// rexcode  ·  Brendan Punsky (dotbmp@github), original author

package rexcode_arm64

// =============================================================================
// INSTRUCTION
// =============================================================================

Instruction_Flags :: bit_field u8 {
	_: u8 | 8,
}

// Sized and aligned to a cache line, deliberately.
//
// The payload is 45 bytes -- shrinking Operand to 10 got it there -- but
// leaving the struct at 48 was measurably worse than padding it back out.
// With `#packed` the struct aligns to 1, so a 48-byte stride straddles a line
// boundary 75% of the time and a 64-byte one straddled 100% of the time (the
// heap base is not line-aligned either). Decode writes whole Instructions, and
// unaligned stores cost enough that on an i7-9750H this layout decodes ~21%
// faster than the 48-byte packed one -- while writing MORE bytes. The gap
// holds even when the whole array is L1-resident, so it is split-store cost at
// the store ports, not cache-line fetches.
//
// Encode is within 1% and a pure read traversal is ~9% slower at working sets
// past L2, both of which the decode win dwarfs for real workloads.
//
// The 19 spare bytes are free: they cost nothing over a 48-byte struct that
// straddles, and new fields land in them without changing the layout.
Instruction :: struct #align(64) {
	ops:           [4]Operand `fmt:"v,operand_count"`, // 4 * size_of(Operand) = 40
	mnemonic:      Mnemonic,                           // 2
	operand_count: u8,                                 // 1
	flags:         Instruction_Flags,                  // 1
	length:        u8,                                 // 1 -- always 4
	_:             [19]u8,
}
#assert(size_of(Instruction)  == 64)
#assert(align_of(Instruction) == 64)

// =============================================================================
// Builders -- the most common shapes; less-common forms can be built
// inline by the caller using the Instruction struct directly.
// =============================================================================

@(require_results)
inst_none :: #force_inline proc "contextless" (m: Mnemonic) -> Instruction {
	return Instruction{mnemonic = m, operand_count = 0, length = 4}
}

// Single-register (e.g. BR, BLR).
@(require_results)
inst_r :: #force_inline proc "contextless" (m: Mnemonic, r: Register) -> Instruction {
	return Instruction{mnemonic = m, operand_count = 1, length = 4,
					   ops = {op_reg(r), {}, {}, {}}}
}

// 2-register (e.g. CLZ, RBIT).
@(require_results)
inst_r_r :: #force_inline proc "contextless" (m: Mnemonic, rd, rn: Register) -> Instruction {
	return Instruction{mnemonic = m, operand_count = 2, length = 4,
					   ops = {op_reg(rd), op_reg(rn), {}, {}}}
}

// 3-register (e.g. ADD shifted, MUL, UDIV, ASRV).
@(require_results)
inst_r_r_r :: #force_inline proc "contextless" (m: Mnemonic, rd, rn, rm: Register) -> Instruction {
	return Instruction{mnemonic = m, operand_count = 3, length = 4,
					   ops = {op_reg(rd), op_reg(rn), op_reg(rm), {}}}
}

// 4-register R4-type (MADD, MSUB, SMADDL, ...).
@(require_results)
inst_r_r_r_r :: #force_inline proc "contextless" (m: Mnemonic, rd, rn, rm, ra: Register) -> Instruction {
	return Instruction{mnemonic = m, operand_count = 4, length = 4,
					   ops = {op_reg(rd), op_reg(rn), op_reg(rm), op_reg(ra)}}
}

// 2-register + immediate (e.g. ADD imm).
@(require_results)
inst_r_r_i :: #force_inline proc "contextless" (m: Mnemonic, rd, rn: Register, imm: i64) -> Instruction {
	return Instruction{mnemonic = m, operand_count = 3, length = 4,
					   ops = {op_reg(rd), op_reg(rn), op_imm(imm), {}}}
}

// 1-register + immediate (e.g. MOVZ).
@(require_results)
inst_r_i :: #force_inline proc "contextless" (m: Mnemonic, rd: Register, imm: i64) -> Instruction {
	return Instruction{mnemonic = m, operand_count = 2, length = 4,
					   ops = {op_reg(rd), op_imm(imm), {}, {}}}
}

// MOVZ/MOVN/MOVK with explicit hw shift (0/16/32/48).
@(require_results)
inst_mov_imm :: #force_inline proc "contextless" (m: Mnemonic, rd: Register, imm: i64, hw: u8) -> Instruction {
	return Instruction{mnemonic = m, operand_count = 3, length = 4,
					   ops = {op_reg(rd), op_imm(imm), op_imm(i64(hw), 1), {}}}
}

// Load/store register: Rt + memory.
@(require_results)
inst_ldst :: #force_inline proc "contextless" (m: Mnemonic, rt: Register, mm: Memory) -> Instruction {
	return Instruction{mnemonic = m, operand_count = 2, length = 4,
					   ops = {op_reg(rt), op_mem(mm), {}, {}}}
}

// Load/store pair: Rt, Rt2, memory.
@(require_results)
inst_ldp_stp :: #force_inline proc "contextless" (m: Mnemonic, rt, rt2: Register, mm: Memory) -> Instruction {
	return Instruction{mnemonic = m, operand_count = 3, length = 4,
					   ops = {op_reg(rt), op_reg(rt2), op_mem(mm), {}}}
}

// PC-relative branch (B, BL).
@(require_results)
inst_branch :: #force_inline proc "contextless" (m: Mnemonic, label_id: u32) -> Instruction {
	return Instruction{mnemonic = m, operand_count = 1, length = 4,
					   ops = {op_label(label_id, 4), {}, {}, {}}}
}

// NOTE: the conditional branches, inst_cbz (+cbnz), inst_tbz (+tbnz) and
// inst_csel (+csinc/csinv/csneg) are generated per-mnemonic in
// mnemonic_builders.odin, so the generator owns those names. A conditional
// branch is one builder per condition -- inst_b_le(label), inst_bc_ne(label)
// -- because the condition is part of the mnemonic, not an operand; the
// condition-operand builders are the select/compare family, which really do
// take one (inst_csinc(rd, rn, rm, cond)).