Commit Graph

1454 Commits

Author SHA1 Message Date
Keith Rothman 4c82ddf394 Add -params mode to force undef parameters in selected cells.
Signed-off-by: Keith Rothman <537074+litghost@users.noreply.github.com>
2019-02-21 10:16:38 -08:00
Clifford Wolf 0e371109b0
Merge pull request #818 from YosysHQ/clifford/dffsrfix
Fix opt_rmdff handling of $_DFFSR_???_ and $_DLATCHSR_???_, fixes #816
2019-02-21 18:58:44 +01:00
Clifford Wolf 893194689d Fix typo in passes/pmgen/README.md
Signed-off-by: Clifford Wolf <clifford@clifford.at>
2019-02-21 18:50:02 +01:00
Clifford Wolf 2da4c9c8f0 Fix opt_rmdff handling of $_DFFSR_???_ and $_DLATCHSR_???_, fixes #816
Signed-off-by: Clifford Wolf <clifford@clifford.at>
2019-02-21 13:49:45 +01:00
Clifford Wolf 2fe1c830eb Bugfix in ice40_dsp
Signed-off-by: Clifford Wolf <clifford@clifford.at>
2019-02-21 13:28:46 +01:00
Clifford Wolf 218e9051bb Add "synth_ice40 -dsp"
Signed-off-by: Clifford Wolf <clifford@clifford.at>
2019-02-20 16:42:27 +01:00
Clifford Wolf 246391200e Add FF support to wreduce
Signed-off-by: Clifford Wolf <clifford@clifford.at>
2019-02-20 16:36:42 +01:00
Clifford Wolf dca65d83a0 Detect and reject cases that do not map well to iCE40 DSPs (yet)
Signed-off-by: Clifford Wolf <clifford@clifford.at>
2019-02-20 11:18:19 +01:00
Jim Lawson 5c4a72c43e Fix normal (non-array) hierarchy -auto-top.
Add simple test.
2019-02-19 14:35:15 -08:00
Clifford Wolf 5a853ed46c Add actual DSP inference to ice40_dsp pass
Signed-off-by: Clifford Wolf <clifford@clifford.at>
2019-02-17 15:35:48 +01:00
Clifford Wolf c06c062469 Merge branch 'master' of github.com:YosysHQ/yosys into pmgen 2019-02-17 12:10:19 +01:00
Jim Lawson 5c504c5ae6 Define basic_cell_type() function and use it to derive the cell type for array references (instead of duplicating the code). 2019-02-15 11:31:37 -08:00
David Shah a4515712cb fsm_opt: Fix runtime error for FSMs without a reset state
Signed-off-by: David Shah <dave@ds0.me>
2019-02-07 10:35:36 +00:00
David Shah 58c22dae31 abc: Improved recovered netnames, also preserve src on nets with dress
Signed-off-by: David Shah <davey1576@gmail.com>
2019-02-06 22:23:13 +01:00
David Shah 8524a479b1 abc: Preserve naming through ABC using 'dress' command
Signed-off-by: David Shah <dave@ds0.me>
2019-02-06 22:23:13 +01:00
whitequark 58d059ccb7 proc_clean: fix critical typo. 2019-01-23 22:08:38 +00:00
whitequark 95b6c35882 proc_clean: fix fully def check to consider compare/signal length.
Fixes #790.
2019-01-18 23:22:19 +00:00
Clifford Wolf 8ddec5d882 Progress in pmgen
Signed-off-by: Clifford Wolf <clifford@clifford.at>
2019-01-15 11:23:25 +01:00
Clifford Wolf 5216735210 Progress in pmgen, add pmgen README
Signed-off-by: Clifford Wolf <clifford@clifford.at>
2019-01-15 11:23:25 +01:00
Clifford Wolf 55ac030382 Fix pmgen "reject" statement
Signed-off-by: Clifford Wolf <clifford@clifford.at>
2019-01-15 11:23:25 +01:00
Clifford Wolf d45379936b Progress in pmgen
Signed-off-by: Clifford Wolf <clifford@clifford.at>
2019-01-15 11:23:25 +01:00
Clifford Wolf 1f8e76f993 Progress in pmgen
Signed-off-by: Clifford Wolf <clifford@clifford.at>
2019-01-15 11:23:25 +01:00
Clifford Wolf b9545aa0e1 Progress in pmgen
Signed-off-by: Clifford Wolf <clifford@clifford.at>
2019-01-15 11:23:25 +01:00
Clifford Wolf ad69c668ce Add mockup .pmg (pattern matcher generator) file
Signed-off-by: Clifford Wolf <clifford@clifford.at>
2019-01-15 11:23:25 +01:00
whitequark e792bd56b7 flowmap: clean up terminology.
* "map": group gates into LUTs;
  * "pack": replace gates with LUTs.

This is important because we have FlowMap and DF-Map, and currently
our messages are ambiguous.

Also clean up some other log messages while we're at it.
2019-01-08 02:05:06 +00:00
whitequark 211c26a4c9 flowmap: implement depth relaxation. 2019-01-08 01:13:05 +00:00
Clifford Wolf 8a63fc51d3 Bugfix in $memrd sharing
Signed-off-by: Clifford Wolf <clifford@clifford.at>
2019-01-07 10:04:47 +01:00
Clifford Wolf dbd51d7bda
Merge pull request #782 from whitequark/flowmap_dfs
flowmap: construct a max-volume max-flow min-cut, not just any one
2019-01-07 09:47:57 +01:00
Clifford Wolf b5f6e786ea Switch "bugpoint" from system() to run_command()
Signed-off-by: Clifford Wolf <clifford@clifford.at>
2019-01-07 09:45:21 +01:00
whitequark a342d6db49 bugpoint: new pass.
A typical use of `bugpoint` would involve a script with a pass under
test, e.g.:

    flowmap -relax -optarea 100

and would be invoked as:

    bugpoint -yosys ./yosys -script flowmap.ys -clean -cells

This replaces the current design with the minimal design that still
crashes the `flowmap.ys` script.

`bugpoint` can also be used to perform generic design minimization
using `select`, e.g. the following script:

    select i:* %x t:$_MUX_ %i -assert-max 0

would remove all parts of the design except for an unbroken path from
an input to an output port that goes through exactly one $_MUX_ cell.
(The condition is inverted.)
2019-01-07 03:13:19 +00:00
whitequark 8b44198e23 flowmap: construct a max-volume max-flow min-cut, not just any one. 2019-01-06 19:51:37 +00:00
Scott Mansell 62c90c4e17 Rename cells based on the wires they drive. 2019-01-06 19:00:16 +13:00
whitequark 2fcc1ee72e flowmap: add -minlut option, to allow postprocessing with opt_lut. 2019-01-04 21:18:03 +00:00
whitequark 9bc5cf0844 flowmap: cleanup for clarity. NFCI. 2019-01-04 13:04:20 +00:00
whitequark fd21564deb flowmap: improve debug graph output. NFC. 2019-01-04 03:30:04 +00:00
whitequark 7850a0c28a flowmap: add link to longer version of paper. NFC. 2019-01-04 02:33:10 +00:00
Clifford Wolf d98fe8ce1f
Merge pull request #775 from whitequark/opt_flowmap
flowmap: new techmap pass
2019-01-03 17:03:18 +01:00
whitequark 07af772a72 flowmap: new techmap pass. 2019-01-03 14:28:19 +00:00
Clifford Wolf 0fc6e2bfcf
Merge pull request #770 from whitequark/opt_expr_cmp
opt_expr: refactor and improve simplification of comparisons
2019-01-02 17:34:04 +01:00
whitequark bf8db55ef3 opt_expr: improve simplification of comparisons with large constants.
The idea behind this simplification is that a N-bit signal X being
compared with an M-bit constant where M>N and the constant has Nth
or higher bit set, it either always succeeds or always fails.

However, the existing implementation only worked with one-hot signals
for some reason. It also printed incorrect messages.

This commit adjusts the simplification to have as much power as
possible, and fixes other bugs.
2019-01-02 15:45:28 +00:00
Clifford Wolf 979de95cf6
Merge pull request #750 from Icenowy/anlogic-ff-init
Initialization of Anlogic DFFs
2019-01-02 15:52:22 +01:00
Clifford Wolf 2e606b1802
Merge pull request #773 from whitequark/opt_lut_elim_fixes
opt_lut: elimination fixes
2019-01-02 15:45:29 +01:00
whitequark efa278e232 Fix typographical and grammatical errors and inconsistencies.
The initial list of hits was generated with the codespell command
below, and each hit was evaluated and fixed manually while taking
context into consideration.

    DIRS="kernel/ frontends/ backends/ passes/ techlibs/"
    DIRS="${DIRS} libs/ezsat/ libs/subcircuit"
    codespell $DIRS -S *.o -L upto,iff,thru,synopsys,uint

More hits were found by looking through comments and strings manually.
2019-01-02 13:12:17 +00:00
whitequark c55dfb8369 opt_lut: reflect changes in sigmap.
Otherwise, some LUTs will be missed during elimination.
2019-01-02 10:21:58 +00:00
whitequark 06143ab33f opt_lut: use a worklist, and revisit cells affected by elimination. 2019-01-02 09:36:32 +00:00
whitequark f7363ac508 opt_lut: count eliminated cells, and set opt.did_something for them. 2019-01-02 09:14:43 +00:00
whitequark 4fd458290c opt_expr: refactor simplification of unsigned X<onehot and X>=onehot. NFCI. 2019-01-02 05:11:29 +00:00
whitequark 9e9846a6ea opt_expr: refactor simplification of signed X>=0 and X<0. NFCI. 2019-01-02 03:01:25 +00:00
whitequark 8e53d2e0bf opt_expr: simplify any unsigned comparisons with all-0 and all-1.
Before this commit, only unsigned comparisons with all-0 would be
simplified. This commit also makes the code handling such comparisons
to be more rigorous and not abort on unexpected input.
2019-01-02 02:45:49 +00:00
whitequark 42c356c49c opt_lut: eliminate LUTs evaluating to constants or inputs. 2018-12-31 23:55:40 +00:00
Clifford Wolf 0a840dd883 Fix handling of (* keep *) wires in wreduce
Signed-off-by: Clifford Wolf <clifford@clifford.at>
2018-12-31 16:37:40 +01:00
whitequark 18291c20d2 proc_clean: remove any empty cases if all cases use all-def compare. 2018-12-23 09:04:30 +00:00
whitequark b784440857 proc_clean: remove any empty cases at the end of the switch.
Previously, only completely empty switches were removed.
2018-12-22 09:04:46 +00:00
whitequark 0c318e7db5 memory_collect: do not truncate 'x from \INIT.
The semantics of an RTLIL constant that has less bits than its
declared bit width is zero padding. Therefore, if the output of
memory_collect will be used for simulation, truncating 'x from
the end of \INIT will produce incorrect simulation results.
2018-12-21 02:01:27 +00:00
David Shah 2b16d4ed3d memory_dff: Fix typo when checking init value
Signed-off-by: David Shah <davey1576@gmail.com>
2018-12-18 17:40:01 +00:00
Icenowy Zheng 256fb8c95c Add "dffinit -noreinit" parameter
Sometimes the FF cell might be initialized during the map process, e.g.
some FPGA platforms (Anlogic Eagle and Lattice ECP5 for example) has
only a "SR" pin for a FF for async reset, that resets the FF to the
initial value, which means the async reset value should be set as the
initial value. In this case the DFFINIT pass shouldn't reinitialize it
to a different value, which will lead to error.

Add a "-noreinit" parameter for the safeguard. If a FF is not
technically initialized before DFFINIT pass, the default value should be
set to x.

Signed-off-by: Icenowy Zheng <icenowy@aosc.io>
2018-12-18 23:10:40 +08:00
Icenowy Zheng fec8b3c81f Add "dffinit -strinit high low"
On some platforms the string to initialize DFF might not be "high" and
"low", e.g. with Anlogic TD it's "SET" and "RESET".

Add a "-strinit" parameter for dffinit to allow specify the strings used
for high and low.

Signed-off-by: Icenowy Zheng <icenowy@aosc.io>
2018-12-18 15:37:43 +08:00
Clifford Wolf 2641a3089b
Revert "Proof-of-concept: preserve naming through ABC using dress" 2018-12-16 21:27:31 +01:00
Clifford Wolf ddff75b60a
Merge pull request #736 from whitequark/select_assert_list
select: print selection if a -assert-* flag causes an error
2018-12-16 16:45:49 +01:00
whitequark f6412d7109 select: print selection if a -assert-* flag causes an error. 2018-12-16 15:44:29 +00:00
Clifford Wolf 0d9c850a07
Merge pull request #735 from daveshah1/trifixes
deminout fixes
2018-12-16 16:02:21 +01:00
Clifford Wolf f53e19cc71 Fix equiv_opt indenting
Signed-off-by: Clifford Wolf <clifford@clifford.at>
2018-12-16 15:57:28 +01:00
Clifford Wolf 2a681909df
Merge pull request #724 from whitequark/equiv_opt
equiv_opt: new command, for verifying optimization passes
2018-12-16 15:54:26 +01:00
Clifford Wolf a2154c1be0
Merge pull request #734 from grahamedgecombe/fix-shuffled-bram-initdata
memory_bram: Fix initdata bit order after shuffling
2018-12-16 15:53:44 +01:00
Clifford Wolf a1fb5b1e4b
Merge pull request #714 from daveshah1/abc_preserve_naming
Proof-of-concept: preserve naming through ABC using dress
2018-12-16 15:41:30 +01:00
Clifford Wolf 19ca4e2ac3
Merge pull request #722 from whitequark/rename_src
rename: add -src, for inferring names from source locations
2018-12-16 15:28:29 +01:00
Clifford Wolf 556341a77f
Merge pull request #720 from whitequark/master
lut2mux: handle 1-bit INIT constant in $lut cells
2018-12-16 15:27:23 +01:00
David Shah 4c59447168 deminout: Consider $tribuf cells
Signed-off-by: David Shah <dave@ds0.me>
2018-12-12 17:17:40 +00:00
David Shah d3fe9465f3 deminout: Don't demote constant-driven inouts to inputs
Signed-off-by: David Shah <dave@ds0.me>
2018-12-12 16:50:46 +00:00
Graham Edgecombe 4fef9689ab memory_bram: Fix initdata bit order after shuffling
In some cases the memory_bram pass shuffles the order of the bits in a
memory's RD_DATA port. Although the order of the bits in the WR_DATA and
WR_EN ports is changed to match the RD_DATA port, the order of the bits
in the initialization data is not.

This causes reads of initialized memories to return invalid data (until
the initialization data is overwritten).

This commit fixes the bug by shuffling the initdata bits in exactly the
same order as the RD_DATA/WR_DATA/WR_EN bits.
2018-12-11 21:02:49 +00:00
whitequark 7ff5a9db2d equiv_opt: pass -D EQUIV when techmapping.
This allows avoiding techmap crashes e.g. because of large memories
in white-box cell models.
2018-12-07 17:20:34 +00:00
whitequark c38ea9ae65 equiv_opt: new command, for verifying optimization passes. 2018-12-07 17:20:34 +00:00
whitequark 7ec740b7ad opt_lut: leave intact LUTs with cascade feeding module outputs. 2018-12-07 17:13:52 +00:00
whitequark 9eb03d458d opt_lut: show original truth table for both cells. 2018-12-07 17:04:41 +00:00
whitequark a8ab722824 opt_lut: add -limit option, for debugging misoptimizations. 2018-12-07 16:36:26 +00:00
David Shah 1dfb2fecab abc: Preserve naming through ABC using 'dress' command
Signed-off-by: David Shah <dave@ds0.me>
2018-12-06 15:05:07 +00:00
Clifford Wolf 643f858acf Bugfix in opt_expr handling of a<0 and a>=0
Signed-off-by: Clifford Wolf <clifford@clifford.at>
2018-12-06 07:29:21 +01:00
whitequark a9baee4b24 rename: add -src, for inferring names from source locations. 2018-12-05 20:35:13 +00:00
whitequark d1f2cb01dc lut2mux: handle 1-bit INIT constant in $lut cells.
This pass already handles INIT constants shorter than 2^width, but
that was not done for the recursion base case.
2018-12-05 19:27:48 +00:00
whitequark 88217d0157 opt_lut: simplify type conversion. NFC. 2018-12-05 19:12:02 +00:00
Clifford Wolf 2d98db73e3 Rename opt_lut.cpp to opt_lut.cc
Signed-off-by: Clifford Wolf <clifford@clifford.at>
2018-12-05 18:03:58 +01:00
whitequark 45cb6200af opt_lut: add -dlogic, to avoid disturbing logic such as carry chains. 2018-12-05 16:30:37 +00:00
whitequark e603484070 opt_lut: always prefer to eliminate 1-LUTs.
These are always either buffers or inverters, and keeping the larger
LUT preserves more source-level information about the design.
2018-12-05 16:30:37 +00:00
whitequark 59eea0183f opt_lut: collect and display statistics. 2018-12-05 16:30:37 +00:00
whitequark e54c7e951c opt_lut: refactor to use a worker. NFC. 2018-12-05 16:30:37 +00:00
whitequark 9e072ec21f opt_lut: new pass, to combine LUTs for tighter packing. 2018-12-05 16:30:37 +00:00
Clifford Wolf c800e3bb16 Fix typo
Signed-off-by: Clifford Wolf <clifford@clifford.at>
2018-12-04 23:30:23 +01:00
Clifford Wolf 70c417174d
Merge pull request #702 from smunaut/min_ce_use
Add option to only use DFFE is the resulting E signal would be use > N times
2018-12-04 14:29:21 -08:00
Clifford Wolf 47c89d600c
Merge pull request #676 from rafaeltp/master
Splits SigSpec into bits before calling check_signal_in_fanout (solves #675)
2018-12-01 04:11:19 +01:00
Sylvain Munaut 8d3ab626ea dff2dffe: Add option for unmap to only remove DFFE with low CE signal use
Signed-off-by: Sylvain Munaut <tnt@246tNt.com>
2018-11-27 21:50:42 +01:00
Clifford Wolf ab97eddee9 Add iteration limit to "opt_muxtree"
Signed-off-by: Clifford Wolf <clifford@clifford.at>
2018-11-20 17:56:47 +01:00
Niels Moseley cfc9b9147c DFFLIBMAP: changed 'missing pin' error into a warning with additional reason/info. 2018-11-06 12:11:52 +01:00
Clifford Wolf 719e29404a Allow square brackets in liberty identifiers
Signed-off-by: Clifford Wolf <clifford@clifford.at>
2018-11-05 12:33:33 +01:00
Niels Moseley 04cd179696 Liberty file newline handling is more relaxed. More descriptive error message 2018-11-03 18:38:49 +01:00
Niels Moseley d1e8249f9a Report an error when a liberty file contains pin references that reference non-existing pins 2018-11-03 18:07:51 +01:00
rafaeltp f8b97e21f3 using [i] to access individual bits of SigSpec and merging bits into a tmp Sig before setting the port to new signal 2018-10-21 11:32:44 -07:00
rafaeltp 7b964bfb83 cleaning up for PR 2018-10-20 18:02:59 -07:00
rafaeltp ce069830c5 fixing code style 2018-10-20 17:57:26 -07:00
rafaeltp 0ad4321781 solves #675 2018-10-20 17:50:21 -07:00
Ruben Undheim 436e3c0a7c Refactor code to avoid code duplication + added comments 2018-10-20 16:06:48 +02:00