diff options
author | rsnikhil <nikhil@acm.org> | 2019-10-24 09:34:21 -0400 |
---|---|---|
committer | rsnikhil <nikhil@acm.org> | 2019-10-24 09:34:21 -0400 |
commit | 1a4e20e5d650e2edc008623b5f52fbd2a40c3684 (patch) | |
tree | 14afba1a61a651649ace811c43d0f286c0bd70ce /doc | |
parent | ee07b9904eaa332da2deebc904065e4bed56a6ab (diff) | |
download | sail-riscv-1a4e20e5d650e2edc008623b5f52fbd2a40c3684.zip sail-riscv-1a4e20e5d650e2edc008623b5f52fbd2a40c3684.tar.gz sail-riscv-1a4e20e5d650e2edc008623b5f52fbd2a40c3684.tar.bz2 |
Added notes_FD_extensions.md to doc/ directory
Diffstat (limited to 'doc')
-rw-r--r-- | doc/notes_FD_extensions.md | 60 |
1 files changed, 60 insertions, 0 deletions
diff --git a/doc/notes_FD_extensions.md b/doc/notes_FD_extensions.md new file mode 100644 index 0000000..72740cc --- /dev/null +++ b/doc/notes_FD_extensions.md @@ -0,0 +1,60 @@ +Notes on F and D extensions (floating point) in the SAIL RISC-V Model +===================================================================== + +Questions? Rishiyur Nikhil (github id: `rsnikhil`), Bluespec, Inc. + +The main original SAIL RISC-V repo is [here](https://github.com/rems-project/sail-riscv) + +[Current working directory](https://github.com/rsnikhil/sail-riscv.git). +This is a fork of the main original repo, on which we have added the SAIL code for F and D extensions. +This will be merged back into the main original repo after testing. + +--- + +Status (2019-10-24) +------------------- + +1. Complete (almost): SAIL coding [Rishiyur Nikhil, Bluespec, Inc.] + Remaining: tweaks to handle illegal rounding modes. +2. In progress: Incorporate Berkeley SoftFloat calls [U.Cambridge] +3. To do: Testing on all ISA tests + +Should not take more than a few days to complete the remaining work. + +--- + +Files (all are in "model/" directory) +------------------------------------- + +Changed file: +Added two one-line predicates `haveFExt()` and `haveDExt()` that test whether MISA.F and MISA.D are set: + + riscv_sys_regs.sail + +New files: +Definition for FLEN and bit-vectors holding floating point values. +The former should be used if only F is supported, +the latter if F and D are supported: + + riscv_flen_F.sail + riscv_flen_D.sail + +New file: A few more definitions for floating point registers. + + riscv_freg_type.sail + +New file: The floating point register file and floating point CSRs: + + riscv_fdext_regs.sail + +New files: The main files for F and D, respectively, containing AST +definitions, binary encode/decode mappings, assembly code generation, +and execution semantics: + + riscv_insts_fext.sail + riscv_insts_dext.sail + +Functions that call out to Berkeley SoftFloat (implemented in C) to +perform the core floating point computations: + + riscv_softfloat_interface.sail |