Commit Graph

39 Commits

Author SHA1 Message Date
whitequark 02b8426f60 firmware: rename libboard to libboard_misoc.
By analogy with libbuild_misoc. Hopefully one day it actually gets
moved to misoc...
2018-05-14 18:09:44 +00:00
whitequark 1ff42683d9 firmware: merge libamp into libboard_artiq.
All gateware interfaces live in libboard_artiq, but libamp predates
libboard_artiq.
2018-05-14 17:33:25 +00:00
whitequark 6a10d54432 firmware: migrate rpc_proto to new libio.
This closes an old and horrible issue in that some code in ksupport
may implicitly try to allocate and we'll never know until it crashes
at runtime inside liballoc_stub.

This also removes liballoc_stub.
This also removes ReadExt and WriteExt from libproto.
Hooray!
2018-05-14 17:26:59 +00:00
whitequark 623614f835 Update LLVM to 6.0.0 and Rust to 1.25.0. 2018-05-01 08:57:37 +00:00
Sebastien Bourdeauducq 928d5dc9b3 drtio: raise RTIOLinkError if operation fails due to link lost (#942) 2018-03-04 01:02:53 +08:00
Sebastien Bourdeauducq dc593ec0f0 Merge branch 'rtio-sed' into sed-merge 2018-01-10 12:04:54 +08:00
whitequark b8f3d28bc0 firmware: update compiler_builtins dependency and move it to libboard.
This is so that we only specify it once.
2017-12-31 09:16:11 +00:00
whitequark 7687a34285 firmware: remove dependency on compiler-rt. 2017-12-28 04:08:29 +00:00
whitequark 4dfe71676e firmware: remove dependency on libbase. 2017-12-28 04:03:27 +00:00
whitequark 3eec15c01d firmware: fix compatibility with newer rustc. NFC. 2017-12-25 17:16:05 +00:00
whitequark 2404a0d8c8 runtime: allow #[cfg(not(has_rtio))] builds. 2017-11-03 15:49:30 +00:00
whitequark 73c76ebb9a firmware: eliminate most transmutes and document the rest. NFC. 2017-10-02 03:09:46 +00:00
Sebastien Bourdeauducq 5437f0e3e3 rtio: make sequence errors consistently asychronous 2017-09-29 14:40:06 +08:00
whitequark 7799413a41 firmware: update Rust from 1.18.0 to 1.20.0. 2017-09-24 22:51:44 +00:00
Sebastien Bourdeauducq 8399f8893d add kernel access to non-realtime SPI buses (#740) 2017-06-18 12:45:07 +08:00
Sebastien Bourdeauducq fe05aede78 firmware: DmaPlayback → DMARetrieve 2017-04-19 11:11:42 +08:00
Sebastien Bourdeauducq e8f7f8ef9c DMA: speed up playback
Time reduced from 1.53µs to 1.37µs.
2017-04-19 10:59:12 +08:00
whitequark 41c4de4556 DMA: add API for a much faster replay using handles. 2017-04-18 08:20:12 +00:00
whitequark 0531dc45c3 DMA: erase trace before re-recording it.
Or we could needlessly OOM replacing a large trace.
2017-04-15 07:48:02 +00:00
whitequark 9dfe9c1248 DMA: improve recording performance.
This commit moves DMA serialization code to the kernel CPU
(to cope with the existence of rtio_output_wide) and batches
the resulting sequences. This results in less data being transferred
between kernel and comms CPUs (24 octets with one pointer before,
18 octets with no pointers now, for the common case of rtio_output),
but most importantly reduces cache flushes, which now happen
once per 64k octets.

On average, it now takes about 15us to record a single RTIO event
in a DMA trace.

Fixes #712.
2017-04-15 07:29:52 +00:00
whitequark 14ae1cc100 runtime: advance now on DMA replay.
Fixes #700.
2017-04-05 18:34:36 +00:00
whitequark 391660e545 gateware: simplify the CRI arbiter to use a plain mux. 2017-04-05 15:09:19 +00:00
Sebastien Bourdeauducq b74d6fb9ba make collision and busy asynchronous errors, and simplify CPU/gateware handshake for output errors and reads 2017-03-27 16:32:23 +08:00
whitequark 924ae73000 firmware: fix DMA trace alignment and flush caches. 2017-03-04 09:40:22 +00:00
whitequark 30ac42de3f ksupport: fix an exception safety issue.
Raising from inside a recv! will never send an acknowledgement.
2017-03-04 09:40:22 +00:00
Sebastien Bourdeauducq 0b4922e0f4 dma: request RTIO control 2017-03-01 18:59:45 +08:00
whitequark a4ece19614 Implement DMA playback. 2017-02-28 21:28:40 +00:00
whitequark 88bf7d2233 firmware: specialize protocol read/write functions.
Before this commit, proto::io::{read,write}_* functions were taking
a &mut {Read,Write}. This means a lot of virtual dispatch.
After this commit, all these functions are specialized for
the specific IO trait.

This could be achieved with just changing the signature from
  fn read_x(reader: &mut Read)
to
  fn read_x<R: Read>(reader: &mut R)
but the functions were also grouped into ReadExt and WriteExt
traits as a refactoring.

Initially, it was expected that the generic traits from
the byteorder crate could be used, but they require endianness
to be specified on every call and thus aren't very ergonomic.
They also lack the equivalent to our read_string and read_bytes.

Thus, it seems fine to just define a slightly different extension
trait.

This also optimized the test_rpc_timing test: 1.7ms→1.2ms.
2017-02-26 18:24:37 +00:00
whitequark 5d3b00cf12 Implement recording of DMA traces on the core device. 2017-02-26 02:50:20 +00:00
whitequark a07bd918f0 firmware: use Rust naming conventions for enum variants. 2017-02-25 22:34:11 +00:00
whitequark 2a81819eb0 firmware: restructure to avoid #[path = "..."] mod ...;
Such code is fragile, introduces mess in dependencies, and
inflates compile times.
2017-02-25 17:54:14 +00:00
whitequark d04e611232 firmware, compiler: rename rpc functions to be more consistent. 2017-02-25 14:12:58 +00:00
whitequark 04ad267055 firmware: rewrite the dynamic linker in Rust. 2017-02-24 18:57:29 +00:00
whitequark 25c9b8827a firmware: mark __artiq_{re,}raise as #[unwind].
This is required for correctness.
2017-02-05 09:04:22 +00:00
whitequark c39394b4d5 firmware: port the exception handling routines to Rust. 2017-02-04 08:21:07 +00:00
whitequark cde2054926 firmware: do not link to C code in runtime and satman. 2017-02-03 12:12:41 +00:00
whitequark fd8b11532f compiler, firmware: use Pascal strings everywhere.
This removes a large amount of very ugly code, and also simplifies
the compiler and runtime.
2017-02-03 11:53:27 +00:00
whitequark c2cf60e7f6 firmware: use the cslice crate to get rid of unsafe code in ksupport. 2017-02-02 00:51:58 +00:00
whitequark bc22d1c009 firmware: libksupport → ksupport.
The idea is that support libraries are called lib*, and the final
products that linked into an executable are not.
2017-01-31 22:24:37 +00:00