home *** CD-ROM | disk | FTP | other *** search
- Path: news.eunet.fi!fipnet!kone!jsaarinen
- Newsgroups: comp.sys.amiga.programmer
- X-NewsReader: IntuiNews 1.2b (31.7.94)
- From: "Jyrki Saarinen" <jsaarinen@kone.fipnet.fi>
- Date: Sun, 18 Feb 96 12:45:14 UT
- Comments: Illegal date header - new date added by quicknews
- X-Original-Date: Sun, 18 Feb 96 13:43:56
- MIME-Version: 1.0
- Content-Type: text/plain; charset=iso-8859-1
- Content-Transfer-Encoding: binary
- Subject: Texture/Gouraud innerloop speedtests
- Message-ID: <38232464@kone.fipnet.fi>
-
-
- Ok, I did a little research. My CPU is a 40MHz 68040,
- a Warp Engine with a very fast memory system, maybe
- this is the reason I did not gain any speed even if
- I turned the data cache and thus data burst off,
- with data burst everything was about 50% slower.
-
- Here is the Texture/Gouraud/Shading table innerloop,
- properly pipelined. The result is 18.81 ms drawing
- 65536 pixels. This one uses 64k aligned shading table
- and texture. As an interesting note, the fastest plain
- mapping loop drew those 65536 pixels in ~12 ms.
-
- So the frame rates were for a 320x256 screen:
- Texture/Gouraud/Shading table, 64k aligned: ~43 fps
- Plain Texture, 64k aligned: ~67 fps
-
- poly REPT 16
- move.w d3,d0
- move.w a1,d1
- move.b d4,d0
- addx.l d5,d3
- move.l d0,a3
- addx.l d6,d4
- move.b (a3),d1
- move.l d1,a3
- add.l a2,a1
- move.b (a3),(a0)+
- ENDR
- dbf d7,poly
- rts
-
- The places were schedeling was most effective were the
- move.l d0,a3
- <something here is a must>
- move.b (a3),d1
- <or>
- move.b (a3),(a0)+
-
- Now, the "normal version" took 23.94 ms to draw the same
- 65536 pixels. Quite a big difference to me..
- I tried to schedule it, but nothing could change the
- 23.94 ms result!
-
- poly REPT 16
- move.l d3,d0
- move.b d4,d0
- move.l a1,d1
- move.b (a3,d0.l),d1
- move.b (a4,d1.l),(a0)+
- addx.l d5,d3
- addx.l d6,d4
- add.l a2,a1
- ENDR
- dbf d7,poly
- rts
-
- -- _
- a Stellar programmer _ //
- "Amiga - back for the future" \X/
-