Move all converters to starch-based implementations (#97)
* Switch all conversion routines to use starch.
main user-visible changes:
* ensure you check out submodules ('git clone --recurse-submodules")
* --version shows the CPU features and DSP implementations in use
* --wisdom allows overriding of the built-in architecture wisdom
* --dcfilter no longer supported
* "starch-benchmark" binary will benchmark all options on the
current machine and can produce a wisdom file to feed to
the --wisdom option
If you have a usecase for --dcfilter, please get in touch and
let me know - it's an edge case and for now there's no starch/DSP
support for it, but support can be written if needed.
In almost all cases the new conversion routines are slightly or
substantially faster than the old conversion routines. The only case
that is slower is SC16/SC16Q11 on a Pi 0, which is around 10% slower
due to changing from heavily approximated lookup tables to higher
quality results (but SC16 is probably already out of reach of a Pi 0)
* No need to build with SC16Q11_TABLE_BITS any more
* Add oneoff/uc8_capture_stats
(reads a UC8 capture; measures min/max/mean I and Q)
* Switch UC8 conversion to 127.4 center, 128 range.
Looking at actual UC8 captures from a RTL2832, the mean I and Q
are actually at 127.4, so use that as the zero point.
This means that the resulting I/Q maximum values could be as large as
127.6. Switch to 128 for simplicity.
* Switch to the new UC8 zero offset in benchmarks, fix some bugs
* Fix some bugs in SC16/SC16Q11 validation, tighten the max error requirements
* Ditch UC8 approximation path, add a NEON VRQSQRTE path.
* Tweak the SC16 exact path, add a new impl that uses a mix of
u32 & floats.
* SC16Q11 impl tweaks:
* add a u32->float exact path
* ditch the approximation path
* add a NEON VRSQRTE path
* add a 12-bit table path (using the full signed I/Q value, not absolute value)
* Ditch SC16 approximation path, add NEON vrsqrte path
* Add oneoff/dsp_error_measurement
This runs sample input through the DSP functions that are
allowed to be inexact and dumps the results as a TSV suitable for
feeding to gnuplot to look at the actual errors.
* Update make clean, make wisdom targets
* Update wisdom based on benchmarking
* Preserve the raw wisdom benchmark data
* Update to latest starch
* Update .gitignore for new wisdom files
* Update starch generated code
* Build starch-benchmark as part of the 'all' target
* Use wisdom from /etc/dump1090-fa/wisdom.local if present
* Package starch-benchmark and a helper script to generate local wisdom data
* Remove submodules in preparation for importing them directly
* Import cpu_features v0.6.0 from https://github.com/google/cpu_features/releases/tag/v0.6.0
* Import starch at commit a725c8491dc33a321565d451b385131e589d8490
from https://github.com/flightaware/starch
2021-01-21 11:45:00 +00:00
|
|
|
|
|
|
|
|
/* starch generated code. Do not edit. */
|
|
|
|
|
|
|
|
|
|
#include <stdint.h>
|
|
|
|
|
|
|
|
|
|
/* mixes */
|
|
|
|
|
|
|
|
|
|
/* ARM */
|
|
|
|
|
#ifdef STARCH_MIX_ARM
|
|
|
|
|
#define STARCH_FLAVOR_ARMV7A_VFPV4
|
|
|
|
|
#define STARCH_FLAVOR_ARMV7A_VFPV3
|
|
|
|
|
#define STARCH_FLAVOR_GENERIC
|
|
|
|
|
#define STARCH_MIX_ALIGNMENT 16
|
|
|
|
|
#endif /* STARCH_MIX_ARM */
|
|
|
|
|
|
2021-02-09 06:05:05 +00:00
|
|
|
/* Generic build, compiler defaults only */
|
|
|
|
|
#ifdef STARCH_MIX_GENERIC
|
|
|
|
|
#define STARCH_FLAVOR_GENERIC
|
|
|
|
|
#define STARCH_MIX_ALIGNMENT 1
|
|
|
|
|
#endif /* STARCH_MIX_GENERIC */
|
|
|
|
|
|
Move all converters to starch-based implementations (#97)
* Switch all conversion routines to use starch.
main user-visible changes:
* ensure you check out submodules ('git clone --recurse-submodules")
* --version shows the CPU features and DSP implementations in use
* --wisdom allows overriding of the built-in architecture wisdom
* --dcfilter no longer supported
* "starch-benchmark" binary will benchmark all options on the
current machine and can produce a wisdom file to feed to
the --wisdom option
If you have a usecase for --dcfilter, please get in touch and
let me know - it's an edge case and for now there's no starch/DSP
support for it, but support can be written if needed.
In almost all cases the new conversion routines are slightly or
substantially faster than the old conversion routines. The only case
that is slower is SC16/SC16Q11 on a Pi 0, which is around 10% slower
due to changing from heavily approximated lookup tables to higher
quality results (but SC16 is probably already out of reach of a Pi 0)
* No need to build with SC16Q11_TABLE_BITS any more
* Add oneoff/uc8_capture_stats
(reads a UC8 capture; measures min/max/mean I and Q)
* Switch UC8 conversion to 127.4 center, 128 range.
Looking at actual UC8 captures from a RTL2832, the mean I and Q
are actually at 127.4, so use that as the zero point.
This means that the resulting I/Q maximum values could be as large as
127.6. Switch to 128 for simplicity.
* Switch to the new UC8 zero offset in benchmarks, fix some bugs
* Fix some bugs in SC16/SC16Q11 validation, tighten the max error requirements
* Ditch UC8 approximation path, add a NEON VRQSQRTE path.
* Tweak the SC16 exact path, add a new impl that uses a mix of
u32 & floats.
* SC16Q11 impl tweaks:
* add a u32->float exact path
* ditch the approximation path
* add a NEON VRSQRTE path
* add a 12-bit table path (using the full signed I/Q value, not absolute value)
* Ditch SC16 approximation path, add NEON vrsqrte path
* Add oneoff/dsp_error_measurement
This runs sample input through the DSP functions that are
allowed to be inexact and dumps the results as a TSV suitable for
feeding to gnuplot to look at the actual errors.
* Update make clean, make wisdom targets
* Update wisdom based on benchmarking
* Preserve the raw wisdom benchmark data
* Update to latest starch
* Update .gitignore for new wisdom files
* Update starch generated code
* Build starch-benchmark as part of the 'all' target
* Use wisdom from /etc/dump1090-fa/wisdom.local if present
* Package starch-benchmark and a helper script to generate local wisdom data
* Remove submodules in preparation for importing them directly
* Import cpu_features v0.6.0 from https://github.com/google/cpu_features/releases/tag/v0.6.0
* Import starch at commit a725c8491dc33a321565d451b385131e589d8490
from https://github.com/flightaware/starch
2021-01-21 11:45:00 +00:00
|
|
|
/* x64-64 */
|
|
|
|
|
#ifdef STARCH_MIX_X86_64
|
|
|
|
|
#define STARCH_FLAVOR_X86_64_AVX2
|
|
|
|
|
#define STARCH_FLAVOR_X86_64_AVX
|
|
|
|
|
#define STARCH_FLAVOR_GENERIC
|
|
|
|
|
#define STARCH_MIX_ALIGNMENT 32
|
|
|
|
|
#endif /* STARCH_MIX_X86_64 */
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
#ifdef STARCH_MIX_ALIGNMENT
|
|
|
|
|
#define STARCH_ALIGNMENT STARCH_MIX_ALIGNMENT
|
|
|
|
|
#define STARCH_IS_ALIGNED(_ptr) (((uintptr_t)(_ptr) & (STARCH_MIX_ALIGNMENT-1)) == 0)
|
|
|
|
|
#else
|
|
|
|
|
/* mix not defined, alignment is unknown, treat everything as unaligned */
|
|
|
|
|
#define STARCH_IS_ALIGNED(_ptr) (0)
|
|
|
|
|
#endif
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
/* entry points and registries */
|
|
|
|
|
|
|
|
|
|
typedef void (* starch_subtract_n_ptr) ( const uint16_t * arg0, unsigned arg1, uint16_t arg2, uint16_t * arg3 );
|
|
|
|
|
extern starch_subtract_n_ptr starch_subtract_n;
|
|
|
|
|
|
|
|
|
|
typedef struct {
|
|
|
|
|
int rank;
|
|
|
|
|
const char *name;
|
|
|
|
|
const char *flavor;
|
|
|
|
|
starch_subtract_n_ptr callable;
|
|
|
|
|
int (*flavor_supported)();
|
|
|
|
|
} starch_subtract_n_regentry;
|
|
|
|
|
|
|
|
|
|
extern starch_subtract_n_regentry starch_subtract_n_registry[];
|
|
|
|
|
starch_subtract_n_regentry * starch_subtract_n_select();
|
|
|
|
|
void starch_subtract_n_set_wisdom( const char * const * received_wisdom );
|
|
|
|
|
|
|
|
|
|
typedef void (* starch_subtract_n_aligned_ptr) ( const uint16_t * arg0, unsigned arg1, uint16_t arg2, uint16_t * arg3 );
|
|
|
|
|
extern starch_subtract_n_aligned_ptr starch_subtract_n_aligned;
|
|
|
|
|
|
|
|
|
|
typedef struct {
|
|
|
|
|
int rank;
|
|
|
|
|
const char *name;
|
|
|
|
|
const char *flavor;
|
|
|
|
|
starch_subtract_n_aligned_ptr callable;
|
|
|
|
|
int (*flavor_supported)();
|
|
|
|
|
} starch_subtract_n_aligned_regentry;
|
|
|
|
|
|
|
|
|
|
extern starch_subtract_n_aligned_regentry starch_subtract_n_aligned_registry[];
|
|
|
|
|
starch_subtract_n_aligned_regentry * starch_subtract_n_aligned_select();
|
|
|
|
|
void starch_subtract_n_aligned_set_wisdom( const char * const * received_wisdom );
|
|
|
|
|
|
|
|
|
|
/* flavors and prototypes */
|
|
|
|
|
|
|
|
|
|
#ifdef STARCH_FLAVOR_ARMV7A_VFPV3
|
|
|
|
|
int supports_neon_vfpv3 (void);
|
|
|
|
|
void starch_subtract_n_generic_armv7a_vfpv3 ( const uint16_t * arg0, unsigned arg1, uint16_t arg2, uint16_t * arg3 );
|
|
|
|
|
void starch_subtract_n_aligned_generic_armv7a_vfpv3 ( const uint16_t * arg0, unsigned arg1, uint16_t arg2, uint16_t * arg3 );
|
|
|
|
|
void starch_subtract_n_unroll_4_armv7a_vfpv3 ( const uint16_t * arg0, unsigned arg1, uint16_t arg2, uint16_t * arg3 );
|
|
|
|
|
void starch_subtract_n_aligned_unroll_4_armv7a_vfpv3 ( const uint16_t * arg0, unsigned arg1, uint16_t arg2, uint16_t * arg3 );
|
|
|
|
|
void starch_subtract_n_bad_implementation_armv7a_vfpv3 ( const uint16_t * arg0, unsigned arg1, uint16_t arg2, uint16_t * arg3 );
|
|
|
|
|
void starch_subtract_n_aligned_bad_implementation_armv7a_vfpv3 ( const uint16_t * arg0, unsigned arg1, uint16_t arg2, uint16_t * arg3 );
|
|
|
|
|
void starch_subtract_n_neon_intrinsics_armv7a_vfpv3 ( const uint16_t * arg0, unsigned arg1, uint16_t arg2, uint16_t * arg3 );
|
|
|
|
|
void starch_subtract_n_aligned_neon_intrinsics_armv7a_vfpv3 ( const uint16_t * arg0, unsigned arg1, uint16_t arg2, uint16_t * arg3 );
|
|
|
|
|
#endif /* STARCH_FLAVOR_ARMV7A_VFPV3 */
|
|
|
|
|
|
|
|
|
|
int starch_read_wisdom (const char * path);
|
|
|
|
|
|
|
|
|
|
#ifdef STARCH_FLAVOR_ARMV7A_VFPV4
|
|
|
|
|
int supports_neon_vfpv4 (void);
|
|
|
|
|
void starch_subtract_n_generic_armv7a_vfpv4 ( const uint16_t * arg0, unsigned arg1, uint16_t arg2, uint16_t * arg3 );
|
|
|
|
|
void starch_subtract_n_aligned_generic_armv7a_vfpv4 ( const uint16_t * arg0, unsigned arg1, uint16_t arg2, uint16_t * arg3 );
|
|
|
|
|
void starch_subtract_n_unroll_4_armv7a_vfpv4 ( const uint16_t * arg0, unsigned arg1, uint16_t arg2, uint16_t * arg3 );
|
|
|
|
|
void starch_subtract_n_aligned_unroll_4_armv7a_vfpv4 ( const uint16_t * arg0, unsigned arg1, uint16_t arg2, uint16_t * arg3 );
|
|
|
|
|
void starch_subtract_n_bad_implementation_armv7a_vfpv4 ( const uint16_t * arg0, unsigned arg1, uint16_t arg2, uint16_t * arg3 );
|
|
|
|
|
void starch_subtract_n_aligned_bad_implementation_armv7a_vfpv4 ( const uint16_t * arg0, unsigned arg1, uint16_t arg2, uint16_t * arg3 );
|
|
|
|
|
void starch_subtract_n_neon_intrinsics_armv7a_vfpv4 ( const uint16_t * arg0, unsigned arg1, uint16_t arg2, uint16_t * arg3 );
|
|
|
|
|
void starch_subtract_n_aligned_neon_intrinsics_armv7a_vfpv4 ( const uint16_t * arg0, unsigned arg1, uint16_t arg2, uint16_t * arg3 );
|
|
|
|
|
#endif /* STARCH_FLAVOR_ARMV7A_VFPV4 */
|
|
|
|
|
|
|
|
|
|
int starch_read_wisdom (const char * path);
|
|
|
|
|
|
2021-02-09 06:05:05 +00:00
|
|
|
#ifdef STARCH_FLAVOR_GENERIC
|
|
|
|
|
void starch_subtract_n_generic_generic ( const uint16_t * arg0, unsigned arg1, uint16_t arg2, uint16_t * arg3 );
|
|
|
|
|
void starch_subtract_n_unroll_4_generic ( const uint16_t * arg0, unsigned arg1, uint16_t arg2, uint16_t * arg3 );
|
|
|
|
|
void starch_subtract_n_bad_implementation_generic ( const uint16_t * arg0, unsigned arg1, uint16_t arg2, uint16_t * arg3 );
|
|
|
|
|
#endif /* STARCH_FLAVOR_GENERIC */
|
|
|
|
|
|
|
|
|
|
int starch_read_wisdom (const char * path);
|
|
|
|
|
|
Move all converters to starch-based implementations (#97)
* Switch all conversion routines to use starch.
main user-visible changes:
* ensure you check out submodules ('git clone --recurse-submodules")
* --version shows the CPU features and DSP implementations in use
* --wisdom allows overriding of the built-in architecture wisdom
* --dcfilter no longer supported
* "starch-benchmark" binary will benchmark all options on the
current machine and can produce a wisdom file to feed to
the --wisdom option
If you have a usecase for --dcfilter, please get in touch and
let me know - it's an edge case and for now there's no starch/DSP
support for it, but support can be written if needed.
In almost all cases the new conversion routines are slightly or
substantially faster than the old conversion routines. The only case
that is slower is SC16/SC16Q11 on a Pi 0, which is around 10% slower
due to changing from heavily approximated lookup tables to higher
quality results (but SC16 is probably already out of reach of a Pi 0)
* No need to build with SC16Q11_TABLE_BITS any more
* Add oneoff/uc8_capture_stats
(reads a UC8 capture; measures min/max/mean I and Q)
* Switch UC8 conversion to 127.4 center, 128 range.
Looking at actual UC8 captures from a RTL2832, the mean I and Q
are actually at 127.4, so use that as the zero point.
This means that the resulting I/Q maximum values could be as large as
127.6. Switch to 128 for simplicity.
* Switch to the new UC8 zero offset in benchmarks, fix some bugs
* Fix some bugs in SC16/SC16Q11 validation, tighten the max error requirements
* Ditch UC8 approximation path, add a NEON VRQSQRTE path.
* Tweak the SC16 exact path, add a new impl that uses a mix of
u32 & floats.
* SC16Q11 impl tweaks:
* add a u32->float exact path
* ditch the approximation path
* add a NEON VRSQRTE path
* add a 12-bit table path (using the full signed I/Q value, not absolute value)
* Ditch SC16 approximation path, add NEON vrsqrte path
* Add oneoff/dsp_error_measurement
This runs sample input through the DSP functions that are
allowed to be inexact and dumps the results as a TSV suitable for
feeding to gnuplot to look at the actual errors.
* Update make clean, make wisdom targets
* Update wisdom based on benchmarking
* Preserve the raw wisdom benchmark data
* Update to latest starch
* Update .gitignore for new wisdom files
* Update starch generated code
* Build starch-benchmark as part of the 'all' target
* Use wisdom from /etc/dump1090-fa/wisdom.local if present
* Package starch-benchmark and a helper script to generate local wisdom data
* Remove submodules in preparation for importing them directly
* Import cpu_features v0.6.0 from https://github.com/google/cpu_features/releases/tag/v0.6.0
* Import starch at commit a725c8491dc33a321565d451b385131e589d8490
from https://github.com/flightaware/starch
2021-01-21 11:45:00 +00:00
|
|
|
#ifdef STARCH_FLAVOR_X86_64_AVX
|
|
|
|
|
int supports_x86_avx (void);
|
|
|
|
|
void starch_subtract_n_generic_x86_64_avx ( const uint16_t * arg0, unsigned arg1, uint16_t arg2, uint16_t * arg3 );
|
|
|
|
|
void starch_subtract_n_aligned_generic_x86_64_avx ( const uint16_t * arg0, unsigned arg1, uint16_t arg2, uint16_t * arg3 );
|
|
|
|
|
void starch_subtract_n_unroll_4_x86_64_avx ( const uint16_t * arg0, unsigned arg1, uint16_t arg2, uint16_t * arg3 );
|
|
|
|
|
void starch_subtract_n_aligned_unroll_4_x86_64_avx ( const uint16_t * arg0, unsigned arg1, uint16_t arg2, uint16_t * arg3 );
|
|
|
|
|
void starch_subtract_n_bad_implementation_x86_64_avx ( const uint16_t * arg0, unsigned arg1, uint16_t arg2, uint16_t * arg3 );
|
|
|
|
|
void starch_subtract_n_aligned_bad_implementation_x86_64_avx ( const uint16_t * arg0, unsigned arg1, uint16_t arg2, uint16_t * arg3 );
|
|
|
|
|
#endif /* STARCH_FLAVOR_X86_64_AVX */
|
|
|
|
|
|
|
|
|
|
int starch_read_wisdom (const char * path);
|
|
|
|
|
|
|
|
|
|
#ifdef STARCH_FLAVOR_X86_64_AVX2
|
|
|
|
|
int supports_x86_avx2 (void);
|
|
|
|
|
void starch_subtract_n_generic_x86_64_avx2 ( const uint16_t * arg0, unsigned arg1, uint16_t arg2, uint16_t * arg3 );
|
|
|
|
|
void starch_subtract_n_aligned_generic_x86_64_avx2 ( const uint16_t * arg0, unsigned arg1, uint16_t arg2, uint16_t * arg3 );
|
|
|
|
|
void starch_subtract_n_unroll_4_x86_64_avx2 ( const uint16_t * arg0, unsigned arg1, uint16_t arg2, uint16_t * arg3 );
|
|
|
|
|
void starch_subtract_n_aligned_unroll_4_x86_64_avx2 ( const uint16_t * arg0, unsigned arg1, uint16_t arg2, uint16_t * arg3 );
|
|
|
|
|
void starch_subtract_n_bad_implementation_x86_64_avx2 ( const uint16_t * arg0, unsigned arg1, uint16_t arg2, uint16_t * arg3 );
|
|
|
|
|
void starch_subtract_n_aligned_bad_implementation_x86_64_avx2 ( const uint16_t * arg0, unsigned arg1, uint16_t arg2, uint16_t * arg3 );
|
|
|
|
|
#endif /* STARCH_FLAVOR_X86_64_AVX2 */
|
|
|
|
|
|
|
|
|
|
int starch_read_wisdom (const char * path);
|
|
|
|
|
|