i965: Add uniform for a CS thread local base ID

v4:
 * Force thread_local_id_index to -1 for now, and have
   fs_visitor::setup_cs_payload look at thread_local_id_index. This
   enables us to more easily cut over from the old local ID layout to
   the new layout, as suggested by Jason.

Cc: "12.0" <mesa-stable@lists.freedesktop.org>
Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>

src/intel/vulkan/anv_pipeline.c

@@ -338,6 +338,10 @@ anv_pipeline_compile(struct anv_pipeline *pipeline,
       pipeline->needs_data_cache = true;
    }
 
+   if (stage == MESA_SHADER_COMPUTE)
+      ((struct brw_cs_prog_data *)prog_data)->thread_local_id_index =
+         prog_data->nr_params++; /* The CS Thread ID uniform */
+
    if (nir->info.num_ssbos > 0)
       pipeline->needs_data_cache = true;
 

src/mesa/drivers/dri/i965/brw_compiler.h

@@ -433,6 +433,7 @@ struct brw_cs_prog_data {
    bool uses_barrier;
    bool uses_num_work_groups;
    unsigned local_invocation_id_regs;
+   int thread_local_id_index;
 
    struct {
       /** @{

src/mesa/drivers/dri/i965/brw_cs.c

@@ -93,6 +93,9 @@ brw_codegen_cs_prog(struct brw_context *brw,
     */
    int param_count = cp->program.Base.nir->num_uniforms / 4;
 
+   /* The backend also sometimes add a param for the thread local id. */
+   prog_data.thread_local_id_index = param_count++;
+
    /* The backend also sometimes adds params for texture size. */
    param_count += 2 * ctx->Const.Program[MESA_SHADER_COMPUTE].MaxTextureImageUnits;
    prog_data.base.param =

src/mesa/drivers/dri/i965/brw_fs.cpp

@@ -5629,7 +5629,8 @@ fs_visitor::setup_cs_payload()
 
    payload.num_regs = 1;
 
-   if (nir->info.system_values_read & SYSTEM_BIT_LOCAL_INVOCATION_ID) {
+   if (nir->info.system_values_read & SYSTEM_BIT_LOCAL_INVOCATION_ID &&
+       prog_data->thread_local_id_index < 0) {
       prog_data->local_invocation_id_regs = dispatch_width * 3 / 8;
       payload.local_invocation_id_reg = payload.num_regs;
       payload.num_regs += prog_data->local_invocation_id_regs;
@@ -6559,6 +6560,21 @@ brw_compile_cs(const struct brw_compiler *compiler, void *log_data,
                                       true);
    brw_nir_lower_cs_shared(shader);
    prog_data->base.total_shared += shader->num_shared;
+
+   /* The driver isn't yet ready to support thread_local_id_index, so we force
+    * it to disabled for now.
+    */
+   prog_data->thread_local_id_index = -1;
+
+   /* Now that we cloned the nir_shader, we can update num_uniforms based on
+    * the thread_local_id_index.
+    */
+   if (prog_data->thread_local_id_index >= 0) {
+      shader->num_uniforms =
+         MAX2(shader->num_uniforms,
+              (unsigned)4 * (prog_data->thread_local_id_index + 1));
+   }
+
    shader = brw_postprocess_nir(shader, compiler->devinfo, true);
 
    prog_data->local_size[0] = shader->info.cs.local_size[0];