Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

support skip old architecture version GPU settings time slice #58

Open
wants to merge 1 commit into
base: main
Choose a base branch
from
Open

support skip old architecture version GPU settings time slice #58

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
30 changes: 30 additions & 0 deletions cmd/nvidia-dra-plugin/sharing.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -22,6 +22,7 @@ import (
"fmt"
"os"
"os/exec"
"strconv"
"strings"
"text/template"
"time"
Expand Down Expand Up @@ -102,6 +103,17 @@ func (t *TimeSlicingManager) SetTimeSlice(devices *PreparedDevices, config *shar
return fmt.Errorf("setting a TimeSlice duration on MIG devices is unsupported")
}

for _, gpu := range devices.Gpu.Devices {
err, isSupportTimeSlice := detectSupportTimeSliceByCudaComputeCapability(gpu.cudaComputeCapability)
if err != nil {
return fmt.Errorf("failed to detectSupportTimeSliceByCudaComputeCapability : %w", err)
}
if !isSupportTimeSlice {
klog.InfoS("the current card does not support setting time slices and will be ignored.", "arch", gpu.architecture, "uuid", gpu.uuid, "cudaComputeCapability", gpu.cudaComputeCapability)
return fmt.Errorf("setting a TimeSlice duration on devices uuid=%v is unsupported", gpu.uuid)
}
}

timeSlice := sharing.DefaultTimeSlice
if config != nil && config.TimeSlice != nil {
timeSlice = *config.TimeSlice
Expand Down Expand Up @@ -390,3 +402,21 @@ func (m *MpsControlDaemon) Stop(ctx context.Context) error {

return nil
}

// detactSupportTimeSliceByArch Determine whether the architecture series
// supports setting time slices based on the gpu cudaComputeCapability.
func detectSupportTimeSliceByCudaComputeCapability(cudaComputeCapability string) (error, bool) {
// ref https://github.com/NVIDIA/k8s-dra-driver/pull/58#discussion_r1469338562
// we believe time-slicing is available on Volta+ architectures, so the check would simply be cudaComputeCapability >= 7.0
// by https://github.com/NVIDIA/go-nvlib/blob/main/pkg/nvlib/device/device.go#L149, We know that cuda major and minor versions are concatenated through `.` .

cudaVersion := strings.Split(cudaComputeCapability, ".")
major, err := strconv.Atoi(cudaVersion[0])
if err != nil {
return fmt.Errorf("error to get cudaComputeCapability major version %v", cudaComputeCapability), false
}
if major >= 7 {
return nil, true
}
return nil, false
}