This enables targets to adapt their pass pipeline to the register allocator in use. For example, with the AArch64 backend, using PBQP with the cortex-a57, the FPLoadBalancing pass is no longer necessary. llvm-svn: 220321
322 B
322 B