[Mesa-dev] [RFC PATCH 13/40] i965: Enable gather push constants

[Abdiel Janulgue]

The 3DSTATE_GATHER_POOL_ALLOC is used to enable or disable the gather
push constants feature within a context. This patch provides the toggle
functionality of using gather push constants to program constant data
within a batch.

In addition, using gather push constants require that a gather pool be
allocated so that the resource streamer can flush the packed constants it
gathered. The pool is later referenced by the 3DSTATE_CONSTANT_* command
to program the push constant data. This patch initializes the gather
pool as well.

Signed-off-by: Abdiel Janulgue <abdiel.janulgue at linux.intel.com>
---
 src/mesa/drivers/dri/i965/brw_binding_tables.c | 40 +++++++++++++++++++++++++-
 src/mesa/drivers/dri/i965/brw_context.c        |  7 +++++
 src/mesa/drivers/dri/i965/brw_context.h        |  7 +++++
 src/mesa/drivers/dri/i965/brw_state.h          |  1 +
 4 files changed, 54 insertions(+), 1 deletion(-)

diff --git a/src/mesa/drivers/dri/i965/brw_binding_tables.c b/src/mesa/drivers/dri/i965/brw_binding_tables.c
index 03e7a4a..b91e5d9 100644
--- a/src/mesa/drivers/dri/i965/brw_binding_tables.c
+++ b/src/mesa/drivers/dri/i965/brw_binding_tables.c
@@ -222,9 +222,44 @@ gen7_update_binding_table_from_array(struct brw_context *brw,
    ADVANCE_BATCH();
 }
 
+static void
+gen7_init_gather_pool(struct brw_context *brw)
+{
+   if (!brw->has_resource_streamer || !brw->use_gather_constants)
+      return;
+
+   if (!brw->gather_pool.bo) {
+      brw->gather_pool.bo = drm_intel_bo_alloc(brw->bufmgr, "gather_pool",
+                                               131072, 4096);
+      brw->gather_pool.next_offset = 0;
+   }
+}
+
+void
+gen7_toggle_gather_constants(struct brw_context *brw, bool enable)
+{
+   if (enable && (!brw->has_resource_streamer || !brw->use_gather_constants))
+      return;
+
+   BEGIN_BATCH(3);
+   OUT_BATCH(_3DSTATE_GATHER_POOL_ALLOC << 16 | (3 - 2));
+   if (enable) {
+      OUT_RELOC(brw->gather_pool.bo, I915_GEM_DOMAIN_SAMPLER, 0,
+                (1 << 11) | (3 << 4) | GEN7_MOCS_L3);
+      OUT_RELOC(brw->gather_pool.bo, I915_GEM_DOMAIN_SAMPLER, 0,
+                brw->gather_pool.bo->size);
+   } else {
+      OUT_BATCH((3 << 4));
+      OUT_BATCH(0);
+   }
+   ADVANCE_BATCH();
+}
+
 void
 gen7_disable_hw_binding_tables(struct brw_context *brw)
 {
+   gen7_toggle_gather_constants(brw, false);
+
    BEGIN_BATCH(3);
    OUT_BATCH(_3DSTATE_BINDING_TABLE_POOL_ALLOC << 16 | (3 - 2));
    OUT_BATCH(3 << 5); /* only in HSW */
@@ -262,6 +297,9 @@ gen7_enable_hw_binding_tables(struct brw_context *brw)
              brw->hw_bt_pool.bo->size);
    ADVANCE_BATCH();
 
+   gen7_init_gather_pool(brw);
+   gen7_toggle_gather_constants(brw, true);
+
    /* Pipe control workaround */
    BEGIN_BATCH(4);
    OUT_BATCH(_3DSTATE_PIPE_CONTROL | (4 - 2));
@@ -275,6 +313,7 @@ void
 gen7_reset_rs_pool_offsets(struct brw_context *brw)
 {
    brw->hw_bt_pool.next_offset = bt_size;
+   brw->gather_pool.next_offset = 0;
 }
 
 const struct brw_tracked_state gen7_hw_binding_tables = {
@@ -358,5 +397,4 @@ const struct brw_tracked_state gen6_binding_table_pointers = {
    },
    .emit = gen6_upload_binding_table_pointers,
 };
-
 /** @} */
diff --git a/src/mesa/drivers/dri/i965/brw_context.c b/src/mesa/drivers/dri/i965/brw_context.c
index b962103..32bbdc2 100644
--- a/src/mesa/drivers/dri/i965/brw_context.c
+++ b/src/mesa/drivers/dri/i965/brw_context.c
@@ -711,6 +711,12 @@ brwCreateContext(gl_api api,
       brw->has_resource_streamer = false;
    }
 
+   if (getenv("INTEL_GATHER")) {
+      brw->use_gather_constants = brw->has_resource_streamer;
+   } else {
+      brw->use_gather_constants = false;
+   }
+
    brw->vs.base.stage = MESA_SHADER_VERTEX;
    brw->gs.base.stage = MESA_SHADER_GEOMETRY;
    brw->wm.base.stage = MESA_SHADER_FRAGMENT;
@@ -853,6 +859,7 @@ brwCreateContext(gl_api api,
       ctx->Const.ContextFlags |= GL_CONTEXT_FLAG_ROBUST_ACCESS_BIT_ARB;
 
    brw->hw_bt_pool.bo = 0;
+   brw->gather_pool.bo = 0;
 
    if (INTEL_DEBUG & DEBUG_SHADER_TIME)
       brw_init_shader_time(brw);
diff --git a/src/mesa/drivers/dri/i965/brw_context.h b/src/mesa/drivers/dri/i965/brw_context.h
index 17fea5b..b205773 100644
--- a/src/mesa/drivers/dri/i965/brw_context.h
+++ b/src/mesa/drivers/dri/i965/brw_context.h
@@ -1081,6 +1081,7 @@ struct brw_context
    bool use_rep_send;
    bool scalar_vs;
    bool has_resource_streamer;
+   bool use_gather_constants;
 
    /**
     * Some versions of Gen hardware don't do centroid interpolation correctly
@@ -1340,6 +1341,12 @@ struct brw_context
       uint32_t next_offset;
    } hw_bt_pool;
 
+   /* Internal storage used by the resource streamer to flush and refer to constant data*/
+   struct {
+      drm_intel_bo *bo;
+      uint32_t next_offset;
+   } gather_pool;
+
    struct {
       uint32_t state_offset;
       uint32_t blend_state_offset;
diff --git a/src/mesa/drivers/dri/i965/brw_state.h b/src/mesa/drivers/dri/i965/brw_state.h
index f985a3a..16355bd 100644
--- a/src/mesa/drivers/dri/i965/brw_state.h
+++ b/src/mesa/drivers/dri/i965/brw_state.h
@@ -309,6 +309,7 @@ void gen7_update_binding_table_from_array(struct brw_context *brw,
 void gen7_enable_hw_binding_tables(struct brw_context *brw);
 void gen7_disable_hw_binding_tables(struct brw_context *brw);
 void gen7_reset_rs_pool_offsets(struct brw_context *brw);
+void gen7_toggle_gather_constants(struct brw_context *brw, bool enable);
 
 #ifdef __cplusplus
 }
-- 
1.9.1