microwatt.git
2 years agoUse FPU for division instructions if we have an FPU
Paul Mackerras [Mon, 9 May 2022 09:18:42 +0000 (19:18 +1000)]
Use FPU for division instructions if we have an FPU

- Arrange for XER to be written for OE=1 forms
- Arrange for condition codes to be set for RC=1 forms
  (including correct handling for 32-bit mode)
- Don't instantiate the divider if we have an FPU.

Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
2 years agoFPU: Add logic for 32-bit integer division
Paul Mackerras [Sat, 7 May 2022 12:34:23 +0000 (22:34 +1000)]
FPU: Add logic for 32-bit integer division

Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
2 years agoFPU: Add integer division logic to FPU
Paul Mackerras [Sat, 7 May 2022 08:28:33 +0000 (18:28 +1000)]
FPU: Add integer division logic to FPU

This adds logic to the FPU to accomplish 64-bit integer divisions.
No instruction actually uses this yet.

The algorithm used is to obtain an estimate of the reciprocal of the
divisor using the lookup table and refine it by one to three
iterations of the Newton-Raphson algorithm (the number of iterations
depends on the number of significant bits in the dividend).  Then the
reciprocal is multiplied by the dividend to get the quotient estimate.
The remainder is calculated as dividend - quotient * divisor.  If the
remainder is greater than or equal to the divisor, the quotient is
incremented, or if a modulo operation is being done, the divisor is
subtracted from the remainder.  The inverse estimate after refinement
is good enough that the quotient estimate is always equal to or one
less than the true quotient.

Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
2 years agoFPU: Convert internal R, A, B, and C registers to 8.56 format
Paul Mackerras [Sun, 1 May 2022 23:39:26 +0000 (09:39 +1000)]
FPU: Convert internal R, A, B, and C registers to 8.56 format

This changes the representation of the R, A, B and C registers in the
FPU from 10.54 format (10 bits to the left of the binary point and 54
bits to the right) to 8.56 format, to match the representation used in
the P and Y registers and the multiplier operands.  This eliminates
the need for shifting when R, A, B or C is an input to the multiplier
and will make it easier to implement integer division in the FPU.

Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
2 years agoTrack hazards explicitly for XER overflow bits
Paul Mackerras [Sat, 9 Jul 2022 08:29:48 +0000 (18:29 +1000)]
Track hazards explicitly for XER overflow bits

This provides a mechanism for tracking updates to the XER overflow
bits (SO, OV, OV32) and stalling instructions which need current
values of those bits (mfxer, integer compare instructions, integer
Rc=1 instructions, addex) or which writes carry bits (since all the
XER common bits are written together, if we are writing CA/CA32 we
need up-to-date values of SO/OV/OV32).

This will enable updates to SO/OV/OV32 to be done at other places
besides the ex1 stage.

Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
2 years agofetch1: Fix debug stop again
Paul Mackerras [Sat, 9 Jul 2022 03:17:18 +0000 (13:17 +1000)]
fetch1: Fix debug stop again

This fixes a bug which prevents the core from stopping properly.  The
same bug was previously fixed in commit e41cb01bca99 ("fetch1: Fix
debug stop", 2020-12-19) and reintroduced by commit 0fb207be6069
("fetch1: Implement a simple branch target cache", 2020-12-19).

Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
2 years agocontrol: Reimplement serialization using tags
Paul Mackerras [Sat, 9 Jul 2022 01:55:13 +0000 (11:55 +1000)]
control: Reimplement serialization using tags

This lets us get rid of r_int and its 'outstanding' counter.  We now
test more directly for excess completions by checking that we don't
get duplicate completions for the same tag.

Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
2 years agodecode1: Remove stash buffer
Paul Mackerras [Fri, 8 Jul 2022 06:37:12 +0000 (16:37 +1000)]
decode1: Remove stash buffer

Now that the timing of the busy signal from decode2 doesn't depend on
register numbers or downstream instruction completion, we no longer
need the stash buffer on the output of decode1.

Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
2 years agodecode2: Rework to make the stall_out signal come from a register
Paul Mackerras [Fri, 8 Jul 2022 04:07:28 +0000 (14:07 +1000)]
decode2: Rework to make the stall_out signal come from a register

At present the busy/stall signal going to decode1 depends on whether
control thinks it can issue the current instruction, and that depends
on completion and bypass signals coming from execute1 and writeback.

To improve the timing of stall_out, this rearranges decode2 so that
stall_out is asserted when we have a valid instruction that couldn't
be issued in the previous cycle.  This means that decode1 could give
us a new instruction when we haven't issued the previous instruction.

This in turn means that we can only use d_in in the first cycle of
processing an instruction.  After the first cycle, we get register
addresses etc. from dc2 rather than d_in.

Then, to avoid the need to read register operands from register_file
in each cycle until the instruction issues, we bring the bypass path
for data being written to the register file into decode2 explicitly
rather than having it in register_file.

A new process called decode2_addrs does the process of calling
decode_input_reg_* and decode_output_reg and sets up the register file
addresses.  This was split out (and decode_input_reg_* reworked) to
try to reduce the number of passes through the decode2_1 process that
need to be done in simulation.

Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
2 years agoRemove support for lq, stq, lqarx and stqcx.
Paul Mackerras [Sat, 4 Jun 2022 07:37:48 +0000 (17:37 +1000)]
Remove support for lq, stq, lqarx and stqcx.

They are optional in SFFS (scalar fixed-point and floating-point
subset), are not needed for running Linux, and add complexity, so
remove them.

Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
2 years agodecode2: Rename 'r' to 'dc2'
Paul Mackerras [Mon, 4 Jul 2022 08:23:03 +0000 (18:23 +1000)]
decode2: Rename 'r' to 'dc2'

Also get rid of a couple of unused variables.

Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
2 years agodecode1: Reduce number of single-issue instructions
Paul Mackerras [Sat, 2 Jul 2022 12:23:35 +0000 (22:23 +1000)]
decode1: Reduce number of single-issue instructions

This reduces the set of instructions marked as single-issue to just
attn and mtspr to "slow" SPRs (those that are not stored in the
register file).

The instructions that were previously single-issue are: isync, dcbf,
dcbst, dcbt, dcbtst, eieio, icbi, mfmsr, mtmsr, mtmsrd, mfspr to slow
SPRS, sync, tlbsync and wait.  The synchronization instructions are
mostly no-ops anyway due to the in-order nature of the core, and the
cache-management instructions are unimplemented (except for icbi).
The MSR ops don't need to be single-issue due to the in-order core and
the fact that MSR updates are effective on the following instruction.

Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
2 years agoFPU: Add stage-2 stall ability to FPU
Paul Mackerras [Sat, 2 Jul 2022 04:17:18 +0000 (14:17 +1000)]
FPU: Add stage-2 stall ability to FPU

This makes the FPU able to stall other units at execute stage 2 and be
stalled by other units (specifically the LSU).

This means that the completion and writeback for an instruction can
now end up being deferred until the second cycle of a following
instruction, i.e. the cycle when the state machine has gone through
IDLE state into one of the DO_* states, which means we need to latch
the destination FPR number, CR mask, etc. from the previous
instruction so that we present the correct information to writeback.

The advantage of this is that we can get rid of the in_progress signal
from the LSU.

Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
2 years agoDo CR0 setting for Rc=1 instructions in execute2 instead of writeback
Paul Mackerras [Tue, 28 Jun 2022 08:18:08 +0000 (18:18 +1000)]
Do CR0 setting for Rc=1 instructions in execute2 instead of writeback

This lets us forward the CR0 result to following instructions that
use CR, meaning they get to issue one cycle earlier.

Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
2 years agoAllow integer instructions and load/store instructions to execute together
Paul Mackerras [Mon, 27 Jun 2022 08:53:04 +0000 (18:53 +1000)]
Allow integer instructions and load/store instructions to execute together

Execute1 and loadstore1 now send each other stall signals that
indicate that a valid instruction in stage 2 can't complete in this
cycle, and hence any valid instruction in stage 1 in the other unit
can't move to stage 2.  With this in place, an ALU instruction can
move into stage 1 while a LSU instruction is in stage 2.

Since the FPU doesn't yet have a way to stall completion, we can't yet
start FPU instructions while any LSU or ALU instruction is in
progress.

Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
2 years agoAdd a bypass path from the execute2 stage
Paul Mackerras [Mon, 27 Jun 2022 22:40:42 +0000 (08:40 +1000)]
Add a bypass path from the execute2 stage

This enables some instructions to issue earlier and thus improves
performance, at the cost of some extra multiplexers in decode2.

Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
2 years agoAdd a second execute stage to the pipeline
Paul Mackerras [Thu, 30 Jun 2022 10:33:33 +0000 (20:33 +1000)]
Add a second execute stage to the pipeline

This adds a second execute stage to the pipeline, in order to match up
the length of the pipeline through loadstore and dcache with the
length through execute1.  This will ultimately enable us to get rid of
the 1-cycle bubble that we currently have when issuing ALU
instructions after one or more LSU instructions.

Most ALU instructions execute in the first stage, except for
count-zeroes and popcount instructions (which take two cycles and do
some of their work in the second stage) and mfspr/mtspr to "slow" SPRs
(TB, DEC, PVR, LOGA/LOGD, CFAR).  Multiply and divide/mod instructions
take several cycles but the instruction stays in the first stage (ex1)
and ex1.busy is asserted until the operation is complete.

There is currently a bypass from the first stage but not the second
stage.  Performance is down somewhat because of that and because this
doesn't yet eliminate the bubble between LSU and ALU instructions.

The forwarding of XER common bits has been changed somewhat because
now there is another pipeline stage between ex1 and the committed
state in cr_file.  The simplest thing for now is to record the last
value written and use that, unless there has been a flush, in which
case the committed state (obtained via e_in.xerc) is used.

Note that this fixes what was previously a benign bug in control.vhdl,
where it was possible for control to forget an instructions dependency
on a value from a previous instruction (a GPR or the CR) if this
instruction writes the value and the instruction gets to the point
where it could issue but is blocked by the busy signal from execute1.
In that situation, control may incorrectly not indicate that a bypass
should be used.  That didn't matter previously because, for ALU and
FPU instructions, there was only one previous instruction in flight
and once the current instruction could issue, the previous instruction
was completing and the correct value would be obtained from
register_file or cr_file.  For loadstore instructions there could be
two being executed, but because there are no bypass paths, failing to
indicate use of a bypass path is fine.

Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
2 years agoexecute1: Rename 'r' to 'ex1'
Paul Mackerras [Sat, 18 Jun 2022 07:29:43 +0000 (17:29 +1000)]
execute1: Rename 'r' to 'ex1'

Maybe this will give us slightly better names in critical path reports
and the like.

Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
2 years agoexecute1: Restructure to separate out execution of side effects
Paul Mackerras [Sat, 18 Jun 2022 06:24:30 +0000 (16:24 +1000)]
execute1: Restructure to separate out execution of side effects

We now have a record that represents the actions taken in executing an
instruction, and a process that computes that for the incoming
instruction.  We no longer have 'current' or 'r.cur_instr', instead
things like the destination register are put into r.e in the first
cycle of an instruction and not reinitialized in subsequent busy
cycles.

For mfspr and mtspr, we now decode "slow" SPR numbers (those SPRs that
are not stored in the register file) to a new "spr_selector" record
in decode1 (excluding those in the loadstore unit).  With this, the
result for mfspr is determined in the data path.

Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
2 years agoMove XER low bits out of register file
Paul Mackerras [Wed, 29 Jun 2022 10:02:36 +0000 (20:02 +1000)]
Move XER low bits out of register file

Besides the overflow and status carry bits, XER has 18 bits which need
to retain the value written by mtxer (in case software wants to
emulate the move-assist instructions (lswi, lswx, stswi, stswx).
Until now these bits (and others) have been stored in the GPR file as
a "fast" SPR, but this causes complications because XER is not really
a fast SPR.

Instead, we now store these 18 bits in the 'ctrl' signal, which exists
in execute1.  This will enable us to simplify the data path in future,
and has the added bonus that with a little bit of plumbing, we can get
the full XER value printed when dumping registers at the end of a
simulation.

Therefore this changes scripts/run_test.sh to remove the greps which
exclude XER from the comparison of actual and expected register
results.

Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
2 years agoSimplify flow control in the dcache and loadstore units
Paul Mackerras [Sat, 11 Jun 2022 09:20:57 +0000 (19:20 +1000)]
Simplify flow control in the dcache and loadstore units

Simplify the flow control by stalling the whole upstream pipeline when
a stage can't proceed, instead of trying to let each stage progress
independently when it can.

Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
2 years agoMerge pull request #353 from tianrui-wei/master
Paul Mackerras [Thu, 16 Jun 2022 23:46:57 +0000 (09:46 +1000)]
Merge pull request #353 from tianrui-wei/master

fix: fix icache_tb not finishing correctly

2 years agoMerge pull request #373 from antonblanchard/icache-insn-u-state
Michael Neuling [Thu, 16 Jun 2022 23:13:49 +0000 (09:13 +1000)]
Merge pull request #373 from antonblanchard/icache-insn-u-state

icache: Don't output X on i_out.insn

2 years agoMerge pull request #376 from antonblanchard/loadstore-init
Michael Neuling [Thu, 16 Jun 2022 06:47:33 +0000 (16:47 +1000)]
Merge pull request #376 from antonblanchard/loadstore-init

loadstore1: reduce U state being output

2 years agoMerge pull request #374 from antonblanchard/icache-unused-sig
Michael Neuling [Thu, 16 Jun 2022 06:45:41 +0000 (16:45 +1000)]
Merge pull request #374 from antonblanchard/icache-unused-sig

core: Remove unused icache_inv signal

2 years agoMerge pull request #364 from shenki/readme-updates
Michael Neuling [Thu, 16 Jun 2022 04:38:12 +0000 (14:38 +1000)]
Merge pull request #364 from shenki/readme-updates

Readme updates

2 years agoMerge pull request #372 from antonblanchard/dcache-unused-sig
Michael Neuling [Thu, 16 Jun 2022 04:36:50 +0000 (14:36 +1000)]
Merge pull request #372 from antonblanchard/dcache-unused-sig

dcache: remove unused do_write signal

2 years agoMerge pull request #371 from antonblanchard/unused-sig
Michael Neuling [Thu, 16 Jun 2022 04:35:10 +0000 (14:35 +1000)]
Merge pull request #371 from antonblanchard/unused-sig

execute1: sub_mux_sel and result_mux_sel are unused

2 years agoMerge pull request #370 from antonblanchard/divider-init
Michael Neuling [Thu, 16 Jun 2022 04:33:45 +0000 (14:33 +1000)]
Merge pull request #370 from antonblanchard/divider-init

divider: Fix d_out.overflow U state issue

2 years agoMerge pull request #368 from antonblanchard/icache-pmu-events
Paul Mackerras [Wed, 15 Jun 2022 01:02:58 +0000 (11:02 +1000)]
Merge pull request #368 from antonblanchard/icache-pmu-events

icache: Hook up PMU events

2 years agoMerge pull request #377 from antonblanchard/fpu-init
Anton Blanchard [Tue, 14 Jun 2022 08:10:37 +0000 (18:10 +1000)]
Merge pull request #377 from antonblanchard/fpu-init

fpu: Reduce uninitialised signals

2 years agofpu: Reduce uninitialised signals
Anton Blanchard [Tue, 14 Jun 2022 05:14:19 +0000 (15:14 +1000)]
fpu: Reduce uninitialised signals

Reduce uninitialised signals coming out of the FPU.

Signed-off-by: Anton Blanchard <anton@linux.ibm.com>
2 years agoMerge pull request #366 from antonblanchard/hello-world-bss
Michael Neuling [Tue, 14 Jun 2022 03:09:57 +0000 (13:09 +1000)]
Merge pull request #366 from antonblanchard/hello-world-bss

Zero BSS in hello world test

2 years agoMerge pull request #375 from antonblanchard/core_debug-init
Anton Blanchard [Sun, 12 Jun 2022 21:15:55 +0000 (07:15 +1000)]
Merge pull request #375 from antonblanchard/core_debug-init

core_debug: Initialise gspr_index

2 years agoloadstore1: reduce U state being output
Anton Blanchard [Sun, 12 Jun 2022 12:15:11 +0000 (22:15 +1000)]
loadstore1: reduce U state being output

While these signals should only be read when valid is true, they
are only a small number of bits and we want to reduce the amount of
U/X state bouncing around the chip.

Signed-off-by: Anton Blanchard <anton@linux.ibm.com>
2 years agocore_debug: Initialise gspr_index
Anton Blanchard [Sun, 12 Jun 2022 11:49:13 +0000 (21:49 +1000)]
core_debug: Initialise gspr_index

Another case of U state being driven out of a module.

Signed-off-by: Anton Blanchard <anton@linux.ibm.com>
2 years agocore: Remove unused icache_inv signal
Anton Blanchard [Sun, 12 Jun 2022 11:04:16 +0000 (21:04 +1000)]
core: Remove unused icache_inv signal

Signed-off-by: Anton Blanchard <anton@linux.ibm.com>
2 years agoicache: Don't output X on i_out.insn
Anton Blanchard [Sun, 12 Jun 2022 01:42:32 +0000 (11:42 +1000)]
icache: Don't output X on i_out.insn

decode1 has a lot of logic that uses i_out.insn without first looking at
i_iout.valid. Play it safe and never output X state.

Signed-off-by: Anton Blanchard <anton@linux.ibm.com>
2 years agodcache: remove unused do_write signal
Anton Blanchard [Sun, 12 Jun 2022 01:39:31 +0000 (11:39 +1000)]
dcache: remove unused do_write signal

Signed-off-by: Anton Blanchard <anton@linux.ibm.com>
2 years agoexecute1: sub_mux_sel and result_mux_sel are unused
Anton Blanchard [Sun, 12 Jun 2022 00:49:26 +0000 (10:49 +1000)]
execute1: sub_mux_sel and result_mux_sel are unused

Remove them.

Signed-off-by: Anton Blanchard <anton@linux.ibm.com>
2 years agodivider: Fix d_out.overflow U state issue
Anton Blanchard [Sun, 12 Jun 2022 00:34:20 +0000 (10:34 +1000)]
divider: Fix d_out.overflow U state issue

While we should only look at this when d_out.valid = 1, we may as remove
some U state across interfaces.

Signed-off-by: Anton Blanchard <anton@linux.ibm.com>
2 years agoMerge pull request #369 from antonblanchard/loadstore-pmu-init
Anton Blanchard [Sun, 12 Jun 2022 00:24:54 +0000 (10:24 +1000)]
Merge pull request #369 from antonblanchard/loadstore-pmu-init

loadstore1: Initialise PMU events

2 years agoloadstore1: Initialise PMU events
Anton Blanchard [Sat, 11 Jun 2022 23:29:46 +0000 (09:29 +1000)]
loadstore1: Initialise PMU events

The loadstore1 PMU events are U state until a load and a store completes.

Signed-off-by: Anton Blanchard <anton@linux.ibm.com>
2 years agoMerge pull request #367 from antonblanchard/fpu-typo
Anton Blanchard [Sat, 11 Jun 2022 23:32:59 +0000 (09:32 +1000)]
Merge pull request #367 from antonblanchard/fpu-typo

fpu: Fix capitalisation of Execute1ToFPUType

2 years agoicache: Hook up PMU events
Anton Blanchard [Sat, 11 Jun 2022 23:21:56 +0000 (09:21 +1000)]
icache: Hook up PMU events

We weren't connecting the icache PMU events up.

Signed-off-by: Anton Blanchard <anton@linux.ibm.com>
2 years agofpu: Fix capitalisation of Execute1ToFPUType
Anton Blanchard [Thu, 9 Jun 2022 22:10:27 +0000 (08:10 +1000)]
fpu: Fix capitalisation of Execute1ToFPUType

While this is not an issue in VHDL, I noticed this when running
a script over the source and we may as well fix it.

Signed-off-by: Anton Blanchard <anton@linux.ibm.com>
2 years agoZero BSS in hello world test
Anton Blanchard [Wed, 8 Jun 2022 05:20:07 +0000 (15:20 +1000)]
Zero BSS in hello world test

While trying to reduce U/X state issues, I notice that our BSS is not
being initialised in the hello world test.

Signed-off-by: Anton Blanchard <anton@linux.ibm.com>
2 years agoMerge pull request #365 from antonblanchard/less-fpga-init
Anton Blanchard [Wed, 8 Jun 2022 04:54:48 +0000 (14:54 +1000)]
Merge pull request #365 from antonblanchard/less-fpga-init

Remove some FPGA style signal inits

2 years agoRemove some FPGA style signal inits
Anton Blanchard [Tue, 7 Jun 2022 10:01:14 +0000 (20:01 +1000)]
Remove some FPGA style signal inits

These don't work on the ASIC flow, so remove them and initialise
them explicitly where required.

Signed-off-by: Anton Blanchard <anton@linux.ibm.com>
2 years agoRemove some FPGA style signal inits
Anton Blanchard [Tue, 7 Jun 2022 07:38:24 +0000 (17:38 +1000)]
Remove some FPGA style signal inits

These don't work on the ASIC flow, so remove them and initialise
them explicitly where required.

Signed-off-by: Anton Blanchard <anton@linux.ibm.com>
2 years agoREADME: Add Linux on Microwatt instructions
Joel Stanley [Tue, 7 Jun 2022 03:20:03 +0000 (12:50 +0930)]
README: Add Linux on Microwatt instructions

These instructions are similar to those at

 https://ozlabs.org/~joel/microwatt/README

except they describe how to build the artifacts from scratch instead of
downloading them.

Signed-off-by: Joel Stanley <joel@jms.id.au>
2 years agoREADME: Add uart to fusesoc instructions
Joel Stanley [Tue, 7 Jun 2022 03:18:42 +0000 (12:48 +0930)]
README: Add uart to fusesoc instructions

The SoC defaults to using the uart16550 so provide instructions on how
to fetch that library when seetting up fusesoc.

Also remove the text about a working directory; fusesoc doesn't need
one.

Signed-off-by: Joel Stanley <joel@jms.id.au>
2 years agoMerge pull request #361 from antonblanchard/alt-reset-address
Michael Neuling [Tue, 22 Mar 2022 00:55:54 +0000 (11:55 +1100)]
Merge pull request #361 from antonblanchard/alt-reset-address

Allow ALT_RESET_ADDRESS to be overridden

2 years agoAllow ALT_RESET_ADDRESS to be overridden
Anton Blanchard [Mon, 21 Mar 2022 22:35:17 +0000 (09:35 +1100)]
Allow ALT_RESET_ADDRESS to be overridden

This allows us to boot from flash for example.

Signed-off-by: Anton Blanchard <anton@linux.ibm.com>
2 years agoMerge pull request #360 from antonblanchard/log2ceil-issue
Michael Neuling [Fri, 18 Mar 2022 07:28:34 +0000 (18:28 +1100)]
Merge pull request #360 from antonblanchard/log2ceil-issue

wishbone_bram_wrapper ram_addr_bits is 1 bit off

2 years agowishbone_bram_wrapper ram_addr_bits is 1 bit off
Anton Blanchard [Thu, 17 Mar 2022 07:03:29 +0000 (18:03 +1100)]
wishbone_bram_wrapper ram_addr_bits is 1 bit off

log2ceil() returns the number of bits required to store a value, so we
need to pass in memory_size-1, not memory_size.

Every other user of log2ceil() gets this right.

Signed-off-by: Anton Blanchard <anton@linux.ibm.com>
2 years agoMerge pull request #358 from antonblanchard/unused-sig
Michael Neuling [Tue, 15 Mar 2022 23:49:47 +0000 (10:49 +1100)]
Merge pull request #358 from antonblanchard/unused-sig

Remove unused sequential signal from Fetch1ToIcacheType

2 years agoMerge pull request #356 from antonblanchard/fpu-constant
Michael Neuling [Tue, 15 Mar 2022 23:49:29 +0000 (10:49 +1100)]
Merge pull request #356 from antonblanchard/fpu-constant

fpu: Make inverse_table a constant

2 years agoMerge pull request #357 from antonblanchard/xics-warning
Michael Neuling [Tue, 15 Mar 2022 23:48:59 +0000 (10:48 +1100)]
Merge pull request #357 from antonblanchard/xics-warning

xics: Fix warning when comparing two std_ulogic_vectors

2 years agoRemove unused sequential signal from Fetch1ToIcacheType
Anton Blanchard [Tue, 15 Mar 2022 07:27:48 +0000 (18:27 +1100)]
Remove unused sequential signal from Fetch1ToIcacheType

GHDL synthesis is flagging a warning about this.

Signed-off-by: Anton Blanchard <anton@linux.ibm.com>
2 years agoxics: Fix warning when comparing two std_ulogic_vectors
Anton Blanchard [Tue, 15 Mar 2022 05:04:18 +0000 (16:04 +1100)]
xics: Fix warning when comparing two std_ulogic_vectors

Use unsigned() to make it clear what we are doing.

Signed-off-by: Anton Blanchard <anton@linux.ibm.com>
2 years agofpu: Make inverse_table a constant
Anton Blanchard [Tue, 15 Mar 2022 05:03:34 +0000 (16:03 +1100)]
fpu: Make inverse_table a constant

GHDL synthesis is complaining that inverse_table is never stored to.
Change it to a constant.

Signed-off-by: Anton Blanchard <anton@linux.ibm.com>
2 years agofix: fix icache_tb not finishing correctly
Tianrui Wei [Tue, 1 Mar 2022 15:51:35 +0000 (23:51 +0800)]
fix: fix icache_tb not finishing correctly

Setting icache to be privileged and accessing physical memory directly.
And set big_endian to 0 to correspond to the testbench result.

Signed-off-by: Tianrui Wei <tianrui@tianruiwei.com>
2 years agoMerge pull request #352 from mkj/static-urjtag
Michael Neuling [Sun, 27 Feb 2022 21:17:50 +0000 (08:17 +1100)]
Merge pull request #352 from mkj/static-urjtag

mw_debug: Add STATIC_URJTAG flag

2 years agomw_debug: Add STATIC_URJTAG flag
Matt Johnston [Fri, 25 Feb 2022 09:43:28 +0000 (17:43 +0800)]
mw_debug: Add STATIC_URJTAG flag

Revert to linking dynamically by default, can statically link with
`make STATIC_URJTAG=1`

Fixes #351

Signed-off-by: Matt Johnston <matt@codeconstruct.com.au>
2 years agoUpdate the README Issues (#350)
Michael Neuling [Fri, 25 Feb 2022 02:18:38 +0000 (13:18 +1100)]
Update the README Issues (#350)

We've had these for a while now:
 - D/I cache
 - GPR bypassing
 - Supervisor state (and can boot linux)

We still need Vector/VMX/VSX (and probably some other things)

Signed-off-by: Michael Neuling <mikey@neuling.org>
2 years agoMerge pull request #349 from madscientist159/master
Michael Neuling [Fri, 25 Feb 2022 00:08:57 +0000 (11:08 +1100)]
Merge pull request #349 from madscientist159/master

Extend LiteDRAM VHDL wrapper to allow more than one clock line

2 years agoExtend LiteDRAM VHDL wrapper to allow more than one clock line
Raptor Engineering Development Team [Tue, 22 Feb 2022 17:49:33 +0000 (11:49 -0600)]
Extend LiteDRAM VHDL wrapper to allow more than one clock line

This is necessary for the upcoming Arctic Tern system enablement,
since Arctic Tern uses two DRAM devices and a separate clock line
is routed to each device.  LiteX handles this behavior correctly,
therefore we assume other hardware exists that uses a similar
DRAM clock design.

Updates from Mikey to fix some compile issues.

Signed-off-by: Timothy Pearson <tpearson@raptorengineering.com>
Signed-off-by: Michael Neuling <mikey@neuling.org>
2 years agoMerge pull request #348 from paulusmack/reduce
Michael Neuling [Wed, 23 Feb 2022 01:03:59 +0000 (12:03 +1100)]
Merge pull request #348 from paulusmack/reduce

Reduce LUT usage

2 years agoxics: Rework the irq_gen process
Paul Mackerras [Tue, 19 Oct 2021 04:13:31 +0000 (15:13 +1100)]
xics: Rework the irq_gen process

At present, the loop in the irq_gen process generates a chain of
comparators and other logic to work out the source number and priority
of the most-favoured (lowest priority number) pending interrupt.
This replaces that chain with (1) logic to generate an array of bits,
one per priority, indicating whether any interrupt is pending at that
priority, (2) a priority encoder to select the most favoured priority
with an interrupt pending, (3) logic to generate an array of bits, one
per source, indicating whether an interrupt is pending at the priority
calculated in step 2, and (4) a priority encoder to work out the
lowest numbered source that has an interrupt pending at the selected
priority.  This reduces LUT utilization.

The priority encoder function implemented here uses the optimized
count-leading-zeroes logic from helpers.vhdl.

Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
2 years agoUse alternative count-leading-zeroes algorithm in the FPU and LSU
Paul Mackerras [Mon, 21 Feb 2022 01:06:11 +0000 (12:06 +1100)]
Use alternative count-leading-zeroes algorithm in the FPU and LSU

Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
2 years agocountzero: Use alternative algorithm for higher bits
Paul Mackerras [Sun, 20 Feb 2022 22:58:07 +0000 (09:58 +1100)]
countzero: Use alternative algorithm for higher bits

This implements an alternative count-leading-zeroes algorithm which
uses less LUTs to generate the higher-order bits (2..5) of the
result.

By doing (v | -v) rather than (v & -v), we get a value which has ones
from the MSB down to the rightmost 1 bit in v and then zeroes down to
the LSB.  This means that we can generate the MSB of the result (the
index of the rightmost 1 bit in v) just by looking at bits 63 and 31
of (v | -v), assuming that v is 64 bits.  Bit 4 of the result requires
looking at bits 63, 47, 31 and 15.  In contrast, each bit of the
result using (v & -v), which has a single 1, requires ORing together
32 bits.

It turns out that the minimum LUT usage comes from using (v & -v) to
generate bits 0 and 1 of the result, and using (v | -v) to generate
bits 2 to 5.  This saves almost 60 6-input LUTs on the Artix-7.

Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
2 years agosoc: Re-do peripheral address decode to improve timing
Paul Mackerras [Mon, 11 Oct 2021 06:46:44 +0000 (17:46 +1100)]
soc: Re-do peripheral address decode to improve timing

This generates a series of io_cycle_* signals which are clean latches
and which become the 'cyc' signals of the wishbone buses going to
various peripherals (syscon, uarts, XICS, GPIO, etc.).  Effectively
this is done by moving the address decoding into the slave_io_latch
process.  The slave_io_type, which drives the multiplexer which
selects which wishbone to look for a response on, is reduced to just 8
values in the expectation that an 8-way multiplexer will use less
logic than one with more than 8 inputs.

With this timing is considerably better on the A7-100T.

Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
2 years agoMerge pull request #346 from mkj/dmi_ecp5
Michael Neuling [Mon, 7 Feb 2022 22:09:22 +0000 (09:09 +1100)]
Merge pull request #346 from mkj/dmi_ecp5

Add DMI and mw_debug for ECP5

2 years agoMerge pull request #343 from mikey/orange-crab-ci
Anton Blanchard [Mon, 7 Feb 2022 06:57:08 +0000 (17:57 +1100)]
Merge pull request #343 from mikey/orange-crab-ci

ci: Add new Orange Crab build

2 years agomw_debug: Add Lattice ECP5 support
Matt Johnston [Fri, 4 Feb 2022 07:29:40 +0000 (15:29 +0800)]
mw_debug: Add Lattice ECP5 support

"-b ecp5" will select ECP5 interface that talks to a JTAGG
primitive.

For example with a FT232H JTAG board:

./mw_debug  -t 'ft2232 vid=0x0403 pid=0x6014'  -s 30000000 -b ecp5 mr ff003888 6
Connected to libftdi driver.
Found device ID: 0x41113043
00000000ff0038886d6f636c65570a0a  ..Welcom
00000000ff00389063694d206f742065  e to Mic
00000000ff0038982120747461776f72  rowatt !
00000000ff0038a00000000000000a0a  ........
00000000ff0038a867697320636f5320   Soc sig
00000000ff0038b0203a65727574616e  nature:
Core: running
 NIA: c0000000000187f8
 MSR: 9000000000001033

Signed-off-by: Matt Johnston <matt@codeconstruct.com.au>
2 years agodmi_dtm_ecp5: Use ECP5 JTAGG for DMI
Matt Johnston [Fri, 26 Nov 2021 02:47:07 +0000 (10:47 +0800)]
dmi_dtm_ecp5: Use ECP5 JTAGG for DMI

This uses the JTAGG primitive which is similar to BSCANE2.
The LUT4 delay approach came from Florian and Greg in
https://github.com/enjoy-digital/litex/pull/1087

Has been tested on an OrangeCrab with 48MHz sysclk
FT232H up to 30MHz (though libusb/urjtag is by far the bottleneck vs
the JTAG clock)

Signed-off-by: Matt Johnston <matt@codeconstruct.com.au>
2 years agomw_debug: Link urjtag statically
Matt Johnston [Fri, 4 Feb 2022 06:40:42 +0000 (14:40 +0800)]
mw_debug: Link urjtag statically

liburjtag isn't in Debian, so usually we're pointing at a urjtag
build directory when building mw_debug

Signed-off-by: Matt Johnston <matt@codeconstruct.com.au>
2 years agomw_debug: use isxdigit for hex arguments
Matt Johnston [Fri, 4 Feb 2022 04:08:07 +0000 (12:08 +0800)]
mw_debug: use isxdigit for hex arguments

Signed-off-by: Matt Johnston <matt@codeconstruct.com.au>
2 years agomw_debug: Add -s frequency argument
Matt Johnston [Fri, 26 Nov 2021 02:43:06 +0000 (10:43 +0800)]
mw_debug: Add -s frequency argument

Chose -s for speed, vs -f for --force

Signed-off-by: Matt Johnston <matt@codeconstruct.com.au>
2 years agomw_debug: pass target parameters to urjtag
Matt Johnston [Thu, 25 Nov 2021 06:12:13 +0000 (14:12 +0800)]
mw_debug: pass target parameters to urjtag

An example

./mw_debug -d -t 'ft2232 vid=0x0403 pid=0x6014'

Signed-off-by: Matt Johnston <matt@codeconstruct.com.au>
2 years agofetch1/icache1: Remove the use_previous logic
Paul Mackerras [Mon, 11 Oct 2021 06:23:08 +0000 (17:23 +1100)]
fetch1/icache1: Remove the use_previous logic

This removes logic that I added some time ago with the thought that it
would enable us to do prefetching in the icache.  This logic detects
when the fetch address is an odd multiple of 4 and the next address in
sequence from the previous cycle.  In that case the instruction we
want is in the output register of the icache RAM already so there is
no need to do another read or any icache tag or TLB lookup.

However, this logic adds complexity, and removing it improves timing,
so this removes it.

Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
2 years agoMerge pull request #345 from antonblanchard/popcnt-go-fast
Paul Mackerras [Fri, 4 Feb 2022 00:43:42 +0000 (11:43 +1100)]
Merge pull request #345 from antonblanchard/popcnt-go-fast

popcnt* timing improvements from Paul

2 years agocore: Make popcnt* take two cycles
Paul Mackerras [Tue, 19 Oct 2021 01:22:10 +0000 (12:22 +1100)]
core: Make popcnt* take two cycles

This moves the calculation of the result for popcnt* into the
countbits unit, renamed from countzero, so that we can take two cycles
to get the result.  The motivation for this is that the popcnt*
calculation was showing up as a critical path.

Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
2 years agoci: Add new Orange Crab build
Michael Neuling [Tue, 18 Jan 2022 01:41:03 +0000 (12:41 +1100)]
ci: Add new Orange Crab build

This builds the Orange Crab v0.21 + litedram image

Signed-off-by: Michael Neuling <mikey@neuling.org>
2 years agoMerge pull request #342 from mkj/orangecrab-merge
Michael Neuling [Tue, 18 Jan 2022 02:27:27 +0000 (13:27 +1100)]
Merge pull request #342 from mkj/orangecrab-merge

Orangecrab working with litedram

Fixed up a few simple merge conflicts in the Makefile.

2 years agoMerge branch 'master' into orangecrab-merge
Michael Neuling [Tue, 18 Jan 2022 01:03:46 +0000 (12:03 +1100)]
Merge branch 'master' into orangecrab-merge

2 years agoMerge pull request #341 from mkj/progtools
Michael Neuling [Tue, 18 Jan 2022 00:51:54 +0000 (11:51 +1100)]
Merge pull request #341 from mkj/progtools

orangecrab programming targets

2 years agoMerge pull request #340 from mkj/orangecrab-ghdl-plugin
Michael Neuling [Tue, 18 Jan 2022 00:50:22 +0000 (11:50 +1100)]
Merge pull request #340 from mkj/orangecrab-ghdl-plugin

Makefile: detect when ghdl is a yosys plugin

2 years agoorangecrab: Fix sdcard wishbone addressing
Matt Johnston [Fri, 19 Nov 2021 05:13:15 +0000 (13:13 +0800)]
orangecrab: Fix sdcard wishbone addressing

Orangecrab missed out on:

Make wishbone addresses be in units of doublewords or words
Author: Paul Mackerras <paulus@ozlabs.org>
Date:   Wed Sep 15 18:18:09 2021 +1000

Signed-off-by: Matt Johnston <matt@codeconstruct.com.au>
2 years agoorangecrab: use litesdcard
Matt Johnston [Fri, 14 Jan 2022 00:04:18 +0000 (08:04 +0800)]
orangecrab: use litesdcard

Currently not working (tested in Linux)

Signed-off-by: Matt Johnston <matt@codeconstruct.com.au>
2 years agolitesdcard: add lattice, regenerate
Matt Johnston [Fri, 14 Jan 2022 00:08:09 +0000 (08:08 +0800)]
litesdcard: add lattice, regenerate

Modifies litescard generate script to take a clock speed.

Regenerated verilog with latest litesdcard
e52c731 ("Bump year.")

Signed-off-by: Matt Johnston <matt@codeconstruct.com.au>
2 years agoorangecrab: No BTC, LOG_LENGTH, dram NUM_LINES
Matt Johnston [Mon, 23 Aug 2021 02:30:40 +0000 (10:30 +0800)]
orangecrab: No BTC, LOG_LENGTH, dram NUM_LINES

Reduce litedram NUM_LINES 64->8
This allows us to meet timing. Can probably
be improved in future with better BRAM usage.

Signed-off-by: Matt Johnston <matt@codeconstruct.com.au>
2 years agoorangecrab: Use litedram
Matt Johnston [Thu, 13 Jan 2022 08:51:57 +0000 (16:51 +0800)]
orangecrab: Use litedram

Signed-off-by: Matt Johnston <matt@codeconstruct.com.au>
2 years agoorangecrab: set HAS_SHORT_MULT
Matt Johnston [Wed, 24 Nov 2021 08:47:16 +0000 (16:47 +0800)]
orangecrab: set HAS_SHORT_MULT

It seems free, generated as a single MULT18X18D

Signed-off-by: Matt Johnston <matt@codeconstruct.com.au>
2 years agoorangecrab: add Orange Crab r0.2 target
Matt Johnston [Wed, 11 Aug 2021 05:11:57 +0000 (13:11 +0800)]
orangecrab: add Orange Crab r0.2 target

top-orangecrab0.2 is a copy of top-arty with various changes.
USRMCLK is added for the SPI clock
ethernet is removed

Signed-off-by: Matt Johnston <matt@codeconstruct.com.au>
2 years agolitedram: Add orangecrab-85-0.2 target
Matt Johnston [Fri, 13 Aug 2021 02:07:15 +0000 (10:07 +0800)]
litedram: Add orangecrab-85-0.2 target

Parameters are based on
https://github.com/gregdavill/OrangeCrab-test-sw/blob/main/hw/OrangeCrab-bitstream.py
and litex-boards orangecrab.py

rtt_nom and cmd_delay are overridden for OrangeCrab, we do the same here.

Generated with litedram and litex
62abf9c ("litedram_gen: Add block_until_ready port parameter to control blocking behaviour.")
add2746a ("tools/litex_cli: Rename wb to bus.")

Signed-off-by: Matt Johnston <matt@codeconstruct.com.au>
2 years agolitedram: set Makefile -Werror
Matt Johnston [Fri, 24 Sep 2021 04:24:29 +0000 (12:24 +0800)]
litedram: set Makefile -Werror

Signed-off-by: Matt Johnston <matt@codeconstruct.com.au>
2 years agolitedram: disable block_until_ready, regenerate
Matt Johnston [Thu, 13 Jan 2022 23:23:35 +0000 (07:23 +0800)]
litedram: disable block_until_ready, regenerate

Recent litedram gets stuck at memtest unless block_until_ready=False.
(discussion in https://github.com/enjoy-digital/litedram/pull/292)

This change regenerates with latest litedram and litex
62abf9c ("litedram_gen: Add block_until_ready port parameter to control blocking behaviour.")
add2746a ("tools/litex_cli: Rename wb to bus.")

Signed-off-by: Matt Johnston <matt@codeconstruct.com.au>
2 years agoMakefile: add ecpprog targets
Matt Johnston [Fri, 26 Nov 2021 02:33:55 +0000 (10:33 +0800)]
Makefile: add ecpprog targets

The 0x80000 offset is specific to the OrangeCrab bootloader.

Signed-off-by: Matt Johnston <matt@codeconstruct.com.au>