Canonical published the results of a performance evaluation study on package optimization through Profile-Guided Optimization (PGO). This form of optimization analyzes program features to generate more efficient code. The study found that implementing PGO optimization led to a 5-7% reduction in CPU load and faster assembly times.
The experiment involved measuring the performance of various software packages such as assembly, GDB, EMACS, and Python3.12 in a Qemu virtual environment emulating a RISC-V system on an X86_64 architecture computer. A comparison was made between default Qemu settings and those with PGO optimization enabled. The selected packages represented different programming languages and had code volumes equivalent to 1-6 hours of assembly time in the emulator.
From a practical standpoint, the experiment demonstrated improved performance in the assembly environment for the RISC-V platform when using an emulator. With PGO optimization, the time taken for assembly work reduced by two hours per day, allowing for two additional full assemblies on the same server.
