From 81c247404a232d17e94ec1b435c0f2be47877778 Mon Sep 17 00:00:00 2001
From: Eric Anholt <eric@anholt.net>
Date: Mon, 10 Dec 2012 10:22:41 -0800
Subject: [PATCH] i965: Print a total time for the different shader stages.

Sometimes I've got a patch for a performance optimization that's not
showing a statistically significant performance difference on reported
FPS, but still seems like a good idea because it ought to reduce time
spent in the shader.  If I can see the total number of cycles spent in
the shader stage being optimized, it may show that the patch is still
worthwhile (or point out that it's actually broken in some way).
---
 src/mesa/drivers/dri/i965/brw_program.c | 48 +++++++++++++++++++------
 1 file changed, 38 insertions(+), 10 deletions(-)

diff --git a/src/mesa/drivers/dri/i965/brw_program.c b/src/mesa/drivers/dri/i965/brw_program.c
index 035824171e3..d7b240a8b9c 100644
--- a/src/mesa/drivers/dri/i965/brw_program.c
+++ b/src/mesa/drivers/dri/i965/brw_program.c
@@ -241,6 +241,21 @@ get_written_and_reset(struct brw_context *brw, int i,
    *reset = brw->shader_time.cumulative[ri];
 }
 
+static void
+print_shader_time_line(const char *name, int shader_num,
+                       uint64_t time, uint64_t total)
+{
+   printf("%s", name);
+   for (int i = strlen(name); i < 10; i++)
+      printf(" ");
+   printf("%4d: ", shader_num);
+
+   printf("%16lld (%7.2f Gcycles)      %4.1f%%\n",
+          (long long)time,
+          (double)time / 1000000000.0,
+          (double)time / total * 100.0);
+}
+
 static void
 brw_report_shader_time(struct brw_context *brw)
 {
@@ -249,13 +264,16 @@ brw_report_shader_time(struct brw_context *brw)
 
    uint64_t scaled[brw->shader_time.num_entries];
    uint64_t *sorted[brw->shader_time.num_entries];
+   uint64_t total_by_type[ST_FS16 + 1];
+   memset(total_by_type, 0, sizeof(total_by_type));
    double total = 0;
    for (int i = 0; i < brw->shader_time.num_entries; i++) {
       uint64_t written = 0, reset = 0;
+      enum shader_time_shader_type type = brw->shader_time.types[i];
 
       sorted[i] = &scaled[i];
 
-      switch (brw->shader_time.types[i]) {
+      switch (type) {
       case ST_VS_WRITTEN:
       case ST_VS_RESET:
       case ST_FS8_WRITTEN:
@@ -288,6 +306,16 @@ brw_report_shader_time(struct brw_context *brw)
          scaled[i] = time;
       }
 
+      switch (type) {
+      case ST_VS:
+      case ST_FS8:
+      case ST_FS16:
+         total_by_type[type] += scaled[i];
+         break;
+      default:
+         break;
+      }
+
       total += scaled[i];
    }
 
@@ -314,24 +342,24 @@ brw_report_shader_time(struct brw_context *brw)
 
       switch (brw->shader_time.types[i]) {
       case ST_VS:
-         printf("vs   %4d: ", shader_num);
+         print_shader_time_line("vs", shader_num, scaled[i], total);
          break;
       case ST_FS8:
-         printf("fs8  %4d: ", shader_num);
+         print_shader_time_line("fs8", shader_num, scaled[i], total);
          break;
       case ST_FS16:
-         printf("fs16 %4d: ", shader_num);
+         print_shader_time_line("fs16", shader_num, scaled[i], total);
          break;
       default:
-         printf("other:     ");
+         print_shader_time_line("other", shader_num, scaled[i], total);
          break;
       }
-
-      printf("%16lld (%7.2f Gcycles)      %4.1f%%\n",
-             (long long)scaled[i],
-             (double)scaled[i] / 1000000000.0,
-             (double)scaled[i] / total * 100.0);
    }
+
+   printf("\n");
+   print_shader_time_line("total vs", -1, total_by_type[ST_VS], total);
+   print_shader_time_line("total fs8", -1, total_by_type[ST_FS8], total);
+   print_shader_time_line("total fs16", -1, total_by_type[ST_FS16], total);
 }
 
 static void
-- 
2.30.2