\frame{
\begin{center}
- \huge{The Libre-SOC Hybrid 3D CPU}\\
- \vspace{32pt}
- \Large{Data-Dependent-Fail-First}\\
+ \huge{Libre-SOC Simple-V Specification \\
+ Advanced features}\\
+ \vspace{24pt}
+ \Large{Data-Dependent Fail-First}\\
\vspace{24pt}
\Large{FOSDEM2024}\\
\vspace{16pt}
- \large{Sponsored by NLnet's PET Programme}\\
+ \large{Funded by NLnet NGI-ASSURE \\
+ EU grant agreement No 957073}\\
\vspace{6pt}
\large{\today}
\end{center}
}
-
-\frame{\frametitle{Why another SoC?}
-
- \begin{itemize}
- \item Intel Management Engine, Apple QA issues, Spectre\vspace{6pt}
- \item Endless proprietary drivers, "simplest" solution: \\
- License proprietary hard macros (with proprietary firmware)\\
- Adversely affects product development cost\\
- due to opaque driver bugs (Samsung S3C6410 / S5P100)
- \vspace{6pt}
- \item Alternative: Intel and Valve-Steam collaboration\\
- "Most productive business meeting ever!"\\
- https://tinyurl.com/valve-steam-intel
- \vspace{6pt}
- \item Because for 30 years I Always Wanted To Design A CPU
- \vspace{6pt}
- \item Ultimately it is a strategic \textit{business} objective to
- develop entirely Libre hardware, firmware and drivers.
- \end{itemize}
-}
-
-
-
-\frame{\frametitle{How can you help?}
-
-\vspace{5pt}
-
- \begin{itemize}
- \item Start here! https://libre-soc.org \\
- Mailing lists https://lists.libre-soc.org \\
- IRC Freenode libre-soc \\
- etc. etc. (it's a Libre project, go figure) \\
- \vspace{3pt}
- \item Can I get paid? Yes! NLnet funded\\
- See https://libre-soc.org/nlnet/\#faq \\
- \vspace{3pt}
- \item Also profit-sharing in any commercial ventures \\
- \vspace{3pt}
- \item How many opportunities to develop Libre SoCs exist,\\
- and actually get paid for it?
- \vspace{3pt}
- \item I'm not a developer, how can I help?\\
- - Plenty of research needed, artwork, website \\
- - Help find customers and OEMs willing to commit (LOI)
- \end{itemize}
-}
-
-
-
-\frame{\frametitle{What goes into a typical SoC?}
-\vspace{9pt}
- \begin{itemize}
- \item 15 to 20mm BGA package: 2.5 to 5 watt power consumption\\
- heat sink normally not required (simplifies overall design)
- \vspace{3pt}
- \item Fully-integrated peripherals (not Northbridge/Southbridge)\\
- USB, HDMI, RGB/TTL, SD/MMC, I2C, UART, SPI, GPIO etc. etc.
- \vspace{3pt}
- \item Built-in GPU (shared memory bus, 3rd party licensed) \vspace{3pt}
- \item Built-in VPU (likewise, proprietary)\vspace{3pt}
- \item Target price between \$2.50 and \$30 depending on market\\
- Radically different from IBM POWER9 Core (200 Watt)
- \vspace{3pt}
- \item We're doing the same, just with a hybrid architecture.\\
- CPU == GPU == VPU
- \end{itemize}
-}
-
-
-
-\frame{\frametitle{Simple SBC-style SoC}
-
-\begin{center}
-\includegraphics[width=0.6\textwidth]{pospopcount.png}
-\end{center}
-
-}
-
-
-
-
\begin{frame}[fragile]
\frametitle{Simple-V CMPI in a nutshell}
\end{frame}
-
-\begin{frame}[fragile]
-\frametitle{Data-Dependent-Fail-First in a nutshell}
-
-\begin{semiverbatim}
- while srcstep < VL - 1
- RT[SS], Rc = DO (MAXS, RA[SS], RB)
- if test (Rc) == FAIL: VL = srcstep + 1, break
-
-\end{semiverbatim}
-
-\begin{itemize}
- \item Above is oversimplified: predication etc. left out
- \item Scalar-scalar and scalar-vector and vector-vector now all in one
- \item OoO may choose to push CMPIs into instr. queue (v. busy!)
-\end{itemize}
-\end{frame}
-
\frame{\frametitle{Load/Store Fault-First}
-
- \begin{itemize}
- \item Problem: vector load and store can cause a page fault
- \item Solution: a protocol that allows optional load/store
- \item instruction \textit{requests} a number of elements
- \item instruction \textit{informs} the number actually loaded
- \item first load/store is not optional
- \vspace{20pt}
- \item Load/Store is Memory to/from Register, what about
+
+ \begin{itemize}
+ \item Problem: vector load and store can cause a page fault
+ \item Solution: a protocol that allows optional load/store
+ \item instruction \textit{requests} a number of elements
+ \item instruction \textit{informs} the number actually loaded
+ \item first element load/store is not optional (cannot fail)
+ \item ARM SVE: https://arxiv.org/pdf/1803.06185.pdf
+ \item more: wikipedia Vector processor page: Fault/Fail First
+ \vspace{10pt}
+ \item Load/Store is Memory to/from Register, what about
Register to Register?
\item Register-to-register: "Data-Dependent Fail-First."
\item Z80 LDIR: Mem-Register, CPIR: Register-Register
- \end{itemize}
+ \end{itemize}
}
+\begin{frame}[fragile]
+ \frametitle{Data-Dependent-Fail-First in a nutshell}
+ \begin{semiverbatim}
+function op\_cmpi(BA, RA, SI) # cmpi not vector-cmpi!
+int i, id=0, ira=0;
+for (i = 0; i < VL; i++)
+ CR[BA+id] <= compare(ireg[RA+ira], SI);
+ if (reg\_is\_vectorised[BA] ) \{ id += 1; \}
+ if (reg\_is\_vectorised[RA]) \{ ira += 1; \}
+ if test (CR[BA+id]) == FAIL: \{ VL = i + 1; break \}
+ \end{semiverbatim}
+
+ \begin{itemize}
+ \item Parallelism still perfectly possible
+ ("hold" writing results until sequential post-analysis
+ carried out. Best done with OoO)
+ \item VL truncation can be inclusive or exclusive
+ (include or exclude a NULL pointer or a
+ string-end character, or overflow result)
+ \item \textit{Truncation can be to zero Vector Length}
+ \end{itemize}
+\end{frame}
-\frame{\frametitle{Additional Simple-V features}
+\frame{\frametitle{Power ISA v3.1 vstribr}
+
+ \lstinputlisting[language={}]{vstribr.txt}
+
+ \begin{itemize}
+ \item ironically this hard-coded instruction is
+ identical to general-purpose Simple-V DD-FFirst...
+ \end{itemize}
- \begin{itemize}
- \item "fail-on-first" (POWER9 VSX strncpy segfaults on boundary!)
- \item "Twin Predication" (covers VSPLAT, VGATHER, VSCATTER, VINDEX etc.)
- \item SVP64: extensive "tag" (Vector context) augmentation
- \item "Context propagation": a VLIW-like context. Allows contexts
- to be repeatedly applied.
- Effectively a "hardware compression algorithm" for ISAs.
- \item Ultimate goal: cut down I-Cache usage, cuts down on power
- \item Typical GPU has its own I-Cache and small shaders.\\
- \textit{We are a Hybrid CPU/GPU: I-Cache is not separate!}
- \item Needs to go through OpenPOWER Foundation `approval'
- \end{itemize}
}
-\frame{\frametitle{maxloc}
+\frame{\frametitle{Pospopcount}
+
\begin{itemize}
- \item "TODO
- \end{itemize}
+ \item Positional popcount adds up the totals of each bit set to 1 in each bit-position, of an array of input values.
+ \item Notoriously difficult to do in SIMD assembler: typically 550 lines
+ \item https://github.com/clausecker/pospop
+
+ \end{itemize}
+
+ \lstinputlisting[language={}]{pospopcount.c}
+
+
}
\frame{\frametitle{Pospopcount}
-
+
+ \begin{center}
+ \includegraphics[width=0.5\textwidth]{pospopcount.png}
+ \end{center}
+ \begin{itemize}
+ \item The challenge is to perform an appropriate transpose of the data (the CPU can only work on registers, horizontally),
+ in blocks that suit the processor and the ISA capacity.
+
+
+ \end{itemize}
+}
+
+\frame{\frametitle{Pospopcount}
+
+ \begin{center}
+ \includegraphics[width=0.6\textwidth]{array_popcnt.png}
+ \end{center}
+
\begin{itemize}
- \item Positional popcount adds up the totals of each bit set to 1 in each bit-position, of an array of input values.
- \item Notoriously difficult to do in SIMD assembler: typically 550 lines
- \end{itemize}
-
- \lstinputlisting[language={}]{pospopcount.c}
-
+
+ \item The draft gbbd instruction implements the transpose (shown above),
+ preparing the data to use standard popcount.
+ (gbbd is based on Power ISA vgbbd, v3.1 p445)
+
+ \end{itemize}
+
}
-\frame{\frametitle{Pospopcount.s}
+
+\frame{\frametitle{pospopcount assembler}
\lstinputlisting[language={}]{pospopcount.s}
}
-
\frame{\frametitle{strncpy}
+ \lstinputlisting[language={}]{strncpy.c}
\begin{itemize}
- \item "TODO
- \end{itemize}
+ \item two simple-looking for-loops,
+ data-dependent in the first.
+ \item sv.cmpi stops at the first zero, /vli includes the zero
+ in VL.
+ \item note the post-increment Load/Store: saves
+ pre-decrementing
+ \item a Vector of CRs is produced which then get tested
+ by the sv.bc/all instruction, counting down CTR
+ per item tested.
+ \item Power ISA added hard-coded data-dependent capacity
+ into vstribr, where SVP64 it is generic (applies
+ to any instruction)
+ \item even the null-ing part is not straightforward as
+ it could be mis-aligned compared to the VSX width.
+ \item end-result: assembler-optimised strncpy on Power
+ ISA v3.0 is a whopping 240 instructions. SVP64 is 10
+ and parallel in HW
+ \end{itemize}
}
+
+
\frame{\frametitle{strncpy assembler}
\lstinputlisting[language={}]{strncpy.s}
}
-\frame{\frametitle{linked-list walking}
+\frame{\frametitle{sv.lbz/ff=RC1/vli *16,1(10)}
+ \begin{center}
+ \includegraphics[width=0.6\textwidth]{lbz_ff_vli.png}
+ \end{center}
+
+ \begin{itemize}
+ \item r10 points to memory address 0x001007
+ \item sv.lbz (Power ISA load byte immediate) multiplies immediate
+ offset by element step index, to get Effective Address (EA)
+ \item LD/ST has no Rc=1 so Data-Dependent Fail-First specified
+ as "ff=RC1". Not LD/ST Fault First! vli: VL inclusive
+ \item Test done after each load. Fails at Memory contents
+ 0x001009. Inclusive Mode: VL is truncated to 5 (FIVE) not 4
+ \end{itemize}
+}
+
+\frame{\frametitle{linked-list walking}
+
\begin{itemize}
- \item "TODO
+ \item "TODO
\end{itemize}
}
+
+\frame{\frametitle{sv.ld/ff=RC1/vli *17, 8(*16)}
+
+ \begin{center}
+ \includegraphics[width=1.0\textwidth]{linked_list_dd.png}
+ \end{center}
+}
+
+\frame{\frametitle{maxloc}
+ \begin{itemize}
+ \item "TODO
+ \end{itemize}
+}
+
\frame{\frametitle{Summary}
\begin{itemize}
- \item Goal is to create a mass-volume low-power embedded SoC suitable
- for use in netbooks, chromebooks, tablets, smartphones, IoT SBCs.
- \item No way we could implement a project of this magnitude without
- nmigen (being able to use python OO to HDL)
- \item Collaboration with OpenPOWER Foundation and Members absolutely
- essential. No short-cuts. Standards to be developed and ratified
- so that everyone benefits.
- \item Riding the wave of huge stability of OpenPOWER ecosystem
- \item Greatly simplified open 3D and Video drivers reduces product
- development costs for customers
- \item It also happens to be fascinating, deeply rewarding technically
- challenging, and funded by NLnet
-
+ \item SIMD fundamentally assumes element independence.
+ \item No provision in SIMD ISAs or Architectures for
+ inter-element inter-dependence, let alone sequential
+ inter-dependence.
+ \item Simple-V adds features such as Data-Dependent
+ Fail-First as \textit{general concepts},
+ exploiting Condition Registers (Vectorised)
+ \item Hardware Parallelism is \textit{still possible}
+ by exploiting the standard capabilities of
+ Speculative Execution: produce results, hold
+ off writing, post-analyse and cancel the results
+ that should not be written. Uses \textit{existing}
+ standard OoO Micro-architecture
+ \item Huge simplification of algorithms, huge "compactification"
+ just like Zilog Z80 and Intel 8086, yet still parallel
+ \item compact deep-expressive assembler brings CISC
+ capability but RISC-RISC (Prefix-Suffix). SIMD remains
+ at the \textit{back-end in hardware} where it belongs.
+ Not exposed at the programmer.
\end{itemize}
}
-
\frame{
\begin{center}
{\Huge The end\vspace{12pt}\\
- Thank you\vspace{12pt}\\
- Questions?\vspace{12pt}
- }
+ Thank you\vspace{12pt}
+ }
\end{center}
-
+
\begin{itemize}
- \item Discussion: http://lists.libre-soc.org
- \item Freenode IRC \#libre-soc
- \item http://libre-soc.org/
- \item http://nlnet.nl/PET
- \item https://libre-soc.org/nlnet/\#faq
+ \item Discussion: http://lists.libre-soc.org
+ \item OFTC.net IRC \#libre-soc
+ \item http://libre-soc.org/
+ \item https://nlnet.nl/project/Libre-SOC-OpenPOWER-ISA
+\item https://bugs.libre-soc.org/show\_bug.cgi?id=676
+\item https://bugs.libre-soc.org/show\_bug.cgi?id=1244
+\item https://libre-soc.org/openpower/sv/cookbook/fortran\_maxloc
+ \item https://libre-soc.org/nlnet/\#faq
\end{itemize}
}