From 70fe3d980f5f14d8125869125ba9a0ea95e09c6b Mon Sep 17 00:00:00 2001 From: Cyril Bur Date: Mon, 29 Feb 2016 17:53:47 +1100 Subject: powerpc: Restore FPU/VEC/VSX if previously used Currently the FPU, VEC and VSX facilities are lazily loaded. This is not a problem unless a process is using these facilities. Modern versions of GCC are very good at automatically vectorising code, new and modernised workloads make use of floating point and vector facilities, even the kernel makes use of vectorised memcpy. All this combined greatly increases the cost of a syscall since the kernel uses the facilities sometimes even in syscall fast-path making it increasingly common for a thread to take an *_unavailable exception soon after a syscall, not to mention potentially taking all three. The obvious overcompensation to this problem is to simply always load all the facilities on every exit to userspace. Loading up all FPU, VEC and VSX registers every time can be expensive and if a workload does avoid using them, it should not be forced to incur this penalty. An 8bit counter is used to detect if the registers have been used in the past and the registers are always loaded until the value wraps to back to zero. Several versions of the assembly in entry_64.S were tested: 1. Always calling C. 2. Performing a common case check and then calling C. 3. A complex check in asm. After some benchmarking it was determined that avoiding C in the common case is a performance benefit (option 2). The full check in asm (option 3) greatly complicated that codepath for a negligible performance gain and the trade-off was deemed not worth it. Signed-off-by: Cyril Bur [mpe: Move load_vec in the struct to fill an existing hole, reword change log] Signed-off-by: Michael Ellerman fixup --- arch/powerpc/kernel/vector.S | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'arch/powerpc/kernel/vector.S') diff --git a/arch/powerpc/kernel/vector.S b/arch/powerpc/kernel/vector.S index 162d0f7..038cff8 100644 --- a/arch/powerpc/kernel/vector.S +++ b/arch/powerpc/kernel/vector.S @@ -91,6 +91,10 @@ _GLOBAL(load_up_altivec) oris r12,r12,MSR_VEC@h std r12,_MSR(r1) #endif + /* Don't care if r4 overflows, this is desired behaviour */ + lbz r4,THREAD_LOAD_VEC(r5) + addi r4,r4,1 + stb r4,THREAD_LOAD_VEC(r5) addi r6,r5,THREAD_VRSTATE li r4,1 li r10,VRSTATE_VSCR -- cgit v1.1 From 6f515d842e8e1b205e54f44b9013bf14870b97a7 Mon Sep 17 00:00:00 2001 From: Cyril Bur Date: Mon, 29 Feb 2016 17:53:50 +1100 Subject: powerpc: Add the ability to save Altivec without giving it up This patch adds the ability to be able to save the VEC registers to the thread struct without giving up (disabling the facility) next time the process returns to userspace. This patch builds on a previous optimisation for the FPU registers in the thread copy path to avoid a possibly pointless reload of VEC state. Signed-off-by: Cyril Bur Signed-off-by: Michael Ellerman --- arch/powerpc/kernel/vector.S | 24 ++++-------------------- 1 file changed, 4 insertions(+), 20 deletions(-) (limited to 'arch/powerpc/kernel/vector.S') diff --git a/arch/powerpc/kernel/vector.S b/arch/powerpc/kernel/vector.S index 038cff8..51b0c17 100644 --- a/arch/powerpc/kernel/vector.S +++ b/arch/powerpc/kernel/vector.S @@ -106,36 +106,20 @@ _GLOBAL(load_up_altivec) blr /* - * __giveup_altivec(tsk) - * Disable VMX for the task given as the argument, - * and save the vector registers in its thread_struct. + * save_altivec(tsk) + * Save the vector registers to its thread_struct */ -_GLOBAL(__giveup_altivec) +_GLOBAL(save_altivec) addi r3,r3,THREAD /* want THREAD of task */ PPC_LL r7,THREAD_VRSAVEAREA(r3) PPC_LL r5,PT_REGS(r3) PPC_LCMPI 0,r7,0 bne 2f addi r7,r3,THREAD_VRSTATE -2: PPC_LCMPI 0,r5,0 - SAVE_32VRS(0,r4,r7) +2: SAVE_32VRS(0,r4,r7) mfvscr v0 li r4,VRSTATE_VSCR stvx v0,r4,r7 - beq 1f - PPC_LL r4,_MSR-STACK_FRAME_OVERHEAD(r5) -#ifdef CONFIG_VSX -BEGIN_FTR_SECTION - lis r3,(MSR_VEC|MSR_VSX)@h -FTR_SECTION_ELSE - lis r3,MSR_VEC@h -ALT_FTR_SECTION_END_IFSET(CPU_FTR_VSX) -#else - lis r3,MSR_VEC@h -#endif - andc r4,r4,r3 /* disable FP for previous task */ - PPC_STL r4,_MSR-STACK_FRAME_OVERHEAD(r5) -1: blr #ifdef CONFIG_VSX -- cgit v1.1 From bf6a4d5b75d1ea87897fe68d0e45d35a2996c678 Mon Sep 17 00:00:00 2001 From: Cyril Bur Date: Mon, 29 Feb 2016 17:53:51 +1100 Subject: powerpc: Add the ability to save VSX without giving it up This patch adds the ability to be able to save the VSX registers to the thread struct without giving up (disabling the facility) next time the process returns to userspace. This patch builds on a previous optimisation for the FPU and VEC registers in the thread copy path to avoid a possibly pointless reload of VSX state. Signed-off-by: Cyril Bur Signed-off-by: Michael Ellerman --- arch/powerpc/kernel/vector.S | 17 ----------------- 1 file changed, 17 deletions(-) (limited to 'arch/powerpc/kernel/vector.S') diff --git a/arch/powerpc/kernel/vector.S b/arch/powerpc/kernel/vector.S index 51b0c17..1c2e7a3 100644 --- a/arch/powerpc/kernel/vector.S +++ b/arch/powerpc/kernel/vector.S @@ -151,23 +151,6 @@ _GLOBAL(load_up_vsx) std r12,_MSR(r1) b fast_exception_return -/* - * __giveup_vsx(tsk) - * Disable VSX for the task given as the argument. - * Does NOT save vsx registers. - */ -_GLOBAL(__giveup_vsx) - addi r3,r3,THREAD /* want THREAD of task */ - ld r5,PT_REGS(r3) - cmpdi 0,r5,0 - beq 1f - ld r4,_MSR-STACK_FRAME_OVERHEAD(r5) - lis r3,MSR_VSX@h - andc r4,r4,r3 /* disable VSX for previous task */ - std r4,_MSR-STACK_FRAME_OVERHEAD(r5) -1: - blr - #endif /* CONFIG_VSX */ -- cgit v1.1