intel/compiler: add new field for storing program size

This will be used by the on disk shader cache.

v2:
 * Set in brw_compile_* rather than brw_codegen_*. (Jason)

Signed-off-by: Timothy Arceri <[email protected]>
[[email protected]: Only add to brw_stage_prog_data]
Signed-off-by: Jordan Justen <[email protected]>
Reviewed-by: Jason Ekstrand <[email protected]>

6 files changed, 35 insertions, 14 deletions

diff --git a/src/intel/compiler/brw_compiler.h b/src/intel/compiler/brw_compiler.h
index 17615c6dd9f..fa5a3bf0274 100644
--- a/src/intel/compiler/brw_compiler.h
+++ b/src/intel/compiler/brw_compiler.h
@@ -587,6 +587,8 @@ struct brw_stage_prog_data {
    unsigned total_scratch;
    unsigned total_shared;
 
+   unsigned program_size;
+
    /**
     * Register where the thread expects to find input data from the URB
     * (typically uniforms, followed by vertex or fragment attributes).
diff --git a/src/intel/compiler/brw_fs.cpp b/src/intel/compiler/brw_fs.cpp
index e5467922553..d355675b2b8 100644
--- a/src/intel/compiler/brw_fs.cpp
+++ b/src/intel/compiler/brw_fs.cpp
@@ -6691,7 +6691,9 @@ brw_compile_fs(const struct brw_compiler *compiler, void *log_data,
       prog_data->reg_blocks_0 = brw_register_blocks(simd16_grf_used);
    }
 
-   return g.get_assembly(final_assembly_size);
+   const unsigned *assembly = g.get_assembly(final_assembly_size);
+   prog_data->base.program_size = *final_assembly_size;
+   return assembly;
 }
 
 fs_reg *
@@ -6890,7 +6892,9 @@ brw_compile_cs(const struct brw_compiler *compiler, void *log_data,
 
    g.generate_code(cfg, prog_data->simd_size);
 
-   return g.get_assembly(final_assembly_size);
+   const unsigned *assembly = g.get_assembly(final_assembly_size);
+   prog_data->base.program_size = *final_assembly_size;
+   return assembly;
 }
 
 /**
diff --git a/src/intel/compiler/brw_shader.cpp b/src/intel/compiler/brw_shader.cpp
index ddb941c1cbf..7253f5d5531 100644
--- a/src/intel/compiler/brw_shader.cpp
+++ b/src/intel/compiler/brw_shader.cpp
@@ -1162,6 +1162,7 @@ brw_compile_tes(const struct brw_compiler *compiler,
 {
    const struct gen_device_info *devinfo = compiler->devinfo;
    const bool is_scalar = compiler->scalar_stage[MESA_SHADER_TESS_EVAL];
+   const unsigned *assembly;
 
    nir_shader *nir = nir_shader_clone(mem_ctx, src_shader);
    nir->info.inputs_read = key->inputs_read;
@@ -1270,7 +1271,7 @@ brw_compile_tes(const struct brw_compiler *compiler,
 
       g.generate_code(v.cfg, 8);
 
-      return g.get_assembly(final_assembly_size);
+      assembly = g.get_assembly(final_assembly_size);
    } else {
       brw::vec4_tes_visitor v(compiler, log_data, key, prog_data,
 			      nir, mem_ctx, shader_time_index);
@@ -1283,8 +1284,11 @@ brw_compile_tes(const struct brw_compiler *compiler,
       if (unlikely(INTEL_DEBUG & DEBUG_TES))
 	 v.dump_instructions();
 
-      return brw_vec4_generate_assembly(compiler, log_data, mem_ctx, nir,
-					&prog_data->base, v.cfg,
-					final_assembly_size);
+      assembly = brw_vec4_generate_assembly(compiler, log_data, mem_ctx, nir,
+                                            &prog_data->base, v.cfg,
+                                            final_assembly_size);
    }
+
+   prog_data->base.base.program_size = *final_assembly_size;
+   return assembly;
 }
diff --git a/src/intel/compiler/brw_vec4.cpp b/src/intel/compiler/brw_vec4.cpp
index c21d3f49c2e..7d8cdfa74a3 100644
--- a/src/intel/compiler/brw_vec4.cpp
+++ b/src/intel/compiler/brw_vec4.cpp
@@ -2901,6 +2901,7 @@ brw_compile_vs(const struct brw_compiler *compiler, void *log_data,
                                             final_assembly_size);
    }
 
+   prog_data->base.base.program_size = *final_assembly_size;
    return assembly;
 }
 
diff --git a/src/intel/compiler/brw_vec4_gs_visitor.cpp b/src/intel/compiler/brw_vec4_gs_visitor.cpp
index 5df6d562ce6..11b73f187cf 100644
--- a/src/intel/compiler/brw_vec4_gs_visitor.cpp
+++ b/src/intel/compiler/brw_vec4_gs_visitor.cpp
@@ -868,7 +868,9 @@ brw_compile_gs(const struct brw_compiler *compiler, void *log_data,
             g.enable_debug(name);
          }
          g.generate_code(v.cfg, 8);
-         return g.get_assembly(final_assembly_size);
+         const unsigned *ret = g.get_assembly(final_assembly_size);
+         prog_data->base.base.program_size = *final_assembly_size;
+         return ret;
       }
    }
 
@@ -897,9 +899,12 @@ brw_compile_gs(const struct brw_compiler *compiler, void *log_data,
          if (v.run()) {
             /* Success! Backup is not needed */
             ralloc_free(param);
-            return brw_vec4_generate_assembly(compiler, log_data, mem_ctx,
-                                              shader, &prog_data->base, v.cfg,
-                                              final_assembly_size);
+            const unsigned *ret =
+               brw_vec4_generate_assembly(compiler, log_data, mem_ctx, shader,
+                                          &prog_data->base, v.cfg,
+                                          final_assembly_size);
+            prog_data->base.base.program_size = *final_assembly_size;
+            return ret;
          } else {
             /* These variables could be modified by the execution of the GS
              * visitor if it packed the uniforms in the push constant buffer.
@@ -967,6 +972,7 @@ brw_compile_gs(const struct brw_compiler *compiler, void *log_data,
    }
 
    delete gs;
+   prog_data->base.base.program_size = *final_assembly_size;
    return ret;
 }
 
diff --git a/src/intel/compiler/brw_vec4_tcs.cpp b/src/intel/compiler/brw_vec4_tcs.cpp
index c4d9f89a91b..04feb6bdf17 100644
--- a/src/intel/compiler/brw_vec4_tcs.cpp
+++ b/src/intel/compiler/brw_vec4_tcs.cpp
@@ -388,6 +388,7 @@ brw_compile_tcs(const struct brw_compiler *compiler,
    const struct gen_device_info *devinfo = compiler->devinfo;
    struct brw_vue_prog_data *vue_prog_data = &prog_data->base;
    const bool is_scalar = compiler->scalar_stage[MESA_SHADER_TESS_CTRL];
+   const unsigned *assembly;
 
    nir_shader *nir = nir_shader_clone(mem_ctx, src_shader);
    nir->info.outputs_written = key->outputs_written;
@@ -487,7 +488,7 @@ brw_compile_tcs(const struct brw_compiler *compiler,
 
       g.generate_code(v.cfg, 8);
 
-      return g.get_assembly(final_assembly_size);
+      assembly = g.get_assembly(final_assembly_size);
    } else {
       vec4_tcs_visitor v(compiler, log_data, key, prog_data,
                          nir, mem_ctx, shader_time_index, &input_vue_map);
@@ -501,10 +502,13 @@ brw_compile_tcs(const struct brw_compiler *compiler,
          v.dump_instructions();
 
 
-      return brw_vec4_generate_assembly(compiler, log_data, mem_ctx, nir,
-                                        &prog_data->base, v.cfg,
-                                        final_assembly_size);
+      assembly = brw_vec4_generate_assembly(compiler, log_data, mem_ctx, nir,
+                                            &prog_data->base, v.cfg,
+                                            final_assembly_size);
    }
+
+   prog_data->base.base.program_size = *final_assembly_size;
+   return assembly;
 }