Commits

Branch selector

 master

User selector

opti-mix

Datepicker

All time

Commit History

Commits on May 10, 2019

• Hoist QuantizationProfile instructions right after the last node that updates their inputs to shorten lifetimes of buffers (pytorch#2698)

  Show description for 3abdf8d
  

  opti-mix

  authored and

  facebook-github-bot

  committed
  3abdf8d

  Copy full SHA for 3abdf8d

• Let backends decide if they support kernel stacking for specific instructions

  Show description for 1e067bf
  

  opti-mix

  authored and

  facebook-github-bot

  committed
  1e067bf

  Copy full SHA for 1e067bf

Commits on May 3, 2019

• [LLVMIRCodeGen] Make LLVMIRCodeGen more cross-compilation friendly

  Show description for 7566736
  opti-mix

  committed
  7566736

  Copy full SHA for 7566736

Commits on Apr 30, 2019

• [ir-optimizer] Add option -instrument-debug-only to instrument only instructions from the provided list

  Show description for a691049
  opti-mix

  committed
  a691049

  Copy full SHA for a691049

Commits on Apr 29, 2019

• [LLVMIRCodeGen] Make it possible for backends to provide custom logic to decide which calls are eligible for specialization

  opti-mix

  committed
  21f67d5

  Copy full SHA for 21f67d5

• [LLVMIRCodeGen] Provide APIs for setting target, arch and cpu properties of LLVMBackend

  Show description for d619859
  opti-mix

  committed
  d619859

  Copy full SHA for d619859

• [LLVMIRCodeGen] Make it possible for backends to provide custom logic to decide which symbols need to be preserved during the LLVM's internalizeModule pass

  Show description for 6d4f9c2
  opti-mix

  committed
  6d4f9c2

  Copy full SHA for 6d4f9c2

• Allow NodeInfo objects to have names

  Show description for b2e8f3d
  opti-mix

  committed
  b2e8f3d

  Copy full SHA for b2e8f3d

Commits on Apr 25, 2019

• [OpenCL] Remove the spurious dependency on CPURuntime

  Show description for 974694b
  opti-mix

  committed
  974694b

  Copy full SHA for 974694b

Commits on Apr 17, 2019

• [LLVMIRCodeGen] Initialize all asm parsers

  Show description for 9d8dae9
  opti-mix

  committed
  9d8dae9

  Copy full SHA for 9d8dae9

Commits on Apr 7, 2019

• Fix some spelling errors. NFC.

  opti-mix

  committed
  0608054

  Copy full SHA for 0608054

Commits on Apr 4, 2019

• [LLVMIRCodeGen] Support using an external LLVM compiler for converting LLVM bitcode files into object files

  Show description for 172bbab
  opti-mix

  committed
  172bbab

  Copy full SHA for 172bbab

Commits on Apr 2, 2019

• Take NodeValue instead of Node * as arguments for createFullyConnected functions

  Show description for da1e9f6
  opti-mix

  committed
  da1e9f6

  Copy full SHA for da1e9f6

Commits on Mar 28, 2019

• [debugging] Make -debug-glow-only option take a list of comma-separated debug types names

  Show description for 8821bce
  opti-mix

  committed
  8821bce

  Copy full SHA for 8821bce

Commits on Mar 20, 2019

• Make more methods in LLVMBackend and LLVMIRGen virtual, so that backends can provide custom implementations

  opti-mix

  committed
  03fd7fd

  Copy full SHA for 03fd7fd

• Move command-line options for selecting LLVM target, arch, cpu and cpu features into a dedicated include file

  Show description for a437cf5
  opti-mix

  committed
  a437cf5

  Copy full SHA for a437cf5

• Move AllocationsInfo.h into the shared includes directory

  Show description for c40c7e8
  opti-mix

  committed
  c40c7e8

  Copy full SHA for c40c7e8

Commits on Mar 18, 2019

• Make createCall a method in LLVMIRGen

  Show description for 50b3e8c
  opti-mix

  committed
  50b3e8c

  Copy full SHA for 50b3e8c

• [function-specializer] Do not try to specialize function declarations

  opti-mix

  committed
  acdfd24

  Copy full SHA for acdfd24

Commits on Mar 6, 2019

• [function-specializer] Fix a bug related to the removal of non-specialized calls

  Show description for 8d55878
  opti-mix

  committed
  8d55878

  Copy full SHA for 8d55878

Commits on Mar 1, 2019

• Move the re-usable pieces of the CPUBackend into a dedicated LLVMIRCodeGen library

  Show description for 43106fc
  opti-mix

  committed
  43106fc

  Copy full SHA for 43106fc

Commits on Feb 26, 2019

• Add some internal caches used by IDEs to .gitignore

  opti-mix

  committed
  fcea7a6

  Copy full SHA for fcea7a6

Commits on Jan 17, 2019

• Add a small support function strFormat to create std::strings using printf-like formatting

  Show description for aea4510
  opti-mix

  committed
  aea4510

  Copy full SHA for aea4510

Commits on Nov 9, 2018

• [jit] Restore the functionality for dumping JIT symbols that got lost during the LLVM6 to LLVM7 transition

  opti-mix

  committed
  8b489be

  Copy full SHA for 8b489be

Commits on Oct 26, 2018

• Add a getter ExecutionEngine::getCompiledFunction

  Show description for 750e63f
  opti-mix

  committed
  750e63f

  Copy full SHA for 750e63f

• Run a round of DCE before any other graph optimizations

  Show description for 17ef797
  opti-mix

  committed
  17ef797

  Copy full SHA for 17ef797

Commits on Oct 25, 2018

• Mark quantize, dequantize and rescale_quantized as data-parallel operations

  Show description for 65e50cc
  opti-mix

  committed
  65e50cc

  Copy full SHA for 65e50cc

Commits on Oct 24, 2018

• [Backend][Interpreter] Add support for quantize/dequantize with fp16 operands

  opti-mix

  committed
  8bd7235

  Copy full SHA for 8bd7235

• Add a createDequantize method that takes an explicit output type argument

  Show description for d65d9db
  opti-mix

  committed
  d65d9db

  Copy full SHA for d65d9db

• Relax the type verification constraints for the operands of quantize and dequantize instructions

  Show description for 0aef28f
  opti-mix

  committed
  0aef28f

  Copy full SHA for 0aef28f

• Support type predicates when auto-generating verifier functions

  Show description for 75f154f
  opti-mix

  committed
  75f154f

  Copy full SHA for 75f154f

• Optimize away conversions between the same types

  opti-mix

  committed
  892e211

  Copy full SHA for 892e211

• Support quantization involving FP16

  opti-mix

  committed
  2d978b3

  Copy full SHA for 2d978b3

• Use isFPType() instead of checking for FP32 in a couple of asserts

  opti-mix

  committed
  984f62b

  Copy full SHA for 984f62b

• Add the isFPType predicate to check if the element type of a tensor is an FP type

  opti-mix

  committed
  c799272

  Copy full SHA for c799272

Pagination

Previous

Next