Use zero_mem instead of a zerointializer for init intrinsic
#21282
Conversation
|
Thanks for the pull request, and welcome! The Rust team is excited to review your changes, and you should hear from @alexcrichton (or someone else) soon. If any changes to this PR are deemed necessary, please add them as extra commits. This ensures that the reviewer can see what has changed since they last reviewed the code. The way Github handles out-of-date commits, this should also make it reasonably obvious what issues have or haven't been addressed. Large or tricky changes may require several passes of review and changes. Please see CONTRIBUTING.md for more information. |
|
Nice! Can you add the test from #21264 as well to make sure we don't regress on the performance again? It should hopefully time out the test suite or set off some alarm if it goes back to the way it was. It may also be worth leaving a comment why we're using As a final bit, I'm not sure how much this matters, but non-optimized builds which use |
|
@alexcrichton I'll add the test case. However, non-optimised builds didn't get translated as immediate constant 0 directly. Instead it would create the stack slot, store a constant 0 and then load the value from that slot. This just changes that middle step to a call to |
| pub fn init<T>() -> T; | ||
| } | ||
|
|
||
| const SIZE: usize = 512 * 1024 * 1024; |
There was a problem hiding this comment.
1024 * 1024 is enough to reliably reproduce the issue. No need to make this test require 512MiB of RAM to run?
|
r=me with the size decrease @nagisa mentioned |
LLVM gets overwhelmed when presented with a zeroinitializer for a large type. In unoptimised builds, it generates a long sequence of stores to memory. In optmised builds, it manages to generate a standard memset of zero values, but takes a long time doing so. Call out to the `llvm.memset` function to zero out the memory instead.
|
@bors: r+ 25a4adc |
|
⌛ Testing commit 25a4adc with merge dd8f887... |
LLVM gets overwhelmed when presented with a zeroinitializer for a large type. In unoptimised builds, it generates a long sequence of stores to memory. In optmised builds, it manages to generate a standard memset of zero values, but takes a long time doing so. Call out to the `llvm.memset` function to zero out the memory instead. Fixes #21264
|
💔 Test failed - auto-mac-64-nopt-t |
|
@bors: retry |
LLVM gets overwhelmed when presented with a zeroinitializer for a large type. In unoptimised builds, it generates a long sequence of stores to memory. In optmised builds, it manages to generate a standard memset of zero values, but takes a long time doing so. Call out to the `llvm.memset` function to zero out the memory instead. Fixes #21264
|
⌛ Testing commit 25a4adc with merge 43f2c19... |
LLVM gets overwhelmed when presented with a zeroinitializer for a large
type. In unoptimised builds, it generates a long sequence of stores to
memory. In optmised builds, it manages to generate a standard memset of
zero values, but takes a long time doing so.
Call out to the
llvm.memsetfunction to zero out the memory instead.Fixes #21264