Merge branch 'master' into i6662-v2p-tlb-simulator

.github/workflows/ci-package.yml

@@ -127,7 +127,8 @@ jobs:
         CI_BRANCH: ${{ github.ref }}
 
     - name: Upload Artifacts
-      uses: actions/upload-artifact@v2
+      # This points to the latest upload-artifact v4.x.x.
+      uses: actions/upload-artifact@v4
       with:
         name: linux-tarball
         path: DynamoRIO-Linux-${{ steps.version.outputs.version_number }}.tar.gz
@@ -215,7 +216,8 @@ jobs:
         CI_BRANCH: ${{ github.ref }}
 
     - name: Upload AArch64
-      uses: actions/upload-artifact@v2
+      # This points to the latest upload-artifact v4.x.x.
+      uses: actions/upload-artifact@v4
       with:
         name: aarch64-tarball
         path: DynamoRIO-AArch64-Linux-${{ steps.version.outputs.version_number }}.tar.gz
@@ -303,7 +305,8 @@ jobs:
         CI_BRANCH: ${{ github.ref }}
 
     - name: Upload ARM
-      uses: actions/upload-artifact@v2
+      # This points to the latest upload-artifact v4.x.x.
+      uses: actions/upload-artifact@v4
       with:
         name: arm-tarball
         path: DynamoRIO-ARM-Linux-EABIHF-${{ steps.version.outputs.version_number }}.tar.gz
@@ -393,7 +396,8 @@ jobs:
         CI_BRANCH: ${{ github.ref }}
 
     - name: Upload Artifacts
-      uses: actions/upload-artifact@v2
+      # This points to the latest upload-artifact v4.x.x.
+      uses: actions/upload-artifact@v4
       with:
         name: android-tarball
         path: DynamoRIO-ARM-Android-EABI-${{ steps.version.outputs.version_number }}.tar.gz
@@ -484,7 +488,8 @@ jobs:
         CI_BRANCH: ${{ github.ref }}
 
     - name: Upload Artifacts
-      uses: actions/upload-artifact@v2
+      # This points to the latest upload-artifact v4.x.x.
+      uses: actions/upload-artifact@v4
       with:
         name: windows-zip
         path: DynamoRIO-Windows-${{ steps.version.outputs.version_number }}.zip
@@ -562,7 +567,8 @@ jobs:
         prerelease: false
 
     - name: Download Linux
-      uses: actions/[email protected]
+      # This points to the latest download-artifact v4.x.x.
+      uses: actions/download-artifact@v4
       with:
         name: linux-tarball
     - name: Upload Linux
@@ -577,7 +583,8 @@ jobs:
         asset_content_type: application/x-gzip
 
     - name: Download AArch64
-      uses: actions/[email protected]
+      # This points to the latest download-artifact v4.x.x.
+      uses: actions/download-artifact@v4
       with:
         name: aarch64-tarball
     - name: Upload AArch64
@@ -592,7 +599,8 @@ jobs:
         asset_content_type: application/x-gzip
 
     - name: Download ARM
-      uses: actions/[email protected]
+      # This points to the latest download-artifact v4.x.x.
+      uses: actions/download-artifact@v4
       with:
         name: arm-tarball
     - name: Upload ARM
@@ -607,7 +615,8 @@ jobs:
         asset_content_type: application/x-gzip
 
     - name: Download Android
-      uses: actions/[email protected]
+      # This points to the latest download-artifact v4.x.x.
+      uses: actions/download-artifact@v4
       with:
         name: android-tarball
     - name: Upload Android
@@ -622,7 +631,8 @@ jobs:
         asset_content_type: application/x-gzip
 
     - name: Download Windows
-      uses: actions/[email protected]
+      # This points to the latest download-artifact v4.x.x.
+      uses: actions/download-artifact@v4
       with:
         name: windows-zip
     - name: Upload Windows

clients/drcachesim/analyzer_multi.cpp

@@ -558,13 +558,17 @@ analyzer_multi_tmpl_t<RecordType, ReaderType>::init_dynamic_schedule()
         op_sched_order_time.get_value() ? sched_type_t::DEPENDENCY_TIMESTAMPS
                                         : sched_type_t::DEPENDENCY_IGNORE,
         sched_type_t::SCHEDULER_DEFAULTS, op_verbose.get_value());
-    sched_ops.quantum_duration = op_sched_quantum.get_value();
-    if (op_sched_time.get_value())
+    sched_ops.time_units_per_us = op_sched_time_per_us.get_value();
+    if (op_sched_time.get_value()) {
         sched_ops.quantum_unit = sched_type_t::QUANTUM_TIME;
+        sched_ops.quantum_duration_us = op_sched_quantum.get_value();
+    } else {
+        sched_ops.quantum_duration_instrs = op_sched_quantum.get_value();
+    }
     sched_ops.syscall_switch_threshold = op_sched_syscall_switch_us.get_value();
     sched_ops.blocking_switch_threshold = op_sched_blocking_switch_us.get_value();
-    sched_ops.block_time_scale = op_sched_block_scale.get_value();
-    sched_ops.block_time_max = op_sched_block_max_us.get_value();
+    sched_ops.block_time_multiplier = op_sched_block_scale.get_value();
+    sched_ops.block_time_max_us = op_sched_block_max_us.get_value();
     sched_ops.randomize_next_input = op_sched_randomize.get_value();
     sched_ops.honor_direct_switches = !op_sched_disable_direct_switches.get_value();
 #ifdef HAS_ZIP

clients/drcachesim/common/options.cpp

@@ -897,13 +897,19 @@ droption_t<bool> op_core_serial(
     "How the scheduling is performed is controlled by a set "
     "of options with the prefix \"sched_\" along with -cores.");
 
+droption_t<double>
+    op_sched_time_per_us(DROPTION_SCOPE_ALL, "sched_time_per_us", 1000.,
+                         "Wall-clock microseconds per simulated microsecond",
+                         "Wall-clock microseconds per simulated microsecond.");
+
 droption_t<int64_t>
-    // We pick 6 million to match 2 instructions per nanosecond with a 3ms quantum.
-    op_sched_quantum(DROPTION_SCOPE_ALL, "sched_quantum", 6 * 1000 * 1000,
+    // We pick 10 million to match 2 instructions per nanosecond with a 5ms quantum.
+    op_sched_quantum(DROPTION_SCOPE_ALL, "sched_quantum", 10 * 1000 * 1000,
                      "Scheduling quantum",
                      "Applies to -core_sharded and -core_serial. "
-                     "Scheduling quantum: in microseconds of wall-clock "
-                     "time if -sched_time is set; otherwise in instructions.");
+                     "Scheduling quantum in instructions, unless -sched_time is set in "
+                     "which case this value is multiplied by -sched_time_per_us to "
+                     "produce a quantum in wall-clock microseconds.");
 
 droption_t<bool>
     op_sched_time(DROPTION_SCOPE_ALL, "sched_time", false,
@@ -932,23 +938,27 @@ droption_t<uint64_t> op_sched_blocking_switch_us(
     "-core_serial. ");
 
 droption_t<double> op_sched_block_scale(
-    DROPTION_SCOPE_ALL, "sched_block_scale", 10., "Input block time scale factor",
-    "The scale applied to the microsecond latency of blocking system calls.  A higher "
-    "value here results in blocking syscalls keeping inputs unscheduled for longer.  "
-    "This should roughly equal the slowdown of instruction record processing versus the "
-    "original (untraced) application execution.");
-
-// We have a max to avoid outlier latencies that are already a second or more from
-// scaling up to tens of minutes.  We assume a cap is representative as the outliers
+    DROPTION_SCOPE_ALL, "sched_block_scale", 0.1, "Input block time scale factor",
+    "A system call considered to block (see -sched_blocking_switch_us) will "
+    "block in the trace scheduler for an amount of simulator time equal to its "
+    "as-traced latency in trace-time microseconds multiplied by this parameter "
+    "and by -sched_time_per_us in simulated microseconds, subject to a "
+    "maximum of --sched_block_max_us. A higher value here results in blocking "
+    "syscalls keeping inputs unscheduled for longer. There is indirect "
+    "overhead inflating the as-traced times, so a value below 1 is typical.");
+
+// We have a max to avoid outlier latencies from scaling up to extreme times.  There is
+// some inflation in the as-traced latencies and some can be inflated more than others.
+// We assume a cap is representative as the outliers
 // likely were not part of key dependence chains.  Without a cap the other threads all
 // finish and the simulation waits for tens of minutes further for a couple of outliers.
 // The cap remains a flag and not a constant as different length traces and different
 // speed simulators need different idle time ranges, so we need to be able to tune this
 // to achieve desired cpu usage targets.  The default value was selected to avoid unduly
 // long idle times with local analyzers; it may need to be increased with more
 // heavyweight analyzers/simulators.
-droption_t<uint64_t> op_sched_block_max_us(DROPTION_SCOPE_ALL, "sched_block_max_us",
-                                           250000,
+// TODO i#6959: Once we have -exit_if_all_unscheduled raise this.
+droption_t<uint64_t> op_sched_block_max_us(DROPTION_SCOPE_ALL, "sched_block_max_us", 2500,
                                            "Maximum blocked input time, in microseconds",
                                            "The maximum blocked time, after scaling with "
                                            "-sched_block_scale.");
@@ -995,6 +1005,13 @@ droption_t<bool> op_sched_disable_direct_switches(
     "switch being determined by latency and the next input in the queue.  The "
     "TRACE_MARKER_TYPE_DIRECT_THREAD_SWITCH markers are not removed from the trace.");
 
+droption_t<double> op_sched_time_units_per_us(
+    DROPTION_SCOPE_ALL, "sched_time_units_per_us", 100.,
+    "Time units per simulated microsecond",
+    "Time units (currently wall-clock time) per simulated microsecond.  This scales all "
+    "of the -sched_*_us values as it converts wall-clock time into the simulated "
+    "microseconds measured by those options.");
+
 // Schedule_stats options.
 droption_t<uint64_t>
     op_schedule_stats_print_every(DROPTION_SCOPE_ALL, "schedule_stats_print_every",

clients/drcachesim/common/options.h

@@ -200,6 +200,7 @@ extern dynamorio::droption::droption_t<int> op_kernel_trace_buffer_size_shift;
 #endif
 extern dynamorio::droption::droption_t<bool> op_core_sharded;
 extern dynamorio::droption::droption_t<bool> op_core_serial;
+extern dynamorio::droption::droption_t<double> op_sched_time_per_us;
 extern dynamorio::droption::droption_t<int64_t> op_sched_quantum;
 extern dynamorio::droption::droption_t<bool> op_sched_time;
 extern dynamorio::droption::droption_t<bool> op_sched_order_time;