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	1047	432	1319	1311	1563	1047	432	1320	4117	2336
dsps_dotprode_f32 for N=256 points with step 1	1307	1307	1314	2581	1819	1308	1309	1317	3609	2087
dsps_dotprod_s16 for N=256 points	437	307	208	3635	3385	437	307	202	6708	4170
FIR Filters
dsps_fir_f32 1024 input samples and 256 coefficients	1078691	443671	1606563	1353122	1606552	1078973	443680	3050985	4497990	3050972
dsps_fird_f32 1024 samples 256 coeffs and decimation 4	276504	115499	339796	350726	415332	277204	115477	339780	1206798	814168
FFTs Radix-2 32 bit Floating Point
dsps_fft2r_fc32 for 64 complex points	4544	3970	4989	7093	6062	4545	3971	4988	8032	8144
dsps_fft2r_fc32 for 128 complex points	10342	8999	11404	16091	13725	10342	8998	11391	18192	18622
dsps_fft2r_fc32 for 256 complex points	23210	20139	25684	36066	30779	23210	20250	25671	40737	41955
dsps_fft2r_fc32 for 512 complex points	51503	44594	57181	79978	68325	51503	44594	57167	90424	93483
dsps_fft2r_fc32 for 1024 complex points	113205	97847	126053	175729	140083	113205	97847	126039	198336	186812
FFTs Radix-4 32 bit Floating Point
dsps_fft4r_fc32 for 64 complex points	3004	2597	3483	5035	3761	3004	2597	3480	5681	3876
dsps_fft4r_fc32 for 256 complex points	15348	13213	18052	25198	19266	15347	13212	18036	28714	19788
dsps_fft4r_fc32 for 1024 complex points	75068	64482	89820	122592	95550	75068	64482	89766	140116	98026
FFTs 16 bit Fixed Point
dsps_fft2r_sc16 for 64 complex points	8775	774	897	9287	9055	8775	775	894	10603	9472
dsps_fft2r_sc16 for 128 complex points	20205	1608	1832	20798	20675	20204	1610	1835	24067	21633
dsps_fft2r_sc16 for 256 complex points	45746	3412	3873	46164	46576	45745	3410	3876	54611	48735
dsps_fft2r_sc16 for 512 complex points	102198	7294	8234	101611	103739	102199	7293	8237	119792	108555
dsps_fft2r_sc16 for 1024 complex points	226154	15623	17523	221955	228808	225852	15624	17526	263431	239447
IIR Filters
dsps_biquad_f32 - biquad filter for 1024 input samples	17442	17552	15391	26651	21544	17441	17441	15403	36883	32789
Matrix Multiplication
dspm_mult_f32 - C[16;16] = A[16;16]*B[16;16]	24659	6280	28276	56915	31481	24660	6282	28239	66482	38913
dspm_mult_s16 - C[16;16] = A[16;16]*B[16;16]	24697	2004	2138	71112	60715	24696	1831	2141	126689	63865
dspm_mult_3x3x1_f32 - C[3;1] = A[3;3]*B[3;1]	70	71	144	247	161	69	131	159	258	164
dspm_mult_3x3x3_f32 - C[3;3] = A[3;3]*B[3;3]	200	200	338	560	309	201	199	332	559	351
dspm_mult_4x4x1_f32 - C[4;1] = A[4;4]*B[4;1]	102	103	206	371	227	103	104	202	398	258
dspm_mult_4x4x4_f32 - C[4;4] = A[4;4]*B[4;4]	395	175	658	1150	608	394	174	658	1189	749
Image processing prototypes
dspi_dotprod_s8/u8 - dotproduct of two images 16x16	2635	161	225	2633	2436	4021	304	228	4019	2373
dspi_dotprod_off_s8/u8 - dotproduct of two images 16x16	2891	228	240	2891	2736	4802	227	243	4801	2584
dspi_dotprod_s8/u8- dotproduct of two images 64x64	37962	689	1662	37962	33723	58885	687	1664	58883	33800
dspi_dotprod_off_s8/u8 - dotproduct of two images 64x64	42059	995	1792	42058	37884	71187	995	1795	71186	37768
dspi_dotprod_s16/u16 - dotproduct of two images 8x8	1189	163	153	1190	994	1749	146	152	1750	1018
dspi_dotprod_off_s16/u16 - dotproduct of two images 8x8	1258	362	145	1258	1056	2647	179	143	2646	1088
dspi_dotprod_s16 - dotproduct of two images 32x32	15181	409	853	15181	13888	24334	407	859	24964	13831
dspi_dotprod_off_s16/u16 - dotproduct of two images 32x32	16210	561	921	16209	14874	28462	558	921	28462	14850

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