[llvm] r260338 - [Orc] Add lazy-JITting support for i386.

Dimitry Andric via llvm-commits llvm-commits at lists.llvm.org
Mon Feb 22 05:01:49 PST 2016


Hi Lang,

Thanks for fixing it.  On i386-freebsd, the OrcLazy tests now all succeed:

-- Testing: 4 tests, 4 threads --
PASS: LLVM :: ExecutionEngine/OrcLazy/anonymous_globals.ll (1 of 4)
PASS: LLVM :: ExecutionEngine/OrcLazy/hello.ll (2 of 4)
PASS: LLVM :: ExecutionEngine/OrcLazy/private_linkage.ll (3 of 4)
PASS: LLVM :: ExecutionEngine/OrcLazy/global_aliases.ll (4 of 4)
Testing Time: 0.10s
  Expected Passes    : 4

-Dimitry

> On 21 Feb 2016, at 23:57, Lang Hames <lhames at gmail.com> wrote:
> 
> Hi Dimitry, Reid,
> 
> I've added stack realignment code to the i386 resolver function in r261503. This should fix the failures due to the fxsave/fxrstor instructions. Hopefully this will fix the BSD bots.
> 
> Reid - this doesn't address the relocation issue you encountered, so we can't turn these tests back on on Windows yet, but it's a step in the right direction.
> 
> Cheers,
> Lang.
> 
> On Sun, Feb 21, 2016 at 12:20 PM, Lang Hames <lhames at gmail.com <mailto:lhames at gmail.com>> wrote:
> Hi Dimitry,
> 
> This looks similar to the issue that Reid ran in to. I suspect it's a stack alignment issue: This code assumes the stack is 16-byte aligned, and I guess that's not true on Windows or BSD.
> 
> I'll add some stack realignment code now and see if it gets the builders green again.
> 
> Cheers,
> Lang.
> 
> 
> On Sun, Feb 21, 2016 at 5:42 AM, Dimitry Andric <dimitry at andric.com <mailto:dimitry at andric.com>> wrote:
> On 10 Feb 2016, at 02:02, Lang Hames via llvm-commits <llvm-commits at lists.llvm.org <mailto:llvm-commits at lists.llvm.org>> wrote:
> >
> > Author: lhames
> > Date: Tue Feb  9 19:02:33 2016
> > New Revision: 260338
> >
> > URL: http://llvm.org/viewvc/llvm-project?rev=260338&view=rev <http://llvm.org/viewvc/llvm-project?rev=260338&view=rev>
> > Log:
> > [Orc] Add lazy-JITting support for i386.
> >
> > This patch adds a new class, OrcI386, which contains the hooks needed to
> > support lazy-JITing on i386 (currently only for Pentium 2 or above, as the JIT
> > re-entry code uses the FXSAVE/FXRSTOR instructions).
> >
> > Support for i386 is enabled in the LLI lazy JIT and the Orc C API, and
> > regression and unit tests are enabled for this architecture.
> 
> Hi,
> 
> I was wondering why I was always getting OrcLazy test failures on i386-freebsd recently, but it appears to have been introduced with this commit. :)
> 
> All the OrcLazy tests result in lli getting a SIGBUS:
> 
> FAIL: LLVM :: ExecutionEngine/OrcLazy/anonymous_globals.ll (18054 of 25776)
> ******************** TEST 'LLVM :: ExecutionEngine/OrcLazy/anonymous_globals.ll' FAILED ********************
> Script:
> --
> /home/dim/obj/llvm-261275M-trunk-freebsd11-i386-ninja-rel-1/./bin/lli -jit-kind=orc-lazy /share/dim/src/llvm/trunk/test/ExecutionEngine/OrcLazy/anonymous_globals.ll
> --
> Exit Code: 138
> 
> Command Output (stderr):
> --
> /home/dim/obj/llvm-261275M-trunk-freebsd11-i386-ninja-rel-1/test/ExecutionEngine/OrcLazy/Output/anonymous_globals.ll.script: line 1: 11213 Bus error               (core dumped) /home/dim/obj/llvm-261275M-trunk-freebsd11-i386-ninja-rel-1/./bin/lli -jit-kind=orc-lazy /share/dim/src/llvm/trunk/test/ExecutionEngine/OrcLazy/anonymous_globals.ll
> 
> --
> 
> ********************
> Testing: 0 .. 10.. 20.. 30.. 40.. 50.. 60..
> FAIL: LLVM :: ExecutionEngine/OrcLazy/global_aliases.ll (18064 of 25776)
> ******************** TEST 'LLVM :: ExecutionEngine/OrcLazy/global_aliases.ll' FAILED ********************
> Script:
> --
> /home/dim/obj/llvm-261275M-trunk-freebsd11-i386-ninja-rel-1/./bin/lli -jit-kind=orc-lazy /share/dim/src/llvm/trunk/test/ExecutionEngine/OrcLazy/global_aliases.ll
> --
> Exit Code: 138
> 
> Command Output (stderr):
> --
> /home/dim/obj/llvm-261275M-trunk-freebsd11-i386-ninja-rel-1/test/ExecutionEngine/OrcLazy/Output/global_aliases.ll.script: line 1: 11214 Bus error               (core dumped) /home/dim/obj/llvm-261275M-trunk-freebsd11-i386-ninja-rel-1/./bin/lli -jit-kind=orc-lazy /share/dim/src/llvm/trunk/test/ExecutionEngine/OrcLazy/global_aliases.ll
> 
> --
> 
> ********************
> Testing: 0 .. 10.. 20.. 30.. 40.. 50.. 60..
> FAIL: LLVM :: ExecutionEngine/OrcLazy/hello.ll (18077 of 25776)
> ******************** TEST 'LLVM :: ExecutionEngine/OrcLazy/hello.ll' FAILED ********************
> Script:
> --
> /home/dim/obj/llvm-261275M-trunk-freebsd11-i386-ninja-rel-1/./bin/lli -jit-kind=orc-lazy -orc-lazy-debug=funcs-to-stdout /share/dim/src/llvm/trunk/test/ExecutionEngine/OrcLazy/hello.ll | /home/dim/obj/llvm-261275M-trunk-freebsd11-i386-ninja-rel-1/./bin/FileCheck /share/dim/src/llvm/trunk/test/ExecutionEngine/OrcLazy/hello.ll
> --
> Exit Code: 2
> 
> Command Output (stderr):
> --
> FileCheck error: '-' is empty.
> 
> --
> 
> ********************
> Testing: 0 .. 10.. 20.. 30.. 40.. 50.. 60..
> FAIL: LLVM :: ExecutionEngine/OrcLazy/private_linkage.ll (18078 of 25776)
> ******************** TEST 'LLVM :: ExecutionEngine/OrcLazy/private_linkage.ll' FAILED ********************
> Script:
> --
> /home/dim/obj/llvm-261275M-trunk-freebsd11-i386-ninja-rel-1/./bin/lli -jit-kind=orc-lazy /share/dim/src/llvm/trunk/test/ExecutionEngine/OrcLazy/private_linkage.ll
> --
> Exit Code: 138
> 
> Command Output (stderr):
> --
> /home/dim/obj/llvm-261275M-trunk-freebsd11-i386-ninja-rel-1/test/ExecutionEngine/OrcLazy/Output/private_linkage.ll.script: line 1: 11218 Bus error               (core dumped) /home/dim/obj/llvm-261275M-trunk-freebsd11-i386-ninja-rel-1/./bin/lli -jit-kind=orc-lazy /share/dim/src/llvm/trunk/test/ExecutionEngine/OrcLazy/private_linkage.ll
> 
> --
> 
> ********************
> Testing: 0 .. 10.. 20.. 30.. 40.. 50.. 60.. 70.. 80.. 90.
> FAIL: LLVM-Unit :: ExecutionEngine/Orc/OrcJITTests/OrcCAPIExecutionTest.TestLazyIRCompilation (24756 of 25776)
> ******************** TEST 'LLVM-Unit :: ExecutionEngine/Orc/OrcJITTests/OrcCAPIExecutionTest.TestLazyIRCompilation' FAILED ********************
> Note: Google Test filter = OrcCAPIExecutionTest.TestLazyIRCompilation
> [==========] Running 1 test from 1 test case.
> [----------] Global test environment set-up.
> [----------] 1 test from OrcCAPIExecutionTest
> [ RUN      ] OrcCAPIExecutionTest.TestLazyIRCompilation
> 
> ********************
> 
> Unfortunately, the debugger seems to have trouble following the execution, so it does not really show anything interesting in the stack trace:
> 
> (gdb) run
> Starting program: /home/dim/obj/llvm-orc-r260338/bin/lli -jit-kind=orc-lazy /share/dim/src/llvm/trunk/test/ExecutionEngine/OrcLazy/private_linkage.ll
> [New Thread 29612000 (LWP 101099)]
> 
> Program received signal SIGBUS, Bus error.
> [Switching to Thread 29612000 (LWP 101099)]
> 0x28fa600f in ?? ()
> (gdb) bt
> #0  0x28fa600f in ?? ()
> #1  0x2913aff5 in ?? ()
> #2  0x0815adcd in main (argc=<optimized out>, argv=<optimized out>, envp=<optimized out>)
>     at /share/dim/src/llvm/trunk/tools/lli/lli.cpp:414
> 
> It appears to die on an fxsave instruction, though:
> 
> (gdb) disassemble 0x28fa6000,0x28fa6020
> Dump of assembler code from 0x28fa6000 to 0x28fa6020:
>    0x28fa6000:  push   %ebp
>    0x28fa6001:  mov    %esp,%ebp
>    0x28fa6003:  push   %eax
>    0x28fa6004:  push   %ebx
>    0x28fa6005:  push   %ecx
>    0x28fa6006:  push   %edx
>    0x28fa6007:  push   %esi
>    0x28fa6008:  push   %edi
>    0x28fa6009:  sub    $0x21c,%esp
> => 0x28fa600f:  fxsave 0x10(%esp)
>    0x28fa6014:  mov    0x4(%ebp),%esi
>    0x28fa6017:  sub    $0x5,%esi
>    0x28fa601a:  mov    %esi,0x4(%esp)
>    0x28fa601e:  movl   $0x296182c0,(%esp)
> End of assembler dump.
> (gdb) info registers
> eax            0xbfbfe6e0       -1077942560
> ecx            0x2913bff0       689160176
> edx            0x3      3
> ebx            0x905e994        151382420
> esp            0xbfbfd58c       0xbfbfd58c
> ebp            0xbfbfd7c0       0xbfbfd7c0
> esi            0xbfbfd7f0       -1077946384
> edi            0xbfbfd820       -1077946336
> eip            0x28fa600f       0x28fa600f
> eflags         0x210292 [ AF SF IF RF ID ]
> cs             0x33     51
> ss             0x3b     59
> ds             0x3b     59
> es             0x3b     59
> fs             0x3b     59
> gs             0x1b     27
> 
> Any ideas on how to debug this further?
> 
> -Dimitry
> 
> 
> 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20160222/ebab8d4e/attachment.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 194 bytes
Desc: Message signed with OpenPGP using GPGMail
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20160222/ebab8d4e/attachment.sig>


More information about the llvm-commits mailing list