Files
archived-ballistic/spec/arm64_xml/smmla_z_zzz.xml
Ronald Caesar 26a677f8b4 decoder: Add ARM specification docs
Signed-off-by: Ronald Caesar <github43132@proton.me>
2025-12-12 18:11:36 -04:00

146 lines
8.3 KiB
XML
Raw Permalink Blame History

This file contains ambiguous Unicode characters

This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

<?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="smmla_z_zzz" title="SMMLA" type="instruction">
<docvars>
<docvar key="instr-class" value="sve" />
<docvar key="isa" value="A64" />
<docvar key="mnemonic" value="SMMLA" />
</docvars>
<heading>SMMLA</heading>
<desc>
<brief>Signed integer matrix multiply-accumulate</brief>
<description>
<para>The signed integer matrix multiply-accumulate instruction multiplies the 2×8 matrix of signed 8-bit integer values held in each 128-bit segment of the first source vector by the 8×2 matrix of signed 8-bit integer values in the corresponding segment of the second source vector. The resulting 2×2 widened 32-bit integer matrix product is then destructively added to the 32-bit integer matrix accumulator held in the corresponding segment of the addend and destination vector. This is equivalent to performing an 8-way dot product per destination element.</para>
<para>This instruction is unpredicated.</para>
<para>ID_AA64ZFR0_EL1.I8MM indicates whether this instruction is implemented.</para>
<para>This instruction is illegal when executed in Streaming SVE mode, unless FEAT_SME_FA64 is implemented and enabled.</para>
</description>
<status>Green</status>
<predicated>False</predicated>
<takes_movprfx>True</takes_movprfx>
<sm_policy>SM_0_only</sm_policy>
</desc>
<alias_list howmany="0"></alias_list>
<classes>
<iclass name="SVE" oneof="1" id="iclass_sve" no_encodings="1" isa="A64">
<docvars>
<docvar key="instr-class" value="sve" />
<docvar key="isa" value="A64" />
<docvar key="mnemonic" value="SMMLA" />
</docvars>
<iclassintro count="1"></iclassintro>
<arch_variants>
<arch_variant name="ARMv8.6" feature="FEAT_I8MM" />
</arch_variants>
<regdiagram form="32" psname="SMMLA-Z.ZZZ-_" tworows="1">
<box hibit="31" width="8" settings="8">
<c>0</c>
<c>1</c>
<c>0</c>
<c>0</c>
<c>0</c>
<c>1</c>
<c>0</c>
<c>1</c>
</box>
<box hibit="23" name="uns&lt;1&gt;" usename="1" settings="1">
<c>0</c>
</box>
<box hibit="22" name="uns&lt;0&gt;" usename="1" settings="1">
<c>0</c>
</box>
<box hibit="21" settings="1">
<c>0</c>
</box>
<box hibit="20" width="5" name="Zm" usename="1">
<c colspan="5"></c>
</box>
<box hibit="15" width="6" settings="6">
<c>1</c>
<c>0</c>
<c>0</c>
<c>1</c>
<c>1</c>
<c>0</c>
</box>
<box hibit="9" width="5" name="Zn" usename="1">
<c colspan="5"></c>
</box>
<box hibit="4" width="5" name="Zda" usename="1">
<c colspan="5"></c>
</box>
</regdiagram>
<encoding name="smmla_z_zzz_" oneofinclass="1" oneof="1" label="">
<docvars>
<docvar key="instr-class" value="sve" />
<docvar key="isa" value="A64" />
<docvar key="mnemonic" value="SMMLA" />
</docvars>
<asmtemplate><text>SMMLA </text><a link="sa_zda" hover="Third source and destination scalable vector register (field &quot;Zda&quot;)">&lt;Zda&gt;</a><text>.S, </text><a link="sa_zn" hover="First source scalable vector register (field &quot;Zn&quot;)">&lt;Zn&gt;</a><text>.B, </text><a link="sa_zm" hover="Second source scalable vector register (field &quot;Zm&quot;)">&lt;Zm&gt;</a><text>.B</text></asmtemplate>
</encoding>
<ps_section howmany="1">
<ps name="SMMLA-Z.ZZZ-_" mylink="SMMLA-Z.ZZZ-_" enclabels="" sections="1" secttype="noheading">
<pstext mayhavelinks="1" section="Decode" rep_section="decode">if !<a link="impl-aarch64.HaveSVE.0" file="shared_pseudocode.xml" hover="function: boolean HaveSVE()">HaveSVE</a>() || !<a link="impl-shared.HaveInt8MatMulExt.0" file="shared_pseudocode.xml" hover="function: boolean HaveInt8MatMulExt()">HaveInt8MatMulExt</a>() then UNDEFINED;
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>(Zda);
boolean op1_unsigned = FALSE;
boolean op2_unsigned = FALSE;</pstext>
</ps>
</ps_section>
</iclass>
</classes>
<explanations scope="all">
<explanation enclist="smmla_z_zzz_" symboldefcount="1">
<symbol link="sa_zda">&lt;Zda&gt;</symbol>
<account encodedin="Zda">
<intro>
<para>Is the name of the third source and destination scalable vector register, encoded in the "Zda" field.</para>
</intro>
</account>
</explanation>
<explanation enclist="smmla_z_zzz_" 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="smmla_z_zzz_" 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="SMMLA-Z.ZZZ-_" mylink="execute" enclabels="" sections="1" secttype="Operation">
<pstext mayhavelinks="1" section="Execute" rep_section="execute"><a link="impl-aarch64.CheckNonStreamingSVEEnabled.0" file="shared_pseudocode.xml" hover="function: CheckNonStreamingSVEEnabled()">CheckNonStreamingSVEEnabled</a>();
constant integer VL = <a link="impl-aarch64.CurrentVL.read.none" file="shared_pseudocode.xml" hover="accessor: integer CurrentVL">CurrentVL</a>;
constant integer segments = VL DIV 128;
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(VL) operand3 = <a link="impl-aarch64.Z.read.2" file="shared_pseudocode.xml" hover="accessor: bits(width) Z[integer n, integer width]">Z</a>[da, VL];
bits(VL) result = <a link="impl-shared.Zeros.1" file="shared_pseudocode.xml" hover="function: bits(N) Zeros(integer N)">Zeros</a>(VL);
bits(128) op1, op2;
bits(128) res, addend;
for s = 0 to segments-1
op1 = <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, s, 128];
op2 = <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, s, 128];
addend = <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, s, 128];
res = <a link="impl-shared.MatMulAdd.5" file="shared_pseudocode.xml" hover="function: bits(N) MatMulAdd(bits(N) addend, bits(N) op1, bits(N) op2, boolean op1_unsigned, boolean op2_unsigned)">MatMulAdd</a>(addend, op1, op2, op1_unsigned, op2_unsigned);
<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, s, 128] = res;
<a link="impl-aarch64.Z.write.2" file="shared_pseudocode.xml" hover="accessor: Z[integer n, integer width] = bits(width) value">Z</a>[da, VL] = result;</pstext>
</ps>
</ps_section>
</instructionsection>