[Issue 10054] New: x86_64 valgrind reports unrecognised instruction (DMD 2.062)

May 09, 2013
Posted by stewart
Permalink
stewart
Permalink
http://d.puremagic.com/issues/show_bug.cgi?id=10054

           Summary: x86_64 valgrind reports unrecognised instruction (DMD
                    2.062)
           Product: D
           Version: D2
          Platform: x86_64
        OS/Version: Linux
            Status: NEW
          Severity: normal
          Priority: P2
         Component: DMD
        AssignedTo: nobody@puremagic.com
        ReportedBy: estewh@gmail.com


--- Comment #0 from stewart <estewh@gmail.com> 2013-05-10 09:00:42 EST ---
Not sure if this is valgrind, me or DMD so thought I should report it. Program runs, so could be just valgrind.

It causes valgrind grief when casting a mutable double variable to a ubyte variable. Casting a literal floating point value worked, as did an immutable double variable.

------------------------
void main()
{
    double dval = 123.456;
    ubyte ubvalFAIL = cast(ubyte)(dval);    // Causes the problem.
    ubyte ubvalOK = cast(ubyte)(123.456); // This cast works  if above line
commented out.

}
------------------------
Given the second cast worked I tried the following and it worked also:

void main() {
    immutable double dval = 123.456;
    ubyte ubvalOK = cast(ubyte)(dval);    // Happy valgrind.
}

=============================================
$ objdump -d quicktest | grep "41652e" -n8
427-  41650c:    f2 48 0f 10 05 ab 77     rex.W movsd 0x177ab(%rip),%xmm0
 # 42dcc0 <__dso_handle+0x8>
428-  416513:    01 00
429-  416515:    f2 48 0f 11 45 f0        rex.W movsd %xmm0,-0x10(%rbp)
430-  41651b:    dd 45 f0                 fldl   -0x10(%rbp)
431-  41651e:    50                       push   %rax
432-  41651f:    d9 7c 24 02              fnstcw 0x2(%rsp)
433-  416523:    66 c7 44 24 04 bf 0f     movw   $0xfbf,0x4(%rsp)
434-  41652a:    d9 6c 24 04              fldcw  0x4(%rsp)
435:  41652e:    48 df 1c 24              rex.W fistp (%rsp)       <<<< HERE
436-  416532:    48 d9 6c 24 02           rex.W fldcw 0x2(%rsp)
437-  416537:    58                       pop    %rax
438-  416538:    88 45 f8                 mov    %al,-0x8(%rbp)
439-  41653b:    31 c0                    xor    %eax,%eax
440-  41653d:    c9                       leaveq
441-  41653e:    c3                       retq
442-  41653f:    90                       nop
443-  416540:    c8 00 00 00              enterq $0x0,$0x0
============================================================================
VALGRIND OUTPUT (abbrev.)
------------------------
vex amd64->IR: unhandled instruction bytes: 0x48 0xDF 0x1C 0x24 0x48 0xD9 0x6C
0x24
vex amd64->IR:   REX=1 REX.W=1 REX.R=0 REX.X=0 REX.B=0
vex amd64->IR:   VEX=0 VEX.L=0 VEX.nVVVV=0x0 ESC=NONE
vex amd64->IR:   PFX.66=0 PFX.F2=0 PFX.F3=0
==2323== valgrind: Unrecognised instruction at address 0x4176fe.
==2323==    at 0x4176FE: _Dmain (/tmp/quicktest.d:18)
...
...
==2323== Your program just tried to execute an instruction that Valgrind
==2323== did not recognise.  There are two possible reasons for this.
==2323== 1. Your program has a bug and erroneously jumped to a non-code
==2323==    location.  If you are running Memcheck and you just saw a
==2323==    warning about a bad jump, it's probably your program's fault.
==2323== 2. The instruction is legitimate but Valgrind doesn't handle it,
==2323==    i.e. it's Valgrind's fault.  If you think this is the case or
==2323==    you are not sure, please let us know and we'll try to fix it.
==2323== Either way, Valgrind will now raise a SIGILL signal which will
==2323== probably kill your program.
===========================

Thanks.

-- 
Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
Forums
