RDP1974

btw. I'll provide intel one api accelerated zlib.so library for linux too

look this -> using zlib accelerated deflate (RDPMM64 repo zip.dll)

msheap, isapi dll

Document Path:          /isapi/testisapi.dll
Document Length:        8416 bytes

Concurrency Level:      100
Time taken for tests:   0.609 seconds
Complete requests:      1000
Failed requests:        0
Total transferred:      8617000 bytes
HTML transferred:       8416000 bytes
Requests per second:    1641.52 [#/sec] (mean)
Time per request:       60.919 [ms] (mean)
Time per request:       0.609 [ms] (mean, across all concurrent requests)
Transfer rate:          13813.49 [Kbytes/sec] received

added:

uses JsonDataObjects, DataSet.Serialize, RDPWebbroker64;
 

in response method:

Response.ContentType:='application/json; charset="UTF-8"';
Response.ZlibDeflate;

end;

response reduced in real time from 43kb to 8kb (thus saving time from server -> to browser and bandwidth cloud cost) while keeping throughput reqs/s unchanged

hi,

I have done a more extended sample, with webbroker indy http server, json parsing, data serialize to firedac memtable and populate a response, using threadvar blobs

single request >
1 request 1 thread: ab -n 1 -c 1 -k http://192.168.1.110:8080/

default
Concurrency Level:      1
Time taken for tests:   0.003 seconds
Complete requests:      1
Failed requests:        0
Keep-Alive requests:    0
Total transferred:      43146 bytes
HTML transferred:       43008 bytes
Requests per second:    287.27 [#/sec] (mean)
Time per request:       3.481 [ms] (mean)
Time per request:       3.481 [ms] (mean, across all concurrent requests)
Transfer rate:          12104.21 [Kbytes/sec] received

rdp64 intel tbb
Concurrency Level:      1
Time taken for tests:   0.003 seconds
Complete requests:      1
Failed requests:        0
Keep-Alive requests:    0
Total transferred:      43146 bytes
HTML transferred:       43008 bytes
Requests per second:    287.44 [#/sec] (mean)
Time per request:       3.479 [ms] (mean)
Time per request:       3.479 [ms] (mean, across all concurrent requests)
Transfer rate:          12111.17 [Kbytes/sec] received

msheap
Concurrency Level:      1
Time taken for tests:   0.005 seconds
Complete requests:      1
Failed requests:        0
Keep-Alive requests:    0
Total transferred:      43146 bytes
HTML transferred:       43008 bytes
Requests per second:    191.57 [#/sec] (mean)
Time per request:       5.220 [ms] (mean)
Time per request:       5.220 [ms] (mean, across all concurrent requests)
Transfer rate:          8071.79 [Kbytes/sec] received

fastmm5
Concurrency Level:      1
Time taken for tests:   0.005 seconds
Complete requests:      1
Failed requests:        0
Keep-Alive requests:    0
Total transferred:      43146 bytes
HTML transferred:       43008 bytes
Requests per second:    191.64 [#/sec] (mean)
Time per request:       5.218 [ms] (mean)
Time per request:       5.218 [ms] (mean, across all concurrent requests)
Transfer rate:          8074.89 [Kbytes/sec] received

> multi thread test >
100 requests 100 threads: ab -n 100 -c 100 -k http://192.168.1.110:8080/ 

default:
Concurrency Level:      100
Time taken for tests:   1.549 seconds
Complete requests:      100
Failed requests:        0
Keep-Alive requests:    0
Total transferred:      4314600 bytes
HTML transferred:       4300800 bytes
Requests per second:    64.56 [#/sec] (mean)
Time per request:       1548.967 [ms] (mean)
Time per request:       15.490 [ms] (mean, across all concurrent requests)
Transfer rate:          2720.18 [Kbytes/sec] received

rdp64 intel tbb
Concurrency Level:      100
Time taken for tests:   0.063 seconds
Complete requests:      100
Failed requests:        0
Keep-Alive requests:    0
Total transferred:      4314600 bytes
HTML transferred:       4300800 bytes
Requests per second:    1596.37 [#/sec] (mean)
Time per request:       62.642 [ms] (mean)
Time per request:       0.626 [ms] (mean, across all concurrent requests)
Transfer rate:          67262.80 [Kbytes/sec] received

msheap
Concurrency Level:      100
Time taken for tests:   0.070 seconds
Complete requests:      100
Failed requests:        0
Keep-Alive requests:    0
Total transferred:      4314600 bytes
HTML transferred:       4300800 bytes
Requests per second:    1431.02 [#/sec] (mean)
Time per request:       69.880 [ms] (mean)
Time per request:       0.699 [ms] (mean, across all concurrent requests)
Transfer rate:          60295.89 [Kbytes/sec] received

fastmm5
Concurrency Level:      100
Time taken for tests:   0.110 seconds
Complete requests:      100
Failed requests:        0
Keep-Alive requests:    0
Total transferred:      4314600 bytes
HTML transferred:       4300800 bytes
Requests per second:    909.90 [#/sec] (mean)
Time per request:       109.902 [ms] (mean)
Time per request:       1.099 [ms] (mean, across all concurrent requests)
Transfer rate:          38338.49 [Kbytes/sec] received

The ab bench is not very granular, so I assume that in a single thread all allocators should be close together.

I also did a test on linux ubuntu 22 in the same host and the results are similar to windows with msheap or tbb.

kind regards

btw. I used libraries from JsonDataObjects, DataSet.Serialize

TestWebbroker.zip

https://learn.microsoft.com/en-us/windows/win32/memory/low-fragmentation-heap

https://illmatics.com/Understanding_the_LFH.pdf

https://www.softwareverify.com/blog/memory-fragmentation-your-worst-nightmare/

https://users.rust-lang.org/t/why-dont-windows-targets-use-malloc-instead-of-heapalloc/57936

I don't know if we need an intermediate allocator or if we can use the Win API heap directly.

this in win2022 server 21h2 hyper-v guest

C:\Exes>cmd3_def
Total Time: 1,062 Seconds
Hands Per Second  2447231,63841808
Hand Evaluation  Expected Actual
Royal Flushes:          4 4
Straight Flushes:      36 36
Four of a Kinds:      624 624
Full Houses:         3744 3744
Flushes:             5108 5108
Straights:          10200 10200
Three of a Kinds:   54912 54912
Two Pairs:         123552 123552
One Pairs:        1098240 1098240
Other:            1302540 1302540
Total Hands:      2598960 2598960

C:\Exes>cmd3_fm5 (fastmm5)
Total Time: 1 Seconds
Hands Per Second  2598960
Hand Evaluation  Expected Actual
Royal Flushes:          4 4
Straight Flushes:      36 36
Four of a Kinds:      624 624
Full Houses:         3744 3744
Flushes:             5108 5108
Straights:          10200 10200
Three of a Kinds:   54912 54912
Two Pairs:         123552 123552
One Pairs:        1098240 1098240
Other:            1302540 1302540
Total Hands:      2598960 2598960

C:\Exes>cmd3_rdp64
Total Time: 1,344 Seconds
Hands Per Second  1933750
Hand Evaluation  Expected Actual
Royal Flushes:          4 4
Straight Flushes:      36 36
Four of a Kinds:      624 624
Full Houses:         3744 3744
Flushes:             5108 5108
Straights:          10200 10200
Three of a Kinds:   54912 54912
Two Pairs:         123552 123552
One Pairs:        1098240 1098240
Other:            1302540 1302540
Total Hands:      2598960 2598960

C:\Exes>cmd3_laz
Total Time: 2,266 Seconds
Hands Per Second  1146937,33451015
Hand Evaluation  Expected Actual
Royal Flushes:          4 4
Straight Flushes:      36 36
Four of a Kinds:      624 624
Full Houses:         3744 3744
Flushes:             5108 5108
Straights:          10200 10200
Three of a Kinds:   54912 54912
Two Pairs:         123552 123552
One Pairs:        1098240 1098240
Other:            1302540 1302540
Total Hands:      2598960 2598960

C:\Exes>cmd3_msheap
Total Time: 0,984 Seconds
Hands Per Second  2641219,51219512
Hand Evaluation  Expected Actual
Royal Flushes:          4 4
Straight Flushes:      36 36
Four of a Kinds:      624 624
Full Houses:         3744 3744
Flushes:             5108 5108
Straights:          10200 10200
Three of a Kinds:   54912 54912
Two Pairs:         123552 123552
One Pairs:        1098240 1098240
Other:            1302540 1302540
Total Hands:      2598960 2598960

hi,

I did a quick benchmark test for single threaded application (using attached poker game, of course it's not exhaustive with a small subset of things)

the win heap manager seems enhanced, now direct heap is faster than default MM or Intel TBB (using host I-9900kf win11 24h2 delphi 12.2.1 fpc 3.2.2 release mode)

single thread x64 console mode->

D12 default:

Total Time: 1,031 Seconds
Hands Per Second  2520814,74296799
Hand Evaluation  Expected Actual
Royal Flushes:          4 4
Straight Flushes:      36 36
Four of a Kinds:      624 624
Full Houses:         3744 3744
Flushes:             5108 5108
Straights:          10200 10200
Three of a Kinds:   54912 54912
Two Pairs:         123552 123552
One Pairs:        1098240 1098240
Other:            1302540 1302540
Total Hands:      2598960 2598960

D12 intel tbb (rdpmm64):

Total Time: 1,281 Seconds
Hands Per Second  2028852,45901639
Hand Evaluation  Expected Actual
Royal Flushes:          4 4
Straight Flushes:      36 36
Four of a Kinds:      624 624
Full Houses:         3744 3744
Flushes:             5108 5108
Straights:          10200 10200
Three of a Kinds:   54912 54912
Two Pairs:         123552 123552
One Pairs:        1098240 1098240
Other:            1302540 1302540
Total Hands:      2598960 2598960

D12 msheap:

Total Time: 0,984 Seconds
Hands Per Second  2641219,51219512
Hand Evaluation  Expected Actual
Royal Flushes:          4 4
Straight Flushes:      36 36
Four of a Kinds:      624 624
Full Houses:         3744 3744
Flushes:             5108 5108
Straights:          10200 10200
Three of a Kinds:   54912 54912
Two Pairs:         123552 123552
One Pairs:        1098240 1098240
Other:            1302540 1302540
Total Hands:      2598960 2598960

latest FPC lazarus:

Total Time: 2,25 Seconds
Hands Per Second  1155093,33333333
Hand Evaluation  Expected Actual
Royal Flushes:          4 4
Straight Flushes:      36 36
Four of a Kinds:      624 624
Full Houses:         3744 3744
Flushes:             5108 5108
Straights:          10200 10200
Three of a Kinds:   54912 54912
Two Pairs:         123552 123552
One Pairs:        1098240 1098240
Other:            1302540 1302540
Total Hands:      2598960 2598960

Many RTL are using directly the heap of windows, as Rust, Clang and others, it resists fragmentation, so I suppose it is okay to use it directly. Also this act very well in multithreading as webbroker apps.

look here if you wish

https://github.com/RDP1974/

Sorry to bore you with these things, just out of curiosity to squeeze the possible performance
btw. do you know a more complete real-world test scenario than this?

wldd49.zip

I have not removed parnassus plugins

now on this win11 24h2 the setup won't complete, both web install or iso

eg. a typical TCP chat client server, with a common protocol as discord or whatsapp or old msn.

On 10/12/2024 at 10:07 PM, Arnaud Bouchez said:

Personal biais: the mORMot 2 Open Source framework has a very efficient JSON library, and several ways to use it:

- from RTTI, using classes, records, collections, dynamic arrays, mORMot generics...
- from variants, and a custom "document" variant type to store JSON objects or arrays...
- from high-level IDocList / IDocDict holders.

 

See https://blog.synopse.info/?post/2024/02/01/Easy-JSON-with-Delphi-and-FPC

 

It is perhaps the fastest library available, working on Delphi and FPC, with unique features, like:

  list := DocList('[{"a":10,"b":20},{"a":1,"b":21},{"a":11,"b":20}]');
  // sort a list/array by the nested objects field(s)
  list.SortByKeyValue(['b', 'a']);
  assert(list.Json = '[{"a":10,"b":20},{"a":11,"b":20},{"a":1,"b":21}]');

  // create a dictionary from key:value pairs supplied from code
  dict := DocDict(['one', 1, 'two', 2, 'three', _Arr([5, 6, 7, 'huit'])]);
  assert(dict.Len = 3); // one dictionary with 3 elements
  assert(dict.Json = '{"one":1,"two":2,"three":[5,6,7,"huit"]}');
  // convert to JSON with nice formatting (line feeds and spaces)
  Memo1.Caption := dic.ToString(jsonHumanReadable);

  // integrated search / filter
  assert(DocList('[{ab:1,cd:{ef:"two"}}]').First('ab<>0').cd.ef = 'two');

 

Hi, is it possible to use a retrieval path method for example on a list with multiple arrays inside?
Many libraries offer a subset of https://goessner.net/articles/JsonPath/ for simple retrievals.

I also did this small test (parsing, path retrieve) over Grijjy.Bson, NesLib.Json, System.Json

of course this test is not exhaustive, anyway Ahausladen is the fastest

(D12.2 x64 Win11 24h2)

also there are many excellent libs (https://github.com/Fr0sT-Brutal/awesome-pascal)

https://github.com/paolo-rossi/delphi-neon

https://github.com/ahausladen/JsonDataObjects

https://github.com/grijjy/GrijjyFoundation

for now I have tested Hausladen JsonDataObjects : that's 3x faster than JsonTools and equally easy

Delphi community is unbeatable!

hi,
can I ask your opinion about this JSON library to adopt?

As much as there are good repositories for Delphi,
in the last period I feel comfortable with this JsonTools (I handled all the blobs correctly from many devices where other libraries gave me errors in parsing)

https://www.getlazarus.org/json/
https://www.getlazarus.org/json/tests/
(all test passed)
btw. I am a loyal Delphi customer, I don't want to advertise Lazarus 🙂

On 8/17/2024 at 6:19 PM, RDP1974 said:

from ver 8 the dll are inside the c:\program files\mysql\bin folder of the server

latest version needs also a pair of dll crypto ssleay or something for the new ssl protocol auth

e.c. these dll are only for 64bit latest mysql (copy them from or use the bin folder of the mysql in FDPhysMySQLDriverLink)

from ver 8 the dll are inside the c:\program files\mysql\bin folder of the server

latest version needs also a pair of dll crypto ssleay or something for the new ssl protocol auth

55 minutes ago, Stefan Glienke said:

You *must not* remove that line - Move in Delphi allowed to be called with negative Count (a possible result of some size calculation), resulting in a no-op in the System implementation. Passing a negative number to most C++ implementations will result in passing a value >2mio because their size parameter is unsigned.

Also, the performance difference is hardly about that little check but the ippsMove_8u implementation.

thanks for the hint

I correct

look, I did only pascal wrappers, here's how to obtain the dll's:

install visual studio c++ (I use the 2019 version)

install intel ipp https://www.intel.com/content/www/us/en/developer/tools/oneapi/ipp.html

install python

the allocator is here:

https://www.oneapi.io/

https://github.com/oneapi-src/oneTBB/tree/master

see cmake dir, run it and build a visual studio project, locate tbbmalloc and compile it (you should select MD multithreading static library)

the simd rtl replacement:

install qt5 python library, if I remember "pip install pyqt5"
python C:\Program Files (x86)\Intel\oneAPI\ipp\latest\opt\ipp\tools\custom_library_tool_python\main.py

this tool will build a vc script to create the custom dll
locate ippsZero_8u, ippsCopy_8u, ippsMove_8u, ippsSet_8u, ippsFind_8u, ippsCompare_8u, ippsUppercaseLatin_8u_I, ippsReplaceC_8u from signal processing, and other select

(indeed we can extend many functions from this tool, especially for image processing, but I didn't have time to do)

run the script to obtain the dll

the zlib accelerated with ipp:

C:\Program Files (x86)\Intel\oneAPI\ipp\2021.12\share\doc\ipp\components_and_examples_win\components\interfaces

here you can see common libraries enhanced, check the folder zlib and open the readme, follow the instructions

tell me if you have troubles in build this

kind regards

maybe in rdpsimd64 can we rem this line?

which cpu do you test?

procedure Move2(const Source; var Dest; Count: NativeInt); inline;
begin
  //if Count > 0 then //>>> useless checking?
    SeaMove(@Source, @Dest, Count);
end;

I removed the check, works ok, thanks for the hint

do you know if somebody has done a webbroker bridge to ms http2 api (http.sys)?

else should be nice build it and add a session mechanism using tdictionary

hints?

ok, I will publish a tutorial how to build the dll's

the allocator is open source here: https://github.com/oneapi-src/oneTBB/tree/master

the simd rtl and the zlib are done using ipp suite: https://www.intel.com/content/www/us/en/developer/tools/oneapi/ipp.html

dll are built on a clean windows 2022 setup

hi,

I have updated the libraries and units of the RTL patches from Intel IPP oneTBB performance suite

these are well suited for web server application scalability on windows architecture

https://github.com/RDP1974

(test on concurrent http calls show 13x speedup)

there are enhancements on zlib 1.3.1 options, and some random av seems solved correctly

I have done testing, but please if you found troubles please notify me at roberto.dellapasqua dot live.com

kind regards

Roberto

hi,

can be possible to use thttpclient in async mode and do 4 get concurrently (as chromium does)? are they overlapped? how to discover the sender?

or we need to run 1 get at once in sequence?

hi,

I want do this with fmx:

1- vertscrollbox

2- tlayout 

3- create at runtime trectangle then fill it with a bitmap, dinamically, when needed (as add tiles)

I did a small test without success

how to do?

btw. I like to add event callbacks too, as onclick()

btw. the best method to draw things, call update, etc.

hi,

I need to develop webrtc ios android app

can someone suggest me which is better among Alcinoe https://github.com/MagicFoundation/Alcinoe or http://www.flashavconverter.com/content/webrtc-delphi-component (this last doesn't patch units and this is preferable to me)

about reliability and quality?

thanks for any hints

mysql oracle from ver.8 is 64bit only

c connectors dll are inside \lib and \bin folders

(without 32bit version)

the solution of Lajos seems a good workaround (32bit odbc in designtime and 64bit native for runtime)

very thanks