i965: Add perf debug for stalls during shader compiles.

v2: fix bad comment from before I gave up and decided to just use doubles.

Reviewed-by: Jordan Justen <[email protected]>
Reviewed-by: Kenneth Graunke <[email protected]>

4 files changed, 45 insertions, 2 deletions

diff --git a/src/mesa/drivers/dri/i965/brw_fs.cpp b/src/mesa/drivers/dri/i965/brw_fs.cpp
index 642d95a9289..3010ed0b435 100644
--- a/src/mesa/drivers/dri/i965/brw_fs.cpp
+++ b/src/mesa/drivers/dri/i965/brw_fs.cpp
@@ -2062,10 +2062,18 @@ brw_wm_fs_emit(struct brw_context *brw, struct brw_wm_compile *c,
 	       struct gl_shader_program *prog)
 {
    struct intel_context *intel = &brw->intel;
+   bool start_busy = false;
+   float start_time = 0;
 
    if (!prog)
       return false;
 
+   if (unlikely(INTEL_DEBUG & DEBUG_PERF)) {
+      start_busy = (intel->batch.last_bo &&
+                    drm_intel_bo_busy(intel->batch.last_bo));
+      start_time = get_time();
+   }
+
    struct brw_shader *shader =
      (brw_shader *) prog->_LinkedShaders[MESA_SHADER_FRAGMENT];
    if (!shader)
@@ -2108,6 +2116,11 @@ brw_wm_fs_emit(struct brw_context *brw, struct brw_wm_compile *c,
       if (shader->compiled_once)
          brw_wm_debug_recompile(brw, prog, &c->key);
       shader->compiled_once = true;
+
+      if (start_busy && !drm_intel_bo_busy(intel->batch.last_bo)) {
+         perf_debug("FS compile took %.03f ms and stalled the GPU\n",
+                    (get_time() - start_time) / 1000);
+      }
    }
 
    return true;
diff --git a/src/mesa/drivers/dri/i965/brw_vec4_emit.cpp b/src/mesa/drivers/dri/i965/brw_vec4_emit.cpp
index aea7d015a0e..d7ff8de3486 100644
--- a/src/mesa/drivers/dri/i965/brw_vec4_emit.cpp
+++ b/src/mesa/drivers/dri/i965/brw_vec4_emit.cpp
@@ -1023,9 +1023,19 @@ extern "C" {
 bool
 brw_vs_emit(struct gl_shader_program *prog, struct brw_vs_compile *c)
 {
+   struct intel_context *intel = &c->func.brw->intel;
+   bool start_busy = false;
+   float start_time = 0;
+
    if (!prog)
       return false;
 
+   if (unlikely(INTEL_DEBUG & DEBUG_PERF)) {
+      start_busy = (intel->batch.last_bo &&
+                    drm_intel_bo_busy(intel->batch.last_bo));
+      start_time = get_time();
+   }
+
    struct brw_shader *shader =
      (brw_shader *) prog->_LinkedShaders[MESA_SHADER_VERTEX];
    if (!shader)
@@ -1037,8 +1047,14 @@ brw_vs_emit(struct gl_shader_program *prog, struct brw_vs_compile *c)
       printf("\n\n");
    }
 
-   if (shader->compiled_once) {
-      perf_debug("Recompiling vertex shader for program %d\n", prog->Name);
+   if (unlikely(INTEL_DEBUG & DEBUG_PERF)) {
+      if (shader->compiled_once) {
+         perf_debug("Recompiling vertex shader for program %d\n", prog->Name);
+      }
+      if (start_busy && !drm_intel_bo_busy(intel->batch.last_bo)) {
+         perf_debug("VS compile took %.03f ms and stalled the GPU\n",
+                    (get_time() - start_time) / 1000);
+      }
    }
 
    vec4_visitor v(c, prog, shader);
diff --git a/src/mesa/drivers/dri/intel/intel_screen.c b/src/mesa/drivers/dri/intel/intel_screen.c
index 3c595bc6e75..e1ec2eb0130 100644
--- a/src/mesa/drivers/dri/intel/intel_screen.c
+++ b/src/mesa/drivers/dri/intel/intel_screen.c
@@ -109,6 +109,19 @@ const GLuint __driNConfigOptions = 15;
 static PFNGLXCREATECONTEXTMODES create_context_modes = NULL;
 #endif /*USE_NEW_INTERFACE */
 
+/**
+ * For debugging purposes, this returns a time in seconds.
+ */
+double
+get_time(void)
+{
+   struct timespec tp;
+
+   clock_gettime(CLOCK_MONOTONIC, &tp);
+
+   return tp.tv_sec + tp.tv_nsec / 1000000000.0;
+}
+
 void
 aub_dump_bmp(struct gl_context *ctx)
 {
diff --git a/src/mesa/drivers/dri/intel/intel_screen.h b/src/mesa/drivers/dri/intel/intel_screen.h
index c0cc2843fcb..f5a374d2293 100644
--- a/src/mesa/drivers/dri/intel/intel_screen.h
+++ b/src/mesa/drivers/dri/intel/intel_screen.h
@@ -81,6 +81,7 @@ intelMakeCurrent(__DRIcontext * driContextPriv,
                  __DRIdrawable * driDrawPriv,
                  __DRIdrawable * driReadPriv);
 
+double get_time(void);
 void aub_dump_bmp(struct gl_context *ctx);
 
 #endif