foreach mesh {
	BASE	0
	OFFSET	from pipe struct

	// call instanced data:
	foreach batch {
		UNPACK
		ITOP		set vertex count
		MSCALF/MSCNT	call micro program

		last batch: FLUSH (earlier: MSKPATH3 enable)
	}
}

VU
==
XITOP	get ITOP (vertex count)
XTOP	get TOP (input address)

VIF
===
BASE:	set BASE
OFFSET:	set OFFSET
	TOPS <- BASE
	DBF <- 0
UNPACK:	transfer to TOPS
ITOP:	set ITOPS
MSCAL:	TOP <- TOPS
	ITOP <- ITOPS
	TOPS <- BASE/BASE+OFFSET (DBF flag)
	DBF <- ~DBF

0 - OFFSET:	input buffer 1

No clipping
===========

	Double buffered
+--------------------+	0
|                    |
|                    |
|   Input 1          |
|                    |
|                    |
+--------------------+	OFFSET
|                    |
|                    |
|   Input 2          |
|                    |                                	Triple buffered
|                    |                               |                    |
+--------------------+	ExpandBuffer1                +--------------------+	ExpandBuffer1
|                    |                               |                    |
|                    |                               |   Output 1         |
|   Output 1         |                               |                    |
|                    |                               +--------------------+	ExpandBuffer2
|                    |                               |                    |
+--------------------+	ExpandBuffer2                |   Output 2         |
|                    |                               |                    |
|                    |                               +--------------------+	ExpandBuffer3
|   Output 2         |                               |                    |
|                    |                               |   Output 3         |
|                    |                               |                    |
+--------------------+	vuSymbStaticDataStart        +--------------------+	vuSymbStaticDataStart
|                    |                               |                    |
|   Static data      |
|                    |
+--------------------+

Clipping
========

	Double buffered                               	Triple buffered
|                    |                               |                    |
+--------------------+	ExpandBuffer1                +--------------------+	ExpandBuffer1
|                    |                               |                    |
|                    |                               |   Output 1         |
|   Output 1         |                               |                    |
|                    |                               +--------------------+	XExpandBuffer2
|                    |                               |                    |
+--------------------+	XExpandBuffer2               |   Output 2         |
|                    |                               |                    |
|                    |                               +--------------------+	XExpandBuffer3
|   Output 2         |                               |                    |
|                    |                               |   Output 3         |
|                    |                               |                    |
+--------------------+	ClippingBuffer               |                    |
|                    |                               +--------------------+	ClippingBuffer
|                    |                               |                    |
|   Clipped Input    |                               |   Clipped Input    |
|                    |                               |                    |
|                    |                               |                    |
+--------------------+	vuSymbStaticDataStart        +--------------------+	vuSymbStaticDataStart
|                    |                               |                    |

Clipping uses one of the idle output buffers to store a clipped polygon.
This is then inserted as input triangles into the clipping buffer.
Whenever the clipping buffer is full, the first pass and FX pass generators
will use it to fill the output buffer(s) and XGKICK them.
Then clipping continues until all input has been processed.



VU program
==========

General flow:

	setup matrix and code path
	for(;;){	// meshes
		light & animate input buffer

		if(clipping){
			while(input buffer not empty){
				foreach primitive while clip buffer not full {
					transform to clip space and clip in idle output buffer
					interpolate vertices, add to clip buffer
				}
				transform clip to screen space
				xgkick pass1; rotate buffers
				(generate pass2
				 xgkick pass2; rotate buffers)
			}
		}else{
			transform model to screen space
			xgkick pass1; rotate buffers
			(generate pass2
			 xgkick pass2; rotate buffers)
		}

		halt	// restart by MSCNT
	}


Static data
===========

set up in _rwSkyPipeSetupForCamera
TODO: what's useFarClip?

	0x3F0	vuSDmat0                   combined matrix; w row = z row
	0x3F1	vuSDmat1                   combined matrix
	0x3F2	vuSDmat2                   combined matrix
	0x3F3	vuSDmat3                   combined matrix
	0x3F4	vuSDnearClip               xyzw: camera near plane			vu1DataNearClip
	0x3F5	vuSDfarClip                xyzw: camera far plane			vu1DataFarClip
	0x3F6	vuSDxMaxyMax               ?, ?, -255/(fogEnd-fogStart), fogEnd		xMaxYMax128
	0x3F7	vuSDcamWcamHzScale         raster width, raster height, zScale, 0	vu1DataXYZScale
	0x3F8	vuSDoffXoffYzShift         0, 0, zShift, 0				vu1DataXYZShift
	0x3F9	vuSDrealOffset             xoff, yoff, 0, 0				vu1DataOffset3D
	0x3FA	vuSDgifTag                 						gifTag128
	0x3FB	vuSDcolScale               material color (scaled by 128/255 when textured)
	0x3FC	vuSDsurfProps              surface properties
	0x3FD	vuSDpingPongAddr           obsolete
	0x3FE	vuSDpingPongCount          obsolete
	0x3FD	vuSDClipvec1               cull: 2047.9374-screenWidth/2, 1/x, zScaleClip, far plane		skyClipVect1
                                           clip: screenWidth/2 + 5, 1/x, zScaleClip, far plane			skyCClipVect1
	0x3FE	vuSDClipvec2               cull: 2047.9374-screenHeight/2, 1/x, zShiftClip, near plane		skyClipVect2
                                           clip: screenHeight/2 + 5, 1/x, zShiftClip, near plane		skyCClipVect2
	0x3FF	vuSDVUSwitch               code path switch, render system, ?, backface cull

xoff = fb->width/2 + fb->offsetX - par->width/2 + 2048
yoff = fb->height/2 + fb->offsetY - par->height/2 + 2048

persp zScaleClip: -2*far*near/(far-near)
persp zShiftClip: (far+near)/(far-near)
paral zScaleClip: 2/(far-near)
paral zShiftClip: -(far+near)/(far-near)



GTA San Andreas
===============

Building:
  DN	0x3CE	x dnbalance; y 1-WetRoads
	0x3CF	xyz building ambient light

BuildingEnvMap:
  DN	0x3C3	x dnbalance; y 1-WetRoads
	0x3C4	xyz building ambient light
	0x3C5	GIF packet env state; FOGCOL, TEX0_2
	0x3C8	GIF packet diffuse state; FOGCOL, TEX0_1
	0x3CB	GIF tag env
	0x3CC	xyz env matrix;
	0x3CD	xyz env matrix; w lightmult
	0x3CE	xyz env matrix; w shininess
	0x3CF	xyzw env transform

BuildingUVA:
	0x3CD	xyz building ambient light
	0x3CE	xy texmat right.xy; zw texmat up.xy
	0x3CF	zw texmat pos.xy

BuildingDNUVA:
	0x3CD	x dnbalance; y 1-WetRoads; zw texmat pos.xy
	0x3CE	xyz building ambient light
	0x3CF	xy texmat right.xy; zw texmat up.xy

Car:
	0x3CD	GIF packet; TEX0_1, TEX2_1

CarEnv:
	0x3BD	GIF packet spec state; FOGCOL, TEX0_2
	0x3C0	GIF packet env state; FOGCOL, TEX0_2
	0x3C3	GIF packet diffuse state; FOGCOL, TEX0_1, TEX2_1
	0x3C7	GIF tag spec/env
	0x3C8	xyz normal view matrix (spec); w specularity
	0x3C9	xyz normal view matrix (spec); effect flags
	0x3CA	xyz normal view matrix (spec)
	0x3CB	xyz view sun direction
	0x3CC	xy normal world matrix (env)
	0x3CD	xy normal world matrix (env); w lightmult
	0x3CE	xy normal world matrix (env); w shininess
	0x3CF	xyzw env transform