similar to: [LLVMdev] LLVMC tests failing when building with clang

Hi, Argyrios Kyrtzidis <kyrtzidis <at> apple.com> writes: > > Hi, > > After building llvm with clang the llvmc tests are failing with: > > llvmc: Node llc is not in graph > > Anyone else see this ? (TOT llvm & clang) It looks like this is due to a bug in clang: // file1.cpp namespace { class Plugin { }; } // file2.cpp namespace { class Plugin {

On Aug 12, 2010, at 11:04 PM, Mikhail Glushenkov wrote: > Argyrios Kyrtzidis <kyrtzidis <at> apple.com> writes: > >> >> Hi, >> >> After building llvm with clang the llvmc tests are failing with: >> >> llvmc: Node llc is not in graph >> >> Anyone else see this ? (TOT llvm & clang) > > It looks like this is due to a

The attached .ll files seem equivalent, but the resulting asm from 'opt-fail.ll' causes a crash to webkit. I suspect the usage of registers is wrong, can someone take a look ? $ llc opt-pass.ll -o - .section __TEXT,__text,regular,pure_instructions .globl __ZN7WebCore6kolos1ERiS0_PKNS_20RenderBoxModelObjectEPNS_10StyleImageE .align 4, 0x90

I attached preprocessed files. $ llvm-g++ gcc-RenderBoxModelObject.ii -fno-exceptions -arch x86_64 -O2 -c -o part.o vs $ clang++ clang-RenderBoxModelObject.ii -fno-exceptions -arch x86_64 -O2 -c -o part.o If I compile with clang, it causes a crash to webkit. -Argiris -------------- next part -------------- A non-text attachment was scrubbed... Name: prepro.zip Type: application/zip Size:

If I use the jump threading pass on the attached IR: $ opt before.ll -jump-threading -o - | llvm-dis -o after.ll a big chunk gets removed, a chunk that is actually necessary. ('before.ll' passes the test in webkit, while 'after.ll' fails) Can someone take a look ? -Argiris -------------- next part -------------- A non-text attachment was scrubbed... Name: before.ll Type:

On Sep 1, 2010, at 6:25 AM, Argyrios Kyrtzidis wrote: > The attached .ll files seem equivalent, but the resulting asm from 'opt-fail.ll' causes a crash to webkit. > I suspect the usage of registers is wrong, can someone take a look ? The difference is that there is a shift right after the multiply, before the divide. In IR, the difference is: %5 = mul nsw i32 %4, %tmp1

Here's the optimized versions: $ opt -std-compile-opts unopt-pass.ll -o - | llvm-dis -o - [...] define %3 @_ZN7WebCore15GraphicsContext19roundToDevicePixelsERKNS_9FloatRectE(%"class.WebCore::GraphicsContext"* %this, %"struct.WebCore::FloatRect"* %rect) nounwind ssp align 2 { %roundedOrigin = alloca %"class.WebCore::FloatSize", align 4 ;

On Sep 2, 2010, at 8:05 AMPDT, Argyrios Kyrtzidis wrote: > If I use the jump threading pass on the attached IR: > > $ opt before.ll -jump-threading -o - | llvm-dis -o after.ll > > a big chunk gets removed, a chunk that is actually necessary. > ('before.ll' passes the test in webkit, while 'after.ll' fails) > > Can someone take a look ? This is fixed in

On 08/05/2010 07:54 PM, Daniel Dunbar wrote: > Hi Argyrios, > > This has been happening for a long time, no one has debugged it yet. > > I personally always configure llvm with --without-llvmgcc and > --without-llvmgxx, which disables these tests. I believe the reason is that llvmc is not built at all with the cmake system. Here is the relevant bug report.

Hi, I've attached 2 .ll files which are supposed to be equivalent but 'unopt-fail.ll' causes a crash in webkit's test suite while 'unopt-pass.ll' does not. I can't give more details about the crash, when I run the crashing test it in isolation it passes, when I run the full suite it crashes; it boggles the mind. Below I provide the optimized asm that is produced from

On Aug 31, 2010, at 1:21 PMPDT, Argyrios Kyrtzidis wrote: > > Just to be clear, are you saying that the fact that, after using llc > on the second IR, the produced asm is using MM registers, indicates > a bug ? Yes. It's not immediately obvious whether it's in the opt or llc, though. Chris was doing work involving <2 x float> and may know about this. >

On Aug 31, 2010, at 11:18 AM, Argyrios Kyrtzidis wrote: > Hi, > > I've attached 2 .ll files which are supposed to be equivalent but 'unopt-fail.ll' causes a crash in webkit's test suite while 'unopt-pass.ll' does not. I can't give more details about the crash, when I run the crashing test it in isolation it passes, when I run the full suite it crashes; it

On Sep 1, 2010, at 9:42 AM, Argyrios Kyrtzidis wrote: > If I compile with clang, it causes a crash to webkit. This should be fixed by r112751, please verify. /jakob

On Sep 1, 2010, at 6:25 AMPDT, Argyrios Kyrtzidis wrote: > The attached .ll files seem equivalent, but the resulting asm from > 'opt-fail.ll' causes a crash to webkit. > I suspect the usage of registers is wrong, can someone take a look ? Yes, the code here is wrong: > movl (%rbx), %ecx > imull %ecx, %eax This computes h*((int32)%1) in %eax. > shrq $32, %rax

On Sep 1, 2010, at 11:14 AM, Dale Johannesen wrote: > > On Sep 1, 2010, at 6:25 AMPDT, Argyrios Kyrtzidis wrote: > >> The attached .ll files seem equivalent, but the resulting asm from >> 'opt-fail.ll' causes a crash to webkit. >> I suspect the usage of registers is wrong, can someone take a look ? > > Yes, the code here is wrong: > >> movl

Using MM registers is wrong unless the user has specifically asked for it, which doesn't seem to be the case here. In the awesome MMX architecture, touching an MM register makes subsequent x87 operations fail unless an EMMS instruction is issued first; none of the compilers here are smart enough to insert EMMS instructions in the right places, so the only safe thing is not to use

I am getting this error while using the command "llvmc -O3 -Wall hello.cpp" where hello.cpp is some cpp file on my machine. llvmc is the driver calling a bunch of tools like llvm-g++, opt, llc etc. I think the llvmc tool could not probably locate the llvm-g++ tool. I would like to know where should I specify the path to llvm-g++ for llvmc to find them. Also can someone give me the link

Hi, Is it intended that llvmc should search PATH for llc? For example, this means that llvmc 2.7 can invoke llc 2.8: % echo $PATH /home/jdenny/installs/llvm/2.8/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games % /home/jdenny/installs/llvm/2.7/bin/llvmc test.ll llc: Unknown command line argument '-f'. Try: 'llc -help' % strace -f

Hi, Apparently there are two bugs in tools/llvmc/c: /.../llvm/tools/llvmc/c:12: Error: Invalid top level configuration item /.../llvm/tools/llvmc/c:28: Error: Expecting a program name /.../llvm/tools/llvmc/c had 2 errors. Terminating. The first error relates to the following line(s): lang.libs=%llvmgccdir%/lib %llvmgccdir%/lib \ %llvmgccdir%/lib/gcc/%llvmgccarch% ... the other one to:

According to the FAQ llvmc is considered experimental/unsupported. But FWIW, here's an issue I found while trying to use it on 32 bit x86 systems. tools/llvmc/plugins/Base/Base.td hardcodes the -relocation-model=pic option into invocations of llc: def llc : Tool< [(in_language "llvm-bitcode"), (out_language "assembler"), (output_suffix "s"), (cmd_line