F16SOP.MAD

Format: F16SOP.MAD dest1, dest2, dest3, dest4, argumentA1, argumentE1, argumentV1, argumentC2, argumentG2, argumentX2, argumentB3, argumentF3, argumentW3, argumentD4, argumentH4, argumentY4

Phase0: F16SOP.MAD W0{.F16}{.E0}{.CLAMP}, W0{.F16}{.E1}{.CLAMP}, W1{.F16}{.E0}{.CLAMP}, W1{.F16}{.E1}{.CLAMP}, Sn{.U8|.F16}{.E0|.E1|.E2|.E3}{.NEG}{.ABS}{.FLR}, (Sn|0){.U8|.F16}{.E0|.E1|.E2|.E3}{.ONEMINUS}{.CLAMP}{.ABS}, (Sn|0){.U8|.F16}{.E0|.E1|.E2|.E3}{.NEG}{.CLAMP}{.ABS}{.FLR}, Sn{.U8|.F16}{.E0|.E1|.E2|.E3}{.NEG}{.ABS}{.FLR}, (Sn|0){.U8|.F16}{.E0|.E1|.E2|.E3}{.ONEMINUS}{.CLAMP}{.ABS}, (Sn|0){.U8|.F16}{.E0|.E1|.E2|.E3}{.NEG}{.CLAMP}{.ABS}{.FLR}, Sn{.U8|.F16}{.E0|.E1|.E2|.E3}{.NEG}{.ABS}{.FLR}, (Sn|0){.U8|.F16}{.E0|.E1|.E2|.E3}{.ONEMINUS}{.CLAMP}{.ABS}, (Sn|0){.U8|.F16}{.E0|.E1|.E2|.E3}{.NEG}{.CLAMP}{.ABS}{.FLR}, Sn{.U8|.F16}{.E0|.E1|.E2|.E3}{.NEG}{.ABS}{.FLR}, (Sn|0){.U8|.F16}{.E0|.E1|.E2|.E3}{.ONEMINUS}{.CLAMP}{.ABS}, (Sn|0){.U8|.F16}{.E0|.E1|.E2|.E3}{.NEG}{.CLAMP}{.ABS}{.FLR}

Phase1: -

Phase2: -

Description: Multiply-add 16-bit floating point sum of products, one source

It is possible to do 4 SOPMADs in one cycle.

Example:

void main()
{
    mediump vec4 a16 = a;
    mediump vec4 b16 = b;
    mediump vec4 c16 = c;
    fragColor = a16 * b16 + c16;
}

2    : sop r0.joutj, sh4, i3, add, sh8, 0.oneminus
       sop r1.koutk, sh5, i1, add, i0, 0.oneminus