Add simplest wm kernel program for no mask picture composite

This is a try to use new gen4asm language, and will finish
composite program for mask picture with or without CA case later.

2 files changed, 141 insertions, 0 deletions

diff --git a/src/Makefile.am b/src/Makefile.am
index f08d9a7a..e7132bc6 100644
--- a/src/Makefile.am
+++ b/src/Makefile.am
@@ -77,6 +77,8 @@ sf_prog.h: packed_yuv_sf.g4a
 	intel-gen4asm -o sf_prog.h packed_yuv_sf.g4a
 wm_prog.h: packed_yuv_wm.g4a
 	intel-gen4asm -o wm_prog.h packed_yuv_wm.g4a
+i965_composite_wm_nomask.h: i965_composite_wm_nomask.g4a
+	intel-gen4asm -o i965_composite_wm_nomask.h i965_composite_wm_nomask.g4a
 endif
 
 if DRI
diff --git a/src/i965_composite_wm_nomask.g4a b/src/i965_composite_wm_nomask.g4a
new file mode 100644
index 00000000..87916318
--- /dev/null
+++ b/src/i965_composite_wm_nomask.g4a
@@ -0,0 +1,139 @@
+/*
+ * This's for exa composite operation in no mask picture case.
+ * The simplest case is just sending what src picture has to dst picture
+ */
+
+/* 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
+ * X0_R is g4
+ * X1_R is g5
+ * Y0_R is g6
+ * Y1_R is g7
+ */
+
+    /* Set up ss0.x coordinates*/
+mov (1) g4<1>F g1.8<0,1,0>UW { align1 };
+add (1) g4.4<1>F g1.8<0,1,0>UW 1UB { align1 };
+mov (1) g4.8<1>F g1.8<0,1,0>UW { align1 };
+add (1) g4.12<1>F g1.8<0,1,0>UW 1UB { align1 };
+    /* Set up ss0.y coordinates */
+mov (1) g6<1>F g1.10<0,1,0>UW { align1 };
+mov (1) g6.4<1>F g1.10<0,1,0>UW { align1 };
+add (1) g6.8<1>F g1.10<0,1,0>UW 1UB { align1 };
+add (1) g6.12<1>F g1.10<0,1,0>UW 1UB { align1 };
+    /* set up ss1.x coordinates */
+mov (1) g4.16<1>F g1.12<0,1,0>UW { align1 };
+add (1) g4.20<1>F g1.12<0,1,0>UW 1UB { align1 };
+mov (1) g4.24<1>F g1.12<0,1,0>UW { align1 };
+add (1) g4.28<1>F g1.12<0,1,0>UW 1UB { align1 };
+    /* set up ss1.y coordinates */
+mov (1) g6.16<1>F g1.14<0,1,0>UW { align1 };
+mov (1) g6.20<1>F g1.14<0,1,0>UW { align1 };
+add (1) g6.24<1>F g1.14<0,1,0>UW 1UB { align1 };
+add (1) g6.28<1>F g1.14<0,1,0>UW 1UB { align1 };
+    /* Set up ss2.x coordinates */
+mov (1) g5<1>F g1.16<0,1,0>UW { align1 };
+add (1) g5.4<1>F g1.16<0,1,0>UW 1UB { align1 };
+mov (1) g5.8<1>F g1.16<0,1,0>UW { align1 };
+add (1) g5.12<1>F g1.16<0,1,0>UW 1UB { align1 };
+    /* Set up ss2.y coordinates */
+mov (1) g7<1>F g1.18<0,1,0>UW { align1 };
+mov (1) g7.4<1>F g1.18<0,1,0>UW { align1 };
+add (1) g7.8<1>F g1.18<0,1,0>UW 1UB { align1 };
+add (1) g7.12<1>F g1.18<0,1,0>UW 1UB { align1 };
+    /* Set up ss3.x coordinates */
+mov (1) g5.16<1>F g1.20<0,1,0>UW { align1 };
+add (1) g5.20<1>F g1.20<0,1,0>UW 1UB { align1 };
+mov (1) g5.24<1>F g1.20<0,1,0>UW { align1 };
+add (1) g5.28<1>F g1.20<0,1,0>UW 1UB { align1 };
+    /* Set up ss3.y coordinates */
+mov (1) g7.16<1>F g1.22<0,1,0>UW { align1 };
+mov (1) g7.20<1>F g1.22<0,1,0>UW { align1 };
+add (1) g7.24<1>F g1.22<0,1,0>UW 1UB { align1 };
+add (1) g7.28<1>F g1.22<0,1,0>UW 1UB { align1 };
+
+    /* Now, map these screen space coordinates into texture coordinates. */
+    /* subtract screen-space X origin of vertex 0. */
+add (8) g4<1>F g4<8,8,1>F -g1<0,1,0>F { align1 };
+add (8) g5<1>F g5<8,8,1>F -g1<0,1,0>F { align1 };
+    /* scale by texture X increment */
+mul (8) g4<1>F g4<8,8,1>F g3<0,1,0>F { align1 };
+mul (8) g5<1>F g5<8,8,1>F g3<0,1,0>F { align1 };
+    /* add in texture X offset */
+add (8) g4<1>F g4<8,8,1>F g3.12<0,1,0>F { align1 };
+add (8) g5<1>F g5<8,8,1>F g3.12<0,1,0>F { align1 };
+    /* subtract screen-space Y origin of vertex 0. */
+add (8) g6<1>F g6<8,8,1>F -g1.4<0,1,0>F { align1 };
+add (8) g7<1>F g7<8,8,1>F -g1.4<0,1,0>F { align1 };
+    /* scale by texture Y increment */
+mul (8) g6<1>F g6<8,8,1>F g3.20<0,1,0>F { align1 };
+mul (8) g7<1>F g7<8,8,1>F g3.20<0,1,0>F { align1 };
+    /* add in texture Y offset */
+add (8) g6<1>F g6<8,8,1>F g3.28<0,1,0>F { align1 };
+add (8) g7<1>F g7<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 g4<8,8,1>F { align1 };
+mov (8) m2<1>F g5<8,8,1>F { align1 };  /* param 0 u in m1, m2 */
+mov (8) m3<1>F g6<8,8,1>F { align1 };
+mov (8) m4<1>F g7<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 */
+send (16) 0 		/* msg reg index */
+	g12<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 */
+
+/* if we set up read-back reg correctly, emit dataport write 'send' cmd with EOT */
+
+/* m0, m1 are all direct passed by PS thread payload */
+mov (8) m1<1>F g1<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 */
+/* g12 -> m2
+   g13 -> m6
+   g14 -> m3
+   g15 -> m7
+   g16 -> m4
+   g17 -> m8
+   g18 -> m5
+   g19 -> m9
+*/
+mov (8) m2<1>F g12<8,8,1>F { align1 };
+mov (8) m3<1>F g14<8,8,1>F { align1 };
+mov (8) m4<1>F g16<8,8,1>F { align1 };
+mov (8) m5<1>F g18<8,8,1>F { align1 };
+mov (8) m6<1>F g13<8,8,1>F { align1 };
+mov (8) m7<1>F g15<8,8,1>F { align1 };
+mov (8) m8<1>F g17<8,8,1>F { align1 };
+mov (8) m9<1>F g19<8,8,1>F { align1 };
+
+/* write */
+send (16) 0 null 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;