optimize Array::concat

[ericlagergren]

The resulting assembly is much nicer. As a bonus, concat no longer has panicking branches.

[ericlagergren]

For reference, here is the old assembly when concatenating Array<u8, U256> and Array<u8, U16>:

sha3_utils::test_concat:
Lfunc_begin0:
	stp x28, x27, [sp, #-64]!
	stp x22, x21, [sp, #16]
	stp x20, x19, [sp, #32]
	stp x29, x30, [sp, #48]
	add x29, sp, #48
	sub sp, sp, #864
	mov x19, x8
	mov x21, #0
	mov x20, #0
	mov x9, #0
	mov x10, #0
	mov x8, #0
	ldp q0, q1, [x0, #192]
	mov x12, sp
	stur q0, [x12, #200]
	stur q1, [x12, #216]
	ldp q0, q1, [x0, #224]
	stur q0, [x12, #232]
	stur q1, [x12, #248]
	ldp q0, q1, [x0, #128]
	stur q0, [x12, #136]
	stur q1, [x12, #152]
	ldp q0, q1, [x0, #160]
	stur q0, [x12, #168]
	stur q1, [x12, #184]
	ldp q0, q1, [x0, #64]
	stur q0, [sp, #72]
	stur q1, [sp, #88]
	ldp q0, q1, [x0, #96]
	stur q0, [sp, #104]
	stur q1, [sp, #120]
	ldp q0, q1, [x0]
	stur q0, [sp, #8]
	stur q1, [sp, #24]
	ldp q0, q1, [x0, #32]
	stur q0, [sp, #40]
	add x11, x12, #8
	stur q1, [sp, #56]
	add x12, x12, #288
	ldr q0, [x1]
	mov w22, #1
	add x13, sp, #592
	mov w14, #1
	str q0, [sp, #288]
	b LBB0_2
LBB0_1:
	add x20, x10, #1
	add x15, x11, x10
	mov w14, #1
	mov x10, x20
	ldrb w15, [x15]
	strb w15, [x13, x8]
	add x8, x8, #1
	cmp x8, #272
	b.eq LBB0_7
LBB0_2:
	tbz w14, #0, LBB0_5
	cmp x10, #256
	b.ne LBB0_1
	mov w22, #0
	cmp x9, #16
	b.eq LBB0_10
	mov w14, #0
	add x21, x9, #1
	add x15, x12, x9
	mov x9, x21
	ldrb w15, [x15]
	strb w15, [x13, x8]
	add x8, x8, #1
	cmp x8, #272
	b.ne LBB0_2
LBB0_7:
	add x0, sp, #320
	add x1, sp, #592
	mov w2, #272
	bl _memcpy
	cmp x20, #256
	cset w8, ne
	and w8, w22, w8
	tbnz w8, #0, LBB0_11
	cmp x21, #16
	b.ne LBB0_11
	add x1, sp, #320
	mov x0, x19
	mov w2, #272
	bl _memcpy
	add sp, sp, #864
	ldp x29, x30, [sp, #48]
	ldp x20, x19, [sp, #32]
	ldp x22, x21, [sp, #16]
	ldp x28, x27, [sp], #64
	ret
LBB0_10:
Lloh0:
	adrp x0, l___unnamed_1@PAGE
Lloh1:
	add x0, x0, l___unnamed_1@PAGEOFF
Lloh2:
	adrp x2, l___unnamed_2@PAGE
Lloh3:
	add x2, x2, l___unnamed_2@PAGEOFF
	mov w1, #47
	bl core::option::expect_failed
LBB0_11:
Lloh4:
	adrp x8, l___unnamed_3@PAGE
Lloh5:
	add x8, x8, l___unnamed_3@PAGEOFF
	mov w9, #1
	str x8, [sp, #592]
	str x9, [sp, #600]
	mov w8, #8
	str xzr, [sp, #624]
	str x8, [sp, #608]
	str xzr, [sp, #616]
Lloh6:
	adrp x1, l___unnamed_4@PAGE
Lloh7:
	add x1, x1, l___unnamed_4@PAGEOFF
	add x0, sp, #592
	bl core::panicking::panic_fmt

and here is the new assembly

sha3_utils::test_concat:
Lfunc_begin0:
	ldp q0, q1, [x0, #192]
	stp q0, q1, [x8, #192]
	ldp q0, q1, [x0, #224]
	stp q0, q1, [x8, #224]
	ldp q0, q1, [x0, #128]
	stp q0, q1, [x8, #128]
	ldp q0, q1, [x0, #160]
	stp q0, q1, [x8, #160]
	ldp q0, q1, [x0, #64]
	stp q0, q1, [x8, #64]
	ldp q0, q1, [x0, #96]
	stp q0, q1, [x8, #96]
	ldp q0, q1, [x0]
	stp q0, q1, [x8]
	ldp q0, q1, [x0, #32]
	stp q0, q1, [x8, #32]
	ldr q0, [x1]
	str q0, [x8, #256]
	ret