-
In low-depth pools, large swaps can incur significant slippage. One solution is to split the swap into a series of ticks, executed however often, so that each individual tick is small compared to the …
-
| | |
| --- | --- |
| Bugzilla Link | [11280](https://llvm.org/bz11280) |
| Version | trunk |
| OS | Linux |
| Reporter | LLVM Bugzilla Contributor |
| CC | @DougGregor,@efriedma-quic,@rjmccall |
#…
-
When creating a pull request, it would be nice to be able to swap the **base** and **compare** dropdowns in one click. Right now, it takes
For example, this icon could be a link:
![screen shot 2016…
-
| | |
|--------------------|----|
| Bugzilla Link | [PR11280](https://bugs.llvm.org/show_bug.cgi?id=11280) |
| Status | REOPENED |
| Importance | P norma…
-
Right now, a DCAS will have two separate traces, which will potentially involve a crash in between. This may be okay for now, as stores won't show up in the shadow region unless it is either flushed a…
-
# Problem
Currently request-based RMA (§11.3.5) is defined for Put, Get, Accumulate, and Get-Accumulate. However, it is not defined for Fetch-and-Op and Compare-and-Swap.
# Proposal
Add `MPI_…
-
Currently torchao QAT has two APIs, [tensor subclasses](https://github.com/pytorch/ao/blob/a4221df5e10ff8c33854f964fe6b4e00abfbe542/torchao/quantization/prototype/qat/api.py#L41) and [module swap](htt…
-
Microblaze toolchain's atomic compare-and-swap returns the inverse boolean value causing at least `ungetc` implementation in `picolibc` to hang.
The issue was fixed in Xilinx's gcc via the below patc…
-
https://github.com/koonwen/flat-combiner/blob/4a9a2df618cfcc08880f69a09694c052166a67ba/src/fc_generic/fc_generic_domains.ml#L64
-
Using __sync intrinsics by default
results into an error.
Changing `-march` solves the problem,
but things should work by default.