Some PowerPC CPU may have slow MFLR instruction, so we need to schedule the MFLR and its store in function prologue away to hidden the long latency for slow MFLR instruction. This patch adds a new feature fastMFLR and the new feature will be used in https://reviews.llvm.org/D137423. Reviewed By: RolandF Differential Revision: https://reviews.llvm.org/D137612
8.6 KiB
8.6 KiB