pulseaudio/src/pulsecore/sconv-s16le.c

442 lines
9.9 KiB
C
Raw Normal View History

/***
This file is part of PulseAudio.
Copyright 2004-2006 Lennart Poettering
PulseAudio is free software; you can redistribute it and/or modify
it under the terms of the GNU Lesser General Public License as published
by the Free Software Foundation; either version 2.1 of the License,
or (at your option) any later version.
PulseAudio is distributed in the hope that it will be useful, but
WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
General Public License for more details.
You should have received a copy of the GNU Lesser General Public License
along with PulseAudio; if not, write to the Free Software
Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307
USA.
***/
#ifdef HAVE_CONFIG_H
#include <config.h>
#endif
2009-01-16 03:15:39 +01:00
/* Despite the name of this file we implement S32 and S24 handling here, too. */
#include <inttypes.h>
#include <stdio.h>
2011-06-13 15:04:33 +02:00
#include <math.h>
#include <pulsecore/sconv.h>
#include <pulsecore/macro.h>
#include <pulsecore/endianmacros.h>
#include "sconv-s16le.h"
#ifndef INT16_FROM
#define INT16_FROM PA_INT16_FROM_LE
#endif
#ifndef UINT16_FROM
#define UINT16_FROM PA_UINT16_FROM_LE
#endif
#ifndef INT16_TO
#define INT16_TO PA_INT16_TO_LE
#endif
#ifndef UINT16_TO
#define UINT16_TO PA_UINT16_TO_LE
#endif
#ifndef INT32_FROM
#define INT32_FROM PA_INT32_FROM_LE
#endif
#ifndef UINT32_FROM
#define UINT32_FROM PA_UINT32_FROM_LE
#endif
#ifndef INT32_TO
#define INT32_TO PA_INT32_TO_LE
#endif
#ifndef UINT32_TO
#define UINT32_TO PA_UINT32_TO_LE
#endif
2009-01-16 03:15:39 +01:00
#ifndef READ24
#define READ24 PA_READ24LE
#endif
#ifndef WRITE24
#define WRITE24 PA_WRITE24LE
#endif
#ifndef SWAP_WORDS
#ifdef WORDS_BIGENDIAN
#define SWAP_WORDS 1
#else
#define SWAP_WORDS 0
#endif
#endif
void pa_sconv_s16le_to_float32ne(unsigned n, const int16_t *a, float *b) {
pa_assert(a);
pa_assert(b);
#if SWAP_WORDS == 1
for (; n > 0; n--) {
int16_t s = *(a++);
*(b++) = INT16_FROM(s) * (1.0f / (1 << 15));
}
#else
for (; n > 0; n--)
*(b++) = *(a++) * (1.0f / (1 << 15));
#endif
}
void pa_sconv_s32le_to_float32ne(unsigned n, const int32_t *a, float *b) {
pa_assert(a);
pa_assert(b);
#if SWAP_WORDS == 1
for (; n > 0; n--) {
int32_t s = *(a++);
*(b++) = INT32_FROM(s) * (1.0f / (1U << 31));
}
#else
for (; n > 0; n--)
*(b++) = *(a++) * (1.0f / (1U << 31));
#endif
}
void pa_sconv_s16le_from_float32ne(unsigned n, const float *a, int16_t *b) {
pa_assert(a);
pa_assert(b);
#if SWAP_WORDS == 1
for (; n > 0; n--) {
int16_t s;
float v = *(a++) * (1 << 15);
s = (int16_t) PA_CLAMP_UNLIKELY(lrintf(v), -0x8000, 0x7FFF);
*(b++) = INT16_TO(s);
}
#else
for (; n > 0; n--) {
float v = *(a++) * (1 << 15);
*(b++) = (int16_t) PA_CLAMP_UNLIKELY(lrintf(v), -0x8000, 0x7FFF);
}
#endif
}
void pa_sconv_s32le_from_float32ne(unsigned n, const float *a, int32_t *b) {
pa_assert(a);
pa_assert(b);
#if SWAP_WORDS == 1
for (; n > 0; n--) {
int32_t s;
float v = *(a++) * (1U << 31);
s = (int32_t) PA_CLAMP_UNLIKELY(llrintf(v), -0x80000000LL, 0x7FFFFFFFLL);
*(b++) = INT32_TO(s);
}
#else
for (; n > 0; n--) {
float v = *(a++) * (1U << 31);
*(b++) = (int32_t) PA_CLAMP_UNLIKELY(llrintf(v), -0x80000000LL, 0x7FFFFFFFLL);
}
#endif
}
void pa_sconv_s16le_to_float32re(unsigned n, const int16_t *a, float *b) {
pa_assert(a);
pa_assert(b);
for (; n > 0; n--) {
int16_t s = *(a++);
float k = INT16_FROM(s) * (1.0f / (1 << 15));
endianmacros: Replace borked PA_FLOAT32_SWAP() with PA_READ_FLOAT32RE() / PA_WRITE_FLOAT32RE() building PA with -O0 leads to test failure in mix-test on i386 issue reported by Felipe, see http://lists.freedesktop.org/archives/pulseaudio-discuss/2014-August/021406.html the problem is the value 0xbeffbd7f: when byte-swapped it becomes 0x7fbdffbe and according to IEEE-754 represents a signalling NaN (starting with s111 1111 10, see http://en.wikipedia.org/wiki/NaN) when this value is assigned to a floating point register, it becomes 0x7ffdffbe, representing a quiet NaN (starting with s111 1111 11) -- a signalling NaN is turned into a quiet NaN! so PA_FLOAT32_SWAP(PA_FLOAT32_SWAP(x)) != x for certain values, uhuh! the following test code can be used; due to volatile, it will always demonstrate the issue; without volatile, it depends on the optimization level (i386, 32-bit, gcc 4.9): // snip static inline float PA_FLOAT32_SWAP(float x) { union { float f; uint32_t u; } t; t.f = x; t.u = bswap_32(t.u); return t.f; } int main() { unsigned x = 0xbeffbd7f; volatile float f = PA_FLOAT32_SWAP(*(float *)&x); printf("%08x %08x %08x %f\n", 0xbeffbd7f, *(unsigned *)&f, bswap_32(*(unsigned *)&f), f); } // snip the problem goes away with optimization when no temporary floating point registers are used the proposed solution is to avoid passing swapped floating point data in a float; this is done with new functions PA_READ_FLOAT32RE() and PA_WRITE_FLOAT32RE() which use uint32_t to dereference a pointer and byte-swap the data, hence no temporary float variable is used also delete PA_FLOAT32_TO_LE()/_BE(), not used Signed-off-by: Peter Meerwald <pmeerw@pmeerw.net> Reported-by: Felipe Sateler <fsateler@debian.org>
2014-09-02 23:53:09 +02:00
PA_WRITE_FLOAT32RE(b++, k);
}
}
void pa_sconv_s32le_to_float32re(unsigned n, const int32_t *a, float *b) {
pa_assert(a);
pa_assert(b);
for (; n > 0; n--) {
int32_t s = *(a++);
float k = INT32_FROM(s) * (1.0f / (1U << 31));
endianmacros: Replace borked PA_FLOAT32_SWAP() with PA_READ_FLOAT32RE() / PA_WRITE_FLOAT32RE() building PA with -O0 leads to test failure in mix-test on i386 issue reported by Felipe, see http://lists.freedesktop.org/archives/pulseaudio-discuss/2014-August/021406.html the problem is the value 0xbeffbd7f: when byte-swapped it becomes 0x7fbdffbe and according to IEEE-754 represents a signalling NaN (starting with s111 1111 10, see http://en.wikipedia.org/wiki/NaN) when this value is assigned to a floating point register, it becomes 0x7ffdffbe, representing a quiet NaN (starting with s111 1111 11) -- a signalling NaN is turned into a quiet NaN! so PA_FLOAT32_SWAP(PA_FLOAT32_SWAP(x)) != x for certain values, uhuh! the following test code can be used; due to volatile, it will always demonstrate the issue; without volatile, it depends on the optimization level (i386, 32-bit, gcc 4.9): // snip static inline float PA_FLOAT32_SWAP(float x) { union { float f; uint32_t u; } t; t.f = x; t.u = bswap_32(t.u); return t.f; } int main() { unsigned x = 0xbeffbd7f; volatile float f = PA_FLOAT32_SWAP(*(float *)&x); printf("%08x %08x %08x %f\n", 0xbeffbd7f, *(unsigned *)&f, bswap_32(*(unsigned *)&f), f); } // snip the problem goes away with optimization when no temporary floating point registers are used the proposed solution is to avoid passing swapped floating point data in a float; this is done with new functions PA_READ_FLOAT32RE() and PA_WRITE_FLOAT32RE() which use uint32_t to dereference a pointer and byte-swap the data, hence no temporary float variable is used also delete PA_FLOAT32_TO_LE()/_BE(), not used Signed-off-by: Peter Meerwald <pmeerw@pmeerw.net> Reported-by: Felipe Sateler <fsateler@debian.org>
2014-09-02 23:53:09 +02:00
PA_WRITE_FLOAT32RE(b++, k);
}
}
void pa_sconv_s16le_from_float32re(unsigned n, const float *a, int16_t *b) {
pa_assert(a);
pa_assert(b);
for (; n > 0; n--) {
int16_t s;
endianmacros: Replace borked PA_FLOAT32_SWAP() with PA_READ_FLOAT32RE() / PA_WRITE_FLOAT32RE() building PA with -O0 leads to test failure in mix-test on i386 issue reported by Felipe, see http://lists.freedesktop.org/archives/pulseaudio-discuss/2014-August/021406.html the problem is the value 0xbeffbd7f: when byte-swapped it becomes 0x7fbdffbe and according to IEEE-754 represents a signalling NaN (starting with s111 1111 10, see http://en.wikipedia.org/wiki/NaN) when this value is assigned to a floating point register, it becomes 0x7ffdffbe, representing a quiet NaN (starting with s111 1111 11) -- a signalling NaN is turned into a quiet NaN! so PA_FLOAT32_SWAP(PA_FLOAT32_SWAP(x)) != x for certain values, uhuh! the following test code can be used; due to volatile, it will always demonstrate the issue; without volatile, it depends on the optimization level (i386, 32-bit, gcc 4.9): // snip static inline float PA_FLOAT32_SWAP(float x) { union { float f; uint32_t u; } t; t.f = x; t.u = bswap_32(t.u); return t.f; } int main() { unsigned x = 0xbeffbd7f; volatile float f = PA_FLOAT32_SWAP(*(float *)&x); printf("%08x %08x %08x %f\n", 0xbeffbd7f, *(unsigned *)&f, bswap_32(*(unsigned *)&f), f); } // snip the problem goes away with optimization when no temporary floating point registers are used the proposed solution is to avoid passing swapped floating point data in a float; this is done with new functions PA_READ_FLOAT32RE() and PA_WRITE_FLOAT32RE() which use uint32_t to dereference a pointer and byte-swap the data, hence no temporary float variable is used also delete PA_FLOAT32_TO_LE()/_BE(), not used Signed-off-by: Peter Meerwald <pmeerw@pmeerw.net> Reported-by: Felipe Sateler <fsateler@debian.org>
2014-09-02 23:53:09 +02:00
float v = PA_READ_FLOAT32RE(a++) * (1 << 15);
s = (int16_t) PA_CLAMP_UNLIKELY(lrintf(v), -0x8000, 0x7FFF);
*(b++) = INT16_TO(s);
}
}
void pa_sconv_s32le_from_float32re(unsigned n, const float *a, int32_t *b) {
pa_assert(a);
pa_assert(b);
for (; n > 0; n--) {
int32_t s;
endianmacros: Replace borked PA_FLOAT32_SWAP() with PA_READ_FLOAT32RE() / PA_WRITE_FLOAT32RE() building PA with -O0 leads to test failure in mix-test on i386 issue reported by Felipe, see http://lists.freedesktop.org/archives/pulseaudio-discuss/2014-August/021406.html the problem is the value 0xbeffbd7f: when byte-swapped it becomes 0x7fbdffbe and according to IEEE-754 represents a signalling NaN (starting with s111 1111 10, see http://en.wikipedia.org/wiki/NaN) when this value is assigned to a floating point register, it becomes 0x7ffdffbe, representing a quiet NaN (starting with s111 1111 11) -- a signalling NaN is turned into a quiet NaN! so PA_FLOAT32_SWAP(PA_FLOAT32_SWAP(x)) != x for certain values, uhuh! the following test code can be used; due to volatile, it will always demonstrate the issue; without volatile, it depends on the optimization level (i386, 32-bit, gcc 4.9): // snip static inline float PA_FLOAT32_SWAP(float x) { union { float f; uint32_t u; } t; t.f = x; t.u = bswap_32(t.u); return t.f; } int main() { unsigned x = 0xbeffbd7f; volatile float f = PA_FLOAT32_SWAP(*(float *)&x); printf("%08x %08x %08x %f\n", 0xbeffbd7f, *(unsigned *)&f, bswap_32(*(unsigned *)&f), f); } // snip the problem goes away with optimization when no temporary floating point registers are used the proposed solution is to avoid passing swapped floating point data in a float; this is done with new functions PA_READ_FLOAT32RE() and PA_WRITE_FLOAT32RE() which use uint32_t to dereference a pointer and byte-swap the data, hence no temporary float variable is used also delete PA_FLOAT32_TO_LE()/_BE(), not used Signed-off-by: Peter Meerwald <pmeerw@pmeerw.net> Reported-by: Felipe Sateler <fsateler@debian.org>
2014-09-02 23:53:09 +02:00
float v = PA_READ_FLOAT32RE(a++) * (1U << 31);
s = (int32_t) PA_CLAMP_UNLIKELY(llrintf(v), -0x80000000LL, 0x7FFFFFFFLL);
*(b++) = INT32_TO(s);
}
}
void pa_sconv_s32le_to_s16ne(unsigned n, const int32_t*a, int16_t *b) {
pa_assert(a);
pa_assert(b);
for (; n > 0; n--) {
*b = (int16_t) (INT32_FROM(*a) >> 16);
a++;
b++;
}
}
void pa_sconv_s32le_to_s16re(unsigned n, const int32_t*a, int16_t *b) {
pa_assert(a);
pa_assert(b);
for (; n > 0; n--) {
int16_t s = (int16_t) (INT32_FROM(*a) >> 16);
*b = PA_INT16_SWAP(s);
a++;
b++;
}
}
void pa_sconv_s32le_from_s16ne(unsigned n, const int16_t *a, int32_t *b) {
pa_assert(a);
pa_assert(b);
for (; n > 0; n--) {
*b = INT32_TO(((int32_t) *a) << 16);
a++;
b++;
}
}
void pa_sconv_s32le_from_s16re(unsigned n, const int16_t *a, int32_t *b) {
pa_assert(a);
pa_assert(b);
for (; n > 0; n--) {
int32_t s = ((int32_t) PA_INT16_SWAP(*a)) << 16;
*b = INT32_TO(s);
a++;
b++;
}
}
2009-01-16 03:15:39 +01:00
void pa_sconv_s24le_to_s16ne(unsigned n, const uint8_t *a, int16_t *b) {
pa_assert(a);
pa_assert(b);
for (; n > 0; n--) {
*b = (int16_t) (READ24(a) >> 8);
a += 3;
b++;
}
}
void pa_sconv_s24le_from_s16ne(unsigned n, const int16_t *a, uint8_t *b) {
pa_assert(a);
pa_assert(b);
for (; n > 0; n--) {
WRITE24(b, ((uint32_t) *a) << 8);
a++;
b += 3;
}
}
void pa_sconv_s24le_to_s16re(unsigned n, const uint8_t *a, int16_t *b) {
pa_assert(a);
pa_assert(b);
for (; n > 0; n--) {
int16_t s = (int16_t) (READ24(a) >> 8);
*b = PA_INT16_SWAP(s);
a += 3;
b++;
}
}
void pa_sconv_s24le_from_s16re(unsigned n, const int16_t *a, uint8_t *b) {
pa_assert(a);
pa_assert(b);
for (; n > 0; n--) {
uint32_t s = ((uint32_t) PA_INT16_SWAP(*a)) << 8;
WRITE24(b, s);
a++;
b += 3;
}
}
void pa_sconv_s24le_to_float32ne(unsigned n, const uint8_t *a, float *b) {
pa_assert(a);
pa_assert(b);
for (; n > 0; n--) {
int32_t s = READ24(a) << 8;
*b = s * (1.0f / (1U << 31));
2009-01-16 03:15:39 +01:00
a += 3;
b++;
2009-01-16 03:15:39 +01:00
}
}
void pa_sconv_s24le_from_float32ne(unsigned n, const float *a, uint8_t *b) {
pa_assert(a);
pa_assert(b);
for (; n > 0; n--) {
int32_t s;
float v = *a * (1U << 31);
s = (int32_t) PA_CLAMP_UNLIKELY(llrint(v), -0x80000000LL, 0x7FFFFFFFLL);
2009-01-16 03:15:39 +01:00
WRITE24(b, ((uint32_t) s) >> 8);
a++;
b += 3;
2009-01-16 03:15:39 +01:00
}
}
void pa_sconv_s24le_to_float32re(unsigned n, const uint8_t *a, float *b) {
pa_assert(a);
pa_assert(b);
for (; n > 0; n--) {
int32_t s = READ24(a) << 8;
float k = s * (1.0f / (1U << 31));
endianmacros: Replace borked PA_FLOAT32_SWAP() with PA_READ_FLOAT32RE() / PA_WRITE_FLOAT32RE() building PA with -O0 leads to test failure in mix-test on i386 issue reported by Felipe, see http://lists.freedesktop.org/archives/pulseaudio-discuss/2014-August/021406.html the problem is the value 0xbeffbd7f: when byte-swapped it becomes 0x7fbdffbe and according to IEEE-754 represents a signalling NaN (starting with s111 1111 10, see http://en.wikipedia.org/wiki/NaN) when this value is assigned to a floating point register, it becomes 0x7ffdffbe, representing a quiet NaN (starting with s111 1111 11) -- a signalling NaN is turned into a quiet NaN! so PA_FLOAT32_SWAP(PA_FLOAT32_SWAP(x)) != x for certain values, uhuh! the following test code can be used; due to volatile, it will always demonstrate the issue; without volatile, it depends on the optimization level (i386, 32-bit, gcc 4.9): // snip static inline float PA_FLOAT32_SWAP(float x) { union { float f; uint32_t u; } t; t.f = x; t.u = bswap_32(t.u); return t.f; } int main() { unsigned x = 0xbeffbd7f; volatile float f = PA_FLOAT32_SWAP(*(float *)&x); printf("%08x %08x %08x %f\n", 0xbeffbd7f, *(unsigned *)&f, bswap_32(*(unsigned *)&f), f); } // snip the problem goes away with optimization when no temporary floating point registers are used the proposed solution is to avoid passing swapped floating point data in a float; this is done with new functions PA_READ_FLOAT32RE() and PA_WRITE_FLOAT32RE() which use uint32_t to dereference a pointer and byte-swap the data, hence no temporary float variable is used also delete PA_FLOAT32_TO_LE()/_BE(), not used Signed-off-by: Peter Meerwald <pmeerw@pmeerw.net> Reported-by: Felipe Sateler <fsateler@debian.org>
2014-09-02 23:53:09 +02:00
PA_WRITE_FLOAT32RE(b, k);
2009-01-16 03:15:39 +01:00
a += 3;
b++;
2009-01-16 03:15:39 +01:00
}
}
void pa_sconv_s24le_from_float32re(unsigned n, const float *a, uint8_t *b) {
pa_assert(a);
pa_assert(b);
for (; n > 0; n--) {
int32_t s;
endianmacros: Replace borked PA_FLOAT32_SWAP() with PA_READ_FLOAT32RE() / PA_WRITE_FLOAT32RE() building PA with -O0 leads to test failure in mix-test on i386 issue reported by Felipe, see http://lists.freedesktop.org/archives/pulseaudio-discuss/2014-August/021406.html the problem is the value 0xbeffbd7f: when byte-swapped it becomes 0x7fbdffbe and according to IEEE-754 represents a signalling NaN (starting with s111 1111 10, see http://en.wikipedia.org/wiki/NaN) when this value is assigned to a floating point register, it becomes 0x7ffdffbe, representing a quiet NaN (starting with s111 1111 11) -- a signalling NaN is turned into a quiet NaN! so PA_FLOAT32_SWAP(PA_FLOAT32_SWAP(x)) != x for certain values, uhuh! the following test code can be used; due to volatile, it will always demonstrate the issue; without volatile, it depends on the optimization level (i386, 32-bit, gcc 4.9): // snip static inline float PA_FLOAT32_SWAP(float x) { union { float f; uint32_t u; } t; t.f = x; t.u = bswap_32(t.u); return t.f; } int main() { unsigned x = 0xbeffbd7f; volatile float f = PA_FLOAT32_SWAP(*(float *)&x); printf("%08x %08x %08x %f\n", 0xbeffbd7f, *(unsigned *)&f, bswap_32(*(unsigned *)&f), f); } // snip the problem goes away with optimization when no temporary floating point registers are used the proposed solution is to avoid passing swapped floating point data in a float; this is done with new functions PA_READ_FLOAT32RE() and PA_WRITE_FLOAT32RE() which use uint32_t to dereference a pointer and byte-swap the data, hence no temporary float variable is used also delete PA_FLOAT32_TO_LE()/_BE(), not used Signed-off-by: Peter Meerwald <pmeerw@pmeerw.net> Reported-by: Felipe Sateler <fsateler@debian.org>
2014-09-02 23:53:09 +02:00
float v = PA_READ_FLOAT32RE(a) * (1U << 31);
s = (int32_t) PA_CLAMP_UNLIKELY(llrint(v), -0x80000000LL, 0x7FFFFFFFLL);
2009-01-16 03:15:39 +01:00
WRITE24(b, ((uint32_t) s) >> 8);
a++;
b+=3;
}
}
void pa_sconv_s24_32le_to_s16ne(unsigned n, const uint32_t *a, int16_t *b) {
pa_assert(a);
pa_assert(b);
for (; n > 0; n--) {
*b = (int16_t) (((int32_t) (UINT32_FROM(*a) << 8)) >> 16);
a++;
b++;
}
}
void pa_sconv_s24_32le_to_s16re(unsigned n, const uint32_t *a, int16_t *b) {
pa_assert(a);
pa_assert(b);
for (; n > 0; n--) {
int16_t s = (int16_t) ((int32_t) (UINT32_FROM(*a) << 8) >> 16);
*b = PA_INT16_SWAP(s);
a++;
b++;
}
}
void pa_sconv_s24_32le_from_s16ne(unsigned n, const int16_t *a, uint32_t *b) {
pa_assert(a);
pa_assert(b);
for (; n > 0; n--) {
*b = UINT32_TO(((uint32_t) ((int32_t) *a << 16)) >> 8);
a++;
b++;
}
}
void pa_sconv_s24_32le_from_s16re(unsigned n, const int16_t *a, uint32_t *b) {
pa_assert(a);
pa_assert(b);
for (; n > 0; n--) {
uint32_t s = ((uint32_t) ((int32_t) PA_INT16_SWAP(*a) << 16)) >> 8;
*b = UINT32_TO(s);
a++;
b++;
}
}
void pa_sconv_s24_32le_to_float32ne(unsigned n, const uint32_t *a, float *b) {
pa_assert(a);
pa_assert(b);
for (; n > 0; n--) {
int32_t s = (int32_t) (UINT32_FROM(*a) << 8);
*b = s * (1.0f / (1U << 31));
a++;
b++;
}
}
void pa_sconv_s24_32le_to_float32re(unsigned n, const uint32_t *a, float *b) {
pa_assert(a);
pa_assert(b);
for (; n > 0; n--) {
int32_t s = (int32_t) (UINT32_FROM(*a) << 8);
float k = s * (1.0f / (1U << 31));
endianmacros: Replace borked PA_FLOAT32_SWAP() with PA_READ_FLOAT32RE() / PA_WRITE_FLOAT32RE() building PA with -O0 leads to test failure in mix-test on i386 issue reported by Felipe, see http://lists.freedesktop.org/archives/pulseaudio-discuss/2014-August/021406.html the problem is the value 0xbeffbd7f: when byte-swapped it becomes 0x7fbdffbe and according to IEEE-754 represents a signalling NaN (starting with s111 1111 10, see http://en.wikipedia.org/wiki/NaN) when this value is assigned to a floating point register, it becomes 0x7ffdffbe, representing a quiet NaN (starting with s111 1111 11) -- a signalling NaN is turned into a quiet NaN! so PA_FLOAT32_SWAP(PA_FLOAT32_SWAP(x)) != x for certain values, uhuh! the following test code can be used; due to volatile, it will always demonstrate the issue; without volatile, it depends on the optimization level (i386, 32-bit, gcc 4.9): // snip static inline float PA_FLOAT32_SWAP(float x) { union { float f; uint32_t u; } t; t.f = x; t.u = bswap_32(t.u); return t.f; } int main() { unsigned x = 0xbeffbd7f; volatile float f = PA_FLOAT32_SWAP(*(float *)&x); printf("%08x %08x %08x %f\n", 0xbeffbd7f, *(unsigned *)&f, bswap_32(*(unsigned *)&f), f); } // snip the problem goes away with optimization when no temporary floating point registers are used the proposed solution is to avoid passing swapped floating point data in a float; this is done with new functions PA_READ_FLOAT32RE() and PA_WRITE_FLOAT32RE() which use uint32_t to dereference a pointer and byte-swap the data, hence no temporary float variable is used also delete PA_FLOAT32_TO_LE()/_BE(), not used Signed-off-by: Peter Meerwald <pmeerw@pmeerw.net> Reported-by: Felipe Sateler <fsateler@debian.org>
2014-09-02 23:53:09 +02:00
PA_WRITE_FLOAT32RE(b, k);
a++;
b++;
}
}
void pa_sconv_s24_32le_from_float32ne(unsigned n, const float *a, uint32_t *b) {
pa_assert(a);
pa_assert(b);
for (; n > 0; n--) {
int32_t s;
float v = *a * (1U << 31);
s = (int32_t) PA_CLAMP_UNLIKELY(llrint(v), -0x80000000LL, 0x7FFFFFFFLL);
*b = UINT32_TO(((uint32_t) s) >> 8);
a++;
b++;
}
}
void pa_sconv_s24_32le_from_float32re(unsigned n, const float *a, uint32_t *b) {
pa_assert(a);
pa_assert(b);
for (; n > 0; n--) {
int32_t s;
endianmacros: Replace borked PA_FLOAT32_SWAP() with PA_READ_FLOAT32RE() / PA_WRITE_FLOAT32RE() building PA with -O0 leads to test failure in mix-test on i386 issue reported by Felipe, see http://lists.freedesktop.org/archives/pulseaudio-discuss/2014-August/021406.html the problem is the value 0xbeffbd7f: when byte-swapped it becomes 0x7fbdffbe and according to IEEE-754 represents a signalling NaN (starting with s111 1111 10, see http://en.wikipedia.org/wiki/NaN) when this value is assigned to a floating point register, it becomes 0x7ffdffbe, representing a quiet NaN (starting with s111 1111 11) -- a signalling NaN is turned into a quiet NaN! so PA_FLOAT32_SWAP(PA_FLOAT32_SWAP(x)) != x for certain values, uhuh! the following test code can be used; due to volatile, it will always demonstrate the issue; without volatile, it depends on the optimization level (i386, 32-bit, gcc 4.9): // snip static inline float PA_FLOAT32_SWAP(float x) { union { float f; uint32_t u; } t; t.f = x; t.u = bswap_32(t.u); return t.f; } int main() { unsigned x = 0xbeffbd7f; volatile float f = PA_FLOAT32_SWAP(*(float *)&x); printf("%08x %08x %08x %f\n", 0xbeffbd7f, *(unsigned *)&f, bswap_32(*(unsigned *)&f), f); } // snip the problem goes away with optimization when no temporary floating point registers are used the proposed solution is to avoid passing swapped floating point data in a float; this is done with new functions PA_READ_FLOAT32RE() and PA_WRITE_FLOAT32RE() which use uint32_t to dereference a pointer and byte-swap the data, hence no temporary float variable is used also delete PA_FLOAT32_TO_LE()/_BE(), not used Signed-off-by: Peter Meerwald <pmeerw@pmeerw.net> Reported-by: Felipe Sateler <fsateler@debian.org>
2014-09-02 23:53:09 +02:00
float v = PA_READ_FLOAT32RE(a) * (1U << 31);
s = (int32_t) PA_CLAMP_UNLIKELY(llrint(v), -0x80000000LL, 0x7FFFFFFFLL);
*b = UINT32_TO(((uint32_t) s) >> 8);
a++;
b++;
}
}