C/C++ vs Assembler

[Gunther]

  as if the subject has never come up before

oh yes, it's a very new topic.  :lol: :lol: :lol:

Gunther

[jj2007]

Behold and see...

Intel(R) Celeron(R) M CPU        420  @ 1.60GHz (SSE3)
41      cycles for Small 1
38      cycles for Small 2
40      cycles for Small 3
42      cycles for Small 3.1
41      cycles for Small 4
50      cycles for C version
53      cycles for C mod JJ
58      hsz2dw (Microsoft 32-Bit C/C++ optimization compiler v16.00.40219.01)
24      hsz2dw2 (unrolled 4 times)

41      cycles for Small 1
39      cycles for Small 2
41      cycles for Small 3
57      cycles for Small 3.1
23      cycles for Small 4
51      cycles for C version
53      cycles for C mod JJ
31      hsz2dw (Microsoft 32-Bit C/C++ optimization compiler v16.00.40219.01)
24      hsz2dw2 (unrolled 4 times)

43      cycles for Small 1
38      cycles for Small 2
41      cycles for Small 3
42      cycles for Small 3.1
41      cycles for Small 4
50      cycles for C version
51      cycles for C mod JJ
31      hsz2dw (Microsoft 32-Bit C/C++ optimization compiler v16.00.40219.01)
24      hsz2dw2 (unrolled 4 times)

Well, a LUT is difficult to beat ;-)

[habran]

 :t

Code Select Expand


Intel(R) Core(TM) i7-3610QM CPU @ 2.30GHz (SSE4)



7       cycles for Fast version
8       cycles for Fast version under AMD
16      cycles for Small 1
18      cycles for Small 2
19      cycles for Small 3
19      cycles for Small 3.1
19      cycles for Small 4
4       cycles for MMX 1
5       cycles for MMX 2
4       cycles for SSE1

Other's Versions:
23      cycles for Axhex2dw improved by Hutch (1)
22      cycles for Axhex2dw improved by Hutch (2)

3       cycles for Lingo's SSE version
7       cycles for Lingo's BIG integer version
21      cycles for Jochen's WORD-Indexed version
12      cycles for Dave's version (with minor changes)


11      cycles for Fast version
14      cycles for Fast version under AMD
17      cycles for Small 1
19      cycles for Small 2
18      cycles for Small 3
19      cycles for Small 3.1
18      cycles for Small 4
5       cycles for MMX 1
4       cycles for MMX 2
4       cycles for SSE1

Other's Versions:
22      cycles for Axhex2dw improved by Hutch (1)
21      cycles for Axhex2dw improved by Hutch (2)

2       cycles for Lingo's SSE version
7       cycles for Lingo's BIG integer version
5       cycles for Jochen's WORD-Indexed version
12      cycles for Dave's version (with minor changes)


10      cycles for Fast version
13      cycles for Fast version under AMD
17      cycles for Small 1
17      cycles for Small 2
18      cycles for Small 3
19      cycles for Small 3.1
19      cycles for Small 4
5       cycles for MMX 1
6       cycles for MMX 2
5       cycles for SSE1

Other's Versions:
23      cycles for Axhex2dw improved by Hutch (1)
22      cycles for Axhex2dw improved by Hutch (2)

4       cycles for Lingo's SSE version
7       cycles for Lingo's BIG integer version
6       cycles for Jochen's WORD-Indexed version
11      cycles for Dave's version (with minor changes)

==========
Codesizes:
Axhex2dw_Unrolled:      396
Axhex2dw_Unrolled_AMD:  396
Axhex2dw1 - 1:  69
Axhex2dw2 - 2:  48
Axhex2dw3 - 3:  57
Axhex2dw3_1 - 3.1:      56
Axhex2dw3 - 4:  61
Axhex2dw_MMX:   128
Axhex2dw_MMX2:  160
Axhex2dw_SSE:   160
Alex_Short_Hutch:       59
Axhex2dw_Hutch2:        54
Hex2dwLingoSSE: 160
lingo_htodw:    1950
ax_jj_htodw:    174
krbhtodw:       547
--- ok ---


[Gunther]

Jochen,

Well, a LUT is difficult to beat ;-)

that's true, but an old wisdom.

Gunther

[Antariy]

OK - I obviously missed the "spirit" of this thread...

No, it's OK, your code is good - fast and well readable, really :t Thanks for posting it It was informational test, too, now in this strange thread (@Dave - ) we do have rolled and unrolled C (and you have provided both) and ASM versions of the code, mixed in a crazy testbeds

[Antariy]

:t

Thank you, habran :t

If your OS is 32 bit, then it really seems as a best idea to run 32 bit proggies under 32 bit OS

[habran]

It is 64 bit Win 7
IMO there is no penalty for running 32 on 64 but 64 is certainly faster because of 64 bit programing :t

[Antariy]

That was my assumption just because your timing results seem to be smaller than other's with the same CPU

[habran]

It is Toshiba Qosmio 16 gig ram laptop with 2.3 gig i7 and 64 bit Windows 7 Home with AVX :t
I think that qWord has got the same one
It is a great toy

[Antariy]

It is a great toy

Sure it is

[Gunther]

Hi habran,

It is Toshiba Qosmio 16 gig ram laptop with 2.3 gig i7 and 64 bit Windows 7 Home with AVX :t
I think that qWord has got the same one
It is a great toy

it's probably the Ivy Bridge, isn't it?

Gunther

[habran]

Yes Gunther, the Ivy Bridge it is

I already posted this before
here are specifications:                 
                                         
Intel® Core™ i7-3610QM Processor         
(6M Cache, up to 3.30 GHz)               
Specifications                           
Essentials                               
Status   Launched                         
Launch Date   Q2'12                       
Processor Number   i7-3610QM             
# of Cores   4                           
# of Threads   8                         
Clock Speed   2.3 GHz                     
Max Turbo Frequency   3.3 GHz             
Intel® Smart Cache   6 MB                 
Bus/Core Ratio   23                       
DMI   5 GT/s                             
Instruction Set   64-bit                 
Instruction Set Extensions   AVX         
Embedded Options Available   No           
Lithography   22 nm                       
Max TDP   45 W                           
Recommended Customer Price   TRAY: $378.00

[Gunther]

Hi habran,

I already posted this before
here are specifications:                 
                                         
Intel® Core™ i7-3610QM Processor         
(6M Cache, up to 3.30 GHz)               
Specifications                           
Essentials                               
Status   Launched                         
Launch Date   Q2'12                       
Processor Number   i7-3610QM             
# of Cores   4                           
# of Threads   8                         
Clock Speed   2.3 GHz                     
Max Turbo Frequency   3.3 GHz             
Intel® Smart Cache   6 MB                 
Bus/Core Ratio   23                       
DMI   5 GT/s                             
Instruction Set   64-bit                 
Instruction Set Extensions   AVX         
Embedded Options Available   No           
Lithography   22 nm                       
Max TDP   45 W                           
Recommended Customer Price   TRAY: $378.00

an excellent machine. Runs Windows 64 as the only OS? How did you manage the new "BIOS"?

Gunther

[habran]

Thanks Gunther
I did not have to touch anything
I just installed my tools and copyed my projects 

[Gunther]

Hi habran,

I did not have to touch anything
I just installed my tools and copyed my projects 

so, do you have an EFI drive, too? Or is your hard disk not over 2.2 GB size?

Gunther