home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
IRIS Development Option 6.2
/
IRIS_Development_Option_6.2_814-0478-001.iso
/
dist
/
gl_dev.idb
/
usr
/
relnotes
/
gl_dev
/
ch3.z
/
ch3
Wrap
Text File
|
1996-03-14
|
8KB
|
265 lines
- 1 -
3. _C_h_a_n_g_e_s__a_n_d__A_d_d_i_t_i_o_n_s
3.1 _C_h_a_n_g_e_s__a_n_d__A_d_d_i_t_i_o_n_s__f_o_r__I_R_I_X__6_._2
IRIX 6.2 adds both OpenGL and IrisGL library support for the
new n32 abi for all platforms including Onyx, Power Onyx and
Crimson. In addition, IRIX 6.2 adds Onyx to the previous
list (Power Onyx, Power Challenge) of platforms supporting
64 bit OpenGL. IrisGL is supported exclusively by the 32
bit ABIs.
IRIX 6.2 also adds support for the new InfiniteReality
graphics adaptor on Onyx systems. InfiniteReality offers a
large number of new features; see chapter 6 (OpenGL) for
more information.
IRIX 6.2 supports all Impact configurations: Max Impact,
High Impact, and Solid Impact. On Max Impact, High Impact,
and Solid Impact, the Pbuffer OpenGL extension is now
supported. See chapter 7 for details about the pbuffer
release on Impact. The Presenter display adapter/projector
is supported in this release as a new feature. In addition,
the OpenGL texture LOD extension is now supported. Some new
timing tables have also been added for impact: the
1600x1200 timing tables support a larger screen size on
appropriate monitors; the new High Impact 1024x768_60_pbuf,
1024x768_72_pbuf, and 1024x768_76_pbuf timing tables differ
from their preexisting counterparts (such as 1024x768_60) in
that they support faster triangle fill rates (in some
operations), and reserve one complete set of framebuffer
bitplanes for pbuffer use. For applications which require a
large number of pbuffers, the 1024x768_XX_pbuf timing tables
provide a new solution. Vertex rendering performance has
been increased substantially in this release by changing the
clipping algorithm.
IrisGL changes have been limited to bug fixes. For more
information, please see chapter 4.
3.2 _C_h_a_n_g_e_s__a_n_d__A_d_d_i_t_i_o_n_s__f_o_r__I_R_I_X__5_._3
OpenGL is now supported on all SGI graphics adaptors except
for the G, GT and GTX.
In this release, several OpenGL extensions have been added.
Most of them are supported on all graphics adaptors while
others are restricted to just RealityEngine, RealityEngine2,
and Impact. For full details on the OpenGL extensions see
chapter 6 of these release notes or the online manual pages
for _g_l_i_n_t_r_o and _g_l_x_i_n_t_r_o.
- 2 -
On Indy (XL), Indigo2 XL, Indigo Entry, and Indigo XS models
without hardware z buffers, an optimization has been added
to OpenGL to improve the speed of zbuffer clears. This
acceleration will occur if the depth function is GL_LESS or
GL_LEQUAL and the z clear value is 1.0 or if the depth
function is GL_GREATER or GL_GEQUAL and the z clear value is
0.0. The GLX visuals have been altered to a 24-bit deep z-
buffer (in 5.2 it was 32-bits). Users that had explicitly
asked for a 32-bit z-buffer will need to alter their code.
The decrease of z resolution is caused by additional
information that needs to be kept on a per-pixel basis.
Other optimizations for OpenGL on the Indy (XL), Indigo2 XL,
Indigo Entry, and Indigo XS models include speedup of z
buffered non-antialised points, faster implementation of
glPolygonOffsetSGI, faster mip-mapped textured polygons, and
faster anti-aliased polygons.
Optimizations for Iris GL on the Indy (XL) and Indigo2 XL
include faster clear times, and faster 2x2 pixel zooms for
lrectwrite.
The IrisGL changes have otherwise been limited to bug fixes.
See Chapter 4 for details.
3.3 _C_h_a_n_g_e_s__a_n_d__A_d_d_i_t_i_o_n_s__f_o_r__I_R_I_X__5_._2
The IrisGL now supports new image processing features on
RealityEngine:
+o Calls to allocate framebuffer space (the ILbuffer) to
be used as image storage/cache and use it as the source
or destination of drawing and pixel transfers. Note
that this is _n_o_t exactly the same as the AUXbuffers in
OpenGL. See _i_l_b_u_f_f_e_r.
+o Modifications to texture loading. Texture memory
represents a warpable, wrappable image store with very
fast fill and resampling (image rotate, zoom, and copy)
rates. The modifications are to allow:
- raster-order load of image data to texture memory
with most _p_i_x_m_o_d_e and pixel transfer operations.
- loading subimages into texture memory. This is
the basic tool that, along with texture coordinate
wrapping, allows fast panning over a large image.
See _s_u_b_t_e_x_l_o_a_d and _f_b_s_u_b_t_e_x_l_o_a_d.
The source for the image data can be either host memory
or the frame buffer.
- 3 -
+o Operators on pixel transfers (_l_r_e_c_t_w_r_i_t_e, _r_e_c_t_c_o_p_y,
_l_r_e_c_t_r_e_a_d). The operations are:
- Convolve 3x3, 5x5, 7x7, separable and general. See
_c_o_n_v_o_l_v_e.
- Lookup tables. See _t_l_u_t_b_i_n_d.
- image statistics: _h_g_r_a_m and _m_i_n_m_a_x
+o The IrisGL can now transfer a single component of a
framebuffer (ABGR) image so that image storage is
extended by a factor of 4.
+o Several OpenGL-like commands were added to _p_i_x_m_o_d_e,
primarily _C_P_M__I_N_P_U_T__T_Y_P_E, _P_M__I_N_P_U_T__F_O_R_M_A_T,
_P_M__O_U_T_P_U_T__T_Y_P_E, _P_M__O_U_T_P_U_T__F_O_R_M_A_T.
Several IrisGL man pages are new or were changed in 5.1 or
5.0.1:
_c_o_n_v_o_l_v_e _i_l_d_r_a_w _r_e_a_d_c_o_m_p_o_n_e_n_t
_f_b_s_u_b_t_e_x_l_o_a_d _i_s_t_e_x_l_o_a_d_e_d _r_e_a_d_s_o_u_r_c_e
_g_e_t_h_g_r_a_m _m_i_n_m_a_x _s_u_b_t_e_x_l_o_a_d
_g_e_t_m_i_n_m_a_x _p_i_x_e_l_m_a_p _t_e_x_b_i_n_d
_h_g_r_a_m _p_i_x_e_l_t_r_a_n_s_f_e_r _t_e_x_d_e_f
_i_l_b_u_f_f_e_r _p_i_x_m_o_d_e _t_e_x_d_e_f_2_d
+o Using _l_i_b_s_p_h_e_r_e, RealityEngine systems can now draw
high-speed ``bitmap'' spheres. They are Phong-shaded
and are limited only by the back-end fill rate. There
are some limitations, but they now work in perspective
as well.
+o On RealityEngine systems, pixmode(PM_SIZE,64) is now
fully supported and documented.
+o On RealityEngine systems, non-multisample primitive
performance has been improved.
+o On RealityEngine systems, the Multi-Channel Option now
supports cursors on multiple screens.
+o Field-sequential RGB format requires the IMP7 ASIC on
the RM4 board. Use /usr/gfx/gfxinfo -v as root and
verify that the last digit of the RM version number is
``2'' or more.
+o On RealityEngine systems, performance was enhanced for
pixel transfer operations using some pixmode features
(PM_ADD24, PM_SHIFT, etc).
- 4 -
+o On RealityEngine systems, performance was enhanced for
mode changing operations such as _t_e_x_b_i_n_d and _l_m_b_i_n_d.