Experiment: New fmt::Arguments implementation

[m-ou-se]

Part of #99012

Continuation of #137294

This is an experimental new implementation of fmt::Arguments. In this implementation, fmt::Arguments is only two pointers in size. (Instead of six, currently.) This makes it the same size as a &str and makes it fit in a register pair.

Additionally, this fmt::Arguments can store a &'static str without any indirection or additional storage. This means that simple cases like print_fmt(format_args!("hello")) are now just as efficient for the caller as print_str("hello"), as shown by this example:

code:

fn main() {
    println!("Hello, world!");
}

before:

main:
 sub     rsp, 56
 lea     rax, [rip + .Lanon_hello_world]
 mov     qword ptr [rsp + 8], rax
 mov     qword ptr [rsp + 16], 1
 mov     qword ptr [rsp + 24], 8
 xorps   xmm0, xmm0
 movups  xmmword ptr [rsp + 32], xmm0
 lea     rdi, [rsp + 8]
 call    qword ptr [rip + std::io::stdio::_print]
 add     rsp, 56
 ret

after:

main:
 lea     rsi, [rip + .Lanon_hello_world]
 mov     edi, 29
 jmp     qword ptr [rip + std::io::stdio::_print]

Similarly, panic!("Hello, world!"); shows the same change.

To do:

• Performance testing
• Figure out why things are slower to compile
• Fix compilation performance
• Documentation / comments

[m-ou-se]

@bors try @rust-timer queue

[rust-bors]

☀️ Try build successful (CI)
Build commit: 89ea077 (89ea077499c3fe06868ef1d91f2c47c8af941dce, parent: 53efb3d4f3b67d189a0c72eb475a52017d79d609)

[rust-timer]

Finished benchmarking commit (89ea077): 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)	0.9%	[0.1%, 8.4%]	95
Regressions ❌ (secondary)	1.1%	[0.0%, 12.0%]	37
Improvements ✅ (primary)	-0.6%	[-2.2%, -0.1%]	39
Improvements ✅ (secondary)	-2.1%	[-35.3%, -0.0%]	55
All ❌✅ (primary)	0.5%	[-2.2%, 8.4%]	134

Max RSS (memory usage)

Results (primary 1.6%, secondary -0.7%)

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

	mean	range	count
Regressions ❌ (primary)	2.7%	[1.0%, 5.5%]	22
Regressions ❌ (secondary)	4.3%	[2.2%, 8.9%]	7
Improvements ✅ (primary)	-2.6%	[-6.5%, -0.6%]	6
Improvements ✅ (secondary)	-3.6%	[-5.6%, -1.5%]	12
All ❌✅ (primary)	1.6%	[-6.5%, 5.5%]	28

Cycles

Results (primary 4.5%, secondary -4.4%)

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

	mean	range	count
Regressions ❌ (primary)	6.8%	[1.7%, 11.0%]	13
Regressions ❌ (secondary)	5.8%	[2.2%, 11.1%]	7
Improvements ✅ (primary)	-3.1%	[-5.8%, -2.1%]	4
Improvements ✅ (secondary)	-8.8%	[-36.8%, -2.4%]	16
All ❌✅ (primary)	4.5%	[-5.8%, 11.0%]	17

Binary size

Results (primary -0.2%, secondary -0.2%)

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

	mean	range	count
Regressions ❌ (primary)	0.4%	[0.0%, 2.2%]	24
Regressions ❌ (secondary)	3.7%	[0.1%, 19.5%]	26
Improvements ✅ (primary)	-0.4%	[-2.2%, -0.0%]	100
Improvements ✅ (secondary)	-1.2%	[-11.2%, -0.0%]	105
All ❌✅ (primary)	-0.2%	[-2.2%, 2.2%]	124

Bootstrap: 473.384s -> 471.178s (-0.47%)
Artifact size: 390.72 MiB -> 387.77 MiB (-0.75%)

[rust-timer]

Finished benchmarking commit (3b95bd4): 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)	1.7%	[0.1%, 22.0%]	120
Regressions ❌ (secondary)	0.9%	[0.1%, 11.2%]	49
Improvements ✅ (primary)	-0.7%	[-3.6%, -0.1%]	45
Improvements ✅ (secondary)	-2.0%	[-36.4%, -0.0%]	62
All ❌✅ (primary)	1.1%	[-3.6%, 22.0%]	165

Max RSS (memory usage)

Results (primary 2.2%, secondary 1.7%)

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

	mean	range	count
Regressions ❌ (primary)	2.9%	[1.1%, 7.5%]	20
Regressions ❌ (secondary)	2.7%	[1.3%, 5.9%]	11
Improvements ✅ (primary)	-2.0%	[-2.9%, -1.0%]	3
Improvements ✅ (secondary)	-3.6%	[-3.7%, -3.6%]	2
All ❌✅ (primary)	2.2%	[-2.9%, 7.5%]	23

Cycles

Results (primary 5.5%, secondary -2.8%)

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

	mean	range	count
Regressions ❌ (primary)	5.9%	[1.6%, 14.2%]	28
Regressions ❌ (secondary)	4.4%	[1.8%, 9.8%]	13
Improvements ✅ (primary)	-4.5%	[-4.5%, -4.5%]	1
Improvements ✅ (secondary)	-10.0%	[-38.0%, -1.8%]	13
All ❌✅ (primary)	5.5%	[-4.5%, 14.2%]	29

Binary size

Results (primary 0.5%, secondary -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%	[0.0%, 5.3%]	71
Regressions ❌ (secondary)	2.9%	[0.0%, 10.6%]	30
Improvements ✅ (primary)	-0.6%	[-2.1%, -0.0%]	51
Improvements ✅ (secondary)	-1.2%	[-11.3%, -0.0%]	101
All ❌✅ (primary)	0.5%	[-2.1%, 5.3%]	122

Bootstrap: 474.482s -> 474.113s (-0.08%)
Artifact size: 390.74 MiB -> 388.33 MiB (-0.62%)

[m-ou-se]

I have no clue what to do with these numbers. Ideas and suggestions very welcome.

[m-ou-se]

@bors try @rust-timer queue

[rust-bors]

☀️ Try build successful (CI)
Build commit: fbb43aa (fbb43aa2e4f4bc9f6b7d8185976b444e49bbef87, parent: c90bcb9571b7aab0d8beaa2ce8a998ffaf079d38)

[rust-timer]

Finished benchmarking commit (fbb43aa): 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)	1.7%	[0.1%, 21.1%]	104
Regressions ❌ (secondary)	0.7%	[0.0%, 5.5%]	43
Improvements ✅ (primary)	-0.6%	[-3.6%, -0.1%]	49
Improvements ✅ (secondary)	-2.0%	[-37.0%, -0.1%]	62
All ❌✅ (primary)	1.0%	[-3.6%, 21.1%]	153

Max RSS (memory usage)

Results (primary 1.0%, secondary -0.7%)

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

	mean	range	count
Regressions ❌ (primary)	2.2%	[0.9%, 4.2%]	16
Regressions ❌ (secondary)	2.5%	[1.2%, 3.4%]	5
Improvements ✅ (primary)	-2.9%	[-4.2%, -1.9%]	5
Improvements ✅ (secondary)	-3.0%	[-4.9%, -1.7%]	7
All ❌✅ (primary)	1.0%	[-4.2%, 4.2%]	21

Cycles

Results (primary 5.3%, secondary -4.9%)

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

	mean	range	count
Regressions ❌ (primary)	6.0%	[2.0%, 13.0%]	11
Regressions ❌ (secondary)	3.9%	[2.1%, 7.9%]	6
Improvements ✅ (primary)	-1.7%	[-1.7%, -1.7%]	1
Improvements ✅ (secondary)	-8.4%	[-38.7%, -1.4%]	15
All ❌✅ (primary)	5.3%	[-1.7%, 13.0%]	12

Binary size

Results (primary 0.5%, secondary -0.3%)

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

	mean	range	count
Regressions ❌ (primary)	1.4%	[0.0%, 5.3%]	72
Regressions ❌ (secondary)	2.9%	[0.0%, 10.0%]	29
Improvements ✅ (primary)	-0.6%	[-2.2%, -0.0%]	60
Improvements ✅ (secondary)	-1.2%	[-11.3%, -0.0%]	103
All ❌✅ (primary)	0.5%	[-2.2%, 5.3%]	132

Bootstrap: 475.185s -> 473.728s (-0.31%)
Artifact size: 390.81 MiB -> 388.34 MiB (-0.63%)

[m-ou-se]

Seems like a lot of time is spent in live_symbols_and_ignored_derived_traits.

generate_crate_metadata also went up somewhat significantly, together with size:crate_metadata going up a lot in many cases. 🤔

[m-ou-se]

@bors try @rust-timer queue

[rust-bors]

☀️ Try build successful (CI)
Build commit: 2f456ca (2f456ca5a02a06d0f2256251c601b1811cc47ac3, parent: 00426d642ad6a8c6192bf517702a68b0a8001547)

[rust-timer]

Finished benchmarking commit (2f456ca): 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)	1.7%	[0.1%, 21.2%]	107
Regressions ❌ (secondary)	0.7%	[0.0%, 5.7%]	46
Improvements ✅ (primary)	-0.7%	[-3.5%, -0.1%]	46
Improvements ✅ (secondary)	-1.9%	[-36.7%, -0.1%]	63
All ❌✅ (primary)	1.0%	[-3.5%, 21.2%]	153

Max RSS (memory usage)

Results (primary 0.9%, secondary 0.5%)

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

	mean	range	count
Regressions ❌ (primary)	2.3%	[1.2%, 4.0%]	18
Regressions ❌ (secondary)	4.2%	[1.1%, 6.9%]	13
Improvements ✅ (primary)	-2.3%	[-3.7%, -0.6%]	8
Improvements ✅ (secondary)	-3.3%	[-6.0%, -2.0%]	13
All ❌✅ (primary)	0.9%	[-3.7%, 4.0%]	26

Cycles

Results (primary 6.3%, secondary -3.7%)

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

	mean	range	count
Regressions ❌ (primary)	6.3%	[3.3%, 13.7%]	9
Regressions ❌ (secondary)	4.8%	[1.6%, 9.8%]	9
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-11.3%	[-38.2%, -2.0%]	10
All ❌✅ (primary)	6.3%	[3.3%, 13.7%]	9

Binary size

Results (primary 0.5%, secondary -0.3%)

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

	mean	range	count
Regressions ❌ (primary)	1.4%	[0.0%, 5.3%]	72
Regressions ❌ (secondary)	2.9%	[0.0%, 10.0%]	29
Improvements ✅ (primary)	-0.6%	[-2.2%, -0.0%]	61
Improvements ✅ (secondary)	-1.2%	[-11.3%, -0.0%]	104
All ❌✅ (primary)	0.5%	[-2.2%, 5.3%]	133

Bootstrap: 476.293s -> 472.782s (-0.74%)
Artifact size: 391.34 MiB -> 388.55 MiB (-0.71%)

[m-ou-se]

Closing this in favor of #148789, which has much better results 🎉