- Counts the number of uppercase characters in a body of text. To simulate a large load, the same sentence is processed 10 million times and observation of JIT activity (C1, C2)
time java -XX:+PrintCompilation CountUppercase.java Hello World > jit-compiler.log
#1 (142 ms)
#2 (99 ms)
#3 (59 ms)
#4 (61 ms)
#5 (60 ms)
#6 (56 ms)
#7 (56 ms)
#8 (65 ms)
#9 (61 ms)
#total: 19999998 (717 ms)
#1,44s user 0,12s system 131% cpu 1,188 total 352848 kb rss
We can observe the activity of C1 (3), via Tiered Compilation. All the profiling data collected, together with the JIT compilation thresholds, triggers C2 (4). The runtime proceeds to the de-optimization phases when the compilation hypotheses are wrong (made not entering).
For more information on compilation logs, see here
- Execution in pure interpretation mode (JIT-less*)
time java -Xint CountUppercase.java Hello World
# 😴😪🥱💤🛌🏼
#1 (3305 ms)
#2 (3317 ms)
#3 (3219 ms)
#4 (3245 ms)
#5 (3272 ms)
#6 (3303 ms)
#7 (3281 ms)
#8 (3260 ms)
#9 (3270 ms)
#total: 19999998 (32722 ms)
#33,34s user 0,21s system 99% cpu 33,769 total 211296 kb rssVery, very serious performance degradation, as we no longer benefit from JIT. The runtime's interpreter mode is very costly.