* Is UBSan supposed to produce a finding for _mm_load_sd and _mm_store_sd
@ 2017-12-08 17:16 Jeffrey Walton
2017-12-08 17:29 ` Marc Glisse
0 siblings, 1 reply; 3+ messages in thread
From: Jeffrey Walton @ 2017-12-08 17:16 UTC (permalink / raw)
To: gcc-help
I have some code that loads a 64-bit integer into a XMM register. It
loads the integer from a byte array:
byte v[8] = ...
__m128i t = _mm_castpd_si128(
_mm_load_sd((const double *)(v)));
It is producing a finding for an unaligned load. I get similar
findings for _mm_load_sd, _mm_store_sd and _mm_loaddup_pd.
According to the Intel Intrinsics Guide (e.g., _mm_load_sd):
Load a double-precision (64-bit) floating-point element from memory
into the lower of dst, and zero the upper element. mem_addr does
not need to be aligned on any particular boundary.
Should GCC be producing a finding in this case? Is there a way to work
around it without an extra memcpy?
A similar Clang issue is at http://bugs.llvm.org/show_bug.cgi?id=20670 .
Thanks in advance,
Jeff
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: Is UBSan supposed to produce a finding for _mm_load_sd and _mm_store_sd
2017-12-08 17:16 Is UBSan supposed to produce a finding for _mm_load_sd and _mm_store_sd Jeffrey Walton
@ 2017-12-08 17:29 ` Marc Glisse
2018-02-22 14:08 ` Jeffrey Walton
0 siblings, 1 reply; 3+ messages in thread
From: Marc Glisse @ 2017-12-08 17:29 UTC (permalink / raw)
To: Jeffrey Walton; +Cc: gcc-help
On Fri, 8 Dec 2017, Jeffrey Walton wrote:
> I have some code that loads a 64-bit integer into a XMM register. It
> loads the integer from a byte array:
>
> byte v[8] = ...
> __m128i t = _mm_castpd_si128(
> _mm_load_sd((const double *)(v)));
>
> It is producing a finding for an unaligned load. I get similar
> findings for _mm_load_sd, _mm_store_sd and _mm_loaddup_pd.
>
> According to the Intel Intrinsics Guide (e.g., _mm_load_sd):
>
> Load a double-precision (64-bit) floating-point element from memory
> into the lower of dst, and zero the upper element. mem_addr does
> not need to be aligned on any particular boundary.
>
> Should GCC be producing a finding in this case? Is there a way to work
> around it without an extra memcpy?
The way _mm_load_sd is currently implemented in gcc, yes, sanitizers are
right to complain. Intel could have named the thing _mm_loadu_sd if that's
what they meant. It would be simple to change if we decide to do so,
please file a PR in bugzilla.
Workaround: define a typedef for double with
__attribute__((__aligned__(1))), and use _mm_set_sd(*(newtype*)p), that's
how it will likely be done if we change emmintrin.h.
--
Marc Glisse
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: Is UBSan supposed to produce a finding for _mm_load_sd and _mm_store_sd
2017-12-08 17:29 ` Marc Glisse
@ 2018-02-22 14:08 ` Jeffrey Walton
0 siblings, 0 replies; 3+ messages in thread
From: Jeffrey Walton @ 2018-02-22 14:08 UTC (permalink / raw)
To: gcc-help
On Fri, Dec 8, 2017 at 12:28 PM, Marc Glisse <marc.glisse@inria.fr> wrote:
> On Fri, 8 Dec 2017, Jeffrey Walton wrote:
> ...
>> According to the Intel Intrinsics Guide (e.g., _mm_load_sd):
>>
>> Load a double-precision (64-bit) floating-point element from memory
>> into the lower of dst, and zero the upper element. mem_addr does
>> not need to be aligned on any particular boundary.
>>
>> Should GCC be producing a finding in this case? Is there a way to work
>> around it without an extra memcpy?
>
> The way _mm_load_sd is currently implemented in gcc, yes, sanitizers are
> right to complain. Intel could have named the thing _mm_loadu_sd if that's
> what they meant. It would be simple to change if we decide to do so, please
> file a PR in bugzilla.
>
> Workaround: define a typedef for double with
> __attribute__((__aligned__(1))), and use _mm_set_sd(*(newtype*)p), that's
> how it will likely be done if we change emmintrin.h.
Thanks again Marc.
I opened an issue at https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84508 .
Jeff
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2018-02-22 2:28 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2017-12-08 17:16 Is UBSan supposed to produce a finding for _mm_load_sd and _mm_store_sd Jeffrey Walton
2017-12-08 17:29 ` Marc Glisse
2018-02-22 14:08 ` Jeffrey Walton
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).