Add tl.recv_no_consume diagnostic API for apples-to-apples pe2pe plot

The pe2pe overview compared IPCQ (tl.send + tl.recv) against raw DMA
(tl.load + tl.store), but DMA is one-sided — DST never reads — while
tl.recv pays a slot-read on DST. The comparison was unfair: IPCQ
looked slower partly because it does more work.

Adds tl.recv_no_consume() — a separate, diagnostic-only entry point
that blocks for slot arrival but skips the slot-read (and bank-hop)
charge on DST. Production tl.recv is unchanged (no `consume` kwarg
on the public API), so the diagnostic flag can never accidentally
leak into real workloads.

Updates test_pe_to_pe_latency to call tl.recv_no_consume so the
overview.png shows IPCQ no-consume vs raw DMA on equal footing.
Also fixes PLOT_DIR back to docs/diagrams/pe2pe_latency_plots/
(was lost in a merge). Adds scripts/replot_pe2pe.py for label-only
re-renders without re-measuring.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

scripts/replot_pe2pe.py

@@ -0,0 +1,141 @@
+"""Re-render pe2pe latency PNGs from the existing summary.csv with the
+current (no-consume) labels. Used after a label-only test edit to avoid
+re-measuring (~5 min) when the data on disk is already correct.
+
+Reads docs/diagrams/pe2pe_latency_plots/summary.csv. Plots 2 curves:
+"IPCQ no-consume" (from the ipcq_no_consume rows if present, else from
+the ipcq rows) and "Raw DMA" (raw rows).
+"""
+from __future__ import annotations
+
+import csv
+from pathlib import Path
+
+import matplotlib.pyplot as plt
+
+ROOT = Path(__file__).resolve().parent.parent
+PLOT_DIR = ROOT / "docs" / "diagrams" / "pe2pe_latency_plots"
+CSV_PATH = PLOT_DIR / "summary.csv"
+
+
+def _load_records():
+    rows = []
+    with open(CSV_PATH, newline="") as f:
+        for r in csv.DictReader(f):
+            rows.append({
+                "hop": r["hop"],
+                "label": r["label"],
+                "size_bytes": int(r["size_bytes"]),
+                "path": r["path"],
+                "total_ns": float(r["total_ns"]),
+            })
+    return rows
+
+
+def _ipcq_rows(records, hop):
+    # Prefer ipcq_no_consume if present (older 3-path CSV); fall back to ipcq
+    # (current single-path CSV where ipcq IS no-consume).
+    nc = [r for r in records
+          if r["hop"] == hop and r["path"] == "ipcq_no_consume"]
+    if nc:
+        return sorted(nc, key=lambda r: r["size_bytes"])
+    return sorted(
+        [r for r in records if r["hop"] == hop and r["path"] == "ipcq"],
+        key=lambda r: r["size_bytes"],
+    )
+
+
+def _raw_rows(records, hop):
+    return sorted(
+        [r for r in records if r["hop"] == hop and r["path"] == "raw"],
+        key=lambda r: r["size_bytes"],
+    )
+
+
+def _hops(records):
+    seen = []
+    for r in records:
+        if r["hop"] not in {h["id"] for h in seen}:
+            seen.append({"id": r["hop"], "label": r["label"]})
+    return seen
+
+
+def _plot_per_hop(records, hop, path):
+    ipcq = _ipcq_rows(records, hop["id"])
+    raw = _raw_rows(records, hop["id"])
+    fig, ax = plt.subplots(figsize=(8, 5))
+    if ipcq:
+        ax.plot(
+            [r["size_bytes"] for r in ipcq],
+            [r["total_ns"] for r in ipcq],
+            marker="o", color="tab:blue",
+            label="IPCQ no-consume (send/recv, no slot read)",
+        )
+    if raw:
+        ax.plot(
+            [r["size_bytes"] for r in raw],
+            [r["total_ns"] for r in raw],
+            marker="s", color="tab:orange",
+            label="Raw DMA (load+store)",
+        )
+    ax.set_xlabel("Data size (bytes)")
+    ax.set_ylabel("Latency (ns)")
+    ax.set_title(hop["label"])
+    ax.grid(True, alpha=0.3)
+    ax.legend()
+    fig.tight_layout()
+    fig.savefig(path, dpi=120)
+    plt.close(fig)
+
+
+def _plot_overview(records, hops, path):
+    fig, axes = plt.subplots(2, 2, figsize=(13, 9))
+    axes = axes.flatten()
+    for i, hop in enumerate(hops):
+        ax = axes[i]
+        ipcq = _ipcq_rows(records, hop["id"])
+        raw = _raw_rows(records, hop["id"])
+        if ipcq:
+            ax.plot(
+                [r["size_bytes"] for r in ipcq],
+                [r["total_ns"] for r in ipcq],
+                marker="o", color="tab:blue",
+                label="IPCQ no-consume",
+            )
+        if raw:
+            ax.plot(
+                [r["size_bytes"] for r in raw],
+                [r["total_ns"] for r in raw],
+                marker="s", color="tab:orange",
+                label="Raw DMA",
+            )
+        ax.set_title(hop["label"], fontsize=10)
+        ax.set_xlabel("bytes")
+        ax.set_ylabel("ns")
+        ax.grid(True, alpha=0.3)
+        ax.legend(fontsize=8)
+    for j in range(len(hops), len(axes)):
+        axes[j].axis("off")
+    fig.suptitle(
+        "PE-to-PE latency: IPCQ no-consume vs raw DMA",
+        fontsize=14,
+    )
+    fig.tight_layout()
+    fig.savefig(path, dpi=120)
+    plt.close(fig)
+
+
+def main():
+    records = _load_records()
+    hops = _hops(records)
+    for hop in hops:
+        out = PLOT_DIR / f"{hop['id']}.png"
+        _plot_per_hop(records, hop, out)
+        print(f"wrote {out}")
+    overview = PLOT_DIR / "overview.png"
+    _plot_overview(records, hops, overview)
+    print(f"wrote {overview}")
+
+
+if __name__ == "__main__":
+    main()