masahi commented on a change in pull request #7833:
URL: https://github.com/apache/tvm/pull/7833#discussion_r612790307
##########
File path: src/runtime/vulkan/vulkan.cc
##########
@@ -60,35 +91,21 @@ class VulkanThreadEntry {
pool.reset();
streams_.clear();
for (const auto& kv : staging_buffers_) {
- if (!kv.second) {
- continue;
- }
- auto& buf = *(kv.second);
- if (buf.host_addr != nullptr) {
- vkUnmapMemory(buf.device, buf.memory);
- }
- if (buf.memory != VK_NULL_HANDLE) {
- vkFreeMemory(buf.device, buf.memory, nullptr);
- }
- if (buf.buffer != VK_NULL_HANDLE) {
- vkDestroyBuffer(buf.device, buf.buffer, nullptr);
- }
+ DeleteHostVisibleBuffer(kv.second.get());
Review comment:
When allocating a buffer, I only have GPU device ID. But I need a CPU ID
for sync.
Sync on staging buffer has a CPU ID info available to it because it is one
of arguments to `CopyDataFromTo`. Note that the existing code that does
`Synchronize` on staging buffer does sync on CPU, below `dev_from` is CPU
https://github.com/apache/tvm/blob/e8369881ef69a46eaafd143a1c9bfa0eba2eb003/src/runtime/vulkan/vulkan.cc#L362
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
