BorlandTalk.com Borland discussion newsgroups   Archives   FAQ   Search   Memberlist   Usergroups   Register   Profile   Log in to check your private messages   Log in  Alphablending Goto page 1, 2, 3  Next          BorlandTalk.com Forum Index -> Delphi Graphics View previous topic :: View next topic   Author Message Roy M Klever Guest Posted: Mon Oct 16, 2006 8:10 pm    Post subject: Alphablending Hi, This is what I use today. // Explanation slSize = ScanLine Size slMain = First ScanLine of bitmap pRGB24 = Is a pointer for a: r,g,b byte array by, bdy, bx, bdy is blending area FColorSelection is the color to blend in // // Routine row := slSize * by + slMain; for y := by to bdy do begin for x := bx to bdx do begin slPnt:= pRGB24(row + (x * 3)); slPnt.r := (FColorSelection.R - slPnt.r) shr 3 + slPnt.r; slPnt.g := (FColorSelection.G - slPnt.g) shr 3 + slPnt.g; slPnt.b := (FColorSelection.B - slPnt.b) shr 3 + slPnt.b; end; inc(row, slSize); end; Do you have any idea for a faster way of doing it. Must be for 24 bit bitmaps not 32 bits. This is for alphablending selection in a own component. Kind Regards Roy Magne Klever Back to top Michael Hansen Guest Posted: Mon Oct 16, 2006 10:14 pm    Post subject: Re: Alphablending Hi Roy, This looks pretty optimal, but switching to MMX could perhaps give a little more (the 24 bit structure may be a problem though). One other thing, you could try shifting the channels instead of using the record approach (IIRC there is a small difference in favour of shifting). Also if you could settle for another blendbalance, you could just do an average of the two colors. Regards, Michael Hansen Back to top Mike Lischke Guest Posted: Mon Oct 16, 2006 11:54 pm    Post subject: Re: Alphablending Roy M Klever wrote: Quote: Do you have any idea for a faster way of doing it. Must be for 24 bit bitmaps not 32 bits. If you want it fast then switch to 32 bits. 24bit requires unaligned access which has a big penalty (particularly with modern CPUs). With 32bit you could also use MMX and speed up things significantly. I could even give you ready-to-use code for that. Mike -- www.soft-gems.net Back to top Lord Crc Guest Posted: Tue Oct 17, 2006 12:32 am    Post subject: Re: Alphablending On 16 Oct 2006 11:54:22 -0700, "Mike Lischke" <newsgroups@soft-gems.net> wrote: Quote: If you want it fast then switch to 32 bits. 24bit requires unaligned access which has a big penalty (particularly with modern CPUs). With 32bit you could also use MMX and speed up things significantly. I could even give you ready-to-use code for that. When using plain delphi code, I found some time ago (on my P3) that using 32bit in general was slightly slower actually. I attributed it to the increased size of the image, causing more cache misses. I guess I should try it again tho, to see how things has changed :) - Asbjørn Back to top Mattias Andersson Guest Posted: Tue Oct 17, 2006 12:38 am    Post subject: Re: Alphablending Roy M Klever wrote: Quote: slPnt.r := (FColorSelection.R - slPnt.r) shr 3 + slPnt.r; slPnt.g := (FColorSelection.G - slPnt.g) shr 3 + slPnt.g; slPnt.b := (FColorSelection.B - slPnt.b) shr 3 + slPnt.b; [snip] Quote: Do you have any idea for a faster way of doing it. Must be for 24 bit bitmaps not 32 bits. You could replace the above code with this (F = FColorSelection, B = slPnt): B := (F shr 3) and $1F1F1F + (B - (B shr 3) and $1F1F1F); Otherwise I think you should consider Mike's suggestions; accessing aligned 32bit data will be much faster. Cheers, Mattias Back to top Lord Crc Guest Posted: Tue Oct 17, 2006 1:04 am    Post subject: Re: Alphablending On Mon, 16 Oct 2006 17:10:06 +0200, Roy M Klever <roy_m_klever (AT) rmklever (DOT) com> wrote: Quote: // Routine row := slSize * by + slMain; for y := by to bdy do begin slPnt:= pRGB24(row); for x := bx to bdx do begin slPnt.r := (FColorSelection.R - slPnt.r) shr 3 + slPnt.r; slPnt.g := (FColorSelection.G - slPnt.g) shr 3 + slPnt.g; slPnt.b := (FColorSelection.B - slPnt.b) shr 3 + slPnt.b; inc(slPnt); // or inc(PChar(slPnt), 3); end; inc(row, slSize); end; Do you have any idea for a faster way of doing it. Must be for 24 bit bitmaps not 32 bits. That's about as fast as it gets on my box. It uses about 220ms on a 4120x8486 bitmap (largest I had handy on my AMD X2 3800+, and <10ms for a 1200x800 bitmap. Perhaps it would be slightly faster to copy FColorSelection to a local variable, that's what I used. - Asbjørn Back to top Roy M Klever Guest Posted: Tue Oct 17, 2006 3:06 am    Post subject: Re: Alphablending Hi Mike, Quote: If you want it fast then switch to 32 bits. 24bit requires unaligned access which has a big penalty (particularly with modern CPUs). With 32bit you could also use MMX and speed up things significantly. I could even give you ready-to-use code for that. Well just switching to 32 bit made it a lot slower, I would say half speed. Just switching is not enough, maybe using MMX is faster, not tested. But thanks for your suggestion. Kind Regards Roy Magne Klever Back to top Roy M Klever Guest Posted: Tue Oct 17, 2006 3:15 am    Post subject: Re: Alphablending Hi Michael, Thanks for your suggestion. Kind Regards Roy Magne Klever Back to top Roy M Klever Guest Posted: Tue Oct 17, 2006 3:26 am    Post subject: Re: Alphablending Hi Lord Crc, Quote: That's about as fast as it gets on my box. It uses about 220ms on a 4120x8486 bitmap (largest I had handy on my AMD X2 3800+, and <10ms for a 1200x800 bitmap. Not to bad I think :) Quote: Perhaps it would be slightly faster to copy FColorSelection to a local variable, that's what I used. Well, not very much of a diffrence, but thanks for suggestion. When using a paintbrush routine I found it to be a lot faster to change to globale vars instead of local vars. Kind Regards Roy Magne Klever Back to top Roy M Klever Guest Posted: Tue Oct 17, 2006 3:33 am    Post subject: Re: Alphablending Hi Mattias, Quote: B := (F shr 3) and $1F1F1F + (B - (B shr 3) and $1F1F1F); I could not make it work... but thanks for suggestion. Kind Regards Roy Magne Klever Back to top Mattias Andersson Guest Posted: Tue Oct 17, 2006 5:08 am    Post subject: Re: Alphablending Roy M Klever wrote: Quote: Hi Mattias, B := (F shr 3) and $1F1F1F + (B - (B shr 3) and $1F1F1F); I could not make it work... but thanks for suggestion. I've attached a small example routine below. It works by first storing the 24bit foreground color in a temporary scanline and then blending this scanline using a 32bit access pattern. I think this should be more efficient than computing each RGB value separately. HTH, Mattias procedure BlendToBitmap(Bitmap: TBitmap; Color: TRGB24); var I, J, DWords: Integer; F, B: PInteger; P, TmpLine: PRGB24; begin DWords := Ceil(Bitmap.Width * 3/4); GetMem(TmpLine, DWords * 4); P := TmpLine; for I := 0 to Bitmap.Width - 1 do begin P^ := Color; Inc(P); end; for J := 0 to Bitmap.Height - 1 do begin F := PInteger(TmpLine); B := Bitmap.ScanLine[J]; for I := 0 to DWords - 1 do begin B^ := (F^ shr 3) and $1F1F1F1F + (B^ - (B^ shr 3) and $1F1F1F1F); Inc(F); Inc(B); end; end; FreeMem(TmpLine); end; Back to top Roy M Klever Guest Posted: Tue Oct 17, 2006 6:31 pm    Post subject: Re: Alphablending Hi Mattias, Quote: After my last post, I realized that this could be done much more efficiently using MMX. I've appended a new MMX routine below, together with the old pascal routines. Excellent will be testing it this afternoon :) Thanks! Kind Regards Roy Magne Klever Back to top Nils Haeck Guest Posted: Tue Oct 17, 2006 8:20 pm    Post subject: Re: Alphablending why not use var TmpLine: array of integer; ... SetLength(TmpLine, Bitmap.Width) ... F := @TmpLine[0]; The "getmem/freemem" without "try/finally" looks fishy to me Using a dynamic array is safer imo, and more modern. Other than that, your suggestion is pretty good, and how it should be done :) Another small optimization can be reached by getting rid of a call to ScanLine for each y-line, and just do it for y=0, y=1 then calculate the delta. Nils Back to top Mattias Andersson Guest Posted: Tue Oct 17, 2006 9:28 pm    Post subject: Re: Alphablending Nils Haeck wrote: Quote: why not use var TmpLine: array of integer; .. SetLength(TmpLine, Bitmap.Width) .. F := @TmpLine[0]; The "getmem/freemem" without "try/finally" looks fishy to me Using a dynamic array is safer imo, and more modern. Ah, yes, a try/finally block would definitely be in place. Seems I put this together in a haste.. I guess it's debatable which one is better -- GetMem/FreeMem or SetLength. If you're only passing the data as an argument to an assembler routine, then I think it doesn't really matter which one you use. However, from a performance perspective, I wouldn't use either of these. For a temporary buffer like this, you could simply allocate memory on the stack (which will avoid the penalty of the memory manager). See e.g. StackAlloc and StackFree in Grids.pas. Quote: Another small optimization can be reached by getting rid of a call to ScanLine for each y-line, and just do it for y=0, y=1 then calculate the delta. Yes, but as you can see from my benchmark results, in order to improve performance you should always focus on the innermost loops. Optimizing the outer loop will give a noticeable performance gain only for small data blocks -- and typically, if the data is small, the routine will be fast enough anyway. :) Mattias Back to top Eric Grange Guest Posted: Tue Oct 17, 2006 9:51 pm    Post subject: Re: Alphablending Quote: BlendLine(TmpLine, Bitmap.ScanLine[I], Bitmap.Width * 3); You'll likely gain some more in the MMX variant if you replace Bitmap.ScanLine[I] with (typecasts left out for clarity) scanLine := Bitmap.ScanLine[0]; scanLineDelta := Bitmap.ScanLine[1]-scanLine; and in the loop Inc(scanLine, scanLineDelta); Eric Back to top Display posts from previous: All Posts1 Day7 Days2 Weeks1 Month3 Months6 Months1 Year Oldest FirstNewest First         BorlandTalk.com Forum Index -> Delphi Graphics All times are GMT Goto page 1, 2, 3  Next Page 1 of 3   Jump to: Select a forum C++ Builder Newsgroups----------------C++ BuilderC++ Builder (ActiveX)C++ Builder (Announcements)C++ Builder (Command Line Tools)C++ Builder DatabasesC++ Builder Databases (ADO)C++ Builder Databases (Desktop)C++ Builder Databases (InterBase Express)C++ Builder Databases (Multi-Tier)C++ Builder Databases (SQL Servers)C++ Builder (Graphics)C++ Builder (IDE)C++ Builder (Internet)C++ Builder (Internet Socket)C++ Builder (Internet Web)C++ Builder (Jobs)C++ Builder (Language)C++ Builder (Language C++)C++ Builder (Compatibility with MFC)C++ Builder (Native API)C++ Builder (Non-Technical)C++ Builder (Students)C++ Builder (Third Party Tools)C++ Builder (Upgrade)C++ Builder (VCL)C++ Builder (VCL Components Usage)C++ Builder (VCL Components Development)C++ Builder (VCL Usage)C++ Builder (WinAPI)C++ Builder Installation Delphi Newsgroups----------------Delphi (General)Delphi ActiveX Controls UsageDelphi ActiveX Controls DevelopmentDelphi As400Delphi DatabasesDelphi Databases (ADO)Delphi Databases (dbExpress)Delphi Databases (Desktop)Delphi Databases (InterBase Express)Delphi Databases (Multi-Tier)Delphi Databases (SQL Servers)Delphi DeploymentDelphi GraphicsDelphi IDEDelphi InternationalizationDelphi InternetDelphi Internet Isapi-WebBrokerDelphi Internet WebSnapDelphi Internet WinsockDelphi JobsDelphi Language BASMDelphi Language ObjectPascalDelphi Model Driven Architecture (General)Delphi Model Driven Architecture (Third Party)Delphi MultimediaDelphi Native APIDelphi Non-TechnicalDelphi ObjectPascalDelphi OLE AutomationDelphi OO designDelphi OpenToolsAPIDelphi Reporting-ChartingDelphi RTLDelphi StudentsDelphi TeamSourceDelphi Thirdparty ToolsDelphi Thirdparty Tools (General)Delphi Thirdparty Tools (IntraWeb)Delphi Thirdparty Tools (RaveReports)Delphi UpgradeDelphi VCLDelphi VCL ComponentsDelphi VCL Components UsingDelphi VCL Components WritingDelphi WebServices SOAPDelphi WinAPIDelphi XMLDelphi Installationcomp.lang.pascal.delphi.components.misccomp.lang.pascal.delphi.components.usagecomp.lang.pascal.delphi.components.writingcomp.lang.pascal.delphi.databasescomp.lang.pascal.delphi.misccomp.sources.delphi JBuilder Newsgroups----------------JBuilderJBuilder AnnouncementsJBuilder Applet IssuesJBuilder As400JBuilder CompilerJBuilder CORBA/RMIJBuilder DatabasesJBuilder DB SwingJBuilder DebuggerJBuilder DeploymentJBuilder DocumentationJBuilder EnterpriseJBuilder HandheldJBuilder IDEJBuilder JavaJBuilder JavaAPIJBuilder Java LanguageJBuilder Java SwingJBuilder Java BeansJBuilder Java Beans UsageJBuilder Java Beans DevelopmentJBuilder JBCLJBuilder JDataStoreJBuilder JobsJBuilder LinuxJBuilder MacintoshJBuilder & MobileSetJBuilder MultiLingual AppsJBuilder Non-TechnicalJBuilder OpenToolsJBuilder Servlets (JSP)JBuilder StudentsJBuilder Team DevelopmentJBuilder Thirdparty ToolsJBuilder UpgradeJBuilder Installation InterBase Newsgroups----------------InterBaseInterBase Bug DiscussionInterBase GeneralInterBase IBConsoleInterBase OpenSourceInterBase ReplicationInterBase SQLInterBase ThirdpartyInterBase Installation Miscellaneous Borland Newsgroups----------------AppCenterAppServerEnteraVisiBroker  You cannot post new topics in this forum You cannot reply to topics in this forum You cannot edit your posts in this forum You cannot delete your posts in this forum You cannot vote in polls in this forum Powered by phpBB © 2001, 2006 phpBB Group SEO toolkit © 2004-2006 webmedic.