Julia versus Intel Fortran

How the programs are written matters! Always look at the source code.

If the fastest programs are flagged * possible hand-written vector instructions or "unsafe" or naked ffi, does the host language matter? You might be more interested in the less optimised programs — more seconds, less gz source code.

fannkuch-redux
	elapsed	Mean or 95% CI
source	secs	cpu secs	mem	gz	make
* Julia #2	8.60	31.98—32.02	333,173	1077	0.11
* Julia #3	9.44	35.58—35.64	333,042	896	0.14
Intel Fortran #3	9.67	38.56	12,128	1164	5.02
Intel Fortran	40.21	40.21	2,179	608	3.95
Julia	42.57	42.62	309,928	441	0.13
n-body
	elapsed	Mean or 95% CI
source	secs	cpu secs	mem	gz	make
Intel Fortran #5	3.39	3.40	2,146	1669	4.48
Julia #5	3.80	3.83—4.13	345,874	1129	0.14
Intel Fortran #2	4.46	4.42—4.69	2,146	1507	4.33
Intel Fortran	4.55	4.55	2,142	1400	4.41
Intel Fortran #3	4.72	4.69—4.82	2,154	1311	4.43
Intel Fortran #6	4.99	4.64—6.08	2,052	1530	4.43
Julia #4	7.19	7.26	414,929	1168	0.15
Julia #3	8.07	8.11—8.17	415,089	1548	0.13
Julia #2	28.68	28.88—29.15	414,970	1084	0.11
Intel Fortran #4		Bad Output
spectral-norm
	elapsed	Mean or 95% CI
source	secs	cpu secs	mem	gz	make
Intel Fortran #2	2.67	2.67	2,331	520	5.05
Intel Fortran #3	0.72	2.88—2.92	5,214	644	4.80
Julia #4	1.31	3.73	315,896	435	0.09
Julia #2	2.84	5.81—5.85	414,818	377	0.14
mandelbrot
	elapsed	Mean or 95% CI
source	secs	cpu secs	mem	gz	make
Julia #7	1.54	4.86	357,138	629	0.12
Julia #2	1.57	4.96	355,762	634	0.10
Julia #3	1.61	5.18	376,463	589	0.11
Julia #4	2.07	5.19	368,173	583	0.14
Julia #5	1.76	5.85	381,813	580	0.12
Julia #6	2.25	5.89	368,599	571	0.13
Intel Fortran #6	2.72	8.55	85,975	669	4.13
Intel Fortran #7	4.09	16.19	85,639	965	5.17
Intel Fortran #4	9.89	25.29—25.33	86,360	626	3.95
Julia	26.66	26.71	339,132	379	0.11
Intel Fortran #5	32.66	32.67	47,174	528	3.72
fasta
	elapsed	Mean or 95% CI
source	secs	cpu secs	mem	gz	make
Julia #2	1.28	1.32	318,083	1094	0.11
Intel Fortran #4	1.36	1.37	2,142	1355	4.44
Julia #5	2.68	2.72	334,770	957	0.10
Julia #4	2.73	2.77	320,528	1053	0.13
* Julia #7	1.98	4.93—5.22	403,165	1189	0.12
* Julia #6	2.26	4.77—4.97	415,330	1213	0.13
Intel Fortran	6.01	6.05—6.11	2,146	1193	4.34
Intel Fortran #3	6.53	6.57—6.65	2,142	1217	4.47
Julia #3	11.71	11.76—11.82	318,497	901	0.13
k-nucleotide
	elapsed	Mean or 95% CI
source	secs	cpu secs	mem	gz	make
Julia #6	6.98	14.54—14.97	524,595	965	0.12
Julia #7	9.93	23.91—24.31	575,861	1379	0.07
Julia	132.79	134.30—135.03	1,696,670	703	0.10
Intel Fortran		Empty
reverse-complement
	elapsed	Mean or 95% CI
source	secs	cpu secs	mem	gz	make
Julia #7	1.79	1.87	1,128,022	531	0.13
Julia #9	2.12	2.20	1,330,082	456	0.11
Intel Fortran	3.35	3.36	1,012,519	1026	4.53
Julia #2	29.48	29.90—30.70	2,259,153	464	0.12
Intel Fortran #2	45.22	45.34—45.58	790,307	778	4.10
binary-trees [ Contentious. Different approaches. ]
	elapsed	Mean or 95% CI
source	secs	cpu secs	mem	gz	make
Intel Fortran #2	2.24	6.89—7.00	148,124	1155	4.32
Julia #2	12.65	13.47—14.55	552,669	430	0.13
Julia #4	14.42	15.81—15.94	550,064	378	0.08
Julia #3	7.40	19.30—20.35	564,216	640	0.14
Intel Fortran	32.74	32.78—32.88	540,066	780	3.98
Julia #5	21.83	61.11—62.09	1,336,107	396	0.13
pidigits [ Contentious. Different libraries. ]
	elapsed	Mean or 95% CI
source	secs	cpu secs	mem	gz	make
Intel Fortran #4	0.74	0.74	6,590	987	3.94
* Julia #3	1.14	1.24	307,749	512	0.10
* Julia #2	1.25	1.31	315,859	686	0.15
Julia #4	3.80	3.88—3.92	414,781	531	0.14
regex-redux [ Contentious. Different libraries. ]
	elapsed	Mean or 95% CI
source	secs	cpu secs	mem	gz	make
Intel Fortran	0.98	3.11—3.15	189,305	2229	4.80
* Julia #5	1.91	3.19	519,877	766	0.16
* Julia #4	2.00	3.29	522,019	772	0.15
* Julia #3	2.19	3.54—3.76	671,166	630	0.12
Julia	5.19	5.26—5.31	714,248	414	0.11
Julia	julia version 1.11.1
Intel Fortran	ifx (IFX) 2024.1.0 20240308