assembly - memcopy so slow... what's wrong?

Question about which tools to use, bugs, the best way to implement a function, etc should go here. Don't forget to see if your question is answered in the wiki first! When in doubt post here.
Post Reply
User avatar
CorruptedByCPU
Member
Member
Posts: 79
Joined: Tue Feb 11, 2014 4:59 pm

assembly - memcopy so slow... what's wrong?

Post by CorruptedByCPU »

I don't know what is wrong with this code:

Processor AMD Phenom x4 965 3.4 GHz
... it's look same on AMD C-60 1 GHz

https://youtu.be/g1oFKPzk5ec
https://www.youtube.com/watch?v=g1oFKPzk5ec[/youtube]

Code: Select all

	; http://forum.4programmers.net/Off-Topic/250332-cyjon_os?p=1268711#id1268711
	shr	rcx,	STATIC_DIVIDE_BY_256

.loop:
	prefetchnta	[rsi + 256]
	prefetchnta	[rsi + 288]
	prefetchnta	[rsi + 320]
	prefetchnta	[rsi + 352]
	prefetchnta	[rsi + 384]
	prefetchnta	[rsi + 416]
	prefetchnta	[rsi + 448]
	prefetchnta	[rsi + 480]

	movdqa	xmm0,	[rsi]
	movdqa	xmm1,	[rsi + 0x10]
	movdqa	xmm2,	[rsi + 0x20]
	movdqa	xmm3,	[rsi + 0x30]
	movdqa	xmm4,	[rsi + 0x40]
	movdqa	xmm5,	[rsi + 0x50]
	movdqa	xmm6,	[rsi + 0x60]
	movdqa	xmm7,	[rsi + 0x70]
	movdqa	xmm8,	[rsi + 0x80]
	movdqa	xmm9,	[rsi + 0x90]
	movdqa	xmm10,	[rsi + 0xA0]
	movdqa	xmm11,	[rsi + 0xB0]
	movdqa	xmm12,	[rsi + 0xC0]
	movdqa	xmm13,	[rsi + 0xD0]
	movdqa	xmm14,	[rsi + 0xE0]
	movdqa	xmm15,	[rsi + 0xF0]

	movntdq	[rdi],	xmm0
	movntdq	[rdi + 0x10],	xmm1
	movntdq	[rdi + 0x20],	xmm2
	movntdq	[rdi + 0x30],	xmm3
	movntdq	[rdi + 0x40],	xmm4
	movntdq	[rdi + 0x50],	xmm5
	movntdq	[rdi + 0x60],	xmm6
	movntdq	[rdi + 0x70],	xmm7
	movntdq	[rdi + 0x80],	xmm8
	movntdq	[rdi + 0x90],	xmm9
	movntdq	[rdi + 0xA0],	xmm10
	movntdq	[rdi + 0xB0],	xmm11
	movntdq	[rdi + 0xC0],	xmm12
	movntdq	[rdi + 0xD0],	xmm13
	movntdq	[rdi + 0xE0],	xmm14
	movntdq	[rdi + 0xF0],	xmm15

	add	rsi,	256
	add	rdi,	256
	dec	rcx
	jnz	.loop
https://blackdev.org/ - system programming, my own 64 bit kernel and software.
Top
Octocontrabass
Member
Member
Posts: 5587
Joined: Mon Mar 25, 2013 7:01 pm

Re: assembly - memcopy so slow... what's wrong?

Post by Octocontrabass »

Where are you copying from?
Top
User avatar
CorruptedByCPU
Member
Member
Posts: 79
Joined: Tue Feb 11, 2014 4:59 pm

Re: assembly - memcopy so slow... what's wrong?

Post by CorruptedByCPU »

Octocontrabass wrote:Where are you copying from?
Video memory at 0xE0000000 from first line to ZERO line (37 lines all)

Resolution 800x600@32 (512 Bytes per char)
https://blackdev.org/ - system programming, my own 64 bit kernel and software.
Top
Octocontrabass
Member
Member
Posts: 5587
Joined: Mon Mar 25, 2013 7:01 pm

Re: assembly - memcopy so slow... what's wrong?

Post by Octocontrabass »

Reading from video memory is very slow. You'll have to find a way to scroll the screen without ever reading video memory.

One easy option is to store an entire copy of the screen in main memory, and then copy it to video memory whenever you update it.
Top
User avatar
CorruptedByCPU
Member
Member
Posts: 79
Joined: Tue Feb 11, 2014 4:59 pm

Re: assembly - memcopy so slow... what's wrong?

Post by CorruptedByCPU »

Octocontrabass wrote:Reading from video memory is very slow. You'll have to find a way to scroll the screen without ever reading video memory.

One easy option is to store an entire copy of the screen in main memory, and then copy it to video memory whenever you update it.
Thank you for answer I will try it :)
https://blackdev.org/ - system programming, my own 64 bit kernel and software.
Top
Post Reply

Return to “OS Development”