This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: Some wide-int review comments
- From: Richard Biener <richard dot guenther at gmail dot com>
- To: Kenneth Zadeck <zadeck at naturalbridge dot com>
- Cc: "Joseph S. Myers" <joseph at codesourcery dot com>, GCC Patches <gcc-patches at gcc dot gnu dot org>, Mike Stump <mikestump at comcast dot net>, Richard Sandiford <rdsandiford at googlemail dot com>
- Date: Wed, 27 Nov 2013 11:50:48 +0100
- Subject: Re: Some wide-int review comments
- Authentication-results: sourceware.org; auth=none
- References: <87mwlfgp8s dot fsf at talisman dot default> <527E456B dot 4050203 at naturalbridge dot com> <CAFiYyc2qJB3onVWWqBu5sYMQE6Otpmys5U3ZpyQtWAfKOFCcCQ at mail dot gmail dot com> <5280E914 dot 2060501 at naturalbridge dot com> <CAFiYyc0tNHO54aS7PaRKLdsWMQ6qN++MdN8=63Eq1AD5GtHrPg at mail dot gmail dot com> <5280F219 dot 6050000 at naturalbridge dot com> <CAFiYyc0LmDSAtMYqAe4kBr=4WcAzNtQm=cjbeX+1OGGyyjS-=g at mail dot gmail dot com> <52825432 dot 5070502 at naturalbridge dot com> <Pine dot LNX dot 4 dot 64 dot 1311121624220 dot 31448 at digraph dot polyomino dot org dot uk> <52825AC9 dot 9050604 at naturalbridge dot com> <CAFiYyc2Zph0ROZcXuapeP-7uXrWV1qXW2Znjseg6Dd0w2e0M2A at mail dot gmail dot com> <528D4649 dot 5020804 at naturalbridge dot com> <5294CD41 dot 1040008 at naturalbridge dot com>
On Tue, Nov 26, 2013 at 5:33 PM, Kenneth Zadeck
<zadeck@naturalbridge.com> wrote:
> Richi,
>
> patch ping
Ok.
Thanks,
Richard.
> also two more pieces of information. With further testing, this seems to
> fix
>
> Tests that now work, but didn't before:
> ===============
> ext/random/hypergeometric_distribution/operators/values.cc (test for excess
> errors)
>
> New tests that PASS:
>
> ext/random/hypergeometric_distribution/operators/values.cc execution test
> ================
>
> also, the corresponding frag for fold-const.c on the wide-int branch will
> look like
> ================
> Index: gcc/fold-const.c
> ===================================================================
> --- gcc/fold-const.c (revision 205224)
> +++ gcc/fold-const.c (working copy)
> @@ -1030,51 +1030,51 @@ int_const_binop_1 (enum tree_code code,
>
> case TRUNC_DIV_EXPR:
> case EXACT_DIV_EXPR:
> - res = wi::div_trunc (arg1, arg2, sign, &overflow);
> - if (overflow)
> + if (arg2 == 0)
> return NULL_TREE;
> + res = wi::div_trunc (arg1, arg2, sign, &overflow);
> break;
>
> case FLOOR_DIV_EXPR:
> - res = wi::div_floor (arg1, arg2, sign, &overflow);
> - if (overflow)
> + if (arg2 == 0)
> return NULL_TREE;
> + res = wi::div_floor (arg1, arg2, sign, &overflow);
> break;
>
> case CEIL_DIV_EXPR:
> - res = wi::div_ceil (arg1, arg2, sign, &overflow);
> - if (overflow)
> + if (arg2 == 0)
> return NULL_TREE;
> + res = wi::div_ceil (arg1, arg2, sign, &overflow);
> break;
>
> case ROUND_DIV_EXPR:
> - res = wi::div_round (arg1, arg2, sign, &overflow);
> - if (overflow)
> + if (arg2 == 0)
> return NULL_TREE;
> + res = wi::div_round (arg1, arg2, sign, &overflow);
> break;
>
> case TRUNC_MOD_EXPR:
> - res = wi::mod_trunc (arg1, arg2, sign, &overflow);
> - if (overflow)
> + if (arg2 == 0)
> return NULL_TREE;
> + res = wi::mod_trunc (arg1, arg2, sign, &overflow);
> break;
>
> case FLOOR_MOD_EXPR:
> - res = wi::mod_floor (arg1, arg2, sign, &overflow);
> - if (overflow)
> + if (arg2 == 0)
> return NULL_TREE;
> + res = wi::mod_floor (arg1, arg2, sign, &overflow);
> break;
>
> case CEIL_MOD_EXPR:
> - res = wi::mod_ceil (arg1, arg2, sign, &overflow);
> - if (overflow)
> + if (arg2 == 0)
> return NULL_TREE;
> + res = wi::mod_ceil (arg1, arg2, sign, &overflow);
> break;
>
> case ROUND_MOD_EXPR:
> - res = wi::mod_round (arg1, arg2, sign, &overflow);
> - if (overflow)
> + if (arg2 == 0)
> return NULL_TREE;
> + res = wi::mod_round (arg1, arg2, sign, &overflow);
> break;
>
> case MIN_EXPR:
>
> ================
>
> On 11/20/2013 06:31 PM, Kenneth Zadeck wrote:
>>
>> On 11/13/2013 04:59 AM, Richard Biener wrote:
>>>
>>> On Tue, Nov 12, 2013 at 5:43 PM, Kenneth Zadeck
>>> <zadeck@naturalbridge.com> wrote:
>>>>
>>>> On 11/12/2013 11:27 AM, Joseph S. Myers wrote:
>>>>>
>>>>> On Tue, 12 Nov 2013, Kenneth Zadeck wrote:
>>>>>
>>>>>> Richi,
>>>>>>
>>>>>> i am having a little trouble putting this back the way that you want.
>>>>>> The
>>>>>> issue is rem.
>>>>>> what is supposed to happen for INT_MIN % -1?
>>>>>>
>>>>>> I would assume because i am failing the last case of
>>>>>> gcc.dg/c90-const-expr-8.c
>>>>>> that INT_MIN %-1 should not overflow even if INT_MIN / -1 does.
>>>>>> however,
>>>>>
>>>>> Given the conclusion in C11 that a%b should be considered undefined if
>>>>> a/b
>>>>> is not representable, I think it's reasonable to say INT_MIN % -1
>>>>> *should*
>>>>> be considered to overflow (for all C standard versions) (and bug 30484
>>>>> is
>>>>> only a bug for -fwrapv).
>>>>>
>>>> however, my local question is what do we want the api to be
>>>> int-const-binop-1? The existing behavior seems to be that at least
>>>> for
>>>> common modes this function silently returns 0 and it is up to the front
>>>> ends
>>>> to put their own spin on it.
>>>
>>> For wide-int you create 1:1 the behavior of current trunk (if a change of
>>> behavior in TImode is not tested in the testsuite then you can ignore
>>> that).
>>> Whatever change you do to semantics of functions you do separately
>>> from wide-int (preferably first on trunk, or at your choice after the
>>> wide-int
>>> merge).
>>>
>>> For this case in question I'd say a % -1 should return 0, but for
>>> INT_MIN % -1 that 0 should have TREE_OVERFLOW set (and
>>> thus you need to adjust that c90-const-expr-8.c testcase).
>>>
>>> Richard.
>>>
>>>> kenny
>>
>> richi,
>> I have done this exactly as you suggested. bootstrapped and regression
>> tested on x86-64.
>>
>> 2013-11-20 Kenneth Zadeck <zadeck@naturalbridge.com>
>>
>> * fold-const.c
>> (int_const_binop_1): Make INT_MIN % -1 return 0 with the overflow
>> bit set.
>>
>>
>> 2013-11-20 Kenneth Zadeck <zadeck@naturalbridge.com>
>>
>> * gcc.dg/c90-const-expr-8.c: Look for overflow on INT_MIN % -1.
>> * gcc.dg/c99-const-expr-8.c: Look for overflow on INT_MIN % -1.
>>
>> ok to commit?
>>
>> kenny
>>
>