- Now using 10x better GCD semaphores
- We now generate Linux-like thread ids
- We now use fast system clock / sleep libraries
- The APE M1 loader now generates Linux-like stacks
llama.com can now load weights that use the new file format which was
introduced a few weeks ago. Note that, unlike llama.cpp, we will keep
support for old file formats in our tool so you don't need to convert
your weights when the upstream project makes breaking changes. Please
note that using ggjt v3 does make avx2 inference go 5% faster for me.
This change progresses our AARCH64 support:
- The AARCH64 build and tests are now passing
- Add 128-bit floating-point support to printf()
- Fix clone() so it initializes cosmo's x28 TLS register
- Fix TLS memory layout issue with aarch64 _Alignas vars
- Revamp microbenchmarking tools so they work on aarch64
- Make some subtle improvements to aarch64 crash reporting
- Make kisdangerous() memory checks more accurate on aarch64
- Remove sys_open() since it's not available on Linux AARCH64
This change makes general improvements to Cosmo and Redbean:
- Introduce GetHostIsa() function in Redbean
- You can now feature check using pledge(0, 0)
- You can now feature check using unveil("",0)
- Refactor some more x86-specific asm comments
- Refactor and write docs for some libm functions
- Make the mmap() API behave more similar to Linux
- Fix WIFSIGNALED() which wrongly returned true for zero
- Rename some obscure cosmo keywords from noFOO to dontFOO
There's a new program named ape/ape-m1.c which will be used to build an
embeddable binary that can load ape and elf executables. The support is
mostly working so far, but still chasing down ABI issues.
- Fix UX issues with llama.com
- Do housekeeping on libm code
- Add more vectorization to GGML
- Get GGJT quantizer programs working well
- Have the quantizer keep the output layer as f16c
- Prefetching improves performance 15% if you use fewer threads
- Perform some housekeeping on scalar math function code
- Import ARM's Optimized Routines for SIMD string processing
- Upgrade to latest Chromium zlib and enable more SIMD optimizations
- Introduce epoll_pwait()
- Rewrite -ftrapv and ffs() libraries in C code
- Use more FreeBSD code in math function library
- Get significantly more tests passing on qemu-aarch64
- Fix many Musl long double functions that were broken on AARCH64
make -j8 o//third_party/radpajama/radpajama.com
make -j8 o//third_party/radpajama/radpajama-chat.com
This change gets the radpajama.mk config working. This package depends
on THIRD_PARTY_GGML but it's configured to call ggjt_v1(), so that the
library will provide the old quantizers. The ggml_quantize_chunk() API
will now dispatch to older quantizers based on the configured version.
This change makes quantized models (e.g. q4_0) go 10% faster on Macs
however doesn't offer much improvement for Intel PC hardware.
This change syncs llama.cpp 699b1ad7fe6f7b9e41d3cb41e61a8cc3ea5fc6b5
which recently made a breaking change to nearly all its file formats
without any migration. Since that'll break hundreds upon hundreds of
models on websites like HuggingFace llama.com will support both file
formats because llama.com will never ever break the GGJT file format
- Utilities like pledge.com now build
- kprintf() will no longer balk at 48-bit addresses
- There's a new aarch64-dbg build mode that should work
- gc() and defer() are mostly pacified; avoid using them on aarch64
- THIRD_PART_STB now has Arm Neon intrinsics for fast image handling
Example use case for JSON completion:
$ m=opt
$ make -j16 m=$m o/$m/third_party/ggml/llama.com
$ o/$m/third_party/ggml/llama.com -m llama.bin -p '{"key": "life", "val": ' -r '}'
42}
This provides better control. More sophisticated facilities for
controlling text generation will be provided soon enough.
We can once again create 2mb statically-linked Python binaries:
$ make -j8 m=tiny o/tiny/examples/pyapp/pyapp.com
$ ls -hal o/tiny/examples/pyapp/pyapp.com
-rwxr-xr-x 1 jart jart 2.1M May 1 14:04 o/tiny/examples/pyapp/pyapp.com
$ o/tiny/examples/pyapp/pyapp.com
cosmopolitan is cool!
The regression was caused by Python thread support in b15f9eb58
- Introduce -v and --verbose flags
- Don't print stats / diagnostics unless -v is passed
- Reduce --top_p default from 0.95 to 0.70
- Change --reverse-prompt to no longer imply --interactive
- Permit --reverse-prompt specifying custom EOS if non-interactive
Python threads are now generally working, however some parts of Python's
regression tests for threads are flaky. This is possibly due to needing
more locking primitives in Cosmo's IO system call wrappers, e.g. close.
make o//third_party/python/Lib/test/test_threading.py.runs
See #747
- enable WITH_THREAD and _POSIX_THREADS
- add headers everywhere
- breaks only two tests (faulthandler and signal)
- disabled terminal completion because it causes segfaults for some
reason (probably could not get the current thread)
- Improve compatibility with Blink virtual machine
- Add non-POSIX APIs for joining threads and signal masks
- Never ever use anything except 32-bit integers for atomics
- Add some `#undef` statements to workaround `ctags` problems
libc/integral changes aren't checked in the build dependency, due to
being explicitly listed in .UNVEIL, which is how this breakage ended
up accidentally slipping through the cracks.
- clock_nanosleep() is now much faster on OpenBSD and NetBSD
- Thread joining is now much faster on NetBSD
- FreeBSD timestamps are now more accurate
- Thread spawning now goes faster on XNU
- Clean up the clone() code
- Clean up sigaction() code
- Add a port scanner example
- Introduce a ParseCidr() API
- Clean up our futex abstraction code
- Fix a harmless integer overflow in ParseIp()
- Use kernel semaphores on NetBSD to make threads much faster
- Exhaustively document cancellation points
- Rename SIGCANCEL to SIGTHR just like BSDs
- Further improve POSIX thread cancellations
- Ensure asynchronous cancellations work correctly
- Elevate the quality of getrandom() and getentropy()
- Make futexes cancel correctly on OpenBSD 6.x and 7.x
- Add reboot.com and shutdown.com to examples directory
- Remove underscore prefix from awesome timespec_*() APIs
- Create assertions that help verify our cancellation points
- Remove bad timespec APIs (cmp generalizes eq/ne/gt/gte/lt/lte)
This change makes some miracle modifications to the System Five system
call support, which lets us have safe, correct, and atomic handling of
thread cancellations. It all turned out to be cheaper than anticipated
because it wasn't necessary to modify the system call veneers. We were
able to encode the cancellability of each system call into the magnums
found in libc/sysv/syscalls.sh. Since cancellations are so waq, we are
also supporting a lovely Musl Libc mask feature for raising ECANCELED.
