use u64 limbs in core::num::bignum

[kiana1kaslana]

View all comments

The code was totally written by myself without any llm.Last time I write description with the polish of llm,but this time I write the description on my own.
This pr use u64 limbs in core::num::bignum

Since the unlocking of u128 in #146277, a discussion happened in #137887. From which I know maybe I can help making a u64 digit path.

changes are as follows:

1. total capacity of bignum remains the same, 1280bits. Big32x40=>Big64x20
2. add u64 entry to SMALL_POW5.
3. update the corresponding coretests to use Big64x20.

performance:

​ I ran benchmarks on my PC,use

./x test library/coretests --test-args bignum
./x test library/coretests --test-args flt2dec
./x test tidy

​ results are as follows:

Benchmark	Big32x40 (ns/iter)	Big64x20 (ns/iter)	Speedup
bench_big_exact_inf	33,568	~21,700	1.55x
bench_big_shortest	2,913	~1,880	1.55x
bench_big_exact_12	1,680	~1,030	1.63x
bench_big_exact_3	878	~585	1.50x
bench_small_exact_inf	1,353	~1,425	~1.0x
bench_small_exact_12	277	~220	1.26x
bench_small_exact_3	170	~134	1.27x
bench_small_shortest	320	~275	1.16x

[rustbot]

Some changes occurred in float parsing

cc @tgross35

[rustbot]

Thanks for the pull request, and welcome! The Rust Project has assigned @clarfonthey (or someone else) to review your changes, you should hear from them (or someone else) within the next two weeks.

Please see the contribution instructions and our LLM policy for more information.

Why was this reviewer chosen?

The reviewer was selected based on:

• Owners of files modified in this PR: libs
• libs expanded to 12 candidates
• Random selection from Darksonn, JohnTitor, Mark-Simulacrum, clarfonthey, jhpratt

[tgross35]

This PR was initially created as #162746

​ I ran benchmarks on my PC,use

./x test library/coretests --test-args bignum
./x test library/coretests --test-args flt2dec
./x test tidy

Could you post your processor type so we have an idea of what this is on? Also these don't look like the invocations to run benchmarks.

[kiana1kaslana]

This PR was initially created as #162746

​ I ran benchmarks on my PC,use

./x test library/coretests --test-args bignum
./x test library/coretests --test-args flt2dec
./x test tidy

Could you post your processor type so we have an idea of what this is on? Also these don't look like the invocations to run benchmarks.

I used the wrong command, it should be ./x bench library/coretests --test-args bignum. My CPU is AMD Ryzen 9 7940HX. After my changes I will rerun the benchmarks and update the data.

[kiana1kaslana]

test command:
./x bench library/coretests --test-args dragon
data(u64 limbs):
benchmarks:
num::flt2dec::strategy::dragon::bench_big_exact_12 655.21ns/iter +/- 227.79
num::flt2dec::strategy::dragon::bench_big_exact_3 358.60ns/iter +/-91.29
num::flt2dec::strategy::dragon::bench_big_exact_inf 12631.63ns/iter +/- 4610.84
num::flt2dec::strategy::dragon::bench_big_shortest 1128.00ns/iter +/- 300.20
num::flt2dec::strategy::dragon::bench_small_exact_12 137.03ns/iter +/-31.65
num::flt2dec::strategy::dragon::bench_small_exact_3 85.67ns/iter +/-19.77
num::flt2dec::strategy::dragon::bench_small_exact_inf 909.91ns/iter +/- 260.49
num::flt2dec::strategy::dragon::bench_small_shortest 174.78ns/iter +/-24.34

I think the comparison bettween u64 and u32 can refer to rust-timer.

[kiana1kaslana]

Well, I've learned the use of unbounded_shr(), thanks.For the macro approach, I'd like to keep the tables as-is for this round

[clarfonthey]

Always happy to find more uses for new libs functions. In terms of the tables, I'm okay with that, but would you mind adding a FIXME comment somewhere around the first table definition and say that maybe we should define these using a macro instead?

[kiana1kaslana]

Ok, I've polished the code as you said,thanks for your guidance.

[clarfonthey]

One final comment, otherwise looks good to me. Thank you for working on this!

[clarfonthey]

@bors r+ rollup

Thank you!

[rust-bors]

📌 Commit 9d8d943 has been approved by clarfonthey

It is now in the queue for this repository.

[tgross35]

I expect the perf improvements in #162989 (comment) may be this, may as well confirm

@rust-timer build 898e43b

[rust-timer]

Finished benchmarking commit (898e43b): comparison URL.

Overall result: no relevant changes - no action needed

Benchmarking means the PR may be perf-sensitive. Consider adding rollup=never if this change is not fit for rolling up.

@rustbot label: -S-waiting-on-perf -perf-regression

Instruction count

This perf run didn't have relevant results for this metric.

Max RSS (memory usage)

Results (primary 3.0%)

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

	mean	range	count
Regressions ❌ (primary)	3.0%	[2.5%, 3.5%]	2
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	3.0%	[2.5%, 3.5%]	2

Cycles

This perf run didn't have relevant results for this metric.

Binary size

Results (primary -0.0%)

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)	-0.0%	[-0.0%, -0.0%]	4
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	-0.0%	[-0.0%, -0.0%]	4

Bootstrap: 498.333s -> 497.77s (-0.11%)
Artifact size: 408.93 MiB -> 408.87 MiB (-0.01%)