matrixStats: Benchmark report

binCounts() benchmarks

This report benchmark the performance of binCounts() against alternative methods.

Alternative methods

hist()

as below

> hist <- graphics::hist
> binCounts_hist <- function(x, bx, right = FALSE, ...) {
+     hist(x, breaks = bx, right = right, include.lowest = TRUE, plot = FALSE)$counts
+ }

Data type “integer”

Non-sorted simulated data

> set.seed(48879)
> nx <- 1e+05
> xmax <- 0.01 * nx
> x <- runif(nx, min = 0, max = xmax)
> storage.mode(x) <- mode
> str(x)
 int [1:100000] 722 285 591 3 349 509 216 91 150 383 ...
> nb <- 10000
> bx <- seq(from = 0, to = xmax, length.out = nb + 1L)
> bx <- c(-1, bx, xmax + 1)

Results

> gc()
          used (Mb) gc trigger (Mb) max used (Mb)
Ncells  807852 43.2    1315788 70.3  1315788 70.3
Vcells 1498901 11.5    8388608 64.0  3562027 27.2
> stats <- microbenchmark(binCounts = binCounts(x, bx = bx), hist = binCounts_hist(x, bx = bx), unit = "ms")

Table: Benchmarking of binCounts() and hist() on integer+unsorted data. The top panel shows times in milliseconds and the bottom panel shows relative times.

	expr	min	lq	mean	median	uq	max
1	binCounts	5.9754	6.43200	6.878917	6.5684	6.68900	13.1917
2	hist	11.5389	12.30655	12.916438	12.4926	12.62675	19.1192

	expr	min	lq	mean	median	uq	max
1	binCounts	1.000000	1.000000	1.000000	1.000000	1.000000	1.000000
2	hist	1.931067	1.913332	1.877685	1.901924	1.887689	1.449336

Figure: Benchmarking of binCounts() and hist() on integer+unsorted data. Outliers are displayed as crosses. Times are in milliseconds.

Sorted simulated data

> x <- sort(x)

> gc()
          used (Mb) gc trigger  (Mb) max used (Mb)
Ncells  969795 51.8    1954774 104.4  1315788 70.3
Vcells 1824343 14.0    8388608  64.0  8386590 64.0
> stats <- microbenchmark(binCounts = binCounts(x, bx = bx), hist = binCounts_hist(x, bx = bx), unit = "ms")

Table: Benchmarking of binCounts() and hist() on integer+sorted data. The top panel shows times in milliseconds and the bottom panel shows relative times.

	expr	min	lq	mean	median	uq	max
1	binCounts	1.2297	1.56915	2.051464	1.65505	1.72965	7.8569
2	hist	3.7337	4.12480	4.656045	4.51275	4.63205	13.4317

	expr	min	lq	mean	median	uq	max
1	binCounts	1.000000	1.000000	1.000000	1.000000	1.000000	1.000000
2	hist	3.036269	2.628684	2.269621	2.726655	2.678027	1.709542

Figure: Benchmarking of binCounts() and hist() on integer+sorted data. Outliers are displayed as crosses. Times are in milliseconds.

Data type “double”

Non-sorted simulated data

> set.seed(48879)
> nx <- 1e+05
> xmax <- 0.01 * nx
> x <- runif(nx, min = 0, max = xmax)
> storage.mode(x) <- mode
> str(x)
 num [1:100000] 722.11 285.54 591.33 3.42 349.14 ...
> nb <- 10000
> bx <- seq(from = 0, to = xmax, length.out = nb + 1L)
> bx <- c(-1, bx, xmax + 1)

Results

> gc()
          used (Mb) gc trigger  (Mb) max used  (Mb)
Ncells  969930 51.8    1954774 104.4  1954774 104.4
Vcells 1874747 14.4    8388608  64.0  8386890  64.0
> stats <- microbenchmark(binCounts = binCounts(x, bx = bx), hist = binCounts_hist(x, bx = bx), unit = "ms")

Table: Benchmarking of binCounts() and hist() on double+unsorted data. The top panel shows times in milliseconds and the bottom panel shows relative times.

	expr	min	lq	mean	median	uq	max
1	binCounts	10.5039	11.7464	12.25321	12.11045	12.20440	19.0747
2	hist	11.6019	13.5030	14.08809	13.69295	13.91275	20.3897

	expr	min	lq	mean	median	uq	max
1	binCounts	1.000000	1.000000	1.000000	1.000000	1.000000	1.000000
2	hist	1.104533	1.149544	1.149747	1.130672	1.139978	1.068939

Figure: Benchmarking of binCounts() and hist() on double+unsorted data. Outliers are displayed as crosses. Times are in milliseconds.

Sorted simulated data

> x <- sort(x)

> gc()
          used (Mb) gc trigger  (Mb) max used  (Mb)
Ncells  969996 51.9    1954774 104.4  1954774 104.4
Vcells 1875303 14.4    8388608  64.0  8388323  64.0
> stats <- microbenchmark(binCounts = binCounts(x, bx = bx), hist = binCounts_hist(x, bx = bx), unit = "ms")

Table: Benchmarking of binCounts() and hist() on double+sorted data. The top panel shows times in milliseconds and the bottom panel shows relative times.

	expr	min	lq	mean	median	uq	max
1	binCounts	1.1680	2.76695	3.048870	2.86545	3.03605	8.1864
2	hist	4.3223	4.88730	5.577062	5.15950	5.44775	13.3829

	expr	min	lq	mean	median	uq	max
1	binCounts	1.000000	1.000000	1.000000	1.00000	1.000000	1.000000
2	hist	3.700599	1.766313	1.829223	1.80059	1.794355	1.634772

Figure: Benchmarking of binCounts() and hist() on double+sorted data. Outliers are displayed as crosses. Times are in milliseconds.

Appendix

Session information

R Under development (unstable) (2023-11-06 r85483 ucrt)
Platform: x86_64-w64-mingw32/x64
Running under: Windows Server 2022 x64 (build 20348)

Matrix products: default


locale:
[1] LC_COLLATE=C                 LC_CTYPE=German_Germany.utf8
[3] LC_MONETARY=C                LC_NUMERIC=C                
[5] LC_TIME=C                   

time zone: Europe/Berlin
tzcode source: internal

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods   base     

other attached packages:
[1] microbenchmark_1.4.10 matrixStats_1.1.0     ggplot2_3.4.4        
[4] knitr_1.45            R.devices_2.17.1      R.utils_2.12.2       
[7] R.oo_1.25.0           R.methodsS3_1.8.2    

loaded via a namespace (and not attached):
 [1] vctrs_0.6.4      cli_3.6.1        rlang_1.1.2      xfun_0.41       
 [5] labeling_0.4.3   glue_1.6.2       colorspace_2.1-0 markdown_1.11   
 [9] scales_1.2.1     fansi_1.0.5      R.cache_0.16.0   grid_4.4.0      
[13] munsell_0.5.0    tibble_3.2.1     R.rsp_0.45.0     base64enc_0.1-3 
[17] lifecycle_1.0.4  compiler_4.4.0   pkgconfig_2.0.3  farver_2.1.1    
[21] digest_0.6.33    R6_2.5.1         utf8_1.2.4       pillar_1.9.0    
[25] magrittr_2.0.3   withr_2.5.2      tools_4.4.0      gtable_0.3.4

Total processing time was 8.67 secs.

Reproducibility

To reproduce this report, do:

html <- matrixStats:::benchmark('binCounts')