ADR-0001 Rev 2: 51-bit PhysAddr layout with concrete sub-unit tables

Remove rack_id (4 bits), rename sip_seg→die_id, shift fields to enable
42-bit local_offset (4 TB per die). Define PE_LOCAL/MCPU_LOCAL/CUBE_SRAM
sub-unit tables for AHBM dies and IOCPU sub-unit table for IOCHIPLET
dies (1 TB window). Supersedes ADR-0031.

Also fixes latent VA/PA confusion in pe_dma pipeline DMA path where
virtual addresses were decoded as physical addresses without MMU
translation — previously masked by coincidental bit-position alignment.

529 passed (+6 recovered), 10 pre-existing failures unchanged.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

src/kernbench/components/builtin/io_cpu.py

@@ -220,10 +220,10 @@ class IoCpuComponent(ComponentBase):
         return []
 
     def _cube_from_pa(self, pa_val: int, fallback: int) -> int:
-        """Extract cube_id from a physical address, with fallback."""
+        """Extract die_id from a physical address, with fallback."""
         from kernbench.policy.address.phyaddr import PhysAddr
         try:
-            return PhysAddr.decode(pa_val).cube_id
+            return PhysAddr.decode(pa_val).die_id
         except Exception:
             return fallback
 

src/kernbench/components/builtin/pe_dma.py

@@ -302,7 +302,16 @@ class PeDmaComponent(PeEngineBase):
             dma_res = self._dma_write if is_write else self._dma_read
             assert dma_res is not None
 
-            pa = PhysAddr.decode(addr)
+            # Translate VA → PA via MMU (same logic as non-pipeline path)
+            target_pa = addr
+            if self._mmu is not None:
+                from kernbench.policy.address.pe_mmu import PageFault
+                try:
+                    target_pa = self._mmu.translate(addr)
+                except PageFault:
+                    target_pa = addr  # fallback: treat as PA directly
+
+            pa = PhysAddr.decode(target_pa)
             dst_node = self.ctx.resolver.resolve(pa)
             path = self.ctx.router.find_path(self._pe_prefix, dst_node)
             drain_ns = self.ctx.compute_drain_ns(path, nbytes)
@@ -314,7 +323,7 @@ class PeDmaComponent(PeEngineBase):
                     correlation_id="pipeline",
                     request_id=f"tile_{token.tile_id}",
                     src_sip=0, src_cube=0, src_pe=0,
-                    dst_pa=addr, nbytes=nbytes,
+                    dst_pa=target_pa, nbytes=nbytes,
                     is_write=is_write,
                 )
                 sub_txn = Transaction(

src/kernbench/components/legacy/builtin/io_cpu.py

@@ -207,10 +207,10 @@ class IoCpuComponent(ComponentBase):
         return []
 
     def _cube_from_pa(self, pa_val: int, fallback: int) -> int:
-        """Extract cube_id from a physical address, with fallback."""
+        """Extract die_id from a physical address, with fallback."""
         from kernbench.policy.address.phyaddr import PhysAddr
         try:
-            return PhysAddr.decode(pa_val).cube_id
+            return PhysAddr.decode(pa_val).die_id
         except Exception:
             return fallback