perf(geekbench): add config defaults for runs, pinning, and output paths

- Provide reproducible defaults for iterations and output directory
- Allow optional CPU pinning and wrapper env overrides
- Document knobs used by LAVA and local runs

Signed-off-by: Srikanth Muppandam <smuppand@qti.qualcomm.com>

Author: smuppand

Runner/suites/Performance/Geekbench/geekbench_baseline.conf

@@ -0,0 +1,71 @@
+# geekbench_baseline.conf
+#
+# Geekbench 6 baseline / gating config for Geekbench
+# - Designed to be simple + human-editable (same spirit as your tiotest baseline)
+# - Use with your runner's --baseline / --delta logic (PASS/FAIL vs thresholds)
+#
+# Units:
+# - Summary scores are plain integers (Geekbench score)
+# - Workload throughput targets use the unit Geekbench prints (e.g., MB/sec)
+#
+# Tips:
+# - Start with summary-only gating (recommended)
+# - Add per-workload goals only when your platform is stable enough
+# - Keep goals realistic; use DELTA (allowed deviation) from YAML to avoid flakiness
+#
+# ------------------------------------------------------------------------------
+# Summary score gating (recommended)
+# ------------------------------------------------------------------------------
+
+# Minimum acceptable Geekbench summary totals (avg across iterations)
+# If these are present, runner should enforce:
+# measured >= goal * (1 - delta)
+goal.geekbench.summary.single.total_score = 1165
+goal.geekbench.summary.multi.total_score = 2881
+
+# Optional: sub-scores (enable if you want tighter gating)
+#goal.geekbench.summary.single.integer_score = 1195
+#goal.geekbench.summary.single.floating_point_score = 1111
+#goal.geekbench.summary.multi.integer_score = 2806
+#goal.geekbench.summary.multi.floating_point_score = 3025
+
+# ------------------------------------------------------------------------------
+# Optional per-workload gating (advanced)
+# ------------------------------------------------------------------------------
+# NOTE:
+# - Workloads are sectioned by single/multi
+# - You can gate on "score" (integer) or "throughput" (float) separately
+#
+# Naming convention (matches long-metric key style):
+# goal.geekbench.<section>.workload.<WorkloadName>.<kind>
+# where:
+# section = single | multi
+# kind = score | throughput
+#
+# WorkloadName must match Geekbench output text exactly (case/spaces).
+# If your implementation sanitizes names, mirror that same sanitization here.
+
+# ---- Single-Core (examples from your sample output) ----
+#goal.geekbench.single.workload.File Compression.score = 1091
+#goal.geekbench.single.workload.File Compression.throughput = 156.6
+#goal.geekbench.single.workload.Navigation.score = 1152
+#goal.geekbench.single.workload.Navigation.throughput = 6.94
+#goal.geekbench.single.workload.HTML5 Browser.score = 1264
+#goal.geekbench.single.workload.HTML5 Browser.throughput = 25.9
+
+# ---- Multi-Core (examples from your sample output) ----
+#goal.geekbench.multi.workload.File Compression.score = 1317
+#goal.geekbench.multi.workload.File Compression.throughput = 189.2
+#goal.geekbench.multi.workload.PDF Renderer.score = 4918
+#goal.geekbench.multi.workload.PDF Renderer.throughput = 113.4
+#goal.geekbench.multi.workload.Clang.score = 4249
+#goal.geekbench.multi.workload.Clang.throughput = 20.9
+
+# ------------------------------------------------------------------------------
+# Optional metadata (ignored by gating, useful for humans/tools)
+# ------------------------------------------------------------------------------
+
+#platform = RB3gen2
+#os = "QCOM Reference Distro with Wayland 2.0"
+#kernel = "Linux 6.19.0-rc4-00414-g65bb05266d47 aarch64"
+#notes = "Baseline captured with --no-upload, default CPU benchmark"