5.3. Compiler Optimization

As was shown in Embecosm Application Note 6 High Performance SoC Modeling with Verilator: A Tutorial for Cycle Accurate SystemC Model Creation and Optimization [10], compiler optimization has a very significant impact on performance.

The effect of compiler optimizations can be seen in Table 5.3. This compares the performance using no optimization, optimization using -Os (the best for single pass optimization) and profile directed optimization using -O3 and profile statistics from a debugging run using the Ackermann's Function test program. To build a model using profile directed optimization with optimization level -O3 use the run-fast target of the Makefile in the distribution.

make run-fast OPT=-O3

Run Description	Cycles	Time	Perf	Load Rate
No optimization(-O0)	8,097,241	257.34 s	31 kHz	164 bytes/s
Single pass optimization (-Os)	8,097,241	124.71 s	65 kHz	383 bytes/s
Profile directed optimization (-O3)	8,097,241	92.69 s	87 kHz	487 bytes/s

Table 5.3. Effect of compiler optimization on GDB server and client performance.

Using a single optimization pass more than doubles performance of both the server and client load times. Using profile directed optimization almost trebles performance.