Skip to content

Conversation

@bibrak
Copy link
Contributor

@bibrak bibrak commented Nov 6, 2025

Enabled multi-gpu madvise for SVM (System Shared Memory) allocated memory.

@bibrak
Copy link
Contributor Author

bibrak commented Nov 6, 2025

Do not Merge!

Implements madvise SVM allocated memory for multi-gpu

Signed-off-by: Chandio, Bibrak Qamar <bibrak.qamar.chandio@intel.com>
@bibrak bibrak force-pushed the multi-gpu-madvise branch from 917fc00 to 1fe058f Compare November 6, 2025 20:12
Implements madvise SVM allocated memory for multi-gpu

Signed-off-by: Chandio, Bibrak Qamar <bibrak.qamar.chandio@intel.com>
// Apply the shared system USM IOCTL to all the VMs of the device
std::vector<uint32_t> vmIds;
vmIds.reserve(subDeviceIds.size());
for (auto subDeviceId : subDeviceIds) {
Copy link
Contributor

Choose a reason for hiding this comment

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

this would be error prone that we pass targetDeviceDrm and subdevices seperatly, can you pass targetDevice and wihin this function extract sub device ids , this way the potential mismatch would be not possible

Copy link
Contributor Author

Choose a reason for hiding this comment

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

Done!!!

@bibrak bibrak marked this pull request as draft November 7, 2025 17:05
Implements madvise SVM allocated memory for multi-gpu

Signed-off-by: Chandio, Bibrak Qamar <bibrak.qamar.chandio@intel.com>
Signed-off-by: Falkowski, John <john.falkowski@intel.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants