For those who have not heard about the AAM - Welcome to a cat.

Things really are not new and many of her in the course. But more people on it have not even heard.

So, I will not pull the cat’s tail - in most modern hard drives have this function, as AAM (Automatic Acoustic Management), allows on the verge of silence and performance. Physical meaning - management level of acoustic noise emitted by a block of magnetic heads for reading / writing. Changing the setting to influence the search path, when the head is moved along the radius of the disc to find the desired cylinder.

In most cases, programs (as the nominal change settings in Windows, I do not know) is a scale from 0 to 255 (sometimes - from 128 to 255). 0 (or 128) corresponds to the silence (minimum performance), all further - faster, higher, stronger!
Some manufacturers do not include this function, others are still at the plant exhibited a maximum for the most proizoditelnosti, and others - on the contrary, to have the glory of the quietest and “slow”. By the way, we should not think that setting to 0 (128) will do very much of the disc brakes. Since the disc has a constant frequency of rotation, linear reading speed is not changed. Increasing the access time to a specific sector (for reduced speed positioner). The minimum value of AAM increases the access time on average by 30%. Do not be afraid of these numbers - is unlikely you will ever need a “high” performance. But in addition to much noise from the falling of the drive to win may be marginally in power consumption and temperature devaysa, as well as its lifetime.

Change AAM setting in Windows can be many means - soft HDTune Pro, HDDLife Pro, Victoria, MHDD, Hitachi Feature Tool (not only on drives Hitachi) and I think, as something else. I take HDTune Pro (the free version does not) - the noise extrema of the difference can be evaluated directly in the program, I think, will be surprised.

More information about AAM can be read on IXBT.

Alternative Fiche this - use heavy body with thick soundproofed walls, a special cradle and boxes, to strengthen the drive all the screws, or using special pads, zoom system block away from the ears, etc.)

Who is not news - it’s indifference) A better answer, how do you refer to my reviews. What they need to add that extra, that should change? Reviews What devices are more interesting? Constructive criticism and fresh Misle will help me become better for you.

—

Check out the analysis of lumosity.com, cinematoday.jp, uniquearticlewizard.com, steadyhealth.com, style.com, mangashare.com, euro-football.ru, twitter.com, interpark.com, pajilleros.com - and much more

Related Posts

• New Lenovo PC quad power
• Broken down Toshiba Satellite L300
• Tuned Asus EEE PC 1000HD photo
• MSI Wind U123
• Bury shots again simulacrum hosting

What to do in this case?

We use the native code from Erlang. Fortunately, for this is as much the whole two ways:
Running native code in third-party process
Using shared library

I prefer the first method. If the native code, if you please break down (lyrics SIGSEGV or SIGFPE), almost no effect on the performance of core applications: the current operation is, of course returns an error, but we will re-launch-party process and will continue to work. But in the case of shared library - drop the whole node entirely.

Task

We have some functions written in C:
/ * Very simple primerchik, take two bytes, return one * /
byte sum (byte addee1, byte addee2) (
return addee1 + addee2;
)
/ * This function “according to legend was written long ago, one very nelyudivym borodotym Dyad’kov,
* With which young officers are afraid to communicate. It was so long ago that nobody remembers how it works … = (
* We only know that it takes an array of bytes, and <s> invert </ s> and something weird with him.
* (Ask people who have devoted their lives cryptography, not to throw tomatoes) * /
void simple_crypt (byte * in_out_chunk, int chunk_size) (
int i;
for (i = 0; i <chunk_size; i + +) (
in_out_chunk [i] ^ = 0xFF;
)
)

I want to be able to do so:
piece_of_code () ->
Encrypted = simple_crypt (Orig),
…

Solution Overview

From the perspective of the operating system we will be running two processes: Erlang and Native. They will communicate via stdin and stdout of the latter.

For the programmer it will look as follows:
module in Erlang exporting simple_crypt / 1 and the sum / 2
Part C programs directly from the functions of interest to us
Part C program, dedicated I / Os
Part C programs dealing marshalling (the dirty and unpleasant work)

Boring technical details about the marshalling

Since communicating with Erlang C in our case through the streams of bytes, will somehow dodge and serialize our data types, so that they are in these flows went to bed.
Components will communicate “proposals”, in which the first two bytes will be official, and will denote the size of the remaining part (carrying semantic meaning).
Size will be written in Little-endian format: [0xAD, 0x0A] - will be denoted 0xAD * 161 + 0×0A.
When prompted, the first byte in the “body” of the request would be caused by an index function. The rest - the arguments for the function.
The answer - just the length, and data return value.
Since in our case the function does not accept / do not return any data types other than bytes - the task much easier: the arguments can be simply read and write, and do not pack and unpack.

The sun was still high … work …

Where do we start?
component number 2 in the review of the decision

/ * Header * /
# define IDX_SUM 1
# define IDX_SIMPLE_CRYPT 3

int sum (byte addee1, byte addee2);
void simple_crypt (byte * in_out_chunk, int chunk_size);
/ * Source * /
int sum (byte addee1, byte addee2) (
return addee1 + addee2;
)
void simple_crypt (byte * in_out_chunk, int chunk_size) (
int i;
for (i = 0; i <chunk_size; i + +) (
in_out_chunk [i] ^ = 0xFF;
)
)

It’s simple. We give a soulless index function names and write functions, which, in the conditions of the problem can not be implemented in Erlang.

Input output

Who cares - read carefully.
Who is not really - just use it, not really thinking: this part will not change in 99.9% of cases.
# include <unistd.h>
# include <stdio.h>
# include <errno.h>
/ *
* Read the command of Erlang
* /
int read_cmd (byte * buff) (
int len;
/ * Try to extort the length of the data from stdin * /
if ((len = read_exact (buff, 2)) = 2) (
return -1;
)
/ * Consider it * /
len = (buff [0] <<8) | buff [1];
/ * Finished reading data from stdin * /
return read_exact (buff, len);
)
/ *
* This is a lower level than the previous one.
* HERE, we read a specified number of bytes.
* /
int read_exact (byte * buff, int len) (
int i, got = 0;
do (
i = read (0 / * STDIN * /, buff + got, len - got);
if (i <= 0) (
return i;
)
got + = i;
) While (got <len);
return len;
)
/ *
* The answer Erlang
* /
int write_cmd (byte * buff, int len) (
/ * Erlang also expects us to answer length * /
byte li;
li = (len>> & 0xff;
write_exact (& li, 1);
li = len & 0xff;
write_exact (& li, 1);
/ * The answer * /
return write_exact (buff, len);
)
/ *
* Write bytes
* /
int write_exact (byte * buff, int len) (
int i; int wrote = 0;
do (
if ((i = write (1 / * STDOUT * /, buff + wrote, len - wrote)) <= 0) (
return i;
)
wrote + = i;
) While (wrote <len);
return len;
)

Marshalling calls

In principle, a little bit and not difficult. But when you add another function, it will be necessary to climb here and add another branch test the index function … Nepriyatnenko.
On the other hand can duplicate pieces “shod” in the macros and enjoy life. =)
int main (int argc, char ** argv) (
int body_size;
int fn_idx;
byte buff [BUFF_SIZE];

while ((body_size = read_cmd (buff))> 0) (
fn_idx = buff [0] / * All as described above: the first byte of the body - index / Haydee called function * /

if (fn_idx == IDX_SUM) (/ * int sum (int, int) * /
int arg1, arg2, result;
arg1 = buff [1];
arg2 = buff [2];
result = sum (arg1, arg2);
buff [0] = result;
write_cmd (buff, 1);
)
else if (fn_idx == IDX_SIMPLE_CRYPT) (/ * void simple_crypt (byte *, int) * /
simple_crypt (
buff + 1, / * one long argument … * /
body_size - 1 / * … that’s such a long * /
)
/ *
* By the way, if you too lazy to deal with GDB: debazhitsya be like this:
fprintf (
stderr / * Erlang it does not take into account, but simply transfers to your console * /
“simple_crypt, chunk_size =% i \ r \ n”, body_size - 1
)
* /
write_cmd (buff + 1, body_size - 1);
)
)
return 0;
)

And for dessert …

Now we write Erlang module, which will provide a convenient interface for a call okoyannyh Legacy Functions
%% Before using the functions - it is necessary to run our c_server
start () ->
spawn_link (fun () ->
register (c_server, self ()),
process_flag (trap_exit, true),
Port = open_port ((spawn, “. / C_server”), [(packet, 2)]),% I turned “. / C_server”, as you compile - you decide
loop (Port)
end).
%% Does the reverse of the previous function =)
stop () ->
c_server! stop.

%% Like this, we will ask c_server call sishnuyu function …
call_c_server (Msg) ->
c_server! (call, self (), Msg),
receive
(c_server, Result) ->
(ok, Result)
after
5000 -> (fail, timeout)
end.

%% … No, better like this.
sum (Addee1, Addee2) ->
call_c_server ((sum, Addee1, Addee2)).
simple_crypt (Chunk) ->
call_c_server ((simple_crypt, Chunk)).

%% Serialization =)
pack_buff ((sum, Addee1, Addee2)) -> [1, Addee1, Addee2];% pack two byte-ints for function # 1
pack_buff ((simple_crypt, Chunk)) -> [3 | Chunk]. % Pack byte array for function # 3
%% Deserialization
unpack_buff ([Result]) -> Result;% int return value
unpack_buff (AnyThing) -> AnyThing. % Byte array return value
%% Yeah … easier than in C … =)

%% Directly, dispatcher
loop (Port) ->
receive
( ‘EXIT’, Port, Reason) ->% in case of “Oh!”
exit ((port_terminated, Reason));
stop ->%% stop
Port! (self (), close),
receive
(Port, closed) ->
exit (normal)
end;
(call, From, Msg) ->% that, for what is being written
Port! (self (), (command, pack_buff (Msg))),% kommanduem …
receive
(Port, (data, Data)) ->% rejoice result
From! (c_server, unpack_buff (Data)),
loop (Port);
AnyThing ->% indignant
io: format ( “received invalid message: ~ p ~ n”, AnyThing),
loop (Port)
end
end.

Let’s go! ©

Check what we got.
Run erl-shell.
Eshell V5.6.5 (abort with ^ G)
1> c_server: sum (1, 5).
** Exception error: bad argument
in function c_server: call_c_server / 1
2> c_server: start ().
<0.35.0>
3> c_server: sum (1, 5).
6
4> c_server: simple_crypt ( “Its bean - a heart dace knight”).
[182,139,140,223,157,154,158,145,223,210,223,158,223,151,
154,158,141,139,223,155,158,156,154,223,148,145,150,152,151 |...]
5> c_server: simple_crypt ([182,139,140,223,157,154,158,145,223,210,223,158,223,151,154,158,141,139,223,155,158,156,154,223,148,145,150,152,151,139]).
“Its bean - a heart dace knight”
6> io: format ( “Hooray!”)

Summarize

We were able to bind together the Native code and Erlang.
One question remains: And Haha?
But, Haha - native code could be:
old and ugly, but very necessary
new and highly-optimized code
general, no, we did it for Lulzim

You see, if you start to learn Erlang, you already have as many as three reasons to learn the material in this article. =)

Appendix A. Really very short troubleshooting guide

For those who did not get to run, tell, with what problem I encountered.
We are looking for a piece of code here:
(Port, (data, Data)) ->% rejoice result
From! (c_server, unpack_buff (Data)),
loop (Port);

Harvesting from the loop (Port), and for understandable reasons, we obtain single-shot-library, which needs the start before each call.
At first I was looking for a problem in Cu. I tried to change the properties through fcntl stdin (it seemed to me that he somehow becomes a non-blocking).
A casket was opened just …

Appendix B. Reference

Programming Erlang. Software for a concurrent world.
Joe Armstrong. ISBN-10: 1-9343560-0-X

Related Posts

• The Europas Liveblog 2009
• Samsung NV100HD Product Image - Gold
• Broken down Toshiba Satellite L300
• To all HDD users
• New Samsung Touch-Screen F490 photo

Related Posts

• New Samsung Touch-Screen F490 photo
• New Lenovo PC quad power
• Hello world!
• BlackKnight i7 WorkStation PC
• New lenovo Laptop X300 photo

The last time we looked, Bing sat at about 9,4% of the search market in July, but the new Nielsen numbers confirm the continuing growth of the search engine through the month of August, up 10.7% of the total market.

Bin Count

The rankings for the top-11 engine shown above, showing an impressive growth Bing month about 22% over the same period of the month. Compare this with the rest of the package, of which no one breaks more than 3% more (Yahoo and Comcast both lose share).

This growth is largely due to huge marketing campaign, but also, of course, that some of the innovations Microsoft brings to the table with Bing. Recently, Visual search function, possibly Flat Out best for viewing certain types of information, as well as social exchange search results will soon come to Binga, as well as an interesting new twist.

Have you used the Bin? What is your impression of the new search engine so far?

Related Posts

• Toshiba 6TP-15
• Tuned Asus EEE PC 1000HD photo
• EEE PC LEDs photo
• Hello world!
• To all HDD users

Related Posts

• To all HDD users
• New Lenovo laptop
• Toshiba 6TP-15
• Broken down Toshiba Satellite L300
• MSI Wind U123

Related Posts

• BlackKnight i7 WorkStation PC
• Free the postcode now! All you need is an iPhone
• Hello world!
• New Samsung Touch-Screen F490 photo
• Tuned Asus EEE PC 1000HD photo

Since the origins the Internet has increased thousand times in size and today the majority of traffic comes from http. There is uncountable number of various protocols to make everything work right and it is so amazing how the world has changed since the Internet burst into our lives.

Related Posts

• Bury shots again simulacrum hosting
• New Lenovo PC quad power
• Toshiba 6TP-15
• EEE PC LEDs photo
• Erlang

Related Posts

• New Samsung Touch-Screen F490 photo

Related Posts

• Samsung NV100HD Product Image - Gold
• New Lenovo PC quad power
• Toshiba 7TP-303
• Toshiba 6TP-15
• Broken down Toshiba Satellite L300

Related Posts

• MSI Wind U123