EXA: mask componentAlpha support

we're getting load of multiple sf/wm programs for i965, might
need to be cleanup in future.

5 files changed, 626 insertions, 20 deletions

diff --git a/src/exa_wm_maskca.g4a b/src/exa_wm_maskca.g4a
new file mode 100644
index 00000000..fa923f6c
--- /dev/null
+++ b/src/exa_wm_maskca.g4a
@@ -0,0 +1,202 @@
+/*
+ * This's for exa composite operation in no mask picture case.
+ * The simplest case is just sending what src picture has to dst picture.
+ * XXX: This is still experimental, and should be fixed to support multiple texture
+ * map, and conditional mul actions. 
+ */
+
+/* I think this should be same as in g4a program for texture video,
+   as we also use 16-pixel dispatch. and SF scale in g3 is useful for us. */
+
+/* The initial payload of the thread is always g0.
+ * WM_URB (incoming URB entries) is g3
+   As mask texture coeffient needs extra setup urb starting from g4, we should
+   shift this location. 
+
+ * X0_R is g4->g6
+ * X1_R is g5->g7
+ * Y0_R is g6->g8
+ * Y1_R is g7->g9
+
+     * X0: {ss0.x, ss0.x+1, ss0.x,   ss0.x+1, ss1.x, ss1.x+1, ss1.x,   ss1.x+y}
+     * Y0: {ss0.y, ss0.y,   ss0.y+1, ss0.y+1, ss1.y, ss1.y,   ss1.y+1, ss1.y+1}
+     * X1: {ss2.x, ss2.x+1, ss2.x,   ss2.x+1, ss3.x, ss3.x+1, ss3.x,   ss3.x+y}
+     * Y1: {ss2.y, ss2.y,   ss2.y+1, ss2.y+1, ss3.y, ss3.y,   ss3.y+1, ss3.y+1}
+ */
+
+/* multitexture program with src and mask texture */
+/* - load src texture */
+/* - load mask texture */
+/* - mul src.X with mask's alpha */
+/* - write out src.X */
+
+    /* Set up ss0.x coordinates*/
+mov (1) g6<1>F g1.8<0,1,0>UW { align1 };
+add (1) g6.4<1>F g1.8<0,1,0>UW 1UB { align1 };
+mov (1) g6.8<1>F g1.8<0,1,0>UW { align1 };
+add (1) g6.12<1>F g1.8<0,1,0>UW 1UB { align1 };
+    /* Set up ss0.y coordinates */
+mov (1) g8<1>F g1.10<0,1,0>UW { align1 };
+mov (1) g8.4<1>F g1.10<0,1,0>UW { align1 };
+add (1) g8.8<1>F g1.10<0,1,0>UW 1UB { align1 };
+add (1) g8.12<1>F g1.10<0,1,0>UW 1UB { align1 };
+    /* set up ss1.x coordinates */
+mov (1) g6.16<1>F g1.12<0,1,0>UW { align1 };
+add (1) g6.20<1>F g1.12<0,1,0>UW 1UB { align1 };
+mov (1) g6.24<1>F g1.12<0,1,0>UW { align1 };
+add (1) g6.28<1>F g1.12<0,1,0>UW 1UB { align1 };
+    /* set up ss1.y coordinates */
+mov (1) g8.16<1>F g1.14<0,1,0>UW { align1 };
+mov (1) g8.20<1>F g1.14<0,1,0>UW { align1 };
+add (1) g8.24<1>F g1.14<0,1,0>UW 1UB { align1 };
+add (1) g8.28<1>F g1.14<0,1,0>UW 1UB { align1 };
+    /* Set up ss2.x coordinates */
+mov (1) g7<1>F g1.16<0,1,0>UW { align1 };
+add (1) g7.4<1>F g1.16<0,1,0>UW 1UB { align1 };
+mov (1) g7.8<1>F g1.16<0,1,0>UW { align1 };
+add (1) g7.12<1>F g1.16<0,1,0>UW 1UB { align1 };
+    /* Set up ss2.y coordinates */
+mov (1) g9<1>F g1.18<0,1,0>UW { align1 };
+mov (1) g9.4<1>F g1.18<0,1,0>UW { align1 };
+add (1) g9.8<1>F g1.18<0,1,0>UW 1UB { align1 };
+add (1) g9.12<1>F g1.18<0,1,0>UW 1UB { align1 };
+    /* Set up ss3.x coordinates */
+mov (1) g7.16<1>F g1.20<0,1,0>UW { align1 };
+add (1) g7.20<1>F g1.20<0,1,0>UW 1UB { align1 };
+mov (1) g7.24<1>F g1.20<0,1,0>UW { align1 };
+add (1) g7.28<1>F g1.20<0,1,0>UW 1UB { align1 };
+    /* Set up ss3.y coordinates */
+mov (1) g9.16<1>F g1.22<0,1,0>UW { align1 };
+mov (1) g9.20<1>F g1.22<0,1,0>UW { align1 };
+add (1) g9.24<1>F g1.22<0,1,0>UW 1UB { align1 };
+add (1) g9.28<1>F g1.22<0,1,0>UW 1UB { align1 };
+
+    /* Now, map these screen space coordinates into texture coordinates. */
+/* This is for src texture */
+/* I don't want to change origin ssX coords, as it will be used later in mask */
+/* so store tex coords in g10, g11, g12, g13 */
+
+    /* subtract screen-space X origin of vertex 0. */
+add (8) g10<1>F g6<8,8,1>F -g1<0,1,0>F { align1 };
+add (8) g11<1>F g7<8,8,1>F -g1<0,1,0>F { align1 };
+    /* scale by texture X increment */
+/* Cx[0] */
+mul (8) g10<1>F g10<8,8,1>F g3<0,1,0>F { align1 };
+mul (8) g11<1>F g11<8,8,1>F g3<0,1,0>F { align1 };
+    /* add in texture X offset */
+/* Co[0] */
+add (8) g10<1>F g10<8,8,1>F g3.12<0,1,0>F { align1 };
+add (8) g11<1>F g11<8,8,1>F g3.12<0,1,0>F { align1 };
+    /* subtract screen-space Y origin of vertex 0. */
+add (8) g12<1>F g8<8,8,1>F -g1.4<0,1,0>F { align1 };
+add (8) g13<1>F g9<8,8,1>F -g1.4<0,1,0>F { align1 };
+    /* scale by texture Y increment */
+/* Cy[0] */
+mul (8) g12<1>F g12<8,8,1>F g3.4<0,1,0>F { align1 };
+mul (8) g13<1>F g13<8,8,1>F g3.4<0,1,0>F { align1 };
+    /* add in texture Y offset */
+/* Co[1] */
+add (8) g12<1>F g12<8,8,1>F g3.28<0,1,0>F { align1 };
+add (8) g13<1>F g13<8,8,1>F g3.28<0,1,0>F { align1 };
+
+/* prepare sampler read back gX register, which would be written back to output */
+
+/* use simd16 sampler, param 0 is u, param 1 is v. */
+/* 'payload' loading, assuming tex coord start from g4 */
+mov (8) m1<1>F g10<8,8,1>F { align1 };
+mov (8) m2<1>F g11<8,8,1>F { align1 }; /* param 0 u in m1, m2 */
+mov (8) m3<1>F g12<8,8,1>F { align1 };
+mov (8) m4<1>F g13<8,8,1>F { align1 }; /* param 1 v in m3, m4 */
+
+/* m0 will be copied with g0, as it contains send desc */
+/* emit sampler 'send' cmd */
+
+/* src texture readback: g14-g21 */
+send (16) 0 		/* msg reg index */
+	g14<1>UW 	/* readback */
+	g0<8,8,1>UW  	/* copy to msg start reg*/
+	sampler (1,0,F)  /* sampler message description, 
+				(binding_table,sampler_index,datatype). 
+			    here(src->dst) we should use src_sampler and 
+			    src_surface */
+	mlen 5 rlen 8 { align1 };   /* required message len 5, readback len 8 */
+
+mov (8) g21<1>UD g21<8,8,1>UD { align1 };  /* wait sampler return */
+
+/* sampler mask texture, use g10, g11, g12, g13 */
+    /* subtract screen-space X origin of vertex 0. */
+add (8) g10<1>F g6<8,8,1>F -g1<0,1,0>F { align1 };
+add (8) g11<1>F g7<8,8,1>F -g1<0,1,0>F { align1 };
+    /* scale by texture X increment */
+/* Cx[2] */
+mul (8) g10<1>F g10<8,8,1>F g4<0,1,0>F { align1 };
+mul (8) g11<1>F g11<8,8,1>F g4<0,1,0>F { align1 };
+    /* add in texture X offset */
+/* Co[2] */
+add (8) g10<1>F g10<8,8,1>F g4.12<0,1,0>F { align1 };
+add (8) g11<1>F g11<8,8,1>F g4.12<0,1,0>F { align1 };
+    /* subtract screen-space Y origin of vertex 0. */
+add (8) g12<1>F g8<8,8,1>F -g1.4<0,1,0>F { align1 };
+add (8) g13<1>F g9<8,8,1>F -g1.4<0,1,0>F { align1 };
+    /* scale by texture Y increment */
+/* Cy[2] */
+mul (8) g12<1>F g12<8,8,1>F g4.4<0,1,0>F { align1 };
+mul (8) g13<1>F g13<8,8,1>F g4.4<0,1,0>F { align1 };
+    /* add in texture Y offset */
+/* Co[3] */
+add (8) g12<1>F g12<8,8,1>F g4.28<0,1,0>F { align1 };
+add (8) g13<1>F g13<8,8,1>F g4.28<0,1,0>F { align1 };
+
+mov (8) m1<1>F g10<8,8,1>F { align1 };
+mov (8) m2<1>F g11<8,8,1>F { align1 }; 
+mov (8) m3<1>F g12<8,8,1>F { align1 };
+mov (8) m4<1>F g13<8,8,1>F { align1 };
+
+/* mask sampler g22-g29 */
+/* binding_table (2), sampler (1) */
+send (16) 0 g22<1>UW g0<8,8,1>UW sampler (2,1,F) mlen 5 rlen 8 { align1 };
+mov (8) g29<1>UD g29<8,8,1>UD { align1 };  /* wait sampler return */
+
+/* mul mask's channel to src, then write out src */
+mul (8) g14<1>F g14<8,8,1>F g22<8,8,1>F { align1 };
+mul (8) g15<1>F g15<8,8,1>F g23<8,8,1>F { align1 };
+mul (8) g16<1>F g16<8,8,1>F g24<8,8,1>F { align1 };
+mul (8) g17<1>F g17<8,8,1>F g25<8,8,1>F { align1 };
+mul (8) g18<1>F g18<8,8,1>F g26<8,8,1>F { align1 };
+mul (8) g19<1>F g19<8,8,1>F g27<8,8,1>F { align1 };
+mul (8) g20<1>F g20<8,8,1>F g28<8,8,1>F { align1 };
+mul (8) g21<1>F g21<8,8,1>F g29<8,8,1>F { align1 };
+
+/* prepare data in m2-m5 for subspan(1,0), m6-m9 for subspan(3,2), then it's ready to write */
+mov (8) m2<1>F g14<8,8,1>F { align1 };
+mov (8) m3<1>F g16<8,8,1>F { align1 };
+mov (8) m4<1>F g18<8,8,1>F { align1 };
+mov (8) m5<1>F g20<8,8,1>F { align1 };
+mov (8) m6<1>F g15<8,8,1>F { align1 };
+mov (8) m7<1>F g17<8,8,1>F { align1 };
+mov (8) m8<1>F g19<8,8,1>F { align1 };
+mov (8) m9<1>F g21<8,8,1>F { align1 };
+
+/* m0, m1 are all direct passed by PS thread payload */
+mov (8) m1<1>UD g1<8,8,1>UD { align1 mask_disable };
+
+/* write */
+send (16) 0 acc0<1>UW g0<8,8,1>UW write (
+	0,  /* binding_table */
+	8,  /* pixel scordboard clear, msg type simd16 single source */
+	4,  /* render target write */
+	0   /* no write commit message */
+	) 
+	mlen 10
+	rlen 0
+	{ align1 EOT };
+
+nop;
+nop;
+nop;
+nop;
+nop;
+nop;
+nop;
+nop;
+nop;
diff --git a/src/exa_wm_maskca_prog.h b/src/exa_wm_maskca_prog.h
new file mode 100644
index 00000000..d936412c
--- /dev/null
+++ b/src/exa_wm_maskca_prog.h
@@ -0,0 +1,95 @@
+   { 0x00000001, 0x20c0013d, 0x00000028, 0x00000000 },
+   { 0x00000040, 0x20c40d3d, 0x00000028, 0x00000001 },
+   { 0x00000001, 0x20c8013d, 0x00000028, 0x00000000 },
+   { 0x00000040, 0x20cc0d3d, 0x00000028, 0x00000001 },
+   { 0x00000001, 0x2100013d, 0x0000002a, 0x00000000 },
+   { 0x00000001, 0x2104013d, 0x0000002a, 0x00000000 },
+   { 0x00000040, 0x21080d3d, 0x0000002a, 0x00000001 },
+   { 0x00000040, 0x210c0d3d, 0x0000002a, 0x00000001 },
+   { 0x00000001, 0x20d0013d, 0x0000002c, 0x00000000 },
+   { 0x00000040, 0x20d40d3d, 0x0000002c, 0x00000001 },
+   { 0x00000001, 0x20d8013d, 0x0000002c, 0x00000000 },
+   { 0x00000040, 0x20dc0d3d, 0x0000002c, 0x00000001 },
+   { 0x00000001, 0x2110013d, 0x0000002e, 0x00000000 },
+   { 0x00000001, 0x2114013d, 0x0000002e, 0x00000000 },
+   { 0x00000040, 0x21180d3d, 0x0000002e, 0x00000001 },
+   { 0x00000040, 0x211c0d3d, 0x0000002e, 0x00000001 },
+   { 0x00000001, 0x20e0013d, 0x00000030, 0x00000000 },
+   { 0x00000040, 0x20e40d3d, 0x00000030, 0x00000001 },
+   { 0x00000001, 0x20e8013d, 0x00000030, 0x00000000 },
+   { 0x00000040, 0x20ec0d3d, 0x00000030, 0x00000001 },
+   { 0x00000001, 0x2120013d, 0x00000032, 0x00000000 },
+   { 0x00000001, 0x2124013d, 0x00000032, 0x00000000 },
+   { 0x00000040, 0x21280d3d, 0x00000032, 0x00000001 },
+   { 0x00000040, 0x212c0d3d, 0x00000032, 0x00000001 },
+   { 0x00000001, 0x20f0013d, 0x00000034, 0x00000000 },
+   { 0x00000040, 0x20f40d3d, 0x00000034, 0x00000001 },
+   { 0x00000001, 0x20f8013d, 0x00000034, 0x00000000 },
+   { 0x00000040, 0x20fc0d3d, 0x00000034, 0x00000001 },
+   { 0x00000001, 0x2130013d, 0x00000036, 0x00000000 },
+   { 0x00000001, 0x2134013d, 0x00000036, 0x00000000 },
+   { 0x00000040, 0x21380d3d, 0x00000036, 0x00000001 },
+   { 0x00000040, 0x213c0d3d, 0x00000036, 0x00000001 },
+   { 0x00600040, 0x214077bd, 0x008d00c0, 0x00004020 },
+   { 0x00600040, 0x216077bd, 0x008d00e0, 0x00004020 },
+   { 0x00600041, 0x214077bd, 0x008d0140, 0x00000060 },
+   { 0x00600041, 0x216077bd, 0x008d0160, 0x00000060 },
+   { 0x00600040, 0x214077bd, 0x008d0140, 0x0000006c },
+   { 0x00600040, 0x216077bd, 0x008d0160, 0x0000006c },
+   { 0x00600040, 0x218077bd, 0x008d0100, 0x00004024 },
+   { 0x00600040, 0x21a077bd, 0x008d0120, 0x00004024 },
+   { 0x00600041, 0x218077bd, 0x008d0180, 0x00000064 },
+   { 0x00600041, 0x21a077bd, 0x008d01a0, 0x00000064 },
+   { 0x00600040, 0x218077bd, 0x008d0180, 0x0000007c },
+   { 0x00600040, 0x21a077bd, 0x008d01a0, 0x0000007c },
+   { 0x00600001, 0x202003be, 0x008d0140, 0x00000000 },
+   { 0x00600001, 0x204003be, 0x008d0160, 0x00000000 },
+   { 0x00600001, 0x206003be, 0x008d0180, 0x00000000 },
+   { 0x00600001, 0x208003be, 0x008d01a0, 0x00000000 },
+   { 0x00800031, 0x21c01d29, 0x008d0000, 0x02580001 },
+   { 0x00600001, 0x22a00021, 0x008d02a0, 0x00000000 },
+   { 0x00600040, 0x214077bd, 0x008d00c0, 0x00004020 },
+   { 0x00600040, 0x216077bd, 0x008d00e0, 0x00004020 },
+   { 0x00600041, 0x214077bd, 0x008d0140, 0x00000080 },
+   { 0x00600041, 0x216077bd, 0x008d0160, 0x00000080 },
+   { 0x00600040, 0x214077bd, 0x008d0140, 0x0000008c },
+   { 0x00600040, 0x216077bd, 0x008d0160, 0x0000008c },
+   { 0x00600040, 0x218077bd, 0x008d0100, 0x00004024 },
+   { 0x00600040, 0x21a077bd, 0x008d0120, 0x00004024 },
+   { 0x00600041, 0x218077bd, 0x008d0180, 0x00000084 },
+   { 0x00600041, 0x21a077bd, 0x008d01a0, 0x00000084 },
+   { 0x00600040, 0x218077bd, 0x008d0180, 0x0000009c },
+   { 0x00600040, 0x21a077bd, 0x008d01a0, 0x0000009c },
+   { 0x00600001, 0x202003be, 0x008d0140, 0x00000000 },
+   { 0x00600001, 0x204003be, 0x008d0160, 0x00000000 },
+   { 0x00600001, 0x206003be, 0x008d0180, 0x00000000 },
+   { 0x00600001, 0x208003be, 0x008d01a0, 0x00000000 },
+   { 0x00800031, 0x22c01d29, 0x008d0000, 0x02580102 },
+   { 0x00600001, 0x23a00021, 0x008d03a0, 0x00000000 },
+   { 0x00600041, 0x21c077bd, 0x008d01c0, 0x008d02c0 },
+   { 0x00600041, 0x21e077bd, 0x008d01e0, 0x008d02e0 },
+   { 0x00600041, 0x220077bd, 0x008d0200, 0x008d0300 },
+   { 0x00600041, 0x222077bd, 0x008d0220, 0x008d0320 },
+   { 0x00600041, 0x224077bd, 0x008d0240, 0x008d0340 },
+   { 0x00600041, 0x226077bd, 0x008d0260, 0x008d0360 },
+   { 0x00600041, 0x228077bd, 0x008d0280, 0x008d0380 },
+   { 0x00600041, 0x22a077bd, 0x008d02a0, 0x008d03a0 },
+   { 0x00600001, 0x204003be, 0x008d01c0, 0x00000000 },
+   { 0x00600001, 0x206003be, 0x008d0200, 0x00000000 },
+   { 0x00600001, 0x208003be, 0x008d0240, 0x00000000 },
+   { 0x00600001, 0x20a003be, 0x008d0280, 0x00000000 },
+   { 0x00600001, 0x20c003be, 0x008d01e0, 0x00000000 },
+   { 0x00600001, 0x20e003be, 0x008d0220, 0x00000000 },
+   { 0x00600001, 0x210003be, 0x008d0260, 0x00000000 },
+   { 0x00600001, 0x212003be, 0x008d02a0, 0x00000000 },
+   { 0x00600201, 0x20200022, 0x008d0020, 0x00000000 },
+   { 0x00800031, 0x24001d28, 0x008d0000, 0x85a04800 },
+   { 0x0000007e, 0x00000000, 0x00000000, 0x00000000 },
+   { 0x0000007e, 0x00000000, 0x00000000, 0x00000000 },
+   { 0x0000007e, 0x00000000, 0x00000000, 0x00000000 },
+   { 0x0000007e, 0x00000000, 0x00000000, 0x00000000 },
+   { 0x0000007e, 0x00000000, 0x00000000, 0x00000000 },
+   { 0x0000007e, 0x00000000, 0x00000000, 0x00000000 },
+   { 0x0000007e, 0x00000000, 0x00000000, 0x00000000 },
+   { 0x0000007e, 0x00000000, 0x00000000, 0x00000000 },
+   { 0x0000007e, 0x00000000, 0x00000000, 0x00000000 },
diff --git a/src/exa_wm_maskca_srcalpha.g4a b/src/exa_wm_maskca_srcalpha.g4a
new file mode 100644
index 00000000..e233968c
--- /dev/null
+++ b/src/exa_wm_maskca_srcalpha.g4a
@@ -0,0 +1,202 @@
+/*
+ * This's for exa composite operation in no mask picture case.
+ * The simplest case is just sending what src picture has to dst picture.
+ * XXX: This is still experimental, and should be fixed to support multiple texture
+ * map, and conditional mul actions. 
+ */
+
+/* I think this should be same as in g4a program for texture video,
+   as we also use 16-pixel dispatch. and SF scale in g3 is useful for us. */
+
+/* The initial payload of the thread is always g0.
+ * WM_URB (incoming URB entries) is g3
+   As mask texture coeffient needs extra setup urb starting from g4, we should
+   shift this location. 
+
+ * X0_R is g4->g6
+ * X1_R is g5->g7
+ * Y0_R is g6->g8
+ * Y1_R is g7->g9
+
+     * X0: {ss0.x, ss0.x+1, ss0.x,   ss0.x+1, ss1.x, ss1.x+1, ss1.x,   ss1.x+y}
+     * Y0: {ss0.y, ss0.y,   ss0.y+1, ss0.y+1, ss1.y, ss1.y,   ss1.y+1, ss1.y+1}
+     * X1: {ss2.x, ss2.x+1, ss2.x,   ss2.x+1, ss3.x, ss3.x+1, ss3.x,   ss3.x+y}
+     * Y1: {ss2.y, ss2.y,   ss2.y+1, ss2.y+1, ss3.y, ss3.y,   ss3.y+1, ss3.y+1}
+ */
+
+/* multitexture program with src and mask texture */
+/* - load src texture */
+/* - load mask texture */
+/* - mul src.X with mask's alpha */
+/* - write out src.X */
+
+    /* Set up ss0.x coordinates*/
+mov (1) g6<1>F g1.8<0,1,0>UW { align1 };
+add (1) g6.4<1>F g1.8<0,1,0>UW 1UB { align1 };
+mov (1) g6.8<1>F g1.8<0,1,0>UW { align1 };
+add (1) g6.12<1>F g1.8<0,1,0>UW 1UB { align1 };
+    /* Set up ss0.y coordinates */
+mov (1) g8<1>F g1.10<0,1,0>UW { align1 };
+mov (1) g8.4<1>F g1.10<0,1,0>UW { align1 };
+add (1) g8.8<1>F g1.10<0,1,0>UW 1UB { align1 };
+add (1) g8.12<1>F g1.10<0,1,0>UW 1UB { align1 };
+    /* set up ss1.x coordinates */
+mov (1) g6.16<1>F g1.12<0,1,0>UW { align1 };
+add (1) g6.20<1>F g1.12<0,1,0>UW 1UB { align1 };
+mov (1) g6.24<1>F g1.12<0,1,0>UW { align1 };
+add (1) g6.28<1>F g1.12<0,1,0>UW 1UB { align1 };
+    /* set up ss1.y coordinates */
+mov (1) g8.16<1>F g1.14<0,1,0>UW { align1 };
+mov (1) g8.20<1>F g1.14<0,1,0>UW { align1 };
+add (1) g8.24<1>F g1.14<0,1,0>UW 1UB { align1 };
+add (1) g8.28<1>F g1.14<0,1,0>UW 1UB { align1 };
+    /* Set up ss2.x coordinates */
+mov (1) g7<1>F g1.16<0,1,0>UW { align1 };
+add (1) g7.4<1>F g1.16<0,1,0>UW 1UB { align1 };
+mov (1) g7.8<1>F g1.16<0,1,0>UW { align1 };
+add (1) g7.12<1>F g1.16<0,1,0>UW 1UB { align1 };
+    /* Set up ss2.y coordinates */
+mov (1) g9<1>F g1.18<0,1,0>UW { align1 };
+mov (1) g9.4<1>F g1.18<0,1,0>UW { align1 };
+add (1) g9.8<1>F g1.18<0,1,0>UW 1UB { align1 };
+add (1) g9.12<1>F g1.18<0,1,0>UW 1UB { align1 };
+    /* Set up ss3.x coordinates */
+mov (1) g7.16<1>F g1.20<0,1,0>UW { align1 };
+add (1) g7.20<1>F g1.20<0,1,0>UW 1UB { align1 };
+mov (1) g7.24<1>F g1.20<0,1,0>UW { align1 };
+add (1) g7.28<1>F g1.20<0,1,0>UW 1UB { align1 };
+    /* Set up ss3.y coordinates */
+mov (1) g9.16<1>F g1.22<0,1,0>UW { align1 };
+mov (1) g9.20<1>F g1.22<0,1,0>UW { align1 };
+add (1) g9.24<1>F g1.22<0,1,0>UW 1UB { align1 };
+add (1) g9.28<1>F g1.22<0,1,0>UW 1UB { align1 };
+
+    /* Now, map these screen space coordinates into texture coordinates. */
+/* This is for src texture */
+/* I don't want to change origin ssX coords, as it will be used later in mask */
+/* so store tex coords in g10, g11, g12, g13 */
+
+    /* subtract screen-space X origin of vertex 0. */
+add (8) g10<1>F g6<8,8,1>F -g1<0,1,0>F { align1 };
+add (8) g11<1>F g7<8,8,1>F -g1<0,1,0>F { align1 };
+    /* scale by texture X increment */
+/* Cx[0] */
+mul (8) g10<1>F g10<8,8,1>F g3<0,1,0>F { align1 };
+mul (8) g11<1>F g11<8,8,1>F g3<0,1,0>F { align1 };
+    /* add in texture X offset */
+/* Co[0] */
+add (8) g10<1>F g10<8,8,1>F g3.12<0,1,0>F { align1 };
+add (8) g11<1>F g11<8,8,1>F g3.12<0,1,0>F { align1 };
+    /* subtract screen-space Y origin of vertex 0. */
+add (8) g12<1>F g8<8,8,1>F -g1.4<0,1,0>F { align1 };
+add (8) g13<1>F g9<8,8,1>F -g1.4<0,1,0>F { align1 };
+    /* scale by texture Y increment */
+/* Cy[0] */
+mul (8) g12<1>F g12<8,8,1>F g3.4<0,1,0>F { align1 };
+mul (8) g13<1>F g13<8,8,1>F g3.4<0,1,0>F { align1 };
+    /* add in texture Y offset */
+/* Co[1] */
+add (8) g12<1>F g12<8,8,1>F g3.28<0,1,0>F { align1 };
+add (8) g13<1>F g13<8,8,1>F g3.28<0,1,0>F { align1 };
+
+/* prepare sampler read back gX register, which would be written back to output */
+
+/* use simd16 sampler, param 0 is u, param 1 is v. */
+/* 'payload' loading, assuming tex coord start from g4 */
+mov (8) m1<1>F g10<8,8,1>F { align1 };
+mov (8) m2<1>F g11<8,8,1>F { align1 }; /* param 0 u in m1, m2 */
+mov (8) m3<1>F g12<8,8,1>F { align1 };
+mov (8) m4<1>F g13<8,8,1>F { align1 }; /* param 1 v in m3, m4 */
+
+/* m0 will be copied with g0, as it contains send desc */
+/* emit sampler 'send' cmd */
+
+/* src texture readback: g14-g21 */
+send (16) 0 		/* msg reg index */
+	g14<1>UW 	/* readback */
+	g0<8,8,1>UW  	/* copy to msg start reg*/
+	sampler (1,0,F)  /* sampler message description, 
+				(binding_table,sampler_index,datatype). 
+			    here(src->dst) we should use src_sampler and 
+			    src_surface */
+	mlen 5 rlen 8 { align1 };   /* required message len 5, readback len 8 */
+
+mov (8) g21<1>UD g21<8,8,1>UD { align1 };  /* wait sampler return */
+
+/* sampler mask texture, use g10, g11, g12, g13 */
+    /* subtract screen-space X origin of vertex 0. */
+add (8) g10<1>F g6<8,8,1>F -g1<0,1,0>F { align1 };
+add (8) g11<1>F g7<8,8,1>F -g1<0,1,0>F { align1 };
+    /* scale by texture X increment */
+/* Cx[2] */
+mul (8) g10<1>F g10<8,8,1>F g4<0,1,0>F { align1 };
+mul (8) g11<1>F g11<8,8,1>F g4<0,1,0>F { align1 };
+    /* add in texture X offset */
+/* Co[2] */
+add (8) g10<1>F g10<8,8,1>F g4.12<0,1,0>F { align1 };
+add (8) g11<1>F g11<8,8,1>F g4.12<0,1,0>F { align1 };
+    /* subtract screen-space Y origin of vertex 0. */
+add (8) g12<1>F g8<8,8,1>F -g1.4<0,1,0>F { align1 };
+add (8) g13<1>F g9<8,8,1>F -g1.4<0,1,0>F { align1 };
+    /* scale by texture Y increment */
+/* Cy[2] */
+mul (8) g12<1>F g12<8,8,1>F g4.4<0,1,0>F { align1 };
+mul (8) g13<1>F g13<8,8,1>F g4.4<0,1,0>F { align1 };
+    /* add in texture Y offset */
+/* Co[3] */
+add (8) g12<1>F g12<8,8,1>F g4.28<0,1,0>F { align1 };
+add (8) g13<1>F g13<8,8,1>F g4.28<0,1,0>F { align1 };
+
+mov (8) m1<1>F g10<8,8,1>F { align1 };
+mov (8) m2<1>F g11<8,8,1>F { align1 }; 
+mov (8) m3<1>F g12<8,8,1>F { align1 };
+mov (8) m4<1>F g13<8,8,1>F { align1 };
+
+/* mask sampler g22-g29 */
+/* binding_table (2), sampler (1) */
+send (16) 0 g22<1>UW g0<8,8,1>UW sampler (2,1,F) mlen 5 rlen 8 { align1 };
+mov (8) g29<1>UD g29<8,8,1>UD { align1 };  /* wait sampler return */
+
+/* src channel has no more use, src.A * mask.C */
+mul (8) g14<1>F g22<8,8,1>F g20<8,8,1>F { align1 };
+mul (8) g15<1>F g23<8,8,1>F g21<8,8,1>F { align1 };
+mul (8) g16<1>F g24<8,8,1>F g20<8,8,1>F { align1 };
+mul (8) g17<1>F g25<8,8,1>F g21<8,8,1>F { align1 };
+mul (8) g18<1>F g26<8,8,1>F g20<8,8,1>F { align1 };
+mul (8) g19<1>F g27<8,8,1>F g21<8,8,1>F { align1 };
+mul (8) g20<1>F g28<8,8,1>F g20<8,8,1>F { align1 };
+mul (8) g21<1>F g29<8,8,1>F g21<8,8,1>F { align1 };
+
+/* prepare data in m2-m5 for subspan(1,0), m6-m9 for subspan(3,2), then it's ready to write */
+mov (8) m2<1>F g14<8,8,1>F { align1 };
+mov (8) m3<1>F g16<8,8,1>F { align1 };
+mov (8) m4<1>F g18<8,8,1>F { align1 };
+mov (8) m5<1>F g20<8,8,1>F { align1 };
+mov (8) m6<1>F g15<8,8,1>F { align1 };
+mov (8) m7<1>F g17<8,8,1>F { align1 };
+mov (8) m8<1>F g19<8,8,1>F { align1 };
+mov (8) m9<1>F g21<8,8,1>F { align1 };
+
+/* m0, m1 are all direct passed by PS thread payload */
+mov (8) m1<1>UD g1<8,8,1>UD { align1 mask_disable };
+
+/* write */
+send (16) 0 acc0<1>UW g0<8,8,1>UW write (
+	0,  /* binding_table */
+	8,  /* pixel scordboard clear, msg type simd16 single source */
+	4,  /* render target write */
+	0   /* no write commit message */
+	) 
+	mlen 10
+	rlen 0
+	{ align1 EOT };
+
+nop;
+nop;
+nop;
+nop;
+nop;
+nop;
+nop;
+nop;
+nop;
diff --git a/src/exa_wm_maskca_srcalpha_prog.h b/src/exa_wm_maskca_srcalpha_prog.h
new file mode 100644
index 00000000..d83b119f
--- /dev/null
+++ b/src/exa_wm_maskca_srcalpha_prog.h
@@ -0,0 +1,95 @@
+   { 0x00000001, 0x20c0013d, 0x00000028, 0x00000000 },
+   { 0x00000040, 0x20c40d3d, 0x00000028, 0x00000001 },
+   { 0x00000001, 0x20c8013d, 0x00000028, 0x00000000 },
+   { 0x00000040, 0x20cc0d3d, 0x00000028, 0x00000001 },
+   { 0x00000001, 0x2100013d, 0x0000002a, 0x00000000 },
+   { 0x00000001, 0x2104013d, 0x0000002a, 0x00000000 },
+   { 0x00000040, 0x21080d3d, 0x0000002a, 0x00000001 },
+   { 0x00000040, 0x210c0d3d, 0x0000002a, 0x00000001 },
+   { 0x00000001, 0x20d0013d, 0x0000002c, 0x00000000 },
+   { 0x00000040, 0x20d40d3d, 0x0000002c, 0x00000001 },
+   { 0x00000001, 0x20d8013d, 0x0000002c, 0x00000000 },
+   { 0x00000040, 0x20dc0d3d, 0x0000002c, 0x00000001 },
+   { 0x00000001, 0x2110013d, 0x0000002e, 0x00000000 },
+   { 0x00000001, 0x2114013d, 0x0000002e, 0x00000000 },
+   { 0x00000040, 0x21180d3d, 0x0000002e, 0x00000001 },
+   { 0x00000040, 0x211c0d3d, 0x0000002e, 0x00000001 },
+   { 0x00000001, 0x20e0013d, 0x00000030, 0x00000000 },
+   { 0x00000040, 0x20e40d3d, 0x00000030, 0x00000001 },
+   { 0x00000001, 0x20e8013d, 0x00000030, 0x00000000 },
+   { 0x00000040, 0x20ec0d3d, 0x00000030, 0x00000001 },
+   { 0x00000001, 0x2120013d, 0x00000032, 0x00000000 },
+   { 0x00000001, 0x2124013d, 0x00000032, 0x00000000 },
+   { 0x00000040, 0x21280d3d, 0x00000032, 0x00000001 },
+   { 0x00000040, 0x212c0d3d, 0x00000032, 0x00000001 },
+   { 0x00000001, 0x20f0013d, 0x00000034, 0x00000000 },
+   { 0x00000040, 0x20f40d3d, 0x00000034, 0x00000001 },
+   { 0x00000001, 0x20f8013d, 0x00000034, 0x00000000 },
+   { 0x00000040, 0x20fc0d3d, 0x00000034, 0x00000001 },
+   { 0x00000001, 0x2130013d, 0x00000036, 0x00000000 },
+   { 0x00000001, 0x2134013d, 0x00000036, 0x00000000 },
+   { 0x00000040, 0x21380d3d, 0x00000036, 0x00000001 },
+   { 0x00000040, 0x213c0d3d, 0x00000036, 0x00000001 },
+   { 0x00600040, 0x214077bd, 0x008d00c0, 0x00004020 },
+   { 0x00600040, 0x216077bd, 0x008d00e0, 0x00004020 },
+   { 0x00600041, 0x214077bd, 0x008d0140, 0x00000060 },
+   { 0x00600041, 0x216077bd, 0x008d0160, 0x00000060 },
+   { 0x00600040, 0x214077bd, 0x008d0140, 0x0000006c },
+   { 0x00600040, 0x216077bd, 0x008d0160, 0x0000006c },
+   { 0x00600040, 0x218077bd, 0x008d0100, 0x00004024 },
+   { 0x00600040, 0x21a077bd, 0x008d0120, 0x00004024 },
+   { 0x00600041, 0x218077bd, 0x008d0180, 0x00000064 },
+   { 0x00600041, 0x21a077bd, 0x008d01a0, 0x00000064 },
+   { 0x00600040, 0x218077bd, 0x008d0180, 0x0000007c },
+   { 0x00600040, 0x21a077bd, 0x008d01a0, 0x0000007c },
+   { 0x00600001, 0x202003be, 0x008d0140, 0x00000000 },
+   { 0x00600001, 0x204003be, 0x008d0160, 0x00000000 },
+   { 0x00600001, 0x206003be, 0x008d0180, 0x00000000 },
+   { 0x00600001, 0x208003be, 0x008d01a0, 0x00000000 },
+   { 0x00800031, 0x21c01d29, 0x008d0000, 0x02580001 },
+   { 0x00600001, 0x22a00021, 0x008d02a0, 0x00000000 },
+   { 0x00600040, 0x214077bd, 0x008d00c0, 0x00004020 },
+   { 0x00600040, 0x216077bd, 0x008d00e0, 0x00004020 },
+   { 0x00600041, 0x214077bd, 0x008d0140, 0x00000080 },
+   { 0x00600041, 0x216077bd, 0x008d0160, 0x00000080 },
+   { 0x00600040, 0x214077bd, 0x008d0140, 0x0000008c },
+   { 0x00600040, 0x216077bd, 0x008d0160, 0x0000008c },
+   { 0x00600040, 0x218077bd, 0x008d0100, 0x00004024 },
+   { 0x00600040, 0x21a077bd, 0x008d0120, 0x00004024 },
+   { 0x00600041, 0x218077bd, 0x008d0180, 0x00000084 },
+   { 0x00600041, 0x21a077bd, 0x008d01a0, 0x00000084 },
+   { 0x00600040, 0x218077bd, 0x008d0180, 0x0000009c },
+   { 0x00600040, 0x21a077bd, 0x008d01a0, 0x0000009c },
+   { 0x00600001, 0x202003be, 0x008d0140, 0x00000000 },
+   { 0x00600001, 0x204003be, 0x008d0160, 0x00000000 },
+   { 0x00600001, 0x206003be, 0x008d0180, 0x00000000 },
+   { 0x00600001, 0x208003be, 0x008d01a0, 0x00000000 },
+   { 0x00800031, 0x22c01d29, 0x008d0000, 0x02580102 },
+   { 0x00600001, 0x23a00021, 0x008d03a0, 0x00000000 },
+   { 0x00600041, 0x21c077bd, 0x008d02c0, 0x008d0280 },
+   { 0x00600041, 0x21e077bd, 0x008d02e0, 0x008d02a0 },
+   { 0x00600041, 0x220077bd, 0x008d0300, 0x008d0280 },
+   { 0x00600041, 0x222077bd, 0x008d0320, 0x008d02a0 },
+   { 0x00600041, 0x224077bd, 0x008d0340, 0x008d0280 },
+   { 0x00600041, 0x226077bd, 0x008d0360, 0x008d02a0 },
+   { 0x00600041, 0x228077bd, 0x008d0380, 0x008d0280 },
+   { 0x00600041, 0x22a077bd, 0x008d03a0, 0x008d02a0 },
+   { 0x00600001, 0x204003be, 0x008d01c0, 0x00000000 },
+   { 0x00600001, 0x206003be, 0x008d0200, 0x00000000 },
+   { 0x00600001, 0x208003be, 0x008d0240, 0x00000000 },
+   { 0x00600001, 0x20a003be, 0x008d0280, 0x00000000 },
+   { 0x00600001, 0x20c003be, 0x008d01e0, 0x00000000 },
+   { 0x00600001, 0x20e003be, 0x008d0220, 0x00000000 },
+   { 0x00600001, 0x210003be, 0x008d0260, 0x00000000 },
+   { 0x00600001, 0x212003be, 0x008d02a0, 0x00000000 },
+   { 0x00600201, 0x20200022, 0x008d0020, 0x00000000 },
+   { 0x00800031, 0x24001d28, 0x008d0000, 0x85a04800 },
+   { 0x0000007e, 0x00000000, 0x00000000, 0x00000000 },
+   { 0x0000007e, 0x00000000, 0x00000000, 0x00000000 },
+   { 0x0000007e, 0x00000000, 0x00000000, 0x00000000 },
+   { 0x0000007e, 0x00000000, 0x00000000, 0x00000000 },
+   { 0x0000007e, 0x00000000, 0x00000000, 0x00000000 },
+   { 0x0000007e, 0x00000000, 0x00000000, 0x00000000 },
+   { 0x0000007e, 0x00000000, 0x00000000, 0x00000000 },
+   { 0x0000007e, 0x00000000, 0x00000000, 0x00000000 },
+   { 0x0000007e, 0x00000000, 0x00000000, 0x00000000 },
diff --git a/src/i965_render.c b/src/i965_render.c
index a2b21db1..8474c483 100644
--- a/src/i965_render.c
+++ b/src/i965_render.c
@@ -133,7 +133,8 @@ static void i965_get_blend_cntl(int op, PicturePtr pMask, CARD32 dst_format,
      * the source blend factor is 0, and the source blend value is the mask
      * channels multiplied by the source picture's alpha.
      */
-    if (pMask && pMask->componentAlpha && i965_blend_op[op].src_alpha) {
+    if (pMask && pMask->componentAlpha && PICT_FORMAT_RGB(pMask->format)
+            && i965_blend_op[op].src_alpha) {
         if (*dblend == BRW_BLENDFACTOR_SRC_ALPHA) {
 	    *dblend = BRW_BLENDFACTOR_SRC_COLOR;
         } else if (*dblend == BRW_BLENDFACTOR_INV_SRC_ALPHA) {
@@ -223,7 +224,8 @@ i965_check_composite(int op, PicturePtr pSrcPicture, PicturePtr pMaskPicture,
     if (op >= sizeof(i965_blend_op) / sizeof(i965_blend_op[0]))
         I830FALLBACK("Unsupported Composite op 0x%x\n", op);
 
-    if (pMaskPicture != NULL && pMaskPicture->componentAlpha) {
+    if (pMaskPicture && pMaskPicture->componentAlpha &&
+            PICT_FORMAT_RGB(pMaskPicture->format)) {
         /* Check if it's component alpha that relies on a source alpha and on
          * the source value.  We can only get one of those into the single
          * source value that we get to blend with.
@@ -234,9 +236,7 @@ i965_check_composite(int op, PicturePtr pSrcPicture, PicturePtr pMaskPicture,
 	    I830FALLBACK("Component alpha not supported with source "
 			 "alpha and source value blending.\n");
 	}
-	/* XXX: fallback now for mask with componentAlpha */
-	I830FALLBACK("mask componentAlpha not ready.\n");
-    }
+    } 
 
     if (!i965_check_composite_texture(pSrcPicture, 0))
         I830FALLBACK("Check Src picture texture\n");
@@ -337,17 +337,19 @@ static const CARD32 sf_kernel_static_mask[][4] = {
 /* ps kernels */
 #define PS_KERNEL_NUM_GRF   32
 #define PS_MAX_THREADS	   32
-/* 1: no mask */
+
 static const CARD32 ps_kernel_static_nomask [][4] = {
 #include "exa_wm_nomask_prog.h"
 };
 
-/* 2: mask with componentAlpha, src * mask color, XXX: later */
 static const CARD32 ps_kernel_static_maskca [][4] = {
-/*#include "i965_composite_wm_maskca.h" */
+#include "exa_wm_maskca_prog.h"
+};
+
+static const CARD32 ps_kernel_static_maskca_srcalpha [][4] = {
+#include "exa_wm_maskca_srcalpha_prog.h"
 };
 
-/* 3: mask without componentAlpha, src * mask alpha */
 static const CARD32 ps_kernel_static_masknoca [][4] = {
 #include "exa_wm_masknoca_prog.h"
 };
@@ -442,13 +444,20 @@ i965_prepare_composite(int op, PicturePtr pSrcPicture,
     else
 	next_offset = sf_kernel_offset + sizeof (sf_kernel_static);
 
-    /* XXX: ps_kernel may be seperated, fix with offset */
     ps_kernel_offset = ALIGN(next_offset, 64);
     if (pMask) {
-	if (pMaskPicture->componentAlpha)
-	    next_offset = ps_kernel_offset + sizeof(ps_kernel_static_maskca);
-	else
-	    next_offset = ps_kernel_offset + sizeof(ps_kernel_static_masknoca);
+	if (pMaskPicture->componentAlpha && 
+                PICT_FORMAT_RGB(pMaskPicture->format)) {
+            if (i965_blend_op[op].src_alpha) {
+                next_offset = ps_kernel_offset + 
+                    sizeof(ps_kernel_static_maskca_srcalpha);
+            } else {
+                next_offset = ps_kernel_offset + 
+                    sizeof(ps_kernel_static_maskca);
+            }
+        } else
+	    next_offset = ps_kernel_offset + 
+                          sizeof(ps_kernel_static_masknoca);
     } else {
    	next_offset = ps_kernel_offset + sizeof (ps_kernel_static_nomask);
     }
@@ -746,7 +755,6 @@ i965_prepare_composite(int op, PicturePtr pSrcPicture,
     vs_state->vs6.vs_enable = 0;
     vs_state->vs6.vert_cache_disable = 1;
 
-    /* XXX: sf_kernel? keep it as now */
     /* Set up the SF kernel to do coord interp: for each attribute,
      * calculate dA/dx and dA/dy.  Hand these interpolation coefficients
      * back to SF which then hands pixels off to WM.
@@ -788,10 +796,15 @@ i965_prepare_composite(int op, PicturePtr pSrcPicture,
 
    /* Set up the PS kernel (dispatched by WM) */
     if (pMask) {
-	if (pMaskPicture->componentAlpha)
-   	    memcpy(ps_kernel, ps_kernel_static_maskca,
-		   sizeof (ps_kernel_static_maskca));
-	else
+	if (pMaskPicture->componentAlpha && 
+                PICT_FORMAT_RGB(pMaskPicture->format)) {
+            if (i965_blend_op[op].src_alpha) 
+                memcpy(ps_kernel, ps_kernel_static_maskca_srcalpha,
+                        sizeof (ps_kernel_static_maskca_srcalpha));
+            else
+                memcpy(ps_kernel, ps_kernel_static_maskca,
+                        sizeof (ps_kernel_static_maskca));
+        } else
    	    memcpy(ps_kernel, ps_kernel_static_masknoca,
 		   sizeof (ps_kernel_static_masknoca));
     } else {
@@ -812,7 +825,6 @@ i965_prepare_composite(int op, PicturePtr pSrcPicture,
     wm_state->thread2.scratch_space_base_pointer = (state_base_offset +
 						    wm_scratch_offset)>>10;
     wm_state->thread2.per_thread_scratch_space = 0;
-    // XXX: urb allocation
     wm_state->thread3.const_urb_entry_read_length = 0;
     wm_state->thread3.const_urb_entry_read_offset = 0;
     /* Each pair of attributes (src/mask coords) is one URB entry */