i965: Make perf_debug() output to GL_ARB_debug_output in a debug context.

I tried to ensure that performance in the non-debug case doesn't change
(we still just check one condition up front), and I think the impact is
small enough in the debug context case to warrant including all of it.

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

6 files changed, 21 insertions, 7 deletions

diff --git a/src/mesa/drivers/dri/intel/intel_buffer_objects.c b/src/mesa/drivers/dri/intel/intel_buffer_objects.c
index f94c6f570ff..34eb7c9ea2c 100644
--- a/src/mesa/drivers/dri/intel/intel_buffer_objects.c
+++ b/src/mesa/drivers/dri/intel/intel_buffer_objects.c
@@ -227,7 +227,7 @@ intel_bufferobj_subdata(struct gl_context * ctx,
 	 drm_intel_bo_unreference(temp_bo);
       }
    } else {
-      if (unlikely(INTEL_DEBUG & DEBUG_PERF)) {
+      if (unlikely(intel->perf_debug)) {
          if (drm_intel_bo_busy(intel_obj->buffer)) {
             perf_debug("Stalling on the GPU in glBufferSubData().\n");
          }
diff --git a/src/mesa/drivers/dri/intel/intel_context.c b/src/mesa/drivers/dri/intel/intel_context.c
index 435f0974445..9e508f7bb9b 100644
--- a/src/mesa/drivers/dri/intel/intel_context.c
+++ b/src/mesa/drivers/dri/intel/intel_context.c
@@ -803,6 +803,8 @@ intelInitContext(struct intel_context *intel,
               "shader_time debugging requires gen7 (Ivybridge) or better.\n");
       INTEL_DEBUG &= ~DEBUG_SHADER_TIME;
    }
+   if (INTEL_DEBUG & DEBUG_PERF)
+      intel->perf_debug = true;
 
    if (INTEL_DEBUG & DEBUG_AUB)
       drm_intel_bufmgr_gem_set_aub_dump(intel->bufmgr, true);
diff --git a/src/mesa/drivers/dri/intel/intel_context.h b/src/mesa/drivers/dri/intel/intel_context.h
index c72176ad416..42db5a7d96d 100644
--- a/src/mesa/drivers/dri/intel/intel_context.h
+++ b/src/mesa/drivers/dri/intel/intel_context.h
@@ -239,6 +239,13 @@ struct intel_context
    bool no_batch_wrap;
    bool tnl_pipeline_running; /**< Set while i915's _tnl_run_pipeline. */
 
+   /**
+    * Set if we're either a debug context or the INTEL_DEBUG=perf environment
+    * variable is set, this is the flag indicating to do expensive work that
+    * might lead to a perf_debug() call.
+    */
+   bool perf_debug;
+
    struct
    {
       GLuint id;
@@ -462,8 +469,14 @@ extern int INTEL_DEBUG;
 } while(0)
 
 #define perf_debug(...) do {					\
-	if (unlikely(INTEL_DEBUG & DEBUG_PERF))			\
-		dbg_printf(__VA_ARGS__);			\
+   static GLuint msg_id = 0;                                    \
+   if (unlikely(INTEL_DEBUG & DEBUG_PERF))                      \
+      dbg_printf(__VA_ARGS__);                                  \
+   if (intel->perf_debug)                                       \
+      _mesa_gl_debug(&intel->ctx, &msg_id,                      \
+                     MESA_DEBUG_TYPE_PERFORMANCE,               \
+                     MESA_DEBUG_SEVERITY_MEDIUM,                \
+                     __VA_ARGS__);                              \
 } while(0)
 
 #define WARN_ONCE(cond, fmt...) do {                            \
diff --git a/src/mesa/drivers/dri/intel/intel_regions.c b/src/mesa/drivers/dri/intel/intel_regions.c
index a5e49d99e9f..7f535d7cdc4 100644
--- a/src/mesa/drivers/dri/intel/intel_regions.c
+++ b/src/mesa/drivers/dri/intel/intel_regions.c
@@ -123,7 +123,7 @@ intel_region_map(struct intel_context *intel, struct intel_region *region,
     * flush is only needed on first map of the buffer.
     */
 
-   if (unlikely(INTEL_DEBUG & DEBUG_PERF)) {
+   if (unlikely(intel->perf_debug)) {
       if (drm_intel_bo_busy(region->bo)) {
          perf_debug("Mapping a busy BO, causing a stall on the GPU.\n");
       }
diff --git a/src/mesa/drivers/dri/intel/intel_tex_copy.c b/src/mesa/drivers/dri/intel/intel_tex_copy.c
index 085b6fdbd35..6043ed26adb 100644
--- a/src/mesa/drivers/dri/intel/intel_tex_copy.c
+++ b/src/mesa/drivers/dri/intel/intel_tex_copy.c
@@ -114,8 +114,7 @@ intel_copy_texsubimage(struct intel_context *intel,
    }
 
    if (!copy_supported && !copy_supported_with_alpha_override) {
-      if (unlikely(INTEL_DEBUG & DEBUG_PERF))
-	 fprintf(stderr, "%s mismatched formats %s, %s\n",
+      perf_debug("%s mismatched formats %s, %s\n",
 		 __FUNCTION__,
 		 _mesa_get_format_name(intelImage->base.Base.TexFormat),
 		 _mesa_get_format_name(intel_rb_format(irb)));
diff --git a/src/mesa/drivers/dri/intel/intel_tex_subimage.c b/src/mesa/drivers/dri/intel/intel_tex_subimage.c
index 7a2f713c88f..b02e5fc549a 100644
--- a/src/mesa/drivers/dri/intel/intel_tex_subimage.c
+++ b/src/mesa/drivers/dri/intel/intel_tex_subimage.c
@@ -230,7 +230,7 @@ intel_texsubimage_tiled_memcpy(struct gl_context * ctx,
       intel_batchbuffer_flush(intel);
    }
 
-   if (unlikely(INTEL_DEBUG & DEBUG_PERF)) {
+   if (unlikely(intel->perf_debug)) {
       if (drm_intel_bo_busy(bo)) {
          perf_debug("Mapping a busy BO, causing a stall on the GPU.\n");
       }