%horizontal-add-vector ( dst src1 src2 rep -- )

Vocabulary

cpu.architecture

Inputs

dst	destination register symbol
src1	first source register
src2	second source register
rep	representation

Outputs

None

Word description

Emits machine code for performing a horizontal add, meaning that adjacent elements in the same operand are added together. So if the two vectors are (a0, a1, a2, a3) and (b0, b1, b2, b3) then the result is (a0 + a1, a2 + a3, b0 + b1, b2 + b3).

Examples

USING: cpu.architecture make tools.disassembler ; [ XMM0 XMM1 XMM2 float-4-rep %horizontal-add-vector ] B{ } make disassemble 0000000002660870: 0f28c1 movaps xmm0, xmm1
0000000002660873: f20f7cc2 haddps xmm0, xmm2

Definition

USING: system ;

IN: cpu.architecture

HOOK: %horizontal-add-vector cpu ( dst src1 src2 rep -- )

Methods

USING: cpu.architecture kernel system ;

M: arm.64 %horizontal-add-vector 4drop ;


USING: combinators cpu.architecture cpu.x86 cpu.x86.assembler
kernel system ;

M: x86 %horizontal-add-vector
    [ two-operand ] keep signed-rep {
        { float-4-rep [ HADDPS ] }
        { double-2-rep [ HADDPD ] }
        { int-4-rep [ PHADDD ] }
        { short-8-rep [ PHADDW ] }
    } case ;