Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.

Deeper Look Into HSAIL And It's Runtime


Published on

Technical overview of the HSAIL and HSAIL runtime from AFDS by Norm Rubin.

Published in: Technology
  • Be the first to comment

Deeper Look Into HSAIL And It's Runtime

  1. 1. HSAILNorm RubinFellowAn introduction to the HSA Intermediate language
  2. 2. Disclaimer & Attribution The information presented in this document is for informational purposes only and may contain technical inaccuracies, omissions and typographical errors. The information contained herein is subject to change and may be rendered inaccurate for many reasons, including but not limited to product and roadmap changes, component and motherboard version changes, new model and/or product releases, product differences between differing manufacturers, software changes, BIOS flashes, firmware upgrades, or the like. There is no obligation to update or otherwise correct or revise this information. However, we reserve the right to revise this information and to make changes from time to time to the content hereof without obligation to notify any person of such revisions or changes. NO REPRESENTATIONS OR WARRANTIES ARE MADE WITH RESPECT TO THE CONTENTS HEREOF AND NO RESPONSIBILITY IS ASSUMED FOR ANY INACCURACIES, ERRORS OR OMISSIONS THAT MAY APPEAR IN THIS INFORMATION. ALL IMPLIED WARRANTIES OF MERCHANTABILITY OR FITNESS FOR ANY PARTICULAR PURPOSE ARE EXPRESSLY DISCLAIMED. IN NO EVENT WILL ANY LIABILITY TO ANY PERSON BE INCURRED FOR ANY DIRECT, INDIRECT, SPECIAL OR OTHER CONSEQUENTIAL DAMAGES ARISING FROM THE USE OF ANY INFORMATION CONTAINED HEREIN, EVEN IF EXPRESSLY ADVISED OF THE POSSIBILITY OF SUCH DAMAGES. AMD, the AMD arrow logo, and combinations thereof are trademarks of Advanced Micro Devices, Inc. All other names used in this presentation are for informational purposes only and may be trademarks of their respective owners. OpenCL is a trademark of Apple Inc. used with permission by Khronos. DirectX is a registered trademark of Microsoft Corporation. © 2012 Advanced Micro Devices, Inc. All rights reserved.2 | hsail AFDS | June 11, 2012
  3. 3. WHAT IS SPLIT COMPILATION?App starts a source program1) A high level compiler (HLC) generates HSAIL2) The HSAIL is shipped to the target machine3) A second compiler (a finalizer) turns HSAIL into ISAUnlike traditional compilers, where optimization is contained in one part or done twiceHSAIL allows optimization to be split into two partsThe heavy lifting goes to the HLC , the quick finish goes to the finalizerHSAIL provides ways for an HLC and a finalizer to cooperate For instance: HSAIL provides a fixed number of registers. HSA implementations might support a different number When the HLC spills registers, it can use special operations that will let the finalizer know where to use extra registers.3 | hsail AFDS | June 11, 2012
  4. 4. SPLIT COMPILATION(MEANS THERE HAS TO BE WAYS TO PASS INFORMATION FROM HLC TO FINALIZER)HLC – High level compiler Lots of time Info from source Lots of aggressive optimizations But limited (or no) knowledge of targetFinalizer Very little time (we estimate that it will take close to linear time) No info not in HSAIL (no back doors (almost) Cannot update regularly (close to bug free) Simple optimizations only But knows the target Exactly how to split some optimizations is still an open problem4 | hsail AFDS | June 11, 2012
  5. 5. WHY A VIRTUAL ISA - WHY NOT JUST TARGET THE REAL ISA?ISA Gains performance Better time to market (because hardware is finished faster)Loses performance (cannot use every hardware trick)No legacy boat anchorReal isa means one vendor/ one chip familyCan fix hardware bugs in softwareOld and new code just works on old and new machinesAllows hardware innovation under the tableFeatures not in HSAIL are not exposed, and are hard to access5 | hsail AFDS | June 11, 2012
  6. 6. Development tools at HSAIL level Today the need for a complete tool chain for each core, each with its own technology, switches etc., is a significant maintenance problem. Debuggability, reproducibility. Because the same application needs to run on different pieces of hardware, current source code contains many conditional preprocessing directives Programmers rely on compiler intrinsic and ad-hoc command line arguments to drive the optimization. This severely impacts code readability and productivity, and the application binary tested and debugged on a workstation is different from the one that eventually runs on the system. Platform openness. Independent software vendors rarely have access to the tool chains needed to program the most powerful parts of the system, namely the DSPs and hardware accelerators. Virtualization can make the whole platform programmable, opening opportunities to third-party high-performance applications .Performance through time to market Because of the finalizer, last minute fixes can happen after the chip is finished. This means that the time to release a new part goes down. Less time per generation translates to better performance6 | hsail AFDS | June 11, 2012
  7. 7. GOALS OF HSAIL1. Can support all of C++ (open up the GPU to mass programming, not only for specialists)2. Avoid constant change (do not change the spec every chip)3. Support accurate IEEE floating point math4. Target lots of different machines5. Allow for packed operations, SSE and friends, bytes/shorts/ints/doubles etc6. Allow packed forms to save power7. Make the model understandable8. Make the finalizer fast (around linear time)9. Make the finalizer simple (do not need monthly updates)10. Less ambiguity in the spec (little undefined behavior)11. Get good performance (little need to write in ISA)12. Support all of OpenCL™ and C++Amp™13. Can ship linkable libraries in HSAIL14. Clean up all nits in AMDIL15. Allow the use of chip specific acceleration when it is a good idea7 | hsail AFDS | June 11, 2012
  8. 8. HSAIL – LOTS OF NEW FEATURESLots of features not in OpenCL and C++ AMP Enough to implement C++ Exceptions/ heterogeneous compute Flat address space (work items on the GPU and agents on the CPU)Because of hand written HSAIL, these features can be exposed earlyFine-grain barriers that work inside control flow, you can implement producer consumer modelsLots of cross wave operations – so you can quickly move data between lanes without loads and storesSpec is available on the web siteThe memory model shows how the CPU and GPU can cooperateSupport for image operations8 | hsail AFDS | June 11, 2012
  9. 9. PARALLELISM MODEL9 | hsail AFDS | June 11, 2012
  10. 10. WAVEFRONTSMost developers will not care about wavefrontsSimilar to cache line sizes Experts can get good performance if they code to the cache line size Compiler has to avoid breaking the developers model HSAIL formalizes the notion of wavefronts you can tell which work item goes into which wavefront you can write producer consumer parallelism between work groups10 | hsail AFDS | June 11, 2012
  11. 11. AN EXAMPLE (IN OPENCL™)__kernel void vec_add (__global const float *a, __global const float *b, __global float *c, const unsigned int n){ // Get our global thread ID int id = get_global_id(0); // Make sure we do not go out of bounds if (id < n) { c[id ] = a[id] + b[id];}11 | hsail AFDS | June 11, 2012
  12. 12. VECTOR ADD A[0:N-1] = B[0:N-1] + C[0:N-1] cur $c0, @BB0_2;version 1:0:$small; brn @BB0_1;kernel &__OpenCL_vec_add_kernel( @BB0_1: // %if.end kernarg_u32 %arg_a ret; kernarg_u32 %arg_b, @BB0_2: // %if.then kernarg_u32 %arg_c, shl_u32 $s1, $s1, 2; kernarg_u32 %arg_n) add_u32 $s2, $s2, $s1;{ @__OpenCL_vec_add_kernel_entry: ld_global_f32 $s2, [$s2]; // BB#0: // %entry add_u32 $s3, $s3, $s1; ld_kernarg_u32 $s0, [%arg_n]; ld_global_f32 $s3, [$s3]; workitemaid $s1, 0; add_f32 $s2, $s3, $s2; cmp_lt_b1_u32 $c0, $s1, $s0; add_u32 $s0, $s0, $s1; ld_kernarg_u32 $s0, [%arg_c]; st_global_f32 $s2, [$s0]; ld_kernarg_u32 $s2, [%arg_b]; brn @BB0_1; ld_kernarg_u32 $s3, [%arg_a]; };12 | hsail AFDS | June 11, 2012
  13. 13. MEMORY SEGMENTS Memory is split into 7 segments kernarg, global, arg, readonly, private, group, and spill There is a single flat address space with everything but its is often advantageous to tell the finalizer which segment to use Load/store machine with registers Some segments are used for intent – – Spill indicates that the slot was used by the HLC for register spilling13 | hsail AFDS | June 11, 2012
  14. 14. SEGMENTS NDRange Work group Work group Work Items Group Private group Arg locations are in private Private Spill locations are in private Agent Flat address space Group within Private within arg memory is within Private flat flat spill memory is within Private privateRW is within Private kernarg is within Global ReadOnly is within Global14 | hsail AFDS | June 11, 2012
  15. 15. HSAIL FEATURES REGISTERS AND TypesTYPES Brigs8, Brigs16, Brigs32, Brigs64,Four classes of registers Brigu8, Brigu16, Brigu32, Brigu64, c/s/d/q Brigf16, Brigf32, Brigf64, Brigb1, 1 bit Brigb8, Brigb16, Brigb32, Brigb64, 32 bits Brigb128, Brigu8x16, 64 bits BrigROImg, BrigRWImg, BrigSamp, 128 bits Brigu8x4, Brigs8x4, Brigu8x8, Brigs8x8,Both Binary (BRIG) and text format Brigs8x16,The binary format is fully specified Brigu16x2, Brigs16x2, Brigf16x2, Brigu16x4, Brigs16x4, Brigf16x4, Brigu16x8,120 opcodes (JavaByte code has 200) Brigs16x8, Brigf16x8, Brigu32x2, Brigs32x2, Brigf32x2, Brigu32x4, Brigs32x4, Brigf32x4, Brigu64x2, Brigs64x2, Brigf64x215 | hsail AFDS | June 11, 2012
  16. 16. WHY DOES HSAIL LOOK THIS WAY?An SIMT model (single instruction, multiple threads) claims that every work-item has a program counterSo branch instructions look pretty naturalA vector machine model looks like sse, one program counter and vector registers, this is like real AMD GPUhardwareSIMT or Vector?16 | hsail AFDS | June 11, 2012
  17. 17. PROS FOR SIMT We want HSAIL to outlast one hardware generation (so at the very least the vector length and real types/number of registers should not get exposed). Even with a vector model the finalizer will still have to map to the real vector length. We expected this to mean that a vector finalizer would not have a much simpler time We want to support lots of machines including ones not built by AMD We can add cross lane operations (like count) to the SIMTmodel so the line between SIMT and vector is blurry We want to open up to 3rd party compiler and tools, all of which can support SIMT but few of which can support vector Work groups is a much more developer friendly model than wavefronts Natural path for OpenCL™/CUDA ™ c++amp™ Graphics is SIMT, so the pressure to make future hardware work well for SIMT is immense17 | hsail AFDS | June 11, 2012
  18. 18. PROS FOR VECTOR Might get more performance, we estimated <10% even in good cases Simpler for expert programmers to reason out what is going on This was a big one for us, the exact rules on wavefront re-convergence are hidden in the SIMTmodel but clear in the vector one In the vector model you can prove some results about code, which cannot be done when the finalizer reorders things On the other hand constructs like C++ virtual functions become very confusing on a vector machine, where the original program was SIMT We think the performance deficits are a reasonable trade for broader adoption, and in many cases can be closed by well written libraries for the cases that really matter.18 | hsail AFDS | June 11, 2012
  19. 19. HSAIL AND FUNCTIONS{ arg_u32 %input1; arg_u32 %input2; // … call &fnWithTwoArgs ()(%input1, %input2); // call of a function // all work-items call the same function }// ...HSAIL supportsVirtual functions,SignaturesJumps via a registerLoad address of code19 | hsail AFDS | June 11, 2012
  20. 20. HSAIL PROVIDES A SERIES OF OPTIMIZATION CONTROLSSometimes you know if an operation is uniform over a rangeld_f32_width(8) $s1, addressWork items in groups of 8 will read the same valuecall_width(64) $s1Even through this is a call through register, work items in groups of 64 will call the same functionld_equiv(3)_u32 $s1, addressA block of memory that cannot alias with other blocks20 | hsail AFDS | June 11, 2012
  21. 21. HSAIL COMPARED TO LLVM-IRHSAIL is low level assumes finalizer does not do as much optimization no phi nodes, finite register count No ssa inputParallelism is built into HSAIL No need to hack the meaning of a barrierNo structures or other high level features21 | hsail AFDS | June 11, 2012
  22. 22. HSAIL COMPARED TO JAVA BYTE CODEHSAIL is more focused on performance,HSAIL has registers not a stackHSAIL has parallelism built inHSAIL is not as focused on security (does not require a formal validator)Not quite write onceHSAIL is less concerned about code compression22 | hsail AFDS | June 11, 2012
  23. 23. HSAIL COMPARED TO AMDILHSAIL supports lots of complex control flow AMDIL provides structured control flow only irreducible flow needed exponential compile timeNo (or limited) graphics features just enough for C++ AMP™ and OpenCL™four sizes of registers 1/32/64/128 bit vs. 4x32 vector registers (no more .x, .y, .z, .w) fieldsHSAIL is extendable (per vendor/per chip extensions)Different cost model23 | hsail AFDS | June 11, 2012
  24. 24. HSAIL COMPARED TO PTXMore formal model of execution possible to write valid programs that pass data between work groupsMore formal model of memory - acq/rel semanticsLess semantics defined by the deviceSupport for libraries and complex calls Interaction between agents and HSAIL code, shared memory, support for GPU to call CPU servicesPer vendor extension mechanismClean separation of core features and per device operationsSupport for linking/ libraries/ separate compilationRemoval of hard to finalize features no predication24 | hsail AFDS | June 11, 2012
  25. 25. MEMORY MODELA memory model defines how writes by one work-item or agent become visible toother work-items and agents.For many implementations, better performance will result if either the hardware or the finalizer is allowed to reordercode. For example, the finalizer might find it more efficient if a write is moved later in the program; so long as theprogram semantics do not change, the finalizer is free to do so. Once a store is deferred, other work-items andagents will not see it until the store actually happens. Hardware might provide a cache that also defers writes.The HSAIL memory model is based on acquire releaseAn ld_acq creates a “downward fence.” This means that normal loads and stores can be moved (by theimplementation) down past the ld_acq but no memory operation (load, store, or atomic) can be moved up above theld_acq.A st_rel creates an “upward fence.” That means that normal loads and stores can be moved (by theimplementation) above the st_rel but no memory operation (load, store, or atomic) can be moved down after thest_rel.25 | hsail AFDS | June 11, 2012
  26. 26. Original Axiomatic Definition [Lamport 1979] A single processor (core) sequentially consistent if “the result of an execution is the same as if the operations had been executed in the order specified by the program.” A multiprocessor sequentially consistent if “the result of any execution is the same as if the operations of all processors (cores) were executed in some sequential order, and the operations of each individual processor (core) appear in this sequence in the order specified by its program.”26 | hsail AFDS | June 11, 2012
  27. 27. SEQUENTIAL CONSISTENCY (SC) OPERATIONAL DEFINITION System P P P 1 memory P simple processors MEMORY Operation: Pick one ready row, do it, & repeat until done Processor 0 ready to load/store of memory … Processor P-1 ready to load/store of memory27 | hsail AFDS | June 11, 2012
  28. 28. SEQUENTIAL CONSISTENCYAny SC implementation must only permit executions allowed by SC operational model (SC executions).The SC operational model is NOT a performance model. SC implementation performance != Counting operation model stepsThe operational model hides most implementation techniques pipelining, out-of-order, speculation, caches, cache coherence, … HW must functional behave “as if” is was like operational modelHW designers & verifiers often most comfortable with operational modelEach processor is eventually selected28 | hsail AFDS | June 11, 2012
  29. 29. HSAIL OPERATIONAL DEFINITION P P P System 1 (host) memory P simple processors Reorder buffer Writes can get held Reads can be satisfied MEMORY Operation: Pick one ready row, do it, & repeat until done Processor 0 ready to load/store of memory … Processor P-1 ready to load/store of memory write values may stay in reorder buffer, reads may come out of the reorder buffer, Rules to move between reorder buffer and memory rel = release the values from the buffer, acq = acquire new values29 | hsail AFDS | June 11, 2012
  30. 30. WITHIN ONE WORK ITEMSEQUENCED BEFOREThis is the order operations appear in the sourceWhat you see looking at the codesingle work item - “as-if-serial” view - each operation appears to happen in the order it appears in the sourceX sb Y - X and Y in same work item, - X sequenced before Ymultiple work items and agents makes this more complex30 | hsail AFDS | June 11, 2012
  31. 31. BETWEEN WORK ITEMSX >> YWhat the memory system seesmemory system must see X before Yglobal visibility orderthis is transitive X >>Y, and Y >> Z, then X >>Z31 | hsail AFDS | June 11, 2012
  32. 32. RULES, SOMETIMESX SB Y => X >> Y•X sb Y, same address, then X >>Y•Different address –If there is a barrier or sync between X and Y then X >>Y•If X is an acquire: – ld_acq, atomic_acq, atomicNoRet_acq, atomic_ar, atomicNoRet_ar –Then X >> Y –This is one sided (Y cannot move before X)The general rule is use acquire and release when you want to force orderAcquire and Release may take extra time, but they give you sequential constancy Compilers can trade performance for simple cross work-item communication32 | hsail AFDS | June 11, 2012
  33. 33. •If Y is a release –st_rel, atomic_ar or atomicNoRet_ar then X >>Y –st rel is another one way fence •Consider a critical region (can use acquire and release to form critical sections) •ld_acq x •Assorted memory operations •st_rel y •No operations can move out, but operations can move in33 | hsail AFDS | June 11, 2012
  34. 34. AN EXAMPLE SB ORDER DOES NOT FORCE MEMORY ORDERWork-item 0 Work-item 1------------------- ------------------------------------@h0: st_u32 1, [&a] @k0: st_u32 1, [&b]@h1: ld_u32 $s0, [&b] @k1: ld_u32 $s1, [&a]Initially, &a and &b = 0. $s0 = 0 and $s1 = 0 is allowed. --constraints added because readers have to follow writers. k1 (the reader)has to happen before h0 changes the value. There are also constraints caused by synchronization h1 >> k1 >> h0 >> k0.Even though h0 appears first (in sequenced-before order) before h1, there is norequirement that the operations appear in text order (sequenced-before order) to thememory system.34 | hsail AFDS | June 11, 2012
  35. 35. EXAMPLE 2 REGISTER DEPENDENCE DOES NOT FORCE MEMORY ORDERWork-item 0 Work-item 1----------------------- ---------------------@h0: ld $s0, [&a] @j0: st 20, [100]@h1: ld $s1, [$s0] @j1: st_rel 100, [&a]Initially, &a and contents of location 100 = 0.$s1 == 0 and $s0 == 100 is allowedIf $s1 == 0 then h1 >> j0. f $s0 == 100 then j1 >> h1.Because this seems to violate dependence order, it is useful to consider how this cancome about.Work-item 0 is allowed to prefetch load h1. One reason it might do this is that code before these operationsreads address 96, and the implementation reads in large cache lines.Later, work-item 1 reads the new value of &a, which is 100. Then it reads the value oflocation 100, but because there is no synchronization, it can use the previously prefetched value of 0.35 | hsail AFDS | June 11, 2012
  36. 36. EXAMPLE 3Work-item 0 Work-item 1@h0: ld_acq $s0, [&a] @j0: st 20, [100]@h1: ld $s1, [$s0] @j1: st_rel 100, [&a]Initially, &a and 100 = 0.HSAIL does not allow $s1 == 0 and $s0 == 100.36 | hsail AFDS | June 11, 2012
  37. 37. QUESTIONS?37 | hsail AFDS | June 11, 2012