Matrix Multiply - 2024.1 English

The following table gives results for the Matrix Multiply function with a wide variety of supported parameters, which are defined in: Matrix Multiply Configuration Parameters.

matrix_mult_benchmark.csv

Table 89 Matrix Multiply benchmark

Library Element	T_DATA_A	T_DATA_B	P_DIM_A	P_DIM_AB	P_DIM_B	P_ADD_TILING_A	P_ADD_TILING_B	P_ADD_DETILING_OUT	P_INPUT_WINDOW_VSIZE_A	P_INPUT_WINDOW_VSIZE_B	P_CASC_LEN	UUT_SSR	AIE_VARIANT	Dynamic Power	Latency	Throughput	NUM_BANKS	NUM_AIE	DATA_MEMORY	PROGRAM_MEMORY
matrix_mult	cint16	cint16	16	64	16	0	0	0	1024	1024	1	1	1	0.797 W	4622 ns	137 MSa/s	7	1	20617	2198
matrix_mult	int16	int16	16	64	16	0	0	0	1024	1024	4	2	1	1.213 W	540 ns	1706 MSa/s	42	8	30762	1788 1820 1820 1916 1788 1820 1820 1916
matrix_mult	int16	int16	16	64	16	0	0	0	1024	1024	1	1	2	0.512 W	859 ns	500 MSa/s	7	1	11406	1536
matrix_mult	int16	int16	16	64	16	1	1	1	1024	1024	1	1	2	0.677 W	2173 ns	411 MSa/s	11	3	24986	1744 1536 1952
matrix_mult	int16	int16	16	64	16	0	0	0	1024	1024	1	2	2	0.644 W	1353 ns	500 MSa/s	14	2	17692	1504 1504
matrix_mult	int16	int16	16	64	16	0	0	0	1024	1024	2	1	2	0.651 W	534 ns	1000 MSa/s	11	2	13587	1456 1584
matrix_mult	int16	int16	16	64	16	0	0	0	1024	1024	4	2	2	1.591 W	457 ns	2000 MSa/s	42	8	30778	1216 1440 1440 1552 1216 1440 1440 1552
matrix_mult	int16	int16	8	64	4	0	0	0	512	256	4	1	1	0.885 W	389 ns	372 MSa/s	22	4	11925	1410 1410 1410 1746
matrix_mult	int16	int32	8	64	4	0	0	0	512	256	4	1	1	0.859 W	419 ns	316 MSa/s	19	4	13082	1650 1682 1682 1770
matrix_mult	int32	int16	16	64	16	0	0	0	1024	1024	1	1	1	0.792 W	2049 ns	250 MSa/s	7	1	16521	2000
matrix_mult	int32	int16	16	64	16	1	1	1	1024	1024	1	1	1	0.821 W	8688 ns	117 MSa/s	12	3	35220	3394 2000 4054
matrix_mult	int32	int16	16	64	16	0	0	0	1024	1024	1	2	1	0.912 W	1131 ns	484 MSa/s	14	2	22802	1932 1932
matrix_mult	int32	int16	16	64	16	0	0	0	1024	1024	2	1	1	0.852 W	1473 ns	410 MSa/s	12	2	18703	1866 1984
matrix_mult	int32	int16	16	64	16	0	0	0	1024	1024	4	2	1	1.269 W	738 ns	1084 MSa/s	40	8	35894	1824 1840 1840 1916 1824 1840 1840 1916
matrix_mult	int16	int16	16	64	16	0	0	0	1024	1024	2	1	1	0.841 W	914 ns	725 MSa/s	12	2	13579	1828 1984
matrix_mult	int32	int16	16	64	16	0	0	0	1024	1024	1	1	2	0.522 W	1613 ns	250 MSa/s	6	1	16526	1776
matrix_mult	int32	int16	16	64	16	0	0	0	1024	1024	1	2	2	0.701 W	843 ns	500 MSa/s	12	2	22812	1728 1728
matrix_mult	int32	int16	16	64	16	0	0	0	1024	1024	2	1	2	0.689 W	944 ns	500 MSa/s	11	2	18707	1648 1824
matrix_mult	int32	int16	16	64	16	0	0	0	1024	1024	4	2	2	1.674 W	482 ns	1815 MSa/s	42	8	35898	1616 1648 1648 1776 1616 1648 1648 1776
matrix_mult	int32	int16	8	64	4	0	0	0	512	256	4	1	1	0.898 W	367 ns	250 MSa/s	22	4	14107	1720 1736 1736 1808
matrix_mult	int32	int32	16	64	16	0	0	0	1024	1024	1	1	1	0.795 W	4622 ns	137 MSa/s	7	1	20617	2198
matrix_mult	int32	int32	16	64	16	1	1	1	1024	1024	1	1	1	0.861 W	9214 ns	121 MSa/s	11	3	43412	3394 2182 3750
matrix_mult	int32	int32	16	64	16	0	0	0	1024	1024	1	2	1	0.893 W	2619 ns	250 MSa/s	14	2	30994	2130 2130
matrix_mult	int32	int32	16	64	16	0	0	0	1024	1024	2	1	1	0.915 W	2885 ns	238 MSa/s	12	2	22801	2140 2280
matrix_mult	int32	int32	16	64	16	0	0	0	1024	1024	4	2	1	1.321 W	1209 ns	703 MSa/s	41	8	44098	2098 2186 2186 2212 2098 2186 2186 2212
matrix_mult	int32	int32	16	64	16	0	0	0	1024	1024	1	1	2	0.58 W	2127 ns	248 MSa/s	6	1	20622	1984
matrix_mult	int32	int32	16	64	16	1	1	1	1024	1024	1	1	2	0.768 W	3223 ns	250 MSa/s	13	3	43417	1344 1536 1984
matrix_mult	int32	int32	16	64	16	0	0	0	1024	1024	1	2	2	0.724 W	2629 ns	250 MSa/s	14	2	31004	1936 1936
matrix_mult	int32	int32	16	64	16	0	0	0	1024	1024	2	1	2	0.773 W	1508 ns	423 MSa/s	11	2	22803	1824 2016
matrix_mult	int32	int16	16	64	16	1	1	1	1024	1024	1	1	2	0.702 W	2582 ns	250 MSa/s	13	3	35225	1344 1776 1776
matrix_mult	int32	int32	16	64	16	0	0	0	1024	1024	4	2	2	1.658 W	901 ns	1000 MSa/s	42	8	44090	1808 1856 1856 1968 1808 1856 1856 1968
matrix_mult	int16	int16	16	64	16	0	0	0	1024	1024	1	2	1	0.844 W	1339 ns	500 MSa/s	14	2	17676	1884 1884
matrix_mult	int16	int16	16	64	16	0	0	0	1024	1024	1	1	1	0.772 W	1177 ns	471 MSa/s	7	1	11398	1952
matrix_mult	cint16	cint16	16	64	16	1	1	1	1024	1024	1	1	1	0.864 W	9214 ns	121 MSa/s	11	3	43412	3394 2182 3750
matrix_mult	cint16	cint16	16	64	16	0	0	0	1024	1024	1	2	1	0.907 W	2619 ns	250 MSa/s	14	2	30994	2130 2130
matrix_mult	cint16	cint16	16	64	16	0	0	0	1024	1024	2	1	1	0.9 W	2885 ns	238 MSa/s	12	2	22801	2140 2280
matrix_mult	cint16	cint16	16	64	16	0	0	0	1024	1024	4	2	1	1.325 W	1209 ns	703 MSa/s	41	8	44098	2098 2186 2186 2212 2098 2186 2186 2212
matrix_mult	cint16	cint16	16	64	16	0	0	0	1024	1024	1	1	2	0.58 W	2047 ns	250 MSa/s	6	1	20622	1728
matrix_mult	cint16	cint16	16	64	16	1	1	1	1024	1024	1	1	2	0.734 W	2902 ns	250 MSa/s	13	3	43417	1088 1744 1728
matrix_mult	cint16	cint16	16	64	16	0	0	0	1024	1024	1	2	2	0.725 W	2629 ns	250 MSa/s	14	2	31004	1728 1728
matrix_mult	cint16	cint16	16	64	16	0	0	0	1024	1024	2	1	2	0.771 W	1421 ns	426 MSa/s	11	2	22803	1664 1792
matrix_mult	cint16	cint16	16	64	16	0	0	0	1024	1024	4	2	2	1.733 W	817 ns	1000 MSa/s	42	8	44090	1648 1664 1664 1792 1648 1664 1664 1792
matrix_mult	cint16	cint16	4	64	8	0	0	0	256	512	4	1	1	0.922 W	553 ns	217 MSa/s	20	4	15137	1730 1752 1752 1822
matrix_mult	cint16	int16	8	64	4	0	0	0	512	256	4	1	1	0.887 W	367 ns	250 MSa/s	22	4	14107	1720 1736 1736 1808
matrix_mult	cint32	cint16	4	64	8	0	0	0	256	512	4	1	1	0.939 W	712 ns	160 MSa/s	22	4	17435	1764 1780 1780 1856
matrix_mult	cint32	cint32	16	32	16	0	0	0	512	512	1	1	1	0.802 W	9936 ns	70 MSa/s	7	1	22662	2158
matrix_mult	int16	int16	16	64	16	1	1	1	1024	1024	1	1	1	0.8 W	7936 ns	129 MSa/s	13	3	24977	4240 1936 3746
matrix_mult	cint32	cint32	16	32	16	1	1	1	512	512	1	1	1	0.841 W	17245 ns	70 MSa/s	12	3	47505	1586 2160 1554
matrix_mult	cint32	cint32	16	32	16	0	0	0	512	512	2	1	1	0.926 W	5918 ns	122 MSa/s	12	2	24845	2072 2192
matrix_mult	cint32	cint32	16	32	16	0	0	0	512	512	4	2	1	1.448 W	2186 ns	372 MSa/s	41	8	46134	2088 2160 2160 2208 2088 2160 2160 2208
matrix_mult	cint32	cint32	16	32	16	0	0	0	512	512	1	1	2	0.61 W	4549 ns	139 MSa/s	7	1	22670	1968
matrix_mult	cint32	cint32	16	32	16	1	1	1	512	512	1	1	2	0.76 W	7781 ns	139 MSa/s	13	3	47513	1088 1776 1968
matrix_mult	cint32	cint32	16	32	16	0	0	0	512	512	1	2	2	0.846 W	2619 ns	246 MSa/s	12	2	33052	1968 1968
matrix_mult	cint32	cint32	16	32	16	0	0	0	512	512	2	1	2	0.818 W	2905 ns	237 MSa/s	10	2	24851	1840 2016
matrix_mult	cint32	cint32	16	32	16	0	0	0	512	512	4	2	2	1.838 W	1199 ns	713 MSa/s	42	8	46138	1840 1872 1872 2032 1840 1872 1872 2032
matrix_mult	cint32	cint32	4	64	8	0	0	0	256	512	4	1	1	0.935 W	1242 ns	100 MSa/s	21	4	21531	2082 2130 2130 2160
matrix_mult	float	cfloat	16	32	16	0	0	0	512	512	1	1	1	0.772 W	11583 ns	63 MSa/s	7	1	18824	2582
matrix_mult	float	cfloat	16	32	16	1	1	1	512	512	1	1	1	0.81 W	19310 ns	63 MSa/s	13	3	40083	1586 2566 2910
matrix_mult	float	cfloat	16	32	16	0	0	0	512	512	1	2	1	0.863 W	5921 ns	125 MSa/s	14	2	29456	2582 2582
matrix_mult	float	cfloat	16	32	16	0	0	0	512	512	2	1	1	0.871 W	6978 ns	109 MSa/s	12	2	21262	2448 2646
matrix_mult	float	cfloat	16	32	16	0	0	0	512	512	4	2	1	1.352 W	3158 ns	257 MSa/s	42	8	44084	2642 2514 2514 3138 2642 2514 2514 3138
matrix_mult	cint32	cint32	16	32	16	0	0	0	512	512	1	2	1	0.943 W	5095 ns	138 MSa/s	14	2	33036	2158 2158
matrix_mult	int32	int32	8	64	4	0	0	0	512	256	4	1	1	0.869 W	556 ns	250 MSa/s	18	4	15137	2006 2094 2094 2138