i965: Move perf_debug code to brw_codegen_*_prog()

We're trying to avoid a libdrm dependency in the core compiler, so let's
move the perf_debug code one level up from the brw_*_emit() helpers to
the brw_codegen_*_prog() helpers.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Signed-off-by: Kristian Høgsberg Kristensen <krh@bitplanet.net>

diff --git a/src/mesa/drivers/dri/i965/brw_cs.c b/src/mesa/drivers/dri/i965/brw_cs.c
index 012c46698e7bf47cb75d7e55b67464855508fbc8..cb3fae66ec35b6f384a92ccaae5887ff14834314 100644 (file)
--- a/src/mesa/drivers/dri/i965/brw_cs.c
+++ b/src/mesa/drivers/dri/i965/brw_cs.c
@@ -63,8 +63,11 @@ brw_codegen_cs_prog(struct brw_context *brw,
    void *mem_ctx = ralloc_context(NULL);
    GLuint program_size;
    struct brw_cs_prog_data prog_data;
+   bool start_busy = false;
+   double start_time = 0;
 
-   struct gl_shader *cs = prog->_LinkedShaders[MESA_SHADER_COMPUTE];
+   struct brw_shader *cs =
+      (struct brw_shader *) prog->_LinkedShaders[MESA_SHADER_COMPUTE];
    assert (cs);
 
    memset(&prog_data, 0, sizeof(prog_data));
@@ -73,8 +76,8 @@ brw_codegen_cs_prog(struct brw_context *brw,
     * prog_data associated with the compiled program, and which will be freed
     * by the state cache.
     */
-   int param_count = cs->num_uniform_components +
-                     cs->NumImages * BRW_IMAGE_PARAM_SIZE;
+   int param_count = cs->base.num_uniform_components +
+                     cs->base.NumImages * BRW_IMAGE_PARAM_SIZE;
 
    /* The backend also sometimes adds params for texture size. */
    param_count += 2 * ctx->Const.Program[MESA_SHADER_COMPUTE].MaxTextureImageUnits;
@@ -83,9 +86,15 @@ brw_codegen_cs_prog(struct brw_context *brw,
    prog_data.base.pull_param =
       rzalloc_array(NULL, const gl_constant_value *, param_count);
    prog_data.base.image_param =
-      rzalloc_array(NULL, struct brw_image_param, cs->NumImages);
+      rzalloc_array(NULL, struct brw_image_param, cs->base.NumImages);
    prog_data.base.nr_params = param_count;
-   prog_data.base.nr_image_params = cs->NumImages;
+   prog_data.base.nr_image_params = cs->base.NumImages;
+
+   if (unlikely(brw->perf_debug)) {
+      start_busy = (brw->batch.last_bo &&
+                    drm_intel_bo_busy(brw->batch.last_bo));
+      start_time = get_time();
+   }
 
    program = brw_cs_emit(brw, mem_ctx, key, &prog_data,
                          &cp->program, prog, &program_size);
@@ -94,6 +103,18 @@ brw_codegen_cs_prog(struct brw_context *brw,
       return false;
    }
 
+   if (unlikely(brw->perf_debug) && cs) {
+      if (cs->compiled_once) {
+         _mesa_problem(&brw->ctx, "CS programs shouldn't need recompiles");
+      }
+      cs->compiled_once = true;
+
+      if (start_busy && !drm_intel_bo_busy(brw->batch.last_bo)) {
+         perf_debug("CS compile took %.03f ms and stalled the GPU\n",
+                    (get_time() - start_time) * 1000);
+      }
+   }
+
    if (prog_data.base.total_scratch) {
       brw_get_scratch_bo(brw, &brw->cs.base.scratch_bo,
                          prog_data.base.total_scratch * brw->max_cs_threads);

diff --git a/src/mesa/drivers/dri/i965/brw_fs.cpp b/src/mesa/drivers/dri/i965/brw_fs.cpp
index d4900f147748ae09e19f794a1ce61d4c92f64896..b9f1051eca42570671b80aa1b41545fc1b91c70c 100644 (file)
--- a/src/mesa/drivers/dri/i965/brw_fs.cpp
+++ b/src/mesa/drivers/dri/i965/brw_fs.cpp
@@ -5140,15 +5140,6 @@ brw_wm_fs_emit(struct brw_context *brw,
                struct gl_shader_program *prog,
                unsigned *final_assembly_size)
 {
-   bool start_busy = false;
-   double start_time = 0;
-
-   if (unlikely(brw->perf_debug)) {
-      start_busy = (brw->batch.last_bo &&
-                    drm_intel_bo_busy(brw->batch.last_bo));
-      start_time = get_time();
-   }
-
    struct brw_shader *shader = NULL;
    if (prog)
       shader = (brw_shader *) prog->_LinkedShaders[MESA_SHADER_FRAGMENT];
@@ -5226,17 +5217,6 @@ brw_wm_fs_emit(struct brw_context *brw,
    if (simd16_cfg)
       prog_data->prog_offset_16 = g.generate_code(simd16_cfg, 16);
 
-   if (unlikely(brw->perf_debug) && shader) {
-      if (shader->compiled_once)
-         brw_wm_debug_recompile(brw, prog, key);
-      shader->compiled_once = true;
-
-      if (start_busy && !drm_intel_bo_busy(brw->batch.last_bo)) {
-         perf_debug("FS compile took %.03f ms and stalled the GPU\n",
-                    (get_time() - start_time) * 1000);
-      }
-   }
-
    return g.get_assembly(final_assembly_size);
 }
 
@@ -5286,15 +5266,6 @@ brw_cs_emit(struct brw_context *brw,
             struct gl_shader_program *prog,
             unsigned *final_assembly_size)
 {
-   bool start_busy = false;
-   double start_time = 0;
-
-   if (unlikely(brw->perf_debug)) {
-      start_busy = (brw->batch.last_bo &&
-                    drm_intel_bo_busy(brw->batch.last_bo));
-      start_time = get_time();
-   }
-
    struct brw_shader *shader =
       (struct brw_shader *) prog->_LinkedShaders[MESA_SHADER_COMPUTE];
 
@@ -5368,17 +5339,5 @@ brw_cs_emit(struct brw_context *brw,
 
    g.generate_code(cfg, prog_data->simd_size);
 
-   if (unlikely(brw->perf_debug) && shader) {
-      if (shader->compiled_once) {
-         _mesa_problem(&brw->ctx, "CS programs shouldn't need recompiles");
-      }
-      shader->compiled_once = true;
-
-      if (start_busy && !drm_intel_bo_busy(brw->batch.last_bo)) {
-         perf_debug("CS compile took %.03f ms and stalled the GPU\n",
-                    (get_time() - start_time) * 1000);
-      }
-   }
-
    return g.get_assembly(final_assembly_size);
 }

diff --git a/src/mesa/drivers/dri/i965/brw_vec4.cpp b/src/mesa/drivers/dri/i965/brw_vec4.cpp
index 181768bddea17c1cda8d55c26d0dfca6822c7e22..ed49cd38987ae73ce778cc21ccce383f4838d929 100644 (file)
--- a/src/mesa/drivers/dri/i965/brw_vec4.cpp
+++ b/src/mesa/drivers/dri/i965/brw_vec4.cpp
@@ -1930,16 +1930,8 @@ brw_vs_emit(struct brw_context *brw,
             struct gl_shader_program *prog,
             unsigned *final_assembly_size)
 {
-   bool start_busy = false;
-   double start_time = 0;
    const unsigned *assembly = NULL;
 
-   if (unlikely(brw->perf_debug)) {
-      start_busy = (brw->batch.last_bo &&
-                    drm_intel_bo_busy(brw->batch.last_bo));
-      start_time = get_time();
-   }
-
    struct brw_shader *shader = NULL;
    if (prog)
       shader = (brw_shader *) prog->_LinkedShaders[MESA_SHADER_VERTEX];
@@ -2029,17 +2021,6 @@ brw_vs_emit(struct brw_context *brw,
       assembly = g.generate_assembly(v.cfg, final_assembly_size);
    }
 
-   if (unlikely(brw->perf_debug) && shader) {
-      if (shader->compiled_once) {
-         brw_vs_debug_recompile(brw, prog, key);
-      }
-      if (start_busy && !drm_intel_bo_busy(brw->batch.last_bo)) {
-         perf_debug("VS compile took %.03f ms and stalled the GPU\n",
-                    (get_time() - start_time) * 1000);
-      }
-      shader->compiled_once = true;
-   }
-
    return assembly;
 }
 

diff --git a/src/mesa/drivers/dri/i965/brw_vs.c b/src/mesa/drivers/dri/i965/brw_vs.c
index 4e0d34f6c6f4ff52c49ab655e918f3c8fdc7f490..465e78f4c747524db14b710a37ad24e2bdebd421 100644 (file)
--- a/src/mesa/drivers/dri/i965/brw_vs.c
+++ b/src/mesa/drivers/dri/i965/brw_vs.c
@@ -98,10 +98,12 @@ brw_codegen_vs_prog(struct brw_context *brw,
    struct brw_stage_prog_data *stage_prog_data = &prog_data.base.base;
    void *mem_ctx;
    int i;
-   struct gl_shader *vs = NULL;
+   struct brw_shader *vs = NULL;
+   bool start_busy = false;
+   double start_time = 0;
 
    if (prog)
-      vs = prog->_LinkedShaders[MESA_SHADER_VERTEX];
+      vs = (struct brw_shader *) prog->_LinkedShaders[MESA_SHADER_VERTEX];
 
    memset(&prog_data, 0, sizeof(prog_data));
 
@@ -121,9 +123,9 @@ brw_codegen_vs_prog(struct brw_context *brw,
        * case being a float value that gets blown up to a vec4, so be
        * conservative here.
        */
-      param_count = vs->num_uniform_components * 4 +
-                    vs->NumImages * BRW_IMAGE_PARAM_SIZE;
-      stage_prog_data->nr_image_params = vs->NumImages;
+      param_count = vs->base.num_uniform_components * 4 +
+                    vs->base.NumImages * BRW_IMAGE_PARAM_SIZE;
+      stage_prog_data->nr_image_params = vs->base.NumImages;
    } else {
       param_count = vp->program.Base.Parameters->NumParameters * 4;
    }
@@ -185,6 +187,12 @@ brw_codegen_vs_prog(struct brw_context *brw,
                               true);
    }
 
+   if (unlikely(brw->perf_debug)) {
+      start_busy = (brw->batch.last_bo &&
+                    drm_intel_bo_busy(brw->batch.last_bo));
+      start_time = get_time();
+   }
+
    /* Emit GEN4 code.
     */
    program = brw_vs_emit(brw, mem_ctx, key, &prog_data,
@@ -194,6 +202,17 @@ brw_codegen_vs_prog(struct brw_context *brw,
       return false;
    }
 
+   if (unlikely(brw->perf_debug) && vs) {
+      if (vs->compiled_once) {
+         brw_vs_debug_recompile(brw, prog, key);
+      }
+      if (start_busy && !drm_intel_bo_busy(brw->batch.last_bo)) {
+         perf_debug("VS compile took %.03f ms and stalled the GPU\n",
+                    (get_time() - start_time) * 1000);
+      }
+      vs->compiled_once = true;
+   }
+
    /* Scratch space is used for register spilling */
    if (prog_data.base.base.total_scratch) {
       brw_get_scratch_bo(brw, &brw->vs.base.scratch_bo,

diff --git a/src/mesa/drivers/dri/i965/brw_wm.c b/src/mesa/drivers/dri/i965/brw_wm.c
index 8dfa1427e4f100f72e0753f70c1672a5c102f24c..35c0908324ffadda1e69e3f0005e08dd2d49e9ce 100644 (file)
--- a/src/mesa/drivers/dri/i965/brw_wm.c
+++ b/src/mesa/drivers/dri/i965/brw_wm.c
@@ -165,11 +165,13 @@ brw_codegen_wm_prog(struct brw_context *brw,
    void *mem_ctx = ralloc_context(NULL);
    struct brw_wm_prog_data prog_data;
    const GLuint *program;
-   struct gl_shader *fs = NULL;
+   struct brw_shader *fs = NULL;
    GLuint program_size;
+   bool start_busy = false;
+   double start_time = 0;
 
    if (prog)
-      fs = prog->_LinkedShaders[MESA_SHADER_FRAGMENT];
+      fs = (struct brw_shader *)prog->_LinkedShaders[MESA_SHADER_FRAGMENT];
 
    memset(&prog_data, 0, sizeof(prog_data));
    /* key->alpha_test_func means simulating alpha testing via discards,
@@ -180,7 +182,7 @@ brw_codegen_wm_prog(struct brw_context *brw,
       fp->program.Base.OutputsWritten & BITFIELD64_BIT(FRAG_RESULT_SAMPLE_MASK);
    prog_data.computed_depth_mode = computed_depth_mode(&fp->program);
 
-   prog_data.early_fragment_tests = fs && fs->EarlyFragmentTests;
+   prog_data.early_fragment_tests = fs && fs->base.EarlyFragmentTests;
 
    /* Use ALT floating point mode for ARB programs so that 0^0 == 1. */
    if (!prog)
@@ -192,9 +194,9 @@ brw_codegen_wm_prog(struct brw_context *brw,
     */
    int param_count;
    if (fs) {
-      param_count = fs->num_uniform_components +
-                    fs->NumImages * BRW_IMAGE_PARAM_SIZE;
-      prog_data.base.nr_image_params = fs->NumImages;
+      param_count = fs->base.num_uniform_components +
+                    fs->base.NumImages * BRW_IMAGE_PARAM_SIZE;
+      prog_data.base.nr_image_params = fs->base.NumImages;
    } else {
       param_count = fp->program.Base.Parameters->NumParameters * 4;
    }
@@ -214,6 +216,12 @@ brw_codegen_wm_prog(struct brw_context *brw,
                                            key->persample_shading,
                                            &fp->program);
 
+   if (unlikely(brw->perf_debug)) {
+      start_busy = (brw->batch.last_bo &&
+                    drm_intel_bo_busy(brw->batch.last_bo));
+      start_time = get_time();
+   }
+
    program = brw_wm_fs_emit(brw, mem_ctx, key, &prog_data,
                             &fp->program, prog, &program_size);
    if (program == NULL) {
@@ -221,6 +229,17 @@ brw_codegen_wm_prog(struct brw_context *brw,
       return false;
    }
 
+   if (unlikely(brw->perf_debug) && fs) {
+      if (fs->compiled_once)
+         brw_wm_debug_recompile(brw, prog, key);
+      fs->compiled_once = true;
+
+      if (start_busy && !drm_intel_bo_busy(brw->batch.last_bo)) {
+         perf_debug("FS compile took %.03f ms and stalled the GPU\n",
+                    (get_time() - start_time) * 1000);
+      }
+   }
+
    if (prog_data.base.total_scratch) {
       brw_get_scratch_bo(brw, &brw->wm.base.scratch_bo,
                         prog_data.base.total_scratch * brw->max_wm_threads);