C / C++ / MFC

17-Jul-10 7:46

My questions is, im trying to make a small benchmark and im not sure if i am correctly doing this.
I am using a high-resolution performance counter, ive placed it in bold

This uses MD5, which you can simply get it at http://www.evenbalance.com/downloads/pbmd5.cpp[^]

im not really sure if im getting the correct bytes per second

#include <windows.h>
#include <stdio.h>
#include <stdlib.h>
#include "md5.h"

#define MUL_BYTES 1024
#define MUL_KB (MUL_BYTES*1024)
#define MUL_MB (MUL_KB*1024)

void Benchmark_MD5(void)
{
   unsigned __int64 iPerf;
   unsigned __int64 iBaseTime;
   double dbFreq;

   unsigned char bzBuffer[] = {0x0a, 0x55, 0x23};
   unsigned __int64 iValue;

   MD5_CTX md5;
   int iPasses = 5000;

   QueryPerformanceFrequency((LARGE_INTEGER *)&iPerf);
   dbFreq = 1.0 / (double)iPerf;
   QueryPerformanceCounter((LARGE_INTEGER *)&iBaseTime);

   for (int i = 0; i < iPasses; i++) {
      MD5Init(&md5, 0);
      MD5Update(&md5, bzBuffer, 3);
      MD5Final(&md5);
   }

   //16 is for the length of the MD5 Digest added by the passes we tested
   //then divided by the Seconds it took to run

   QueryPerformanceCounter((LARGE_INTEGER *)&iValue);
   double seconds = ((double)(iValue - iBaseTime) * dbFreq);
   double bps = (double)((double)(16 + iPasses) / seconds);


   printf("DEBUG: [Passes = %d] %5.3f Byte/s\n", iPasses, bps);

   printf("RESULT:\t");
   if (bps < MUL_BYTES)
      printf(" %5.2f Bytes/sec\n", bps);
   else    if (bps < MUL_KB)
      printf(" %5.2f KB/sec\n", bps / 1024);
   else    if (bps < MUL_MB)
      printf(" %5.2f MB/sec\n", bps / 1024 / 1024);
   else
      printf(" %5.2f GB/sec\n", bps / 1024 / 1024 / 1024);
}

void main(void)
{
   Benchmark_MD5();
}

C^hris L^osinger17-Jul-10 8:37

17-Jul-10 8:37

what are your results ?

saiyuk6=717-Jul-10 8:45

17-Jul-10 8:45

This is the results

DEBUG: [Passes = 5000] 311191.953 Byte/s
RESULT:  303.90 KB/sec

DEBUG: [Passes = 5000] 734956.927 Byte/s
RESULT:  717.73 KB/sec

DEBUG: [Passes = 5000] 391962.053 Byte/s
RESULT:  382.78 KB/sec

DEBUG: [Passes = 5000] 274082.350 Byte/s
RESULT:  267.66 KB/sec

DEBUG: [Passes = 5000] 281520.063 Byte/s
RESULT:  274.92 KB/sec

Re: calculating bytes per second [modified]

David Crow19-Jul-10 5:46

David Crow

19-Jul-10 5:46

saiyuk6=7 wrote:
double seconds = ((double)(iValue - iBaseTime) * dbFreq);

To get seconds, shouldn't you be dividing by the timer's frequency?

"One man's wage rise is another man's price increase." - Harold Wilson

"Fireproof doesn't mean the fire will never come. It means when the fire comes that you will be able to withstand it." - Michael Simmons

"Man who follows car will be exhausted." - Confucius

saiyuk6=719-Jul-10 14:15

19-Jul-10 14:15

I've updated the code somewhat now, i have taken your advice by dividing the timers freq

i've changed how the bps is calculated as well
double bps = (double)((ulSize * iPasses) / tm.Seconds());
i've multiplied the Size of the Buffer by the number of passes that runs in the for loop

results are

<br />
DEBUG: 1074556249.047 Byte/s<br />
MEMSET:  0.00/sec  0.98/ms  1.00 GB/sec<br />
RESULT:  6.91/sec 6905.56/ms 72.41 MB/sec<br />

class CTimer {
public:
   CTimer() {
	   m_TimeStart.QuadPart = 0;
	   m_TimeStop.QuadPart = 0;
	    QueryPerformanceFrequency(&m_Frequency);
   }

   void Start(void) {
      QueryPerformanceCounter(&m_TimeStart);
   }

   void Stop(void) {
      QueryPerformanceCounter(&m_TimeStop);
   }

   double Seconds() const {
      return (double)(m_TimeStop.QuadPart - m_TimeStart.QuadPart) / (double)m_Frequency.QuadPart;;
   }

   double Milliseconds() const {
      return Seconds() * 1000.0;
   }

private:
   double m_dbFreq;
   LARGE_INTEGER m_Frequency;
   LARGE_INTEGER m_TimeStart;
   LARGE_INTEGER m_TimeStop;
};

#define MUL_BYTES 1024.0
#define MUL_KB (MUL_BYTES*1024.0)
#define MUL_MB (MUL_KB*1024.0)
#define MUL_GB (MUL_MB*1024.0)

std::string BPStoString(double dbPerSec)
{
   char *pTemp[] = {
      "Bytes", "KB", "MB",
      "GB", "TB", "PB",
      "EB", "ZB", "YB"
   };
   char szTemp[30];

   int i = 0;
   while (dbPerSec >= 0.9 * MUL_BYTES) {
      dbPerSec /= MUL_BYTES;
      i++;
   }

   sprintf(szTemp, "%5.2f %s/sec", dbPerSec, pTemp[i]);
   return std::string(szTemp);
}

void Benchmark_MD5(void)
{
   CTimer tm;

   tm.Start();
   unsigned long ulSize = MUL_KB;
   unsigned char *bzBuffer = new unsigned char[ulSize];
   if (bzBuffer != NULL) {
      memset(bzBuffer, 0, ulSize);
      tm.Stop();
      <code>double bps = (double)((ulSize * 1) / tm.Seconds());</code>
      printf("DEBUG: %5.3f Byte/s\n", bps);
      printf("MEMSET: %5.2f/sec %5.2f/ms %s\n", tm.Seconds(), tm.Milliseconds(),BPStoString(bps).c_str());
   }

   tm.Start();

   MD5_CTX md5;
   int iPasses = 500;
   for (int i = 0; i < iPasses; i++) {
      MD5Init(&md5);
      MD5Update(&md5, bzBuffer, ulSize);
      MD5Final(&md5);
   }
   tm.Stop();
  <code> double bps = (double)((ulSize * iPasses) / tm.Seconds());</code>
   printf("RESULT: %5.2f/sec %5.2f/ms %s\n", tm.Seconds(), tm.Milliseconds(),BPStoString(bps).c_str());

   delete bzBuffer;
}

modified on Monday, July 19, 2010 8:57 PM

Re: calculating bytes per second [modified]

David Crow20-Jul-10 2:22

David Crow

20-Jul-10 2:22

saiyuk6=7 wrote:
unsigned char *bzBuffer = new unsigned char[ulSize];

Why not:

unsigned char bzBuffer[MUL_KB];

saiyuk6=720-Jul-10 5:36

Re: OutputDebugString in x64

20-Jul-10 5:36

well i've tried doing that but i get this error

bm.cpp
bm.cpp(40) : error C2057: expected constant expression
bm.cpp(40) : error C2466: cannot allocate an array of constant size 0
bm.cpp(40) : error C2133: 'bzBuffer' : unknown size

Well i had started to allocate memory for it because later on i was going to start using a hash driver that i was going to make that would provide me with several hash algo's,.... example
there might be errors below, but its just a example of what i had planned to do

unsigned char *bzBuffer;
for (int i = HT_MD5; i < (HT_WHIRLPOOL + 1); i++) {
   hash.Init(i);
   bzBuffer = new unsigned char[hash.DigestSize()];
   if (bzBuffer == NULL) {
      //err code here
      return 0;
   }
   hash.Update(hashTest[i].buffer, hashTest[i].length);
   hash.Final(bzBuffer);
   for (int h = 0; h < hash.DigestSize(); h++)
      printf("%02.2%", bzBuffer[i]);
   delete bzBuffer;
}

modified on Tuesday, July 20, 2010 12:28 PM

OutputDebugString in x64

Pascal Ganaye17-Jul-10 0:51

Pascal Ganaye

17-Jul-10 0:51

int _cdecl main(int argc, char *argv[])
{
    OutputDebugString("Hello World");
}

When I compile and debug with Target = Win32 I get:
Hello World
(plus a lot of 'instlsp.exe': Loaded 'C:\WINDOWS\SysWOW64\....dll', Cannot find or open the PDB file)

When I compile and debug with Target = x64 I get nothing
It runs but the output is filtered somewhat.
I tried this

[HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\Session Manager\Debug Print Filter]
"DEFAULT"=dword:0000000f

And also 0xFFFFFFFF
No success even after reboot.

I feel like I am stuck in chapter 1 of the Kernighan and Ritchie C book.
I am using Visual C++ Express under Windows XP Pro x64

Is this just a bad day?
Going to the zoo now. Perhaps the monkeys will have a clue, I should take a couple of typewriters with me.

Rajesh R Subramanian17-Jul-10 8:06

Rajesh R Subramanian

17-Jul-10 8:06

Pascal Ganaye wrote:
Going to the zoo now. Perhaps the monkeys will have a clue,

Sorry, but I have no clue.

Seriously though, I'm on Windows 7 x64 and it works just right for me. Which platform are you on?

It's time for a new sig. Seriously.

how to know the downloaded data from the net for a PC

Sakhalean16-Jul-10 23:25

Sakhalean

16-Jul-10 23:25

Hi I am developing a application
knowing the size of downloaded data from the net
for a month for a PC.

How to do this?

Re: how to know the downloaded data from the net for a PC

Spawn@Melmac17-Jul-10 4:32

Spawn@Melmac

17-Jul-10 4:32

You need to be more specific. What 'data'? Where is it stored? Is it part of your application?

If you have a more intelligent router connecting you to the Internet then you could use SNMP to pull statistics from the router.

Sorry but while you are so vague it is impossible to advise further.

P.S. this is more a general IT question.

Alan

Re: how to know the downloaded data from the net for a PC

Niklas L17-Jul-10 9:16

Niklas L

17-Jul-10 9:16

Have a look at WinPcap[^] libraries.

home

Compiling 64bit

Fareed Rizkalla16-Jul-10 10:32

Fareed Rizkalla

16-Jul-10 10:32

I tried compiling my application in 64bit, during compilation I got errors. Cause my application uses UNICODE, does 64bit lack UNICODE support???

Re: Compiling 64bit

Emilio Garavaglia16-Jul-10 22:50

Emilio Garavaglia

16-Jul-10 22:50

Fareed Rizkalla wrote:
Cause my application uses UNICODE

The sentence is not clear.
Do you mean "because I'm using Unicode strings" or "because it seems the UNICODE preprocessor symbol is not defined"?

In the second case, just look in the compiler options if the Unicode support is activated ad if UNICODE and _UNICODE are in the defines symbol specifications.

2 bugs found.
> recompile ...
65534 bugs found.
D'Oh! | :doh:

Re: Compiling 64bit

Richard MacCutchan16-Jul-10 23:00

Richard MacCutchan

16-Jul-10 23:00

It would help if you showed the relevant lines of code and the exact error messages produced by the compiler.

It's time for a new signature.

proper conversion of unsigned char [16] -> unsigned long [4] [modified]

saiyuk6=716-Jul-10 8:45

Re: proper conversion of unsigned char [16] -> unsigned long [4]

16-Jul-10 8:45

I am stuck actually, im trying to write a md5 / blowfish function
im not sure if i am correctly converting the MD5[16] buffer into a unsigned long[4]
this doesnt really make too much sense too me

typedef struct _tag2Long {
   unsigned long _1;
   unsigned long _2;
   unsigned long _3;
   unsigned long _4;
} _2LONG;

void SignBuffer(unsigned char *pbzKey, unsigned char *pbzBuffer,
                unsigned int iSize, unsigned long pSignature[4])
{
   MD5_CTX md5;
   BLOWFISH_CTX bf;

   MD5Init(&md5);
   MD5Update(&md5, pbzBuffer, iSize);
   MD5Final(&md5);

   _2LONG *tl = (_2LONG*) md5.digest;

   printf("BLE 1\t\t %x %x %x %x\n", tl->_1, tl->_2, tl->_3, tl->_4);

   Blowfish_Init(&bf, pbzKey, 16);
   Blowfish_Encrypt(&bf, &tl->_1, &tl->_2);
   Blowfish_Encrypt(&bf, &tl->_3, &tl->_4);

   pSignature[0] = tl->_1;
   pSignature[1] = tl->_2;
   pSignature[2] = tl->_3;
   pSignature[3] = tl->_4;

   printf("BLE 2\t\t %x %x %x %x\n", pSignature[0], pSignature[1],
          pSignature[2], pSignature[3]);
}


bool VerifyBuffer(unsigned char *pbzKey, unsigned char *pSignature,
                  unsigned int iSize, unsigned long pBuffer[4])
{
   BLOWFISH_CTX bf;
   MD5_CTX md5;
   bool bResult = false;

   printf("BLD 1\t\t %x %x %x %x\n", pBuffer[0], pBuffer[1],
          pBuffer[2], pBuffer[3]);

   Blowfish_Init(&bf, pbzKey, 16);
   Blowfish_Decrypt(&bf, &pBuffer[0], &pBuffer[1]);
   Blowfish_Decrypt(&bf, &pBuffer[2], &pBuffer[3]);

   printf("BLD 2\t\t %x %x %x %x\n", pBuffer[0], pBuffer[1],
          pBuffer[2], pBuffer[3]);

   MD5Init(&md5);
   MD5Update(&md5, pSignature, iSize);
   MD5Final(&md5);

   _2LONG *tl = (_2LONG*) md5.digest;
//   printf("BLD 3\t\t %x %x %x %x\n", tl->_1, tl->_2, tl->_3, tl->_4);

   if (tl->_1 == pBuffer[0] && tl->_2 == pBuffer[1] &&
       tl->_3 == pBuffer[2] && tl->_4 == pBuffer[4])
      bResult = true;

   return bResult;
}

void main(void)
{
   unsigned char bszBuffer[2] = {'A', 'B'};
   unsigned long ulSig[4];
   SignBuffer((unsigned char*)"key", bszBuffer, 2, ulSig);
   printf("\n");
   if (VerifyBuffer((unsigned char*)"key", bszBuffer, 2, ulSig))
      printf("\nok");
   else printf("\nerr");
}

and this is the output of them

BLE 1            b0c66fb8 733df651 4c2d26de a9a0e334
BLE 2            f12b0ed9 782fae18 a87960ac f37b342e

BLD 1            f12b0ed9 782fae18 a87960ac f37b342e
BLD 2            dc95818a 65685d43 8d7d54b9 8827b217

err

modified on Friday, July 16, 2010 3:32 PM

C^hris L^osinger16-Jul-10 9:14

16-Jul-10 9:14

the conversion looks fine.

but doesn't Blowfish use 64-bit (8-byte) blocks ?

so this Blowfish_Encrypt(&bf, &tl->_1, &tl->_2); will encrypt the 8 bytes starting at tl->_1 ? since _1 is only four bytes, it's going to run over into _2.

so, you're doing BF(_1 + _2) -> _2 + _3.

or maybe not. i don't know what Blowfish_Encrypt looks like.

Re: proper conversion of unsigned char [16] -> unsigned long [4]

saiyuk6=716-Jul-10 9:28

Re: proper conversion of unsigned char [16] -> unsigned long [4]

16-Jul-10 9:28

well it would be total 8 blocks for both _1 and _2, because a unsigned long is 4 bytes
i am using Paul Kocher Blowfish Algorithm http://www.schneier.com/code/bfsh-koc.zip[^]

C^hris L^osinger16-Jul-10 9:40

16-Jul-10 9:40

ah, ok. it's expecting 2 32-bit parts. i thought the params were input and output.

Re: proper conversion of unsigned char [16] -> unsigned long [4]

saiyuk6=716-Jul-10 9:53

Re: proper conversion of unsigned char [16] -> unsigned long [4]

16-Jul-10 9:53

do you happen to know how i can fix it? i am lost

C^hris L^osinger16-Jul-10 10:27

16-Jul-10 10:27

you're doing this, right:

Sign:
Sig = BFE(MD5(data))

Verify:
X = BFD(Sig)
Sig2 = MD5(Sig)

Compare:
X == Sig2 ?

but shouldn't that bold line be Sig2 = MD5(data) ? the MD5 of the signature doesn't get you anything.

Re: proper conversion of unsigned char [16] -> unsigned long [4]

saiyuk6=716-Jul-10 10:40

Re: proper conversion of unsigned char [16] -> unsigned long [4]

16-Jul-10 10:40

i fixed it, thank you very much for your help, i noticed

if (tl->_1 == pBuffer[0] && tl->_2 == pBuffer[1] &&
    tl->_3 == pBuffer[2] && tl->_4 == pBuffer[4])
   bResult = true;

pBuffer[4] should of been pBuffer[3] i was going out of range of the array which didnt exist

here is the updated code

int _ustrlen(unsigned char *pbz)
{
   const unsigned char *ptr;
   for (ptr = pbz; *ptr; ++ptr);
   return (ptr - pbz);
}

void SignBuffer(unsigned char *pbzKey, unsigned char *pbzBuffer,
                unsigned int iSize, unsigned long *pSignature)
{
   MD5_CTX md5, md5key;
   BLOWFISH_CTX bf;

   MD5Init(&md5);
   MD5Update(&md5, pbzBuffer, iSize);
   MD5Final(&md5);

   _2LONG *tl = (_2LONG*) md5.digest;

   printf("BLE 1\t\t %x %x %x %x\n", tl->_1, tl->_2, tl->_3, tl->_4);

   MD5Init(&md5key);
   MD5Update(&md5key, pbzKey, _ustrlen(pbzKey));
   MD5Final(&md5key);

   Blowfish_Init(&bf, md5key.digest, 16);
   Blowfish_Encrypt(&bf, &tl->_1, &tl->_2);
   Blowfish_Encrypt(&bf, &tl->_3, &tl->_4);

   pSignature[0] = tl->_1;
   pSignature[1] = tl->_2;
   pSignature[2] = tl->_3;
   pSignature[3] = tl->_4;

   printf("BLE 2\t\t %x %x %x %x\n", pSignature[0], pSignature[1],
          pSignature[2], pSignature[3]);
}

bool VerifyBuffer(unsigned char *pbzKey, unsigned char *pSignature,
                  unsigned int iSize, unsigned long *pBuffer)
{
   BLOWFISH_CTX bf;
   MD5_CTX md5, md5key;
   bool bResult = false;

   printf("BLD 1\t\t %x %x %x %x\n", pBuffer[0], pBuffer[1],
          pBuffer[2], pBuffer[3]);

   MD5Init(&md5key);
   MD5Update(&md5key, pbzKey, _ustrlen(pbzKey));
   MD5Final(&md5key);

   Blowfish_Init(&bf, md5key.digest, 16);
   Blowfish_Decrypt(&bf, &pBuffer[0], &pBuffer[1]);
   Blowfish_Decrypt(&bf, &pBuffer[2], &pBuffer[3]);

   printf("BLD 2\t\t %x %x %x %x\n", pBuffer[0], pBuffer[1],
          pBuffer[2], pBuffer[3]);

   MD5Init(&md5);
   MD5Update(&md5, pSignature, iSize);
   MD5Final(&md5);

   _2LONG *tl = (_2LONG*) md5.digest;
   printf("BLD 3\t\t %x %x %x %x\n", tl->_1, tl->_2, tl->_3, tl->_4);

   if (tl->_1 == pBuffer[0] && tl->_2 == pBuffer[1] &&
       tl->_3 == pBuffer[2] && tl->_4 == pBuffer[3])
      bResult = true;

   return bResult;
}

void main(void)
{
   unsigned char bszBuffer[2] = {'A', 'B'};
   unsigned long ulSig[4];
   SignBuffer((unsigned char*)"key", bszBuffer, 2, ulSig);
   printf("\n");
   if (VerifyBuffer((unsigned char*)"key", bszBuffer, 2, ulSig))
      printf("\nok");
   else printf("\nerr");
}

I had made _ustrlen because i got tired of using casting for (unsigned char*), unless there is a different way?

Code-o-mat16-Jul-10 9:18

Code-o-mat

16-Jul-10 9:18

Hello!

It might only be me but it is not obvious (to me) where you are trying to do this conversion you are talking about, you should at least highlight the part of interest in your code so people who want to help don't have to study your whole program trying to grasp the way you think.

> The problem with computers is that they do what you tell them to do and not what you want them to do. <
> "It doesn't work, fix it" does not qualify as a bug report. <
> Amazing what new features none of the programmers working on the project ever heard of you can learn about when reading what the marketing guys wrote about it. <

Re: proper conversion of unsigned char [16] -> unsigned long [4]

saiyuk6=716-Jul-10 9:39