[openal] [PATCH] Add some mixer SSE2/4.1 optimisations
Timothy Arceri
t_arceri at yahoo.com.au
Mon Jun 2 18:08:30 EDT 2014
This is missing support for windows in helpers.c as I do not have a
windows machine to test on.
On Tue, 2014-06-03 at 08:05 +1000, Timothy Arceri wrote:
> When SSE4.1 is enabled these changes can reduce cpu time spent in OpenAL
> by 45% during the OpenArena benchmark of the Phoronix Test Suite.
>
> ---
> Alc/ALc.c | 4 +-
> Alc/ALu.c | 8 +++
> Alc/helpers.c | 8 ++-
> Alc/mixer.c | 15 +++--
> Alc/mixer_c.c | 13 +++++
> Alc/mixer_defs.h | 16 ++++++
> Alc/mixer_sse2.c | 139 ++++++++++++++++++++++++++++++++++++++++++++++
> Alc/mixer_sse41.c | 82 +++++++++++++++++++++++++++
> CMakeLists.txt | 60 +++++++++++++++++++-
> OpenAL32/Include/alMain.h | 3 +-
> config.h.in | 2 +
> 11 files changed, 338 insertions(+), 12 deletions(-)
> create mode 100644 Alc/mixer_sse2.c
> create mode 100644 Alc/mixer_sse41.c
>
> diff --git a/Alc/ALc.c b/Alc/ALc.c
> index 3b3e539..3438c23 100644
> --- a/Alc/ALc.c
> +++ b/Alc/ALc.c
> @@ -910,7 +910,7 @@ static void alc_initconfig(void)
>
> capfilter = 0;
> #ifdef HAVE_SSE
> - capfilter |= CPU_CAP_SSE | CPU_CAP_SSE2;
> + capfilter |= CPU_CAP_SSE | CPU_CAP_SSE2 | CPU_CAP_SSE4_1;
> #endif
> #ifdef HAVE_NEON
> capfilter |= CPU_CAP_NEON;
> @@ -940,6 +940,8 @@ static void alc_initconfig(void)
> capfilter &= ~CPU_CAP_SSE;
> else if(len == 4 && strncasecmp(str, "sse2", len) == 0)
> capfilter &= ~CPU_CAP_SSE2;
> + else if(len == 6 && strncasecmp(str, "sse4.1", len) == 0)
> + capfilter &= ~CPU_CAP_SSE4_1;
> else if(len == 4 && strncasecmp(str, "neon", len) == 0)
> capfilter &= ~CPU_CAP_NEON;
> else
> diff --git a/Alc/ALu.c b/Alc/ALu.c
> index 82932cf..6a8a061 100644
> --- a/Alc/ALu.c
> +++ b/Alc/ALu.c
> @@ -87,6 +87,14 @@ static ResamplerFunc SelectResampler(enum Resampler Resampler, ALuint increment)
> case PointResampler:
> return Resample_point32_C;
> case LinearResampler:
> +#ifdef HAVE_SSE4_1
> + if((CPUCapFlags&CPU_CAP_SSE4_1))
> + return Resample_lerp32_SSE41;
> +#endif
> +#ifdef HAVE_SSE2
> + if((CPUCapFlags&CPU_CAP_SSE2))
> + return Resample_lerp32_SSE2;
> +#endif
> return Resample_lerp32_C;
> case CubicResampler:
> return Resample_cubic32_C;
> diff --git a/Alc/helpers.c b/Alc/helpers.c
> index a0230b7..74893f2 100644
> --- a/Alc/helpers.c
> +++ b/Alc/helpers.c
> @@ -138,8 +138,11 @@ void FillCPUCaps(ALuint capfilter)
> if((cpuinf[0].regs[3]&(1<<25)))
> {
> caps |= CPU_CAP_SSE;
> - if((cpuinf[0].regs[3]&(1<<26)))
> + if((cpuinf[0].regs[3]&(1<<26))) {
> caps |= CPU_CAP_SSE2;
> + if((cpuinf[0].regs[2]&(1<<19)))
> + caps |= CPU_CAP_SSE4_1;
> + }
> }
> }
> }
> @@ -164,8 +167,9 @@ void FillCPUCaps(ALuint capfilter)
> caps |= CPU_CAP_NEON;
> #endif
>
> - TRACE("Got caps:%s%s%s%s\n", ((caps&CPU_CAP_SSE)?((capfilter&CPU_CAP_SSE)?" SSE":" (SSE)"):""),
> + TRACE("Got caps:%s%s%s%s%s\n", ((caps&CPU_CAP_SSE)?((capfilter&CPU_CAP_SSE)?" SSE":" (SSE)"):""),
> ((caps&CPU_CAP_SSE2)?((capfilter&CPU_CAP_SSE2)?" SSE2":" (SSE2)"):""),
> + ((caps&CPU_CAP_SSE4_1)?((capfilter&CPU_CAP_SSE4_1)?" SSE4.1":" (SSE4.1)"):""),
> ((caps&CPU_CAP_NEON)?((capfilter&CPU_CAP_NEON)?" Neon":" (Neon)"):""),
> ((!caps)?" -none-":""));
> CPUCapFlags = caps & capfilter;
> diff --git a/Alc/mixer.c b/Alc/mixer.c
> index ef1517b..b8d862e 100644
> --- a/Alc/mixer.c
> +++ b/Alc/mixer.c
> @@ -35,6 +35,7 @@
> #include "alAuxEffectSlot.h"
> #include "alu.h"
> #include "bs2b.h"
> +#include "mixer_defs.h"
>
>
> static inline ALfloat Sample_ALbyte(ALbyte val)
> @@ -383,12 +384,14 @@ ALvoid MixSource(ALactivesource *src, ALCdevice *Device, ALuint SamplesToDo)
> }
> }
> /* Update positions */
> - for(j = 0;j < DstBufferSize;j++)
> - {
> - DataPosFrac += increment;
> - DataPosInt += DataPosFrac>>FRACTIONBITS;
> - DataPosFrac &= FRACTIONMASK;
> - }
> +#ifdef HAVE_SSE2
> + if((CPUCapFlags&CPU_CAP_SSE2))
> + UpdatePositions_SSE2(DstBufferSize, &DataPosInt, &DataPosFrac, increment);
> + else
> + UpdatePositions_C(DstBufferSize, &DataPosInt, &DataPosFrac, increment);
> +#else
> + UpdatePositions_C(DstBufferSize, &DataPosInt, &DataPosFrac, increment);
> +#endif
> OutPos += DstBufferSize;
> src->Offset += DstBufferSize;
> src->Direct.Counter = maxu(src->Direct.Counter, DstBufferSize) - DstBufferSize;
> diff --git a/Alc/mixer_c.c b/Alc/mixer_c.c
> index 6dd01e7..aeb3d7c 100644
> --- a/Alc/mixer_c.c
> +++ b/Alc/mixer_c.c
> @@ -59,6 +59,19 @@ void ALfilterState_processC(ALfilterState *filter, ALfloat *restrict dst, const
> }
>
>
> +void UpdatePositions_C(ALuint DstBufferSize, ALuint *DataPosInt,
> + ALuint *DataPosFrac, ALuint increment)
> +{
> + ALuint j;
> + for(j = 0;j < DstBufferSize;j++)
> + {
> + *DataPosFrac += increment;
> + *DataPosInt += *DataPosFrac>>FRACTIONBITS;
> + *DataPosFrac &= FRACTIONMASK;
> + }
> +}
> +
> +
> static inline void ApplyCoeffsStep(ALuint Offset, ALfloat (*restrict Values)[2],
> const ALuint IrSize,
> ALfloat (*restrict Coeffs)[2],
> diff --git a/Alc/mixer_defs.h b/Alc/mixer_defs.h
> index 04fd1f5..52a8fdb 100644
> --- a/Alc/mixer_defs.h
> +++ b/Alc/mixer_defs.h
> @@ -30,6 +30,10 @@ void MixSend_C(ALfloat (*restrict OutBuffer)[BUFFERSIZE], const ALfloat *data,
> struct MixGainMono *Gain, ALuint Counter, ALuint OutPos,
> ALuint BufferSize);
>
> +/* C helpers */
> +void UpdatePositions_C(ALuint DstBufferSize, ALuint *DataPosInt,
> + ALuint *DataPosFrac, ALuint increment);
> +
> /* SSE mixers */
> void MixDirect_Hrtf_SSE(ALfloat (*restrict OutBuffer)[BUFFERSIZE], const ALfloat *data,
> ALuint Counter, ALuint Offset, ALuint OutPos, const ALuint IrSize,
> @@ -42,6 +46,18 @@ void MixSend_SSE(ALfloat (*restrict OutBuffer)[BUFFERSIZE], const ALfloat *data,
> struct MixGainMono *Gain, ALuint Counter, ALuint OutPos,
> ALuint BufferSize);
>
> +/* SSE resamplers */
> +const ALfloat *Resample_lerp32_SSE2(const ALfloat *src, ALuint frac, ALuint increment,
> + ALfloat *restrict dst, ALuint numsamples);
> +const ALfloat *Resample_lerp32_SSE41(const ALfloat *src, ALuint frac, ALuint increment,
> + ALfloat *restrict dst, ALuint numsamples);
> +
> +/* SSE helpers */
> +void InitiatePositionArrays(ALuint frac, ALuint increment, ALuint pos,
> + ALuint *frac_arr, ALuint *pos_arr);
> +void UpdatePositions_SSE2(ALuint DstBufferSize, ALuint *DataPosInt,
> + ALuint *DataPosFrac, ALuint increment);
> +
> /* Neon mixers */
> void MixDirect_Hrtf_Neon(ALfloat (*restrict OutBuffer)[BUFFERSIZE], const ALfloat *data,
> ALuint Counter, ALuint Offset, ALuint OutPos, const ALuint IrSize,
> diff --git a/Alc/mixer_sse2.c b/Alc/mixer_sse2.c
> new file mode 100644
> index 0000000..e4d4c3a
> --- /dev/null
> +++ b/Alc/mixer_sse2.c
> @@ -0,0 +1,139 @@
> +/**
> + * OpenAL cross platform audio library
> + * Copyright (C) 1999-2014 by authors.
> + * This library is free software; you can redistribute it and/or
> + * modify it under the terms of the GNU Library General Public
> + * License as published by the Free Software Foundation; either
> + * version 2 of the License, or (at your option) any later version.
> + *
> + * This library 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
> + * Library General Public License for more details.
> + *
> + * You should have received a copy of the GNU Library General Public
> + * License along with this library; if not, write to the
> + * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
> + * Boston, MA 02111-1307, USA.
> + * Or go to http://www.gnu.org/copyleft/lgpl.html
> + */
> +
> +#include "config.h"
> +
> +#include <emmintrin.h>
> +
> +#include "alu.h"
> +#include "mixer_defs.h"
> +
> +
> +void InitiatePositionArrays(ALuint frac, ALuint increment, ALuint pos,
> + ALuint *frac_arr, ALuint *pos_arr)
> +{
> + ALuint frac_tmp;
> +
> + pos_arr[0] = pos;
> + frac_arr[0] = frac;
> +
> + frac_tmp = frac_arr[0] + increment;
> + pos_arr[1] = pos_arr[0] + (frac_tmp>>FRACTIONBITS);
> + frac_arr[1] = frac_tmp & FRACTIONMASK;
> +
> + frac_tmp = frac_arr[1] + increment;
> + pos_arr[2] = pos_arr[1] + (frac_tmp>>FRACTIONBITS);
> + frac_arr[2] = frac_tmp & FRACTIONMASK;
> +
> + frac_tmp = frac_arr[2] + increment;
> + pos_arr[3] = pos_arr[2] + (frac_tmp>>FRACTIONBITS);
> + frac_arr[3] = frac_tmp & FRACTIONMASK;
> +}
> +
> +
> +void UpdatePositions_SSE2(ALuint DstBufferSize, ALuint *DataPosInt,
> + ALuint *DataPosFrac, ALuint increment)
> +{
> +
> + ALuint pos_arr[4];
> + ALuint frac_arr[4];
> + ALuint j;
> +
> + __m128i frac4, pos4;
> + const __m128i increment4 = _mm_set1_epi32(increment*4);
> + const __m128i fracMask4 = _mm_set1_epi32(FRACTIONMASK);
> +
> + InitiatePositionArrays(*DataPosFrac, increment, *DataPosInt, frac_arr, pos_arr);
> +
> + frac4 = _mm_set_epi32(frac_arr[3], frac_arr[2], frac_arr[1], frac_arr[0]);
> + pos4 = _mm_set_epi32(pos_arr[3], pos_arr[2], pos_arr[1], pos_arr[0]);
> +
> + for(j = 0;j < DstBufferSize-3;j+=4)
> + {
> + frac4 = _mm_add_epi32(frac4, increment4);
> + pos4 = _mm_add_epi32(pos4, _mm_srli_epi32(frac4, FRACTIONBITS));
> + frac4 = _mm_and_si128(frac4, fracMask4);
> + }
> +
> + *DataPosFrac = _mm_cvtsi128_si32(_mm_shuffle_epi32(frac4, _MM_SHUFFLE(3,3,3,3)));
> + *DataPosInt = _mm_cvtsi128_si32(_mm_shuffle_epi32(pos4, _MM_SHUFFLE(3,3,3,3)));
> +
> + for(;j < DstBufferSize;j++)
> + {
> + *DataPosFrac += increment;
> + *DataPosInt += *DataPosFrac>>FRACTIONBITS;
> + *DataPosFrac &= FRACTIONMASK;
> + }
> +}
> +
> +const ALfloat *Resample_lerp32_SSE2(const ALfloat *src, ALuint frac, ALuint increment,
> + ALfloat *restrict dst, ALuint numsamples)
> +{
> + ALuint i;
> + ALuint pos = 0;
> + ALuint pos_arr[4];
> + ALuint frac_arr[4];
> +
> + __m128i frac4, pos4;
> + const __m128i increment4 = _mm_set1_epi32(increment*4);
> + const __m128 fracOne4 = _mm_set1_ps(1.0f/FRACTIONONE);
> + const __m128i fracMask4 = _mm_set1_epi32(FRACTIONMASK);
> +
> + InitiatePositionArrays(frac, increment, pos, frac_arr, pos_arr);
> +
> + frac4 = _mm_set_epi32(frac_arr[3], frac_arr[2], frac_arr[1], frac_arr[0]);
> + pos4 = _mm_set_epi32(pos_arr[3], pos_arr[2], pos_arr[1], pos_arr[0]);
> +
> + for(i = 0;i < numsamples-3;i += 4)
> + {
> + __m128 val1 = _mm_set_ps(src[pos_arr[3]], src[pos_arr[2]], src[pos_arr[1]], src[pos_arr[0]]);
> + __m128 val2 = _mm_set_ps(src[pos_arr[3]+1], src[pos_arr[2]+1], src[pos_arr[1]+1], src[pos_arr[0]+1]);
> +
> + /* val1 + (val2-val1)*mu */
> + const __m128 r0 = _mm_sub_ps(val2, val1);
> + const __m128 mu = _mm_mul_ps(_mm_cvtepi32_ps(frac4), fracOne4);
> + const __m128 r1 = _mm_mul_ps(mu, r0);
> + const __m128 out = _mm_add_ps(val1, r1);
> +
> + _mm_store_ps(&dst[i], out);
> +
> + frac4 = _mm_add_epi32(frac4, increment4);
> + pos4 = _mm_add_epi32(pos4, _mm_srli_epi32(frac4, FRACTIONBITS));
> + frac4 = _mm_and_si128(frac4, fracMask4);
> +
> + pos_arr[0] = _mm_cvtsi128_si32(pos4);
> + pos_arr[1] = _mm_cvtsi128_si32(_mm_shuffle_epi32(pos4, _MM_SHUFFLE(1,1,1,1)));
> + pos_arr[2] = _mm_cvtsi128_si32(_mm_shuffle_epi32(pos4, _MM_SHUFFLE(2,2,2,2)));
> + pos_arr[3] = _mm_cvtsi128_si32(_mm_shuffle_epi32(pos4, _MM_SHUFFLE(3,3,3,3)));
> + }
> +
> + pos = pos_arr[0];
> + frac = _mm_cvtsi128_si32(frac4);
> +
> + for(;i < numsamples;i++)
> + {
> + dst[i] = lerp(src[pos], src[pos+1], frac * (1.0f/FRACTIONONE));
> +
> + frac += increment;
> + pos += frac>>FRACTIONBITS;
> + frac &= FRACTIONMASK;
> + }
> + return dst;
> +}
> diff --git a/Alc/mixer_sse41.c b/Alc/mixer_sse41.c
> new file mode 100644
> index 0000000..7f70ee8
> --- /dev/null
> +++ b/Alc/mixer_sse41.c
> @@ -0,0 +1,82 @@
> +/**
> + * OpenAL cross platform audio library
> + * Copyright (C) 1999-2014 by authors.
> + * This library is free software; you can redistribute it and/or
> + * modify it under the terms of the GNU Library General Public
> + * License as published by the Free Software Foundation; either
> + * version 2 of the License, or (at your option) any later version.
> + *
> + * This library 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
> + * Library General Public License for more details.
> + *
> + * You should have received a copy of the GNU Library General Public
> + * License along with this library; if not, write to the
> + * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
> + * Boston, MA 02111-1307, USA.
> + * Or go to http://www.gnu.org/copyleft/lgpl.html
> + */
> +
> +#include "config.h"
> +
> +#include <smmintrin.h>
> +
> +#include "alu.h"
> +#include "mixer_defs.h"
> +
> +
> +const ALfloat *Resample_lerp32_SSE41(const ALfloat *src, ALuint frac, ALuint increment,
> + ALfloat *restrict dst, ALuint numsamples)
> +{
> + ALuint i;
> + ALuint pos = 0;
> + ALuint pos_arr[4];
> + ALuint frac_arr[4];
> +
> + __m128i frac4, pos4;
> + const __m128i increment4 = _mm_set1_epi32(increment*4);
> + const __m128 fracOne4 = _mm_set1_ps(1.0f/FRACTIONONE);
> + const __m128i fracMask4 = _mm_set1_epi32(FRACTIONMASK);
> +
> + InitiatePositionArrays(frac, increment, pos, frac_arr, pos_arr);
> +
> + frac4 = _mm_set_epi32(frac_arr[3], frac_arr[2], frac_arr[1], frac_arr[0]);
> + pos4 = _mm_set_epi32(pos_arr[3], pos_arr[2], pos_arr[1], pos_arr[0]);
> +
> + for(i = 0;i < numsamples-3;i += 4)
> + {
> + __m128 val1 = _mm_set_ps(src[pos_arr[3]], src[pos_arr[2]], src[pos_arr[1]], src[pos_arr[0]]);
> + __m128 val2 = _mm_set_ps(src[pos_arr[3]+1], src[pos_arr[2]+1], src[pos_arr[1]+1], src[pos_arr[0]+1]);
> +
> + /* val1 + (val2-val1)*mu */
> + const __m128 r0 = _mm_sub_ps(val2, val1);
> + const __m128 mu = _mm_mul_ps(_mm_cvtepi32_ps(frac4), fracOne4);
> + const __m128 r1 = _mm_mul_ps(mu, r0);
> + const __m128 out = _mm_add_ps(val1, r1);
> +
> + _mm_store_ps(&dst[i], out);
> +
> + frac4 = _mm_add_epi32(frac4, increment4);
> + pos4 = _mm_add_epi32(pos4, _mm_srli_epi32(frac4, FRACTIONBITS));
> + frac4 = _mm_and_si128(frac4, fracMask4);
> +
> + pos_arr[0] = _mm_extract_epi32(pos4, 0);
> + pos_arr[1] = _mm_extract_epi32(pos4, 1);
> + pos_arr[2] = _mm_extract_epi32(pos4, 2);
> + pos_arr[3] = _mm_extract_epi32(pos4, 3);
> + }
> +
> + pos = pos_arr[0];
> + frac = _mm_cvtsi128_si32(frac4);
> +
> + for(;i < numsamples;i++)
> + {
> + dst[i] = lerp(src[pos], src[pos+1], frac * (1.0f/FRACTIONONE));
> +
> + frac += increment;
> + pos += frac>>FRACTIONBITS;
> + frac &= FRACTIONMASK;
> + }
> + return dst;
> +}
> diff --git a/CMakeLists.txt b/CMakeLists.txt
> index 27ec720..3d5dcc9 100644
> --- a/CMakeLists.txt
> +++ b/CMakeLists.txt
> @@ -300,6 +300,8 @@ ELSE()
> ENDIF()
>
> SET(SSE_SWITCH "")
> +SET(SSE2_SWITCH "")
> +SET(SSE4_1_SWITCH "")
> IF(MSVC)
> CHECK_C_COMPILER_FLAG(/arch:SSE HAVE_ARCHSSE_SWITCH)
> IF(HAVE_ARCHSSE_SWITCH)
> @@ -312,6 +314,18 @@ IF(NOT SSE_SWITCH)
> SET(SSE_SWITCH "-msse")
> ENDIF()
> ENDIF()
> +IF(NOT SSE2_SWITCH)
> + CHECK_C_COMPILER_FLAG(-msse2 HAVE_MSSE2_SWITCH)
> + IF(HAVE_MSSE2_SWITCH)
> + SET(SSE2_SWITCH "-msse2")
> + ENDIF()
> +ENDIF()
> +IF(NOT SSE4_1_SWITCH)
> + CHECK_C_COMPILER_FLAG(-msse4.1 HAVE_MSSE4_1_SWITCH)
> + IF(HAVE_MSSE4_1_SWITCH)
> + SET(SSE4_1_SWITCH "-msse4.1")
> + ENDIF()
> +ENDIF()
>
> CHECK_C_SOURCE_COMPILES("int foo(const char *str, ...) __attribute__((format(printf, 1, 2)));
> int main() {return 0;}" HAVE_GCC_FORMAT)
> @@ -547,13 +561,15 @@ SET(ALC_OBJS Alc/ALc.c
>
> SET(CPU_EXTS "Default")
> SET(HAVE_SSE 0)
> +SET(HAVE_SSE2 0)
> +SET(HAVE_SSE4_1 0)
> SET(HAVE_NEON 0)
>
> # Check for SSE support
> -OPTION(ALSOFT_REQUIRE_SSE "Require SSE/SSE2 support" OFF)
> +OPTION(ALSOFT_REQUIRE_SSE "Require SSE support" OFF)
> CHECK_INCLUDE_FILE(xmmintrin.h HAVE_XMMINTRIN_H "${SSE_SWITCH}")
> IF(HAVE_XMMINTRIN_H)
> - OPTION(ALSOFT_CPUEXT_SSE "Enable SSE/SSE2 support" ON)
> + OPTION(ALSOFT_CPUEXT_SSE "Enable SSE support" ON)
> IF(ALSOFT_CPUEXT_SSE)
> IF(ALIGN_DECL OR HAVE_C11_ALIGNAS)
> SET(HAVE_SSE 1)
> @@ -570,6 +586,46 @@ IF(ALSOFT_REQUIRE_SSE AND NOT HAVE_SSE)
> MESSAGE(FATAL_ERROR "Failed to enabled required SSE CPU extensions")
> ENDIF()
>
> +OPTION(ALSOFT_REQUIRE_SSE2 "Require SSE2 support" OFF)
> +CHECK_INCLUDE_FILE(emmintrin.h HAVE_EMMINTRIN_H "${SSE2_SWITCH}")
> +IF(HAVE_EMMINTRIN_H)
> + OPTION(ALSOFT_CPUEXT_SSE2 "Enable SSE2 support" ON)
> + IF(ALSOFT_CPUEXT_SSE2)
> + IF(ALIGN_DECL OR HAVE_C11_ALIGNAS)
> + SET(HAVE_SSE2 1)
> + SET(ALC_OBJS ${ALC_OBJS} Alc/mixer_sse2.c)
> + IF(SSE2_SWITCH)
> + SET_SOURCE_FILES_PROPERTIES(Alc/mixer_sse2.c PROPERTIES
> + COMPILE_FLAGS "${SSE2_SWITCH}")
> + ENDIF()
> + SET(CPU_EXTS "${CPU_EXTS}, SSE2")
> + ENDIF()
> + ENDIF()
> +ENDIF()
> +IF(ALSOFT_REQUIRE_SSE2 AND NOT HAVE_SSE2)
> + MESSAGE(FATAL_ERROR "Failed to enabled required SSE2 CPU extensions")
> +ENDIF()
> +
> +OPTION(ALSOFT_REQUIRE_SSE4_1 "Require SSE4.1 support" OFF)
> +CHECK_INCLUDE_FILE(smmintrin.h HAVE_SMMINTRIN_H "${SSE4_1_SWITCH}")
> +IF(HAVE_SMMINTRIN_H)
> + OPTION(ALSOFT_CPUEXT_SSE4_1 "Enable SSE4.1 support" ON)
> + IF(ALSOFT_CPUEXT_SSE4_1)
> + IF(ALIGN_DECL OR HAVE_C11_ALIGNAS)
> + SET(HAVE_SSE4_1 1)
> + SET(ALC_OBJS ${ALC_OBJS} Alc/mixer_sse41.c)
> + IF(SSE4_1_SWITCH)
> + SET_SOURCE_FILES_PROPERTIES(Alc/mixer_sse41.c PROPERTIES
> + COMPILE_FLAGS "${SSE4_1_SWITCH}")
> + ENDIF()
> + SET(CPU_EXTS "${CPU_EXTS}, SSE4.1")
> + ENDIF()
> + ENDIF()
> +ENDIF()
> +IF(ALSOFT_REQUIRE_SSE4_1 AND NOT HAVE_SSE4_1)
> + MESSAGE(FATAL_ERROR "Failed to enabled required SSE4.1 CPU extensions")
> +ENDIF()
> +
> # Check for ARM Neon support
> OPTION(ALSOFT_REQUIRE_NEON "Require ARM Neon support" OFF)
> CHECK_INCLUDE_FILE(arm_neon.h HAVE_ARM_NEON_H)
> diff --git a/OpenAL32/Include/alMain.h b/OpenAL32/Include/alMain.h
> index 57b0811..232c438 100644
> --- a/OpenAL32/Include/alMain.h
> +++ b/OpenAL32/Include/alMain.h
> @@ -887,7 +887,8 @@ extern ALuint CPUCapFlags;
> enum {
> CPU_CAP_SSE = 1<<0,
> CPU_CAP_SSE2 = 1<<1,
> - CPU_CAP_NEON = 1<<2,
> + CPU_CAP_SSE4_1 = 1<<2,
> + CPU_CAP_NEON = 1<<3,
> };
>
> void FillCPUCaps(ALuint capfilter);
> diff --git a/config.h.in b/config.h.in
> index 090c00a..3fdc0c7 100644
> --- a/config.h.in
> +++ b/config.h.in
> @@ -25,6 +25,8 @@
>
> /* Define if we have SSE CPU extensions */
> #cmakedefine HAVE_SSE
> +#cmakedefine HAVE_SSE2
> +#cmakedefine HAVE_SSE4_1
>
> /* Define if we have ARM Neon CPU extensions */
> #cmakedefine HAVE_NEON
More information about the openal
mailing list