Skip to content

Detect TPU using PCI devices #659

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

Merged
merged 2 commits into from
Jan 30, 2025
Merged

Detect TPU using PCI devices #659

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
338aac5
Detect TPU using PCI devices
Pangoraw Jan 30, 2025
dd39fdf
add source
Pangoraw Jan 30, 2025
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
39 changes: 39 additions & 0 deletions src/Devices.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -23,3 +23,42 @@ function addressable_devices(client::XLA.Client=XLA.default_backend[])
ndevices = XLA.ClientNumAddressableDevices(client)
return [XLA.ClientGetAddressableDevice(client, i - 1) for i in 1:ndevices]
end

# https://github.com/jax-ml/jax/blob/152099ee0ef31119f16f4c2dac50d84fcb1575ef/jax/_src/hardware_utils.py#L19-L55
const _GOOGLE_PCI_VENDOR_ID = "0x1ae0"
const _TPU_PCI_DEVICE_IDS = (
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Can you link the source where this comes from?

# TPU v2, v3
"0x0027",
# No public name (plc)
"0x0056",
# TPU v4
"0x005e",
# TPU v5p
"0x0062",
# TPU v5e
"0x0063",
# TPU v6e
"0x006f",
)

function has_tpu()
Sys.islinux() || return false

devices_dir = "/sys/bus/pci/devices/"
isdir(devices_dir) || return false

try
for path in readdir(devices_dir; join=true, sort=false)
if trim(read(joinpath(path, "vendor"), String)) == _GOOGLE_PCI_VENDOR_ID &&
trim(read(joinpath(path, "device"), String)) in _TPU_PCI_DEVICE_IDS
return true
end
end
catch ex
@warn "failed to query PCI device information" maxlog = 1 exception = (
ex, catch_backtrace()
)
end

return false
end
2 changes: 1 addition & 1 deletion src/XLA.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -148,7 +148,7 @@ function __init__()
end

@static if !Sys.isapple()
if isfile("/usr/lib/libtpu.so")
if Reactant.has_tpu()
dataset_dir = @get_scratch!("libtpu")
if !isfile(dataset_dir * "/libtpu.so")
Downloads.download(
Expand Down
Loading