'[cfe-dev] Size suffix ignored in atomic builtins?'

[prev in list] [next in list] [prev in thread] [next in thread] 

List:       cfe-dev
Subject:    [cfe-dev] Size suffix ignored in atomic builtins?
From:       Mikhail Zolotukhin via cfe-dev <cfe-dev () lists ! llvm ! org>
Date:       2015-08-24 19:52:48
Message-ID: 0129EF0C-8A29-4F23-8697-FA31C1DD8D4B () apple ! com
[Download RAW message or body]

[Attachment #2 (multipart/alternative)]

Hi,

Currently atomic intrinsics are overloaded by the datatype. That allow us to use a \
non-suffixed version for any data-type, like this:  signed char sc;
  void foo(void) {
    (void) __sync_fetch_and_add(&sc, 1);
  }
instead of explicitly specifying data-type, as here:
  signed char sc;
  void foo(void) {
    (void) __sync_fetch_and_add_1(&sc, 1);
  }

For both of these examples clang generates
  %0 = atomicrmw add i8* @sc, i8 1 seq_cst

However, it looks like due to this overloading, we completely ignore the size suffix. \
For instance, for   signed char sc;
  void foo(void) {
    (void) __sync_fetch_and_add_4(&sc, 1);
  }

clang will generate the same IR with i8:
  %0 = atomicrmw add i8* @sc, i8 1 seq_cst ; << i8, instead of i32

Is it an expected behavior, or a bug?

Thanks,
Michael

[Attachment #5 (unknown)]

<html><head><meta http-equiv="Content-Type" content="text/html \
charset=us-ascii"></head><body style="word-wrap: break-word; -webkit-nbsp-mode: \
space; -webkit-line-break: after-white-space;" class="">Hi,<div class=""><br \
class=""></div><div class="">Currently atomic intrinsics are overloaded by the \
datatype. That allow us to use a non-suffixed version for any data-type, like \
this:</div><div class=""><div class=""><font face="Menlo" class="">&nbsp; signed char \
sc;</font></div><div class=""><font face="Menlo" class="">&nbsp; void foo(void) \
{</font></div><div class=""><font face="Menlo" class="">&nbsp; &nbsp; (void) \
__sync_fetch_and_add(&amp;sc, 1);</font></div><div class=""><font face="Menlo" \
class="">&nbsp; }</font></div></div><div class="">instead of explicitly specifying \
data-type, as here:</div><div class=""><div class=""><div class=""><font face="Menlo" \
class="">&nbsp; signed char sc;</font></div><div class=""><font face="Menlo" \
class="">&nbsp; void foo(void) {</font></div><div class=""><font face="Menlo" \
class="">&nbsp; &nbsp; (void) __sync_fetch_and_add_1(&amp;sc, 1);</font></div><div \
class=""><font face="Menlo" class="">&nbsp; }</font></div></div></div><div \
class=""><font face="Menlo" class=""><br class=""></font></div><div class="">For both \
of these examples clang generates</div><div class=""><font face="Menlo" \
class="">&nbsp; %0 = atomicrmw add i8* @sc, i8 1 seq_cst</font></div><div \
class=""><br class=""></div><div class="">However, it looks like due to this \
overloading, we completely ignore the size suffix. For instance, for&nbsp;</div><div \
class=""><div class=""><font face="Menlo" class="">&nbsp; signed char \
sc;</font></div><div class=""><font face="Menlo" class="">&nbsp; void foo(void) \
{</font></div><div class=""><font face="Menlo" class="">&nbsp; &nbsp; (void) \
__sync_fetch_and_add_4(&amp;sc, 1);</font></div><div class=""><font face="Menlo" \
class="">&nbsp; }</font></div></div><div class=""><br class=""></div><div \
class="">clang will generate the same IR with i8:</div><div class=""><div \
class=""><font face="Menlo" class="">&nbsp; %0 = atomicrmw add i8* @sc, i8 1 seq_cst \
; &lt;&lt; i8, instead of i32</font></div></div><div class=""><br class=""></div><div \
class=""><br class=""></div><div class="">Is it an expected behavior, or a \
bug?</div><div class=""><br class=""></div><div class="">Thanks,</div><div \
class="">Michael</div></body></html>

[Attachment #6 (text/plain)]

_______________________________________________
cfe-dev mailing list
cfe-dev@lists.llvm.org
http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-dev

[prev in list] [next in list] [prev in thread] [next in thread] 