collab/polyglot-riscv-arm

what if the same machine code was valid riscv and arm asm

git clone https://git.julia.blue/collab/polyglot-riscv-arm

Jonathan Dönszelmannwebsite start885cc4e

main
9.7 KiB68 linesraw
1<?xml version="1.0" encoding="utf-8"?>
2<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.1//EN" "http://www.w3.org/TR/xhtml11/DTD/xhtml11.dtd">
3<html xmlns="http://www.w3.org/1999/xhtml"><head><link rel="stylesheet" type="text/css" href="insn.css"/><meta name="generator" content="iform.xsl"/><title>VMMLA -- AArch32</title></head><body><table style="margin: 0 auto;"><tr><td><div class="topbar"><a href="index.html">Base Instructions</a></div></td><td><div class="topbar"><a href="fpsimdindex.html">SIMD&amp;FP Instructions</a></div></td><td><div class="topbar"><a href="a32_encindex.html">A32 Instructions by Encoding</a></div></td><td><div class="topbar"><a href="t32_encindex.html">T32 Instructions by Encoding</a></div></td><td><div class="topbar"><a href="shared_pseudocode.html">Shared Pseudocode</a></div></td><td><div class="topbar"><a href="notice.html">Proprietary Notice</a></div></td></tr></table><hr/><h2 class="instruction-section">VMMLA</h2>
4      <p class="aml">BFloat16 floating-point matrix multiply-accumulate. This instruction multiplies the 2x4 matrix of BF16 values in the first 128-bit source vector by the 4x2 BF16 matrix in the second 128-bit source vector. The resulting 2x2 single-precision matrix product is then added destructively to the 2x2 single-precision matrix in the 128-bit destination vector. This is equivalent to performing a 4-way dot product per destination element. The instruction does not update the <a class="armarm-xref" title="Reference to Armv8 ARM section" data-linkend="AArch32.fpscr">FPSCR</a> exception status.</p>
5      <div class="note"><hr class="note"/><h4>Note</h4>
6        <p class="aml">Arm expects that the VMMLA instruction will deliver a peak BF16 multiply throughput that is at least as high as can be achieved using two VDOT instructions, with a goal that it should have significantly higher throughput.</p>
7      <hr class="note"/></div>
8    
9    <p class="desc">
10      It has encodings from the following instruction sets:
11       A32 (
12      <a href="#iclass_a1">A1</a>
13      )
14       and 
15       T32 (
16      <a href="#iclass_t1">T1</a>
17      )
18      .
19    </p>
20    <h3 class="classheading"><a id="iclass_a1"/>A1<span style="font-size:smaller;"><br/>(FEAT_AA32BF16)
21          </span></h3><div class="regdiagram-32"><table class="regdiagram"><thead><tr><td>31</td><td>30</td><td>29</td><td>28</td><td>27</td><td>26</td><td>25</td><td>24</td><td>23</td><td>22</td><td>21</td><td>20</td><td>19</td><td>18</td><td>17</td><td>16</td><td>15</td><td>14</td><td>13</td><td>12</td><td>11</td><td>10</td><td>9</td><td>8</td><td>7</td><td>6</td><td>5</td><td>4</td><td>3</td><td>2</td><td>1</td><td>0</td></tr></thead><tbody><tr class="firstrow"><td class="l">1</td><td>1</td><td>1</td><td>1</td><td>1</td><td>1</td><td class="r">0</td><td class="l">0</td><td class="r">0</td><td class="lr">D</td><td class="lr">0</td><td class="lr">0</td><td colspan="4" class="lr">Vn</td><td colspan="4" class="lr">Vd</td><td class="lr">1</td><td class="lr">1</td><td class="lr">0</td><td class="lr">0</td><td class="lr">N</td><td class="lr">1</td><td class="lr">M</td><td class="lr">0</td><td colspan="4" class="lr">Vm</td></tr></tbody></table></div><div class="encoding"><h4 class="encoding">A1</h4><a id="VMMLA_A1_Q"/><p class="asm-code">VMMLA{<a href="#sa_q" title="See {xref{ARMARM_Babbefhf}{Standard assembler syntax fields}}">&lt;q&gt;</a>}.BF16 <a href="#sa_qd" title="128-bit SIMD&amp;FP destination register (field &quot;D:Vd&quot;)">&lt;Qd&gt;</a>, <a href="#sa_qn" title="First 128-bit SIMD&amp;FP source register (field &quot;N:Vn&quot;)">&lt;Qn&gt;</a>, <a href="#sa_qm" title="Second 128-bit SIMD&amp;FP source register (field &quot;M:Vm&quot;)">&lt;Qm&gt;</a></p></div><p class="pseudocode">if !IsFeatureImplemented(FEAT_AA32BF16) then UNDEFINED;
22if Vd&lt;0&gt; == '1' || Vn&lt;0&gt; == '1' || Vm&lt;0&gt; == '1' then UNDEFINED;
23constant integer d = <a href="shared_pseudocode.html#impl-shared.UInt.1" title="function: integer UInt(bits(N) x)">UInt</a>(D:Vd);
24constant integer n = <a href="shared_pseudocode.html#impl-shared.UInt.1" title="function: integer UInt(bits(N) x)">UInt</a>(N:Vn);
25constant integer m = <a href="shared_pseudocode.html#impl-shared.UInt.1" title="function: integer UInt(bits(N) x)">UInt</a>(M:Vm);
26constant integer regs = 2;</p>
27    <h3 class="classheading"><a id="iclass_t1"/>T1<span style="font-size:smaller;"><br/>(FEAT_AA32BF16)
28          </span></h3><div class="regdiagram-16x2"><table class="regdiagram"><thead><tr><td>15</td><td>14</td><td>13</td><td>12</td><td>11</td><td>10</td><td>9</td><td>8</td><td>7</td><td>6</td><td>5</td><td>4</td><td>3</td><td>2</td><td>1</td><td>0</td><td>15</td><td>14</td><td>13</td><td>12</td><td>11</td><td>10</td><td>9</td><td>8</td><td>7</td><td>6</td><td>5</td><td>4</td><td>3</td><td>2</td><td>1</td><td>0</td></tr></thead><tbody><tr class="firstrow"><td class="l">1</td><td>1</td><td>1</td><td>1</td><td>1</td><td>1</td><td class="r">0</td><td class="l">0</td><td class="r">0</td><td class="lr">D</td><td class="lr">0</td><td class="lr">0</td><td colspan="4" class="lr">Vn</td><td colspan="4" class="lr">Vd</td><td class="lr">1</td><td class="lr">1</td><td class="lr">0</td><td class="lr">0</td><td class="lr">N</td><td class="lr">1</td><td class="lr">M</td><td class="lr">0</td><td colspan="4" class="lr">Vm</td></tr></tbody></table></div><div class="encoding"><h4 class="encoding">T1</h4><a id="VMMLA_T1_Q"/><p class="asm-code">VMMLA{<a href="#sa_q" title="See {xref{ARMARM_Babbefhf}{Standard assembler syntax fields}}">&lt;q&gt;</a>}.BF16 <a href="#sa_qd" title="128-bit SIMD&amp;FP destination register (field &quot;D:Vd&quot;)">&lt;Qd&gt;</a>, <a href="#sa_qn" title="First 128-bit SIMD&amp;FP source register (field &quot;N:Vn&quot;)">&lt;Qn&gt;</a>, <a href="#sa_qm" title="Second 128-bit SIMD&amp;FP source register (field &quot;M:Vm&quot;)">&lt;Qm&gt;</a></p></div><p class="pseudocode">if <a href="shared_pseudocode.html#impl-aarch32.InITBlock.0" title="function: boolean InITBlock()">InITBlock</a>() then UNPREDICTABLE;
29if !IsFeatureImplemented(FEAT_AA32BF16) then UNDEFINED;
30if Vd&lt;0&gt; == '1' || Vn&lt;0&gt; == '1' || Vm&lt;0&gt; == '1' then UNDEFINED;
31constant integer d = <a href="shared_pseudocode.html#impl-shared.UInt.1" title="function: integer UInt(bits(N) x)">UInt</a>(D:Vd);
32constant integer n = <a href="shared_pseudocode.html#impl-shared.UInt.1" title="function: integer UInt(bits(N) x)">UInt</a>(N:Vn);
33constant integer m = <a href="shared_pseudocode.html#impl-shared.UInt.1" title="function: integer UInt(bits(N) x)">UInt</a>(M:Vm);
34constant integer regs = 2;</p>
35  <div class="encoding-notes"/><h3 class="explanations">Assembler Symbols</h3><div class="explanations"><table><col class="asyn-l"/><col class="asyn-r"/><tr><td>&lt;q&gt;</td><td><a id="sa_q"/>
36        
37          <p class="aml">See <a class="armarm-xref" title="Reference to Armv8 ARM section" data-linkend="Babbefhf">Standard assembler syntax fields</a>.</p>
38        
39      </td></tr></table><table><col class="asyn-l"/><col class="asyn-r"/><tr><td>&lt;Qd&gt;</td><td><a id="sa_qd"/>
40        
41          <p class="aml">Is the 128-bit name of the SIMD&amp;FP destination register, encoded in the "D:Vd" field as &lt;Qd&gt;*2.</p>
42        
43      </td></tr></table><table><col class="asyn-l"/><col class="asyn-r"/><tr><td>&lt;Qn&gt;</td><td><a id="sa_qn"/>
44        
45          <p class="aml">Is the 128-bit name of the first SIMD&amp;FP source register, encoded in the "N:Vn" field as &lt;Qn&gt;*2.</p>
46        
47      </td></tr></table><table><col class="asyn-l"/><col class="asyn-r"/><tr><td>&lt;Qm&gt;</td><td><a id="sa_qm"/>
48        
49          <p class="aml">Is the 128-bit name of the second SIMD&amp;FP source register, encoded in the "M:Vm" field as &lt;Qm&gt;*2.</p>
50        
51      </td></tr></table></div><div class="syntax-notes"/>
52    <div class="ps"><a id="execute"/><h3 class="pseudocode">Operation</h3>
53      <p class="pseudocode"><a href="shared_pseudocode.html#impl-aarch32.CheckAdvSIMDEnabled.0" title="function: CheckAdvSIMDEnabled()">CheckAdvSIMDEnabled</a>();
54
55constant bits(128) op1 = <a href="shared_pseudocode.html#impl-aarch32.Q.read.1" title="accessor: bits(128) Q[integer n]">Q</a>[n&gt;&gt;1];
56constant bits(128) op2 = <a href="shared_pseudocode.html#impl-aarch32.Q.read.1" title="accessor: bits(128) Q[integer n]">Q</a>[m&gt;&gt;1];
57constant bits(128) acc = <a href="shared_pseudocode.html#impl-aarch32.Q.read.1" title="accessor: bits(128) Q[integer n]">Q</a>[d&gt;&gt;1];
58constant <a href="shared_pseudocode.html#FPCR_Type" title="type FPCR_Type">FPCR_Type</a> fpcr = <a href="shared_pseudocode.html#impl-shared.EffectiveFPCR.0" title="function: FPCR_Type EffectiveFPCR()">EffectiveFPCR</a>();
59
60<a href="shared_pseudocode.html#impl-aarch32.Q.write.1" title="accessor: Q[integer n] = bits(128) value">Q</a>[d&gt;&gt;1] = <a href="shared_pseudocode.html#impl-shared.BFMatMulAdd.4" title="function: bits(N) BFMatMulAdd(bits(N) addend, bits(N) op1, bits(N) op2, FPCR_Type fpcr)">BFMatMulAdd</a>(acc, op1, op2, fpcr);</p>
61    </div>
62  <hr/><table style="margin: 0 auto;"><tr><td><div class="topbar"><a href="index.html">Base Instructions</a></div></td><td><div class="topbar"><a href="fpsimdindex.html">SIMD&amp;FP Instructions</a></div></td><td><div class="topbar"><a href="a32_encindex.html">A32 Instructions by Encoding</a></div></td><td><div class="topbar"><a href="t32_encindex.html">T32 Instructions by Encoding</a></div></td><td><div class="topbar"><a href="shared_pseudocode.html">Shared Pseudocode</a></div></td><td><div class="topbar"><a href="notice.html">Proprietary Notice</a></div></td></tr></table><p class="versions">
63      Internal version only: isa v01_31, pseudocode v2024-09_rel
64      ; Build timestamp: 2024-09-26T13:26
65    </p><p class="copyconf">
66      Copyright © 2010-2024 Arm Limited or its affiliates. All rights reserved.
67      This document is Non-Confidential.
68    </p></body></html>