Monte Carlo Simulation with RDRAND

[Gunther]

Attached is the file mc.zip. It contains the following files:

• b_mc.bat:            batch file to build the application
• mc.asm:              assembly language source
• mc.obj:               object file
• mc.pdf:               algorithmic background
• montecarlo.c:      C source
• montecarlo.exe:   application
• montecarlo.o:      object file

A typical output of the program is:

Code Select Expand


Generating 200 Million random numbers with RDRAND.
That'll take a little while ...

Area           = 0.250001985000000
Absolute Error = 0.000001985000000
Elapsed Time   = 15.71 Seconds

Generating 200 Million random numbers with C.
That'll take a little while ...

Area           = 0.250020340000000
Absolute Error = 0.000020340000000
Elapsed Time   = 5.02 Seconds

A special thanks goes to Dave aka dedndave. :t He did the proof reading for the algorithmic background and he made an excellent job.

Gunther

[dedndave]

turned out great, Gunther   :t

something i didn't catch in the edit, though   :(
in the table of contents, it says "Literature"
then, that section is named "References" (i like this one better)   :P
sorry about that

[Gunther]

Dave,

something i didn't catch in the edit, though   :(
in the table of contents, it says "Literature"
then, that section is named "References" (i like this one better)   :P
sorry about that

That's a bagatelle; no big deal. You were a great help.

Gunther

[jj2007]

Hi Gunther,
Can't test it with my 32-bit OS but it looks good. Just one question: Is RDRAND really slower than the C implementation??

[GoneFishing]

Hi Gunther !
Thank you for your work  :t
Can't test it on my machine ... but it seems like RDRAND is doing a lot of work behind the scene 

BTW  What's the CPU / OS you  ran your test on ?

[Gunther]

Jochen,

Hi Gunther,
Can't test it with my 32-bit OS but it looks good. Just one question: Is RDRAND really slower than the C implementation??

yes, it tends to be slow.

Gunther

[sinsi]

Any reason for doing this in 64-bit?

[Gunther]

Any reason for doing this in 64-bit?

Yes. The generated random number is 64 bit wide.

BTW  What's the CPU / OS you  ran your test on ?

Windows 7 (64 bit), i7, Ivy Bridge.

Gunther

[sinsi]

But you can use ax/eax/rax, would that change the time taken? No need for 64-bit Windows if it's 32-bit code.

[dedndave]

RDRAND retrieves a hardware generated random value from the DRNG and
stores it in the destination register given as an argument to the instruction. The
size of the random value (16-, 32-, or 64-bits) is determined by the size of the
register given. The carry flag (CF) must be checked to determine whether a
random value was available at the time of instruction execution.
Note that RDRAND is available to any system or application software running on
the platform. That is, there are no hardware ring requirements that restrict
access based on process privilege level. As such, RDRAND may be invoked as
part of an operating system or hypervisor system library, a shared software
library, or directly by an application.
To determine programmatically whether a given Intel platform supports RDRAND,
the user can use the CPUID instruction to examine bit 30 of the ECX register.

i found a document that said it was 128-bits wide - lol
guess they were wrong - or they were testing the new 128-bit processor   :P

[TWell]

Is that function CheckRdRand correct ?
In amd64 Win7 64-bit it crash when trying to use RdRnd.
In PellesC this function works:

Code Select Expand

int CheckRdRandC(void) {
int r[4]; // EAX, EBX, ECX and EDX
_cpuid(r, 1);
if (r[2] & 0x40000000) return 1;
return 0;
}EDIT Thank's qWord !

[dedndave]

Code Select Expand

        push       rbx                   ; ebx is affected by CPUID
        mov        eax, 1                ; get version information
        cpuid

        ; check RDRAND support

        bt         ecx, 30               ; RDRAND supported?
        jnc        .failed               ; no: jump

.done:

        mov        rax, 1                ; indicate success
        pop        rbx
        ret

.failed:

        sub        rax, rax              ; indicate failure
        jmp        short .done

little oops, there

Code Select Expand

        mov        rax, 1                ; indicate success

.done:
        pop        rbx
        ret

even better

Code Select Expand

mov  rax,1
push rbx
cpuid
xor  rax,rax
bt   ecx,30
pop  rbx
rcl  rax,1
ret

[qWord]

Is that function CheckRdRand correct ?

   if ((r[2] & 0x40000000) == 0x40000000)

[Gunther]

Hi TWell,

Is that function CheckRdRand correct ?
In amd64 Win7 64-bit it crash when trying to use RdRnd.

I think so.

Gunther

[dedndave]

it has a slight bug, Gunther
see reply #11

the result is always set to 1