GCC 5 Through GCC 10 Compiler Benchmarks - Five Years Worth Of C/C++ Compiler Performance

[phoronix]

Phoronix: GCC 5 Through GCC 10 Compiler Benchmarks - Five Years Worth Of C/C++ Compiler Performance

As part of our end-of-year benchmark comparisons, the latest results are looking at how the GNU Compiler Collection has evolved with the past five years of performance in testing GCC 5 through GCC 9 stable and the latest GCC 10 development compiler from the same system.

http://www.phoronix.com/vr.php?view=28679

[FireBurn]

What was -march=native picking for the older versions of GCC? This command should tell you:

 
 gcc -march=native -E -v - </dev/null 2>&1 | grep cc1

[r08z]

Anyone using GCC 8.3 / 9.2 with weekly git patches are the smart ones using the best compiler.

[rene]

so noch much performance progress, much slower compile times, ..? :-/ https://www.youtube.com/watch?v=qP5JcmjfCBY

[eva2000]

Would of thought using a Intel Skylake or newer cpu would better gauge GCC 8+ improvements than an old Intel Haswell cpu ?

[Michael]

Would of thought using a Intel Skylake or newer cpu would better gauge GCC 8+ improvements than an old Intel Haswell cpu ?

For GCC 8+, sure, but this comparison went back to GCC 5.

[frief]

the compiler metric "size of compiled binary" might be interesting too.
(if not then this might be an indicator that hardware has won over software)

[pyler]

Compile times increase for no real gain :/ Confirmed.

GCC devs should really care about reasonable compile times too. Instead, they are totally focused on SPEC benchmarks :/

[eva2000]

For GCC 8+, sure, but this comparison went back to GCC 5.

Yeah. Round 2 with such a cpu would be interesting

[hubicka]

Compile times increase for no real gain :/ Confirmed.

GCC devs should really care about reasonable compile times too. Instead, they are totally focused on SPEC benchmarks :/

I usually spend about half of my GCC development efforts on optimizing compile time and memory use
One problem is that every release new ISA extensions and similar features are added which bloats the compiler up and makes it slower on building small files. One way to deal the problem is to build you compiler with

configure --with-build-config=bootstrap-lto-lean
make profiledbootstrap

this helps to optimize code placement and optimize cold sections for size and helps to reduce the collateral damage of GCC getting bigger.