Aurelien
/
xf86-video-intel
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Fix the sf_mask program to compute and pass corrected uvw cooefficients 

sf_mask is the same as sf except that it must compute both src and mask uvw
cooefficients, which are conveniently adjacent in the same registers, and so
need only an extended execution width
This commit is contained in:
Keith Packard 2008-03-30 19:14:18 -07:00
parent 6304b38423
commit 492ff1494f
2 changed files with 72 additions and 52 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

104
src/exa_sf_mask.g4a
View File

@ -21,52 +21,82 @@
* IN THE SOFTWARE.
*
* Authors:
* Keith Packard <keithp@keithp.com>
* Eric Anholt <eric@anholt.net>
* Wang Zhenyu <zhenyu.z.wang@intel.com>
*/
/* FIXME how to setup second coeffient for mask tex coord */
/*
g3 (v0) { u0, v0, 1.0, 1.0 } ==> {u0, v0, 1.0, 1.0, mu0, mv0, 1.0, 1.0} Co[0](u0) Co[1](v0) Co[2](mu0) Co[3](mv0)
g4 (v1) { u1, v1, 1.0, 1.0 } ==> {u1, v1, 1.0, 1.0, mu1, mv1, 1.0, 1.0}
g5 (v2) { u2, v2 } ==> (u2, v2, mu2, mv2}
g6 { 1/(x1-x0), 1/(y1-y0) }
g7 { u1-u0, v1-v0, 0, 0} ==>{u1-u0, v1-v0,0, 0, mu1-mu0, mv1-mv0, 0, 0}
-> { (u1-u0)/(x1-x0), (v1-v0)/(y1-y0) } ==>{(u1-u0)/(x1-x0), (v1-v0)/(y1-y0),(mu1-mu0)/(x1-x0), (mv1-mv0)/(y1-y0)
Cx, Cy Cx[0], Cy[0], Cx[1], Cy[1]
/*
* Inputs (note all sub-register addresses are bytes, not float indices)
*
* Note that the vertices will have been reordered:
*
* V0 is topmost (leftmost among topmost) (upper left)
* V1 is next clockwise (lower right)
* V2 is remaining (lower left)
*
* V0 ...................... XX
* | .
* | .
* | .
* V2------------------------V1
*
* G0 thread state -- just pass along
*
* G1 and G2 are fixed by SF spec
*
* G1.0 reserved
* G1.4 Provoking vertex
* G1.8 Determinant
* G1.12 X1 - X0
* G1.16 X2 - X0
* G1.20 Y1 - Y0
* G1.24 Y2 - Y0
* G1.30 reserved
*
* G2.0 Z0
* G2.4 1/W0
* G2.8 Z1
* G2.12 1/W1
* G2.16 Z2
* G2.20 1/W2
* G2.24 reserved
* G2.30 reserved
*
* G3 is V0 Vertex Attribute Data from URB (upper left)
*
* G3.0 u0
* G3.4 v0
*
* G4 is V1 Vertex Attribute Data from URB (lower right)
*
* G4.0 u1
* G4.4 v1
*
* G5 is V2 Vertex Attribute Data from URB (lower left)
*
*/
/* assign Cx[0], Cx[1] to src, same to Cy, Co
Cx[2], Cx[3] to mask, same to Cy, Co */
/* Compute inverses of the input deltas */
send (4) 0 g6<1>F g1.12<4,4,1>F math inv mlen 1 rlen 1 { align1 };
send (1) 0 g6<1>F g1.12<0,1,0>F math inv scalar mlen 1 rlen 1 { align1 };
send (1) 0 g6.4<1>F g1.20<0,1,0>F math inv scalar mlen 1 rlen 1 { align1 };
add (8) g7<1>F g4<8,8,1>F -g3<8,8,1>F { align1 };
/* Cx[0] */
mul (1) g7<1>F g7<0,1,0>F g6<0,1,0>F { align1 };
/* Cy[0] */
mul (1) g7.4<1>F g7.4<0,1,0>F g6.4<0,1,0>F { align1 };
/* Cx[2] */
mul (1) g7.8<1>F g7.8<0,1,0>F g6<0,1,0>F { align1 };
/* Cy[2] */
mul (1) g7.12<1>F g7.12<0,1,0>F g6.4<0,1,0>F { align1 };
/* src Cx[0], Cx[1] */
mov (8) m1<1>F g7<0,1,0>F { align1 };
/* mask Cx[2], Cx[3] */
mov (1) m1.8<1>F g7.8<0,1,0>F { align1 };
mov (1) m1.12<1>F g7.8<0,1,0>F { align1 };
/* src Cy[0], Cy[1] */
mov (8) m2<1>F g7.4<0,1,0>F { align1 };
/* mask Cy[2], Cy[3] */
mov (1) m2.8<1>F g7.12<0,1,0>F { align1 };
mov (1) m2.12<1>F g7.12<0,1,0>F { align1 };
/* src Co[0], Co[1] */
/* texture location at V0 */
mov (8) m3<1>F g3<8,8,1>F { align1 };
/* mask Co[2], Co[3] */
mov (1) m3.8<1>F g3.8<0,1,0>F { align1 };
mov (1) m3.12<1>F g3.12<0,1,0>F { align1 };
/* compute V1 - V2 (motion in X) for texture coordinates */
add (8) g7<1>F g4<8,8,1>F -g5<8,8,1>F { align1 };
/* multiply by 1/dx */
mul (8) m1<1>F g7<8,8,1>F g6.0<0,1,0>F { align1 };
/* Compute V2 - V0 (motion in Y) for texture coordinates */
add (8) g7<1>F g5<8,8,1>F -g3<8,8,1>F { align1 };
/* multiply by 1/dy */
mul (8) m2<1>F g7<8,8,1>F g6.8<0,1,0>F {align1 };
/* and we're done */
send (8) 0 null g0<8,8,1>F urb 0 transpose used complete mlen 4 rlen 0 { align1 EOT };
nop;
nop;

20
src/exa_sf_mask_prog.h
View File

@ -1,19 +1,9 @@
{ 0x00000031, 0x20c01fbd, 0x0000002c, 0x01110081 },
{ 0x00000031, 0x20c41fbd, 0x00000034, 0x01110081 },
{ 0x00600040, 0x20e077bd, 0x008d0080, 0x008d4060 },
{ 0x00000041, 0x20e077bd, 0x000000e0, 0x000000c0 },
{ 0x00000041, 0x20e477bd, 0x000000e4, 0x000000c4 },
{ 0x00000041, 0x20e877bd, 0x000000e8, 0x000000c0 },
{ 0x00000041, 0x20ec77bd, 0x000000ec, 0x000000c4 },
{ 0x00600001, 0x202003be, 0x000000e0, 0x00000000 },
{ 0x00000001, 0x202803be, 0x000000e8, 0x00000000 },
{ 0x00000001, 0x202c03be, 0x000000e8, 0x00000000 },
{ 0x00600001, 0x204003be, 0x000000e4, 0x00000000 },
{ 0x00000001, 0x204803be, 0x000000ec, 0x00000000 },
{ 0x00000001, 0x204c03be, 0x000000ec, 0x00000000 },
{ 0x00400031, 0x20c01fbd, 0x0069002c, 0x01110001 },
{ 0x00600001, 0x206003be, 0x008d0060, 0x00000000 },
{ 0x00000001, 0x206803be, 0x00000068, 0x00000000 },
{ 0x00000001, 0x206c03be, 0x0000006c, 0x00000000 },
{ 0x00600040, 0x20e077bd, 0x008d0080, 0x008d40a0 },
{ 0x00600041, 0x202077be, 0x008d00e0, 0x000000c0 },
{ 0x00600040, 0x20e077bd, 0x008d00a0, 0x008d4060 },
{ 0x00600041, 0x204077be, 0x008d00e0, 0x000000c8 },
{ 0x00600031, 0x20001fbc, 0x008d0000, 0x8640c800 },
{ 0x0000007e, 0x00000000, 0x00000000, 0x00000000 },
{ 0x0000007e, 0x00000000, 0x00000000, 0x00000000 },