<div dir="ltr"><div class="gmail_extra"><div class="gmail_quote">On Mon, Aug 17, 2015 at 1:59 PM, David Blaikie <span dir="ltr"><<a href="mailto:dblaikie@gmail.com" target="_blank">dblaikie@gmail.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div dir="ltr"><br><div class="gmail_extra"><br><div class="gmail_quote"><span class="">On Mon, Aug 17, 2015 at 11:07 AM, Evgeniy Stepanov via cfe-commits <span dir="ltr"><<a href="mailto:cfe-commits@lists.llvm.org" target="_blank">cfe-commits@lists.llvm.org</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">eugenis created this revision.<br>
eugenis added reviewers: chandlerc, rsmith.<br>
eugenis added a subscriber: cfe-commits.<br>
eugenis set the repository for this revision to rL LLVM.<br>
<br>
Currently always_inline definitions are emitted as (in most cases) an available_externally llvm function with an alwaysinline attribute. </blockquote></span><div><br>Naive dude missing some context here: How can that ^ be correct? <br><br>A basic test case doesn't seem to indicate that that's the case:<br><br>__attribute__((always_inline)) void f1() { }<br>void f2() { f1(); }<br><br>Compile with LLVM optimizations disabled, etc - the bitcode has no mention of available_externally in it... </div></div></div></div></blockquote><div><br></div><div>1) __attribute__((always_inline)) does not imply inline; you need to add 'inline' to f1 or you're not in the "in most cases" case<br></div><div>2) always-inlining is performed even at -O0; you need -mllvm -disable-llvm-optzns to see the always_inline body</div><div>3) The "in most cases" case is for C or for C++ functions with __attribute__((gnu_inline)); in plain C++ you get a linkonce_odr function definition instead.</div><div> </div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div dir="ltr"><div class="gmail_extra"><div class="gmail_quote"><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div><div class="h5">This is not exactly right: always_inline functions are NOT available externally, and, for example, libc++ uses this semantics to preserve ABI stability.<br>
<br>
Emitting an undefined symbol for an always_inline function is always a bug. The current code can still do it in certain cases.<br>
a. Inliner is an SCC pass. It traverses the graph starting from the roots, which are either main() function, or all externally-visible functions. Inlining does not happen in functions that are not reachable.<br>
b. Dead code elimination is not perfect. There are cases where a function will become unreachable due to some late optimizations and will still be emitted into the binary.<br>
<br>
This patch changes the way always_inline functions are emitted in the Clang codegen to ensure this never happens. A function F is emitted as a pair of<br>
a. internal F.inlinefunction() alwaysinline { **original function body** }<br>
and, depending on the function visibility, either<br>
b1. declare F()<br>
or<br>
b2. define external F() { musttail call F.inlinefunction() }<br>
<br>
Frontend ensures that all direct calls go to F.inlinefunction().<br>
<br>
This provides a simple invariant that all alwaysinline functions are internal, which can be checked in the IR verifier. Another invariant would be that alwaysinline functions never reach the backend.<br>
<br>
This patch is based on ideas by Chandler Carruth and Richard Smith.<br>
<br>
<br>
Repository:<br>
rL LLVM<br>
<br>
<a href="http://reviews.llvm.org/D12087" rel="noreferrer" target="_blank">http://reviews.llvm.org/D12087</a><br>
<br>
Files:<br>
lib/CodeGen/CGCXX.cpp<br>
lib/CodeGen/CodeGenModule.cpp<br>
lib/CodeGen/CodeGenModule.h<br>
lib/CodeGen/ItaniumCXXABI.cpp<br>
test/CodeGen/2008-05-19-AlwaysInline.c<br>
test/CodeGen/always-inline.c<br>
test/CodeGen/always_inline.c<br>
test/CodeGen/alwaysinline.c<br>
test/CodeGen/dllimport.c<br>
test/CodeGen/function-attributes.c<br>
test/CodeGen/pr9614.c<br>
test/CodeGenCXX/alwaysinline.cpp<br>
test/CodeGenCXX/dllimport.cpp<br>
test/Frontend/optimization-remark-line-directive.c<br>
test/Frontend/optimization-remark.c<br>
test/Modules/cxx-irgen.cpp<br>
<br>
<br></div></div><span class="">_______________________________________________<br>
cfe-commits mailing list<br>
<a href="mailto:cfe-commits@lists.llvm.org" target="_blank">cfe-commits@lists.llvm.org</a><br>
<a href="http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits" rel="noreferrer" target="_blank">http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits</a><br>
<br></span></blockquote></div><br></div></div>
</blockquote></div><br></div></div>