In article <9eemm8$a26$1@news.inet.tele.dk>, LP@BONUS.DK says...
> Alt er nu forsøgt uden held - også Detonator 11., så det er godnat for nu.
>
Hvis du ikke har opgivet og stadig har energi til at pille og god tid, kan du jo prøve nogle af
nedenstående kombinationer hvor du så kører efter laveste fælles nævner.
Bemærk endvidere at Workarounds settings fungerer som akumulerede bit-værdier
dvs. Workaround = 20001 er summen af
0: 0x00000001: treat implementation as OpenGL version 1.0 even though it may claim to be 1.1
og
17: 0x00020000: don't use stippling in combination with texturing (incl. textured text)
Du kan også køre software emulering. Du skal så blot afvikle følgende fil som du kan finde på
Inventor CD'en: "hardwareLibrary_No_Accel.reg"
Herefter går det langsomt men sikkert
Som du sikert nu fornemmer har Inventor en yderst avanceret grafikdel, med ret omfattende
justeringsmuligheder, der korrekt justeret kan trække alt ud af et grafikkort.
mvh Lars Sandberg
Her er listen.
R4 cataloged workarounds:
0: 0x00000001: treat implementation as OpenGL version 1.0 even though it may claim to be 1.1
1: 0x00000002: Matrox Millennium single-pixel wide unpatterned lines with non-source raster op
2: 0x00000004: Matrox Millennium drawing single-pixel wide unpatterned lines with
glDrawElements
drops some line segments
3: 0x00000008: workaround for Oxygen 202 v2.8 bug turning off stippling
4: 0x00000010: force setting of modeling matrix even when it seems redundant (is this used?)
5: 0x00000020: Permedia 2 bug that spuriously connects disjoint line segments
6: 0x00000040: don't use GL_POLYGON_OFFSET_FILL -- crash on FireGL1000Pro driver v244 et. al.
7: 0x00000080: don't use glPolygonStipple (screen door transparency) -- crash on Matrox
Millennium
8: 0x00000100: don't use GL_POLYGON_OFFSET_FILL in combination with culling turned off
(FireGL1000Pro)
9: 0x00000200: don't use vertex array constructs (OpenGL 1.1 only)
10: 0x00000400: don't use back-face culling (problems on TNT2?)
11: 0x00000800: use anti-aliased lines by default
12: 0x00001000: implement non-lighted Colors as emissive Materials (for HP Visualizer?)
13: 0x00002000: don't release GLRC's when dtoring DrawTarget(!), for 3Dlabs VX-1 "share-with"
bug
14: 0x00004000: don't use bound textures
15: 0x00008000: don't use stencil buffer (not implemented [set internally] or not safe)
16: 0x00010000: okay to use stencil buffer when drawing to front buffer
17: 0x00020000: don't use stippling in combination with texturing (incl. textured text)
[Nvidia]
18: 0x00040000: don't use LazyAttributeDrawTarget [E&S]
19: 0x00080000: clear stencil buffer each frame (in BeginDraw.)If a card has 80000, try 40000
instead.
20: 0x00100000: okay to use EXT_compiled_vertex_array extension if present
21: 0x00200000: okay to use NV_vertex_array_range extension if present
22: 0x00400000: don't free offscreen regions -- just leak them (!) (ELSA crash)
23: 0x00800000: add a little jitter to slicing planes to avoid placing them exactly on vertices
(Permedia2)
24: 0x01000000: call CleanUpNearlyAlignedMatrix on all sets of Modeling Matrix (crash in
Generic OpenGL driver on NT)
30: 0x40000000: DO assert on OpenGL errors
31: 0x80000000: don't look for vendor-renderer-version specific settings and workarounds in
registry
New in R4SP1, R5:
25: 0x02000000: do try/except around image calls when drawing 2D marker (ATI) DID #276460 and
276762
26: 0x04000000: do simulated SwapType 2 flush the old way (glFlush) rather than the new
(glFinish) way
Note on WA 24:
See bug 265629: crash in GDI Generic driver during IDBuffer rendering for Drawings
Appears to be a float/double precision problem with numbers in modelview matrix
near the smallest resolvable unit for doubles (1e-16)
See also bug 263811, in which
a look-at of a certain face on a certain part causes a crash soon afterward
in the graphics card driver code.
device to model transforms hosed after LookAt command --
near zero terms in the viewing matrix cause gluUnproject to return nonsense!
Currently available swap strategies:
0: use strategy 1 if driver says it does swap-copy, else use strategy 3.
1: force use of regular Windows SwapBuffers call and assume back buffer remains valid after
swap
2: simulate copy-swap using glCopyPixels
3: draw all graphics in back buffer and swap to front
4: do save/restore of back buffer using KTX_buffer_region extension if available; otherwise
not recommended