Use buffer

[blegat]

Part of https://github.com/JuliaAlgebra/MultivariatePolynomials.jl/issues/194

Needs

• [x] https://github.com/jump-dev/MutableArithmetics.jl/pull/196

Int64

Benchmark 0

Before

	Time	Alloc	Memory
SIMDPolynomials	317.813 ns	6	272 bytes
DynamicPolynomials	5.072 μs	170	9.62 KiB
TypedPolynomials	117.847 ns	2	208 bytes

After

	Time	Alloc	Memory
SIMDPolynomials	301.179 ns	6	272 bytes
DynamicPolynomials	3.124 μs	147	8.56 KiB
TypedPolynomials	118.545 ns	2	208 bytes

Benchmark 1

Before

	Time	Alloc	Memory
SIMDPolynomials	122.177 μs	1316	468.48 KiB
DynamicPolynomials	692.039 μs	15658	1.24 MiB
TypedPolynomials	97.478 μs	502	473.70 KiB

After

	Time	Alloc	Memory
SIMDPolynomials	117.127 μs	1354	442.55 KiB
DynamicPolynomials	611.045 μs	14205	1.18 MiB
TypedPolynomials	89.858 μs	476	445.50 KiB

Benchmark 2

Before

	Time	Alloc	Memory
SIMDPolynomials	3.761 μs	218	8.31 KiB
DynamicPolynomials	31.402 μs	1361	92.11 KiB
TypedPolynomials	4.284 μs	137	6.09 KiB

After

	Time	Alloc	Memory
SIMDPolynomials	3.709 μs	232	8.52 KiB
DynamicPolynomials	29.634 μs	1359	92.86 KiB
TypedPolynomials	4.310 μs	135	5.77 KiB

Benchmark 3

Before

	Time	Alloc	Memory
SIMDPolynomials	461.465 μs	26515	1.02 MiB
DynamicPolynomials	2.307 ms	89587	6.01 MiB
TypedPolynomials	92.893 μs	547	353.50 KiB

After

	Time	Alloc	Memory
SIMDPolynomials	443.949 μs	26563	934.09 KiB
DynamicPolynomials	2.203 ms	90077	6.07 MiB
TypedPolynomials	85.777 μs	595	184.81 KiB

BigInt

Benchmark 0

Before

	Time	Alloc	Memory
SIMDPolynomials	750.315 ns	34	784 bytes
DynamicPolynomials	5.634 μs	202	10.19 KiB
TypedPolynomials	555.119 ns	30	720 bytes

After

	Time	Alloc	Memory
SIMDPolynomials	774.748 ns	47	1.12 KiB
DynamicPolynomials	3.867 μs	187	9.27 KiB
TypedPolynomials	509.906 ns	27	672 bytes

Benchmark 1

Before

	Time	Alloc	Memory
SIMDPolynomials	338.276 μs	12454	1001.43 KiB
DynamicPolynomials	880.820 μs	27690	1.85 MiB
TypedPolynomials	324.824 μs	11554	1004.15 KiB

After

	Time	Alloc	Memory
SIMDPolynomials	316.109 μs	11307	969.02 KiB
DynamicPolynomials	846.564 μs	28317	1.83 MiB
TypedPolynomials	294.887 μs	9975	960.13 KiB

Benchmark 2

Before

	Time	Alloc	Memory
SIMDPolynomials	5.636 μs	307	9.92 KiB
DynamicPolynomials	35.949 μs	1471	95.04 KiB
TypedPolynomials	5.281 μs	197	6.98 KiB

After

	Time	Alloc	Memory
SIMDPolynomials	5.531 μs	323	10.16 KiB
DynamicPolynomials	34.546 μs	1479	95.95 KiB
TypedPolynomials	5.347 μs	213	7.10 KiB

Benchmark 3

Before

	Time	Alloc	Memory
SIMDPolynomials	945.529 μs	59541	1.54 MiB
DynamicPolynomials	2.873 ms	111267	6.50 MiB
TypedPolynomials	515.503 μs	33525	885.83 KiB

After

	Time	Alloc	Memory
SIMDPolynomials	927.409 μs	59597	1.43 MiB
DynamicPolynomials	2.712 ms	112985	6.58 MiB
TypedPolynomials	483.811 μs	33581	717.30 KiB

Rational{BigInt}

Benchmark 0

Before

	Time	Alloc	Memory
SIMDPolynomials	2.932 μs	217	5.33 KiB
DynamicPolynomials	8.069 μs	386	15.14 KiB
TypedPolynomials	2.792 μs	213	5.27 KiB

After

	Time	Alloc	Memory
SIMDPolynomials	2.976 μs	233	5.77 KiB
DynamicPolynomials	7.941 μs	386	15.14 KiB
TypedPolynomials	2.653 μs	213	5.27 KiB

Benchmark 1

Before

	Time	Alloc	Memory
SIMDPolynomials	1.445 ms	68042	2.67 MiB
DynamicPolynomials	1.997 ms	77748	3.53 MiB
TypedPolynomials	1.446 ms	67322	2.68 MiB

After

	Time	Alloc	Memory
SIMDPolynomials	1.417 ms	68761	2.67 MiB
DynamicPolynomials	1.970 ms	79636	3.59 MiB
TypedPolynomials	1.380 ms	67361	2.66 MiB

Benchmark 2

Before

	Time	Alloc	Memory
SIMDPolynomials	10.845 μs	692	20.28 KiB
DynamicPolynomials	45.825 μs	1968	108.00 KiB
TypedPolynomials	10.151 μs	580	17.30 KiB

After

	Time	Alloc	Memory
SIMDPolynomials	10.468 μs	706	20.42 KiB
DynamicPolynomials	42.614 μs	2007	109.67 KiB
TypedPolynomials	10.233 μs	594	17.31 KiB

Benchmark 3

Before

	Time	Alloc	Memory
SIMDPolynomials	2.562 ms	163028	4.71 MiB
DynamicPolynomials	5.302 ms	243218	9.64 MiB
TypedPolynomials	2.083 ms	135981	4.01 MiB

After

	Time	Alloc	Memory
SIMDPolynomials	2.502 ms	163076	4.55 MiB
DynamicPolynomials	5.180 ms	249218	9.84 MiB
TypedPolynomials	2.001 ms	136029	3.79 MiB

[blegat]

I don't understand where the added allocations come from. It seems to be because I changed operate!! into buffered_operate!!. Actually, changing to operate! further reduces it while it shouldn't. promote_operation does not allocate anything though. I checked on Julia v1.9 beta and there is the same issue.