10 The vectorization transformation can be rather complicated, involving several
11 potential alternatives, especially for outer-loops [1]_ but also possibly for
12 innermost loops. These alternatives may have significant performance impact,
13 both positive and negative. A cost model is therefore employed to identify the
14 best alternative, including the alternative of avoiding any transformation
17 The Vectorization Plan is an explicit model for describing vectorization
18 candidates. It serves for both optimizing candidates including estimating their
19 cost reliably, and for performing their final translation into IR. This
20 facilitates dealing with multiple vectorization candidates.
25 Vectorization Workflow
26 ----------------------
27 VPlan-based vectorization involves three major steps, taking a "scenario-based
28 approach" to vectorization planning:
30 1. Legal Step: check if a loop can be legally vectorized; encode constraints and
34 a. Build initial VPlans following the constraints and decisions taken by
35 Legal Step 1, and compute their cost.
36 b. Apply optimizations to the VPlans, possibly forking additional VPlans.
37 Prune sub-optimal VPlans having relatively high cost.
38 3. Execute Step: materialize the best VPlan. Note that this is the only step
43 In what follows, the term "input IR" refers to code that is fed into the
44 vectorizer whereas the term "output IR" refers to code that is generated by the
45 vectorizer. The output IR contains code that has been vectorized or "widened"
46 according to a loop Vectorization Factor (VF), and/or loop unroll-and-jammed
47 according to an Unroll Factor (UF).
48 The design of VPlan follows several high-level guidelines:
50 1. Analysis-like: building and manipulating VPlans must not modify the input IR.
51 In particular, if the best option is not to vectorize at all, the
52 vectorization process terminates before reaching Step 3, and compilation
53 should proceed as if VPlans had not been built.
55 2. Align Cost & Execute: each VPlan must support both estimating the cost and
56 generating the output IR code, such that the cost estimation evaluates the
57 to-be-generated code reliably.
59 3. Support vectorizing additional constructs:
61 a. Outer-loop vectorization. In particular, VPlan must be able to model the
62 control-flow of the output IR which may include multiple basic-blocks and
65 c. Combinations of the above, including nested vectorization: vectorizing
66 both an inner loop and an outer-loop at the same time (each with its own
67 VF and UF), mixed vectorization: vectorizing a loop with SLP patterns
68 inside [4]_, (re)vectorizing input IR containing vector code.
69 d. Function vectorization [2]_.
71 4. Support multiple candidates efficiently. In particular, similar candidates
72 related to a range of possible VF's and UF's must be represented efficiently.
73 Potential versioning needs to be supported efficiently.
75 5. Support vectorizing idioms, such as interleaved groups of strided loads or
76 stores. This is achieved by modeling a sequence of output instructions using
77 a "Recipe", which is responsible for computing its cost and generating its
80 6. Encapsulate Single-Entry Single-Exit regions (SESE). During vectorization
81 such regions may need to be, for example, predicated and linearized, or
82 replicated VF*UF times to handle scalarized and predicated instructions.
83 Innerloops are also modelled as SESE regions.
87 The low-level design of VPlan comprises of the following classes.
89 :LoopVectorizationPlanner:
90 A LoopVectorizationPlanner is designed to handle the vectorization of a loop
91 or a loop nest. It can construct, optimize and discard one or more VPlans,
92 each VPlan modelling a distinct way to vectorize the loop or the loop nest.
93 Once the best VPlan is determined, including the best VF and UF, this VPlan
94 drives the generation of output IR.
97 A model of a vectorized candidate for a given input IR loop or loop nest. This
98 candidate is represented using a Hierarchical CFG. VPlan supports estimating
99 the cost and driving the generation of the output IR code it represents.
102 A control-flow graph whose nodes are basic-blocks or Hierarchical CFG's. The
103 Hierarchical CFG data structure is similar to the Tile Tree [5]_, where
104 cross-Tile edges are lifted to connect Tiles instead of the original
105 basic-blocks as in Sharir [6]_, promoting the Tile encapsulation. The terms
106 Region and Block are used rather than Tile [5]_ to avoid confusion with loop
110 The building block of the Hierarchical CFG. A pure-virtual base-class of
111 VPBasicBlock and VPRegionBlock, see below. VPBlockBase models the hierarchical
112 control-flow relations with other VPBlocks. Note that in contrast to the IR
113 BasicBlock, a VPBlockBase models its control-flow successors and predecessors
114 directly, rather than through a Terminator branch or through predecessor
115 branches that "use" the VPBlockBase.
118 VPBasicBlock is a subclass of VPBlockBase, and serves as the leaves of the
119 Hierarchical CFG. It represents a sequence of output IR instructions that will
120 appear consecutively in an output IR basic-block. The instructions of this
121 basic-block originate from one or more VPBasicBlocks. VPBasicBlock holds a
122 sequence of zero or more VPRecipes that model the cost and generation of the
123 output IR instructions.
126 VPRegionBlock is a subclass of VPBlockBase. It models a collection of
127 VPBasicBlocks and VPRegionBlocks which form a SESE subgraph of the output IR
128 CFG. A VPRegionBlock may indicate that its contents are to be replicated a
129 constant number of times when output IR is generated, effectively representing
130 a loop with constant trip-count that will be completely unrolled. This is used
131 to support scalarized and predicated instructions with a single model for
132 multiple candidate VF's and UF's.
135 A pure-virtual base class modeling a sequence of one or more output IR
136 instructions, possibly based on one or more input IR instructions. These
137 input IR instructions are referred to as "Ingredients" of the Recipe. A Recipe
138 may specify how its ingredients are to be transformed to produce the output IR
139 instructions; e.g., cloned once, replicated multiple times or widened
140 according to selected VF.
143 Stores information used for generating output IR, passed from
144 LoopVectorizationPlanner to its selected VPlan for execution, and used to pass
145 additional information down to VPBlocks and VPRecipes.
147 Related LLVM components
148 -----------------------
149 1. SLP Vectorizer: one can compare the VPlan model with LLVM's existing SLP
150 tree, where TSLP [3]_ adds Plan Step 2.b.
152 2. RegionInfo: one can compare VPlan's H-CFG with the Region Analysis as used by
157 .. [1] "Outer-loop vectorization: revisited for short SIMD architectures", Dorit
158 Nuzman and Ayal Zaks, PACT 2008.
160 .. [2] "Proposal for function vectorization and loop vectorization with function
161 calls", Xinmin Tian, [`cfe-dev
162 <http://lists.llvm.org/pipermail/cfe-dev/2016-March/047732.html>`_].,
164 See also `review <https://reviews.llvm.org/D22792>`_.
166 .. [3] "Throttling Automatic Vectorization: When Less is More", Vasileios
167 Porpodas and Tim Jones, PACT 2015 and LLVM Developers' Meeting 2015.
169 .. [4] "Exploiting mixed SIMD parallelism by reducing data reorganization
170 overhead", Hao Zhou and Jingling Xue, CGO 2016.
172 .. [5] "Register Allocation via Hierarchical Graph Coloring", David Callahan and
173 Brian Koblenz, PLDI 1991
175 .. [6] "Structural analysis: A new approach to flow analysis in optimizing
176 compilers", M. Sharir, Journal of Computer Languages, Jan. 1980
178 .. [7] "Enabling Polyhedral Optimizations in LLVM", Tobias Grosser, Diploma
181 .. [8] "Introducing VPlan to the Loop Vectorizer", Gil Rapaport and Ayal Zaks,
182 European LLVM Developers' Meeting 2017.