Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]


Groups > linux.kernel > #1242584

Re: [PATCH v2 06/22] arm64: sys_reg: Define System register encoding

From "Suzuki K. Poulose" <Suzuki.Poulose@arm.com>
Newsgroups linux.kernel
Subject Re: [PATCH v2 06/22] arm64: sys_reg: Define System register encoding
Date 2015-10-08 18:20 +0200
Message-ID <qhm5c-bp-3@gated-at.bofh.it> (permalink)
References <qghqV-5ar-3@gated-at.bofh.it> <qghqW-5ar-39@gated-at.bofh.it> <qgZV1-1TQ-37@gated-at.bofh.it> <qh0o2-2H0-15@gated-at.bofh.it> <qhkG6-6u9-35@gated-at.bofh.it>
Organization linux.* mail to news gateway

Show all headers | View raw


On 08/10/15 15:43, Catalin Marinas wrote:
> On Wed, Oct 07, 2015 at 06:03:34PM +0100, Suzuki K. Poulose wrote:
>> On 07/10/15 17:36, Catalin Marinas wrote:
>>> On Mon, Oct 05, 2015 at 06:01:55PM +0100, Suzuki K. Poulose wrote:
>>>>    * (Ref: ARMv8 ARM, Section: "System instruction class encoding overview",
>>>>    *  C5.2, version:ARM DDI 0487A.f)
>>>>    *	[20-19] : Op0
>>>> @@ -34,15 +34,40 @@
>>>>    *	[15-12] : CRn
>>>>    *	[11-8]  : CRm
>>>>    *	[7-5]   : Op2
>>>> + * Hence we use [ sys_reg() << 5 ] in the mrs/msr instructions.
>>>
>>> Do we really need to have all the ids shifted right by 5? I can't see
>>> where it helps. OTOH, it makes the code more complicated by having to
>>> remember to shift the id left by 5.
>>
>> This is a cosmetic change, to reuse the sys_reg() definitions for both
>> mrs_s/msr_s macros and the CPU ID. The (existing)left shift is only needed
>> for the mrs_s/msr_s, so the existing users don't have to worry about the shift
>> unless they have hard-open-coded values for the register. On the plus
>> side it becomes slightly easier to use the same encoding for CPU id
>> tracking (and manual decoding). If you think this is superfluous, I could
>> change the CPU-id to use right shifted values from sys_reg.
>
> You may be right but I still fail to see whether the shifted values are
> useful to the CPUID code. Are you referring to the 'switch' statements?
>

Yes. Again we can adjust the macros to get the right fields for switch().
Its also about the meaning of sys_reg(). The name kind of implies the system
register encoding defined by ARM ARM. However the current version
defines the encoding as it appears in the mrs/msr instructions, which can
be confusing if we start using the same encoding for representing the sys_reg
elsewhere(in this case CPUID infrastructure). Again, I am not too particular
about this change.

Suzuki


--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Back to linux.kernel | Previous | NextPrevious in thread | Find similar | Unroll thread


Thread

Re: [PATCH v2 06/22] arm64: sys_reg: Define System register encoding Catalin Marinas <catalin.marinas@arm.com> - 2015-10-07 18:40 +0200
  Re: [PATCH v2 06/22] arm64: sys_reg: Define System register encoding "Suzuki K. Poulose" <Suzuki.Poulose@arm.com> - 2015-10-07 19:10 +0200
    Re: [PATCH v2 06/22] arm64: sys_reg: Define System register encoding Catalin Marinas <catalin.marinas@arm.com> - 2015-10-08 16:50 +0200
      Re: [PATCH v2 06/22] arm64: sys_reg: Define System register encoding "Suzuki K. Poulose" <Suzuki.Poulose@arm.com> - 2015-10-08 18:20 +0200

csiph-web