Instruction Class Clean-Up

[FinnWilkinson]

Generally tidies the instruction classes and moves some common functions & member variables into the base class.

Also includes new classes in Container.hh which mitigates slowdown caused by changing sourceValues_, results_, sourceRegisters_, destinationRegisters_ to std::vector when SME was first supported. Said class has uses std::variant which contains std::array (used by default) and std::vector (used on call of makeSME()). This means most instructions can use std::array and get speed benefits, and SME instructions get the flexibility of std::vecotr (needed due to variable and large amount of operands).

Closes #355

AArch64 speeds with new Constants.hh (run on M1 MacBook Pro). TX2.yaml used for tx2 binaries, a64fx.yaml for armv8.4-a(+sve) binaries, a64fx_SME.yaml for all SME binaries. SME binaries run matmul kernel for 100 iterations on square inputs.

Binary	MIPS dev	time/ms dev	MIPS insn-class-clean	time/ms insn-class-clean
bude-armclang20.0-armv8.4-a	1.2	27033	1.4	22432
bude-armclang20.0-armv8.4-a+sve	0.87	12522	1.1	10078
bude-armclang20.0-tx2	1.2	27276	1.3	25262
bude-gcc10.3.0-armv8.4-a	1.2	27048	1.2	26280
bude-gcc10.3.0-armv8.4-a+sve	0.86	11912	0.92	11195
bude-gcc10.3.0-tx2	1.2	26344	1.3	24695
bude-gcc8.3.0-armv8.4-a	1.1	28607	1.2	26451
bude-gcc8.3.0-armv8.4-a+sve	0.86	12303	0.97	10880
bude-gcc8.3.0-tx2	1.2	27657	1.3	25550
bude-gcc9.3.0-armv8.4-a	1.1	28521	1.2	26678
bude-gcc9.3.0-armv8.4-a+sve	0.8	12520	0.94	10590
bude-gcc9.3.0-tx2	1.2	27505	1.3	26040
stream-armclang20.0-armv8.4-a	0.78	16179	1	12644
stream-armclang20.0-armv8.4-a+sve	0.82	9180	0.99	7610
stream-armclang20.0-tx2	0.64	15039	0.76	12591
stream-gcc10.3.0-armv8.4-a	0.97	15709	1.2	12856
stream-gcc10.3.0-armv8.4-a+sve	0.68	11046	0.85	8809
stream-gcc10.3.0-tx2	1	14667	1.2	13210
stream-gcc8.3.0-armv8.4-a	1	14808	1.1	13695
stream-gcc8.3.0-armv8.4-a+sve	0.68	10956	0.76	9871
stream-gcc8.3.0-tx2	0.99	15327	1.1	13459
stream-gcc9.3.0-armv8.4-a	1	14642	1.1	13574
stream-gcc9.3.0-armv8.4-a+sve	0.72	10311	0.75	9877
stream-gcc9.3.0-tx2	1	14942	1.1	13881
sme_matmul_fp32 m=n=k=64	0.72	2138	0.75	2057
sme_matmul_fp32 m=n=k=128	0.8	10099	0.85	9561
sme_matmul_fp32 m=n=k=256	0.82	59515	0.86	56645

[ABenC377]

Looks good other than a comment typo

[JosephMoore25]

Looks good, and yields impressive speedups. Happy to approve once other's comments have been handled, but nothing extra that I could find.

[JosephMoore25]

Nice work. I like the change to using just operandContainer.

[ABenC377]

👏🏻

[JosephMoore25]

Looks good. A shame we couldn't get isInstruction cleaner, but I'll have a think for next release.