xcorr(_x, _y, xcorr, len, max_pitch, arch) \
- ((*CELT_PITCH_XCORR_IMPL[(arch)&OPUS_ARCHMASK])(_x, _y, \
- xcorr, len, max_pitch))
-# else
-
#ifdef FIXED_POINT
opus_val32
#else
@@ -214,7 +195,6 @@ void
celt_pitch_xcorr(const opus_val16 *_x, const opus_val16 *_y,
opus_val32 *xcorr, int len, int max_pitch, int arch);
-# endif
#endif
#endif
diff --git a/thirdparty/opus/celt/rate.c b/thirdparty/opus/celt/rate.c
index b28d8feccd..7dfa5be8a6 100644
--- a/thirdparty/opus/celt/rate.c
+++ b/thirdparty/opus/celt/rate.c
@@ -296,7 +296,7 @@ static OPUS_INLINE int interp_bits2pulses(const CELTMode *m, int start, int end,
done = 0;
for (j=end;j-->start;)
{
- int tmp = bits1[j] + (lo*bits2[j]>>ALLOC_STEPS);
+ int tmp = bits1[j] + ((opus_int32)lo*bits2[j]>>ALLOC_STEPS);
if (tmp < thresh[j] && !done)
{
if (tmp >= alloc_floor)
diff --git a/thirdparty/opus/celt/vq.c b/thirdparty/opus/celt/vq.c
index f358396065..d29f38fd8e 100644
--- a/thirdparty/opus/celt/vq.c
+++ b/thirdparty/opus/celt/vq.c
@@ -271,7 +271,7 @@ unsigned alg_quant(celt_norm *X, int N, int K, int spread, int B, ec_enc *enc
best_id = 0;
/* The squared magnitude term gets added anyway, so we might as well
add it outside the loop */
- yy = ADD32(yy, 1);
+ yy = ADD16(yy, 1);
j=0;
do {
opus_val16 Rxy, Ryy;
diff --git a/thirdparty/opus/celt/x86/pitch_sse.h b/thirdparty/opus/celt/x86/pitch_sse.h
index d4cbeb8b9c..e5f87ab51a 100644
--- a/thirdparty/opus/celt/x86/pitch_sse.h
+++ b/thirdparty/opus/celt/x86/pitch_sse.h
@@ -102,21 +102,21 @@ opus_val32 celt_inner_prod_sse(
#if defined(OPUS_X86_PRESUME_SSE4_1) && defined(FIXED_POINT)
#define OVERRIDE_CELT_INNER_PROD
#define celt_inner_prod(x, y, N, arch) \
- ((void)arch, celt_inner_prod_sse4_1(x, y, N))
+ ((void)arch, celt_inner_prod_sse4_1(x, y, N))
#elif defined(OPUS_X86_PRESUME_SSE2) && defined(FIXED_POINT) && !defined(OPUS_X86_MAY_HAVE_SSE4_1)
#define OVERRIDE_CELT_INNER_PROD
#define celt_inner_prod(x, y, N, arch) \
- ((void)arch, celt_inner_prod_sse2(x, y, N))
+ ((void)arch, celt_inner_prod_sse2(x, y, N))
#elif defined(OPUS_X86_PRESUME_SSE) && !defined(FIXED_POINT)
#define OVERRIDE_CELT_INNER_PROD
#define celt_inner_prod(x, y, N, arch) \
- ((void)arch, celt_inner_prod_sse(x, y, N))
+ ((void)arch, celt_inner_prod_sse(x, y, N))
#elif ((defined(OPUS_X86_MAY_HAVE_SSE4_1) || defined(OPUS_X86_MAY_HAVE_SSE2)) && defined(FIXED_POINT)) || \
- (defined(OPUS_X86_MAY_HAVE_SSE) && !defined(FIXED_POINT))
+ (defined(OPUS_X86_MAY_HAVE_SSE) && !defined(FIXED_POINT))
extern opus_val32 (*const CELT_INNER_PROD_IMPL[OPUS_ARCHMASK + 1])(
const opus_val16 *x,
@@ -138,19 +138,19 @@ extern opus_val32 (*const CELT_INNER_PROD_IMPL[OPUS_ARCHMASK + 1])(
#undef comb_filter_const
void dual_inner_prod_sse(const opus_val16 *x,
- const opus_val16 *y01,
- const opus_val16 *y02,
- int N,
- opus_val32 *xy1,
- opus_val32 *xy2);
+ const opus_val16 *y01,
+ const opus_val16 *y02,
+ int N,
+ opus_val32 *xy1,
+ opus_val32 *xy2);
void comb_filter_const_sse(opus_val32 *y,
- opus_val32 *x,
- int T,
- int N,
- opus_val16 g10,
- opus_val16 g11,
- opus_val16 g12);
+ opus_val32 *x,
+ int T,
+ int N,
+ opus_val16 g10,
+ opus_val16 g11,
+ opus_val16 g12);
#if defined(OPUS_X86_PRESUME_SSE)
@@ -169,7 +169,7 @@ extern void (*const DUAL_INNER_PROD_IMPL[OPUS_ARCHMASK + 1])(
opus_val32 *xy1,
opus_val32 *xy2);
-#define dual_inner_prod(x, y01, y02, N, xy1, xy2, arch) \
+#define dual_inner_prod(x, y01, y02, N, xy1, xy2, arch) \
((*DUAL_INNER_PROD_IMPL[(arch) & OPUS_ARCHMASK])(x, y01, y02, N, xy1, xy2))
extern void (*const COMB_FILTER_CONST_IMPL[OPUS_ARCHMASK + 1])(
@@ -181,7 +181,7 @@ extern void (*const COMB_FILTER_CONST_IMPL[OPUS_ARCHMASK + 1])(
opus_val16 g11,
opus_val16 g12);
-#define comb_filter_const(y, x, T, N, g10, g11, g12, arch) \
+#define comb_filter_const(y, x, T, N, g10, g11, g12, arch) \
((*COMB_FILTER_CONST_IMPL[(arch) & OPUS_ARCHMASK])(y, x, T, N, g10, g11, g12))
#define NON_STATIC_COMB_FILTER_CONST_C
diff --git a/thirdparty/opus/celt/x86/x86_celt_map.c b/thirdparty/opus/celt/x86/x86_celt_map.c
index 8e5e449275..47ba41b9ee 100644
--- a/thirdparty/opus/celt/x86/x86_celt_map.c
+++ b/thirdparty/opus/celt/x86/x86_celt_map.c
@@ -72,7 +72,7 @@ void (*const XCORR_KERNEL_IMPL[OPUS_ARCHMASK + 1])(
#endif
#if (defined(OPUS_X86_MAY_HAVE_SSE4_1) && !defined(OPUS_X86_PRESUME_SSE4_1)) || \
- (!defined(OPUS_X86_MAY_HAVE_SSE_4_1) && defined(OPUS_X86_MAY_HAVE_SSE2) && !defined(OPUS_X86_PRESUME_SSE2))
+ (!defined(OPUS_X86_MAY_HAVE_SSE_4_1) && defined(OPUS_X86_MAY_HAVE_SSE2) && !defined(OPUS_X86_PRESUME_SSE2))
opus_val32 (*const CELT_INNER_PROD_IMPL[OPUS_ARCHMASK + 1])(
const opus_val16 *x,
diff --git a/thirdparty/opus/celt/x86/x86cpu.c b/thirdparty/opus/celt/x86/x86cpu.c
index 555a576b8a..080eb25e41 100644
--- a/thirdparty/opus/celt/x86/x86cpu.c
+++ b/thirdparty/opus/celt/x86/x86cpu.c
@@ -46,7 +46,7 @@
#include <intrin.h>
static _inline void cpuid(unsigned int CPUInfo[4], unsigned int InfoType)
{
- __cpuid((int*)CPUInfo, InfoType);
+ __cpuid((int*)CPUInfo, InfoType);
}
#else
generated by cgit v1.2.3 (git 2.25.1) at 2024-11-23 05:30:30 +0000