i965: Use state streaming on programs, and state base address on gen5+.

There will be a little bit of thrashing of the program cache BO as the
cache warms up, but once the application is in steady state, this
reduces relocations on gen5 and later.

On my T420 laptop, cairogl firefox-talos-gfx performance improves 2.6%
+/- 1.3% (n=6).  No statistically significant performance difference
on nexuiz (n=5).

1 files changed, 3 insertions, 7 deletions

diff --git a/src/mesa/drivers/dri/i965/gen6_wm_state.c b/src/mesa/drivers/dri/i965/gen6_wm_state.c
index 43e651db3ef..024a1d879e7 100644
--- a/src/mesa/drivers/dri/i965/gen6_wm_state.c
+++ b/src/mesa/drivers/dri/i965/gen6_wm_state.c
@@ -183,7 +183,7 @@ upload_wm_state(struct brw_context *brw)
 
    BEGIN_BATCH(9);
    OUT_BATCH(_3DSTATE_WM << 16 | (9 - 2));
-   OUT_RELOC(brw->wm.prog_bo, I915_GEM_DOMAIN_INSTRUCTION, 0, 0);
+   OUT_BATCH(brw->wm.prog_offset);
    OUT_BATCH(dw2);
    if (brw->wm.prog_data->total_scratch) {
       OUT_RELOC(brw->wm.scratch_bo, I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER,
@@ -195,12 +195,8 @@ upload_wm_state(struct brw_context *brw)
    OUT_BATCH(dw5);
    OUT_BATCH(dw6);
    OUT_BATCH(0); /* kernel 1 pointer */
-   if (brw->wm.prog_data->prog_offset_16) {
-      OUT_RELOC(brw->wm.prog_bo, I915_GEM_DOMAIN_INSTRUCTION, 0,
-		brw->wm.prog_data->prog_offset_16);
-   } else {
-      OUT_BATCH(0); /* kernel 2 pointer */
-   }
+   /* kernel 2 pointer */
+   OUT_BATCH(brw->wm.prog_offset + brw->wm.prog_data->prog_offset_16);
    ADVANCE_BATCH();
 }