Compute pixel values directly into data port

Instead of leaving pixel values in src_sample registers, compute the pixel
values directl to the data port to save 8 moves. This cannot work when no
computation is done as there is both no way to wait for the sampler to
finish and because the sampler returns data in a different order from that
required by the data port (sigh).

1 files changed, 0 insertions, 18 deletions

diff --git a/src/exa_wm_write.g4a b/src/exa_wm_write.g4a
index b16e6497..c46023e8 100644
--- a/src/exa_wm_write.g4a
+++ b/src/exa_wm_write.g4a
@@ -27,24 +27,6 @@
 
 include(`exa_wm.g4i')
 
-/*
- * 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) 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) data_port_msg_1<1>UD	g1<8,8,1>UD		{ align1 };