Author Topic: The fastest way to fill a dword array with string values (Read 74942 times)

MichaelW · « **Reply #30 on:** December 09, 2012, 12:32:20 PM »

For Dave’s code I get 2292 2299 2303 2302 2290 on a P4 Northwood and 2674 2676 2674 2677 2690 on a P3. Frank’s EXE will not run on my P3 Windows 2000 system:

“Test_FillArray.exe is not a valid Win32 application.”

But on my P4 Northwood Windows XP system I typically get:

Code: [Select]

------------------------------------------------------------------------
Intel(R) Pentium(R) 4 CPU 3.00GHz

Instructions: MMX, SSE1, SSE2
------------------------------------------------------------------------
2676    cycles for Dedndave code
------------------------------------------------------------------------
2655    cycles for Dedndave code
------------------------------------------------------------------------
2306    cycles for Dedndave code
------------------------------------------------------------------------
2277    cycles for Dedndave code
------------------------------------------------------------------------

But running it multiple times the first one or two results are significantly higher than the others, suggesting that the code is not delaying long enough after it loads before it starts counting cycles. After adding a:

invoke Sleep, 5000

Below the start label and assembling and linking with ML 6.15.8803 and Link 5.12.8078, typical results are:

Code: [Select]

------------------------------------------------------------------------
Intel(R) Pentium(R) 4 CPU 3.00GHz

Instructions: MMX, SSE1, SSE2
------------------------------------------------------------------------
2278    cycles for Dedndave code
------------------------------------------------------------------------
2296    cycles for Dedndave code
------------------------------------------------------------------------
2263    cycles for Dedndave code
------------------------------------------------------------------------
2285    cycles for Dedndave code
------------------------------------------------------------------------

And the EXE will then run on my P3 system:

Code: [Select]

------------------------------------------------------------------------
Pentium III

Instructions: MMX, SSE1
------------------------------------------------------------------------
2457    cycles for Dedndave code
------------------------------------------------------------------------
2437    cycles for Dedndave code
------------------------------------------------------------------------
2437    cycles for Dedndave code
------------------------------------------------------------------------
2439    cycles for Dedndave code
------------------------------------------------------------------------

jj2007 · « **Reply #31 on:** December 09, 2012, 01:21:46 PM »

That will be a close race between Dave and Frank...

Intel(R) Celeron(R) M CPU 420 @ 1.60GHz (SSE3)
loop overhead is approx. 189/100 cycles1556 cycles for 100 * FA Dave 541 cycles for 100 * FA Jochen 1557 cycles for 100 * FA Frank 1556 cycles for 100 * FA Dave 539 cycles for 100 * FA Jochen 1557 cycles for 100 * FA Frank 1558 cycles for 100 * FA Dave 542 cycles for 100 * FA Jochen 1564 cycles for 100 * FA Frank 230 bytes for FA Dave 281 bytes for FA Jochen 350 bytes for FA Frank

frktons · « **Reply #32 on:** December 09, 2012, 01:54:39 PM »

Quote from: jj2007 on December 09, 2012, 01:21:46 PM

That will be a close race between Dave and Frank...

Intel(R) Celeron(R) M CPU 420 @ 1.60GHz (SSE3)
loop overhead is approx. 189/100 cycles1556 cycles for 100 * FA Dave 541 cycles for 100 * FA Jochen 1557 cycles for 100 * FA Frank 1556 cycles for 100 * FA Dave 539 cycles for 100 * FA Jochen 1557 cycles for 100 * FA Frank 1558 cycles for 100 * FA Dave 542 cycles for 100 * FA Jochen 1564 cycles for 100 * FA Frank 230 bytes for FA Dave 281 bytes for FA Jochen 350 bytes for FA Frank

Actually I didn't write any code. I only posted Dave code in
another testbed. Well with your routine you demonstrated that
the good job of Dave could be optimized with better registers
and parallel computing.

dedndave · « **Reply #33 on:** December 09, 2012, 01:55:28 PM »

i don't think Frank has any working code, per se
looks like you got the speed thing :t
did you verify the resulting table ? :P

Code: [Select]

Intel(R) Pentium(R) 4 CPU 3.00GHz (SSE3)
++18 of 20 tests valid, loop overhead is approx. 246/100 cycles

2069    cycles for 100 * FA Dave
915     cycles for 100 * FA Jochen
2230    cycles for 100 * FA Frank

2029    cycles for 100 * FA Dave
913     cycles for 100 * FA Jochen
2592    cycles for 100 * FA Frank

2034    cycles for 100 * FA Dave
930     cycles for 100 * FA Jochen
2217    cycles for 100 * FA Frank

frktons · « **Reply #34 on:** December 09, 2012, 01:59:51 PM »

Jochen,
the source you posted is for MasmBasic editor, I suppose,
couldn't you post a normal ascii text file, in order to see what it
actually does?

Frank

dedndave · « **Reply #35 on:** December 09, 2012, 02:04:35 PM »

it's a "rich" text file, Frank
i think you can open it with WordPad, if you don't have anything else :t

dedndave · « **Reply #36 on:** December 09, 2012, 02:09:17 PM »

this is Jochen's data and code

Code: [Select]

align 16
TestB_s:
Src0123	db "   0   1   2   3"
Add4444	dd 04000000h, 04000000h, 04000000h, 04000000h	; xmm1
Add44xx	dd 04000000h, 04000000h, 0FA110000h, 0FA110000h	; xmm2
Addxx44	dd 0FA110000h, 0FA110000h, 04000000h, 04000000h	; xmm3
Addxxxx	dd 0FA010000h, 0FA010000h, 0FA010000h, 0FA010000h	; xmm4

Add244xx	dd 04000000h, 04000000h, 0FA010000h, 0FA010000h	; xmm2
Add2xx44	dd 0FA010000h, 0FA010000h, 04000000h, 04000000h	; xmm3
Add2xxxx	dd 0FA010000h, 0FA010000h, 0FA010000h, 0FA010000h	; xmm4

Sub100a	dd 00009EF00h, 00009EF00h, 00009EF00h, 00009EF00h
Sub100b	dd 00009FF00h, 00009FF00h, 00009FF00h, 00009FF00h

NameB equ FA Jochen	; assign a descriptive name here
TestB proc
	mov esi, offset Src0123
	mov edi, offset MyArray
	push edi
	xor ecx, ecx
	movaps xmm0, [esi]
	movaps xmm1, [esi+16]
	movaps xmm2, [esi+32]
	movaps xmm3, [esi+48]
	movaps xmm4, [esi+64]
	lea edx, [edi+4000]
	m2m ecx, -5
;	align 4
	.Repeat
		movaps [edi], xmm0
		paddd xmm0, xmm1	; 4444
		movaps [edi+16], xmm0
		paddd xmm0, xmm2	; 44xx
		movaps [edi+32], xmm0
		paddd xmm0, xmm3	; xx44
		movaps [edi+48], xmm0
		paddd xmm0, xmm1	; 4444
		movaps [edi+64], xmm0
		paddd xmm0, xmm4	; xxxx
		inc ecx
		.if Zero?
		    psubd xmm0, oword ptr Sub100a
		.elseif ecx==-4
			movaps xmm2, [esi+80]
			movaps xmm3, [esi+96]
			movaps xmm4, [esi+112]
		.elseif ecx==5
		    psubd xmm0, oword ptr Sub100b
		    xor ecx, ecx
		.endif
		add edi, 80
	.Until edi>=edx
	pop eax
  ret
TestB endp
TestB_endp:

jj2007 · « **Reply #37 on:** December 09, 2012, 06:26:42 PM »

Quote from: dedndave on December 09, 2012, 02:04:35 PM

it's a "rich" text file, Frank
i think you can open it with WordPad, if you don't have anything else :t

Frank does have something else: \Masm32\RichMasm\RichMasm.exe
Drag *.asc over RichMasm.exe, then click on the bookmarks "Test A" and "Test B" on the right :icon_mrgreen:

By the way: performance drops dramatically if you replace movaps with movups. But there is a well-known workaround, see attachment.

For MisAlign=0 and MisAlign=8, performance is almost like movaps (ca. 8 cycles slower, see below). For all other values, it is still almost 500 cycles faster than Dave's non-SSE code.

MisAlign=0:
Intel(R) Celeron(R) M CPU 420 @ 1.60GHz (SSE3)
loop overhead is approx. 189/100 cycles
1558 cycles for 100 * FA Dave 540 cycles for 100 * FA Jochen 547 cycles for 100 * FA Jochen unaligned 1557 cycles for 100 * FA Dave 539 cycles for 100 * FA Jochen 548 cycles for 100 * FA Jochen unaligned

MisAlign=3:
Intel(R) Celeron(R) M CPU 420 @ 1.60GHz (SSE3)
loop overhead is approx. 189/100 cycles
1845 cycles for 100 * FA Dave 1365 cycles for 100 * FA Jochen unaligned 1853 cycles for 100 * FA Dave 1362 cycles for 100 * FA Jochen unaligned

sinsi · « **Reply #38 on:** December 09, 2012, 06:44:29 PM »

I got times around 850 for my code (no sse) but half the size.
Here's your latest jj
AMD Phenom(tm) II X6 1100T Processor (SSE3) loop overhead is approx. 231/100 cycles 856 cycles for 100 * FA Dave 449 cycles for 100 * FA Jochen 436 cycles for 100 * FA Jochen unaligned 803 cycles for 100 * FA Dave 445 cycles for 100 * FA Jochen 435 cycles for 100 * FA Jochen unaligned 809 cycles for 100 * FA Dave 454 cycles for 100 * FA Jochen 433 cycles for 100 * FA Jochen unaligned 230 bytes for FA Dave 281 bytes for FA Jochen 141 bytes for FA Jochen unaligned

TouEnMasm · « **Reply #39 on:** December 09, 2012, 07:03:32 PM »

Here is a source code to verify mmx is faster:

Quote

   .data
   array dd 1000 dup (0)
   controle dd 0
   Nomfichier db "resultat.txt",0
   Hfile dd 0
   NumberOfBytesWritten dd 0
   retourligne db 13,10,0
   .code

   start:
   mov eax,30303020h ; 000
   mov edx,offset array
   unit:
   mov [edx],eax   ;000
   add eax,1000000h
   add edx,4
   mov [edx],eax   ;1
   add eax,1000000h
   add edx,4
   mov [edx],eax   ;2
   add eax,1000000h
   add edx,4
   mov [edx],eax   ;3
   add eax,1000000h
   add edx,4
   mov [edx],eax   ;4
   add eax,1000000h
   add edx,4
   mov [edx],eax   ;5
   add eax,1000000h
   add edx,4
   mov [edx],eax   ;6
   add eax,1000000h
   add edx,4
   mov [edx],eax   ;7
   add eax,1000000h
   add edx,4
   mov [edx],eax   ;8
   add eax,1000000h
   add edx,4
   mov [edx],eax   ;9
   ;-----------------------------
   sub eax,9000000h
   add eax,10000h
   add edx,4
   mov ecx,eax
   shr ecx,16
   .if cl == 3Ah ;1-
      sub eax,0A0000h
      inc ah   ;303030h + 10000h
      .if ah == 3Ah
         ;fini
         jmp fin
      .else
         jmp unit
      .endif
   .else
      jmp unit
   .endif
   fin:
   lea edx,controle   ;debug limit must be NULL
   lea edx,array   ;debug view what is in memory
   invoke CreateFile,addr Nomfichier,GENERIC_WRITE,NULL,\
         NULL,CREATE_ALWAYS,FILE_ATTRIBUTE_NORMAL,0
   mov Hfile,eax
   mov edx,offset array
   mov ecx,0
   ecrire:
   push edx
   push ecx
   invoke WriteFile,Hfile,edx,400,addr NumberOfBytesWritten,NULL
   invoke WriteFile,Hfile,addr retourligne,2,addr NumberOfBytesWritten,NULL
   pop ecx
   pop edx
   add edx,400
   add ecx,100
   .if ecx != 1000
      jmp ecrire
   .endif

   invoke CloseHandle,Hfile
   invoke ExitProcess,0
;################################################################
   end start

ascii table is put in the texte file 000 to 999

sinsi · « **Reply #40 on:** December 09, 2012, 07:15:06 PM »

116 bytes of slackware

Code: [Select]

TestA proc 
    pusha
    
    mov edi,offset MyArray
    mov ebp,00303030h
    mov ecx,10
l2: mov edx,10
    push ebp
l1: call proc0
    add ebp,00000100h
    sub edx,1
    jnz l1
    pop ebp
    inc ebp ;add ebp,00000001h
    sub ecx,1
    jnz l2
    popa
    mov eax,offset MyArray
    ret
proc0:
    push ecx
    push edx
    
    lea eax,[ebp]
    lea ebx,[ebp+00010000h]
    lea ecx,[ebp+00020000h]
    lea edx,[ebp+00030000h]
    lea esi,[ebp+00040000h]

    mov ebp,00050000h
    
    mov [edi],eax
    mov [edi+4],ebx
    mov [edi+8],ecx
    mov [edi+12],edx
    mov [edi+16],esi
    
    add eax,ebp
    add ebx,ebp
    add ecx,ebp
    add edx,ebp
    add esi,ebp

    mov [edi+20],eax
    mov [edi+24],ebx
    mov [edi+28],ecx
    mov [edi+32],edx
    mov [edi+36],esi
    
    add edi,40
    pop edx
    pop ecx
    ret
    


TestA endp

This would be nice to do in 64-bit with all those regs, but the xmm's are much better.

jj2007 · « **Reply #41 on:** December 09, 2012, 07:35:07 PM »

Quote from: ToutEnMasm on December 09, 2012, 07:03:32 PM

Here is a source code to verify mmx is faster:

Where's da mmx, Yves?

Quote from: sinsi on December 09, 2012, 07:15:06 PM

116 bytes of slackware
...
This would be nice to do in 64-bit with all those regs, but the xmm's are much better.

SSE2 is difficult to beat indeed...

Intel(R) Celeron(R) M CPU 420 @ 1.60GHz (SSE3)
loop overhead is approx. 189/100 cycles
1557 cycles for 100 * FA Dave 540 cycles for 100 * FA Jochen 1838 cycles for 100 * FA Sinsi 547 cycles for 100 * FA Jochen unaligned 1540 cycles for 100 * FA Yves

sinsi · « **Reply #42 on:** December 09, 2012, 08:05:45 PM »

Just goes to show, these tests are of academic interest only

AMD Phenom(tm) II X6 1100T Processor (SSE3) loop overhead is approx. 228/100 cycles 786 cycles for 100 * FA Dave 441 cycles for 100 * FA Jochen 849 cycles for 100 * FA Sinsi 419 cycles for 100 * FA Jochen unaligned 1021 cycles for 100 * FA Yves 786 cycles for 100 * FA Dave 437 cycles for 100 * FA Jochen 849 cycles for 100 * FA Sinsi 431 cycles for 100 * FA Jochen unaligned 1029 cycles for 100 * FA Yves 779 cycles for 100 * FA Dave 435 cycles for 100 * FA Jochen 846 cycles for 100 * FA Sinsi 417 cycles for 100 * FA Jochen unaligned 1023 cycles for 100 * FA Yves 230 bytes for FA Dave 281 bytes for FA Jochen 116 bytes for FA Sinsi 141 bytes for FA Jochen unaligned 141 bytes for FA Yves 4208864 = eax FA Dave 4208864 = eax FA Jochen 4208864 = eax FA Sinsi 4208864 = eax FA Jochen unaligned 4208864 = eax FA Yves

dedndave · « **Reply #43 on:** December 09, 2012, 08:11:22 PM »

it should generate leading-zero suppressed strings, right-justified in a field of 3 spaces

Intel(R) Pentium(R) 4 CPU 3.00GHz (SSE3) loop overhead is approx. 240/100 cycles 2040 cycles for 100 * FA Dave 919 cycles for 100 * FA Jochen 2609 cycles for 100 * FA Sinsi 1226 cycles for 100 * FA Jochen unaligned 1854 cycles for 100 * FA Yves 2040 cycles for 100 * FA Dave 943 cycles for 100 * FA Jochen 2606 cycles for 100 * FA Sinsi 1197 cycles for 100 * FA Jochen unaligned 1816 cycles for 100 * FA Yves 2037 cycles for 100 * FA Dave 920 cycles for 100 * FA Jochen 2602 cycles for 100 * FA Sinsi 1196 cycles for 100 * FA Jochen unaligned 1892 cycles for 100 * FA Yves

jj2007 · « **Reply #44 on:** December 09, 2012, 08:25:09 PM »

Quote from: sinsi on December 09, 2012, 08:05:45 PM

Just goes to show, these tests are of academic interest only

Absolutely

New version:

Intel(R) Celeron(R) M CPU 420 @ 1.60GHz (SSE3)
loop overhead is approx. 189/100 cycles
1556 cycles for 100 * FA Dave 1811 cycles for 100 * FA Sinsi 441 cycles for 100 * FA Jochen unaligned 1541 cycles for 100 * FA Yves

P.S.: If you don't agree with the suffix "FINAL", write a faster algo

The MASM Forum

News:

Author Topic: The fastest way to fill a dword array with string values (Read 74942 times)

MichaelW

Re: The fastest way to fill a dword array with string values

jj2007

Re: The fastest way to fill a dword array with string values

frktons

Re: The fastest way to fill a dword array with string values

dedndave

Re: The fastest way to fill a dword array with string values

frktons

Re: The fastest way to fill a dword array with string values

dedndave

Re: The fastest way to fill a dword array with string values

dedndave

Re: The fastest way to fill a dword array with string values

jj2007

Re: The fastest way to fill a dword array with string values

sinsi

Re: The fastest way to fill a dword array with string values

TouEnMasm

Re: The fastest way to fill a dword array with string values

sinsi

Re: The fastest way to fill a dword array with string values

jj2007

Re: The fastest way to fill a dword array with string values

sinsi

Re: The fastest way to fill a dword array with string values

dedndave

Re: The fastest way to fill a dword array with string values

jj2007

Re: The fastest way to fill a dword array with string values