similar to: [LLVMdev] Bug: GCCLD bugs section of How to Submit a Bug

On Fri, Apr 22, 2005 at 02:32:25AM +0200, Markus F.X.J. Oberhumer wrote: > Ok, after pasting the output from "gccas -debug-pass=Arguments > </dev/null -o - >/dev/null" I get the following: > > bugpoint -verify -lowersetjmp -funcresolve -raiseallocs -simplifycfg > -mem2reg -globalopt -globaldce -ipconstprop -deadargelim -instcombine > -simplifycfg -prune-eh

Jack Howarth wrote: > I've had better luck compiling all of pymol 1.1r2 with > -O4 on darwin9. Everythink links and there appears to be > no regressions in the resulting code. I take it that LTO > in llvm 2.5 is still limited to dead code elimination, > correct? No. libLTO does the equivalent to opt -internalize -ipsccp -globalopt -constmerge -deadargelim -instcombine

Which transformation passes are 'safe', meaning it does not worsens the effectiveness of a later pass or the generated code? I imagine all passes which either removes data or add attributes are included in this list, plus some simplification passes: -adce -argpromotion -constmerge -constprop -deadargelim -dse -functionattrs -globaldce -globalopt -gvn -instcombine -internalize

Hi, Op 19-mei-07, om 00:39 heeft Chris Lattner het volgende geschreven: > Anton is right. You should be able to use -fno-builtins to disable > this. Thanks, that did the trick. Some final remarks (my app works again :-)): * llvm.va_start and similar intrinsics now have an i8* arg instead of an sbyte** * For some reason the Arguments of a Function are now circularly linked,

Which transformation passes are 'safe', meaning it does not worsens the effectiveness of a later pass or the generated code? I imagine all passes which either removes data or add attributes are included in this list, plus some simplification passes: -adce -argpromotion -constmerge -constprop -deadargelim -dse -functionattrs -globaldce -globalopt -gvn -instcombine -internalize

On Mon, Aug 01, 2005 at 05:13:15PM -0500, Nicholas Riley wrote: > The attached patch fixes this. ...but had tabs in it. Misha kindly reminded me off-list that this was bad. Try this one instead. -- Nicholas Riley <njriley at uiuc.edu> | <http://www.uiuc.edu/ph/www/njriley> -------------- next part -------------- Index: tools/gccld/GenerateCode.cpp

gccld passes -shared through if it's generating a shared library, but if you're compiling a program that needs to have its symbols externally accessible, it doesn't pass -export-dynamic through to gcc for the final link. The attached patch fixes this. I've tested with a small test case I sent Chris, and with Python; both seem to work. I also fixed some inaccurate comments in

hi, I'm really new to llvm. I've successfully bootstrapped llvm-14 on my system and am able to successfully compile c code to llvm. the problem is now that gccld is complaining that it can't find the libraries, like "c" or "crtend". [1] all is fine, if I just use intrinsified functions like printf and friends, but I want to use the clock_gettime function and

Hi Jing, I am cc ing to LLVMdev (LLVM developer's mailing list). You get more accurate and faster responses that way. <snip> > Actually LLVM is well documented and the coding style is very friendly. We have rarely had problems since we started one month ago. Unfortunately, here comes something I am not very clear, but it is critical to our evaluation. I appreciate if you could

Here is the latest version of the native linker patch. This differs from the previous patch in one major way... It no longer takes the crude hack of screening for 'llvm' in the -L path. Instead, gccld now actively screens for -L paths that contain bytecode shared libraries or true llvm bytecode archives. Those -L paths that do contain them are omitted from the call to the system

> What I want to do is to insert a call funcation into each basic > block, which is for statistic some information. I recommend that you look at the various pieces of code under llvm/lib/Transforms/Instrumentation, e.g. BlockProfiling.cpp and TraceBasicBlocks.cpp. They do essentially the same thing as you are trying to do. > 1) I implement call function in another c/cpp file and >

Misha Brukman wrote: > Hey, Jakob -- > > On Tue, Mar 01, 2005 at 05:55:07PM +0100, Jakob Praher wrote: > >>I'm really new to llvm. I've successfully bootstrapped llvm-14 on my >>system and am able to successfully compile c code to llvm. >> >>the problem is now that gccld is complaining that it can't find the >>libraries, like "c" or

Hi all, I would like to know which optimization passes are performed at -O2 by opt. So I used following command: llvm-as < /dev/null | opt -O2 -std-compile-opts -disable-output -debug-pass=Arguments I've got following output for LLVM opt 2.9: Pass Arguments: -no-aa -tbaa -basicaa -simplifycfg -domtree -scalarrepl -early-cse Pass Arguments: -targetlibinfo -no-aa -tbaa -basicaa

Chris Lattner wrote: > On Tue, 1 Mar 2005, Jakob Praher wrote: > >> thanks for the pointer. Yes I've done that, but in the new shell >> session I apparently forgot to set the LLVM_LIB_SEARCH_PATH. >> >> now gccld isn't complaining anymore but the interpreter doesn't seem >> to like it still: > > > It looks like the jit doesn't find

Hello, I am trying to simplify the CFG of a given code and eliminate the conditionals, even though I will obtain codes that are not semantically equivalent. For example, given a simple loop: for(i=0; i<N; i++){    a[i] = i;     if (i%2==0)       a[i] += 12; } I would keep only the loop, without the if statement: for(i=0; i<N; i++){    a[i] = i; } I can eliminate such conditionals on

gccld would barf if you tried to specify the full pathname of an archive on the command line to gcc, (e.g., instead of gcc ... -L/full/path/to -lFOO you say gcc ... /full/path/to/libFOO.a) which many GNU Makefiles try to do. This patch fixes that. Ok to commit? -- gaeke at uiuc.edu gccld.cpp: Fix typo in header. Add IsArchive static method. Roll LoadLibraryFromDirectory() into

> Hi LLVMDev, > > The GCC frontend for LLVM inserts a call to a __main() in the > main() function. Whats the purpose of this? And which LLVM > library do I need to link to satisfy the reference? __main is used primarily to call static constructors. The way you satisfy this dependency is to link the bytecode file against crtend.o (aka libcrtend.bc). This is normally accomplished

Hi, I've been thinking about how to keep the line number with the llvm transform/Analysis passes. Basically, I agree with Chris's notes ( http://www.nondot.org/sabre/LLVMNotes/DebugInfoImprovements.txt), and I will follow his way to turn on the line number information when optimization enabled. Here is a detailed proposal: 1. Introduction At the time of this writing, LLVM's

> That's essentially what I did to get it to work but instead of > appending, I just opted for LinkOnce. Sounds ok. > > I don't quite understand why you need an appending global to do this, if you > > can only have one main. I would just set it to the correct initializer in the > > module that has main, and leave it external everywhere else. > > Its

Hi, Thank Brian Gaeke so much. Following TraceBasicBlocks.cpp, I modified the code as below and could insert instruction or function I need into anywhere in Module. But it works well without BB->getInstList().push_back(InstrCall), and if I add the BB->getInstList().push_back() following new CallInst(), I got error information when runing opt. What is the reason for it? And is it necessary