AXI IDs and Response Order

Where the interviewer goes next. Open a node to practise that follow-up.

WhyWhat IfProve ItCompareScale It

1. AXI IDs and Response Order
2. WhyWhy Same-ID Responses Stay in OrderL1 Junior

   Why does AXI force same-ID responses into order at all? Why not let everything reorder?

   Reveal the Model Answer

   A response carries only the ID. With two transactions open on one ID, the manager can pair each response with its request only by order. Ordering per ID also lets a simple manager use one ID and never build a reorder buffer. A capable manager uses several IDs as several in-order streams.

   Self-Check

   Tick each key point you covered in your answer.

   Key point 1 A response carries only the ID. Key point 2 With two transactions open on one ID, the manager can pair each response with its request only by order. Key point 3 Ordering per ID also lets a simple manager use one ID and never build a reorder buffer. Key point 4 A capable manager uses several IDs as several in-order streams.
3. What IfInterleaved Read DataL2 Mid-Level

   Two 4-beat reads on IDs 1 and 2 come back as 1, 2, 1, 2, 2, 1, 2, 1. Legal? What must the monitor do?

   Reveal the Model Answer

   Legal: different IDs may interleave beat by beat. Each transaction's beats must still stay in order and end with its own RLAST. The monitor must assemble beats per RID and file each beat under the oldest open read of that ID. It closes the transaction on RLAST. A monitor that assumes contiguous bursts glues beats of different reads together.

   Self-Check

   Tick each key point you covered in your answer.

   Key point 1 Legal: different IDs may interleave beat by beat. Key point 2 Each transaction's beats must still stay in order and end with its own RLAST. Key point 3 The monitor must assemble beats per RID and file each beat under the oldest open read of that ID. Key point 4 It closes the transaction on RLAST. Key point 5 A monitor that assumes contiguous bursts glues beats of different reads together.

   • Lesson P2.11Different IDs: Out of Order and Interleaving
   • Lesson P2.17An AXI Agent, Monitor and Scoreboard
   • Problem P2.17.1The AXI Monitor
   • Problem P2.18.6A Beat-to-Transaction Monitor
4. What IfA Manager That Cannot Take InterleavingL2 Mid-LevelL3 Senior

   A manager team says their block cannot handle interleaved read data. Is that allowed, and what does the spec offer?

   Reveal the Model Answer

   Not if the manager issues different IDs: such a manager must accept interleaved data. The Read_Interleaving_Disabled property only lets it switch interleaving off as an optimisation. That works only when the attached subordinate supports it. So the test plan must cover interleaved returns for every manager that uses several IDs.

   Self-Check

   Tick each key point you covered in your answer.

   Key point 1 Not if the manager issues different IDs: such a manager must accept interleaved data. Key point 2 The Read_Interleaving_Disabled property only lets it switch interleaving off as an optimisation. Key point 3 That works only when the attached subordinate supports it. Key point 4 So the test plan must cover interleaved returns for every manager that uses several IDs.

   Spec Misquote

   "Interleaving is optional, so our manager need not handle it."

   Every manager that issues different IDs must accept it. Disabling it is an optimisation that needs the subordinate to support it.

   Source AMBA AXI Protocol Specification IHI 0022L §A5.6.1 Read data interleaving

   • Lesson P2.11Different IDs: Out of Order and Interleaving
   • Lesson P2.17An AXI Agent, Monitor and Scoreboard
   • Problem P2.17.1The AXI Monitor
5. Scale ItChoosing IDs for ThroughputL2 Mid-LevelL3 Senior

   A manager uses one ID for everything. The latency of its short reads is terrible whenever a long read is in flight. Why, and what would you change?

   Reveal the Model Answer

   Same-ID responses must return in order, so a short read queues behind a slow one (head-of-line blocking). Give independent streams their own IDs, so the subordinate or interconnect can answer the fast one first. Keep one ID only for accesses that must stay ordered. The cost is reorder storage in the manager and more tracking downstream. Lesson P2.10 measured a fast read waiting seven edges on a shared ID, and three on its own.

   Both answers carry ID 5. Only their order tells the manager which is which.

   1. Cycle 1

      The first read with ID 5 goes to the slow path, which needs six cycles. The manager notes: the first answer on ID 5 is this one.

   2. Cycle 2

      The second read with ID 5 goes to the fast path.

   3. Cycle 4

      The fast path has its data, 50, two cycles later. It may not send it: an older read with ID 5 is still open.

   4. Cycle 8

      The slow answer, d0, leaves first and lands on the first request. The manager has no other way to pair them.

   5. Cycle 9

      Now 50 leaves and lands on the second request. Five cycles of waiting bought the only guarantee the ID can give.

   The read data channel carries the ID and nothing else that names a request. Same ID, same order: that is the whole rule.

   Self-Check

   Tick each key point you covered in your answer.

   Key point 1 Same-ID responses must return in order, so a short read queues behind a slow one (head-of-line blocking). Key point 2 Give independent streams their own IDs, so the subordinate or interconnect can answer the fast one first. Key point 3 Keep one ID only for accesses that must stay ordered. Key point 4 The cost is reorder storage in the manager and more tracking downstream. Key point 5 Lesson P2.10 measured a fast read waiting seven edges on a shared ID, and three on its own.

   What Interviewers Listen For (Mid-Level)

   A good mid-level answer names head-of-line blocking, where a fast response waits behind a slow one, and gives the fix.

6. CompareA Read Issued Before the Write ResponseL1 JuniorL2 Mid-Level

   Asked if they assume reads and writes share the order

   A manager writes a word, and right after its last W beat it reads the same address with the same ID. Is it guaranteed to see the new value?

   Reveal the Model Answer

   No. AXI never orders a read against a write, whatever the IDs, so the read may return the old value. The manager's tool is to wait for the write's B response before issuing the read. Lesson P2.12 measured exactly this, with a posted write and a six-cycle write latency. The read issued after the last data beat returned the old word.

   Self-Check

   Tick each key point you covered in your answer.

   Key point 1 No. Key point 2 AXI never orders a read against a write, whatever the IDs, so the read may return the old value. Key point 3 The manager's tool is to wait for the write's B response before issuing the read. Key point 4 Lesson P2.12 measured exactly this, with a posted write and a six-cycle write latency. Key point 5 The read issued after the last data beat returned the old word.

   Misconception

   "Same ID means same ordering stream, so the read waits for the write."

   Same-ID ordering applies within reads and within writes. Read and write channels have no ordering guarantee between them.

   Source AMBA AXI Protocol Specification IHI 0022L §A5.3 Request ordering

   • Lesson P2.12The Ordering Model: Reads Versus Writes
   • Problem P2.12.1The Stale Read
7. Scale ItOrdering a Flag After a BufferL2 Mid-LevelL3 Senior

   A DMA writes a 4 KB buffer, then writes a "done" flag that the CPU polls. How do you make sure the CPU never sees the flag before the data?

   Reveal the Model Answer

   Before writing the flag, wait for the B responses of every buffer write. Only a completion response makes an access observable. Same-ID writes stay in order only to the same location, so one ID does not order the buffer against the flag. A Bufferable B can come from a buffer partway along the path. It then guarantees observability only inside the write's Domain, so the CPU must sit in that Domain.

   Self-Check

   Tick each key point you covered in your answer.

   Key point 1 Before writing the flag, wait for the B responses of every buffer write. Key point 2 Only a completion response makes an access observable. Key point 3 Same-ID writes stay in order only to the same location, so one ID does not order the buffer against the flag. Key point 4 A Bufferable B can come from a buffer partway along the path. Key point 5 It then guarantees observability only inside the write's Domain, so the CPU must sit in that Domain.

   What Interviewers Listen For (Mid-Level)

   A good mid-level answer says the manager must wait for every B response before it writes the flag.

8. Prove ItSolve P2.12.1: The Stale ReadSimulator-graded

   Show it, don't just say it: this Silicon Lab problem is graded by the simulator.

   Open Problem P2.12.1

9. Prove ItSolve P2.10.1: The Same-ID Order CheckerSimulator-graded

   Show it, don't just say it: this Silicon Lab problem is graded by the simulator.

   Open Problem P2.10.1