3d_gpu/layouts/coriolis2_180nm.mdwn

   1 # Coriolis2 180nm layout
   2
   3 * <http://bugs.libre-riscv.org/show_bug.cgi?id=138> - toplevel
   4 * <http://bugs.libre-riscv.org/show_bug.cgi?id=199> - main layout
   5 * <https://ftp.libre-soc.org/course_18oct2021/>
   6 * [[180nm_Oct2020]]
   7
   8 # Simple floorplan
   9
  10 [[!img simple_floorplan.png size="500x"]]
  11
  12 ## Register files
  13
  14 There are 5 register files: SPR, INT, CR, XER and FAST.
  15
  16 Access to each of the ports is managed via a "Priority Picker" - an
  17 unary-in but one-hot unary-out picker - which allows one and only one
  18 "user" of a given regfile port at any one time.
  19
  20 ## Computation Units
  21
  22 There are 8 Function Units: ALU, Logical, Condition, Branch, ShiftRot, LDST,
  23 Trap, and SPRs.
  24
  25 Each Function Unit has operand inputs and operand outputs.  Across *all*
  26 pipelines there are multiple Function Units that require "RA" (Register A
  27 Integer Register File).  All of such "RA" read requests are (surprise)
  28 connected to the same "Priority Picker" mentioned above: likewise
  29 all Function Units requiring write to the "RT" register are connected
  30 to the exact same "RT-managing" Write Priority Picker.
  31
  32 ### Load Store Computation Unit(s)
  33
  34 Load/Store is a special type of Computation Unit that additionally has
  35 access to external memory.  In the case where multiple LDSTCompUnits
  36 are added, L0CacheBuffer is responsible for "merging" these into single
  37 requests.
  38
  39 There are however *two* L0 Caches (both 128-bit wide), with a split
  40 on address bit 4 for selecting either the odd L0 Cache or the even L0 Cache.
  41
  42 Each of the two L0 caches has dual 64-bit Wishbone interfaces giving
  43 a total of *four* 64-bit Memory Bus requests that will be merged through
  44 an Arbiter down onto the same Memory Bus that the I-Cache is also connected
  45 to.
  46
  47 ## Instructions
  48
  49 Instructions are decoded by PowerDecoder2, after being read by the
  50 simple core FSM from the Instruction Cache.  Currently this is an
  51 extremely simple memory block, to be replaced by a proper I-Cache
  52 with a proper connection to the Memory Bus (wishbone).
  53
  54 # IO Ring and JTAG
  55
  56 The IO Ring is autogenerated from the same pinmux program
  57 that created the [[180nm_Oct2020/pinouts]] and the SVG
  58 image. The image was used by Greatek for packaging as well as
  59 a PCB designed by Professor Galayko of Sorbonne University.
  60
  61 The exact same pinmux program's output, specifying all interfaces,
  62 was also used to autogenerate the HDL for the JTAG Boundary Scan.
  63
  64 By strictly using the exact same *machine readable* specification
  65 for all Interfaces using only autogenerated techniques it was possible
  66 to ensure complete consistency across
  67
  68 * Markdown file
  69 * SVG Image for packaging
  70 * IO Ring
  71 * JTAG Boundary Scan
  72
  73 JTAG
  74
  75 [[!img 180nm_Oct2020/ls180.svg size="400x" ]]
  76
  77 # Building
  78
  79 To build see [[HDL_workflow/coriolis2]]. A tag has been used and the
  80 build instructions specify it.  The soclayout repository is standalone,
  81 containing a snapshot of the verilog autogenerated output.
  82
  83 # About coriolis2
  84
  85 There are several talks online now.
  86
  87 * [[conferences/fosdem2022]]
  88 * <https://m.youtube.com/playlist?list=PLBtNqZjUZB80uByfZVm0gGYEtmTG0mZzm>