archived-ballistic/spec/arm64_xml/bfmopa_za_pp_zz.xml

<?xml version="1.0" encoding="utf-8"?>
<?xml-stylesheet type="text/xsl" encoding="UTF-8" href="iform.xsl" version="1.0"?>
<!DOCTYPE instructionsection PUBLIC "-//ARM//DTD instructionsection //EN" "iform-p.dtd">
<!-- Copyright (c) 2010-2022 Arm Limited or its affiliates. All rights reserved. -->
<!-- This document is Non-Confidential. This document may only be used and distributed in accordance with the terms of the agreement entered into by Arm and the party that Arm delivered this document to. -->

<instructionsection id="bfmopa_za_pp_zz" title="BFMOPA (non-widening)" type="instruction">
  <docvars>
    <docvar key="instr-class" value="mortlach2" />
    <docvar key="isa" value="A64" />
    <docvar key="mnemonic" value="BFMOPA" />
  </docvars>
  <heading>BFMOPA (non-widening)</heading>
  <desc>
    <brief>BFloat16 floating-point outer product and accumulate</brief>
    <description>
      <para>This instruction works with a 16-bit element ZA tile.</para>
      <para>These instructions generate an outer product of the first source vector and the second source vector. The first source is SVL<sub>H</sub>×1 vector and the second source is 1×SVL<sub>H</sub> vector.</para>
      <para>Each source vector is independently predicated by a corresponding governing predicate. When either source vector element is Inactive the corresponding destination tile element remains unmodified.</para>
      <para>The resulting outer product, SVL<sub>H</sub>×SVL<sub>H</sub>, is then destructively added to the destination tile. This is equivalent to performing a single multiply-accumulate to each of the destination tile elements.</para>
      <para>This instruction follows SME2.1 ZA-targeting non-widening BFloat16 numerical behaviors.</para>
      <para>ID_AA64SMFR0_EL1.B16B16 indicates whether this instruction is implemented.</para>
    </description>
    <status>Green</status>
    <predicated>True</predicated>
    <sm_policy>SM_1_only</sm_policy>
    <is_gov_pred_pair>True</is_gov_pred_pair>
  </desc>
  <alias_list howmany="0"></alias_list>
  <classes>
    <iclass name="SME2" oneof="1" id="iclass_mortlach2" no_encodings="1" isa="A64">
      <docvars>
        <docvar key="instr-class" value="mortlach2" />
        <docvar key="isa" value="A64" />
        <docvar key="mnemonic" value="BFMOPA" />
      </docvars>
      <iclassintro count="1"></iclassintro>
      <arch_variants>
        <arch_variant name="FEAT_B16B16" feature="FEAT_B16B16" />
      </arch_variants>
      <regdiagram form="32" psname="BFMOPA-ZA.PP.ZZ-16" tworows="1">
        <box hibit="31" width="2" settings="2">
          <c>1</c>
          <c>0</c>
        </box>
        <box hibit="29" width="9" settings="9">
          <c>0</c>
          <c>0</c>
          <c>0</c>
          <c>0</c>
          <c>0</c>
          <c>1</c>
          <c>1</c>
          <c>0</c>
          <c>1</c>
        </box>
        <box hibit="20" width="5" name="Zm" usename="1">
          <c colspan="5"></c>
        </box>
        <box hibit="15" width="3" name="Pm" usename="1">
          <c colspan="3"></c>
        </box>
        <box hibit="12" width="3" name="Pn" usename="1">
          <c colspan="3"></c>
        </box>
        <box hibit="9" width="5" name="Zn" usename="1">
          <c colspan="5"></c>
        </box>
        <box hibit="4" name="S" usename="1" settings="1">
          <c>0</c>
        </box>
        <box hibit="3" settings="1">
          <c>1</c>
        </box>
        <box hibit="2" width="2" settings="2">
          <c>0</c>
          <c>0</c>
        </box>
        <box hibit="0" name="ZAda" usename="1">
          <c></c>
        </box>
      </regdiagram>
      <encoding name="bfmopa_za_pp_zz_16" oneofinclass="1" oneof="1" label="">
        <docvars>
          <docvar key="instr-class" value="mortlach2" />
          <docvar key="isa" value="A64" />
          <docvar key="mnemonic" value="BFMOPA" />
        </docvars>
        <asmtemplate><text>BFMOPA  </text><a link="sa_zada" hover="ZA tile ZA0-ZA1 (field &quot;ZAda&quot;)">&lt;ZAda&gt;</a><text>.H, </text><a link="sa_pn" hover="First governing scalable predicate register P0-P7 (field &quot;Pn&quot;)">&lt;Pn&gt;</a><text>/M, </text><a link="sa_pm" hover="Second governing scalable predicate register P0-P7 (field &quot;Pm&quot;)">&lt;Pm&gt;</a><text>/M, </text><a link="sa_zn" hover="First source scalable vector register (field &quot;Zn&quot;)">&lt;Zn&gt;</a><text>.H, </text><a link="sa_zm" hover="Second source scalable vector register (field &quot;Zm&quot;)">&lt;Zm&gt;</a><text>.H</text></asmtemplate>
      </encoding>
      <ps_section howmany="1">
        <ps name="BFMOPA-ZA.PP.ZZ-16" mylink="BFMOPA-ZA.PP.ZZ-16" enclabels="" sections="1" secttype="noheading">
          <pstext mayhavelinks="1" section="Decode" rep_section="decode">if !<a link="impl-aarch64.HaveSMEB16B16.0" file="shared_pseudocode.xml" hover="function: boolean HaveSMEB16B16()">HaveSMEB16B16</a>() then UNDEFINED;
integer a = <a link="impl-shared.UInt.1" file="shared_pseudocode.xml" hover="function: integer UInt(bits(N) x)">UInt</a>(Pn);
integer b = <a link="impl-shared.UInt.1" file="shared_pseudocode.xml" hover="function: integer UInt(bits(N) x)">UInt</a>(Pm);
integer n = <a link="impl-shared.UInt.1" file="shared_pseudocode.xml" hover="function: integer UInt(bits(N) x)">UInt</a>(Zn);
integer m = <a link="impl-shared.UInt.1" file="shared_pseudocode.xml" hover="function: integer UInt(bits(N) x)">UInt</a>(Zm);
integer da = <a link="impl-shared.UInt.1" file="shared_pseudocode.xml" hover="function: integer UInt(bits(N) x)">UInt</a>(ZAda);
boolean sub_op = FALSE;</pstext>
        </ps>
      </ps_section>
    </iclass>
  </classes>
  <explanations scope="all">
    <explanation enclist="bfmopa_za_pp_zz_16" symboldefcount="1">
      <symbol link="sa_zada">&lt;ZAda&gt;</symbol>
      <account encodedin="ZAda">
        <intro>
          <para>Is the name of the ZA tile ZA0-ZA1, encoded in the "ZAda" field.</para>
        </intro>
      </account>
    </explanation>
    <explanation enclist="bfmopa_za_pp_zz_16" symboldefcount="1">
      <symbol link="sa_pn">&lt;Pn&gt;</symbol>
      <account encodedin="Pn">
        <intro>
          <para>Is the name of the first governing scalable predicate register P0-P7, encoded in the "Pn" field.</para>
        </intro>
      </account>
    </explanation>
    <explanation enclist="bfmopa_za_pp_zz_16" symboldefcount="1">
      <symbol link="sa_pm">&lt;Pm&gt;</symbol>
      <account encodedin="Pm">
        <intro>
          <para>Is the name of the second governing scalable predicate register P0-P7, encoded in the "Pm" field.</para>
        </intro>
      </account>
    </explanation>
    <explanation enclist="bfmopa_za_pp_zz_16" symboldefcount="1">
      <symbol link="sa_zn">&lt;Zn&gt;</symbol>
      <account encodedin="Zn">
        <intro>
          <para>Is the name of the first source scalable vector register, encoded in the "Zn" field.</para>
        </intro>
      </account>
    </explanation>
    <explanation enclist="bfmopa_za_pp_zz_16" symboldefcount="1">
      <symbol link="sa_zm">&lt;Zm&gt;</symbol>
      <account encodedin="Zm">
        <intro>
          <para>Is the name of the second source scalable vector register, encoded in the "Zm" field.</para>
        </intro>
      </account>
    </explanation>
  </explanations>
  <ps_section howmany="1">
    <ps name="BFMOPA-ZA.PP.ZZ-16" mylink="execute" enclabels="" sections="1" secttype="Operation">
      <pstext mayhavelinks="1" section="Execute" rep_section="execute"><a link="impl-aarch64.CheckStreamingSVEAndZAEnabled.0" file="shared_pseudocode.xml" hover="function: CheckStreamingSVEAndZAEnabled()">CheckStreamingSVEAndZAEnabled</a>();
constant integer VL = <a link="impl-aarch64.CurrentVL.read.none" file="shared_pseudocode.xml" hover="accessor: integer CurrentVL">CurrentVL</a>;
constant integer PL = VL DIV 8;
constant integer dim = VL DIV 16;
bits(PL) mask1 = <a link="impl-aarch64.P.read.2" file="shared_pseudocode.xml" hover="accessor: bits(width) P[integer n, integer width]">P</a>[a, PL];
bits(PL) mask2 = <a link="impl-aarch64.P.read.2" file="shared_pseudocode.xml" hover="accessor: bits(width) P[integer n, integer width]">P</a>[b, PL];
bits(VL) operand1 = <a link="impl-aarch64.Z.read.2" file="shared_pseudocode.xml" hover="accessor: bits(width) Z[integer n, integer width]">Z</a>[n, VL];
bits(VL) operand2 = <a link="impl-aarch64.Z.read.2" file="shared_pseudocode.xml" hover="accessor: bits(width) Z[integer n, integer width]">Z</a>[m, VL];
bits(dim*dim*16) operand3 = <a link="impl-aarch64.ZAtile.read.3" file="shared_pseudocode.xml" hover="accessor: bits(width) ZAtile[integer tile, integer esize, integer width]">ZAtile</a>[da, 16, dim*dim*16];
bits(dim*dim*16) result;

for row = 0 to dim-1
    for col = 0 to dim-1
        bits(16) element1 = <a link="impl-shared.Elem.read.3" file="shared_pseudocode.xml" hover="accessor: bits(size) Elem[bits(N) vector, integer e, integer size]">Elem</a>[operand1, row, 16];
        bits(16) element2 = <a link="impl-shared.Elem.read.3" file="shared_pseudocode.xml" hover="accessor: bits(size) Elem[bits(N) vector, integer e, integer size]">Elem</a>[operand2, col, 16];
        bits(16) element3 = <a link="impl-shared.Elem.read.3" file="shared_pseudocode.xml" hover="accessor: bits(size) Elem[bits(N) vector, integer e, integer size]">Elem</a>[operand3, row*dim+col, 16];

        if <a link="impl-aarch64.ActivePredicateElement.3" file="shared_pseudocode.xml" hover="function: boolean ActivePredicateElement(bits(N) pred, integer e, integer esize)">ActivePredicateElement</a>(mask1, row, 16) &amp;&amp; <a link="impl-aarch64.ActivePredicateElement.3" file="shared_pseudocode.xml" hover="function: boolean ActivePredicateElement(bits(N) pred, integer e, integer esize)">ActivePredicateElement</a>(mask2, col, 16) then
            if sub_op then element1 = <a link="impl-shared.BFNeg.1" file="shared_pseudocode.xml" hover="function: bits(16) BFNeg(bits(16) op)">BFNeg</a>(element1);
            <a link="impl-shared.Elem.write.3" file="shared_pseudocode.xml" hover="accessor: Elem[bits(N) &amp;vector, integer e, integer size] = bits(size) value">Elem</a>[result, row*dim+col, 16] = <a link="impl-shared.BFMulAdd_ZA.4" file="shared_pseudocode.xml" hover="function: bits(16) BFMulAdd_ZA(bits(16) addend, bits(16) op1, bits(16) op2, FPCRType fpcr_in)">BFMulAdd_ZA</a>(element3, element1, element2, FPCR[]);
        else
            <a link="impl-shared.Elem.write.3" file="shared_pseudocode.xml" hover="accessor: Elem[bits(N) &amp;vector, integer e, integer size] = bits(size) value">Elem</a>[result, row*dim+col, 16] = element3;

<a link="impl-aarch64.ZAtile.write.3" file="shared_pseudocode.xml" hover="accessor: ZAtile[integer tile, integer esize, integer width] = bits(width) value">ZAtile</a>[da, 16, dim*dim*16] = result;</pstext>
    </ps>
  </ps_section>
</instructionsection>