Parameters: Maximum Voxel algorithm
===============================================================================
		Timing tests with balls.kf

phoenix: compile flags: gcc -fpcc-struct-return -O
1 CPU: 10.6 s (real)

deepthought: gcc -fpcc-struct-return -O
2 90 MHz HyperSPARC CPUs:
  Job split into 8 pieces, main thread helps: 5.1 s (real)
  Job split into 8 pieces, main thread idle:  4.1 s (real)
  Job split per line, main thread helps:      4.9 s (real)
  Job split per line, main thread idle:       4.8 s (real)
  Single job, one thread (one CPU):           7.5 s (real)


===============================================================================
	Timing tests with tile1_balls.kf (80x300x300) (bottom tile: 10x10x10)

Permanent algorithmic changes:

Added offset (0.01) only once per ray.				11-JUN-1995
Wireframe computation disabled for most shaders			12-JUN-1995
Split shaders into codesets					18-JUN-1995

Temporary algorithmic changes:
Compute _sh_t by only incrementing (called HACK2)		12-JUN-1995
Compute _sh_x and _sh_y only by incrementing (called HACK3)	12-JUN-1993
-------------------------------------------------------------------------------
Platform: Sun 4/670 model 51

gcc -fpcc-struct-return -O
1 CPU: 8.5 s (real)

gcc -fpcc-struct-return -O2					28-FEB-1995
1 CPU: 7.2 s (real)

gcc -fpcc-struct-return -O2					1-APR-1995
1 CPU: 7.1 s (real)

acc -fast							30-APR-1995
1 CPU: 6.6 s (real)

As above but compile  raycast_3d.c  with: acc -O2		30-APR-1995
1 CPU: 7.1 s (real)
This tells me that GCC did just as well as acc for the shaders.

acc -O2								30-APR-1995
1 CPU: 8.8 s (real)

acc -O2 -cg92							30-APR-1995
1 CPU: 8.8 s (real)

acc -O2 -cg92 -fnonstd						30-APR-1995
1 CPU: 8.8 s (real)

acc -O2 -cg89							30-APR-1995
1 CPU: 8.8 s (real)

acc -O2 -cg89 -fnonstd -dalign					30-APR-1995
1 CPU: 9.0 s (real)
-------------------------------------------------------------------------------
Platform: Sun 4/670 4xHyperSPARC 90 SunOS 4.1.3

gcc -fpcc-struct-return -O2					15-MAY-1995
  Single job, one process (one CPU):		4.6 s (real)
  Job split into 16 pieces, GUI process idle:	1.4 s (real)

-------------------------------------------------------------------------------
Platform: Sun 4/670 4xHyperSPARC 90 Solaris 2.4

gcc -fpcc-struct-return -O2 (Solaris 2.3)			12-JUN-1995
Wireframe computation disabled for most shaders
Added offset (0.01) only once per ray.
Pass array of plane pointers to shaders rather than cube and plane offsets.
1 CPU  used: 4.2 s (real)	verified 13-JUN-1995 after bug fix
2 CPUs used: 2.1 s (real)
4 CPUs used: 1.2 s (real)

gcc -fpcc-struct-return -O2 (Solaris 2.3)			13-JUN-1995
Wireframe computation disabled for most shaders
Added offset (0.01) twice per loop iteration.
Pass array of plane pointers to shaders rather than cube and plane offsets.
1 CPU  used: 4.4 s (real)

gcc -fpcc-struct-return -O2 -DHACK3 (Solaris 2.3)		13-JUN-1995
Wireframe computation disabled for most shaders
Pass array of plane pointers to shaders rather than cube and plane offsets.
1 CPU  used: 2.8 s (real)	(35 cycles/voxel)
2 CPUs used: 1.4 s (real)
4 CPUs used: 0.8 s (real)

gcc -fpcc-struct-return -O2 -DHACK1 -DHACK3 (Solaris 2.3)	13-JUN-1995
Wireframe computation disabled for most shaders
Pass array of plane pointers to shaders rather than cube and plane offsets.
1 CPU  used: 2.9 s (real)	(36 cycles/voxel)
2 CPUs used: 1.4 s (real)
4 CPUs used: 0.8 s (real)

gcc -fpcc-struct-return -O2 -DHACK3 (Solaris 2.3)		14-JUN-1995
Wireframe computation disabled for most shaders
All memory references in inner loop removed.
1 CPU: 1.5 s (real)		(19 cycles/voxel)	THEORETICAL CPU LIMIT

gcc -fpcc-struct-return -O2 -DHACK3 (Solaris 2.3)		14-JUN-1995
Wireframe computation disabled for most shaders
All memory references in inner loop removed.
Removed float to integer conversion, replaced with 2 Iincs and 2 Fadds.
1 CPU: 1.4 s (real)		(17.5 cycles/voxel)

gcc -fpcc-struct-return -O2 -DHACK3 (Solaris 2.3)		14-JUN-1995
Wireframe computation disabled for most shaders
Shader returns after setup: inner loop skipped.
1 CPU: 0.6 s (real)		(7.5 cycles/voxel)

gcc -fpcc-struct-return -O2 (Solaris 2.3)			18-JUN-1995
Shader codeset: generic
No depth plane writing.
1 CPU  used: 2.7 s (real)	(35 cycles/voxel)

gcc -fpcc-struct-return -O2 (Solaris 2.3)			18-JUN-1995
Shader codeset: generic
Dummy depth plane writing.
1 CPU  used: 2.7 s (real)	(35 cycles/voxel)

gcc -fpcc-struct-return -O2 (Solaris 2.3)			18-JUN-1995
Shader codeset: integer
Dummy depth plane writing.
1 CPU  used: 2.3 s (real)	(29 cycles/voxel)
2 CPUs used: 1.2 s (real)
4 CPUs used: 0.7 s (real)

-------------------------------------------------------------------------------
Platform: Sparc 20 clone 2xHyperSPARC 90 Solaris 2.3

gcc -fpcc-struct-return -O
  Job split into 8 pieces, main thread helps:  3.4 s (real)
  Job split into 8 pieces, main thread idle:   2.8 s (real)
  Job split per line, main thread helps:       3.4 s (real)
  Job split per line, main thread idle:        3.8 s (real)
  Single job, one thread (one CPU):            5.5 s (real)

-------------------------------------------------------------------------------
Platform: Sparc 20 clone 4xHyperSPARC 90 Solaris 2.3

4 90 MHz HyperSPARC CPUs:
  Job split into 16 pieces, main thread idle:  1.6 s (real)

gcc -fpcc-struct-return -O					28-FEB-1995
  Single job, one thread (one CPU):            5.4 s (real)
  Job split into 16 pieces, main thread idle:  1.5 s (real)

gcc -fpcc-struct-return -O2					28-FEB-1995
  Single job, one thread (one CPU):            4.7 s (real)
  Job split into 16 pieces, main thread idle:  1.4 s (real)

-------------------------------------------------------------------------------
Platform: Sun Sparc 20 2xHyperSPARC 150 Solaris 2.4

gcc -fpcc-struct-return -O2 (Solaris 2.4)			22-FEB-1996
Shader codeset: integer
1 CPU  used: 1.5 s (real)
2 CPUs used: 0.8 s (real)

-------------------------------------------------------------------------------
Platform: Sun Sparc Ultra 140 Solaris 2.5

gcc -fpcc-struct-return -xO3 (Solaris 2.5)			?-JAN-1996
Shader codeset: integer
1 CPU  used: 1.5 s (real)

-------------------------------------------------------------------------------
Platform: SGI power challenge (4x75 MHz r8000s)

cc -xansi -signed -O
1 CPU  used: 5.0 s (real)
2 CPUs used: 2.5 s (real)
4 CPUs used: 1.4 s (real)

-------------------------------------------------------------------------------
Platform: SGI Indigo2 Extreme (75 MHz rs8000)

cc -xansi -signed -O3		c_dev: 6.0.2			30-MAY-1995
1 CPU: 4.2 s (real)

cc -xansi -signed -O3 -DHACK1	c_dev: 6.0.2			30-MAY-1995
Result suspect due to bug in rough.cast
1 CPU: 2.5 s (real)

cc -xansi -signed -O3 -DHACK1	c_dev: 6.0.2			13-JUN-1995
Wireframe computation disabled for most shaders.
Added offset (0.01) twice per loop iteration.
Pass array of plane pointers to shaders rather than cube and plane offsets.
1 CPU: 2.3 s (real)

cc -xansi -signed -O3 -DHACK1 -DHACK3	c_dev: 6.0.2		13-JUN-1995
Wireframe computation disabled for most shaders.
Pass array of plane pointers to shaders rather than cube and plane offsets.
1 CPU: 2.1 s (real)

cc -xansi -signed -O3 -DHACK1 -DHACK3	c_dev: 6.0.2		13-JUN-1995
Wireframe computation disabled for most shaders.
All memory references in inner loop removed.
1 CPU: 2.2 s (real)	(23 cycles/voxel)	THEORETICAL CPU LIMIT

cc -xansi -signed -O3 -DHACK1 -DHACK3	c_dev: 6.0.2		13-JUN-1995
Wireframe computation disabled for most shaders.
All memory references in inner loop removed.
Removed float to integer conversion, replaced with 2 Iincs and 2 Fadds.
1 CPU: 1.1 s (real)	(11.5 cycles/voxel)

cc -xansi -signed -O3	c_dev: 6.0.2				18-JUN-1995
Shader codeset: mips4	No longer pipelines.
No depth plane writing.
1 CPU: 3.2 s (real)

cc -xansi -signed -O3	c_dev: 6.0.2				18-JUN-1995
Shader codeset: mips4	No longer pipelines.
Dummy depth plane writing.
1 CPU: 3.2 s (real)

cc -xansi -signed -O3	c_dev: 6.0.2				18-JUN-1995
Shader codeset: generic
No depth plane writing.
1 CPU: 3.7 s (real)

cc -xansi -signed -O3	c_dev: 6.0.2				18-JUN-1995
Shader codeset: integer
No depth plane writing.
1 CPU: 2.2 s (real)

-------------------------------------------------------------------------------
Platform: SGI Power Challenge L (4x75 MHz r8000s)

cc -xansi -signed -O3 -DHACK1					11-JUN-1995
Result suspect due to bug in rough.cast
1 CPU  used: 2.3 s (real)
2 CPUs used: 1.1 s (real)
4 CPUs used: 0.7 s (real)

cc -xansi -signed -O3 -DHACK1	c_dev: 6.0.2			13-JUN-1995
Wireframe computation disabled for most shaders.
Added offset (0.01) only once per ray.
Pass array of plane pointers to shaders rather than cube and plane offsets.
1 CPU  used: 2.1 s (real)
2 CPUs used: 1.1 s (real)
4 CPUs used: 0.7 s (real)

cc -xansi -signed -O3 -DHACK1 -DHACK3	c_dev: 6.0.2		13-JUN-1995
Wireframe computation disabled for most shaders.
Pass array of plane pointers to shaders rather than cube and plane offsets.
1 CPU  used: 2.1 s (real)
2 CPUs used: 1.0 s (real)
4 CPUs used: 0.7 s (real)

cc -xansi -signed -O3	c_dev: 6.0.2				18-JUN-1995
Shader codeset: integer
No depth plane writing.
1 CPU  used: 2.2 s (real)

-------------------------------------------------------------------------------
Platform: SGI Power Challenge XL (10x75 MHz r8000s)

cc -xansi -signed -O3	c_dev: 6.0.?				11-JUL-1995
Shader codeset: integer
Dummy depth plane writing.
1  CPU  used: 2.3 s (real)
2  CPUs used: 1.1 s (real)
4  CPUs used: 0.6 s (real)
10 CPUs used: 0.3 s (real)					27-JUL-1995

-------------------------------------------------------------------------------
Platform: 486DX50 256k secondary cache

cc -O
1 CPU: 39.2 s (real)

-------------------------------------------------------------------------------
Platform: Pentium 100 MHz 256k secondary cache

cc -O2 (GCC 2.6.4-950518 ELF)					21-MAY-1995
1 CPU: 8.2 s (real)

cc -O2 (GCC 2.6.4-950530 ELF)					9-JUN-1995
1 CPU: 7.9 s (real)

cc -O2 -DHACK1 (GCC 2.6.4-950530 ELF)				9-JUN-1995
1 CPU: 8.1 s (real)

cc -O2 (GCC 2.6.4-950530 ELF)					12-JUN-1995
Wireframe computation disabled.
Added offset (0.01) only once per ray.
1 CPU: 7.5 s (real)

cc -O2 (GCC 2.6.4-950530 ELF)					12-JUN-1995
Wireframe computation disabled.
Added offset (0.01) only once per ray.
Pass array of plane pointers to shaders rather than cube and plane offsets.
1 CPU: 7.5 s (real)

cc -O2 -DHACK2 (GCC 2.6.4-950530 ELF)				12-JUN-1995
Wireframe computation disabled for most shaders.
Added offset (0.01) only once per ray.
Pass array of plane pointers to shaders rather than cube and plane offsets.
1 CPU: 7.3 s (real)

cc -O2 -DHACK3 (GCC 2.6.4-950530 ELF)				13-JUN-1995
Wireframe computation disabled for most shaders.
Pass array of plane pointers to shaders rather than cube and plane offsets.
1 CPU: 7.0 s (real)	(97 cycles/voxel)

cc -O2 -DHACK3 (GCC 2.6.4-950530 ELF)				15-JUN-1995
Wireframe computation disabled for most shaders.
Pass array of plane pointers to shaders rather than cube and plane offsets.
1 CPU: 6.8 s (real)	(94 cycles/voxel)

cc -O2 -DHACK3 (GCC 2.6.4-950612 ELF)				16-JUN-1995
Wireframe computation disabled for most shaders.
Pass array of plane pointers to shaders rather than cube and plane offsets.
1 CPU: 6.7 s (real)	(93 cycles/voxel)

cc -O2 (GCC 2.6.4-950612 ELF)					18-JUN-1995
Shader codeset: generic
No depth plane writing.
1 CPU: 6.8 s (real)

cc -O2 (GCC 2.6.4-950612 ELF)					18-JUN-1995
Shader codeset: generic
Dummy depth plane writing.
1 CPU: 6.8 s (real)

cc -O2 (GCC 2.6.4-950612 ELF)					18-JUN-1995
Shader codeset: integer
Dummy depth plane writing.
1 CPU: 3.9 s (real)	(54 cycles/voxel)

cc -O2 (GCC 2.7.0 ELF)						19-JUN-1995
Shader codeset: generic
Dummy depth plane writing.
1 CPU: 6.7 s (real)

cc -O2 (GCC 2.7.0 ELF)						19-JUN-1995
Shader codeset: integer
Dummy depth plane writing.
1 CPU: 3.8 s (real)

-------------------------------------------------------------------------------
Platform: Pentium 100 MHz 512k secondary cache

cc -O2 (GCC 2.6.4-950518 ELF)					25-MAY-1995
1 CPU: 7.7 s (real)

cc -O2 (GCC 2.6.4-950523 ELF)					11-JUN-1995
1 CPU: 7.7 s (real)

cc -O2 (GCC 2.6.4-950523 ELF)					11-JUN-1995
Wireframe computation disabled.
1 CPU: 7.4 s (real)

cc -O2 (GCC 2.6.4-950523 ELF)					11-JUN-1995
Wireframe computation disabled.
Added offset (0.01) only once per ray.
1 CPU: 7.1 s (real)

cc -O2 -DHACK1 -DHACK3 (GCC 2.6.4-950523 ELF)			12-JUN-1995
Wireframe computation disabled for most shaders.
Pass array of plane pointers to shaders rather than cube and plane offsets.
1 CPU: 7.0 s (real)

cc -O2 -DHACK3 (GCC 2.6.4-950523 ELF)				12-JUN-1995
Wireframe computation disabled for most shaders.
Pass array of plane pointers to shaders rather than cube and plane offsets.
1 CPU: 6.5 s (real)	(90 cycles/voxel)	verified 13-JUN-1995

cc -O2 (GCC 2.6.4-950523 ELF)					13-JUN-1995
Wireframe computation disabled for most shaders.
Added offset (0.01) twice per loop iteration.
Pass array of plane pointers to shaders rather than cube and plane offsets.
1 CPU: 7.4 s (real)

cc -O2 -DHACK3 (GCC 2.6.4-950523 ELF)				13-JUN-1995
Wireframe computation disabled for most shaders.
All memory references in inner loop removed.
1 CPU: 5.5 s (real)	(65 cycles/voxel)	THEORETICAL CPU LIMIT

cc -O2 -DHACK3 (GCC 2.6.4-950523 ELF)				13-JUN-1995
Wireframe computation disabled for most shaders.
All memory references in inner loop removed.
Removed float to integer conversion, replaced with 2 Iincs and 2 Fadds.
1 CPU: 2.6 s (real)	(25 cycles/voxel)

cc -O2 -DHACK3 (GCC 2.6.4-950523 ELF)				13-JUN-1995
Wireframe computation disabled for most shaders.
All memory references in inner loop removed.
Removed float to integer conversion, replaced with 2 integer increments.
1 CPU: 2.6 s (real)	(25 cycles/voxel)

cc -O2 -DHACK3 (GCC 2.6.4-950523 ELF)				13-JUN-1995
Wireframe computation disabled for most shaders.
All memory references in inner loop removed.
Removed float to integer conversion, no replacement integer increments.
1 CPU: 2.3 s (real)	(21 cycles/voxel)

cc -O2 -DHACK3 (GCC 2.6.4-950523 ELF)				13-JUN-1995
Wireframe computation disabled for most shaders.
All memory references in inner loop removed.
Removed float to integer conversion, no replacement integer increments.
Extra floating point add.
1 CPU: 2.6 s (real)	(25 cycles/voxel)

cc -O2 -DHACK3 (GCC 2.6.4-950523 ELF)				13-JUN-1995
Shader returns after setup: inner loop skipped.
1 CPU: 0.8 s (real)

cc -O2 (GCC 2.6.4-950523 ELF)					19-JUN-1995
Shader codeset: generic
Dummy depth plane writing.
1 CPU: 6.4 s (real)

cc -O2 (GCC 2.6.4-950523 ELF)					19-JUN-1995
Shader codeset: integer
Dummy depth plane writing.
1 CPU: 3.4 s (real)	(47 cycles/voxel)

-------------------------------------------------------------------------------
Platform: Pentium 100 MHz 256k secondary cache two machines

cc -O2 -DHACK3 (GCC 2.6.4-950530 ELF)				13-JUN-1995
Wireframe computation disabled for most shaders.
Pass array of plane pointers to shaders rather than cube and plane offsets.
Compute time: 3.8 s (real)					16-JUN-1995
-------------------------------------------------------------------------------
Platform: DEC alpha/OSF1

cc -O								6-DEC-1995
Shader codeset: integer
Dummy depth plane writing.
Compute time: 1.7 s (CPU: realtime large due to heavily loaded machine) 6-DEC-1995
-------------------------------------------------------------------------------
Platform: Dual Pentium 133 MHz 256k pipeline secondary cache

cc -O2 (GCC 2.7.2)						29-JUL-1996
Shader codeset: integer
1 CPU  used: 2.8 s (real)
2 CPUs used: 1.5 s (real)

*******************************************************************************
Quick speed comparison with tile1_balls.kf and integer codeset.Single CPU times
No cache.

max_voxel:
----------
AMD Athalon 500 MHz (Linux 2.3.28):	0.5 s	6-DEC-1999
Sparc Ultra 2 (336 MHz, Solaris 2.6):	0.6 s	3-SEP-1998
Pentium III 500 MHz (Linux 2.3.28):	0.7 s	6-DEC-1999
Pentium II 400 MHz (Linux 2.1.130-SMP):	0.7 s	3-DEC-1998
alpha EV5/400 2M cache (kaputar, V4.0):	0.8 s	12-OCT-1998
Pentium II 266 MHz, EDO RAM:		1.0 s	12-DEC-1997
alpha 21164@266MHz 2M cache (siamang):	1.0 s
Sparc Ultra 2 (200 MHz, Solaris 2.5.1):	1.0 s	27-JUN-1997
Sparc Ultra 170E (cc -fast -xO5):	1.3 s
Sparc Ultra 170E:			1.4 s	27-JUN-1997
Dec alpha (kaputar):			1.4 s
PentiumPro/180:				1.5 s
Sparc 20 HyperSPARC150:			1.5 s
Sparc Ultra 140:			1.5 s
Sparc 10 HyperSPARC150 (SunOS 4.1.4):	1.6 s
Sparc 10 HyperSPARC150 (Solaris 2.5):	1.8 s
Power Challenge XL/r8000 75 MHz:	2.2 s
P166/256 kByte L2 cache:		2.3 s	13-NOV-1996
Cyrix/IBM 6x86MX PR233 (188 MHz):	2.3 s	24-AUG-1998
Sparc 4/670 HyperSPARC90:		2.3 s
hp9000/J200:				2.6 s
Pentium120/256 kByte L2 cache:		3.3 s
Pentium100/256 kByte L2 cache (Triton):	3.3 s
Sparc 5/110:				3.3 s
Pentium100/512 kByte L2 cache:		3.4 s
hp9000/755:				3.5 s
Pentium100/256 kByte L2 cache:		3.8 s
Sparc 5/85:				4.3 s
Cyrix/IBM 6x86 100 MHz no L2, 430VX:	4.6 s	24-AUG-1998
Sparc2/Weitek:				5.3 s
Cyrix 5x86/100 Laptop:			6.7 s
hp9000/715-33:				9.7 s
486DX2-66/256 kByte L2 cache:		9.9 s  10.9 s

hot gas mono:
-------------
AMD Athalon 500 MHz (Linux 2.3.28):	0.7 s	6-DEC-1999
Pentium III 500 MHz (Linux 2.3.28):	0.9 s	6-DEC-1999
Pentium II 400 MHz (Linux 2.1.130-SMP):	0.9 s	3-DEC-1998
alpha EV5/400 2M cache (kaputar, V4.0):	1.1 s	12-OCT-1998
Sparc Ultra 2 (336 MHz, Solaris 2.6):	1.2 s	3-SEP-1998
Pentium II 266 MHz, EDO RAM:		1.4 s	12-DEC-1997
alpha 21164@266MHz 2M cache (siamang):	1.5 s
Sparc Ultra 170E (cc -fast -xO5):	1.7 s
PentiumPro/180:				2.0 s
Dec alpha (kaputar):			2.0 s
Sparc Ultra 2 (200 MHz, Solaris 2.5.1):	2.5 s	27-JUN-1997
Sparc 10 HyperSPARC150 (SunOS 4.1.4):	2.6 s
Sparc Ultra 140:			2.9 s
Sparc Ultra 170E:			3.1 s	27-JUN-1997
hp9000/J200:				3.2 s
P166/256 kByte L2 cache:		3.5 s	13-NOV-1996
hp9000/755:				3.5 s
Power Challenge XL/r8000 75 MHz:	4.8 s	?? max_voxel was 2.7 s ??
Cyrix/IBM 6x86MX PR233 (188 MHz):	4.9 s	24-AUG-1998
Pentium120/256 kByte L2 cache:		5.1 s
Pentium100/256 kByte L2 cache (Triton):	5.3 s	13-NOV-1996
Pentium100/256 kByte L2 cache:		6.1 s	>> max_voxel was 4.0 s <<
Sparc 20 HyperSPARC150:			6.1 s	>> max_voxel was 1.7 s <<
Sparc 10 HyperSPARC150:			6.9 s
Sparc 5/110:				8.0 s
Sparc2/Weitek:				8.4 s
Cyrix/IBM 6x86 100 MHz no L2, 430VX:	9.7 s	24-AUG-1998
Sparc 4/670 HyperSPARC90:		9.8 s	>> max_voxel was 2.8 s <<
Sparc 5/85:				10.2 s
hp9000/715-33:				11.7 s
Cyrix 5x86/100 Laptop:			13.3 s
486DX2-66/256 kByte L2 cache:		22.0 s

max_voxel, smooth:
------------------
alpha EV5/400 2M cache (kaputar, V4.0):	7.9 s	12-OCT-1998
AMD Athalon 500 MHz (Linux 2.3.28):	8.8 s	6-DEC-1999
alpha 21164@266MHz 2M cache (siamang):	10.7 s
Sparc Ultra 2 (336 MHz, Solaris 2.6):	11.7 s	3-SEP-1998
Pentium III 500 MHz (Linux 2.3.28):	12.3 s	6-DEC-1999
Pentium II 400 MHz (Linux 2.1.130-SMP):	16.8 s	3-DEC-1998
Dec alpha (kaputar):			17.1 s
Sparc 10 HyperSPARC150 (SunOS 4.1.4):	19.4 s
Sparc Ultra 2 (200 MHz, Solaris 2.5.1):	19.6 s	27-JUN-1997
Sparc Ultra 170E (cc -xO5):		23.1 s
Sparc Ultra 170E:			23.9 s	27-JUN-1997
Pentium II 266 MHz, EDO RAM:		24.2 s	12-DEC-1997
Sparc 10 HyperSPARC150:			25.2 s
Sparc 20 HyperSPARC150:			26.0 s  >> max_voxel was 1.7 s <<
Sparc Ultra 140:			27.8 s
Power Challenge XL/r8000 75 MHz:	29.6 s  ?? max_voxel was 2.7 s ??
P166/256 kByte L2 cache:		31.0 s  13-NOV-1996
PentiumPro/180:				35.3 s
Sparc 4/670 HyperSPARC90:		39.1 s  >> max_voxel was 2.8 s <<
Pentium120/256 kByte L2 cache:		43.5 s
Cyrix/IBM 6x86MX PR233 (188 MHz):	43.6 s	24-AUG-1998
Sparc 5/110:				47.7 s
Pentium100/256 kByte L2 cache (Triton):	49.5 s  13-NOV-1996
Pentium100/256 kByte L2 cache:		53.4 s  >> max_voxel was 4.0 s <<
hp9000/J200:				55.9 s
hp9000/755:				59.9 s
Sparc 5/85:				64.9 s
Sparc2/Weitek:				75.6 s
Cyrix/IBM 6x86 100 MHz no L2, 430VX:	86.2 s	24-AUG-1998
hp9000/715-33:				190.9 s
486DX2-66/256 kByte L2 cache:		

*******************************************************************************
Parallelisation tests with various algorithms.
Datasets used may be different for each platform.
Times rounded to nearest 0.1 s. Speedups computed with original data.
-------------------------------------------------------------------------------
Platform: Sparc Ultra Enterprise 6000 with 10 Ultra 200 MHz CPUs

max_voxel shader:
1  CPU  used: 5.0 s (real)	1.0x
2  CPUs used: 2.3 s (real)	2.1x
5  CPUs used: 1.0 s (real)	4.7x
10 CPUs used: 0.8 s (real)	7.3x				12-MAY-1997

hot_gas_mono shader:
1  CPU  used: 7.3 s (real)	1.0x
2  CPUs used: 4.8 s (real)	1.5x
5  CPUs used: 2.1 s (real)	3.5x
10 CPUs used: 1.3 s (real)	5.7x				12-MAY-1997

max_voxel shader with smoothing:
1  CPU  used: 43.8 s (real)	1.0x
2  CPUs used: 23.3 s (real)	1.9x
5  CPUs used: 10.6 s (real)	4.1x
10 CPUs used: 8.6  s (real)	5.1x				12-MAY-1997
