i965/cs: Add a binding table entry for gl_NumWorkGroups

If glDispatchComputeIndirect is used, then the value for this variable
must be read from the indirect BO.

To allow the same generated code to support indirect and
glDispatchCompute, we will also setup a BO for the number of work
groups using the intel_upload_data mechanism. This will only be
required if the gl_NumWorkGroups variable is accessed.

Signed-off-by: Jordan Justen <[email protected]>
Reviewed-by: Kristian Høgsberg <[email protected]>

3 files changed, 29 insertions, 5 deletions

diff --git a/src/mesa/drivers/dri/i965/brw_context.h b/src/mesa/drivers/dri/i965/brw_context.h
index a65cac10d98..e407dec8eab 100644
--- a/src/mesa/drivers/dri/i965/brw_context.h
+++ b/src/mesa/drivers/dri/i965/brw_context.h
@@ -497,6 +497,14 @@ struct brw_cs_prog_data {
    unsigned local_size[3];
    unsigned simd_size;
    bool uses_barrier;
+
+   struct {
+      /** @{
+       * surface indices the CS-specific surfaces
+       */
+      uint32_t work_groups_start;
+      /** @} */
+   } binding_table;
 };
 
 /**
@@ -758,7 +766,8 @@ struct brw_vs_prog_data {
                             12 + /* ubo */                              \
                             BRW_MAX_ABO +                               \
                             BRW_MAX_IMAGES +                            \
-                            2 /* shader time, pull constants */)
+                            2 + /* shader time, pull constants */       \
+                            1 /* cs num work groups */)
 
 #define SURF_INDEX_GEN6_SOL_BINDING(t) (t)
 
diff --git a/src/mesa/drivers/dri/i965/brw_fs.cpp b/src/mesa/drivers/dri/i965/brw_fs.cpp
index 49dc7f65b48..64215ae5a6a 100644
--- a/src/mesa/drivers/dri/i965/brw_fs.cpp
+++ b/src/mesa/drivers/dri/i965/brw_fs.cpp
@@ -4750,7 +4750,7 @@ fs_visitor::setup_cs_payload()
 }
 
 void
-fs_visitor::assign_binding_table_offsets()
+fs_visitor::assign_fs_binding_table_offsets()
 {
    assert(stage == MESA_SHADER_FRAGMENT);
    brw_wm_prog_data *prog_data = (brw_wm_prog_data*) this->prog_data;
@@ -4767,6 +4767,20 @@ fs_visitor::assign_binding_table_offsets()
 }
 
 void
+fs_visitor::assign_cs_binding_table_offsets()
+{
+   assert(stage == MESA_SHADER_COMPUTE);
+   brw_cs_prog_data *prog_data = (brw_cs_prog_data*) this->prog_data;
+   uint32_t next_binding_table_offset = 0;
+
+   /* May not be used if the gl_NumWorkGroups variable is not accessed. */
+   prog_data->binding_table.work_groups_start = next_binding_table_offset;
+   next_binding_table_offset++;
+
+   assign_common_binding_table_offsets(next_binding_table_offset);
+}
+
+void
 fs_visitor::calculate_register_pressure()
 {
    invalidate_live_intervals();
@@ -5020,7 +5034,7 @@ fs_visitor::run_fs(bool do_rep_send)
 
    sanity_param_count = prog->Parameters->NumParameters;
 
-   assign_binding_table_offsets();
+   assign_fs_binding_table_offsets();
 
    if (devinfo->gen >= 6)
       setup_payload_gen6();
@@ -5108,7 +5122,7 @@ fs_visitor::run_cs()
 
    sanity_param_count = prog->Parameters->NumParameters;
 
-   assign_common_binding_table_offsets(0);
+   assign_cs_binding_table_offsets();
 
    setup_cs_payload();
 
diff --git a/src/mesa/drivers/dri/i965/brw_fs.h b/src/mesa/drivers/dri/i965/brw_fs.h
index 50ce4155f16..a8b6726b9f0 100644
--- a/src/mesa/drivers/dri/i965/brw_fs.h
+++ b/src/mesa/drivers/dri/i965/brw_fs.h
@@ -128,7 +128,8 @@ public:
    bool run_cs();
    void optimize();
    void allocate_registers();
-   void assign_binding_table_offsets();
+   void assign_fs_binding_table_offsets();
+   void assign_cs_binding_table_offsets();
    void setup_payload_gen4();
    void setup_payload_gen6();
    void setup_vs_payload();