This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [wide-int] resolve bootstrap issue
- From: Mike Stump <mikestump at comcast dot net>
- To: Richard Sandiford <rsandifo at linux dot vnet dot ibm dot com>
- Cc: "gcc-patches\ at gcc dot gnu dot org Patches" <gcc-patches at gcc dot gnu dot org>
- Date: Wed, 15 Jan 2014 13:51:39 -0800
- Subject: Re: [wide-int] resolve bootstrap issue
- Authentication-results: sourceware.org; auth=none
- References: <925647AB-F4A2-484B-8CDA-010C0DE847BF at comcast dot net> <874n56k2le dot fsf at sandifor-thinkpad dot stglab dot manchester dot uk dot ibm dot com>
On Jan 14, 2014, at 7:25 AM, Richard Sandiford <rsandifo@linux.vnet.ibm.com> wrote:
> Mike Stump <mikestump@comcast.net> writes:
>> diff --git a/gcc/expmed.c b/gcc/expmed.c
>> index ce063eb..720d8c1 100644
>> --- a/gcc/expmed.c
>> +++ b/gcc/expmed.c
>> @@ -4963,6 +4963,7 @@ make_tree (tree type, rtx x)
>> return t;
>>
>> case CONST_DOUBLE:
>> + gcc_assert (HOST_BITS_PER_WIDE_INT * 2 <= MAX_BITSIZE_MODE_ANY_INT);
>> if (TARGET_SUPPORTS_WIDE_INT == 0 && GET_MODE (x) == VOIDmode)
>> t = wide_int_to_tree (type,
>> wide_int::from_array (&CONST_DOUBLE_LOW (x), 2,
>
> I think this would be better as a STATIC_ASSERT.
The cost should be the same in generated code. The C folks can read and understand the gcc_assert easier, and the static property of the check we don't make any use of. I could change it, but there are 0 of these in *.c presently, and I think we want to tread lightly on those folks that know and like C, and don't want gcc to be `different' for no good reason. That said, if people think it would be good to use it, I'd be happy to change it.
>> @@ -1440,10 +1442,10 @@ real_to_integer (const REAL_VALUE_TYPE *r, bool *fail, int precision)
>> }
>> #endif
>> w = SIGSZ * HOST_BITS_PER_LONG + words * HOST_BITS_PER_WIDE_INT;
>> - result = wide_int::from_array
>> + tmp = real_int::from_array
>> (val, (w + HOST_BITS_PER_WIDE_INT - 1) / HOST_BITS_PER_WIDE_INT, w);
>> - result = wi::lrshift (result, (words * HOST_BITS_PER_WIDE_INT) - exp);
>> - result = wide_int::from (result, precision, UNSIGNED);
>> + tmp = wi::lrshift<real_int> (tmp, (words * HOST_BITS_PER_WIDE_INT) - exp);
>> + result = wide_int::from (tmp, precision, UNSIGNED);
>
> Why did you need the <real_int>? It was supposed to work without.
The code in question needs something that is max int + max significand real in size, we made the max int smaller (smaller than this quantity on x86) so, this code needs a special wide int that is bigger. The type is free as vrp uses the same type. As for why Kenny choose this method, I'd defer to him.
>> diff --git a/gcc/simplify-rtx.c b/gcc/simplify-rtx.c
>> index 00b5439..7c21afa 100644
>> --- a/gcc/simplify-rtx.c
>> +++ b/gcc/simplify-rtx.c
>> @@ -5384,6 +5384,7 @@ simplify_immed_subreg (enum machine_mode outermode, rtx op,
>> tmp[u] = buf;
>> base += HOST_BITS_PER_WIDE_INT;
>> }
>> + gcc_assert (GET_MODE_PRECISION (outer_submode) <= MAX_BITSIZE_MODE_ANY_INT);
>> r = wide_int::from_array (tmp, units,
>> GET_MODE_PRECISION (outer_submode));
>> elems[elem] = immed_wide_int_const (r, outer_submode);
>
> Long line.
Fixed:
Index: simplify-rtx.c
===================================================================
--- simplify-rtx.c (revision 206601)
+++ simplify-rtx.c (working copy)
@@ -5384,7 +5384,8 @@ simplify_immed_subreg (enum machine_mode
tmp[u] = buf;
base += HOST_BITS_PER_WIDE_INT;
}
- gcc_assert (GET_MODE_PRECISION (outer_submode) <= MAX_BITSIZE_MODE_ANY_INT);
+ gcc_assert (GET_MODE_PRECISION (outer_submode)
+ <= MAX_BITSIZE_MODE_ANY_INT);
r = wide_int::from_array (tmp, units,
GET_MODE_PRECISION (outer_submode));
elems[elem] = immed_wide_int_const (r, outer_submode);
Thanks for the review.