Stadium relaunch: full 127-page deep-dive report -- per-cell, per-ISA,

every factor interaction, and a raw-data appendix

Expanded the campaign-mechanism validation report from a condensed
6-page summary into the full depth Captain Bob asked for: analyze all 9
cells as a conglomerate Latin square, then dive into each cell's own
data, then cover every within-ISA and cross-factor interaction
explicitly rather than averaging it away.

Report structure (127 pages, compiled clean, no undefined references):
- Front matter: context, methodology, the SWAP-MTX bug narrative
  (console-interleaving fix + the Fisher-Yates correctness bug and its
  fix, both already committed separately)
- Layer 1: aggregate 3x3 Latin square (heatmap, invariant-metrics table)
- Per-Cell Deep Dive (9 sections): cfg-level distribution, summary
  table, and a rep-order execution-trajectory chart per cell -- the
  trajectory charts are what actually visualize the order-dependence
  finding rather than just stating it
- Per-ISA Deep Dive (3 sections): within-architecture seed comparison
  (violin plots, Kruskal-Wallis, per-factor main effects)
- Factor Interactions (6 sections, every pairwise combination of the 4
  L8 binary factors): both infer_dec_q and early_exit interaction plots
  faceted by architecture, plus the three-way
  factor x factor x architecture significance test
- Per-Factor Response (4 sections): linear response by architecture,
  with an explicit note that a true quadratic term isn't identifiable
  from this 2-level factorial design
- Appendix: full run_id-ordered raw data, all 4,320 rows across all 9
  cells, as the primary-source backing for every statistic above

Generated programmatically (analyse_stadium_relaunch_fixed.R for the
aggregate layer, generate_stadium_deepdive.R for the per-cell/per-ISA/
interaction/appendix layers) rather than hand-authored, since content at
this scale needs to be data-driven to stay honest.

Also includes analyse_stadium_relaunch.R, the earlier script built
against the pre-fix (buggy-shuffle) dataset -- superseded but kept for
the record, matching how the underlying data commits were handled.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

experiments/bare_metal/analysis/tables/stadium_fixed_anova_full.txt

@@ -0,0 +1,14 @@
+Analysis of Variance Table
+
+Response: infer_dec_q
+             Df  Sum Sq Mean Sq F value Pr(>F)
+ent_f         1    1829 1829.10  1.3689 0.2421
+cv_f          1       5    5.42  0.0041 0.9492
+tmp_f         1     260  259.60  0.1943 0.6594
+stb_f         1     605  605.25  0.4530 0.5010
+arch          2       0    0.00  0.0000 1.0000
+ent_f:arch    2       0    0.00  0.0000 1.0000
+cv_f:arch     2       0    0.00  0.0000 1.0000
+tmp_f:arch    2       0    0.00  0.0000 1.0000
+stb_f:arch    2       0    0.00  0.0000 1.0000
+Residuals  4305 5752465 1336.23               

experiments/bare_metal/analysis/tables/stadium_fixed_cell_summary.csv

@@ -0,0 +1,10 @@
+"arch","seed","n_rows","n_cfg","n_run_id","mean_infer_dec_q","sd_infer_dec_q","early_exit_rate","l8_mode_const","win_div_const","fit_q_const"
+"amd64","12345",480,16,480,56.8541666666667,29.2966182458014,0.9875,TRUE,TRUE,TRUE
+"amd64","67890",480,16,480,28.0854166666667,38.3427977824838,0.989583333333333,TRUE,TRUE,TRUE
+"amd64","13579",480,16,480,44.5,35.4835961394778,0.991666666666667,TRUE,TRUE,TRUE
+"aarch64","12345",480,16,480,56.8541666666667,29.2966182458014,0.9875,TRUE,TRUE,TRUE
+"aarch64","67890",480,16,480,28.0854166666667,38.3427977824838,0.989583333333333,TRUE,TRUE,TRUE
+"aarch64","13579",480,16,480,44.5,35.4835961394778,0.991666666666667,TRUE,TRUE,TRUE
+"riscv64","12345",480,16,480,56.8541666666667,29.2966182458014,0.9875,TRUE,TRUE,TRUE
+"riscv64","67890",480,16,480,28.0854166666667,38.3427977824838,0.989583333333333,TRUE,TRUE,TRUE
+"riscv64","13579",480,16,480,44.5,35.4835961394778,0.991666666666667,TRUE,TRUE,TRUE

experiments/bare_metal/analysis/tables/stadium_fixed_kw_within_isa.csv

@@ -0,0 +1,4 @@
+"arch","H","df","p"
+"amd64",50.7569110985713,2,9.51210850642286e-12
+"aarch64",50.7569110985713,2,9.51210850642286e-12
+"riscv64",50.7569110985713,2,9.51210850642286e-12

experiments/bare_metal/analysis/tables/stadium_fixed_logit_early_exit.txt

@@ -0,0 +1,33 @@
+
+Call:
+glm(formula = early_exit ~ (ent_f + cv_f + tmp_f + stb_f) * arch, 
+    family = binomial(), data = all_data)
+
+Coefficients:
+                      Estimate Std. Error z value Pr(>|z|)    
+(Intercept)          4.118e+00  5.411e-01   7.612 2.71e-14 ***
+ent_flo              4.105e-01  5.302e-01   0.774    0.439    
+cv_flo               7.011e-01  5.505e-01   1.273    0.203    
+tmp_flo              4.105e-01  5.302e-01   0.774    0.439    
+stb_flo             -4.105e-01  5.302e-01  -0.774    0.439    
+archaarch64         -7.403e-15  7.652e-01   0.000    1.000    
+archriscv64         -7.626e-15  7.652e-01   0.000    1.000    
+ent_flo:archaarch64  1.672e-14  7.498e-01   0.000    1.000    
+ent_flo:archriscv64  1.472e-14  7.498e-01   0.000    1.000    
+cv_flo:archaarch64   6.210e-15  7.786e-01   0.000    1.000    
+cv_flo:archriscv64   7.805e-15  7.786e-01   0.000    1.000    
+tmp_flo:archaarch64  3.743e-15  7.498e-01   0.000    1.000    
+tmp_flo:archriscv64  5.025e-15  7.498e-01   0.000    1.000    
+stb_flo:archaarch64 -1.461e-15  7.498e-01   0.000    1.000    
+stb_flo:archriscv64  1.453e-16  7.498e-01   0.000    1.000    
+---
+Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
+
+(Dispersion parameter for binomial family taken to be 1)
+
+    Null deviance: 500.32  on 4319  degrees of freedom
+Residual deviance: 489.67  on 4305  degrees of freedom
+AIC: 519.67
+
+Number of Fisher Scoring iterations: 7
+