Fixed-point arithmetic c library inspired by Atmel Software Framework, and en.wikipedia.org/wiki/Q_(number_format)
Q is a fixed point number format where the number of fractional bits (and optionally the number of integer bits) is specified. For example, a Q15 number has 15 fractional bits; a Q1.14 number has 1 integer bit and 14 fractional bits. and in applications that require constant resolution.
- from wikipedia.com
Q14.1 - "SIIIIIIIIIIIIIIF" <- -16384 < float < 16384
Q13.2 - "SIIIIIIIIIIIIIFF" <- -8192 < float < 8192
Q12.3 - "SIIIIIIIIIIIIFFF" <- -4096 < float < 4096
Q11.4 - "SIIIIIIIIIIIFFFF" <- -2048 < float < 2048
Q10.5 - "SIIIIIIIIIIFFFFF" <- -1024 < float < 1024
Q9.6 - "SIIIIIIIIIFFFFFF" <- -512 < float < 512
Q8.7 - "SIIIIIIIIFFFFFFF" <- -256 < float < 256
Q7.8 - "SIIIIIIIFFFFFFFF" <- -128 < float < 128
Q6.9 - "SIIIIIIFFFFFFFFF" <- -64 < float < 64
Q5.10 - "SIIIIIFFFFFFFFFF" <- -32 < float < 32
Q4.11 - "SIIIIFFFFFFFFFFF" <- -16 < float < 16
Q3.12 - "SIIIFFFFFFFFFFFF" <- -8 < float < 8
Q2.13 - "SIIFFFFFFFFFFFFF" <- -4 < float < 4
Q1.14 - "SIFFFFFFFFFFFFFF" <- -2 < float < 2
Q0.15 - "SFFFFFFFFFFFFFFF" <- -1 < float < 1 where S is sign bit, I is integer bit, and F is fractional bit.
In float-typed conatiner,
Their unit resolution are;
Q14.1: 5.000000000000000000e-01
Q13.2: 2.500000000000000000e-01
Q12.3: 1.250000000000000000e-01
Q11.4: 6.250000000000000000e-02
Q10.5: 3.125000000000000000e-02
Q9.6: 1.562500000000000000e-02
Q8.7: 7.812500000000000000e-03
Q7.8: 3.906250000000000000e-03
Q6.9: 1.953125000000000000e-03
Q5.10: 9.765625000000000000e-04
Q4.11: 4.882812500000000000e-04
Q3.12: 2.441406250000000000e-04
Q2.13: 1.220703125000000000e-04
Q1.14: 6.103515625000000000e-05
Q0.15: 3.051757812500000000e-05And the exact covering ranges are;
Q14.1: [-1.638400000000000000e+04, 1.638350000000000000e+04]
Q13.2: [-8.192000000000000000e+03, 8.191750000000000000e+03]
Q12.3: [-4.096000000000000000e+03, 4.095875000000000000e+03]
Q11.4: [-2.048000000000000000e+03, 2.047937500000000000e+03]
Q10.5: [-1.024000000000000000e+03, 1.023968750000000000e+03]
Q9.6: [-5.120000000000000000e+02, 5.119843750000000000e+02]
Q8.7: [-2.560000000000000000e+02, 2.559921875000000000e+02]
Q7.8: [-1.280000000000000000e+02, 1.279960937500000000e+02]
Q6.9: [-6.400000000000000000e+01, 6.399804687500000000e+01]
Q5.10: [-3.200000000000000000e+01, 3.199902343750000000e+01]
Q4.11: [-1.600000000000000000e+01, 1.599951171875000000e+01]
Q3.12: [-8.000000000000000000e+00, 7.999755859375000000e+00]
Q2.13: [-4.000000000000000000e+00, 3.999877929687500000e+00]
Q1.14: [-2.000000000000000000e+00, 1.999938964843750000e+00]
Q0.15: [-1.000000000000000000e+00, 9.999694824218750000e-01]In double-typed conatiner,
Their unit resolution are;
Q14.1: 5.000000000000000000e-01
Q13.2: 2.500000000000000000e-01
Q12.3: 1.250000000000000000e-01
Q11.4: 6.250000000000000000e-02
Q10.5: 3.125000000000000000e-02
Q9.6: 1.562500000000000000e-02
Q8.7: 7.812500000000000000e-03
Q7.8: 3.906250000000000000e-03
Q6.9: 1.953125000000000000e-03
Q5.10: 9.765625000000000000e-04
Q4.11: 4.882812500000000000e-04
Q3.12: 2.441406250000000000e-04
Q2.13: 1.220703125000000000e-04
Q1.14: 6.103515625000000000e-05
Q0.15: 3.051757812500000000e-05And the exact covering ranges are;
Q14.1: [-1.638400000000000000e+04, 1.638350000000000000e+04]
Q13.2: [-8.192000000000000000e+03, 8.191750000000000000e+03]
Q12.3: [-4.096000000000000000e+03, 4.095875000000000000e+03]
Q11.4: [-2.048000000000000000e+03, 2.047937500000000000e+03]
Q10.5: [-1.024000000000000000e+03, 1.023968750000000000e+03]
Q9.6: [-5.120000000000000000e+02, 5.119843750000000000e+02]
Q8.7: [-2.560000000000000000e+02, 2.559921875000000000e+02]
Q7.8: [-1.280000000000000000e+02, 1.279960937500000000e+02]
Q6.9: [-6.400000000000000000e+01, 6.399804687500000000e+01]
Q5.10: [-3.200000000000000000e+01, 3.199902343750000000e+01]
Q4.11: [-1.600000000000000000e+01, 1.599951171875000000e+01]
Q3.12: [-8.000000000000000000e+00, 7.999755859375000000e+00]
Q2.13: [-4.000000000000000000e+00, 3.999877929687500000e+00]
Q1.14: [-2.000000000000000000e+00, 1.999938964843750000e+00]
Q0.15: [-1.000000000000000000e+00, 9.999694824218750000e-01] Q30_1 - SIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIF <- -1073741824 < float < 1073741824
Q29_2 - SIIIIIIIIIIIIIIIIIIIIIIIIIIIIIFF <- -536870912 < float < 536870912
Q28_3 - SIIIIIIIIIIIIIIIIIIIIIIIIIIIIFFF <- -268435456 < float < 268435456
Q27_4 - SIIIIIIIIIIIIIIIIIIIIIIIIIIIFFFF <- -134217728 < float < 134217728
Q26_5 - SIIIIIIIIIIIIIIIIIIIIIIIIIIFFFFF <- -67108864 < float < 67108864
Q25_6 - SIIIIIIIIIIIIIIIIIIIIIIIIIFFFFFF <- -33554432 < float < 33554432
Q24_7 - SIIIIIIIIIIIIIIIIIIIIIIIIFFFFFFF <- -16777216 < float < 16777216
Q23_8 - SIIIIIIIIIIIIIIIIIIIIIIIFFFFFFFF <- -8388608 < float < 8388608
Q22_9 - SIIIIIIIIIIIIIIIIIIIIIIFFFFFFFFF <- -4194304 < float < 4194304
Q21_10 - SIIIIIIIIIIIIIIIIIIIIIFFFFFFFFFF <- -2097152 < float < 2097152
Q20_11 - SIIIIIIIIIIIIIIIIIIIIFFFFFFFFFFF <- -1048576 < float < 1048576
Q19_12 - SIIIIIIIIIIIIIIIIIIIFFFFFFFFFFFF <- -524288 < float < 524288
Q18_13 - SIIIIIIIIIIIIIIIIIIFFFFFFFFFFFFF <- -262144 < float < 262144
Q17_14 - SIIIIIIIIIIIIIIIIIFFFFFFFFFFFFFF <- -131072 < float < 131072
Q16_15 - SIIIIIIIIIIIIIIIIFFFFFFFFFFFFFFF <- -65536 < float < 65536
Q15_16 - SIIIIIIIIIIIIIIIFFFFFFFFFFFFFFFF <- -32768 < float < 32768
Q14_17 - SIIIIIIIIIIIIIIFFFFFFFFFFFFFFFFF <- -16384 < float < 16384
Q13_18 - SIIIIIIIIIIIIIFFFFFFFFFFFFFFFFFF <- -8192 < float < 8192
Q12_19 - SIIIIIIIIIIIIFFFFFFFFFFFFFFFFFFF <- -4096 < float < 4096
Q11_20 - SIIIIIIIIIIIFFFFFFFFFFFFFFFFFFFF <- -2048 < float < 2048
Q10_21 - SIIIIIIIIIIFFFFFFFFFFFFFFFFFFFFF <= -1024 < float < 1024
Q9_22 - SIIIIIIIIIFFFFFFFFFFFFFFFFFFFFFF <= -512 < float < 512
Q8_23 - SIIIIIIIIFFFFFFFFFFFFFFFFFFFFFFF <= -256 < float < 256
Q7_24 - SIIIIIIIFFFFFFFFFFFFFFFFFFFFFFFF <= -128 < float < 128
Q6_25 - SIIIIIIFFFFFFFFFFFFFFFFFFFFFFFFF <= -64 < float < 64
Q5_26 - SIIIIIFFFFFFFFFFFFFFFFFFFFFFFFFF <= -32 < float < 32
Q4_27 - SIIIIFFFFFFFFFFFFFFFFFFFFFFFFFFF <= -16 < float < 16
Q3_28 - SIIIFFFFFFFFFFFFFFFFFFFFFFFFFFFF <= -8 < float < 8
Q2_29 - SIIFFFFFFFFFFFFFFFFFFFFFFFFFFFFF <= -4 < float < 4
Q1_30 - SIFFFFFFFFFFFFFFFFFFFFFFFFFFFFFF <= -2 < float < 2
Q0_31 - SFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFF <= -1 < float < 1, where S is sign bit, I is integer bit, and F is fractional bit.
In float-typed conatiner,
Their unit resolution are;
Q30.1: 5.000000000000000000e-01
Q29.2: 2.500000000000000000e-01
Q28.3: 1.250000000000000000e-01
Q27.4: 6.250000000000000000e-02
Q26.5: 3.125000000000000000e-02
Q25.6: 1.562500000000000000e-02
Q24.7: 7.812500000000000000e-03
Q23.8: 3.906250000000000000e-03
Q22.9: 1.953125000000000000e-03
Q21.10: 9.765625000000000000e-04
Q20.11: 4.882812500000000000e-04
Q19.12: 2.441406250000000000e-04
Q18.13: 1.220703125000000000e-04
Q17.14: 6.103515625000000000e-05
Q16.15: 3.051757812500000000e-05
Q15.16: 1.525878906250000000e-05
Q14.17: 7.629394531250000000e-06
Q13.18: 3.814697265625000000e-06
Q12.19: 1.907348632812500000e-06
Q11.20: 9.536743164062500000e-07
Q10.21: 4.768371582031250000e-07
Q9.22: 2.384185791015625000e-07
Q8.23: 1.192092895507812500e-07
Q7.24: 5.960464477539062500e-08
Q6.25: 2.980232238769531250e-08
Q5.26: 1.490116119384765625e-08
Q4.27: 7.450580596923828125e-09
Q3.28: 3.725290298461914062e-09
Q2.29: 1.862645149230957031e-09
Q1.30: 9.313225746154785156e-10
Q0.31: 4.656612873077392578e-10And the exact covering ranges are;
Q30.1: [-1.073741824000000000e+09, 1.073741824000000000e+09]
Q29.2: [-5.368709120000000000e+08, 5.368709120000000000e+08]
Q28.3: [-2.684354560000000000e+08, 2.684354560000000000e+08]
Q27.4: [-1.342177280000000000e+08, 1.342177280000000000e+08]
Q26.5: [-6.710886400000000000e+07, 6.710886400000000000e+07]
Q25.6: [-3.355443200000000000e+07, 3.355443200000000000e+07]
Q24.7: [-1.677721600000000000e+07, 1.677721600000000000e+07]
Q23.8: [-8.388608000000000000e+06, 8.388608000000000000e+06]
Q22.9: [-4.194304000000000000e+06, 4.194304000000000000e+06]
Q21.10: [-2.097152000000000000e+06, 2.097152000000000000e+06]
Q20.11: [-1.048576000000000000e+06, 1.048576000000000000e+06]
Q19.12: [-5.242880000000000000e+05, 5.242880000000000000e+05]
Q18.13: [-2.621440000000000000e+05, 2.621440000000000000e+05]
Q17.14: [-1.310720000000000000e+05, 1.310720000000000000e+05]
Q16.15: [-6.553600000000000000e+04, 6.553600000000000000e+04]
Q15.16: [-3.276800000000000000e+04, 3.276800000000000000e+04]
Q14.17: [-1.638400000000000000e+04, 1.638400000000000000e+04]
Q13.18: [-8.192000000000000000e+03, 8.192000000000000000e+03]
Q12.19: [-4.096000000000000000e+03, 4.096000000000000000e+03]
Q11.20: [-2.048000000000000000e+03, 2.048000000000000000e+03]
Q10.21: [-1.024000000000000000e+03, 1.024000000000000000e+03]
Q9.22: [-5.120000000000000000e+02, 5.120000000000000000e+02]
Q8.23: [-2.560000000000000000e+02, 2.560000000000000000e+02]
Q7.24: [-1.280000000000000000e+02, 1.280000000000000000e+02]
Q6.25: [-6.400000000000000000e+01, 6.400000000000000000e+01]
Q5.26: [-3.200000000000000000e+01, 3.200000000000000000e+01]
Q4.27: [-1.600000000000000000e+01, 1.600000000000000000e+01]
Q3.28: [-8.000000000000000000e+00, 8.000000000000000000e+00]
Q2.29: [-4.000000000000000000e+00, 4.000000000000000000e+00]
Q1.30: [-2.000000000000000000e+00, 2.000000000000000000e+00]
Q0.31: [-1.000000000000000000e+00, 1.000000000000000000e+00]In double-typed conatiner,
Their unit resolution are;
Q30.1: 5.000000000000000000e-01
Q29.2: 2.500000000000000000e-01
Q28.3: 1.250000000000000000e-01
Q27.4: 6.250000000000000000e-02
Q26.5: 3.125000000000000000e-02
Q25.6: 1.562500000000000000e-02
Q24.7: 7.812500000000000000e-03
Q23.8: 3.906250000000000000e-03
Q22.9: 1.953125000000000000e-03
Q21.10: 9.765625000000000000e-04
Q20.11: 4.882812500000000000e-04
Q19.12: 2.441406250000000000e-04
Q18.13: 1.220703125000000000e-04
Q17.14: 6.103515625000000000e-05
Q16.15: 3.051757812500000000e-05
Q15.16: 1.525878906250000000e-05
Q14.17: 7.629394531250000000e-06
Q13.18: 3.814697265625000000e-06
Q12.19: 1.907348632812500000e-06
Q11.20: 9.536743164062500000e-07
Q10.21: 4.768371582031250000e-07
Q9.22: 2.384185791015625000e-07
Q8.23: 1.192092895507812500e-07
Q7.24: 5.960464477539062500e-08
Q6.25: 2.980232238769531250e-08
Q5.26: 1.490116119384765625e-08
Q4.27: 7.450580596923828125e-09
Q3.28: 3.725290298461914062e-09
Q2.29: 1.862645149230957031e-09
Q1.30: 9.313225746154785156e-10
Q0.31: 4.656612873077392578e-10And the exact covering ranges are;
Q30.1: [-1.073741824000000000e+09, 1.073741823500000000e+09]
Q29.2: [-5.368709120000000000e+08, 5.368709117500000000e+08]
Q28.3: [-2.684354560000000000e+08, 2.684354558750000000e+08]
Q27.4: [-1.342177280000000000e+08, 1.342177279375000000e+08]
Q26.5: [-6.710886400000000000e+07, 6.710886396875000000e+07]
Q25.6: [-3.355443200000000000e+07, 3.355443198437500000e+07]
Q24.7: [-1.677721600000000000e+07, 1.677721599218750000e+07]
Q23.8: [-8.388608000000000000e+06, 8.388607996093750000e+06]
Q22.9: [-4.194304000000000000e+06, 4.194303998046875000e+06]
Q21.10: [-2.097152000000000000e+06, 2.097151999023437500e+06]
Q20.11: [-1.048576000000000000e+06, 1.048575999511718750e+06]
Q19.12: [-5.242880000000000000e+05, 5.242879997558593750e+05]
Q18.13: [-2.621440000000000000e+05, 2.621439998779296875e+05]
Q17.14: [-1.310720000000000000e+05, 1.310719999389648438e+05]
Q16.15: [-6.553600000000000000e+04, 6.553599996948242188e+04]
Q15.16: [-3.276800000000000000e+04, 3.276799998474121094e+04]
Q14.17: [-1.638400000000000000e+04, 1.638399999237060547e+04]
Q13.18: [-8.192000000000000000e+03, 8.191999996185302734e+03]
Q12.19: [-4.096000000000000000e+03, 4.095999998092651367e+03]
Q11.20: [-2.048000000000000000e+03, 2.047999999046325684e+03]
Q10.21: [-1.024000000000000000e+03, 1.023999999523162842e+03]
Q9.22: [-5.120000000000000000e+02, 5.119999997615814209e+02]
Q8.23: [-2.560000000000000000e+02, 2.559999998807907104e+02]
Q7.24: [-1.280000000000000000e+02, 1.279999999403953552e+02]
Q6.25: [-6.400000000000000000e+01, 6.399999997019767761e+01]
Q5.26: [-3.200000000000000000e+01, 3.199999998509883881e+01]
Q4.27: [-1.600000000000000000e+01, 1.599999999254941940e+01]
Q3.28: [-8.000000000000000000e+00, 7.999999996274709702e+00]
Q2.29: [-4.000000000000000000e+00, 3.999999998137354851e+00]
Q1.30: [-2.000000000000000000e+00, 1.999999999068677425e+00]
Q0.31: [-1.000000000000000000e+00, 9.999999995343387127e-01] // Return the Q format with the range larger than given range
fixed16_qformat_t fixed16_get_qformat_covering_range(double fmin, double fmax);
// Return the Q format with the unit resolution finer than given epsilon
fixed16_qformat_t fixed16_get_qformat_covering_resolution(double epsilon);
// Generate the unit resolutions of Q16 format in float-typed container
void fixed16_generate_resolutions_in_float(void);
// Generate the covering ranges of Q16 format in float-typed container
void fixed16_generate_ranges_in_float(void);
// Generate the unit resolutions of Q16 format in double-typed container
void fixed16_generate_resolutions_in_double(void);
// Generate the covering ranges of Q16 format in double-typed container
void fixed16_generate_ranges_in_double(void);
// Conversion between 16bit Q formats
fixed16_t fixed16_convert(fixed16_t fix16, fixed16_qformat_t q1, fixed16_qformat_t q2);
// Conversion between 16bit Q format, and standard c-type
static inline fixed16_t uint8_to_fixed16(uint8_t v, fixed16_qformat_t q) { return (fixed16_t)v << q; }
static inline fixed16_t int8_to_fixed16(int8_t v, fixed16_qformat_t q) { return (fixed16_t)v << q; }
static inline fixed16_t uint16_to_fixed16(uint16_t v, fixed16_qformat_t q) { return (fixed16_t)v << q; }
static inline fixed16_t int16_to_fixed16(int16_t v, fixed16_qformat_t q) { return (fixed16_t)v << q; }
static inline fixed16_t uint32_to_fixed16(uint32_t v, fixed16_qformat_t q) { return (fixed16_t)v << q; }
static inline fixed16_t int32_to_fixed16(int32_t v, fixed16_qformat_t q) { return (fixed16_t)v << q; }
static inline fixed16_t float_to_fixed16(float v, fixed16_qformat_t q) { return (fixed16_t)(v * (1UL << q)); }
static inline fixed16_t double_to_fixed16(double v, fixed16_qformat_t q) { return (fixed16_t)(v * (1UL << q)); }
static inline uint8_t fixed16_to_uint8(fixed16_t fix16, fixed16_qformat_t q) { return (uint8_t)(fix16 >> q); }
static inline int8_t fixed16_to_int8(fixed16_t fix16, fixed16_qformat_t q) { return (int8_t)(fix16 >> q); }
static inline uint16_t fixed16_to_uint16(fixed16_t fix16, fixed16_qformat_t q) { return (uint16_t)(fix16 >> q); }
static inline int16_t fixed16_to_int16(fixed16_t fix16, fixed16_qformat_t q) { return (int16_t)(fix16 >> q); }
static inline uint32_t fixed16_to_uint32(fixed16_t fix16, fixed16_qformat_t q) { return (uint32_t)(fix16 >> q); }
static inline int32_t fixed16_to_int32(fixed16_t fix16, fixed16_qformat_t q) { return (int32_t)(fix16 >> q); }
static inline float fixed16_to_float(fixed16_t fix16, fixed16_qformat_t q) { return (float)fix16 / (1UL << q); }
static inline double fixed16_to_double(fixed16_t fix16, fixed16_qformat_t q) { return (double)fix16 / (1UL << q); }
// Arithmetic operations of 16bit Q Format
static inline fixed16_t fixed16_add(fixed16_t a, fixed16_t b) { fixed16_t c; c = a + b; return c; }
fixed16_t fixed16_add_with_trim(fixed16_t a, fixed16_t b);
static inline fixed16_t fixed16_subtract(fixed16_t a, fixed16_t b) { fixed16_t c; c = a - b; return c; }
fixed16_t fixed16_subtract_with_trim(fixed16_t a, fixed16_t b);
fixed16_t fixed16_multiply(fixed16_t a, fixed16_t b, fixed16_qformat_t q);
fixed16_t fixed16_divide(fixed16_t a, fixed16_t b, fixed16_qformat_t q); // Return the Q format with the range larger than given range
fixed32_qformat_t fixed32_get_qformat_covering_range(double fmin, double fmax);
// Return the Q format with the unit resolution finer than given epsilon
fixed32_qformat_t fixed32_get_qformat_covering_resolution(double epsilon);
// Generate the unit resolutions of Q32 format in float-typed container
void fixed32_generate_resolutions_in_float(void);
// Generate the covering ranges of Q32 format in float-typed container
void fixed32_generate_ranges_in_float(void);
// Generate the unit resolutions of Q32 format in double-typed container
void fixed32_generate_resolutions_in_double(void);
// Generate the covering ranges of Q32 format in double-typed container
void fixed32_generate_ranges_in_double(void);
// Conversion between 32bit Q formats
fixed32_t fixed32_convert(fixed32_t fix32, fixed32_qformat_t q1, fixed32_qformat_t q2);
// Conversion between 32bit Q format, and standard c-type
static inline fixed32_t uint8_to_fixed32(uint8_t v, fixed32_qformat_t q) { return (fixed32_t)v << q; }
static inline fixed32_t int8_to_fixed32(int8_t v, fixed32_qformat_t q) { return (fixed32_t)v << q; }
static inline fixed32_t uint16_to_fixed32(uint16_t v, fixed32_qformat_t q) { return (fixed32_t)v << q; }
static inline fixed32_t int16_to_fixed32(int16_t v, fixed32_qformat_t q) { return (fixed32_t)v << q; }
static inline fixed32_t uint32_to_fixed32(uint32_t v, fixed32_qformat_t q) { return (fixed32_t)v << q; }
static inline fixed32_t int32_to_fixed32(int32_t v, fixed32_qformat_t q) { return (fixed32_t)v << q; }
static inline fixed32_t float_to_fixed32(float v, fixed32_qformat_t q) { return (fixed32_t)(v * (1UL << q)); }
static inline fixed32_t double_to_fixed32(double v, fixed32_qformat_t q) { return (fixed32_t)(v * (1UL << q)); }
static inline uint8_t fixed32_to_uint8(fixed32_t fix32, fixed32_qformat_t q) { return (uint8_t)(fix32 >> q); }
static inline int8_t fixed32_to_int8(fixed32_t fix32, fixed32_qformat_t q) { return (int8_t)(fix32 >> q); }
static inline uint16_t fixed32_to_uint16(fixed32_t fix32, fixed32_qformat_t q) { return (uint16_t)(fix32 >> q); }
static inline int16_t fixed32_to_int16(fixed32_t fix32, fixed32_qformat_t q) { return (int16_t)(fix32 >> q); }
static inline uint32_t fixed32_to_uint32(fixed32_t fix32, fixed32_qformat_t q) { return (uint32_t)(fix32 >> q); }
static inline int32_t fixed32_to_int32(fixed32_t fix32, fixed32_qformat_t q) { return (int32_t)(fix32 >> q); }
static inline float fixed32_to_float(fixed32_t fix32, fixed32_qformat_t q) { return (float)fix32 / (1UL << q); }
static inline double fixed32_to_double(fixed32_t fix32, fixed32_qformat_t q) { return (double)fix32 / (1UL << q); }
// Arithmetic operations of 32bit Q Format
static inline fixed32_t fixed32_add(fixed32_t a, fixed32_t b) { fixed32_t c; c = a + b; return c; }
fixed32_t fixed32_add_with_trim(fixed32_t a, fixed32_t b);
static inline fixed32_t fixed32_subtract(fixed32_t a, fixed32_t b) { fixed32_t c; c = a - b; return c; }
fixed32_t fixed32_subtract_with_trim(fixed32_t a, fixed32_t b);
fixed32_t fixed32_multiply(fixed32_t a, fixed32_t b, fixed32_qformat_t q);
fixed32_t fixed32_divide(fixed32_t a, fixed32_t b, fixed32_qformat_t q);CORDIC-based