Mesa (7.10): i965: bump VS thread number to 60 on SNB

[Ian Romanick]

Module: Mesa
Branch: 7.10
Commit: c21a44463a402a886bf58d0e689d5a098eda1ff4
URL:    http://cgit.freedesktop.org/mesa/mesa/commit/?id=c21a44463a402a886bf58d0e689d5a098eda1ff4

Author: Zou Nan hai <nanhai.zou at intel.com>
Date:   Tue Mar  1 10:39:35 2011 +0800

i965: bump VS thread number to 60 on SNB

Signed-off-by: Zou Nan hai <nanhai.zou at intel.com>
(cherry picked from commit 6c324777a685d28d0a81d23157e4863240552999)

---

 src/mesa/drivers/dri/i965/brw_vs_emit.c   |   10 +++++++++-
 src/mesa/drivers/dri/i965/gen6_vs_state.c |    3 ++-
 2 files changed, 11 insertions(+), 2 deletions(-)

diff --git a/src/mesa/drivers/dri/i965/brw_vs_emit.c b/src/mesa/drivers/dri/i965/brw_vs_emit.c
index 0411ce0..6ec6255 100644
--- a/src/mesa/drivers/dri/i965/brw_vs_emit.c
+++ b/src/mesa/drivers/dri/i965/brw_vs_emit.c
@@ -1561,6 +1561,7 @@ static void emit_vertex_write( struct brw_vs_compile *c)
    int eot;
    GLuint len_vertex_header = 2;
    int next_mrf, i;
+   int msg_len;
 
    if (c->key.copy_edgeflag) {
       brw_MOV(p, 
@@ -1727,13 +1728,20 @@ static void emit_vertex_write( struct brw_vs_compile *c)
 
    eot = (c->first_overflow_output == 0);
 
+   msg_len = c->nr_outputs + 2 + len_vertex_header; 
+   if (intel->gen >= 6) {
+	   /* interleaved urb write message length for gen6 should be multiple of 2 */
+	   if ((msg_len % 2) != 0)
+		msg_len++;
+   }
+
    brw_urb_WRITE(p, 
 		 brw_null_reg(), /* dest */
 		 0,		/* starting mrf reg nr */
 		 c->r0,		/* src */
 		 0,		/* allocate */
 		 1,		/* used */
-		 MIN2(c->nr_outputs + 1 + len_vertex_header, (BRW_MAX_MRF-1)), /* msg len */
+		 MIN2(msg_len - 1, (BRW_MAX_MRF - 1)), /* msg len */
 		 0,		/* response len */
 		 eot, 		/* eot */
 		 eot, 		/* writes complete */
diff --git a/src/mesa/drivers/dri/i965/gen6_vs_state.c b/src/mesa/drivers/dri/i965/gen6_vs_state.c
index ed132bd..20dfb13 100644
--- a/src/mesa/drivers/dri/i965/gen6_vs_state.c
+++ b/src/mesa/drivers/dri/i965/gen6_vs_state.c
@@ -136,7 +136,8 @@ upload_vs_state(struct brw_context *brw)
    OUT_BATCH((1 << GEN6_VS_DISPATCH_START_GRF_SHIFT) |
 	     (brw->vs.prog_data->urb_read_length << GEN6_VS_URB_READ_LENGTH_SHIFT) |
 	     (0 << GEN6_VS_URB_ENTRY_READ_OFFSET_SHIFT));
-   OUT_BATCH((0 << GEN6_VS_MAX_THREADS_SHIFT) |
+
+   OUT_BATCH(((60 - 1) << GEN6_VS_MAX_THREADS_SHIFT) | /* max 60 threads for gen6 */
 	     GEN6_VS_STATISTICS_ENABLE |
 	     GEN6_VS_ENABLE);
    ADVANCE_BATCH();