ggml: Fix internal overflow in ggml_time_us on Windows #1702
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
The implementation of ggml_time_us() on Windows can realistically overflow and produce wrong data (It happened to me and messed up a bunch of benchmark results).
With a timer frequency of 10,000,000 ticks per second (which I have on my machine) the internal result of the multiplication (t.QuadPart * 1000000) overflows the 63 bits after 2^63 / 1000000 / 10000000 seconds. That's about 10.5 days. Because that product is then divided by the timer frequency, the upper bits are essentially lost and the calculated timings from using this function will be wrong.
This PR changes the code so that the startup time is subtracted. That means it won't overflow when the uptime of the OS crosses these 10 day, but only when the process runs longer than 10 days.
Ideally, you'd want to use something like MulDiv for 64-bits to prevent the overflow entirely, but I couldn't find a simple implementation for that.