conferences/fosdem2024/fosdem2024_ddffirst/fosdem2024_ddffirst.tex

   1 \documentclass[slidestop]{beamer}
   2 \usepackage{beamerthemesplit}
   3 \usepackage{graphics}
   4 \usepackage{pstricks}
   5 \usepackage{pgffor}
   6 \usepackage{listings}
   7
   8 \graphicspath{{./}}
   9
  10 \title{Data-Dependent-Fail-First}
  11 \author{Luke Kenneth Casson Leighton and Shriya Sharma}
  12
  13
  14 \begin{document}
  15
  16 \frame{
  17    \begin{center}
  18     \huge{The Libre-SOC Hybrid 3D CPU}\\
  19     \vspace{32pt}
  20     \Large{Data-Dependent-Fail-First}\\
  21
  22     \vspace{24pt}
  23     \Large{FOSDEM2024}\\
  24     \vspace{16pt}
  25     \large{Sponsored by NLnet's PET Programme}\\
  26     \vspace{6pt}
  27     \large{\today}
  28   \end{center}
  29 }
  30
  31
  32 \frame{\frametitle{Why another SoC?}
  33
  34  \begin{itemize}
  35    \item Intel Management Engine, Apple QA issues, Spectre\vspace{6pt}
  36    \item Endless proprietary drivers, "simplest" solution: \\
  37          License proprietary hard macros (with proprietary firmware)\\
  38                  Adversely affects product development cost\\
  39                 due to opaque driver bugs (Samsung S3C6410 / S5P100)
  40                  \vspace{6pt}
  41    \item Alternative: Intel and Valve-Steam collaboration\\
  42          "Most productive business meeting ever!"\\
  43          https://tinyurl.com/valve-steam-intel
  44                 \vspace{6pt}
  45    \item Because for 30 years I Always Wanted To Design A CPU
  46                 \vspace{6pt}
  47    \item Ultimately it is a strategic \textit{business} objective to
  48          develop entirely Libre hardware, firmware and drivers.
  49   \end{itemize}
  50 }
  51
  52
  53
  54 \frame{\frametitle{How can you help?}
  55
  56 \vspace{5pt}
  57
  58  \begin{itemize}
  59    \item Start here! https://libre-soc.org \\
  60              Mailing lists https://lists.libre-soc.org \\
  61              IRC Freenode libre-soc \\
  62              etc. etc. (it's a Libre project, go figure) \\
  63                    \vspace{3pt}
  64    \item Can I get paid? Yes!  NLnet funded\\
  65                  See https://libre-soc.org/nlnet/\#faq \\
  66                  \vspace{3pt}
  67    \item Also profit-sharing in any commercial ventures \\
  68              \vspace{3pt}
  69    \item How many opportunities to develop Libre SoCs exist,\\
  70              and actually get paid for it?
  71                      \vspace{3pt}
  72    \item I'm not a developer, how can I help?\\
  73                 - Plenty of research needed, artwork, website \\
  74                 - Help find customers and OEMs willing to commit (LOI)
  75   \end{itemize}
  76 }
  77
  78
  79
  80 \frame{\frametitle{What goes into a typical SoC?}
  81 \vspace{9pt}
  82  \begin{itemize}
  83    \item 15 to 20mm BGA package: 2.5 to 5 watt power consumption\\
  84                 heat sink normally not required (simplifies overall design)
  85                 \vspace{3pt}
  86    \item Fully-integrated peripherals (not Northbridge/Southbridge)\\
  87          USB, HDMI, RGB/TTL, SD/MMC, I2C, UART, SPI, GPIO etc. etc.
  88          \vspace{3pt}
  89    \item Built-in GPU (shared memory bus, 3rd party licensed) \vspace{3pt}
  90    \item Built-in VPU (likewise, proprietary)\vspace{3pt}
  91    \item Target price between \$2.50 and \$30 depending on market\\
  92          Radically different from IBM POWER9 Core (200 Watt)
  93          \vspace{3pt}
  94    \item We're doing the same, just with a hybrid architecture.\\
  95                  CPU == GPU == VPU
  96   \end{itemize}
  97 }
  98
  99
 100
 101 \begin{frame}[fragile]
 102 \frametitle{Simple-V CMPI in a nutshell}
 103
 104 \begin{semiverbatim}
 105 function op\_cmpi(BA, RA, SI) # cmpi not vector-cmpi!
 106   (assuming you know power-isa)
 107   int i, id=0, ira=0;
 108   for (i = 0; i < VL; i++)
 109     CR[BA+id] <= compare(ireg[RA+ira], SI);
 110     if (reg\_is\_vectorised[BA] ) \{ id += 1; \}
 111     if (reg\_is\_vectorised[RA])  \{ ira += 1; \}
 112 \end{semiverbatim}
 113
 114   \begin{itemize}
 115    \item Above is oversimplified: predication etc. left out
 116    \item Scalar-scalar and scalar-vector and vector-vector now all in one
 117    \item OoO may choose to push CMPIs into instr. queue (v. busy!)
 118   \end{itemize}
 119 \end{frame}
 120
 121
 122 \frame{\frametitle{Load/Store Fault-First}
 123
 124         \begin{itemize}
 125                 \item Problem: vector load and store can cause a page fault
 126                 \item Solution: a protocol that allows optional load/store
 127                 \item instruction \textit{requests} a number of elements
 128                 \item instruction \textit{informs} the number actually loaded
 129                 \item first element load/store is not optional (cannot fail)
 130         \item ARM SVE: https://arxiv.org/pdf/1803.06185.pdf
 131         \item more: wikipedia Vector processor page: Fault/Fail First
 132         \vspace{10pt}
 133                 \item Load/Store is Memory to/from Register, what about
 134               Register to Register?
 135         \item Register-to-register: "Data-Dependent Fail-First."
 136         \item Z80 LDIR: Mem-Register, CPIR: Register-Register
 137         \end{itemize}
 138 }
 139
 140 \begin{frame}[fragile]
 141         \frametitle{Data-Dependent-Fail-First in a nutshell}
 142
 143         \begin{semiverbatim}
 144 function op\_cmpi(BA, RA, SI) # cmpi not vector-cmpi!
 145 int i, id=0, ira=0;
 146 for (i = 0; i < VL; i++)
 147     CR[BA+id] <= compare(ireg[RA+ira], SI);
 148     if (reg\_is\_vectorised[BA] ) \{ id += 1; \}
 149     if (reg\_is\_vectorised[RA])  \{ ira += 1; \}
 150     if test (CR[BA+id]) == FAIL: \{ VL = i + 1; break \}
 151         \end{semiverbatim}
 152
 153         \begin{itemize}
 154                 \item Parallelism still perfectly possible
 155                       ("hold" writing results until sequential post-analysis
 156                        carried out. Best done with OoO)
 157                 \item VL truncation can be inclusive or exclusive
 158                       (include or exclude a NULL pointer or a
 159                       string-end character, or overflow result)
 160                 \item \textit{Truncation can be to zero Vector Length}
 161         \end{itemize}
 162 \end{frame}
 163
 164 \frame{\frametitle{Power ISA v3.1 vstribr}
 165
 166         \lstinputlisting[language={}]{vstribr.txt}
 167
 168         \begin{itemize}
 169                 \item ironically this hard-coded instruction is
 170                 identical to general-purpose Simple-V DD-FFirst...
 171         \end{itemize}
 172
 173 }Po
 174
 175 \frame{\frametitle{maxloc}
 176   \begin{itemize}
 177                 \item "TODO
 178   \end{itemize}
 179 }
 180
 181 \frame{\frametitle{Pospopcount}
 182
 183   \begin{itemize}
 184         \item   Positional popcount adds up the totals of each bit set to 1 in each bit-position, of an array of input values.
 185         \item   Notoriously difficult to do in SIMD assembler: typically 550 lines
 186     \item https://github.com/clausecker/pospop
 187
 188    \end{itemize}
 189
 190         \lstinputlisting[language={}]{pospopcount.c}
 191
 192
 193 }
 194
 195 \frame{\frametitle{Pospopcount}
 196
 197         \begin{center}
 198                 \includegraphics[width=0.5\textwidth]{pospopcount.png}
 199         \end{center}
 200           \begin{itemize}
 201                 \item   The challenge is to perform an appropriate transpose of the data (the CPU can only work on registers, horizontally),
 202                 in blocks that suit the processor and the ISA capacity.
 203
 204
 205         \end{itemize}
 206 }
 207
 208 \frame{\frametitle{Pospopcount}
 209
 210         \begin{center}
 211                 \includegraphics[width=0.6\textwidth]{array_popcnt.png}
 212         \end{center}
 213
 214   \begin{itemize}
 215
 216                 \item   The draft gbbd instruction implements the transpose (shown above),
 217                                 preparing the data to use the standard popcount instruction.
 218                            (gbbd is based on Power ISA vgbbd)
 219
 220         \end{itemize}
 221
 222 }
 223
 224 \frame{\frametitle{Pospopcount.s}
 225
 226
 227 \lstinputlisting[language={}]{pospopcount.s}
 228
 229 }
 230
 231
 232 \frame{\frametitle{strncpy}
 233
 234         \lstinputlisting[language={}]{strncpy.c}
 235   \begin{itemize}
 236         \item "TODO
 237  \end{itemize}
 238 }
 239
 240
 241
 242 \frame{\frametitle{strncpy assembler}
 243
 244 \lstinputlisting[language={}]{strncpy.s}
 245
 246 }
 247
 248 \frame{\frametitle{linked-list walking}
 249   \begin{itemize}
 250         \item "TODO
 251  \end{itemize}
 252 }
 253 \frame{\frametitle{Summary}
 254
 255  \begin{itemize}
 256    \item Goal is to create a mass-volume low-power embedded SoC suitable
 257          for use in netbooks, chromebooks, tablets, smartphones, IoT SBCs.
 258    \item No way we could implement a project of this magnitude without
 259          nmigen (being able to use python OO to  HDL)
 260    \item Collaboration with OpenPOWER Foundation and Members absolutely
 261          essential. No short-cuts.  Standards to be developed and ratified
 262          so that everyone benefits.
 263    \item Riding the wave of huge stability of OpenPOWER ecosystem
 264    \item Greatly simplified open 3D and Video drivers reduces product
 265          development costs for customers
 266    \item It also happens to be fascinating, deeply rewarding technically
 267          challenging, and funded by NLnet
 268
 269   \end{itemize}
 270 }
 271
 272
 273 \frame{
 274   \begin{center}
 275     {\Huge The end\vspace{12pt}\\
 276                    Thank you\vspace{12pt}\\
 277                    Questions?\vspace{12pt}
 278         }
 279   \end{center}
 280
 281   \begin{itemize}
 282         \item Discussion: http://lists.libre-soc.org
 283         \item Freenode IRC \#libre-soc
 284         \item http://libre-soc.org/
 285         \item http://nlnet.nl/PET
 286         \item https://libre-soc.org/nlnet/\#faq
 287   \end{itemize}
 288 }
 289
 290
 291 \end{document}