Streamline rent/return on ArrayPool

[stephentoub]

A variety of tweaks to reduce overheads:

• Stop storing and using a _bucketArraySizes. It's cheaper to recompute the shift on each use than it is to index into the array (with a bounds check). Plus less memory.
• The 99% case is renting a positive length for pooled array sizes (especially now that we've bumped the limit up to a gig). Move the checks for lengths <= 0 to after the check for whether the length is poolable.
• Move arrays/counts into locals to enable the JIT to eliminate some bounds checks.
• Use ThrowHelpers where we already have them
• Move non-generic helpers out of generic class into Utilities
• Consolidate buffer allocation in Rent to a single line
• Reorganize TLS checks to be as early as possible
• Use FastMod instead of % in per-core stacks

Method	Toolchain	Mean	Error	StdDev	Ratio	RatioSD	Code Size
PopPush	\main\corerun.exe	17.58 ns	0.357 ns	0.334 ns	1.00	0.00	2,221 B
PopPush	\pr\corerun.exe	16.64 ns	0.021 ns	0.019 ns	0.95	0.02	1,980 B
PopPushMulti	\main\corerun.exe	1,127.07 ns	7.303 ns	6.098 ns	1.00	0.00	2,167 B
PopPushMulti	\pr\corerun.exe	1,082.29 ns	3.208 ns	2.679 ns	0.96	0.01	1,875 B

[ghost]

I couldn't figure out the best area label to add to this PR. If you have write-permissions please help me learn by adding exactly one area label.

[ghost]

Tagging subscribers to this area: @GrabYourPitchforks, @dotnet/area-system-buffers
See info in area-owners.md if you want to be subscribed.

Issue Details

---

A variety of tweaks to reduce overheads:

• Stop storing and using a _bucketArraySizes. It's cheaper to recompute the shift on each use than it is to index into the array (with a bounds check). Plus less memory.
• The 99% case is renting a positive length for pooled array sizes (especially now that we've bumped the limit up to a gig). Move the checks for lengths <= 0 to after the check for whether the length is poolable.
• Move arrays/counts into locals to enable the JIT to eliminate some bounds checks.
• Use ThrowHelpers where we already have them
• Move non-generic helpers out of generic class into Utilities
• Consolidate buffer allocation in Rent to a single line
• Reorganize TLS checks to be as early as possible
• Use FastMod instead of % in per-core stacks

Method	Toolchain	Mean	Error	StdDev	Ratio	RatioSD	Code Size
PopPush	\main\corerun.exe	17.58 ns	0.357 ns	0.334 ns	1.00	0.00	2,221 B
PopPush	\pr\corerun.exe	16.64 ns	0.021 ns	0.019 ns	0.95	0.02	1,980 B
PopPushMulti	\main\corerun.exe	1,127.07 ns	7.303 ns	6.098 ns	1.00	0.00	2,167 B
PopPushMulti	\pr\corerun.exe	1,082.29 ns	3.208 ns	2.679 ns	0.96	0.01	1,875 B

Author:	stephentoub
Assignees:	-
Labels:	area-System.Buffers
Milestone:	6.0.0