Re: powerpc/64s: exception optimise MSR handling

2016-09-20 Thread Michael Ellerman
On Thu, 2016-15-09 at 09:04:46 UTC, Nicholas Piggin wrote: > mtmsrd with L=1 only affects MSR_EE and MSR_RI bits, and we always > know what state those bits are, so the kernel MSR does not need to be > loaded when modifying them. > > mtmsrd is often in the critical execution path, so avoiding depe

Re: [PATCH] powerpc/64s: exception optimise MSR handling

2016-09-19 Thread Nicholas Piggin
On Tue, 20 Sep 2016 14:25:48 +1000 Michael Ellerman wrote: > Nicholas Piggin writes: > > > mtmsrd with L=1 only affects MSR_EE and MSR_RI bits, and we always > > know what state those bits are, so the kernel MSR does not need to be > > loaded when modifying them. > > > > mtmsrd is often in the

Re: [PATCH] powerpc/64s: exception optimise MSR handling

2016-09-19 Thread Michael Ellerman
Nicholas Piggin writes: > mtmsrd with L=1 only affects MSR_EE and MSR_RI bits, and we always > know what state those bits are, so the kernel MSR does not need to be > loaded when modifying them. > > mtmsrd is often in the critical execution path, so avoiding dependency > on even L1 load is notica

[PATCH] powerpc/64s: exception optimise MSR handling

2016-09-15 Thread Nicholas Piggin
mtmsrd with L=1 only affects MSR_EE and MSR_RI bits, and we always know what state those bits are, so the kernel MSR does not need to be loaded when modifying them. mtmsrd is often in the critical execution path, so avoiding dependency on even L1 load is noticable. On a POWER8 this saves about 3 c