SG against C on Windows
The programs of SG against C, measured again on Windows. How the numbers are taken is in How it is measured.
How it was run
Measured on Windows (python3 bench/run.py, the hundred-and-second slice): Windows 11 on 64-bit Arm in a
virtual machine on the Mac of SG against C, Clang 23 with C's library in the program (/MT, as sg build
links), best of five after a run that is not timed, which also leaves out the first start of a new program
(which on Windows took seconds now and then; why was not looked into); spawn, spawnasync, capture,
pipelines and pipesasync are one timed run each (--runs 1), since five thousand starts take over twenty
seconds there. The last column is the same ratio on macOS, from SG against C or from the
program's choice. Windows on x86-64 is not measured: there is no such machine here, and the machines of the CI are
shared and give no steady numbers.
The table
| Program | SG, release | SG, overflow checked | C | release / C | on macOS |
|---|---|---|---|---|---|
fib | 0.276 s | 0.393 s | 0.273 s | 1.01 | 0.98 |
primes | 0.488 s | 0.475 s | 0.471 s | 1.04 | 0.99 |
collatz | 0.339 s | 0.540 s | 0.340 s | 1.00 | 1.00 |
words | 1.155 s | 1.094 s | 1.100 s | 1.05 | 1.02 |
list | 0.433 s | 0.444 s | 0.431 s | 1.00 | 1.01 |
expr | 0.342 s | 0.346 s | 0.338 s | 1.01 | 1.02 |
parse | 0.286 s | 0.315 s | 0.284 s | 1.01 | 0.99 |
vector | 2.117 s | 2.371 s | 2.114 s | 1.00 | 1.02 |
counts | 0.408 s | 0.404 s | 0.343 s | 1.19 | 1.20 |
show | 2.876 s | 3.025 s | 8.363 s | 0.34 | 0.47 |
lookup | 0.501 s | 0.488 s | 0.483 s | 1.04 | 1.01 |
buffer | 0.287 s | 0.300 s | 0.287 s | 1.00 | 0.99 |
bits | 0.339 s | 0.303 s | 0.369 s | 0.92 | 0.79 |
hash | 0.132 s | 0.134 s | 0.131 s | 1.01 | 1.00 |
nbody | 0.669 s | 0.669 s | 0.673 s | 0.99 | 0.97 |
numbers | 0.222 s | 0.225 s | 0.498 s | 0.45 | 0.95 |
floats | 0.182 s | 0.195 s | 0.735 s | 0.25 | 0.23 |
fields | 1.349 s | 1.358 s | 1.487 s | 0.91 | 1.02 |
lines | 0.072 s | 0.098 s | 0.454 s | 0.16 | 0.82 |
copy | 0.160 s | 0.187 s | 0.910 s | 0.18 | 0.54 |
bytes | 0.061 s | 0.059 s | 0.064 s | 0.95 | 0.94 |
table | 2.524 s | 2.476 s | 3.325 s | 0.76 | 0.52 |
blocks | 0.107 s | 0.120 s | 0.117 s | 0.91 | 1.00 |
keywords | 0.233 s | 0.234 s | 0.216 s | 1.08 | 1.03 |
classes | 0.209 s | 0.269 s | 0.234 s | 0.89 | 0.87 |
rewrite | 0.575 s | 0.625 s | 0.720 s | 0.80 | 0.94 |
tree | 0.194 s | 0.244 s | 0.114 s | 1.71 | 1.53 |
tasks | 0.121 s | 0.189 s | 0.123 s | 0.99 | 1.00 |
async | 0.236 s | 0.354 s | 0.208 s | 1.14 | 1.08 |
asynclines | 0.153 s | 0.171 s | 0.458 s | 0.33 | 1.28 |
tasklist | 0.534 s | 0.746 s | 0.445 s | 1.20 | 1.24 |
asynccopy | 0.384 s | 0.434 s | 0.920 s | 0.42 | 1.04 |
echo | 0.861 s | 0.850 s | 0.769 s | 1.12 | 0.89 |
echosync | 0.873 s | 0.778 s | 0.769 s | 1.13 | 0.99 |
echowork | 0.456 s | 0.457 s | 0.551 s | 0.83 | 1.02 |
netlines | 0.096 s | 0.132 s | 0.111 s | 0.86 | 1.14 |
ring | 0.343 s | 0.342 s | 0.343 s | 1.00 | 1.00 |
crc | 2.800 s | 2.798 s | 2.798 s | 1.00 | 0.99 |
deflate | 1.312 s | 1.311 s | 1.279 s | 1.03 | 1.04 |
clock | 0.237 s | 0.228 s | 0.236 s | 1.01 | 1.01 |
dates | 0.123 s | 0.303 s | 0.106 s | 1.16 | 0.29 |
rng | 0.115 s | 0.115 s | 0.115 s | 1.00 | 1.00 |
secure | 0.137 s | 0.137 s | 0.139 s | 0.98 | 1.00 |
listing | 0.316 s | 0.316 s | 0.276 s | 1.14 | 0.99 |
status | 11.688 s | 11.795 s | 15.677 s | 0.75 | 1.00 |
spawn | 22.829 s | 22.798 s | 22.761 s | 1.00 | 0.70 |
capture | 1.096 s | 1.279 s | 1.337 s | 0.82 | 1.49 |
pipelines | 0.091 s | 0.113 s | 0.502 s | 0.18 | 0.85 |
pipesasync | 0.153 s | 0.166 s | 0.507 s | 0.30 | 1.03 |
spawnasync | 22.644 s | 23.007 s | 22.968 s | 0.99 | 0.71 |
case | 0.176 s | 0.179 s | 5.273 s | 0.03 | 0.23 |
letters | 0.295 s | 0.376 s | 4.692 s | 0.06 | 0.10 |
udpecho | 1.089 s | 1.080 s | 1.365 s | 0.80 | 1.19 |
udpechosync | 1.299 s | 1.306 s | 1.284 s | 1.01 | 1.00 |
inline | 1.015 s | 1.012 s | 1.012 s | 1.00 | 1.03 |
callback | 0.914 s | 0.925 s | 0.913 s | 1.00 | 1.00 |
sqlite | 0.300 s | 0.301 s | 0.301 s | 1.00 | 1.00 |
curl | 0.366 s | 0.367 s | 0.365 s | 1.00 | 1.02 |
signals | 0.008 s | 0.007 s | 0.246 s | 0.03 | 1.27 |
bitfields | 0.371 s | 0.335 s | 0.352 s | 1.05 | 0.91 |
atomics | 0.226 s | 0.212 s | 0.260 s | 0.87 | 1.02 |
queue | 0.247 s | 0.250 s | 0.255 s | 0.97 | 1.39 |
library | 0.582 s | 0.576 s | 0.568 s | 1.02 | 1.01 |
kernel (size) | 5168 B | 2236 B | 2.31 | 2.31 |
How the C differs on Windows
The C is the same program wherever it can be. Where POSIX has a function Windows has not, the C takes
Windows' own (bench/windows/posix.h): threads of _beginthreadex and locks of SRWLOCK, as SG's runtime
has them there; Winsock; getline written over fgets, and the input read as bytes; popen through
cmd.exe. clock reads QueryPerformanceCounter, secure calls BCryptGenRandom, listing lists with
FindFirstFileExW, spawn and capture start CreateProcessW, netlines cuts its lines out of what
recv gives (Windows' C library cannot open a FILE over a socket), and signals is C's signal and
raise with SIGTERM and a pipe of C's library, since Windows has no SIGUSR1. dates takes the moments
apart by hand (Howard Hinnant's civil_from_days), since gmtime_s refuses a moment before 1970: so on
Windows it is the same algorithm in both, where on macOS it is the C library's gmtime_r. Windows has no
cat and no true, so run.py builds both from bench/windows and puts them first on PATH, for SG and
C alike. case is compared by the size of its text only: its C maps a character to one by the tables of
the C library, and Windows' writes three bytes more each line.
What the numbers say
Most of the rest is the time of C, as on macOS. Where SG is much faster, the difference is Windows' C
library: printf (show, numbers, floats), fgets (lines, copy, asynccopy, pipelines,
pipesasync) and the functions of wide characters (case, letters) are several times slower there than
on macOS, and SG uses none of them. signals is 0.03 because Signals.raise on Windows hands the signal on
inside the program, where C's raise calls the handler, which writes into a pipe. Where SG is slower:
tree, counts, async and tasklist as on macOS; dates, listing, echo and echosync on Windows
only, in Slower than C.