xps.1.1

; this code performs the YV12 -> RGB conversion.
; Y, U, and V textures are inputted (pre-scaled by the hardware)
; RGB values are then calculated.

; Formula is as follows:

; We start with separate Y, U, V textures.  In byte notation, they have the following allowed ranges:

; Y in [16,235]
; U in [16,239]
; V in [16,239]

; Step 1: These are converted to a 0..1 range for Y and a -0.5..0.5 range for U and V:

; we let Y' = (Y-16)/219
; we let U' = (U-128)/223
; we let V' = (V-128)/223

; Step 2: We then calculate using the known YUV->RGB formula:

; r0_R = Y'              + 1.403 * V'
; r0_G = Y' - 0.344 * U' - 0.714 * V'
; r0_B = Y' + 1.770 * U'

; see http://www.fourcc.org/fccyvrgb.php for more info

; Set up u and v co-efficients, to be doubled before applying
;def c0, 0      ,0.18664,0.96032,0
;def c1, 0.76120,0.38738,0      ,0
; FIXME: coefficient corresponding to 1.770 should be 1.012 
def c0, 0      ,0.19668,1,0
def c1, 0.80216,0.40823,0,0
; setup sign multiplier value as constants can't be negative
def c2, 1,0,1,1 
; y adjust
def c3, 0.0625,0.0625,0.0625,0
; y co-efficient, halved
def c4, 0.58219,0.58219,0.58219,0.5
; texture read
tex t0
tex t1
tex t2

; try to keep this to 4 or less ops!

; r0 = 2*(u-0.5)*c0, r1 = 2*(v-0.5)*c1;
xmma_x2 r0,r1,discard, t1_bias,c0, t2_bias,c1

; r0 = r0 + r1, with green channel negate
xmma discard,discard,r0, r0,c2_bx2, r1,c2_bx2

; r1 = 2*yfac*(y-yadj)
; r2 = y-yadj, r1 = 2*yfac*r2
xmma_x2 discard,discard,r1, t0,c4, -c3,c4

; r0 = r0 + r1 (alpha set from y, a = 1.0)
add r0, r0, r1