DFT - 2023.2 English

Following table gives results for the DFT function with a wide variety of supported parameters, which are defined in: DFT configuration parameters.

dft_benchmark.csv

Table 71 DFT benchmark

Library Element	AIE_VARIANT	DATA_TYPE	TWIDDLE_TYPE	POINT_SIZE	FFT_NIFFT	CASC_LEN	NUM_FRAMES	WINDOW_VSIZE	API_IO	PARALLEL_POWER	Latency	Throughput	NUM_BANKS	NUM_AIE	DATA_MEMORY	PROGRAM_MEMORY
dft	1	cfloat	cfloat	15	0	1	16	256	0	0	6876 ns	104 MSa/s	6	1	14639	2036
dft	1	cfloat	cfloat	15	0	1	48	768	0	0	20134 ns	106 MSa/s	6	1	31023	2036
dft	1	cfloat	cfloat	16	0	4	16	256	0	0	2528 ns	303 MSa/s	17	4	26809	1740 1714 1714 1826
dft	1	cfloat	cfloat	16	0	4	32	512	0	0	4781 ns	312 MSa/s	18	4	35001	1740 1714 1714 1826
dft	1	cfloat	cfloat	16	0	4	48	768	0	0	7034 ns	316 MSa/s	18	4	43193	1740 1714 1714 1826
dft	1	cfloat	cfloat	27	0	1	1	28	0	0	1494 ns	56 MSa/s	6	1	11471	2740
dft	1	cfloat	cfloat	27	0	1	16	448	0	0	20690 ns	62 MSa/s	6	1	24911	2766
dft	1	cfloat	cfloat	32	0	1	1	32	0	0	1655 ns	57 MSa/s	6	1	13743	1832
dft	1	cfloat	cfloat	32	0	4	16	512	0	0	8021 ns	191 MSa/s	18	4	36537	1756 1764 1764 1876
dft	1	cfloat	cfloat	32	0	4	32	1024	0	0	15621 ns	193 MSa/s	18	4	52921	1740 1764 1764 1876
dft	1	cfloat	cfloat	32	0	4	48	1536	0	0	23212 ns	194 MSa/s	20	4	69305	1740 1764 1764 1876
dft	1	cfloat	cfloat	48	0	1	1	48	0	0	3362 ns	42 MSa/s	8	1	24495	1832
dft	1	cfloat	cfloat	48	0	1	8	384	0	0	25329 ns	44 MSa/s	8	1	35247	1848
dft	1	cfloat	cfloat	64	0	1	1	64	0	0	5740 ns	33 MSa/s	9	1	39343	1832
dft	1	cfloat	cfloat	64	0	12	32	2048	0	0	19612 ns	311 MSa/s	52	12	158249	1914 1968 1968 1968 1884 1884 1884 1884 1884 1884 1884 1996
dft	1	cfloat	cfloat	64	0	1	8	512	0	0	44325 ns	33 MSa/s	9	1	53679	1848
dft	1	cfloat	cfloat	64	0	8	16	1024	0	0	15682 ns	197 MSa/s	33	8	73073	1756 1764 1764 1764 1764 1764 1764 1876
dft	1	cint16	cint16	15	0	1	1	16	0	0	531 ns	153 MSa/s	6	1	5488	1662
dft	1	cint16	cint16	15	0	1	48	768	0	0	7355 ns	285 MSa/s	6	1	17520	1794
dft	1	cint16	cint16	15	0	1	1	16	0	0	531 ns	153 MSa/s	6	1	5488	1662
dft	1	cint16	cint16	15	0	1	16	256	0	0	2607 ns	274 MSa/s	6	1	9328	1794
dft	1	cint16	cint16	16	0	4	16	256	0	0	817 ns	1066 MSa/s	18	4	23482	1734 1758 1758 1868
dft	1	cint16	cint16	16	0	4	32	512	0	0	1580 ns	1091 MSa/s	17	4	29626	1718 1758 1758 1868
dft	1	cint16	cint16	16	0	4	48	768	0	0	2355 ns	1001 MSa/s	18	4	35770	1718 1758 1758 1868
dft	1	cint16	cint16	27	0	1	1	32	0	0	711 ns	179 MSa/s	6	1	8240	1992
dft	1	cint16	cint16	27	0	1	16	512	0	0	6103 ns	234 MSa/s	6	1	15920	2024
dft	1	cint16	cint16	32	0	4	16	512	0	0	2439 ns	626 MSa/s	18	4	26298	1750 1798 1798 1920
dft	1	cint16	cint16	32	0	4	32	1024	0	0	4616 ns	646 MSa/s	18	4	34490	1734 1798 1798 1920
dft	1	cint16	cint16	32	0	4	48	1536	0	0	6785 ns	654 MSa/s	18	4	42682	1734 1798 1798 1920
dft	1	cint16	cint16	32	0	1	1	32	0	0	756 ns	168 MSa/s	6	1	8880	1856
dft	1	cint16	cint16	48	0	1	1	48	0	0	1031 ns	141 MSa/s	7	1	14256	1856
dft	1	cint16	cint16	48	0	1	8	384	0	0	6838 ns	160 MSa/s	7	1	19632	1888
dft	1	cint16	cint16	64	0	12	32	2048	0	0	6802 ns	899 MSa/s	50	12	103210	1798 1862 1862 1862 1782 1782 1782 1782 1782 1782 1782 1888
dft	1	cint16	cint16	64	0	12	36	2304	0	0	7610 ns	902 MSa/s	52	12	109354	1798 1862 1862 1862 1782 1782 1782 1782 1782 1782 1782 1888
dft	1	cint16	cint16	64	0	8	16	1024	0	0	4232 ns	739 MSa/s	34	8	52594	1750 1798 1798 1798 1798 1798 1798 1920
dft	1	cint16	cint16	64	0	1	1	64	0	0	1633 ns	117 MSa/s	7	1	21680	1856
dft	1	cint16	cint16	64	0	1	8	512	0	0	11660 ns	126 MSa/s	7	1	28848	1888
dft	1	cint16	cint16	72	0	1	1	72	0	0	2019 ns	106 MSa/s	8	1	26160	2790
dft	1	cint16	cint16	72	0	1	8	576	0	0	14588 ns	114 MSa/s	8	1	34224	2808
dft	1	cint16	cint16	87	0	1	1	88	0	0	4896 ns	53 MSa/s	9	1	36304	2990
dft	1	cint16	cint16	87	0	1	8	704	0	0	37697 ns	55 MSa/s	9	1	46160	3006
dft	1	cint32	cint16	15	0	1	1	16	0	0	547 ns	115 MSa/s	6	1	5744	1776
dft	1	cint32	cint16	15	0	1	48	768	0	0	11844 ns	172 MSa/s	6	1	29808	1808
dft	1	cint32	cint16	15	0	1	1	16	0	0	547 ns	115 MSa/s	6	1	5744	1776
dft	1	cint32	cint16	15	0	1	16	256	0	0	4099 ns	168 MSa/s	6	1	13424	1808
dft	1	cint32	cint16	16	0	4	16	256	0	0	1878 ns	402 MSa/s	18	4	25530	1566 1598 1598 1696
dft	1	cint32	cint16	16	0	4	32	512	0	0	3501 ns	419 MSa/s	18	4	33722	1566 1598 1598 1696
dft	1	cint32	cint16	16	0	4	48	768	0	0	5135 ns	425 MSa/s	18	4	41914	1566 1598 1598 1696
dft	1	cint32	cint16	27	0	1	1	28	0	0	815 ns	104 MSa/s	6	1	8192	2152
dft	1	cint32	cint16	27	0	1	16	448	0	0	9937 ns	124 MSa/s	6	1	21632	2164
dft	1	cint32	cint16	32	0	4	16	512	0	0	5559 ns	270 MSa/s	18	4	34490	1584 1600 1600 1716
dft	1	cint32	cint16	32	0	4	32	1024	0	0	10811 ns	274 MSa/s	18	4	50874	1568 1600 1600 1716
dft	1	cint32	cint16	32	0	4	48	1536	0	0	16066 ns	275 MSa/s	20	4	67258	1568 1600 1600 1716
dft	1	cint32	cint16	32	0	1	1	32	0	0	969 ns	99 MSa/s	6	1	9392	1648
dft	1	cint32	cint16	48	0	1	1	48	0	0	1836 ns	76 MSa/s	7	1	15024	1648
dft	1	cint32	cint16	48	0	1	8	384	0	0	13282 ns	82 MSa/s	7	1	25776	1664
dft	1	cint32	cint16	64	0	12	32	2048	0	0	12107 ns	500 MSa/s	52	12	152362	1770 1846 1846 1846 1766 1766 1766 1766 1766 1766 1766 1888
dft	1	cint32	cint16	64	0	8	16	1024	0	0	10824 ns	282 MSa/s	34	8	68978	1584 1600 1600 1600 1600 1600 1600 1716
dft	1	cint32	cint16	64	0	1	1	64	0	0	3041 ns	61 MSa/s	7	1	22704	1648
dft	1	cint32	cint16	64	0	1	8	512	0	0	22926 ns	64 MSa/s	7	1	37040	1664
dft	1	cint32	cint16	72	0	1	1	72	0	0	3773 ns	56 MSa/s	8	1	27312	1648
dft	1	cint32	cint16	72	0	1	8	576	0	0	28746 ns	57 MSa/s	8	1	43440	1664
dft	1	cint32	cint16	87	0	1	1	88	0	0	5466 ns	47 MSa/s	9	1	37712	1776
dft	1	cint32	cint16	87	0	1	8	704	0	0	42354 ns	48 MSa/s	9	1	57424	1808
dft	2	cint16	cint16	16	0	1	10	160	0	0	1369 ns	328 MSa/s	5	1	7832	1584
dft	2	cint16	cint16	16	0	1	1	16	0	0	481 ns	202 MSa/s	5	1	5528	1552
dft	2	cint16	cint16	32	0	4	40	1280	0	0	7445 ns	504 MSa/s	18	4	38470	1504 1520 1520 1584
dft	2	cint16	cint16	36	0	8	100	4000	0	0	14674 ns	787 MSa/s	36	8	118558	1408 1424 1424 1424 1392 1392 1392 1408
dft	2	cint16	cint16	67	0	5	1	72	0	0	2111 ns	229 MSa/s	21	5	30196	1712 1744 1744 1744 1824
dft	2	cint16	cint16	67	0	5	4	288	0	0	3795 ns	241 MSa/s	22	5	33844	1760 1776 1760 1760 1840
dft	2	cint32	cint16	16	0	1	10	160	0	0	1361 ns	298 MSa/s	5	1	10392	1552
dft	2	cint32	cint16	16	0	1	1	16	0	0	342 ns	192 MSa/s	5	1	5784	1520
dft	2	cint32	cint16	32	0	4	40	1280	0	0	7418 ns	500 MSa/s	19	4	58950	1456 1472 1472 1504
dft	2	cint32	cint16	36	0	8	100	4000	0	0	26040 ns	499 MSa/s	40	8	201758	1376 1392 1392 1392 1360 1360 1360 1424
dft	2	cint32	cint16	67	0	5	1	72	0	0	1519 ns	224 MSa/s	21	5	31412	1664 1664 1648 1648 1712
dft	2	cint32	cint16	67	0	5	4	288	0	0	3765 ns	240 MSa/s	22	5	38708	1680 1712 1696 1696 1760