Files
Jesse Gross 2e036e7cdf mlx, mlxrunner: move the MLX engine out of x/
The MLX runner is the only Go inference runner left and is no longer
experimental, so its packages leave x/. The bindings become a top-level
mlx package beside the carried patches in mlx/compat, mirroring how
llama/ holds the llama.cpp integration, and the runner becomes mlxrunner
with the architectures nested under the package they implement.
Subpackages move with their parent unless listed.

  x/mlxrunner/mlx            mlx
  x/internal/mlxthread       mlx/mlxthread
  x/internal/mlxthreadtest   mlx/mlxthread/mlxthreadtest
  x/internal/mlxtest         mlx/mlxtest
  x/quant                    mlx/quant
  mlx/compat/*.patch         mlx/compat/mlx-c   (MLX patches go in mlx/compat/mlx)
  x/mlxrunner                mlxrunner
  x/models/nn                mlxrunner/nn
  x/models/<arch>            mlxrunner/model/<arch>
  x/mlxrunner/imports.go     mlxrunner/model/architectures   (new package)
  x/create                   create
  x/safetensors              fs/safetensors
  x/tokenizer                mlxrunner/tokenizer

Every package keeps its name, so the Go changes are the import path
rewrites the moves force, and the CMake, Dockerfile, CI cache keys, drift
check and Darwin payload script follow the new paths. Four edits are not
paths: the runner's blank architecture imports become the package
mlxrunner/model/architectures, so the list to extend for a new model sits
beside the architecture directories; a depguard rule keeps the two test
harnesses out of non-test code, as the x/internal placement used to; the
CI change filter's two entries for the long-deleted x/imagegen/mlx now
name the bindings' CMake project and the carried patches, so a change to
either builds the payload; and the tokenizer parity test reads its
fixtures from its own testdata instead of walking out of x/.

x/server and x/imagegen/manifest stay for the next two commits.
2026-09-16 14:06:08 -07:00

110 lines
2.2 KiB
Go

package mlxrunner
import (
"context"
"log/slog"
"sync"
"time"
)
const statusMemoryRefreshWait = 50 * time.Millisecond
type statusMemoryRefreshFunc func() (uint64, error)
// statusMemoryCache keeps health checks from depending synchronously on the
// serialized MLX worker while still refreshing memory telemetry opportunistically.
type statusMemoryCache struct {
done <-chan struct{}
wait time.Duration
refresh statusMemoryRefreshFunc
mu sync.Mutex
memory uint64
refreshedAt time.Time
inFlight chan struct{}
}
func newStatusMemoryCache(ctx context.Context, memory uint64, refreshedAt time.Time, wait time.Duration, refresh statusMemoryRefreshFunc) *statusMemoryCache {
return &statusMemoryCache{
done: ctx.Done(),
wait: wait,
refresh: refresh,
memory: memory,
refreshedAt: refreshedAt,
}
}
func (c *statusMemoryCache) Memory() uint64 {
done := c.startRefresh()
if c.wait <= 0 {
<-done
memory, _ := c.snapshot()
return memory
}
timer := time.NewTimer(c.wait)
defer timer.Stop()
select {
case <-done:
case <-timer.C:
memory, refreshedAt := c.snapshot()
if refreshedAt.IsZero() {
slog.Debug("using cached MLX memory status before first refresh")
} else {
slog.Debug("using cached MLX memory status", "stale", time.Since(refreshedAt))
}
return memory
case <-c.done:
}
memory, _ := c.snapshot()
return memory
}
func (c *statusMemoryCache) startRefresh() chan struct{} {
c.mu.Lock()
if c.inFlight != nil {
done := c.inFlight
c.mu.Unlock()
return done
}
refreshDone := make(chan struct{})
c.inFlight = refreshDone
refresh := c.refresh
lifecycleDone := c.done
c.mu.Unlock()
go func() {
memory, err := refresh()
now := time.Now()
c.mu.Lock()
defer c.mu.Unlock()
defer close(refreshDone)
if err != nil {
select {
case <-lifecycleDone:
default:
slog.Debug("failed to refresh MLX memory status", "error", err)
}
c.inFlight = nil
return
}
c.memory = memory
c.refreshedAt = now
c.inFlight = nil
}()
return refreshDone
}
func (c *statusMemoryCache) snapshot() (uint64, time.Time) {
c.mu.Lock()
defer c.mu.Unlock()
return c.memory, c.refreshedAt
}