- shared common code path float16/32/64
- well commented and easy to follow code
- added a bunch of float16 support
While some operations are slower the key ones exercised by the
floating point dbt-bench are the same: https://i.imgur.com/oXNJNql.png
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1
iQEcBAABAgAGBQJajU1+AAoJEPvQ2wlanipEg3gH/jzOmmJm5URYPmBBYqG4LbOO
jZVUP8mfRT/xEHvD0BExXFevF2Be+YzqKoEyspcJTgjKKjvhc/t0rgC3umEib7H+
r/kzpGB4C9YodO4sXqPUJQHWnfjAvRE6VJKW+KaRhvxQAmo3pt2e01xxPCcqCfLx
ww8ASGxnPn8NgWmOOt+8BIO5wm/9bU/2Fg0blyqVEh8x97wGR+YV/H2B6PtEjkHo
3Wba/zK0kq3wyw+J4VeHg+ODaR0kfBGjF+VQ3NJOhk7IW1Tztm+Jb80ShnIWzA3y
WhiY1sbj009uJEbaG6ImUm1WN0rMvMnV2b72bK1bc/HREU7SJfQxAoSHGh54Pc8=
=/kJP
-----END PGP SIGNATURE-----
Merge remote-tracking branch 'remotes/stsquad/tags/pull-softfloat-refactor-210218-1' into staging
This is the re-factor of softfloat:
- shared common code path float16/32/64
- well commented and easy to follow code
- added a bunch of float16 support
While some operations are slower the key ones exercised by the
floating point dbt-bench are the same: https://i.imgur.com/oXNJNql.png
# gpg: Signature made Wed 21 Feb 2018 10:44:14 GMT
# gpg: using RSA key FBD0DB095A9E2A44
# gpg: Good signature from "Alex Bennée (Master Work Key) <alex.bennee@linaro.org>"
# Primary key fingerprint: 6685 AE99 E751 67BC AFC8 DF35 FBD0 DB09 5A9E 2A44
* remotes/stsquad/tags/pull-softfloat-refactor-210218-1: (22 commits)
fpu/softfloat: re-factor sqrt
fpu/softfloat: re-factor compare
fpu/softfloat: re-factor minmax
fpu/softfloat: re-factor scalbn
fpu/softfloat: re-factor int/uint to float
fpu/softfloat: re-factor float to int/uint
fpu/softfloat: re-factor round_to_int
fpu/softfloat: re-factor muladd
fpu/softfloat: re-factor div
fpu/softfloat: re-factor mul
fpu/softfloat: re-factor add/sub
fpu/softfloat: define decompose structures
fpu/softfloat: move the extract functions to the top of the file
fpu/softfloat: improve comments on ARM NaN propagation
include/fpu/softfloat: add some float16 constants
include/fpu/softfloat: implement float16_set_sign helper
include/fpu/softfloat: implement float16_chs helper
include/fpu/softfloat: implement float16_abs helper
target/*/cpu.h: remove softfloat.h
fpu/softfloat-types: new header to prevent excessive re-builds
...
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
This is a little bit of a departure from softfloat's original approach
as we skip the estimate step in favour of a straight iteration. There
is a minor optimisation to avoid calculating more bits of precision
than we need however this still brings a performance drop, especially
for float64 operations.
Suggested-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
The compare function was already expanded from a macro. I keep the
macro expansion but move most of the logic into a compare_decomposed.
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Let's do the same re-factor treatment for minmax functions. I still
use the MACRO trick to expand but now all the checking code is common.
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
This is one of the simpler manipulations you could make to a floating
point number.
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
These are considerably simpler as the lower order integers can just
use the higher order conversion function. As the decomposed fractional
part is a full 64 bit rounding and inexact handling comes from the
pack functions.
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
We share the common int64/uint64_pack_decomposed function across all
the helpers and simply limit the final result depending on the final
size.
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
We can now add float16_round_to_int and use the common round_decomposed and
canonicalize functions to have a single implementation for
float16/32/64 round_to_int functions.
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
We can now add float16_muladd and use the common decompose and
canonicalize functions to have a single implementation for
float16/32/64 muladd functions.
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
We can now add float16_div and use the common decompose and
canonicalize functions to have a single implementation for
float16/32/64 versions.
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
We can now add float16_mul and use the common decompose and
canonicalize functions to have a single implementation for
float16/32/64 versions.
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
We can now add float16_add/sub and use the common decompose and
canonicalize functions to have a single implementation for
float16/32/64 add and sub functions.
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
These structures pave the way for generic softfloat helper routines
that will operate on fully decomposed numbers.
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
This is pure code-motion during re-factoring as the helpers will be
needed earlier.
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Mention the pseudo-code fragment from which this is based.
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
This defines the same set of common constants for float 16 as defined
for 32 and 64 bit floats. These are often used by target helper
functions. I've also removed constants that are not used by anybody.
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
This will be required when expanding the MINMAX() macro for 16
bit/half-precision operations.
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
As cpu.h is another typically widely included file which doesn't need
full access to the softfloat API we can remove the includes from here
as well. Where they do need types it's typically for float_status and
the rounding modes so we move that to softfloat-types.h as well.
As a result of not having softfloat in every cpu.h call we now need to
add it to various helpers that do need the full softfloat.h
definitions.
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
[For PPC parts]
Acked-by: David Gibson <david@gibson.dropbear.id.au>
The main culprit here is bswap.h which pulled in softfloat.h so it
could use the types in its CPU_Float* and ldfl/stfql functions. As
bswap.h is very widely included this added a compile dependency every
time we touch softfloat.h. Move the typedefs for each float type into
their own file so we don't re-build the world every time we tweak the
main softfloat.h header.
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
It's not actively built and when enabled things fail to compile. I'm
not sure the type-checking is really helping here. Seeing as we "own"
our softfloat now lets remove the cruft.
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
This will be required when expanding the MINMAX() macro for 16
bit/half-precision operations.
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Fill the terminal box from right to left to avoid
Gtk-WARNING **: Allocating size to GtkScrollbar 0x55f6d54b0200 without
calling gtk_widget_get_preferred_width/height(). How does the code
know the size to allocate?
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Message-id: 902aaef8-d20e-0530-dea2-cdfe3db33ff3@web.de
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Switch over all leftover users to qapi DisplayType.
Then delete the unused display_type variable.
Add 'default' DisplayType, which isn't an actual display type but
a placeholder for "user didn't specify a display". It will be replaced
by the DisplayType actually used, which in turn depends on the
DisplayTypes availabel in the particular build.
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-id: 20180202111022.19269-13-kraxel@redhat.com
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Switch over the one leftover user to qapi DisplayType.
The delete the unused request_opengl variable.
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-id: 20180202111022.19269-12-kraxel@redhat.com
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Not used any more, delete it.
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-id: 20180202111022.19269-11-kraxel@redhat.com
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Not used any more, delete it.
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Message-id: 20180202111022.19269-7-kraxel@redhat.com
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Add QAPI DisplayType enum, DisplayOptions union and DisplayGTK struct.
Switch gtk configuration to use the qapi type.
Some bookkeeping (fullscreen for example) is done twice now, this is
temporary until more/all UIs are switched over to qapi configuration.
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-id: 20180202111022.19269-5-kraxel@redhat.com
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
qapi DisplayType will replace the current enum. For the transition both
will coexist though, so rename it so we don't have a name clash.
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-id: 20180202111022.19269-4-kraxel@redhat.com
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Leak found thanks to ASAN:
Direct leak of 8 byte(s) in 1 object(s) allocated from:
#0 0x55995789ac90 in __interceptor_malloc (/home/elmarco/src/qemu/build/x86_64-softmmu/qemu-system-x86_64+0x1510c90)
#1 0x7f0a91190f0c in g_malloc /home/elmarco/src/gnome/glib/builddir/../glib/gmem.c:94
#2 0x5599580a281c in v9fs_path_copy /home/elmarco/src/qemu/hw/9pfs/9p.c:196:17
#3 0x559958f9ec5d in coroutine_trampoline /home/elmarco/src/qemu/util/coroutine-ucontext.c:116:9
#4 0x7f0a8766ebbf (/lib64/libc.so.6+0x50bbf)
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Signed-off-by: Greg Kurz <groug@kaod.org>
lhs/rhs doesn't tell much about how argument are handled, dst/src is
and const arguments is clearer in my mind. Use g_memdup() while at it.
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Signed-off-by: Greg Kurz <groug@kaod.org>
SystemTap's dtrace(1) produces the following warning when it encounters
"char const" instead of "const char":
Warning: /usr/bin/dtrace:trace-dtrace-root.dtrace:66: syntax error near:
probe flatview_destroy_rcu
Warning: Proceeding as if --no-pyparsing was given.
This is a limitation in current SystemTap releases. I have sent a patch
upstream to accept "char const" since it is valid C:
https://sourceware.org/ml/systemtap/2018-q1/msg00017.html
In QEMU we still wish to avoid warnings in the current SystemTap
release. It's simple enough to replace "char const" with "const char".
I'm not changing the documentation or implementing checks to prevent
this from occurring again in the future. The next release of SystemTap
will hopefully resolve this issue.
Cc: Daniel P. Berrange <berrange@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Daniel P. Berrange <berrange@redhat.com>
Message-id: 20180201162625.4276-1-stefanha@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Previously functions having arguments of type bool was not traced
properly. The bool arguments were missing from the trace.
Signed-off-by: Jon Emil Jahren <jonemilj@gmail.com>
Message-id: 20180129041648.30884-3-jonemilj@gmail.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Using the greedy star matching, arguments like "...%"PRIx64 caused issues
for functions with multiple PRI formats.
The issue was only seen with the ust backend, as it is the only one
using the format regex.
The result for many functions was that the arguments coming after the
greedy star end was left out of the tracepoint, and in some cases some
of the arguments that was traced had the wrong format.
Signed-off-by: Jon Emil Jahren <jonemilj@gmail.com>
Message-id: 20180129041648.30884-2-jonemilj@gmail.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
PVRDMA is the QEMU implementation of VMware's paravirtualized RDMA device.
It works with its Linux Kernel driver AS IS, no need for any special
guest modifications.
While it complies with the VMware device, it can also communicate with
bare metal RDMA-enabled machines and does not require an RDMA HCA in the
host, it can work with Soft-RoCE (rxe).
It does not require the whole guest RAM to be pinned allowing memory
over-commit and, even if not implemented yet, migration support will be
possible with some HW assistance.
Implementation is divided into 2 components, rdma general and pvRDMA
specific functions and structures.
The second PVRDMA sub-module - interaction with PCI layer.
- Device configuration and setup (MSIX, BARs etc).
- Setup of DSR (Device Shared Resources)
- Setup of device ring.
- Device management.
Reviewed-by: Dotan Barak <dotanb@mellanox.com>
Reviewed-by: Zhu Yanjun <yanjun.zhu@oracle.com>
Signed-off-by: Yuval Shaia <yuval.shaia@oracle.com>
Signed-off-by: Marcel Apfelbaum <marcel@redhat.com>
This layer is composed of two sub-modules, backend and resource manager.
Backend sub-module is responsible for all the interaction with IB layers
such as ibverbs and umad (external libraries).
Resource manager is a collection of functions and structures to manage
RDMA resources such as QPs, CQs and MRs.
Reviewed-by: Dotan Barak <dotanb@mellanox.com>
Reviewed-by: Zhu Yanjun <yanjun.zhu@oracle.com>
Signed-off-by: Yuval Shaia <yuval.shaia@oracle.com>
Signed-off-by: Marcel Apfelbaum <marcel@redhat.com>
Definition of various structures and constants used in backend and
resource manager modules.
Reviewed-by: Dotan Barak <dotanb@mellanox.com>
Reviewed-by: Zhu Yanjun <yanjun.zhu@oracle.com>
Signed-off-by: Yuval Shaia <yuval.shaia@oracle.com>
Signed-off-by: Marcel Apfelbaum <marcel@redhat.com>
