Rename src/mask/data registers to indicate channel

9 files changed, 76 insertions, 60 deletions

diff --git a/src/exa_wm.g4i b/src/exa_wm.g4i
index 10e630e1..ee8e3ad0 100644
--- a/src/exa_wm.g4i
+++ b/src/exa_wm.g4i
@@ -102,21 +102,39 @@ define(`mask_w_0',  `src_w_0')
 define(`mask_w_1',  `src_w_1')
 
 /* sample src to these registers */
-define(`src_sample0',	`g14')
-define(`src_sample1',	`g15')
-define(`src_sample2',	`g16')
-define(`src_sample3',	`g17')
-define(`src_sample4',	`g18')
-define(`src_sample5',	`g19')
-define(`src_sample6',	`g20')
-define(`src_sample7',	`g21')
+define(`src_sample_base',	`g14')
+define(`src_sample_r_01',	`g14')
+define(`src_sample_r_23',	`g15')
+define(`src_sample_g_01',	`g16')
+define(`src_sample_g_23',	`g17')
+define(`src_sample_b_01',	`g18')
+define(`src_sample_b_23',	`g19')
+define(`src_sample_a_01',	`g20')
+define(`src_sample_a_23',	`g21')
 
 /* sample mask to these registers */
-define(`mask_sample0',	`g22')
-define(`mask_sample1',	`g23')
-define(`mask_sample2',	`g24')
-define(`mask_sample3',	`g25')
-define(`mask_sample4',	`g26')
-define(`mask_sample5',	`g27')
-define(`mask_sample6',	`g28')
-define(`mask_sample7',	`g29')
+define(`mask_sample_base',	`g22')
+define(`mask_sample_r_01',	`g22')
+define(`mask_sample_r_23',	`g23')
+define(`mask_sample_g_01',	`g24')
+define(`mask_sample_g_23',	`g25')
+define(`mask_sample_b_01',	`g26')
+define(`mask_sample_b_23',	`g27')
+define(`mask_sample_a_01',	`g28')
+define(`mask_sample_a_23',	`g29')
+
+/* data port SIMD16 send registers */
+
+define(`data_port_msg_0',	`m0')
+define(`data_port_msg_0_ind',	`0')
+define(`data_port_msg_1',	`m1')
+define(`data_port_r_01',	`m2')
+define(`data_port_g_01',	`m3')
+define(`data_port_b_01',	`m4')
+define(`data_port_a_01',	`m5')
+
+define(`data_port_r_23',	`m6')
+define(`data_port_g_23',	`m7')
+define(`data_port_b_23',	`m8')
+define(`data_port_a_23',	`m9')
+
diff --git a/src/exa_wm_ca.g4a b/src/exa_wm_ca.g4a
index 955c68c2..5d982b38 100644
--- a/src/exa_wm_ca.g4a
+++ b/src/exa_wm_ca.g4a
@@ -32,7 +32,7 @@
 include(`exa_wm.g4i')
 
 /* mul mask rgba channels to src */
-mul (16)    src_sample0<1>F src_sample0<8,8,1>F	mask_sample0<8,8,1>F	{ compr align1 };
-mul (16)    src_sample2<1>F src_sample2<8,8,1>F	mask_sample2<8,8,1>F	{ compr align1 };
-mul (16)    src_sample4<1>F src_sample4<8,8,1>F	mask_sample4<8,8,1>F	{ compr align1 };
-mul (16)    src_sample6<1>F src_sample6<8,8,1>F	mask_sample6<8,8,1>F	{ compr align1 };
+mul (16)    src_sample_r_01<1>F	src_sample_r_01<8,8,1>F	mask_sample_r_01<8,8,1>F { compr align1 };
+mul (16)    src_sample_g_01<1>F src_sample_g_01<8,8,1>F	mask_sample_g_01<8,8,1>F { compr align1 };
+mul (16)    src_sample_b_01<1>F src_sample_b_01<8,8,1>F	mask_sample_b_01<8,8,1>F { compr align1 };
+mul (16)    src_sample_a_01<1>F src_sample_a_01<8,8,1>F	mask_sample_a_01<8,8,1>F { compr align1 };
diff --git a/src/exa_wm_ca_srcalpha.g4a b/src/exa_wm_ca_srcalpha.g4a
index e252e19b..d1f847fd 100644
--- a/src/exa_wm_ca_srcalpha.g4a
+++ b/src/exa_wm_ca_srcalpha.g4a
@@ -31,7 +31,7 @@
 
 include(`exa_wm.g4i')
 
-mul (16)    src_sample0<1>F mask_sample0<8,8,1>F    src_sample6<8,8,1>F	{ compr align1 };
-mul (16)    src_sample2<1>F mask_sample2<8,8,1>F    src_sample6<8,8,1>F	{ compr align1 };
-mul (16)    src_sample4<1>F mask_sample4<8,8,1>F    src_sample6<8,8,1>F	{ compr align1 };
-mul (16)    src_sample6<1>F mask_sample6<8,8,1>F    src_sample6<8,8,1>F	{ compr align1 };
+mul (16)    src_sample_r_01<1>F mask_sample_r_01<8,8,1>F src_sample_a_01<8,8,1>F { compr align1 };
+mul (16)    src_sample_g_01<1>F mask_sample_g_01<8,8,1>F src_sample_a_01<8,8,1>F { compr align1 };
+mul (16)    src_sample_b_01<1>F mask_sample_b_01<8,8,1>F src_sample_a_01<8,8,1>F { compr align1 };
+mul (16)    src_sample_a_01<1>F mask_sample_a_01<8,8,1>F src_sample_a_01<8,8,1>F { compr align1 };
diff --git a/src/exa_wm_mask_sample_a.g4a b/src/exa_wm_mask_sample_a.g4a
index c06611d5..bbb19d7a 100644
--- a/src/exa_wm_mask_sample_a.g4a
+++ b/src/exa_wm_mask_sample_a.g4a
@@ -40,7 +40,7 @@ mov (1) g0.8<1>UD	0x00007000UD { align1 mask_disable };
 /* mask_msg will be copied with g0, as it contains send desc */
 /* emit sampler 'send' cmd */
 send (16) mask_msg_ind		/* msg reg index */
-	mask_sample6<1>UW 	/* readback */
+	mask_sample_a_01<1>UW 	/* readback */
 	g0<8,8,1>UW		/* copy to msg start reg*/
 	sampler (2,1,F)		/* sampler message description, (binding_table,sampler_index,datatype)
 				/* here(src->dst) we should use src_sampler and src_surface */
diff --git a/src/exa_wm_mask_sample_argb.g4a b/src/exa_wm_mask_sample_argb.g4a
index 7f0815f2..def4cfe4 100644
--- a/src/exa_wm_mask_sample_argb.g4a
+++ b/src/exa_wm_mask_sample_argb.g4a
@@ -40,7 +40,7 @@ mov (1) g0.8<1>UD	0x00000000UD { align1 mask_disable };
 /* mask_msg will be copied with g0, as it contains send desc */
 /* emit sampler 'send' cmd */
 send (16) mask_msg_ind		/* msg reg index */
-	mask_sample0<1>UW 	/* readback */
+	mask_sample_base<1>UW 	/* readback */
 	g0<8,8,1>UW		/* copy to msg start reg*/
 	sampler (2,1,F)		/* sampler message description, (binding_table,sampler_index,datatype)
 				/* here(src->dst) we should use src_sampler and src_surface */
diff --git a/src/exa_wm_noca.g4a b/src/exa_wm_noca.g4a
index 7dd12247..d0d60faa 100644
--- a/src/exa_wm_noca.g4a
+++ b/src/exa_wm_noca.g4a
@@ -32,7 +32,7 @@
 include(`exa_wm.g4i')
 /* mul mask's alpha channel to src */
 
-mul (16)    src_sample0<1>F src_sample0<8,8,1>F	mask_sample6<8,8,1>F	{ compr align1 };
-mul (16)    src_sample2<1>F src_sample2<8,8,1>F	mask_sample6<8,8,1>F	{ compr align1 };
-mul (16)    src_sample4<1>F src_sample4<8,8,1>F	mask_sample6<8,8,1>F	{ compr align1 };
-mul (16)    src_sample6<1>F src_sample6<8,8,1>F	mask_sample6<8,8,1>F	{ compr align1 };
+mul (16)    src_sample_r_01<1>F	src_sample_r_01<8,8,1>F	mask_sample_a_01<8,8,1>F { compr align1 };
+mul (16)    src_sample_g_01<1>F src_sample_g_01<8,8,1>F	mask_sample_a_01<8,8,1>F { compr align1 };
+mul (16)    src_sample_b_01<1>F src_sample_b_01<8,8,1>F	mask_sample_a_01<8,8,1>F { compr align1 };
+mul (16)    src_sample_a_01<1>F src_sample_a_01<8,8,1>F	mask_sample_a_01<8,8,1>F { compr align1 };
diff --git a/src/exa_wm_src_sample_a.g4a b/src/exa_wm_src_sample_a.g4a
index 803c358a..552aaeeb 100644
--- a/src/exa_wm_src_sample_a.g4a
+++ b/src/exa_wm_src_sample_a.g4a
@@ -40,7 +40,7 @@ mov (1) g0.8<1>UD	0x00007000UD { align1 mask_disable };
 /* src_msg will be copied with g0, as it contains send desc */
 /* emit sampler 'send' cmd */
 send (16) src_msg_ind		/* msg reg index */
-	src_sample6<1>UW 	/* readback */
+	src_sample_a_01<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 */
diff --git a/src/exa_wm_src_sample_argb.g4a b/src/exa_wm_src_sample_argb.g4a
index 4fcf276c..c20f53f2 100644
--- a/src/exa_wm_src_sample_argb.g4a
+++ b/src/exa_wm_src_sample_argb.g4a
@@ -40,7 +40,7 @@ mov (1) g0.8<1>UD	0x00000000UD { align1 mask_disable };
 /* src_msg will be copied with g0, as it contains send desc */
 /* emit sampler 'send' cmd */
 send (16) src_msg_ind		/* msg reg index */
-	src_sample0<1>UW 	/* readback */
+	src_sample_base<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 */
diff --git a/src/exa_wm_write.g4a b/src/exa_wm_write.g4a
index 5d3e6b1e..b16e6497 100644
--- a/src/exa_wm_write.g4a
+++ b/src/exa_wm_write.g4a
@@ -25,41 +25,39 @@
  *    Keith Packard <keithp@keithp.com>
  */
 
-/* 
- * Once the data are ready, write them to the destination
- */
-
 include(`exa_wm.g4i')
 
-/* prepare data in m2-m5 for subspan(1,0), m6-m9 for subspan(3,2), then it's ready to write */
-/* src_sample0 -> m2
-   src_sample1 -> m6
-   src_sample2 -> m3
-   src_sample3 -> m7
-   src_sample4 -> m4
-   src_sample5 -> m8
-   src_sample6 -> m5
-   src_sample7 -> m9
-*/
+/*
+ * Prepare data in m2-m5 for subspan(1,0), m6-m9 for subspan(3,2),
+ *
+ * Note that the SIMD16 write message takes data for the first
+ * two sub-spans followed by the data for the second two sub-spans
+ * instead of having the two sub-spans interleaved by channel. Weird.
+ */
+
+mov (8) data_port_r_01<1>F	src_sample_r_01<8,8,1>F { align1 };
+mov (8) data_port_g_01<1>F	src_sample_g_01<8,8,1>F { align1 };
+mov (8) data_port_b_01<1>F	src_sample_b_01<8,8,1>F { align1 };
+mov (8) data_port_a_01<1>F	src_sample_a_01<8,8,1>F { align1 };
 
-mov (8) m2<1>F src_sample0<8,8,1>F { align1 };
-mov (8) m3<1>F src_sample2<8,8,1>F { align1 };
-mov (8) m4<1>F src_sample4<8,8,1>F { align1 };
-mov (8) m5<1>F src_sample6<8,8,1>F { align1 };
-mov (8) m6<1>F src_sample1<8,8,1>F { align1 };
-mov (8) m7<1>F src_sample3<8,8,1>F { align1 };
-mov (8) m8<1>F src_sample5<8,8,1>F { align1 };
-mov (8) m9<1>F src_sample7<8,8,1>F { align1 };
+mov (8) data_port_r_23<1>F	src_sample_r_23<8,8,1>F { align1 };
+mov (8) data_port_g_23<1>F	src_sample_g_23<8,8,1>F { align1 };
+mov (8) data_port_b_23<1>F	src_sample_b_23<8,8,1>F { align1 };
+mov (8) data_port_a_23<1>F 	src_sample_a_23<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 };
+mov (8) data_port_msg_1<1>UD	g1<8,8,1>UD		{ align1 };
 
 /* 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 */
+send (16) 
+	data_port_msg_0_ind 
+	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