add option for dbg prints

4 files changed, 50 insertions, 31 deletions

diff --git a/src/abstract_hardware_model.h b/src/abstract_hardware_model.h
index 3f7eac3..698c896 100644
--- a/src/abstract_hardware_model.h
+++ b/src/abstract_hardware_model.h
@@ -1404,7 +1404,7 @@ class checkpoint {
   unsigned radnom;
 };
 // Debug print macro for ITS (disabled by default)
-#define AWARE_DEBUG_PRINT 1
+#define AWARE_DEBUG_PRINT 0
 #define AWARE_DPRINTF(...)      \
   if (AWARE_DEBUG_PRINT) {      \
     printf(__VA_ARGS__);        \
diff --git a/src/gpgpu-sim/gpu-sim.cc b/src/gpgpu-sim/gpu-sim.cc
index 7b947fb..cd38857 100644
--- a/src/gpgpu-sim/gpu-sim.cc
+++ b/src/gpgpu-sim/gpu-sim.cc
@@ -703,6 +703,14 @@ void shader_core_config::reg_options(class OptionParser *opp) {
       "2=inter-first (matches pre-Change-4 behavior), 3=fewest-lanes, "
       "4=same-PC (default 2).",
       "2");
+  option_parser_register(
+      opp, "-gpgpu_simd_partitioning_debug", OPT_BOOL,
+      &gpgpu_simd_partitioning_debug,
+      "Enable per-cycle `SIMD_SETS:` printf spam from the co-issue "
+      "scheduler (default off). Aggregate counts are always available in "
+      "the `gpu_tot_inter_warp_coissue` / `gpu_tot_intra_warp_coissue` "
+      "stats; only turn this on for quick_tests that grep the raw lines.",
+      "0");
 
   for (unsigned j = 0; j < SPECIALIZED_UNIT_NUM; ++j) {
     std::stringstream ss;
diff --git a/src/gpgpu-sim/shader.cc b/src/gpgpu-sim/shader.cc
index 921fa60..d433b6a 100644
--- a/src/gpgpu-sim/shader.cc
+++ b/src/gpgpu-sim/shader.cc
@@ -1747,13 +1747,15 @@ void scheduler_unit::try_inter_warp_coissue(
         continue;
       }
 
-      printf(
-          "SIMD_SETS: cycle %llu, core %u, sched %u: warp %u "
-          "CO-ISSUED with primary warp %u (%u sets)\n",
-          m_shader->get_gpu()->gpu_sim_cycle +
-              m_shader->get_gpu()->gpu_tot_sim_cycle,
-          get_sid(), m_id, cand_warp_id, co_issue_primary_warp_id,
-          cand_sets_needed);
+      if (m_shader->m_config->gpgpu_simd_partitioning_debug) {
+        printf(
+            "SIMD_SETS: cycle %llu, core %u, sched %u: warp %u "
+            "CO-ISSUED with primary warp %u (%u sets)\n",
+            m_shader->get_gpu()->gpu_sim_cycle +
+                m_shader->get_gpu()->gpu_tot_sim_cycle,
+            get_sid(), m_id, cand_warp_id, co_issue_primary_warp_id,
+            cand_sets_needed);
+      }
 
       m_shader->co_issue_warp(co_issue_composite, cand_inst, cand_mask,
                               cand_warp_id, m_id, next_free_set);
@@ -1824,12 +1826,14 @@ void scheduler_unit::try_intra_warp_coissue(
       continue;
     }
 
-    printf(
-        "SIMD_SETS: cycle %llu, core %u, sched %u: INTRA-WARP warp %u "
-        "split %u CO-ISSUED with primary split (%u sets)\n",
-        m_shader->get_gpu()->gpu_sim_cycle +
-            m_shader->get_gpu()->gpu_tot_sim_cycle,
-        get_sid(), m_id, primary_warp_id, sec_split_id, sec_sets_needed);
+    if (m_shader->m_config->gpgpu_simd_partitioning_debug) {
+      printf(
+          "SIMD_SETS: cycle %llu, core %u, sched %u: INTRA-WARP warp %u "
+          "split %u CO-ISSUED with primary split (%u sets)\n",
+          m_shader->get_gpu()->gpu_sim_cycle +
+              m_shader->get_gpu()->gpu_tot_sim_cycle,
+          get_sid(), m_id, primary_warp_id, sec_split_id, sec_sets_needed);
+    }
 
     m_shader->co_issue_warp(co_issue_composite, sec_inst, sec_mask,
                             primary_warp_id, m_id, next_free_set,
@@ -2007,12 +2011,14 @@ void scheduler_unit::try_utilization_max_coissue(
     }
 
     if (c.is_intra) {
-      printf(
-          "SIMD_SETS: cycle %llu, core %u, sched %u: INTRA-WARP warp %u "
-          "split %u CO-ISSUED with primary split (%u sets)\n",
-          m_shader->get_gpu()->gpu_sim_cycle +
-              m_shader->get_gpu()->gpu_tot_sim_cycle,
-          get_sid(), m_id, c.warp_id, c.split_id, needed);
+      if (m_shader->m_config->gpgpu_simd_partitioning_debug) {
+        printf(
+            "SIMD_SETS: cycle %llu, core %u, sched %u: INTRA-WARP warp %u "
+            "split %u CO-ISSUED with primary split (%u sets)\n",
+            m_shader->get_gpu()->gpu_sim_cycle +
+                m_shader->get_gpu()->gpu_tot_sim_cycle,
+            get_sid(), m_id, c.warp_id, c.split_id, needed);
+      }
 
       m_shader->co_issue_warp(co_issue_composite, c.inst, c.mask, c.warp_id,
                               m_id, next_free_set, c.split_id);
@@ -2036,12 +2042,14 @@ void scheduler_unit::try_utilization_max_coissue(
         }
       }
     } else {
-      printf(
-          "SIMD_SETS: cycle %llu, core %u, sched %u: warp %u "
-          "CO-ISSUED with primary warp %u (%u sets)\n",
-          m_shader->get_gpu()->gpu_sim_cycle +
-              m_shader->get_gpu()->gpu_tot_sim_cycle,
-          get_sid(), m_id, c.warp_id, co_issue_primary_warp_id, needed);
+      if (m_shader->m_config->gpgpu_simd_partitioning_debug) {
+        printf(
+            "SIMD_SETS: cycle %llu, core %u, sched %u: warp %u "
+            "CO-ISSUED with primary warp %u (%u sets)\n",
+            m_shader->get_gpu()->gpu_sim_cycle +
+                m_shader->get_gpu()->gpu_tot_sim_cycle,
+            get_sid(), m_id, c.warp_id, co_issue_primary_warp_id, needed);
+      }
 
       m_shader->co_issue_warp(co_issue_composite, c.inst, c.mask, c.warp_id,
                               m_id, next_free_set);
@@ -2412,11 +2420,13 @@ void scheduler_unit::cycle() {
     }
 
     if (available_sets > 0) {
-      printf("SIMD_SETS: cycle %llu, core %u, sched %u: %u sets available "
-             "after primary warp %u\n",
-             m_shader->get_gpu()->gpu_sim_cycle +
-                 m_shader->get_gpu()->gpu_tot_sim_cycle,
-             get_sid(), m_id, available_sets, co_issue_primary_warp_id);
+      if (m_shader->m_config->gpgpu_simd_partitioning_debug) {
+        printf("SIMD_SETS: cycle %llu, core %u, sched %u: %u sets available "
+               "after primary warp %u\n",
+               m_shader->get_gpu()->gpu_sim_cycle +
+                   m_shader->get_gpu()->gpu_tot_sim_cycle,
+               get_sid(), m_id, available_sets, co_issue_primary_warp_id);
+      }
 
       switch (m_shader->m_config->gpgpu_co_issue_priority) {
         case 0:  // greedy (utilization-max, unified pool)
diff --git a/src/gpgpu-sim/shader.h b/src/gpgpu-sim/shader.h
index 70de529..42d807c 100644
--- a/src/gpgpu-sim/shader.h
+++ b/src/gpgpu-sim/shader.h
@@ -1960,6 +1960,7 @@ class shader_core_config : public core_config {
   unsigned gpgpu_compaction_mode;       // 0=none, 1=xor-static, 2=full
   unsigned gpgpu_opndcoll_read_latency; // extra cycles to collector
   unsigned gpgpu_co_issue_priority;     // 0=greedy..4=same-PC
+  bool gpgpu_simd_partitioning_debug;   // verbose SIMD_SETS printf gate
 
   unsigned n_simt_cores_per_cluster;
   unsigned n_simt_clusters;