RGBCube/serenity
1
Fork 0
mirror of https://github.com/RGBCube/serenity synced 2025-05-31 10:28:10 +00:00
Code Issues Activity Actions

TmpFS: Grow the underlying inode buffer with 2x factor when written to

Browse source 
Before this, we would end up in memcpy() churn hell when a program was
doing repeated write() calls to a file in /tmp.

An even better solution will be to only grow the VM allocation of the
underlying buffer and keep using the same physical pages. This would
eliminate all the memcpy() work.

I've benchmarked this using g++ to compile Kernel/Process.cpp.
With these changes, compilation goes from ~35 sec to ~31 sec. :^)
This commit is contained in:
Andreas Kling 2020-01-19 13:58:16 +01:00
parent 1d02ac35fc
commit 167b57a6b7
1 changed files with 9 additions and 1 deletions
Download patch file Download diff file Expand all files Collapse all files

10
Kernel/FileSystem/TmpFS.cpp
View file

@ -208,7 +208,15 @@ ssize_t TmpFSInode::write_bytes(off_t offset, ssize_t size, const u8* buffer, Fi
if (m_content.has_value() && m_content.value().capacity() >= (size_t)new_size) {
m_content.value().set_size(new_size);
} else {
auto tmp = KBuffer::create_with_size(new_size);
// Grow the content buffer 2x the new sizeto accomodate repeating write() calls.
// Note that we're not actually committing physical memory to the buffer
// until it's needed. We only grow VM here.
// FIXME: Fix this so that no memcpy() is necessary, and we can just grow the
// KBuffer and it will add physical pages as needed while keeping the
// existing ones.
auto tmp = KBuffer::create_with_size(new_size * 2);
tmp.set_size(new_size);
if (m_content.has_value())
memcpy(tmp.data(), m_content.value().data(), old_size);
m_content = move(tmp);