Implement jump threading MIR opt

[cjgillot]

This pass is an attempt to generalize ConstGoto and SeparateConstSwitch passes into a more complete jump threading pass.

This pass is rather heavy, as it performs a truncated backwards DFS on MIR starting from each SwitchInt terminator. This backwards DFS remains very limited, as it only walks through Goto terminators.

It is build to support constants and discriminants, and a propagating through a very limited set of operations.

The pass successfully manages to disentangle the Some(x?) use case and the DFA use case. It still needs a few tests before being ready.

[rustbot]

r? @eholk

(rustbot has picked a reviewer for you, use r? to override)

[saethlin]

Is there any particular reason we need limits on here? Is the runtime potentially infinite, or just O(big) if we have none?

In general, I think it would be neat if we could have some mechanism to turn off all such limits. For DataflowConstProp I suggested that pass be enabled with limits at opt-level 3, and enabled with no limits at level 4. Of course I'm not aware of any agreement on what opt levels are for so maybe I'm the odd one here.

[cjgillot]

@bors try @rust-timer queue

[bors]

⌛ Trying commit 36155639f1e6bd248624a49537c61e11be91f233 with merge 64cc3853a551ea0838ed2abeceb8bcc254b99600...

[bors]

☀️ Try build successful - checks-actions
Build commit: 64cc3853a551ea0838ed2abeceb8bcc254b99600 (64cc3853a551ea0838ed2abeceb8bcc254b99600)

[rust-timer]

Finished benchmarking commit (64cc3853a551ea0838ed2abeceb8bcc254b99600): comparison URL.

Overall result: ❌✅ regressions and improvements - ACTION NEEDED

Benchmarking this pull request likely means that it is perf-sensitive, so we're automatically marking it as not fit for rolling up. While you can manually mark this PR as fit for rollup, we strongly recommend not doing so since this PR may lead to changes in compiler perf.

Next Steps: If you can justify the regressions found in this try perf run, please indicate this with @rustbot label: +perf-regression-triaged along with sufficient written justification. If you cannot justify the regressions please fix the regressions and do another perf run. If the next run shows neutral or positive results, the label will be automatically removed.

@bors rollup=never
@rustbot label: -S-waiting-on-perf +perf-regression

Instruction count

This is a highly reliable metric that was used to determine the overall result at the top of this comment.

	mean	range	count
Regressions ❌ (primary)	0.6%	[0.2%, 1.3%]	11
Regressions ❌ (secondary)	1.5%	[0.2%, 5.4%]	9
Improvements ✅ (primary)	-0.7%	[-1.8%, -0.2%]	7
Improvements ✅ (secondary)	-0.6%	[-1.6%, -0.2%]	14
All ❌✅ (primary)	0.1%	[-1.8%, 1.3%]	18

Max RSS (memory usage)

Results

This is a less reliable metric that may be of interest but was not used to determine the overall result at the top of this comment.

	mean	range	count
Regressions ❌ (primary)	3.1%	[0.1%, 7.1%]	3
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-1.4%	[-3.7%, -0.6%]	6
Improvements ✅ (secondary)	-2.4%	[-4.1%, -0.9%]	3
All ❌✅ (primary)	0.1%	[-3.7%, 7.1%]	9

Cycles

Results

This is a less reliable metric that may be of interest but was not used to determine the overall result at the top of this comment.

	mean	range	count
Regressions ❌ (primary)	1.2%	[1.2%, 1.2%]	1
Regressions ❌ (secondary)	3.4%	[3.3%, 3.6%]	2
Improvements ✅ (primary)	-2.2%	[-2.2%, -2.2%]	1
Improvements ✅ (secondary)	-2.4%	[-2.4%, -2.4%]	1
All ❌✅ (primary)	-0.5%	[-2.2%, 1.2%]	2

[rustbot]

Some changes occurred to MIR optimizations

cc @rust-lang/wg-mir-opt

[nikic]

Just to check, does this implementation prevent the formation of irreducible control flow?

[cjgillot]

@nikic I don't know. I'm not familiar enough with the literature, and I haven't put anything specific to prevent them, so I guess they are not prevented. Do you have pointers on prior work so I can verify/modify the pass properly?

[JakobDegen]

https://en.wikipedia.org/wiki/Control-flow_graph#Reducibility

See the Wikipedia entry for the definition of reducibility. I'll review this in a bit though and can make a suggestion then as to how to preserve it

[bors]

☔ The latest upstream changes (presumably #101692) made this pull request unmergeable. Please resolve the merge conflicts.

[eholk]

I'm probably not enough of a MIR expert to review this... Re-rolling.

@bors r? compiler

[pnkfelix]

We only recurse into a switchInt if it is the current block's single predecessor.

Hmm, I think I see what you are saying. The emphasis in the above quote is on "current", since the recursion doesn't itself change the notion of the "current block" (which I think makes sense, given what the transformation is doing), and so you won't recur more than once.

Okay. That seems fine to me; maybe someone later on can look into whether its worth attempting to generalize the code for that case.

[pnkfelix]

@bors r+

[bors]

📌 Commit dd08dd4 has been approved by pnkfelix

It is now in the queue for this repository.

[cjgillot]

Okay. That seems fine to me; maybe someone later on can look into whether its worth attempting to generalize the code for that case.

I'll think about it. I had an idea how to avoid the current backtracking algorithm, replaced by a single pass on the CFG. But there may be some time before I get around to write it.

[bors]

⌛ Testing commit dd08dd4 with merge 1322f92...

[bors]

☀️ Test successful - checks-actions
Approved by: pnkfelix
Pushing 1322f92 to master...

[rust-timer]

Finished benchmarking commit (1322f92): comparison URL.

Overall result: no relevant changes - no action needed

@rustbot label: -perf-regression

Instruction count

This benchmark run did not return any relevant results for this metric.

Max RSS (memory usage)

Results

This is a less reliable metric that may be of interest but was not used to determine the overall result at the top of this comment.

	mean	range	count
Regressions ❌ (primary)	-	-	0
Regressions ❌ (secondary)	2.7%	[2.7%, 2.7%]	1
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-5.1%	[-5.1%, -5.1%]	1
All ❌✅ (primary)	-	-	0

Cycles

Results

This is a less reliable metric that may be of interest but was not used to determine the overall result at the top of this comment.

	mean	range	count
Regressions ❌ (primary)	-	-	0
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-3.5%	[-4.2%, -2.9%]	6
All ❌✅ (primary)	-	-	0

Binary size

This benchmark run did not return any relevant results for this metric.

Bootstrap: missing data
Artifact size: 304.29 MiB -> 304.31 MiB (0.00%)