skip codegen for intrinsics with big fallback bodies if backend does not need them

[RalfJung]

This hopefully fixes the perf regression from #148478. I only added the intrinsics with big fallback bodies to the list; it doesn't seem worth the effort of going through the entire list.

Fixes #149945
Cc @scottmcm @bjorn3

[rustbot]

r? @jieyouxu

rustbot has assigned @jieyouxu.
They will have a look at your PR within the next two weeks and either review your PR or reassign to another reviewer.

Use r? to explicitly pick a reviewer

[RalfJung]

It seems there is no way to get the current codegen backend from a tcx. I wasn't sure what the best way is to make this list of symbols available to monomorphization, and went for a new field in Session -- does that make sense?

[scottmcm]

I don't know enough about how all this should be structured to know what the best option is here.

This seems at least plausible, since at worst it stays empty and that doesn't hurt anything (other than perf).

[RalfJung]

@bjorn3 do you have any suggestions for how to deal with this?

[bjorn3]

I am not the biggest fan of another Session field, but don't have any other suggestions either.

[RalfJung]

@bors try
@rust-timer queue

[rust-bors]

☀️ Try build successful (CI)
Build commit: 4763a83 (4763a83f81ae539aaa6f6e5e773ba1fc73de0a10, parent: 8a24a202aa02f677fc2a3e0e1a69af7545803952)

[rust-timer]

Finished benchmarking commit (4763a83): comparison URL.

Overall result: ❌✅ regressions and improvements - please read the text below

Benchmarking this pull request means it may be perf-sensitive – we'll automatically label it not fit for rolling up. You can override this, but we strongly advise not to, due to possible changes in compiler perf.

Next Steps: If you can justify the regressions found in this try perf run, please do so in sufficient writing along with @rustbot label: +perf-regression-triaged. If not, please fix the regressions and do another perf run. If its results are neutral or positive, the label will be automatically removed.

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

Instruction count

Our most reliable metric. Used to determine the overall result above. However, even this metric can be noisy.

	mean	range	count
Regressions ❌ (primary)	0.6%	[0.6%, 0.6%]	1
Regressions ❌ (secondary)	0.1%	[0.1%, 0.1%]	1
Improvements ✅ (primary)	-1.8%	[-2.8%, -0.8%]	2
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	-1.0%	[-2.8%, 0.6%]	3

Max RSS (memory usage)

Results (primary -1.5%, secondary 3.5%)

A less reliable metric. May be of interest, but not used to determine the overall result above.

	mean	range	count
Regressions ❌ (primary)	1.3%	[0.7%, 1.9%]	2
Regressions ❌ (secondary)	3.5%	[3.5%, 3.5%]	1
Improvements ✅ (primary)	-4.3%	[-7.2%, -1.4%]	2
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	-1.5%	[-7.2%, 1.9%]	4

Cycles

Results (primary -3.9%)

A less reliable metric. May be of interest, but not used to determine the overall result above.

	mean	range	count
Regressions ❌ (primary)	-	-	0
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-3.9%	[-3.9%, -3.9%]	1
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	-3.9%	[-3.9%, -3.9%]	1

Binary size

Results (primary 0.2%)

A less reliable metric. May be of interest, but not used to determine the overall result above.

	mean	range	count
Regressions ❌ (primary)	1.4%	[1.4%, 1.4%]	1
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-0.0%	[-0.1%, -0.0%]	7
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	0.2%	[-0.1%, 1.4%]	8

Bootstrap: 473.485s -> 474.195s (0.15%)
Artifact size: 390.77 MiB -> 390.79 MiB (0.01%)

[RalfJung]

@bors try
@rust-timer queue

[rust-bors]

☀️ Try build successful (CI)
Build commit: c75310a (c75310a5c412df8835187dd0ef37361b2f00d085, parent: 5497a36a7faf3d2af37beebcff7008e493202902)

[rust-timer]

Finished benchmarking commit (c75310a): comparison URL.

Overall result: ❌✅ regressions and improvements - please read the text below

Benchmarking this pull request means it may be perf-sensitive – we'll automatically label it not fit for rolling up. You can override this, but we strongly advise not to, due to possible changes in compiler perf.

Next Steps: If you can justify the regressions found in this try perf run, please do so in sufficient writing along with @rustbot label: +perf-regression-triaged. If not, please fix the regressions and do another perf run. If its results are neutral or positive, the label will be automatically removed.

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

Instruction count

Our most reliable metric. Used to determine the overall result above. However, even this metric can be noisy.

	mean	range	count
Regressions ❌ (primary)	0.7%	[0.7%, 0.7%]	1
Regressions ❌ (secondary)	0.1%	[0.1%, 0.1%]	1
Improvements ✅ (primary)	-1.8%	[-2.9%, -0.8%]	2
Improvements ✅ (secondary)	-0.4%	[-0.4%, -0.4%]	1
All ❌✅ (primary)	-1.0%	[-2.9%, 0.7%]	3

Max RSS (memory usage)

Results (primary -4.1%)

A less reliable metric. May be of interest, but not used to determine the overall result above.

	mean	range	count
Regressions ❌ (primary)	-	-	0
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-4.1%	[-7.3%, -1.7%]	3
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	-4.1%	[-7.3%, -1.7%]	3

Cycles

Results (primary -3.9%, secondary 15.2%)

A less reliable metric. May be of interest, but not used to determine the overall result above.

	mean	range	count
Regressions ❌ (primary)	-	-	0
Regressions ❌ (secondary)	15.2%	[15.0%, 15.4%]	2
Improvements ✅ (primary)	-3.9%	[-3.9%, -3.9%]	1
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	-3.9%	[-3.9%, -3.9%]	1

Binary size

Results (primary 0.2%)

A less reliable metric. May be of interest, but not used to determine the overall result above.

	mean	range	count
Regressions ❌ (primary)	1.4%	[1.4%, 1.4%]	1
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-0.0%	[-0.1%, -0.0%]	7
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	0.2%	[-0.1%, 1.4%]	8

Bootstrap: 471.287s -> 473.923s (0.56%)
Artifact size: 390.83 MiB -> 390.83 MiB (-0.00%)

[jieyouxu]

@rustbot reroll

[RalfJung]

Seems like I had no luck with that reroll.
@rustbot reroll

@scottmcm or could you review this?

[mati865]

Cool idea!

I'll wait a few days to give @scottmcm time to respond respond as the much more knowledgeable person.

Do you know if there is a list of similarly optimised intrinsics somewhere?

[RalfJung]

In principle one could go over all the intrinsics that have fallback bodies, and then check whether the LLVM backend has implementations for them.

But most fallback bodies are small so the cost of monomorphizing them is tiny. Not sure if it's worth going through the entire list. I think I got all the ones that have big fallback bodies where we really don't want to pay the monomorphization cost.