From a769f53472a159223f5587d8f81a50e68dfe03f5 Mon Sep 17 00:00:00 2001
From: Christoph Urlacher <christoph.urlacher@protonmail.com>
Date: Mon, 31 Aug 2026 21:04:48 +0200
Subject: [PATCH] Add resultsdata query without "de-pruning"

---
 scripts/Queries/ResultsDataPruned.pm | 58 ++++++++++++++++++++++++++++
 scripts/menu.pl                      |  6 +++
 2 files changed, 64 insertions(+)
 create mode 100644 scripts/Queries/ResultsDataPruned.pm

diff --git a/scripts/Queries/ResultsDataPruned.pm b/scripts/Queries/ResultsDataPruned.pm
new file mode 100644
index 0000000..28a20c0
--- /dev/null
+++ b/scripts/Queries/ResultsDataPruned.pm
@@ -0,0 +1,58 @@
+package ResultsDataPruned;
+
+use strict;
+use warnings;
+use diagnostics;
+
+use FindBin;
+use lib "$FindBin::Bin/../Modules";
+use Filters;
+
+use feature 'say';
+
+sub query {
+    my ( $experiment, $experiment_dir, @filter_config_names ) = @_;
+
+    my $extra =
+      Filters::build_filter_clause( $experiment_dir, @filter_config_names );
+
+    # This query basically asks: How many equivalence classes
+    # ended up in each possible outcome?
+    # In contrast to ResultsData.pm it doesn't do any "de-pruning", e.g.
+    # no expansion of the "data liveness interval length"
+    # SUM(t.time2 - t.time1 + 1).
+    #
+    # We run with --inject-single-bit, so the bits of one class can
+    # end up in different outcomes.
+    # This would be counted multiple times,
+    # so we'll have more results than pilots.
+    #
+    # The column is named "faults" so the charts still work with this.
+    my $querystring = "SELECT
+benchmark, resulttype, COUNT(DISTINCT g.pilot_id) AS faults
+FROM variant v
+JOIN trace t ON v.id = t.variant_id
+JOIN fspgroup g ON g.variant_id = t.variant_id AND g.instr2 = t.instr2 AND g.data_physical_address = t.data_physical_address
+JOIN result_GenericExperimentMessage r ON r.pilot_id = g.pilot_id
+JOIN fsppilot p ON r.pilot_id = p.id
+WHERE v.variant = '$experiment'$extra
+GROUP BY v.id, resulttype
+ORDER BY variant, benchmark, resulttype;";
+
+    say $querystring;
+
+    return $querystring;
+}
+
+sub args { return "--batch --raw"; }
+
+sub filename {
+    my @filter_config_names = grep { defined && length } @_;
+    my $suffix =
+      @filter_config_names ? "_" . join( "+", sort @filter_config_names ) : "";
+    return "resultsdata_pruned${suffix}.csv";
+}
+
+sub postprocess { $_[0] =~ s/\t/,/g; }
+
+1;
diff --git a/scripts/menu.pl b/scripts/menu.pl
index ced09b6..424fe48 100644
--- a/scripts/menu.pl
+++ b/scripts/menu.pl
@@ -95,6 +95,9 @@ my %handlers = (
             ResultsData =>
               'Faults summary per benchmark with resulttype (resultsdata.csv)',
 
+            ResultsDataPruned =>
+'Faults summary per benchmark with resulttype, using the pruned data without expansion (resultsdata_pruned.csv)',
+
             TargetClass =>
 'Faults by code type with data type (stack/heap/bss/...) and resulttype -> targetclass.csv',
 
@@ -662,6 +665,9 @@ my %handlers = (
             combined_fault_probability =>
 'marker probability per fault space (resultsdata.csv [+ traceweight.csv]).',
 
+            combined_fault_probability_merged =>
+'marker probability per fault space, ip/mem/regs merged (resultsdata.csv).',
+
             combined_fault_rates_per_instruction =>
 'faults normalised by instruction count (faults.csv + mnemonics.csv).',