Espressif DSP Library Benchmarks

The table bellow contains benchmarks of functions provided by ESP-DSP library. The values are CPU cycle counts taken to execute each of the functions. The Values in the column “O2” are made with compiler optimization for speed, and in the column “Os” column are made with compiler optimization for size. The values in “ESP32”, “ESP32S3” and “ESP32P4” columns are for the optimized (assembly) implementation, values in “ANSI” column are for the non-optimized implementation.

Function Optimization	ESP32 O2	ESP32S3 O2	ESP32P4 O2	ANSI Xtensa O2	ANSI Risc-V O2	ESP32 Os	ESP32S3 Os	ESP32P4 Os	ANSI Xtensa Os	ANSI Risc-V Os
Dot Product
dsps_dotprod_f32 for N=256 points	1058	449	1319	1325	1563	1058	448	1320	4129	2336
dsps_dotprode_f32 for N=256 points with step 1	1317	1326	1314	2853	1819	1317	1326	1317	3621	2087
dsps_dotprod_s16 for N=256 points	447	323	208	3647	3385	447	322	202	6466	4170
FIR Filters
dsps_fir_f32 1024 input samples and 256 coefficients	1079312	443690	1606563	2150685	1606552	1079599	443688	3050985	5147785	3050972
dsps_fird_f32 1024 samples 256 coeffs and decimation 4	350915	115494	339796	614234	415332	350520	115495	339780	1317367	814168
FFTs Radix-2 32 bit Floating Point
dsps_fft2r_fc32 for 64 complex points	6079	5142	4989	7037	6062	5452	5140	4988	8333	8144
dsps_fft2r_fc32 for 128 complex points	13031	11706	11404	15907	13725	12399	11705	11391	19035	18622
dsps_fft2r_fc32 for 256 complex points	27828	26303	25684	35562	30779	27828	26303	25671	42922	41955
dsps_fft2r_fc32 for 512 complex points	61753	58435	57181	78705	68325	61753	58437	57167	95673	93483
dsps_fft2r_fc32 for 1024 complex points	135742	128697	126053	172664	140083	135742	128585	126039	211252	186812
FFTs Radix-4 32 bit Floating Point
dsps_fft4r_fc32 for 64 complex points	3125	3345	3483	5185	3761	3247	3176	3480	5631	3876
dsps_fft4r_fc32 for 256 complex points	15551	15522	18052	26115	19266	16056	15792	18036	28397	19788
dsps_fft4r_fc32 for 1024 complex points	75547	75273	89820	127669	95550	77587	76554	89766	138522	98026
FFTs 16 bit Fixed Point
dsps_fft2r_sc16 for 64 complex points	8786	793	897	14575	9055	8786	794	894	15861	9472
dsps_fft2r_sc16 for 128 complex points	20214	1828	1832	33121	20675	20214	1627	1835	36238	21633
dsps_fft2r_sc16 for 256 complex points	45755	3429	3873	74290	46576	45755	3428	3876	81638	48735
dsps_fft2r_sc16 for 512 complex points	102208	7512	8234	164803	103739	102416	7311	8237	181758	108555
dsps_fft2r_sc16 for 1024 complex points	225862	15641	17523	362193	228808	225861	15642	17526	400853	239447
IIR Filters
dsps_biquad_f32 - biquad filter for 1024 input samples	17450	17459	15391	24613	21544	17451	17630	15403	36895	32789
Matrix Multiplication
dspm_mult_f32 - C[16;16] = A[16;16]*B[16;16]	24669	6297	28276	51502	31481	24670	6498	28239	78197	38913
dspm_mult_s16 - C[16;16] = A[16;16]*B[16;16]	24707	1848	2138	83699	60715	24707	1848	2141	99353	63865
dspm_mult_3x3x1_f32 - C[3;1] = A[3;3]*B[3;1]	79	275	144	226	161	80	86	159	271	164
dspm_mult_3x3x3_f32 - C[3;3] = A[3;3]*B[3;3]	211	323	338	492	309	210	217	332	611	351
dspm_mult_4x4x1_f32 - C[4;1] = A[4;4]*B[4;1]	112	119	206	334	227	113	121	202	425	258
dspm_mult_4x4x4_f32 - C[4;4] = A[4;4]*B[4;4]	405	192	658	1008	608	404	194	658	1335	749
Image processing prototypes
dspi_dotprod_s8/u8 - dotproduct of two images 16x16	3827	381	225	3828	2436	4010	179	228	4011	2373
dspi_dotprod_off_s8/u8 - dotproduct of two images 16x16	4142	243	240	4142	2736	4772	245	243	4774	2584
dspi_dotprod_s8/u8- dotproduct of two images 64x64	58069	705	1662	58068	33723	58825	705	1664	58826	33800
dspi_dotprod_off_s8/u8 - dotproduct of two images 64x64	62365	1010	1792	62366	37884	71233	1177	1795	71062	37768
dspi_dotprod_s16/u16 - dotproduct of two images 8x8	1455	162	153	1453	994	1804	162	152	1806	1018
dspi_dotprod_off_s16/u16 - dotproduct of two images 8x8	1529	196	145	1531	1056	2074	195	143	2074	1088
dspi_dotprod_s16 - dotproduct of two images 32x32	20190	565	853	20029	13888	25300	566	859	25301	13831
dspi_dotprod_off_s16/u16 - dotproduct of two images 32x32	21090	576	921	21089	14874	29432	743	921	29432	14850

The benchmark test could be reproduced by executing test cases found in /test/test_dsp.c.