NPRG054 Assignment - Deep Convolutional Neural Network

Deep Convolutional Neural Network

Deadline: 2026/06/30 23:59
This assignment is tested with -n 1 -c 64 Slurm settings
Points will be awarded for speed-up wrt. the public initial solution, logarithmically (400 pts for ten-fold speed-up)
Doxymentation
Layer sizes and complexity

Note: If your solution does not appear in the table, it is because it did not compile.

If it compiles well in your worktree, the difference is because your worktree is not up-to-date with respect to the asgn (upstream) repository. In this case, you should update your worktree using the following command, resolve the conflicts, make sure it compiles, and push to your (origin) repository.

git pull git@gitlab.mff.cuni.cz:teaching/nprg054/asgn.git

machine	author	commit	commit date	platform					loss	time	check	rel time	points
mpi-homo	andrejcd	a1a831b	"2026-06-25 00:13:05 +0200"	avx512	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.128019	OK	0.747821	50.481
mpi-homo	asgn	6ef8d97	"2026-04-18 21:17:08 +0200"	avx512	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.171312	OK	1.00072	-0.124164
mpi-homo	asgn	6ef8d97	"2026-04-18 21:17:08 +0200"	avx	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.169272	OK	1.00219	-0.379234
mpi-homo	balkoad	4696c1d	"2026-04-25 02:42:27 +0200"	avx512	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.170799	OK	0.99772	0.396509
mpi-homo	balkoad	4696c1d	"2026-04-25 02:42:27 +0200"	avx	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.16888	OK	0.999865	0.0234209
mpi-homo	bambasz	4d2e6c0	"2026-03-07 14:47:14 +0100"	avx512	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.171122	OK	0.999603	0.0689583
mpi-homo	bambasz	4d2e6c0	"2026-03-07 14:47:14 +0100"	avx	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.168959	OK	1.00033	-0.0569748
mpi-homo	bendaja1	6f40a23	"2026-05-11 15:27:31 +0200"	avx512	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.170902	OK	0.998323	0.291623
mpi-homo	bendaja1	6f40a23	"2026-05-11 15:27:31 +0200"	avx	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.168841	OK	0.999632	0.0638535
mpi-homo	beranst	5f13dd8	"2026-05-10 18:47:47 +0200"	avx512	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.1708	OK	0.997728	0.395213
mpi-homo	beranst	5f13dd8	"2026-05-10 18:47:47 +0200"	avx	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.168815	OK	0.99948	0.0902952
mpi-homo	bittnem1	4145a4d	"2026-06-24 16:24:32 +0200"	avx512	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.170994	OK	0.99886	0.198132
mpi-homo	bittnem1	131dd04	"2026-05-11 19:47:48 +0200"	avx	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.168888	OK	0.999913	0.0151953
mpi-homo	bitzanf	0d2037b	"2026-06-30 18:14:21 +0200"	avx512	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.112512	OK	0.657238	72.911
mpi-homo	bitzanf	0d2037b	"2026-06-30 18:14:21 +0200"	avx	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.151669	OK	0.897962	18.6968
mpi-homo	blahetad	5cbd5ea	"2026-04-27 15:02:02 +0200"	avx512	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.170688	OK	0.99707	0.509697
mpi-homo	blahetad	5cbd5ea	"2026-04-27 15:02:02 +0200"	avx	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.168894	OK	0.999947	0.00914997
mpi-homo	bubakf	5a23c16	"2026-04-18 13:28:08 +0200"	avx512	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.170749	OK	0.997425	0.447905
mpi-homo	bubakf	5a23c16	"2026-04-18 13:28:08 +0200"	avx	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.169391	OK	1.00289	-0.500623
mpi-homo	cernohj3	aaae8c4	"2026-06-30 21:25:45 +0200"	avx512	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.0695857	OK	0.406484	156.383
mpi-homo	cimermmi	90f20ec	"2026-06-30 23:53:52 +0200"	avx512	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.130371	OK	0.761561	47.3182
mpi-homo	cimermmi	90f20ec	"2026-06-30 23:53:52 +0200"	avx	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.13711	OK	0.811764	36.2281
mpi-homo	dostama4	89ca229	"2026-05-10 21:20:02 +0200"	avx512	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.170844	OK	0.997983	0.350661
mpi-homo	dostama4	89ca229	"2026-05-10 21:20:02 +0200"	avx	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.16899	OK	1.00052	-0.0896101
mpi-homo	drlikmar	9270c7b	"2026-06-21 22:55:07 +0200"	avx512	(16,16,5)	(2048,2048,2)	(8,8,2)	all	2	0.063257	OK	0.369514	172.948
mpi-homo	drlikmar	9270c7b	"2026-06-21 22:55:07 +0200"	avx	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.188153	OK	1.11397	-18.7493
mpi-homo	farkasm2	fd78b5b	"2026-07-04 02:22:52 +0200"	avx512	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.098946	OK	0.577991	95.2314
mpi-homo	farkasm2	fd78b5b	"2026-07-04 02:22:52 +0200"	avx	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.113498	OK	0.67197	69.06
mpi-homo	fedorekd	4d2e6c0	"2026-03-07 14:47:14 +0100"	avx	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.168817	OK	0.999492	0.0883385
mpi-homo	fedorekd	4d2e6c0	"2026-03-07 14:47:14 +0100"	avx512	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.171131	OK	0.999658	0.0594921
mpi-homo	havelh	5835e56	"2026-04-17 15:26:57 +0200"	avx512	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.170793	OK	0.997685	0.402566
mpi-homo	havelh	5835e56	"2026-04-17 15:26:57 +0200"	avx	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.168881	OK	0.999869	0.0227418
mpi-homo	hrdinap1	4d2e6c0	"2026-03-07 14:47:14 +0100"	avx512	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.170747	OK	0.997418	0.449174
mpi-homo	hrdinap1	4d2e6c0	"2026-03-07 14:47:14 +0100"	avx	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.168845	OK	0.999657	0.0595369
mpi-homo	hubatama	4d2e6c0	"2026-03-07 14:47:14 +0100"	avx512	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.171124	OK	0.999616	0.0666572
mpi-homo	hubatama	4d2e6c0	"2026-03-07 14:47:14 +0100"	avx	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.16895	OK	1.00028	-0.0484665
mpi-homo	huzickam	e949a67	"2026-05-06 15:33:07 +0200"	avx512	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.170763	OK	0.997509	0.433227
mpi-homo	huzickam	e949a67	"2026-05-06 15:33:07 +0200"	avx	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.16891	OK	1.00004	-0.00714848
mpi-homo	janicej	33941a5	"2026-05-11 14:11:58 +0200"	avx512	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.170719	OK	0.997253	0.477774
mpi-homo	janicej	33941a5	"2026-05-11 14:11:58 +0200"	avx	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.168944	OK	1.00024	-0.0418285
mpi-homo	jaworskr	ef8ab75	"2026-06-20 18:51:37 +0200"	avx512	(16,16,5)	(2048,2048,2)	(8,8,2)	all	4	0.118184	OK	0.690371	64.3671
mpi-homo	jaworskr	ef8ab75	"2026-06-20 18:51:37 +0200"	avx	(16,16,5)	(2048,2048,2)	(8,8,2)	all	4	0.136707	OK	0.809382	36.7385
mpi-homo	kanev	4d2e6c0	"2026-03-07 14:47:14 +0100"	avx512	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.170888	OK	0.998242	0.305736
mpi-homo	kanev	4d2e6c0	"2026-03-07 14:47:14 +0100"	avx	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.168879	OK	0.999856	0.0250046
mpi-homo	kapylouma	4d2e6c0	"2026-03-07 14:47:14 +0100"	avx512	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.171123	OK	0.999613	0.0672245
mpi-homo	kapylouma	4d2e6c0	"2026-03-07 14:47:14 +0100"	avx	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.169249	OK	1.00205	-0.355504
mpi-homo	karaseto	d0a3297	"2026-05-11 17:14:22 +0200"	avx512	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.170889	OK	0.998246	0.30503
mpi-homo	karaseto	d0a3297	"2026-05-11 17:14:22 +0200"	avx	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.168951	OK	1.00029	-0.0495114
mpi-homo	khotchai	c7bc803	"2026-06-30 13:56:23 +0200"	avx	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.0641525	OK	0.379818	168.17
mpi-homo	khotchai	c7bc803	"2026-06-30 13:56:23 +0200"	avx512	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.065267	OK	0.381256	167.513
mpi-homo	kolnika	4d2e6c0	"2026-03-07 14:47:14 +0100"	avx512	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.1709	OK	0.99831	0.293895
mpi-homo	kolnika	4d2e6c0	"2026-03-07 14:47:14 +0100"	avx	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.168824	OK	0.999533	0.0811689
mpi-homo	komarem4	fea02b5	"2026-05-20 00:52:25 +0200"	avx512	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.0883581	OK	0.516142	114.892
mpi-homo	komarem4	fea02b5	"2026-05-20 00:52:25 +0200"	avx	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.0966061	OK	0.571961	97.0533
mpi-homo	koribans	c2085eb	"2026-05-24 05:53:38 +0200"	avx512	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.0776424	OK	0.453547	137.351
mpi-homo	koribans	c2085eb	"2026-05-24 05:53:38 +0200"	avx	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.0812749	OK	0.481192	127.073
mpi-homo	kotlint	118fefa	"2026-06-29 14:55:36 +0200"	avx512	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.130068	OK	0.759788	47.7229
mpi-homo	kotlint	118fefa	"2026-06-29 14:55:36 +0200"	avx	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.138129	OK	0.817802	34.9408
mpi-homo	kotvami	36531e6	"2026-07-08 00:24:28 +0200"	avx512	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.12912	OK	0.754255	48.9927
mpi-homo	kotvami	36531e6	"2026-07-08 00:24:28 +0200"	avx	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.136807	OK	0.809972	36.6121
mpi-homo	krenmar	d1c7bbb	"2026-05-17 13:06:43 +0200"	avx	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.168766	OK	0.999188	0.141115
mpi-homo	krenmar	d1c7bbb	"2026-05-17 13:06:43 +0200"	avx512	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.171171	OK	0.999891	0.0189444
mpi-homo	kroupad1	4d2e6c0	"2026-03-07 14:47:14 +0100"	avx512	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.170875	OK	0.998161	0.319683
mpi-homo	kroupad1	4d2e6c0	"2026-03-07 14:47:14 +0100"	avx	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.169242	OK	1.00201	-0.348624
mpi-homo	kruchij	20747c5	"2026-06-30 12:19:05 +0200"	avx512	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.130466	OK	0.762112	47.1924
mpi-homo	kruchij	20747c5	"2026-06-30 12:19:05 +0200"	avx	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.138097	OK	0.81761	34.9814
mpi-homo	kunovaer	695f8ac	"2026-06-08 12:33:46 +0200"	avx512	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.110521	OK	0.645609	76.0121
mpi-homo	kunovaer	695f8ac	"2026-06-08 12:33:46 +0200"	avx	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.141617	OK	0.838453	30.6084
mpi-homo	kurzs	ec83dbb	"2026-05-17 12:18:20 +0200"	avx512	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.0781705	OK	0.456631	136.174
mpi-homo	kurzs	ec83dbb	"2026-05-17 12:18:20 +0200"	avx	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.168991	OK	1.00052	-0.0903302
mpi-homo	kyselaj1	d2f74d4	"2026-05-25 23:15:07 +0200"	avx512	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.170791	OK	0.997674	0.40454
mpi-homo	kyselaj1	d2f74d4	"2026-05-25 23:15:07 +0200"	avx	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.169215	OK	1.00184	-0.320018
mpi-homo	lamplotm	b92102f	"2026-06-27 10:50:44 +0200"	avx512	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.0577709	OK	0.337468	188.707
mpi-homo	lamplotm	b92102f	"2026-06-27 10:50:44 +0200"	avx	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.0796385	OK	0.471504	130.606
mpi-homo	lejkom	e71a7e2	"2026-05-29 15:20:25 +0200"	avx512	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.0748247	OK	0.437087	143.773
mpi-homo	lejkom	e71a7e2	"2026-05-29 15:20:25 +0200"	avx	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.0745007	OK	0.441085	142.191
mpi-homo	lipovss	a161d4a	"2026-06-30 13:54:34 +0200"	avx	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.101135	OK	0.598772	89.0953
mpi-homo	lipovss	a161d4a	"2026-06-30 13:54:34 +0200"	avx512	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.105433	OK	0.615884	84.2004
mpi-homo	lopatad	5173af2	"2026-05-04 21:38:11 +0200"	avx512	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.170813	OK	0.9978	0.382517
mpi-homo	lopatad	5173af2	"2026-05-04 21:38:11 +0200"	avx	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.168791	OK	0.999335	0.115572
mpi-homo	lovisekd	6ef8d97	"2026-04-18 21:17:08 +0200"	avx512	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.170799	OK	0.997717	0.397064
mpi-homo	lovisekd	6ef8d97	"2026-04-18 21:17:08 +0200"	avx	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.168822	OK	0.999518	0.0836685
mpi-homo	lukasej1	abc5dfc	"2026-06-29 01:12:35 +0200"	avx	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.100226	OK	0.593395	90.6624
mpi-homo	lukasej1	abc5dfc	"2026-06-29 01:12:35 +0200"	avx512	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.106154	OK	0.620097	83.0162
mpi-homo	machhy	318ce68	"2026-07-01 14:44:41 +0200"	avx512	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.130782	OK	0.763959	46.772
mpi-homo	machhy	318ce68	"2026-07-01 14:44:41 +0200"	avx	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.138453	OK	0.819717	34.5344
mpi-homo	machotao	a67b323	"2026-05-29 11:36:13 +0200"	avx512	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.0825059	OK	0.481957	126.797
mpi-homo	machotao	a67b323	"2026-05-29 11:36:13 +0200"	avx	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.0906797	OK	0.536874	108.051
mpi-homo	maliarm	f248943	"2026-03-11 03:09:48 +0100"	avx512	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.171114	OK	0.999559	0.0766851
mpi-homo	maliarm	f248943	"2026-03-11 03:09:48 +0100"	avx	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.16885	OK	0.999685	0.0547763
mpi-homo	malydo	7605690	"2026-06-29 20:31:45 +0200"	avx512	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.0675949	OK	0.394855	161.425
mpi-homo	malydo	7605690	"2026-06-29 20:31:45 +0200"	avx	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.117595	OK	0.69623	62.8989
mpi-homo	masekmi2	0d4b782	"2026-04-27 15:45:15 +0200"	avx512	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.170759	OK	0.997483	0.437854
mpi-homo	masekmi2	0d4b782	"2026-04-27 15:45:15 +0200"	avx	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.168967	OK	1.00038	-0.0653527
mpi-homo	meszard	9da3e77	"2026-06-30 21:46:55 +0200"	avx512	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.0821921	OK	0.480124	127.459
mpi-homo	meszard	9da3e77	"2026-06-30 21:46:55 +0200"	avx	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.0896642	OK	0.530862	110.007
mpi-homo	mojikm	861a978	"2026-05-11 23:53:01 +0200"	avx512	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.170729	OK	0.997312	0.467506
mpi-homo	mojikm	861a978	"2026-05-11 23:53:01 +0200"	avx	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.168833	OK	0.999584	0.0723259
mpi-homo	mudraja1	3624aee	"2026-04-15 13:50:02 +0200"	avx512	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.17079	OK	0.997665	0.406174
mpi-homo	mudraja1	3624aee	"2026-04-15 13:50:02 +0200"	avx	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.169031	OK	1.00076	-0.131202
mpi-homo	novakj80	b59c4da	"2026-07-07 22:54:06 +0200"	avx512	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.0823161	OK	0.480848	127.197
mpi-homo	novakj80	b59c4da	"2026-07-07 22:54:06 +0200"	avx	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.0844502	OK	0.499992	120.415
mpi-homo	pajonkf	4d2e6c0	"2026-03-07 14:47:14 +0100"	avx	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.168871	OK	0.999812	0.0326659
mpi-homo	pajonkf	4d2e6c0	"2026-03-07 14:47:14 +0100"	avx512	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.171568	OK	1.00221	-0.383606
mpi-homo	pelikam3	52b02c6	"2026-05-11 23:59:26 +0200"	avx512	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.170801	OK	0.997733	0.394216
mpi-homo	pelikam3	52b02c6	"2026-05-11 23:59:26 +0200"	avx	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.169133	OK	1.00136	-0.236502
mpi-homo	petruchm	3bdb7d5	"2026-06-30 21:18:17 +0200"	avx512	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.130751	OK	0.76378	46.8128
mpi-homo	petruchm	3bdb7d5	"2026-06-30 21:18:17 +0200"	avx	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.138208	OK	0.818268	34.8419
mpi-homo	popovaks	19a1f56	"2026-06-25 22:08:51 +0200"	avx512	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.109031	OK	0.636901	78.3713
mpi-homo	popovaks	19a1f56	"2026-06-25 22:08:51 +0200"	avx	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.122372	OK	0.724513	55.9816
mpi-homo	posaa	ea5de54	"2026-07-01 16:18:17 +0200"	avx	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.125139	OK	0.740894	52.0975
mpi-homo	posaa	ea5de54	"2026-07-01 16:18:17 +0200"	avx512	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.153542	OK	0.896912	18.9001
mpi-homo	ref	83d4b77	"2026-04-18 21:35:42 +0200"	avx512	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.0680516	OK	0.397522	160.255
mpi-homo	ref	83d4b77	"2026-04-18 21:35:42 +0200"	avx	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.0730303	OK	0.432379	145.654
mpi-homo	reznico	b087bd6	"2026-05-08 13:38:29 +0200"	avx512	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.170802	OK	0.997739	0.393169
mpi-homo	reznico	b087bd6	"2026-05-08 13:38:29 +0200"	avx	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.168968	OK	1.00039	-0.067019
mpi-homo	sadovskv	5bacd1a	"2026-05-11 13:49:50 +0200"	avx512	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.170748	OK	0.997422	0.448373
mpi-homo	sadovskv	5bacd1a	"2026-05-11 13:49:50 +0200"	avx	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.169218	OK	1.00186	-0.323539
mpi-homo	schrodm	85d095e	"2026-05-10 09:23:11 +0200"	avx512	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.171013	OK	0.998967	0.179535
mpi-homo	schrodm	85d095e	"2026-05-10 09:23:11 +0200"	avx	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.168923	OK	1.00012	-0.0207434
mpi-homo	sebaned	223517d	"2026-05-03 14:51:39 +0200"	avx512	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.170831	OK	0.997904	0.36457
mpi-homo	sebaned	223517d	"2026-05-03 14:51:39 +0200"	avx	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.16881	OK	0.999449	0.095767
mpi-homo	shtabovv	240339b	"2026-05-30 21:48:46 +0200"	avx	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.121761	OK	0.720891	56.8522
mpi-homo	shtabovv	240339b	"2026-05-30 21:48:46 +0200"	avx512	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.142216	OK	0.830753	32.2112
mpi-homo	snajdrvl	4d2e6c0	"2026-03-07 14:47:14 +0100"	avx512	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.17071	OK	0.9972	0.48714
mpi-homo	snajdrvl	4d2e6c0	"2026-03-07 14:47:14 +0100"	avx	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.168853	OK	0.999705	0.0513051
mpi-homo	stefanm4	df73f4b	"2026-05-12 11:40:06 +0200"	avx512	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.171086	OK	0.999394	0.105366
mpi-homo	stefanm4	df73f4b	"2026-05-12 11:40:06 +0200"	avx	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.168846	OK	0.999663	0.0586159
mpi-homo	stehlisi	d6d9e0e	"2026-06-28 09:57:15 +0200"	avx512	(16,16,5)	(2048,2048,2)	(8,8,2)	all	5	0.0705102	OK	0.411884	154.09
mpi-homo	stehlisi	d6d9e0e	"2026-06-28 09:57:15 +0200"	avx	(16,16,5)	(2048,2048,2)	(8,8,2)	all	5	0.0719657	OK	0.426077	148.205
mpi-homo	strecans	8445aeb	"2026-06-01 18:43:33 +0200"	avx512	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.0681144	OK	0.397889	160.095
mpi-homo	strecans	8445aeb	"2026-06-01 18:43:33 +0200"	avx	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.0725476	OK	0.429522	146.806
mpi-homo	stupkava	3fc7866	"2026-07-16 21:39:04 +0200"	avx512	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.0847202	OK	0.494891	122.196
mpi-homo	stupkava	3fc7866	"2026-07-16 21:39:04 +0200"	avx	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.0953309	OK	0.564412	99.3617
mpi-homo	tomam	1ce1671	"2026-05-22 01:18:08 +0200"	avx512	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.0731566	OK	0.427343	147.689
mpi-homo	tomam	1ce1671	"2026-05-22 01:18:08 +0200"	avx	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.0902451	OK	0.534301	108.886
mpi-homo	tomisz	4d2e6c0	"2026-03-07 14:47:14 +0100"	avx512	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.170838	OK	0.997946	0.357178
mpi-homo	tomisz	4d2e6c0	"2026-03-07 14:47:14 +0100"	avx	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.169171	OK	1.00159	-0.275262
mpi-homo	vernem	20bc26b	"2026-06-30 20:08:59 +0200"	avx512	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.0921517	OK	0.538302	107.589
mpi-homo	vivtonea	15996de	"2026-05-10 17:24:04 +0000"	avx512	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.171075	OK	0.999334	0.11582
mpi-homo	vivtonea	15996de	"2026-05-10 17:24:04 +0000"	avx	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.168957	OK	1.00032	-0.0554711
mpi-homo	yuriicho	4d2e6c0	"2026-03-07 14:47:14 +0100"	avx512	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.171067	OK	0.999287	0.123851
mpi-homo	yuriicho	4d2e6c0	"2026-03-07 14:47:14 +0100"	avx	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.169067	OK	1.00097	-0.168625
mpi-homo	zapotocf	45c974e	"2026-06-17 12:04:33 +0200"	avx512	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.0825533	OK	0.482233	126.697
mpi-homo	zapotocf	45c974e	"2026-06-17 12:04:33 +0200"	avx	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.084041	OK	0.497569	121.259
mpi-homo	zatloum1	e9f312c	"2026-04-20 20:19:44 +0200"	avx512	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.170882	OK	0.998205	0.312058
mpi-homo	zatloum1	e9f312c	"2026-04-20 20:19:44 +0200"	avx	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.16929	OK	1.00229	-0.397148
mpi-homo	zavodsv	4d2e6c0	"2026-03-07 14:47:14 +0100"	avx512	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.170838	OK	0.997949	0.356594
mpi-homo	zavodsv	4d2e6c0	"2026-03-07 14:47:14 +0100"	avx	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.169352	OK	1.00266	-0.461315
mpi-homo	zelezno	00e5729	"2026-06-25 13:52:29 +0200"	avx512	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.0592418	OK	0.34606	184.339
mpi-homo	zelezno	00e5729	"2026-06-25 13:52:29 +0200"	avx	(16,16,5)	(2048,2048,2)	(8,8,2)	all	6	0.0888096	OK	0.525802	111.671

Your job is to improve the performance of the existing code, by altering the code in sol/dcnnsol.hpp. (You may also write some code in sol/dcnnsol.cpp; however, all the existing code is templated and must remain in the header file.)

Input data

The folder containing input data is set by the command-line parameter --data-folder.

The input data are already available at parlab, in the folder /home/_teaching/hiperf/dcnndata. Therefore, the program shall be invoked as:

srun -p mpi-homo-short -n 1 -c 64 ./dcnn --data-folder=/home/_teaching/hiperf/dcnndata

The input data may be downloaded from parlab via scp or compressed from here:

simplenet.model.tgz (5.3 MB) - Pretrained weights

imagenet.input.tgz (307.9 MB) - Test images and expected results

Test parameters

data-folder - the folder containing the input data files (default: data).

minibatch - the number of images in a testing minibatch (processed in one call to the forward functions). Default: 16.

superbatch - the number of minibatches in a testing batch (each minibatch is assigned to a different thread). Default: 8 (1 in Debug mode).

total - the total number of images submitted into testing (shall be divisible by minibatch*superbatch). Default: 2048 (16 in Debug mode).

Credits

The DCNN architecture was taken from [Hasanpour 2016]. The original implementation used the Caffe framework and was later converted to Pytorch.

Both the pretrained weights and the test images were converted from publicly available data:

SimpleNet - ImageNet pretrained weights

ILSVRC2012 - ImageNet Object Localization Challenge

References

[Hasanpour 2016] Hasanpour, Seyyed Hossein, et al. Lets keep it simple, using simple architectures to outperform deeper and more complex architectures. arXiv:1608.06037.