| Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings. |
| Owners-Override | +1 |
Kalvin, can you take a look? (Both Keishi-san and Sakamoto-san are OOO)
| Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings. |
| Code-Review | +1 |
Sorry about the long long delay. One non-blocking nit, but feel free to overrule.
constexpr size_t kLargeEmptySlotSpanRingSize = 1 << 7;Nit: this magic constant appears a fair bit in this CL. This is personal preference, but I might do something like
```
enum SlotSpanRingMaxSize : int16_t {
kSmaller = 1 << 7,
kLarger = 1 << 10,
};
inline constexpr size_t kMaxEmptySlotSpanRingSize =
#if PA_BUILDFLAG(USE_LARGE_EMPTY_SLOT_SPAN_RING)
SlotSpanRingMaxSize::kLarger;
#else
SlotSpanRingMaxSize::kSmaller;
#endif
// and in PA unit test, Blink, and `partition_alloc_features`,
// we can utter `SlotSpanRingMaxSize::kSmaller`.
```
and then we don't get the naked `1 << 7` traveling around the codebase.
I think most files here are allowed to include `partition_alloc_constants.h`, so hopefully we don't get any build / dependency complications out of this.
If I have misunderstood the constant (or you anticipate this will be a relatively short-lived change), feel free to ignore this comment 😊
| Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings. |
| Commit-Queue | +2 |
| Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings. |
4 is the latest approved patch-set.
The change was submitted with unreviewed changes in the following files:
```
The name of the file: base/allocator/partition_alloc_features.cc
Insertions: 8, Deletions: 12.
The diff is too large to show. Please review the diff.
```
```
The name of the file: base/allocator/partition_allocator/src/partition_alloc/partition_alloc_constants.h
Insertions: 11, Deletions: 3.
The diff is too large to show. Please review the diff.
```
```
The name of the file: base/allocator/partition_allocator/src/partition_alloc/partition_alloc_unittest.cc
Insertions: 3, Deletions: 3.
The diff is too large to show. Please review the diff.
```
```
The name of the file: third_party/blink/renderer/platform/wtf/allocator/partitions.cc
Insertions: 2, Deletions: 1.
The diff is too large to show. Please review the diff.
```
[PA] Parametrize Empty Slot Span Ring Buffer size
This helps to fine-tune the constant through an experiment.
Ring size configuration is now handled outside PartitionAlloc, because
concept of "foreground" / "background" is Chrome-specific and should be
taken care of at `//base` layer.
| Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings. |
Are you now parameterizing the value and working on the linked bug now? We are using the older foreground notify function (in the shim) in d8 and need d8 to be aligned with what's going on here to get meaningful profiles.
| Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings. |
This also broke V8 downstream and blocked a DEPS roll. How does PA ensure that it doesn't block embedders? Are we holding things wrong in V8?
In fact, we had to switch to an `internal` version now: https://chromium-review.googlesource.com/c/v8/v8/+/7414928
| Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings. |