diff --git "a/D9E4T4oBgHgl3EQffA2Y/content/tmp_files/load_file.txt" "b/D9E4T4oBgHgl3EQffA2Y/content/tmp_files/load_file.txt"
new file mode 100644--- /dev/null
+++ "b/D9E4T4oBgHgl3EQffA2Y/content/tmp_files/load_file.txt"
@@ -0,0 +1,957 @@
+filepath=/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf,len=956
+page_content='Learning to compile smartly for program size reduction  Youwei Liang∗ 1 Kevin Stone∗ 2 Ali Shameli 2 Chris Cummins 2 Mostafa Elhoushi 2 Jiadong Guo 2  Benoit Steiner 2 Pengtao Xie 1 Hugh Leather 2 Yuandong Tian 2  Abstract  Compiler optimization passes are an important  tool for improving program efficiency and reduc-  ing program size, but manually selecting opti-  mization passes can be time-consuming and error-  prone.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' While human experts have identified a few  fixed sequences of optimization passes (e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', the  Clang -Oz passes) that perform well for a wide  variety of programs, these sequences are not con-  ditioned on specific programs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' In this paper, we  propose a novel approach that learns a policy to  select passes for program size reduction, allow-  ing for customization and adaptation to specific  programs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' Our approach uses a search mecha-  nism that helps identify useful pass sequences  and a GNN with customized attention that se-  lects the optimal sequence to use.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' Crucially it  is able to generalize to new, unseen programs,  making it more flexible and general than previ-  ous approaches.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' We evaluate our approach on a  range of programs and show that it leads to size  reduction compared to traditional optimization  techniques.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' Our results demonstrate the potential  of a single policy that is able to optimize many  programs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' Introduction  Finding the right compiler optimization ordering for a given  program in order to execute them more efficiently with a  smaller amount of resources (e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', memory, CPU and stor-  age) is an important yet challenging problem.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' Traditionally,  to tune configurations, human effort and expert knowledge  are needed, which is a time-consuming and error-prone  process that often yields sub-par results.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  In recent years, machine learning-guided compiler optimiza-  tion has emerged as an interesting field to replace this labori-  ous process (Wang & O’Boyle, 2018).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' Along this line, many  Equal contribution  1University of California, San Diego  2Meta AI.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' Correspondence to:  Kevin Stone  <kevinlee-  stone@meta.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='com>.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  works show promising results using various machine learn-  ing (ML) techniques and optimizers (e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', reinforcement  learning (Haj-Ali et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', 2020a), language modelling (Cum-  mins et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', 2017), evolutionary algorithms (Kulkarni &  Cavazos, 2012), etc).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' They focus on automatic decision-  making of specific components in compilation, e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', op-  timizing computational graph of neural network for ML  compilers (Zhou et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', 2020), optimizing the loop structure  of neural network computations (Steiner et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', 2021), deter-  mination of a function to be inlining (MLGO) (Trofin et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=',  2021), etc.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' There are only a few works targeting generic  optimization of compilation (Haj-Ali et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', 2020b;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' Almakki  et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', 2022).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  In this work, we take a different path and explore the pos-  sibility of learning policies for more a general aspect of  compiler optimization.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' More specifically, we treat compila-  tion optimization as a general sequential decision process,  in which each step is to find an additional compiler pass  so that when combined with existing passes, can improve  specific metrics (e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', the binary size / running speed of the  codebase).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' Our goal is to learn a policy, named pass policy,  to help quickly find the right compiler passes to be used,  given the current program and existing passes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' We aim to  compare with existing handtuned policies in the compiler,  e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', -Oz for code size reduction and “-O3” for running  speed optimization.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' Such policies have been tuned by the  domain experts for decades and used extensively in all the  computer systems, but are invariant to the program being  compiled.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  As the first contribution, we propose a novel evaluation  protocol for compiler pass optimization, named zero-shot  generalizability (ZSG), to evaluate learned policies applied  to unseen programs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' Here “zero-shot” means that each  unseen program has never been seen before in the training  process, and we allow a fixed number of optimization passes  (say 45).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' The policy is allowed to adjust the passes based  on previous outcomes (This includes backtracking if an  exploratory step turns out to be sub-optimal.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=') Compared  to previous works (e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', Autophase (Haj-Ali et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', 2020b),  CompilerGym (Cummins et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', 2022), MLGO (Trofin et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=',  2021), MLGoPerf (Ashouri et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', 2022)) that run adaptive  search algorithms to optimize a set of programs for many  hours, we argue that ZSG is a realistic setting for practical  arXiv:2301.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='05104v1  [cs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='PL]  9 Jan 2023  Learning to compile smartly for program size reduction  deployment: for most programs except the critical ones,  there is a limited time budget to optimize them, and often  there is not enough computational resources to fine-tune  learned policies for each unseen task separately (e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', fine-  tune setting in GO (Zhou et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', 2020)).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' While similar metric  has been used in previous works (e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', MLGO and GO), we  are the first to test it in more general ML-guided compiler  optimizations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  With the ZSG metric, we then evaluate multiple existing  techniques applied to compiler optimization, with LLVM  compiler and on large-scale datasets provided by Compiler-  Gym (Cummins et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', 2022).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' We use code-size reduction  as the metric since it is a deterministic quality and relatively  easy/cheap to compute, and we leave program run-time  optimization for future work.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  Given the metric, we discover a universal core subset of  LLVM pass sequences, of size 50, that leads to very strong  performance across multiple sets of programs from diverse  domains, ranging from Linux Kernel to BLAS library.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' The  total number of steps in the 50 pass sequences is 625 and  the average length of the sequences is 12.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' Specifically,  for one unseen program, there exists one of the 50 pass  sequences that leads to an average code size reduction of  5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='8% compared to the default -Oz setting, across 10 diverse  codebases of over one million programs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' In other words,  after running trying the 625 steps on an unseen program, the  smallest code size during the compilation is 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='8% smaller  than that of -Oz.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' Considering the huge search space of com-  piler flags (10104), this is a very surprising finding.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' We find  this coreset in a training set and it can generalize reasonably  well to the evaluation set including datasets not including in  the training set.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  While it can be time-consuming to find such a compiler  flag configuration with an exhaustive enumeration of the  core subset, we find that the optimal pass can be directly  predicted with high accuracy via GNN with a customized  attention layer to control information flow with the Pro-  GraML (Cummins et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', 2021) graph as the input.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' The  prediction top-1 and top-5 accuracy is 75% and 95%.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' There-  fore, we can run a few sequences selected by the model on  an unseen program to obtain a good code size reduction.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  This enables us to find a good flag configuration that leads  to 4% improvement on average, with just 45 compilation  passes (this is roughly 3 sequences since the average length  of the sequences in the coreset is 12.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='5).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  We have compared our approach with extensive baselines,  including RL-based methods such as PPO and Q-learning  and black-box optimizers such as evolutionary algorithm.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  It turns out that ML approaches often suffer from severe  overfitting to the training set and does not perform well  for new program categories, even combined with heuristic  search, and optimizers may get lost in the exponential action  space due to lack of domain knowledge.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' In comparison, our  approach is simple, effective and generalizable to unseen  programs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' Related Work  Graph structured data are present in numerous applications  and it has been shown that taking advantage of this data  can help us train very effective machine learning models.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  (Brauckmann et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', 2020) use abstract syntax trees and con-  trol flow graphs for learning compiler optimization goals.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  They show that using such graphs allows them to outper-  form state-of-the-art in the task of heterogeneous OpenCL  mapping.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' (Guo et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', 2020) uses a transformer based model  with a graph guided masked attention that incorporates the  data flow graph into the training.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' They achieve state of the  art performance in four tasks including code search, clone  detection, code translation, and code refinement.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  As a contender to graph neural networks, (Mialon et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=',  2021) uses transformers to process graphs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' They show that  if we effectively encode positional and local sub-structures  of graphs and feed them to the transformer, then the trans-  former can outperform the classical GNN models.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' They  test their model on classification and regression tasks and  achieve state of the art performance.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  In (Srinivas et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=',  2020), they used an unsupervised model to learn embed-  dings of high dimensional pixel data using contrastive learn-  ing.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' They then use this embedding for downstream rein-  forcement learning tasks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' Methodology  3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' Action space  The CompilerGym framework provides a convenient in-  terface for the compiler pass ordering problem.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' The de-  fault environment allows choosing one of 124 discrete ac-  tions at each step corresponding to running a sequence of  specific compiler pass.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' Given that our trajectories have a  length of 45 steps, this means we have 12445 ∼ 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='6 × 1094  possible action trajectories to explore.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' To find an opti-  mal action sequence for a program, we can apply some  existing reinforcement learning methods including Q learn-  ing like DQN (Mnih et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', 2015) and policy gradient like  PPO (Schulman et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', 2017).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  Action Sequences However for this problem it turns out  that certain action sequences are good at optimizing many  different programs (where “good” is defined as better than  the compiler default -Oz).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' We found that constraining the  action space to a learned set of action sequences enables  state of the performance and also significantly reduces the  challenge of exploration.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' This allows us to cast the prob-  lem as one of supervised learning over this set of action  Learning to compile smartly for program size reduction  Figure 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' An example of a small action tree created by keeping  track of the common prefixes of the action sequences.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  sequences.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' We use the following algorithm to find a good  set of action sequences.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  Random search We seed the list of candidate action  trajectories (action sequences) by running a random  policy on a subset of the training programs (N).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' For  each program we run M episodes and keep track of  the best action sequence for each program.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  Canonicalize During the search process we find that  trajectories often revisit the same state.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' Whenever this  happens we truncate all previous actions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' On average  this reduces the trajectory length by a factor of 1/5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  All-to-all We then test all N best action sequences  against all N programs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' This gives us a N×N matrix  where each value is the cumulative reward (return).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' We  then normalize this matrix by the maximum return for  each program.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' A value of 1 represents that an action  sequences was optimal for a program (optimal in the  sense of this limited set of action sequences).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  Greedy assignment Many action sequences are opti-  mal for more than one program.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' We take advantage  of this to reduce the number of action sequences by  greedily picking action sequences that are optimal for  the the largest number of programs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' This results in a  much smaller list of action sequences.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' We can visual-  ize this by creating a prefix tree that shows common  prefixes as single node.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' See Figure 1 for truncated tree  for illustration.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  It is interesting to note that some actions are common across  the beginning of multiple action sequences.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' These popular  actions such as 27 (-reg2mem) as seen in Figure 1 are  pivotal for many programs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' For a complete list of all LLVM  passes refer to Table 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' Offline behavior cloning methods  Normalized Value Prediction After discovering the “good”  action sequences (i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', the coreset), we can turn the problem  of the sequential decision-making on compiler passes into a  problem of supervised classification.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' The target is to train  a model to predict the best action sequence conditioned on  the program, where the label of the program is the index  of the action sequence that results in the highest code size  reduction.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' However, one important observation we have  is that there are typically multiple action sequences in the  coreset that all result in the highest code size reduction.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  Therefore, instead of using the single-class classification  method with cross entropy loss, we leverage the fact we have  access to the values for all action sequences.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' We predict  the softmax normalized value of each action sequence with  a cross entropy loss detailed below.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' We call this approach  behavior cloning (BC) over the coreset.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  For a program o, we roll out all the predefined action se-  quences on it, obtaining a return ro  i for the i-th sequence  (i.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', the highest cumulative reward observed during the  rollout of the action sequence), which forms a value vec-  tor ro = [ro  1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' , ro  n].' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' Then, the normalized values of the  action sequences are defined by  vo = Softmax(ro/T)  (1)  where T is temperature.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  For an initial observation so  0 of a program, our model out-  puts a probability distribution, p = f(so  0), over the action  sequences.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' The target of the training is to make p close to  the normalized values of the action sequences.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' We use the  cross entropy loss to supervise the model  L(po, vo) = −  n  �  i  po  i log vo  i  (2)  3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' Program Representation  We are considering LLVM optimization passes on that op-  erate on the Intermediate Representation (IR) of a program.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  The IR contains very rich structures, while its size can be  quite large for large programs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' It also contains informa-  tion (e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', strings and constants) irrelevant to the task we  are considering.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' We found that working with compressed  representations made this problem tractable and run in a  reasonable amount of time.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  ProGraML We leverage a graph based representation that  encodes semantic information of the program covering three  layers: control flow, data flow, and data types.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' This rep-  resentation has the advantage that it is not a fixed size - it  does oversimplify large programs - and yet it is still a more  compact format than the original IR format.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' Network Architecture  One of the ways to model our policy function is to use a  graph neural network (GNN).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' To achieve this goal, we use  53,122,31,36,111,10,97  10  64,31,10,52,111,116,36,40,48,54,30,53,114,29,120,10  36,103,24,53,97,53,38,69,97,57,10,29  39,64,55,53,38,122,31,111,64,10,39,21,105,36  27  104,55,57,26,103,10,29,31,36,120,102,53  root  29  55,39,61,27,41,36,25,103,10  30,48,29,120,103,96,47,29,78,21,122,41,36,10  72,55,103,36,122,59,30,65,53,10  103,102,30,36,61,29,41,71,10,61,41,52Learning to compile smartly for program size reduction  1  2  5  3  𝑋!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='"  #  𝑋!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='$  #  𝑋!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='%  #  𝑋"&  #  𝑋"!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  𝑋%!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  𝑋$!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  𝑋&"  4  Figure 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' Graph attention.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' Circles denote nodes and solid arrows  denote edges.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' Squares are the calculated features, and dash arrows  represent feature aggregation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' The orange/green squares denote the  features to be aggregated in the target/source nodes of the edges.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  the ProGraML graph structure proposed in (Cummins et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=',  2021) in which individual statements are connected to other  statements through relational dependencies.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' Compared to  the approach of treating the program scripts as text and  encoding the programs with a language model, using the  ProGraML graph representations and encoding them with  GNNs has several advantages.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' Firstly, the long-range de-  pendencies of instructions are automatically captured by  the edges in the graphs, whereas the NLP approaches need  to use an LSTM or Transformer to capture the dependen-  cies.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' An LSTM could lose early memory when the program  is long, and a Transformer could cause out of memory is-  sues in such cases.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' Secoundly, changing the names of the  variables/constants/functions/classes in a program will not  affect its ProGraML representation but will change the rep-  resentations in texts.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' Our goal is to use the structure and  relational dependencies of this graph to learn an embed-  ding which allows us to learn a better policy.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' We experi-  mented with several different architectures such as Gated  Graph Convolutions (Li et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', 2015), Graph Attention Net-  works (Veliˇckovi´c et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', 2017), as well as our own custom  variation described bellow.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  Edges types ProGraML supports multiple directed edge  types representing control flow, data flow, function call, and  data type definition.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' The other edge features include the  integer position of an edge among all the edges pointing  to the same node, a boolean feature indicating whether the  Notation  Meaning  E  The set of edges in the graph  (i, j)  Edge from node i pointing to node j  X(t)  i  Repr.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' of node i at layer t  E(t)  i→j  Repr.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' of the edge (i, j) at layer t  X(t)  ij  Repr.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' for node i associated with edge (i, j)  X′(t)  ij  Repr.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' for node i associated with edge (j, i)  a(t)  ij  Raw attention associated with repr.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' X(t)  ij  a′(t)  ij  Raw attention associated with repr.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' X′(t)  ij  α(t)  ij  Normalized attention associated with a(t)  ij  α′(t)  ij  Normalized attention associated with a′(t)  ij  Ti  Target neighbors of node i: {j|(i, j) ∈ E}  Si  Source neighbors of node i: {j|(j, i) ∈ E}  Table 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' The notations in GNN (“Repr.” means representation)  two nodes connected by the edge are in the same LLVM  basic block, and the distance of the two nodes if they are  in the same basic block.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' Equipped with the rich features of  the edges in ProGraML graphs, we propose a dynamic edge  encoding approach to capture the edge representations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  Dynamic edge representation Most existing GNNs that  exploit the edge feature basically use a static edge feature,  which means the same edge feature is repeatedly used for  all layers.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' It turns out that it is important to use a dynamic  edge representation during graph encoding, where the edge  representation gets updated in each GNN layer.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' The initial  edge representations are the concatenations of the embed-  ding of edge types, edge positions, and other edge features  discussed in the last paragraph.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  Attention with edge features It was also helpful to modify  the default attention mechanism to support these custom  edge types.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' We propose to incorporate the edge features  by encoding a triplet containing the features of two nodes  and the feature of the edge that connects them.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' For clar-  ity, we show a table containing the notations used in the  GNN in Table 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' Then, the feature update process can be  mathematically defined by the following equations, where  Mi, i = 1, .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' , 5 is an encoding MLP.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  X′(t+1)  ij  = M1(X(t)  i ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' E(t)  i→j,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' X(t)  j ),' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  (3)  a′(t+1)  ij  = M2(X(t)  i ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' E(t)  i→j,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' X(t)  j ),' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  (4)  X(t+1)  ji  = M3(X(t)  i ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' E(t)  i→j,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' X(t)  j ),' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  (5)  a(t+1)  ji  = M4(X(t)  i ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' E(t)  i→j,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' X(t)  j ),' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  (6)  E(t+1)  i→j  = M5(X(t)  i ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' E(t)  i→j,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' X(t)  j ),' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  (7)  In words,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' the 3-tuple,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' (X(t)  i ,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' E(t)  i→j,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' X(t)  j ),' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' associated with  Learning to compile smartly for program size reduction  edge (i,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' j),' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' is encoded by MLPs to output 5 features,' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' includ-  ing X′(t+1)  ij  and a′(t+1)  ij  (a representation and attention to  be aggregated in node i),' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' and X(t+1)  ji  and a(t+1)  ji  (a repre-  sentation and attention to be aggregated in node j),' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' and the  updated edge representation E(t+1)  i→j .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' Note that the features  to be aggregated to a target node are marked with the ′, and  those to a source node are without the ′.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' After the feature  encoding, we perform an attention-weighted neighborhood  aggregation for each node, which can be mathematically  described by the following equations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  �  {α(t+1)  ij  }j∈Ti ∪ {α′(t+1)  ij  }j∈Si  �  = Softmax  �  {a(t+1)  ij  }j∈Ti ∪ {a′(t+1)  ij  }j∈Si  �  (8)  X(t+1)  i  =  �  j∈Ti  α(t+1)  ij  X(t+1)  ij  +  �  j∈Si  α′(t+1)  ij  X′(t+1)  ij  (9)  3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' Dataset preparation  Overfitting issues could happen if training is performed on  a small subset of programs, or the set of programs is not  diverse enough.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' To mitigate this we found it helpful to  create an aggregate dataset that uses many different public  datasets as curated by CompilerGym.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' CompilerGym gives  us access to 14 different datasets constructed using two  different methods.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  Curated These are small collections of hand-picked  programs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' They are curated to be distinct from one  another, so splitting curated suites can be challenging.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  Typically programs are larger as they may comprise  multiple source files combined into a single program.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  These are commonly used for evaluating compiler op-  timization improvements.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  Uncurated These are comprised of individual source  files scraped from open source repositories such as  Linux, Tensorflow, or synthetically generated pro-  grams, normally targeted for compiler testing (not op-  timization).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' They may not be as ”representative” of  human written test programs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  For our aggregate dataset we decided to hold-out the entirety  of the four curated datasets for use as an out-of-domain test  set.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' This is important because they represent the types of  programs we expect to see in the wild.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' We also split the  uncurated datasets into train, validaton, and test programs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' Evaluation  For all our metrics and rewards we leverage the IR instruc-  tion count as value we are trying to minimize.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' We also  report metrics on each CompilerGym dataset as well as  Type  Dataset  Splits  Uncurated  anghabench-v1  train,val,test  blas-v0  train,val,test  github-v0  train,val,test  linux-v0  train,val,test  opencv-v0  train,val,test  poj104-v1  train,val,test  tensorflow-v0  train,val,test  clgen-v0  train,val,test  csmith-v0  train,val,test  llvm-stress-v0  train,val,test  Curated  cbench-v1  test  chstone-v0  test  mibench-v1  test  npb-v0  test  Table 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' CompilerGym dataset types and training splits.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  the mean over datasets to get a single number to compare  overall results.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  The mean percent improved over -Oz (MeanOverOz)  defined as following:  MeanOverOz :=  1  |P|  �  p  IOz  p  − Iπθ  p  IOz  p  ,  (10)  where p is a specific program from the set of programs  P in the dataset.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' IOz  p  is the number of IR instructions  in program after running the default compiler pass -Oz.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  Iπθ  p  is the number of IR instruction in the program after  applying the policy under consideration.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' We can think  of this as a simple average of the percent improvement  over -Oz.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  We  also  look  compare  the  geometric  mean  (GMeanOverOz) of final size relative to -Oz.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  This metric is less sensitive to outliers and is used  by (Cummins et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', 2022).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  GMeanOverOz :=  ��  p  IOz  p  Iπθ  p  �  1  |P|  (11)  4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' Experiments  4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' Baselines  Oracle-All We consider a brute-force search over the  action tree in order to find the best action sequence for  a given program.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' This gives us an upper-bound of the  downstream policy network.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' In this case the action tree  has a total of 625 nodes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  Learning to compile smartly for program size reduction  Oracle-Top-45 We also consider how well we would  do if the oracle is only allowed to use the most popular  action sequences but limited to 45 steps.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' We use 45  steps because this is maximum allowed for our all other  baselines and our proposed method.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  Autophase-RL We start with the strong baseline of us-  ing Autophase features.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' Autophase is a mapping from  a program to fixed size feature vector of 54 dimensions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  It contains integer counts of various program proper-  ties such as number of instructions, maximum loop  depth, etc.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' This is used in combined with a 2-layer  MLP model and trained with the PPO algorithm.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' This  is the approach presented in (Haj-Ali et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', 2020b).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  Autophase-BC We also consider the performance of  using the action sequences combined with a MLP  model using Autophase features to isolate the contribu-  tion of the GNN from the action sequences search.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  GNN-RL We compare with using ProGraML features  and our GNN model, but trained with the PPO algo-  rithm.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' This helps motivate the reason for performing  the search for action sequences in the first phase.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' Results  In Table 3 we present the results of our experiments com-  paring our proposed model GNN-BC as compared to the  various baselines.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' The test programs were completely held  out during both data-driven learning phases (action sequence  search and model training).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  The results show that our model achieves strong perfor-  mance over the prior method (Autophase-RL) proposed in  (Haj-Ali et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', 2020b).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' Additionally we can see that both  the GNN model and the action sequences were needed to  achieve our final performance.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' See Figure 3 for a visualiza-  tion of the improvement in program size over the 45 steps  on some of the programs from the holdout set.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  The Oracle-All shows strong performance but requires a  large number of interactions with the compiler.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' But, this  shows that the action sequence search generalizes to new  unseen programs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' This is somewhat unsurprising given that  the compilers built-in hand tuned pass list (-Oz) works  reasonably well for most programs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  The performance of Oracle-Top-45 by itself is weak show-  ing that in order to achieve good results in a reasonable  number of passes (45) we need to leverage a general pol-  icy and search to select the most likely candidate action  sequences to evaluate.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  Both RL baselines using the original action space of single  passes Autophase-RL and GNN-RL performed poorly on  the generalization task.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' We hypothesis that this is partly  Method  Test MeanOverOz  Test GMeanOverOz  Compiler (-Oz)  0%  0  Autophase-RL  16.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='3%  0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='960  Autophase-BC  4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='2%  1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='056  GNN-RL  9.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='6%  1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='005  GNN-BC  4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='7%  1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='062  Oracle-Top-45  7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='5%  0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='992  Oracle-All  5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='8%  1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='075  Table 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' Evaluation results on held out test set averaged over all  datasets.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  due to the challenging exploration problem of the large  search space (12445).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' This is also a very hard setting for RL  because each program has its own cumulative reward upper  bound (which is unknown even for the training set).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' This  makes approximating the value function very difficult for  the baseline RL methods.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  References  Almakki, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', Izzeldin, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', Huang, Q.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', Ali, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', and  Cummins, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' Autophase v2: Towards function level  phase ordering optimization.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' In ISCA 2022 Workshop on  MLArchSys, 2022.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  Ashouri, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', Elhoushi, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', Hua, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', Wang, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', Manzoor,  M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', Chan, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', and Gao, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' Mlgoperf: An ml guided  inliner to optimize performance, 2022.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' URL https:  //arxiv.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='org/abs/2207.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='08389.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  Brauckmann, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', Goens, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', Ertel, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', and Castrillon, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  Compiler-based graph representations for deep learning  models of code.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' In CC, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' 201–211, 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  Cummins, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', Petoumenos, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', Wang, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', and Leather, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  End-to-end deep learning of optimization heuristics.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' In  26th International Conference on Parallel Architectures  and Compilation Techniques (PACT).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' IEEE, 2017.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  Cummins, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', Fisches, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', Ben-Nun, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', Hoefler, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=',  O’Boyle, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', and Leather, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' ProGraML: A Graph-  based Program Representation for Data Flow Analysis  and Compiler Optimizations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' CoRR, ICML, 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  Cummins, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', Wasti, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', Guo, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', Cui, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', Ansel, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', Gomez,  S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', Jain, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', Liu, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', Teytaud, O.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', Steiner, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' Com-  pilergym: robust, performant compiler optimization en-  vironments for ai research.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' In 2022 IEEE/ACM Interna-  tional Symposium on Code Generation and Optimization  (CGO), pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' 92–105.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' IEEE, 2022.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  Guo, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', Ren, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', Lu, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', Feng, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', Tang, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', Liu, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', Zhou,  L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', Duan, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', Svyatkovskiy, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', Fu, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' Graphcode-  Learning to compile smartly for program size reduction  bert: Pre-training code representations with data flow.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  arXiv preprint arXiv:2009.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='08366, 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  Haj-Ali, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', Ahmed, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', Willke, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', Shao, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', Asanovic,  K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', and Stoica, I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' Neurovectorizer: End-to-end vectoriza-  tion with deep reinforcement learning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' In Proceedings of  the 18th ACM/IEEE International Symposium on Code  Generation and Optimization, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' 242–255, 2020a.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  Haj-Ali, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', Huang, Q.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', Xiang, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', Moses, W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', Asanovic,  K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', Wawrzynek, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', and Stoica, I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  Autophase: Jug-  gling hls phase orderings in random forests with  deep reinforcement learning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  In Dhillon, I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', Pa-  pailiopoulos, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', and Sze, V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' (eds.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' ), Proceedings  of Machine Learning and Systems, volume 2, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  70–81, 2020b.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  URL https://proceedings.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  mlsys.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='org/paper/2020/file/  4e732ced3463d06de0ca9a15b6153677-Paper.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  pdf.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  Kulkarni, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' and Cavazos, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' Mitigating the compiler opti-  mization phase-ordering problem using machine learning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  In Proceedings of the ACM international conference on  Object oriented programming systems languages and ap-  plications, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' 147–162, 2012.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  Li, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', Tarlow, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', Brockschmidt, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', and Zemel, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  Gated graph sequence neural networks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' arXiv preprint  arXiv:1511.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='05493, 2015.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  Mialon, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', Chen, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', Selosse, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', and Mairal, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' Graphit:  Encoding graph structure in transformers, 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  Mnih, V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', Kavukcuoglu, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', Silver, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', Rusu, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', Veness,  J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', Bellemare, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', Graves, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', Riedmiller, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', Fidje-  land, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', Ostrovski, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' Human-level control  through deep reinforcement learning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' nature, 518(7540):  529–533, 2015.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  Schulman, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', Wolski, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', Dhariwal, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', Radford, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', and  Klimov, O.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' Proximal policy optimization algorithms.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  arXiv preprint arXiv:1707.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='06347, 2017.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  Srinivas, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', Laskin, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', and Abbeel, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' Curl: Contrastive  unsupervised representations for reinforcement learning,  2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  Steiner, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', Cummins, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', He, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', and Leather, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' Value  Learning for Throughput Optimization of Deep Learning  Workloads.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' In MLSys, 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  Trofin, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', Qian, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', Brevdo, E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', Lin, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', Choromanski,  K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', and Li, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' Mlgo: a machine learning guided com-  piler optimizations framework, 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  URL https:  //arxiv.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='org/abs/2101.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='04808.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  Veliˇckovi´c, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', Cucurull, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', Casanova, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', Romero, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=',  Lio, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', and Bengio, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' Graph attention networks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' arXiv  preprint arXiv:1710.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='10903, 2017.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  Wang, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' and O’Boyle, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' Machine learning in compiler  optimization.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' Proceedings of the IEEE, 106(11):1879–  1901, 2018.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  Zhou, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', Roy, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', Abdolrashidi, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', Wong, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', Ma, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=',  Xu, Q.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', Liu, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', Phothilimtha, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', Wang, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=', Goldie, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=',  et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' Transferable graph optimizers for ml compilers.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  Advances in Neural Information Processing Systems, 33:  13844–13855, 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  Learning to compile smartly for program size reduction  Figure 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' Program optimization example over many steps comparing the Autophase-RL (blue) approach with our GNN-BC (orange)  approach.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' The dashed line represents the compiler default -Oz performance and higher is better.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  benchmark://cbench-v1/diikstra  benchmark://cbench-v1/stringsearch  1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='5  1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='0  0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='5  0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='0  benchmark://cbench-v1/blowifish  benchmark://cbench-v1/stringsearch2  1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='5  1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='0  0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='5  0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='0  benchmark://cbench-vl/gsort  benchmark://cbench-v1/bitcount  1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='5  1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='0  0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='5  0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='0  benchmark://cbench-y1/rindae!' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  benchmark://cbench-v1/sha  1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='5  1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='0  0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='5  0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='0  i  0  5  10  15  20  25  30  35  40  0  5  10  15  20  25  30  35  40Learning to compile smartly for program size reduction  Dataset  Oracle-All  Oracle-Top-45  Autophase-RL  Autophase-BC  GNN-RL  GNN-BC  anghabench-v1  0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='7%/1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='011  1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='0%/0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='996  15.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='9%/0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='974  0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='1%/1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='002  2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='5%/0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='988  0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='0%/1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='003  blas-v0  2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='6%/1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='028  0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='4%/0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='997  1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='7%/0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='984  1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='2%/1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='013  1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='2%/0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='989  2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='4%/1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='026  cbench-v1  3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='5%/1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='041  2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='4%/0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='984  10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='1%/0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='925  1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='5%/1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='021  2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='4%/1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='030  2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='2%/1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='028  chstone-v0  9.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='3%/1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='106  1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='2%/1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='016  1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='3%/1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='018  7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='0%/1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='079  6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='4%/1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='071  8.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='8%/1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='101  clgen-v0  5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='4%/1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='060  3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='1%/1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='034  0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='5%/0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='998  4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='5%/1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='050  2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='2%/1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='024  5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='0%/1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='056  csmith-v0  21.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='2%/1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='320  96.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='3%/0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='851  116.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='0%/0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='954  21.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='1%/1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='318  125.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='4%/0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='994  21.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='1%/1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='320  github-v0  1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='0%/1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='011  0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='2%/1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='002  0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='1%/1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='001  0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='9%/1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='009  0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='1%/1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='002  0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='9%/1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='010  linux-v0  0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='6%/1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='007  0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='4%/0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='998  0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='5%/0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='997  0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='6%/1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='006  0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='9%/0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='996  0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='6%/1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='007  llvm-stress-v0  6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='3%/1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='087  18.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='9%/0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='885  67.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='0%/0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='731  1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='6%/1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='040  22.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='0%/0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='872  2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='1%/1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='045  mibench-v1  1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='7%/1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='020  0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='0%/1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='003  2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='8%/0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='976  0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='4%/0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='999  0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='6%/1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='008  0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='1%/1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='003  npb-v0  9.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='8%/1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='159  5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='7%/1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='085  0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='9%/1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='035  6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='0%/1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='088  4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='8%/1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='074  5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='5%/1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='085  opencv-v0  5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='2%/1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='061  1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='0%/1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='013  0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='5%/1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='007  4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='5%/1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='054  0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='7%/1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='009  4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='8%/1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='057  poj104-v1  7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='8%/1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='105  3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='9%/1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='055  17.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='5%/0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='876  5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='7%/1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='075  0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='1%/1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='014  6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='3%/1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='082  tensorflow-v0  6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='1%/1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='077  0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='2%/0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='998  0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='2%/1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='004  5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='1%/1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='063  0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='8%/1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='011  5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='9%/1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='075  Average  5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='8%/1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='075  7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='5%/0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='992  16.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='3%/0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='960  4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='2%/1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='056  9.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='6%/1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='005  4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='7%/1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='062  Table 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' Evaluation results on held out test set averaged over all datasets.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='Learning to compile smartly for program size reduction  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='Index  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='Flag  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='Index  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='Flag  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='Index  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='Flag  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='0  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='add-discriminators  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='42  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='globalsplit  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='84  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='lower-expect  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='1  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='adce  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='43  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='guard-widening  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='85  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='lower-guard-intrinsic  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='2  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='aggressive-instcombine  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='44  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='hotcoldsplit  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='86  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='lowerinvoke  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='3  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='alignment-from-assumptions  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='45  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='ipconstprop  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='87  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='lower-matrix-intrinsics  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='4  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='always-inline  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='46  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='ipsccp  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='88  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='lowerswitch  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='5  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='argpromotion  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='47  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='indvars  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='89  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='lower-widenable-condition  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='6  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='attributor  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='48  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='irce  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='90  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='memcpyopt  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='7  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='barrier  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='49  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='infer-address-spaces  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='91  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='mergefunc  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='8  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='bdce  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='50  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='inferattrs  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='92  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='mergeicmps  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='9  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='break-crit-edges  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='51  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='inject-tli-mappings  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='93  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='mldst-motion  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='10  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='simplifycfg  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='52  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='instsimplify  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='94  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='sancov  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='11  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='callsite-splitting  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='53  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='instcombine  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='95  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='name-anon-globals  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='12  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='called-value-propagation  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='54  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='instnamer  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='96  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='nary-reassociate  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='13  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='canonicalize-aliases  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='55  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='jump-threading  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='97  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='newgvn  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='14  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='consthoist  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='56  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='lcssa  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='98  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='pgo-memop-opt  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='15  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='constmerge  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='57  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='licm  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='99  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='partial-inliner  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='16  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='constprop  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='58  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='libcalls-shrinkwrap  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='100  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='partially-inline-libcalls  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='17  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='coro-cleanup  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='59  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='load-store-vectorizer  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='101  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='post-inline-ee-instrument  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='18  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='coro-early  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='60  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='loop-data-prefetch  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='102  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='functionattrs  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='19  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='coro-elide  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='61  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='loop-deletion  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='103  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='mem2reg  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='20  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='coro-split  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='62  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='loop-distribute  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='104  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='prune-eh  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='21  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='correlated-propagation  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='63  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='loop-fusion  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='105  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='reassociate  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='22  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='cross-dso-cfi  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='64  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='loop-guard-widening  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='106  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='redundant-dbg-inst-elim  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='23  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='deadargelim  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='65  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='loop-idiom  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='107  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='rpo-functionattrs  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='24  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='dce  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='66  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='loop-instsimplify  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='108  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='rewrite-statepoints-for-gc  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='25  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='die  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='67  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='loop-interchange  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='109  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='sccp  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='26  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='dse  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='68  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='loop-load-elim  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='110  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='slp-vectorizer  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='27  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='reg2mem  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='69  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='loop-predication  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='111  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='sroa  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='28  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='div-rem-pairs  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='70  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='loop-reroll  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='112  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='scalarizer  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='29  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='early-cse-memssa  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='71  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='loop-rotate  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='113  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='separate-const-offset-from-gep  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='30  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='early-cse  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='72  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='loop-simplifycfg  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='114  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='simple-loop-unswitch  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='31  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='elim-avail-extern  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='73  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='loop-simplify  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='115  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='sink  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='32  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='ee-instrument  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='74  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='loop-sink  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='116  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='speculative-execution  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='33  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='flattencfg  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='75  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='loop-reduce  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='117  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='slsr  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='34  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='float2int  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='76  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='loop-unroll-and-jam  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='118  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='strip-dead-prototypes  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='35  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='forceattrs  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='77  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='loop-unroll  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='119  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='strip-debug-declare  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='36  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='inline  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='78  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='loop-unswitch  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='120  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='strip-nondebug  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='37  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='insert-gcov-profiling  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='79  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='loop-vectorize  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='121  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='strip  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='38  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='gvn-hoist  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='80  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='loop-versioning-licm  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='122  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='tailcallelim  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='39  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='gvn  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='81  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='loop-versioning  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='123  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='mergereturn  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='40  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='globaldce  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='82  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='loweratomic  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='41  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='globalopt  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='83  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='lower-constant-intrinsics  ' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content='Table 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}
+page_content=' A list of LLVM compiler pass indices and their corresponding command line flag.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/D9E4T4oBgHgl3EQffA2Y/content/2301.05104v1.pdf'}