// Tile size: 1x8
	// Accumulators: 0-7
	// Col regs: 14 then 8
	// Row regs: 15

    vmovups         xmm15, [rax]
    vmovups         xmm14, [rbx]

    vcvtph2ps       ymm15, xmm15
    vcvtph2ps       ymm14, xmm14

    vbroadcastss    ymm8, xmm14 
    vfmadd231ps     ymm0, ymm15, ymm8

    pshufd          xmm8, xmm14, 1
    vbroadcastss    ymm8, xmm8
    vfmadd231ps     ymm1, ymm15, ymm8

    pshufd          xmm8, xmm14, 2
    vbroadcastss    ymm8, xmm8
    vfmadd231ps     ymm2, ymm15, ymm8

    pshufd          xmm8, xmm14, 3
    vbroadcastss    ymm8, xmm8
    vfmadd231ps     ymm3, ymm15, ymm8

    vperm2f128      ymm14, ymm14, ymm14, 1

    vbroadcastss    ymm8, xmm14 
    vfmadd231ps     ymm4, ymm15, ymm8

    pshufd          xmm8, xmm14, 1
    vbroadcastss    ymm8, xmm8 
    vfmadd231ps     ymm5, ymm15, ymm8

    pshufd          xmm8, xmm14, 2
    vbroadcastss    ymm8, xmm8 
    vfmadd231ps     ymm6, ymm15, ymm8

    pshufd          xmm8, xmm14, 3
    vbroadcastss    ymm8, xmm8
    vfmadd231ps     ymm7, ymm15, ymm8