Change Details

The old CPU model only had MLA->MLA forwarding. I added some more accumulate op to accumulate op, and non-accumulate op (e.g. MUL, shift) to relevant accumulate op (e.g. MLA, SRA) forwarding according to the Cortex A57 Software Optimization Guide. The patch improves performance in some internal benchmarks and causes no significant regressions (none in SPEC).

The old CPU model only had MLA->MLA forwarding. I added some missing MUL->MLA read advances and a missing absolute diff accumulator read advance according to the Cortex A57 Software Optimization Guide. Also added missing schedules for ASIMD shift by immed basic, and ASIMD shift by register, basic, D-form. The patch improves performance in EEMBC benchmark and causes no significant regressions (none in SPEC).

The old CPU model only had MLA->MLA forwarding. I added some moreissing accumulate op to accumulate op, and non-accumulate op (e.g. MUL, shift)MUL->MLA read advances and a missing absolute diff accumulator read to relevant accumulate op (e.g. MLAadvance according to the Cortex A57 Software Optimization Guide. Also added missing schedules for ASIMD shift by immed basic, SRA) forwarding according to theand Cortex A57 Software Optimization GuideASIMD shift by register, basic, D-form. The patch improves performance in some internal benchmarks andEEMBC benchmark and causes no causes no significant regressions (none in SPEC).