-
Minotaur: A SIMD-Oriented Synthesizing Superoptimizer
Authors:
Zhengyang Liu,
Stefan Mada,
John Regehr
Abstract:
Minotaur is a superoptimizer for LLVM's intermediate representation that focuses on integer SIMD instructions, both portable and specific to x86-64. We created it to attack problems in finding missing peephole optimizations for SIMD instructions-this is challenging because there are many such instructions and they can be semantically complex. Minotaur runs a hybrid synthesis algorithm where instru…
▽ More
Minotaur is a superoptimizer for LLVM's intermediate representation that focuses on integer SIMD instructions, both portable and specific to x86-64. We created it to attack problems in finding missing peephole optimizations for SIMD instructions-this is challenging because there are many such instructions and they can be semantically complex. Minotaur runs a hybrid synthesis algorithm where instructions are enumerated concretely, but literal constants are generated by the solver. We use Alive2 as a verification engine; to do this we modified it to support synthesis and also to support a large subset of Intel's vector instruction sets (SSE, AVX, AVX2, and AVX-512). Minotaur finds many profitable optimizations that are missing from LLVM. It achieves limited speedups on the integer parts of SPEC CPU2017, around 1.3%, and it speeds up the test suite for the libYUV library by 2.2%, on average, and by 1.64x maximum, when targeting an Intel Cascade Lake processor.
△ Less
Submitted 12 July, 2023; v1 submitted 31 May, 2023;
originally announced June 2023.
-
Future Directions for Optimizing Compilers
Authors:
Nuno P. Lopes,
John Regehr
Abstract:
As software becomes larger, programming languages become higher-level, and processors continue to fail to be clocked faster, we'll increasingly require compilers to reduce code bloat, eliminate abstraction penalties, and exploit interesting instruction sets. At the same time, compiler execution time must not increase too much and also compilers should never produce the wrong output. This paper exa…
▽ More
As software becomes larger, programming languages become higher-level, and processors continue to fail to be clocked faster, we'll increasingly require compilers to reduce code bloat, eliminate abstraction penalties, and exploit interesting instruction sets. At the same time, compiler execution time must not increase too much and also compilers should never produce the wrong output. This paper examines the problem of making optimizing compilers faster, less buggy, and more capable of generating high-quality output.
△ Less
Submitted 6 September, 2018;
originally announced September 2018.
-
Souper: A Synthesizing Superoptimizer
Authors:
Raimondas Sasnauskas,
Yang Chen,
Peter Collingbourne,
Jeroen Ketema,
Gratian Lup,
Jubi Taneja,
John Regehr
Abstract:
If we can automatically derive compiler optimizations, we might be able to sidestep some of the substantial engineering challenges involved in creating and maintaining a high-quality compiler. We developed Souper, a synthesizing superoptimizer, to see how far these ideas might be pushed in the context of LLVM. Along the way, we discovered that Souper's intermediate representation was sufficiently…
▽ More
If we can automatically derive compiler optimizations, we might be able to sidestep some of the substantial engineering challenges involved in creating and maintaining a high-quality compiler. We developed Souper, a synthesizing superoptimizer, to see how far these ideas might be pushed in the context of LLVM. Along the way, we discovered that Souper's intermediate representation was sufficiently similar to the one in Microsoft Visual C++ that we applied Souper to that compiler as well. Ship**, or about-to-ship, versions of both compilers contain optimizations suggested by Souper but implemented by hand. Alternately, when Souper is used as a fully automated optimization pass it compiles a Clang compiler binary that is about 3 MB (4.4%) smaller than the one compiled by LLVM.
△ Less
Submitted 5 April, 2018; v1 submitted 13 November, 2017;
originally announced November 2017.
