
MOTOROLA
Chapter 6. Memory Management
6-29
Page Table Search Operation
3. The PTE in the selected PTEG is tested for a match with the virtual page number
(VPN) of the access. The VPN is the VSID concatenated with the page index field
of the virtual address. For a match to occur, the following must be true:
— PTE[H] = 0
— PTE[V] = 1
— PTE[VSID] = VA[0–23]
— PTE[API] = VA[24–29]
4. If a match is not found, step 3 is repeated for each of the other seven PTEs in the
primary PTEG. If a match is found, the table search process continues as described
in step 8. If a match is not found within the eight PTEs of the primary PTEG, the
address of the secondary PTEG is generated.
5. The first PTE (PTE0) in the secondary PTEG is read from memory. Again, because
PTE reads typically have a WIM bit combination of 0b001, an entire cache line is
burst into the on-chip cache.
6. The PTE in the selected secondary PTEG is tested for a match with the virtual page
number (VPN) of the access. For a match to occur, the following must be true:
— PTE[H] = 1
— PTE[V] = 1
— PTE[VSID] = VA[0–23]
— PTE[API] = VA[24–29]
7. If a match is not found, step 6 is repeated for each of the other seven PTEs in the
secondary PTEG.
8. If a match is found, the PTE is written into the on-chip TLB and the R bit is updated
in the PTE in memory (if necessary). If there is no memory protection violation, the
C bit is also updated in memory and the table search is complete.
9. If no match is found in the eight PTEs of the secondary PTEG, the search fails and a
page fault exception condition occurs (either an ISI exception or a DSI exception).
Note that the software routines that implement this algorithm must synthesize this
condition by appropriately setting the SRR1 or DSISR and branching to the ISI or
DSI handler routine.
Reads from memory for table search operations should be performed as global (but not
exclusive), cacheable operations, and can be loaded into the on-chip cache.
Figure 6-9 and Figure 6-10 provide conceptual flow diagrams of primary and secondary
page table search operations as described in the OEA for 32-bit processors. Recall that the
architecture allows implementations to perform the page table search operations
automatically (in hardware) or with software assistance (may be required), as is the case
with the G2 core. Also, the elements in the figure that apply to TLBs are shown as optional
because TLBs are not required by the architecture.
F
Freescale Semiconductor, Inc.
For More Information On This Product,
Go to: www.freescale.com
n
.