i#4857: Add instr count to drmemtrace filtered traces (#4864)

Adds tracking of the dynamic (pre-filtered) instruction count for each thread.
Stores the count in a marker at thread exit (storing prior to each miss entry
might be useful but in experiments it adds too much overhead to include
without a clear mandate and possibly under an off-by-default runtime option).

Adds a check to the filter tests confirming that the instruction count
marker is present at thread exit for filtered traces.  This is done by
invoked test_mode for these traces.  To simplify testing for online
and offline, the trace type marker is added to online traces.

Adds a new tool.drcachesim.filter-asm test which ensures the instr
count is the precise count expected for that fully-deterministic
tiny static binary.

A larger-than-static-asm application was manually tested, confirming
the need for the rep string expansion special case in the code to
limit to 1 instruction for that block.

Fixes #4857

clients/drcachesim/analyzer_multi.cpp

@@ -1,5 +1,5 @@
 /* **********************************************************
- * Copyright (c) 2016-2020 Google, Inc.  All rights reserved.
+ * Copyright (c) 2016-2021 Google, Inc.  All rights reserved.
  * **********************************************************/
 
 /*
@@ -174,8 +174,8 @@ analyzer_multi_t::create_analysis_tools()
     num_tools_ = 1;
 #ifdef DEBUG
     if (op_test_mode.get_value()) {
-        tools_[1] =
-            new trace_invariants_t(op_offline.get_value(), op_verbose.get_value());
+        tools_[1] = new trace_invariants_t(op_offline.get_value(), op_verbose.get_value(),
+                                           op_test_mode_name.get_value());
         if (tools_[1] == NULL)
             return false;
         if (!!*tools_[1])

clients/drcachesim/common/options.cpp

@@ -161,13 +161,13 @@ droption_t<std::string> op_LL_miss_file(
 
 droption_t<bool> op_L0_filter(
     DROPTION_SCOPE_CLIENT, "L0_filter", false,
-    "Filter out zero-level hits during tracing",
+    "Filter out first-level cache hits during tracing",
     "Filters out instruction and data hits in a 'zero-level' cache during tracing "
-    "itself, "
-    "shrinking the final trace to only contain instruction and data accesses that miss "
-    "in "
-    "this initial cache.  This cache is direct-mapped with sizes equal to -L0I_size and "
-    "-L0D_size.  It uses virtual addresses regardless of -use_physical.");
+    "itself, shrinking the final trace to only contain instruction and data accesses "
+    "that miss in this initial cache.  This cache is direct-mapped with sizes equal to "
+    "-L0I_size and -L0D_size.  It uses virtual addresses regardless of -use_physical. "
+    "The dynamic (pre-filtered) per-thread instruction count is tracked and supplied "
+    "via a #TRACE_MARKER_TYPE_INSTRUCTION_COUNT marker at thread exit.");
 
 droption_t<bytesize_t> op_L0I_size(
     DROPTION_SCOPE_CLIENT, "L0I_size", 32 * 1024U,
@@ -327,6 +327,9 @@ droption_t<bool>
 #ifdef DEBUG
 droption_t<bool> op_test_mode(DROPTION_SCOPE_ALL, "test_mode", false, "Run sanity tests",
                               "Run extra analyses for sanity checks on the trace.");
+droption_t<std::string> op_test_mode_name(
+    DROPTION_SCOPE_ALL, "test_mode_name", "", "Run custom sanity tests",
+    "Run extra analyses for specific sanity checks by name on the trace.");
 #endif
 droption_t<bool> op_disable_optimizations(
     DROPTION_SCOPE_ALL, "disable_optimizations", false,

clients/drcachesim/common/options.h

@@ -1,5 +1,5 @@
 /* **********************************************************
- * Copyright (c) 2015-2020 Google, Inc.  All rights reserved.
+ * Copyright (c) 2015-2021 Google, Inc.  All rights reserved.
  * **********************************************************/
 
 /*
@@ -106,6 +106,7 @@ extern droption_t<bool> op_show_func_trace;
 extern droption_t<int> op_jobs;
 #ifdef DEBUG
 extern droption_t<bool> op_test_mode;
+extern droption_t<std::string> op_test_mode_name;
 #endif
 extern droption_t<bool> op_disable_optimizations;
 extern droption_t<std::string> op_dr_root;

clients/drcachesim/common/trace_entry.h

@@ -1,5 +1,5 @@
 /* **********************************************************
- * Copyright (c) 2015-2020 Google, Inc.  All rights reserved.
+ * Copyright (c) 2015-2021 Google, Inc.  All rights reserved.
  * **********************************************************/
 
 /*
@@ -280,6 +280,13 @@ typedef enum {
      */
     TRACE_MARKER_TYPE_CACHE_LINE_SIZE,
 
+    /**
+     * The marker value contains the count of dynamic instruction executions in
+     * this software thread since the start of the trace.  This marker type is only
+     * present in online-cache-filtered traces and is placed at thread exit.
+     */
+    TRACE_MARKER_TYPE_INSTRUCTION_COUNT,
+
     // ...
     // These values are reserved for future built-in marker types.
     // ...
@@ -403,7 +410,8 @@ typedef enum {
 
 /**
  * Bitfields used to describe the high-level characteristics of both an
- * offline final trace and a raw not-yet-postprocessed trace.
+ * offline final trace and a raw not-yet-postprocessed trace, as well as
+ * (despite the OFFLINE_ prefix) an online trace.
  * In a final trace these are stored in a marker of type #TRACE_MARKER_TYPE_FILETYPE.
  */
 typedef enum {

clients/drcachesim/drcachesim.dox.in

@@ -1010,6 +1010,10 @@ closely match the actual hardware if so desired.
 Traces also include markers indicating disruptions in user mode control
 flow such as signal handler entry and exit.
 
+Filtered traces (filtered via -L0_filter) include the dynamic (pre-filtered)
+per-thread instruction count in a #TRACE_MARKER_TYPE_INSTRUCTION_COUNT marker at
+thread exit.
+
 A final feature that aids core simulators is the pair of interfaces
 module_mapper_t::get_loaded_modules() and
 module_mapper_t::find_mapped_trace_address(), which facilitate reading the raw

clients/drcachesim/tests/filter-asm.templatex

@@ -0,0 +1,39 @@
+Hello world!
+Hello world!
+Hello world!
+Hello world!
+Hello world!
+Hello world!
+Hello world!
+Hello world!
+Hello world!
+Hello world!
+Hello world!
+Hello world!
+Hello world!
+Hello world!
+Hello world!
+Hello world!
+---- <application exited with code 0> ----
+Cache simulation results:
+Core #0 \(1 thread\(s\)\)
+  L1I stats:
+    Hits:                                0
+    Misses:                             .*
+    Invalidations:                       0
+    Miss rate:                      100.00%
+  L1D stats:
+    Hits:                                0
+    Misses:                              0
+    Invalidations:                       0
+Core #1 \(0 thread\(s\)\)
+Core #2 \(0 thread\(s\)\)
+Core #3 \(0 thread\(s\)\)
+LL stats:
+    Hits:                                0
+    Misses:                             .*
+    Invalidations:                       0
+    Miss rate:                      100.00%
+
+===========================================================================
+Trace invariant checks passed

clients/drcachesim/tests/filter-no-d.templatex

@@ -20,4 +20,7 @@ LL stats:
     Invalidations:                *0
 .*   Local miss rate:        *[0-9,.]*%
     Child hits:                   *[0-9,\.]*
-    Total miss rate:              *[1-9]?[0-9][,\.]..%
+    Total miss rate:              *[1-9]?[0-9][,\.]..%(
+
+===========================================================================
+Trace invariant checks passed)?

clients/drcachesim/tests/filter-no-i.templatex

@@ -20,4 +20,7 @@ LL stats:
     Invalidations:                *0
 .*   Local miss rate:        *[0-9,.]*%
     Child hits:                   *[0-9,\.]*
-    Total miss rate:              *[1-9]?[0-9][,\.]..%
+    Total miss rate:              *[1-9]?[0-9][,\.]..%(
+
+===========================================================================
+Trace invariant checks passed)?

clients/drcachesim/tests/filter-simple.templatex

@@ -21,4 +21,7 @@ LL stats:
     Invalidations:                *0
 .*   Local miss rate:        *[0-9,.]*%
     Child hits:                   *[0-9,\.]*
-    Total miss rate:              *[1-9]?[0-9][,\.]..%
+    Total miss rate:              *[1-9]?[0-9][,\.]..%(
+
+===========================================================================
+Trace invariant checks passed)?

clients/drcachesim/tests/trace_invariants.cpp

@@ -1,5 +1,5 @@
 /* **********************************************************
- * Copyright (c) 2017-2020 Google, Inc.  All rights reserved.
+ * Copyright (c) 2017-2021 Google, Inc.  All rights reserved.
  * **********************************************************/
 
 /*
@@ -35,9 +35,11 @@
 #include <iostream>
 #include <string.h>
 
-trace_invariants_t::trace_invariants_t(bool offline, unsigned int verbose)
+trace_invariants_t::trace_invariants_t(bool offline, unsigned int verbose,
+                                       std::string test_name)
     : knob_offline_(offline)
     , knob_verbose_(verbose)
+    , knob_test_name_(test_name)
     , instrs_until_interrupt_(-1)
     , memrefs_until_interrupt_(-1)
     , app_handler_pc_(0)
@@ -83,12 +85,28 @@ trace_invariants_t::process_memref(const memref_t &memref)
         app_handler_pc_ = prev_entry_.instr.addr;
     }
 
+    if (memref.marker.type == TRACE_TYPE_MARKER &&
+        memref.marker.marker_type == TRACE_MARKER_TYPE_FILETYPE) {
+        file_type_ = static_cast<offline_file_type_t>(memref.marker.marker_value);
+    }
+    if (memref.marker.type == TRACE_TYPE_MARKER &&
+        memref.marker.marker_type == TRACE_MARKER_TYPE_INSTRUCTION_COUNT) {
+        found_instr_count_marker_[memref.marker.tid] = true;
+        if (knob_test_name_ == "filter_asm_instr_count") {
+#ifndef NDEBUG
+            static constexpr int ASM_INSTR_COUNT = 133;
+#endif
+            assert(memref.marker.marker_value == ASM_INSTR_COUNT);
+        }
+    }
     if (memref.marker.type == TRACE_TYPE_MARKER &&
         memref.marker.marker_type == TRACE_MARKER_TYPE_CACHE_LINE_SIZE) {
         found_cache_line_size_marker_[memref.marker.tid] = true;
     }
 
     if (memref.exit.type == TRACE_TYPE_THREAD_EXIT) {
+        assert(!TESTALL(OFFLINE_FILE_TYPE_FILTERED, file_type_) ||
+               found_instr_count_marker_[memref.exit.tid]);
         assert(found_cache_line_size_marker_[memref.exit.tid]);
         thread_exited_[memref.exit.tid] = true;
     }
@@ -122,7 +140,9 @@ trace_invariants_t::process_memref(const memref_t &memref)
         if (prev_instr_.instr.addr != 0 /*first*/ &&
             prev_instr_.instr.tid == memref.instr.tid &&
             !type_is_instr_branch(prev_instr_.instr.type)) {
-            assert( // Regular fall-through.
+            assert( // Filtered.
+                TESTALL(OFFLINE_FILE_TYPE_FILTERED, file_type_) ||
+                // Regular fall-through.
                 (prev_instr_.instr.addr + prev_instr_.instr.size == memref.instr.addr) ||
                 // String loop.
                 (prev_instr_.instr.addr == memref.instr.addr &&

clients/drcachesim/tests/trace_invariants.h

@@ -1,5 +1,5 @@
 /* **********************************************************
- * Copyright (c) 2016-2020 Google, Inc.  All rights reserved.
+ * Copyright (c) 2016-2021 Google, Inc.  All rights reserved.
  * **********************************************************/
 
 /*
@@ -42,7 +42,8 @@
 
 class trace_invariants_t : public analysis_tool_t {
 public:
-    trace_invariants_t(bool offline = true, unsigned int verbose = 0);
+    trace_invariants_t(bool offline = true, unsigned int verbose = 0,
+                       std::string test_name = "");
     virtual ~trace_invariants_t();
     bool
     process_memref(const memref_t &memref) override;
@@ -52,6 +53,7 @@ class trace_invariants_t : public analysis_tool_t {
 protected:
     bool knob_offline_;
     unsigned int knob_verbose_;
+    std::string knob_test_name_;
     memref_t prev_instr_;
     memref_t prev_xfer_marker_;
     memref_t prev_entry_;
@@ -60,8 +62,10 @@ class trace_invariants_t : public analysis_tool_t {
     int instrs_until_interrupt_;
     int memrefs_until_interrupt_;
     addr_t app_handler_pc_;
+    offline_file_type_t file_type_ = OFFLINE_FILE_TYPE_DEFAULT;
     std::unordered_map<memref_tid_t, bool> thread_exited_;
     std::unordered_map<memref_tid_t, bool> found_cache_line_size_marker_;
+    std::unordered_map<memref_tid_t, bool> found_instr_count_marker_;
 };
 
 #endif /* _TRACE_INVARIANTS_H_ */

clients/drcachesim/tracer/instru.h

@@ -1,5 +1,5 @@
 /* **********************************************************
- * Copyright (c) 2016-2020 Google, Inc.  All rights reserved.
+ * Copyright (c) 2016-2021 Google, Inc.  All rights reserved.
  * **********************************************************/
 
 /*
@@ -312,6 +312,8 @@ class online_instru_t : public instru_t {
     append_iflush(byte *buf_ptr, addr_t start, size_t size) override;
     int
     append_thread_header(byte *buf_ptr, thread_id_t tid) override;
+    virtual int
+    append_thread_header(byte *buf_ptr, thread_id_t tid, offline_file_type_t file_type);
     int
     append_unit_header(byte *buf_ptr, thread_id_t tid) override;
 

clients/drcachesim/tracer/instru_online.cpp

@@ -1,5 +1,5 @@
 /* **********************************************************
- * Copyright (c) 2016-2020 Google, Inc.  All rights reserved.
+ * Copyright (c) 2016-2021 Google, Inc.  All rights reserved.
  * **********************************************************/
 
 /*
@@ -138,17 +138,25 @@ online_instru_t::append_iflush(byte *buf_ptr, addr_t start, size_t size)
 }
 
 int
-online_instru_t::append_thread_header(byte *buf_ptr, thread_id_t tid)
+online_instru_t::append_thread_header(byte *buf_ptr, thread_id_t tid,
+                                      offline_file_type_t file_type)
 {
     byte *new_buf = buf_ptr;
     new_buf += append_tid(new_buf, tid);
     new_buf += append_pid(new_buf, dr_get_process_id());
 
+    new_buf += append_marker(new_buf, TRACE_MARKER_TYPE_FILETYPE, file_type);
     new_buf += append_marker(new_buf, TRACE_MARKER_TYPE_CACHE_LINE_SIZE,
                              proc_get_cache_line_size());
     return (int)(new_buf - buf_ptr);
 }
 
+int
+online_instru_t::append_thread_header(byte *buf_ptr, thread_id_t tid)
+{
+    return append_thread_header(buf_ptr, tid, OFFLINE_FILE_TYPE_DEFAULT);
+}
+
 int
 online_instru_t::append_unit_header(byte *buf_ptr, thread_id_t tid)
 {