1. EECC722 - Shaaban #1 lec # 10 Fall 2003 10-22-2003 A New Approach to Speculation in the Stanford Hydra Chip Multiprocessor (CMP) A Chip Multiprocessor microarchitecture/compiler effort at Stanford that provides hardware/software support for Data/Thread Level Speculation (TLS) to extract parallel speculated threads from sequential code augmented with software thread speculation handlers

2. EECC722 - Shaaban #2 lec # 10 Fall 2003 10-22-2003 Chip Multiprocessors (CMPs) A CMP offers implementation benefits –High-speed signals are localized in individual CPUs –A proven CPU design may be replicated across the die Overcomes diminishing performance/transistor return problem in uniprocessors –Transistors are used today mostly for ILP extraction –MPs use transistors to run multiple threads... On parallelized programs With multiprogrammed workloads Fast inter-processor communication eases parallelization of code (Shared L2 cache)

3. EECC722 - Shaaban #3 lec # 10 Fall 2003 10-22-2003 Stanford Hydra CMP Approach Exploit all levels of program parallelism. Develop a single-chip multiprocessor architecture that simplifies microprocessor design and achieves high performance. Make the multiprocessor transparent to the average user. Integrate use of parallelizing compiler technology in the design of microarchitecture that supports data/thread speculation.

4. EECC722 - Shaaban #4 lec # 10 Fall 2003 10-22-2003 The Basic Hydra CMP 4 processors and secondary cache on a chip 2 buses connect processors and memory Coherence: writes are broadcast on write bus

5. EECC722 - Shaaban #5 lec # 10 Fall 2003 10-22-2003 Hydra Memory Hierarchy Characteristics

6. EECC722 - Shaaban #6 lec # 10 Fall 2003 10-22-2003 CMP Parallel Performance Varying levels of performance –Multiprogrammed workloads work well –Very parallel apps (matrix-based FP and multimedia) are excellent –Acceptable only with a few less parallel (i.e. integer) applications

7. EECC722 - Shaaban #7 lec # 10 Fall 2003 10-22-2003 Problem: Limited Parallel Software Current parallel software is limited –Some programs just don’t have significant parallelism –Parallel compilers generally require dense matrix applications Many applications only hand-parallelizable –Parallelism may exist in algorithm, but code hides it –Compilers must statically verify parallelism –Data dependencies require synchronization –Pointer disambiguation is a major problem for this! Can hardware help the situation?

8. EECC722 - Shaaban #8 lec # 10 Fall 2003 10-22-2003 Possible Limited Parallel Software Solution: Data Speculation & Thread Level Speculation (TLS) Data speculation and Thread Level Speculation (TLS) enable parallelization without regard for data dependencies –Normal sequential program is broken up into speculative threads –Speculative threads are now run in parallel on CPUs –Speculation hardware ensures correctness Parallel software implications –Loop parallelization is now easily automated –More “arbitrary” threads are possible (subroutines) –Add synchronization only for performance Speculation support mechanisms –Speculative thread control mechanism –Five basic speculation hardware/memory system requirements for correct data/thread speculation

9. EECC722 - Shaaban #9 lec # 10 Fall 2003 10-22-2003 Subroutine Speculation

10. EECC722 - Shaaban #10 lec # 10 Fall 2003 10-22-2003 Loop Iteration Speculative Threads A Simple Example of A speculatively Executed Loop using Data/Thread Speculation

11. EECC722 - Shaaban #11 lec # 10 Fall 2003 10-22-2003 Overview of Loop-Iteration Thread Speculation Parallel regions (loop iterations) are annotated by the compiler. The hardware uses these annotations to run loop iterations in parallel as speculated threads on a number of CPUs. Each CPU knows which loop iteration it is running CPUs dynamically prevent data dependency violations –“later” iterations can’t use data before write by “earlier” iterations (RAW) –“earlier” iterations never see writes by “later” iterations (WAW) If a “later” iteration has used data that an “earlier” iteration writes (RAW hazard), it is restarted –All following iterations are halted and restarted, also –All writes by the later iteration are discarded (undo speculated work).

12. EECC722 - Shaaban #12 lec # 10 Fall 2003 10-22-2003 Hydra’s Data & Thread Speculation Operations

13. EECC722 - Shaaban #13 lec # 10 Fall 2003 10-22-2003 Hydra Loop Compiling for Speculation

14. EECC722 - Shaaban #14 lec # 10 Fall 2003 10-22-2003 Loop Execution with Thread Speculation

15. EECC722 - Shaaban #15 lec # 10 Fall 2003 10-22-2003 Speculative Thread Creation in Hydra Register Passing Buffer (RPB)

16. EECC722 - Shaaban #16 lec # 10 Fall 2003 10-22-2003 Speculative Data Access in Speculated Threads i Less Speculated thread i+1 More speculated thread WAR RAW WAW

17. EECC722 - Shaaban #17 lec # 10 Fall 2003 10-22-2003 To provide the desired memory behavior, the data speculation hardware must provide: 1. A method for detecting true memory dependencies, in order to determine when a dependency has been violated. 2. A method for backing up and re-executing speculative loads and any instructions that may be dependent upon them when the load causes a violation. 3. A method for buffering any data written during a speculative region of a program so that it may be discarded when a violation occurs or permanently committed at the right time. Speculative Data Access in Speculated Threads

18. EECC722 - Shaaban #18 lec # 10 Fall 2003 10-22-2003 Five Basic Speculation Hardware Requirements For Correct Data/Thread Speculation 1. Forward data between parallel threads (RAW). A speculative system must be able to forward shared data quickly and efficiently from an earlier thread running on one processor to a later thread running on another. 2. Detect when reads occur too early (RAW hazards). If a data value is read by a later thread and subsequently written by an earlier thread, the hardware must notice that the read retrieved incorrect data since a true dependence violation has occurred. 3. Safely discard speculative state after violations. All speculative changes to the machine state must be discarded after a violation, while no permanent machine state may be lost in the process. 4. Retire speculative writes in the correct order (WAW hazards). Once speculative threads have completed successfully, their state must be added to the permanent state of the machine in the correct program order, considering the original sequencing of the threads. 5. Provide memory renaming (WAR hazards). The speculative hardware must ensure that the older thread cannot “see” any changes made by later threads, as these would not have occurred yet in the original sequential program.

19. EECC722 - Shaaban #19 lec # 10 Fall 2003 10-22-2003 Speculative Hardware/Memory Requirements 1-2 (RAW) (RAW hazard or violation)

20. EECC722 - Shaaban #20 lec # 10 Fall 2003 10-22-2003 Speculative Hardware/Memory Requirements 3-4 (RAW hazard). (WAW hazards). Restart

21. EECC722 - Shaaban #21 lec # 10 Fall 2003 10-22-2003 Speculative Hardware/Memory Requirement 5 Memory Renaming to prevent WAR hazards.

22. EECC722 - Shaaban #22 lec # 10 Fall 2003 10-22-2003 Hydra Speculation Hardware

23. EECC722 - Shaaban #23 lec # 10 Fall 2003 10-22-2003 Hydra Speculation Support

24. EECC722 - Shaaban #24 lec # 10 Fall 2003 10-22-2003 L1 Cache Tag Details

25. EECC722 - Shaaban #25 lec # 10 Fall 2003 10-22-2003 L2 Buffer Details

26. EECC722 - Shaaban #26 lec # 10 Fall 2003 10-22-2003 The operation of Speculative Loads

27. EECC722 - Shaaban #27 lec # 10 Fall 2003 10-22-2003 Reading L2 Cache Speculative Buffers

28. EECC722 - Shaaban #28 lec # 10 Fall 2003 10-22-2003 The Operation of Speculative Stores

29. EECC722 - Shaaban #29 lec # 10 Fall 2003 10-22-2003 Hydra’s Handling of Five Basic Speculation Hardware Requirements For Correct Data/Thread Speculation 1. Forward data between parallel threads (RAW). –When a speculative thread writes data over the write bus, all more-speculative threads that may need the data have their current copy of that cache line invalidated. –This is similar to the way the system works during non- speculative operation (invalidate cache coherency protocol). –If any of the threads subsequently need the new speculative data forwarded to them, they will miss in their primary cache and access the secondary cache. The speculative data contained in the write buffers of the current or older threads replaces data returned from the secondary cache on a byte-by-byte basis just before the composite line is returned to the processor and primary cache.

30. EECC722 - Shaaban #30 lec # 10 Fall 2003 10-22-2003 2. Detect when reads occur too early (RAW hazards). –Primary cache bits are set to mark any reads that may cause violations. –Subsequently, if a write to that address from an earlier thread invalidates the address, a violation is detected, and the thread is restarted. 3. Safely discard speculative state after violations. –Since all permanent machine state in Hydra is always maintained within the secondary cache, anything in the primary caches and secondary cache buffers may be invalidated at any time without risking a loss of permanent state. –As a result, any lines in the primary cache containing speculative data (marked with a special modified bit) may simply be invalidated all at once to clear any speculative state from a primary cache. –In parallel with this operation, the secondary cache buffer for the thread may be emptied to discard any speculative data written by the thread. Hydra’s Handling of Five Basic Speculation Hardware Requirements For Correct Data/Thread Speculation

31. EECC722 - Shaaban #31 lec # 10 Fall 2003 10-22-2003 4. Retire speculative writes in the correct order (WAW hazards). –Separate secondary cache buffers are maintained for each thread. As long as these are drained into the secondary cache in the original program sequence of the threads, they will reorder speculative memory references correctly. 5. Provide memory renaming (WAR hazards). –Each processor can only read data written by itself or earlier threads when reading its own primary cache or the secondary cache buffers. –Writes from later threads don’t cause immediate invalidations in the primary cache, since these writes should not be visible to earlier threads. –However, these “ignored” invalidations are recorded using an additional pre- invalidate primary cache bit associated with each line. This is because they must be processed before a different speculative or non-speculative thread executes on this processor. –If future threads have written to a particular line in the primary cache, the pre- invalidate bit for that line is set. When the current thread completes, these bits allow the processor to quickly simulate the effect of all stored invalidations caused by all writes from later processors all at once, before a new thread begins execution on this processor. Hydra’s Handling of Five Basic Speculation Hardware Requirements For Correct Data/Thread Speculation

32. EECC722 - Shaaban #32 lec # 10 Fall 2003 10-22-2003 Thread Speculation Performance Results representative of entire uniprocessor applications Simulated with accurate modeling of Hydra’s memory and hardware speculation support.

33. EECC722 - Shaaban #33 lec # 10 Fall 2003 10-22-2003 Hydra Prototype Overview CPU core and cache. Speculative coprocessor –Speculative memory reference controller –Speculative interrupt screening mechanism –Statistics mechanisms for performance evaluation and to allow feedback for code tuning Memory system –Read and write buses –Controllers for all resources –On-chip L2 cache –Simple off-chip main memory controller –I/O and debugging interface

34. EECC722 - Shaaban #34 lec # 10 Fall 2003 10-22-2003 Hydra Prototype Layout 250 MHz clock rate target

35. EECC722 - Shaaban #35 lec # 10 Fall 2003 10-22-2003 Hydra Conclusions Hydra offers a number of advantages –Good performance on parallel applications –Promising performance on difficult to parallelize applications uniprocessor applications using data/thread speculation mechanisms. –Scalable, modular design –Low hardware overhead support for speculative thread parallelism, yet greatly increases the number of parallel applications.

36. EECC722 - Shaaban #36 lec # 10 Fall 2003 10-22-2003 Other Thread Level Speculation (TLS) Efforts: Wisconsin Multiscalar This CMP design proposed the first reasonable hardware to implement TLS. Unlike Hydra, Multiscalar implements a ring-like network between all of the processors to allow direct register-to-register communication. –Along with hardware-based thread sequencing, this type of communication allows much smaller threads to be exploited at the expense of more complex processor cores. The designers proposed two different speculative memory systems to support the Multiscalar core. –The first was a unified primary cache, or address resolution buffer (ARB). Unfortunately, the ARB has most of the complexity of Hydra’s secondary cache buffers at the primary cache level, making it difficult to implement. –Later, they proposed the speculative versioning cache (SVC). The SVC uses write-back primary caches to buffer speculative writes in the primary caches, using a sophisticated coherence scheme.

37. EECC722 - Shaaban #37 lec # 10 Fall 2003 10-22-2003 This CMP-with-TLS proposal is very similar to Hydra, including the use of software speculation handlers. However, the hardware is simpler than Hydra’s. The design uses write-back primary caches to buffer writes— similar to those in the SVC—and sophisticated compiler technology to explicitly mark all memory references that require forwarding to another speculative thread. Their simplified SVC must drain its speculative contents as each thread completes, unfortunately resulting in heavy bursts of bus activity. Other Thread Level Speculation (TLS) Efforts: Carnegie-Mellon Stampede

38. EECC722 - Shaaban #38 lec # 10 Fall 2003 10-22-2003 This CMP design has three processors that share a primary cache and can communicate register-to-register through a crossbar. Each processor can also switch dynamically among several threads. (TLS & SMT??) As a result, the hardware connecting processors together is quite complex and slow. However, programs executed on the M-machine can be parallelized using very fine-grain mechanisms that are impossible on an architecture that shares outside of the processor cores, like Hydra. Performance results show that on typical applications extremely fine-grained parallelization is often not as effective as parallelism at the levels that Hydra can exploit. The overhead incurred by frequent synchronizations reduces the effectiveness. Other Thread Level Speculation (TLS) Efforts: MIT M-machine