[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [PATCH 5/9] xen/bitops: Introduce generic_hweightl() and hweightl()


  • To: Jan Beulich <jbeulich@xxxxxxxx>
  • From: Andrew Cooper <andrew.cooper3@xxxxxxxxxx>
  • Date: Tue, 27 Aug 2024 13:30:53 +0100
  • Autocrypt: addr=andrew.cooper3@xxxxxxxxxx; keydata= xsFNBFLhNn8BEADVhE+Hb8i0GV6mihnnr/uiQQdPF8kUoFzCOPXkf7jQ5sLYeJa0cQi6Penp VtiFYznTairnVsN5J+ujSTIb+OlMSJUWV4opS7WVNnxHbFTPYZVQ3erv7NKc2iVizCRZ2Kxn srM1oPXWRic8BIAdYOKOloF2300SL/bIpeD+x7h3w9B/qez7nOin5NzkxgFoaUeIal12pXSR Q354FKFoy6Vh96gc4VRqte3jw8mPuJQpfws+Pb+swvSf/i1q1+1I4jsRQQh2m6OTADHIqg2E ofTYAEh7R5HfPx0EXoEDMdRjOeKn8+vvkAwhviWXTHlG3R1QkbE5M/oywnZ83udJmi+lxjJ5 YhQ5IzomvJ16H0Bq+TLyVLO/VRksp1VR9HxCzItLNCS8PdpYYz5TC204ViycobYU65WMpzWe LFAGn8jSS25XIpqv0Y9k87dLbctKKA14Ifw2kq5OIVu2FuX+3i446JOa2vpCI9GcjCzi3oHV e00bzYiHMIl0FICrNJU0Kjho8pdo0m2uxkn6SYEpogAy9pnatUlO+erL4LqFUO7GXSdBRbw5 gNt25XTLdSFuZtMxkY3tq8MFss5QnjhehCVPEpE6y9ZjI4XB8ad1G4oBHVGK5LMsvg22PfMJ ISWFSHoF/B5+lHkCKWkFxZ0gZn33ju5n6/FOdEx4B8cMJt+cWwARAQABzSlBbmRyZXcgQ29v cGVyIDxhbmRyZXcuY29vcGVyM0BjaXRyaXguY29tPsLBegQTAQgAJAIbAwULCQgHAwUVCgkI CwUWAgMBAAIeAQIXgAUCWKD95wIZAQAKCRBlw/kGpdefoHbdD/9AIoR3k6fKl+RFiFpyAhvO 59ttDFI7nIAnlYngev2XUR3acFElJATHSDO0ju+hqWqAb8kVijXLops0gOfqt3VPZq9cuHlh IMDquatGLzAadfFx2eQYIYT+FYuMoPZy/aTUazmJIDVxP7L383grjIkn+7tAv+qeDfE+txL4 SAm1UHNvmdfgL2/lcmL3xRh7sub3nJilM93RWX1Pe5LBSDXO45uzCGEdst6uSlzYR/MEr+5Z JQQ32JV64zwvf/aKaagSQSQMYNX9JFgfZ3TKWC1KJQbX5ssoX/5hNLqxMcZV3TN7kU8I3kjK mPec9+1nECOjjJSO/h4P0sBZyIUGfguwzhEeGf4sMCuSEM4xjCnwiBwftR17sr0spYcOpqET ZGcAmyYcNjy6CYadNCnfR40vhhWuCfNCBzWnUW0lFoo12wb0YnzoOLjvfD6OL3JjIUJNOmJy RCsJ5IA/Iz33RhSVRmROu+TztwuThClw63g7+hoyewv7BemKyuU6FTVhjjW+XUWmS/FzknSi dAG+insr0746cTPpSkGl3KAXeWDGJzve7/SBBfyznWCMGaf8E2P1oOdIZRxHgWj0zNr1+ooF /PzgLPiCI4OMUttTlEKChgbUTQ+5o0P080JojqfXwbPAyumbaYcQNiH1/xYbJdOFSiBv9rpt TQTBLzDKXok86M7BTQRS4TZ/ARAAkgqudHsp+hd82UVkvgnlqZjzz2vyrYfz7bkPtXaGb9H4 Rfo7mQsEQavEBdWWjbga6eMnDqtu+FC+qeTGYebToxEyp2lKDSoAsvt8w82tIlP/EbmRbDVn 7bhjBlfRcFjVYw8uVDPptT0TV47vpoCVkTwcyb6OltJrvg/QzV9f07DJswuda1JH3/qvYu0p vjPnYvCq4NsqY2XSdAJ02HrdYPFtNyPEntu1n1KK+gJrstjtw7KsZ4ygXYrsm/oCBiVW/OgU g/XIlGErkrxe4vQvJyVwg6YH653YTX5hLLUEL1NS4TCo47RP+wi6y+TnuAL36UtK/uFyEuPy wwrDVcC4cIFhYSfsO0BumEI65yu7a8aHbGfq2lW251UcoU48Z27ZUUZd2Dr6O/n8poQHbaTd 6bJJSjzGGHZVbRP9UQ3lkmkmc0+XCHmj5WhwNNYjgbbmML7y0fsJT5RgvefAIFfHBg7fTY/i kBEimoUsTEQz+N4hbKwo1hULfVxDJStE4sbPhjbsPCrlXf6W9CxSyQ0qmZ2bXsLQYRj2xqd1 bpA+1o1j2N4/au1R/uSiUFjewJdT/LX1EklKDcQwpk06Af/N7VZtSfEJeRV04unbsKVXWZAk uAJyDDKN99ziC0Wz5kcPyVD1HNf8bgaqGDzrv3TfYjwqayRFcMf7xJaL9xXedMcAEQEAAcLB XwQYAQgACQUCUuE2fwIbDAAKCRBlw/kGpdefoG4XEACD1Qf/er8EA7g23HMxYWd3FXHThrVQ HgiGdk5Yh632vjOm9L4sd/GCEACVQKjsu98e8o3ysitFlznEns5EAAXEbITrgKWXDDUWGYxd pnjj2u+GkVdsOAGk0kxczX6s+VRBhpbBI2PWnOsRJgU2n10PZ3mZD4Xu9kU2IXYmuW+e5KCA vTArRUdCrAtIa1k01sPipPPw6dfxx2e5asy21YOytzxuWFfJTGnVxZZSCyLUO83sh6OZhJkk b9rxL9wPmpN/t2IPaEKoAc0FTQZS36wAMOXkBh24PQ9gaLJvfPKpNzGD8XWR5HHF0NLIJhgg 4ZlEXQ2fVp3XrtocHqhu4UZR4koCijgB8sB7Tb0GCpwK+C4UePdFLfhKyRdSXuvY3AHJd4CP 4JzW0Bzq/WXY3XMOzUTYApGQpnUpdOmuQSfpV9MQO+/jo7r6yPbxT7CwRS5dcQPzUiuHLK9i nvjREdh84qycnx0/6dDroYhp0DFv4udxuAvt1h4wGwTPRQZerSm4xaYegEFusyhbZrI0U9tJ B8WrhBLXDiYlyJT6zOV2yZFuW47VrLsjYnHwn27hmxTC/7tvG3euCklmkn9Sl9IAKFu29RSo d5bD8kMSCYsTqtTfT6W4A3qHGvIDta3ptLYpIAOD2sY3GYq2nf3Bbzx81wZK14JdDDHUX2Rs 6+ahAA==
  • Cc: Roger Pau Monné <roger.pau@xxxxxxxxxx>, Wei Liu <wl@xxxxxxx>, Stefano Stabellini <sstabellini@xxxxxxxxxx>, Julien Grall <julien@xxxxxxx>, Volodymyr Babchuk <Volodymyr_Babchuk@xxxxxxxx>, Bertrand Marquis <bertrand.marquis@xxxxxxx>, Michal Orzel <michal.orzel@xxxxxxx>, Oleksii Kurochko <oleksii.kurochko@xxxxxxxxx>, Shawn Anastasio <sanastasio@xxxxxxxxxxxxxxxxxxxxx>, Xen-devel <xen-devel@xxxxxxxxxxxxxxxxxxxx>
  • Delivery-date: Tue, 27 Aug 2024 12:31:06 +0000
  • List-id: Xen developer discussion <xen-devel.lists.xenproject.org>

On 27/08/2024 12:41 pm, Jan Beulich wrote:
> On 27.08.2024 12:39, Andrew Cooper wrote:
>> On 26/08/2024 12:40 pm, Jan Beulich wrote:
>>> On 23.08.2024 01:06, Andrew Cooper wrote:
>>> --- a/xen/include/xen/bitops.h
>>> +++ b/xen/include/xen/bitops.h
>>> @@ -35,6 +35,12 @@ extern void __bitop_bad_size(void);
>>>  unsigned int __pure generic_ffsl(unsigned long x);
>>>  unsigned int __pure generic_flsl(unsigned long x);
>>>  
>>>> +/*
>>>> + * Hamming Weight, also called Population Count.  Returns the number of 
>>>> set
>>>> + * bits in @x.
>>>> + */
>>>> +unsigned int __pure generic_hweightl(unsigned long x);
>>> Aren't this and ...
>>>
>>>> @@ -284,6 +290,18 @@ static always_inline __pure unsigned int 
>>>> fls64(uint64_t x)
>>>>          (_v & (_v - 1)) != 0;                   \
>>>>      })
>>>>  
>>>> +static always_inline __pure unsigned int hweightl(unsigned long x)
>>> ... this even __attribute_const__?
>> Hmm.  This is following fls(), but on further consideration, they should
>> be const too.
>>
>> I'll do a prep patch fixing that, although I'm going to rename it to
>> __attr_const for brevity.
>>
>> Much as I'd prefer __const, I expect that is going too far, making it
>> too close to regular const.
> I was actually going to suggest using that name, if we want to shorten
> __attribute_const__. Yes, gcc treats __const (and __const__) as
> keywords, but do we care (much)? All it requires is that we don't start
> using __const as a (real) keyword.

Well also we'll get into more MISRA fun for overriding keywords.

But yes - the fact that GCC treats __const to mean const is precisely
why we shouldn't give it an unrelated meaning.

>
> Of course __const is a good example of why really we shouldn't use
> double-underscore prefixed names anywhere. Any of them can be assigned
> a meaning by the compiler, and here that's clearly the case. Therefore,
> taking your planned rename, maybe better make it attr_const then? And
> eventually switch stuff like __packed, __pure, and __weak to attr_* as
> well? Or even introduce something like
>
> #define attr(attr...) __attribute__((attr))
>
> and use attr(const) here?

Hmm - that's an interesting approach, and for other attributes which we
can use unconditionally.  It will end up shorter than multiple separate
__-prefixed names.

As a tangent, I've got some work from playing with -fanalyzer which
sprinkles some attr malloc/alloc_{size,align}()/free around.  It does
improve code generation (abeit marginally), but the function declaration
size suffers.

It won't work for attributes which are conditionally nothing (e.g.
cf_check), or ones that contain multiple aspects (e.g. __constructer
conataining cf_check).

In practice this means we're always going to end up with a mix, so maybe
attr_const is better for consistency.

>
>>>> +{
>>>> +    if ( __builtin_constant_p(x) )
>>>> +        return __builtin_popcountl(x);
>>> How certain are you that compilers (even old ones) will reliably fold
>>> constant expressions here, and never emit a libgcc call instead? The
>>> conditions look to be more tight than just __builtin_constant_p(); a
>>> pretty absurd example:
>>>
>>> unsigned ltest(void) {
>>>     return __builtin_constant_p("") ? __builtin_popcountl((unsigned 
>>> long)"") : ~0;
>>> }
>> How do you express that in terms of a call to hweightl()?
> hweightl((unsigned long)"");
>
> Yet as said - it's absurd. It merely serves to make the point that what
> __builtin_constant_p() returns true for doesn't necessarily constant-
> fold in expressions.

Yes, but as shown in the godbolt link, this form changes GCC's mind
about the __builtin_const-ness of the expression.

>
>> Again, this is following the layout started with fls() in order to avoid
>> each arch opencoding different versions of constant folding.
>>
>> https://godbolt.org/z/r544c49oY
>>
>> When it's forced through the hweightl() interface, even GCC 4.1 decides
>> that it's non-constant and falls back to generic_hweightl().
>>
>>
>> I did spend a *lot* of time with the fls() series checking that all
>> compilers we supported did what we wanted in this case, so I don't
>> expect it to be a problem.
> Right, and I guess I was pointlessly more concerned about popcount than
> I was for ffs() / fls(). The criteria upon which gcc decides whether to
> constant-fold the uses is exactly the same.
>
>>   But, if a library call is emitted, it will
>> be very obvious (link failure), and we can re-evaluate.
> Indeed, we certainly would notice, albeit the diagnostic may be cryptic
> to people.
>
> Bottom line - keep it as is.

Thanks.

~Andrew



 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.