Commits

Breadcrumbs

History for

onnix

User selector

All users

All time

Commit History

Commits on Oct 17, 2025

• Resolve #75

  pizlonator

  committed
  b0db5dc

  Copy full SHA for b0db5dc

Commits on Oct 12, 2025

• Resolve #74.

  pizlonator

  committed
  9212f38

  Copy full SHA for 9212f38

Commits on Oct 11, 2025

• Make OpenSSL use the assembly implementations of crypto so that we have higher assurance regarding timing side channels.

  pizlonator

  committed
  9323262

  Copy full SHA for 9323262

Commits on Oct 7, 2025

• Resolve #72

  pizlonator

  committed
  9691c2b

  Copy full SHA for 9691c2b

Commits on Oct 5, 2025

• Resolve #69.

  pizlonator

  committed
  dc63cd5

  Copy full SHA for dc63cd5

• Resolve issue #70

  pizlonator

  committed
  e6ccc1d

  Copy full SHA for e6ccc1d

Commits on Oct 4, 2025

• Add some more primitives that are necessary for building OpenSSL with its assembly-based crypto implementations. This includes a zcheck primitive so that when we expose the assembly functions to Fi…

  Show description for 7bfaaad
  pizlonator

  committed
  7bfaaad

  Copy full SHA for 7bfaaad

• Add limited support for calling to Yolo-land from Fil-C. This is Fil-C's first escape hatch. The intended use case is to allow OpenSSL to use its constant time crypto assembly code.

  pizlonator

  committed
  e5e8e0b

  Copy full SHA for e5e8e0b

Commits on Sep 15, 2025

• Fixed a bunch of cases of obvious UB causing LLVM to optimize an operation that should have panicked into some kind of no-op. Resolves issue #56.

  pizlonator

  committed
  d950cf0

  Copy full SHA for d950cf0

Commits on Sep 13, 2025

• Add an escape analysis to FilPizlonator that turns local variables that don't escape into stack allocations. Some of those stack allocations (the ones that are not random accessed) are laid out in …

  Show description for 7092c7b
  pizlonator

  committed
  7092c7b

  Copy full SHA for 7092c7b

Commits on Sep 8, 2025

• Use builtin memmove instead of my own janky memmove implementation. This is perf neutral on PizBench9016. Also fix typos in comments in stdfil.h

  pizlonator

  committed
  035d3dd

  Copy full SHA for 035d3dd

Commits on Sep 6, 2025

• Add a useful assert

  pizlonator

  committed
  69e9fa7

  Copy full SHA for 69e9fa7

Commits on Sep 1, 2025

• Correctly support comdat/linkonce/odr. Specifically, drop the ODR flags since they are not safe. And, put the constellation of internal details behind a symbol in the same comdat as the pizlonated_…

  Show description for 8681267
  pizlonator

  committed
  8681267

  Copy full SHA for 8681267

Commits on Aug 22, 2025

• Rebase Fil-C to clang/llvm 20.1.8

  pizlonator

  committed
  d292f5d

  Copy full SHA for d292f5d

Commits on Aug 12, 2025

• Make all uses of memcpy preserve capabilities. Added a few tests to illustrate the issue. 1% regression on PizBench9013P, mostly due to a 30% regression on xz compression. It's likely this regressi…

  Show description for fb6763c
  pizlonator

  committed
  fb6763c

  Copy full SHA for fb6763c

Commits on Aug 8, 2025

• Extend memmove inlining limit to 40

  pizlonator

  committed
  9939cda

  Copy full SHA for 9939cda

Commits on Aug 7, 2025

• Inline sufficiently small memmoves. 3% faster on PizBench9012P.

  pizlonator

  committed
  bf1f582

  Copy full SHA for bf1f582

Commits on Aug 1, 2025

• Add missing cstdint include to llvm/lib/Target/X86/MCTargetDesc/X86MCTargetDesc.h

  wmww

  committed
  28a2a8c

  Copy full SHA for 28a2a8c

• [ADT] Add `<cstdint>` to SmallVector (#101761)

  Show description for 8338806
  

  thesamesam

  authored and

  wmww

  committed
  8338806

  Copy full SHA for 8338806

Commits on Jul 30, 2025

• Make backwards propagation of checks work across sufficiently harmless calls. In the process, I realized that the ASSERT_BOUNDS nonsense I previously added to the runtime is wrong, since freeing an…

  Show description for d3dafff
  pizlonator

  committed
  d3dafff

  Copy full SHA for d3dafff

Commits on Jul 29, 2025

• First step of inlining zmemmove_union: inline the bounds checks that it does. Also add a bunch of tests specifically for zmemmove_union, and expand the testing of memmoves generally.

  pizlonator

  committed
  dea49a5

  Copy full SHA for dea49a5

Commits on Jul 27, 2025

• Make the handling of ensureaux a bit more precise in the AI. 0.5% speedup on PizBench9012P.

  pizlonator

  committed
  6021c7c

  Copy full SHA for 6021c7c

Commits on Jul 24, 2025

• Short-circuit calls to memmove, allocation, and freeing. 2.6% speed-up on PizBench9011P.

  pizlonator

  committed
  203217b

  Copy full SHA for 203217b

Commits on Jul 23, 2025

• Fold the size class lookup and inline part of the allocation path. 1.2% speed-up on PizBench9011P. 35% speed-up on reversecomplement, 9% speed-up on simdutf.

  pizlonator

  committed
  9c4881b

  Copy full SHA for 9c4881b

Commits on Jul 18, 2025

• Make it possible to call setjmp in a loop and then jump to all of the buffers created by that loop. Also add more tests of threads and timers. Also make weak functionality treat freed objects as de…

  Show description for 96a4384
  pizlonator

  committed
  96a4384

  Copy full SHA for 96a4384

Commits on Jul 12, 2025

• Fix typo

  pizlonator

  committed
  c27b84e

  Copy full SHA for c27b84e

• Fix bug in backwards propagation pass

  pizlonator

  committed
  7504541

  Copy full SHA for 7504541

Commits on Jul 7, 2025

• Remove changes to clang and llvm that are not necessary. These changes were needed back when I used previous capability models (PLUT, SideCap, and MonoCap), all of which enlarged pointers. But Invi…

  Show description for 7b9ae97
  pizlonator

  committed
  7b9ae97

  Copy full SHA for 7b9ae97

Commits on Jun 5, 2025

• Fix almost all libffi test failures by getting rid of the autopromotion of argument types.

  pizlonator

  committed
  bf29709

  Copy full SHA for bf29709

Commits on Jun 1, 2025

• Fix issue #7. By-val struct arguments are now copied, not aliased. Also, va_arg now works with structs. No perf change on PizBench9011P.

  pizlonator

  committed
  17e0225

  Copy full SHA for 17e0225

Commits on May 26, 2025

• Make the closure API a bit more powerful and add a lot more tests.

  pizlonator

  committed
  57c71ef

  Copy full SHA for 57c71ef

Commits on May 25, 2025

• Add closure support. This allows you to create a function packaged with some data, and that function can retrieve that data when called. Needed for supporting libffi's closure API. 0.4% regression …

  Show description for c7d2592
  pizlonator

  committed
  c7d2592

  Copy full SHA for c7d2592

Commits on May 16, 2025

• Add initial support for weak maps.

  pizlonator

  committed
  37889dc

  Copy full SHA for 37889dc

Commits on May 5, 2025

• Add weak references to FUGC and expose them via stdfil.h API.

  pizlonator

  committed
  3c90f1d

  Copy full SHA for 3c90f1d

Commits on Apr 27, 2025

• Add get/set_rounding intrinsics.

  pizlonator

  committed
  9fab3e1

  Copy full SHA for 9fab3e1