Error profiling with embedded assembly language in C++ code

Question

I found this article on the efficiency of std::vector::push_back, the associated code can be found here. I tried it myself and I got an illegal instruction (core dumped), gdb indicates the error occurs on line 37.

I compiled using gcc 4.7.2, on a computer with:

$ sudo dmidecode -t processor
# dmidecode 2.11
SMBIOS 2.5 present.

Handle 0x0400, DMI type 4, 40 bytes
Processor Information
    Socket Designation: CPU
    Type: Central Processor
    Family: Core 2 Duo
    Manufacturer: Intel
    ID: 7A 06 01 00 FF FB EB BF
    Signature: Type 0, Family 6, Model 23, Stepping 10
    Flags:
        FPU (Floating-point unit on-chip)
        VME (Virtual mode extension)
        DE (Debugging extension)
        PSE (Page size extension)
        TSC (Time stamp counter)
        MSR (Model specific registers)
        PAE (Physical address extension)
        MCE (Machine check exception)
        CX8 (CMPXCHG8 instruction supported)
        APIC (On-chip APIC hardware supported)
        SEP (Fast system call)
        MTRR (Memory type range registers)
        PGE (Page global enable)
        MCA (Machine check architecture)
        CMOV (Conditional move instruction supported)
        PAT (Page attribute table)
        PSE-36 (36-bit page size extension)
        CLFSH (CLFLUSH instruction supported)
        DS (Debug store)
        ACPI (ACPI supported)
        MMX (MMX technology supported)
        FXSR (FXSAVE and FXSTOR instructions supported)
        SSE (Streaming SIMD extensions)
        SSE2 (Streaming SIMD extensions 2)
        SS (Self-snoop)
        HTT (Multi-threading)
        TM (Thermal monitor supported)
        PBE (Pending break enabled)
    Version: Not Specified
    Voltage: 1.2 V
    External Clock: 1333 MHz
    Max Speed: 5200 MHz
    Current Speed: 3000 MHz
    Status: Populated, Enabled
    Upgrade: Socket LGA775
    L1 Cache Handle: 0x0700
    L2 Cache Handle: 0x0701
    L3 Cache Handle: Not Provided
    Serial Number: Not Specified
    Asset Tag: Not Specified
    Part Number: Not Specified
    Core Count: 2
    Core Enabled: 2
    Thread Count: 2
    Characteristics:
        64-bit capable

What is the problem here? How can I get this code to work? I also tried with icpc 13.1.0 but this also failed.

Edit: I'm using Ubuntu 12.10 64-bit.

Works fine for me with gcc 4.2 on a Core i7 - what OS are you using ? — Paul R
– Paul R, Commented Apr 12, 2013 at 9:30
The error on line 37 doesn't mean the error is precisely there. You need to look at the disassembly and see which instruction exactly is causing it. — Alexey Frunze
– Alexey Frunze, Commented Apr 12, 2013 at 9:32
@AlexeyFrunze thanks, so how do I do that? Also is it odd that the error is not reported in the startRDTSC function since that contains all the asm calls as stopRDTSC but is called first and seems to work. My entire stack trace is the following: 0 in stopRDTSCP of testvector.cpp:37 1 in CPUBenchmark::stop of testvector.cpp:55 2 in main of testvector.cpp:177 — okmatija
– okmatija, Commented Apr 12, 2013 at 9:41
Actually stopRDTSC has the line "RDTSCP\n\t" which is not in startRDTSC, perhaps that is the problem? It seems to works when I changed that to "RDTSC\n\t" as in startRDTSC. — okmatija
– okmatija, Commented Apr 12, 2013 at 9:42

Community · Accepted Answer · 2017-05-23 12:18:54Z

1

Your CPU doesn't support the RDTSCP instruction. It's a Core i7 instruction, and your processor is an earlier generation (Merom-L).

You should be able to use RDTSC instead. See, for example, Difference between rdtscp, rdtsc : memory and cpuid / rdtsc?

edited May 23, 2017 at 12:18

CommunityBot

11 silver badge

answered Apr 12, 2013 at 9:40

NPE

503k114 gold badges970 silver badges1k bronze badges

Sign up to request clarification or add additional context in comments.

5 Comments

Alexey Frunze Over a year ago

Or the use of this instruction is disabled via CR4.TSD=0.

NPE Over a year ago

@AlexeyFrunze: It's a Merom-L CPU, and predates Core i7.

okmatija Over a year ago

@NPE Oh, I just tried this too and it seems to work. Thanks for the link, if I understand correctly it seems like changing to RDTSC can have a performance penalty, does that then spoil the usefulness of this code as a profiling tool?

okmatija Over a year ago

Perhaps someone could try it modified and unmodified to see if it makes a difference for a compatible CPU, if not then perhaps it would be okay for my CPU also?

NPE Over a year ago

@mkm: As long as you measure a non-trivial amount of work, you'll be fine.

Collectives™ on Stack Overflow

Error profiling with embedded assembly language in C++ code

1 Answer 1

5 Comments

Your Answer

Linked

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

5 Comments

Your Answer

Sign up or log in

Post as a guest

Linked

Related