* [0/5] C-SKY port
@ 2018-07-24 4:17 Sandra Loosemore
2018-07-24 4:20 ` [1/5] C-SKY port: Configury Sandra Loosemore
` (5 more replies)
0 siblings, 6 replies; 30+ messages in thread
From: Sandra Loosemore @ 2018-07-24 4:17 UTC (permalink / raw)
To: gcc-patches; +Cc: Han Mao, Yunhai Shang
This patch series is a new port for C-SKY ABI V2 processors. It's
based on the original work by C-SKY with cleanup, modernization, and
bug-fixing by Mentor Graphics. The supported targets are csky-elf and
csky-linux-* (both glibc and uClibc). At this time we're only
supporting the C and C++ languages, but we've built all the other
front ends except Ada where we're still trying to sort out build
environment issues.
C-SKY proposes Han Mao <han_mao@c-sky.com> and Yunhai Shang
<yunhai_shang@c-sky.com> as maintainers for this port. We expect that
C-SKY will also be providing a public link to the processor and ABI
documentation at some point. Meanwhile, here's brief overview
of the architecture from the compiler's point of view:
This is a 32-bit target with mixed 16- and 32-bit instructions and
support for both endiannesses. There are 5 different architecture
variants. The ck801 variant is substantially different from the
others in terms of code generation -- it has only a few 32-bit
instructions and a restricted register set, so it's almost analagous
to a Thumb-only ARM core. ck802 and ck803 have more 32-bit
instructions and 16 registers, while ck807 and ck810 are Linux-capable
with 32 registers.
The ABI is fairly standard; arguments are passed in r0-r3 with the
overflow on the stack, return values in r0-r1, etc. There is no
dedicated frame pointer register.
The ELF target is configured to build 16(!) multilibs and the Linux
targets 8, to account for hard/soft float variants and endiannesses.
We've been testing all of them with QEMU (user-mode for the Linux
targets, since we don't have a stable kernel yet) and the results are
quite good now.
I've split the patch set up into 5 pieces, with part 2 having the bulk
of the new code:
[1] Configury
[2] Backend implementation
[3] Documentation
[4] Testsuite
[5] libgcc
We also have a couple small patches to target-independent code to fix
bugs, which we'll post separately.
-Sandra
^ permalink raw reply [flat|nested] 30+ messages in thread
* [1/5] C-SKY port: Configury
2018-07-24 4:17 [0/5] C-SKY port Sandra Loosemore
@ 2018-07-24 4:20 ` Sandra Loosemore
2018-07-24 15:09 ` Jeff Law
2018-07-24 4:21 ` [2/5] C-SKY port: Backend implementation Sandra Loosemore
` (4 subsequent siblings)
5 siblings, 1 reply; 30+ messages in thread
From: Sandra Loosemore @ 2018-07-24 4:20 UTC (permalink / raw)
To: gcc-patches; +Cc: Han Mao, Yunhai Shang
[-- Attachment #1: Type: text/plain, Size: 540 bytes --]
2018-07-23 Jojo <jijie_rong@c-sky.com>
Huibin Wang <huibin_wang@c-sky.com>
Sandra Loosemore <sandra@codesourcery.com>
Chung-Lin Tang <cltang@codesourcery.com>
Andrew Jenner <andrew@codesourcery.com>
C-SKY port: Configury
gcc/
* config.gcc (csky-*-*): New.
* configure.ac: Add csky to targets for dwarf2 debug_line support.
* configure: Regenerated.
contrib/
* config-list.mk (LIST): Add csky-elf and csky-linux-gnu.
[-- Attachment #2: csky-gcc-1.patch --]
[-- Type: text/x-patch, Size: 3908 bytes --]
diff --git a/contrib/config-list.mk b/contrib/config-list.mk
index c3537d2..d9e48a9 100644
--- a/contrib/config-list.mk
+++ b/contrib/config-list.mk
@@ -40,6 +40,7 @@ LIST = aarch64-elf aarch64-linux-gnu aarch64-rtems \
arm-symbianelf avr-elf \
bfin-elf bfin-uclinux bfin-linux-uclibc bfin-rtems bfin-openbsd \
c6x-elf c6x-uclinux cr16-elf cris-elf cris-linux crisv32-elf crisv32-linux \
+ csky-elf csky-linux-gnu \
epiphany-elf epiphany-elfOPT-with-stack-offset=16 fido-elf \
fr30-elf frv-elf frv-linux ft32-elf h8300-elf hppa-linux-gnu \
hppa-linux-gnuOPT-enable-sjlj-exceptions=yes hppa64-linux-gnu \
diff --git a/gcc/config.gcc b/gcc/config.gcc
index 78e84c2..cd98836 100644
--- a/gcc/config.gcc
+++ b/gcc/config.gcc
@@ -1278,6 +1278,70 @@ crisv32-*-linux* | cris-*-linux*)
;;
esac
;;
+csky-*-*)
+ if test x${with_endian} != x; then
+ case ${with_endian} in
+ big|little) ;;
+ *)
+ echo "with_endian=${with_endian} not supported."
+ exit 1
+ ;;
+ esac
+ fi
+ if test x${with_float} != x; then
+ case ${with_float} in
+ soft | hard) ;;
+ *) echo
+ "Unknown floating point type used in --with-float=$with_float"
+ exit 1
+ ;;
+ esac
+ fi
+ tm_file="csky/csky.h"
+ md_file="csky/csky.md"
+ out_file="csky/csky.c"
+ tm_p_file="${tm_p_file} csky/csky-protos.h"
+ extra_options="${extra_options} csky/csky_tables.opt"
+
+ if test x${enable_tpf_debug} = xyes; then
+ tm_defines="${tm_defines} ENABLE_TPF_DEBUG"
+ fi
+
+ case ${target} in
+ csky-*-elf*)
+ tm_file="dbxelf.h elfos.h newlib-stdint.h ${tm_file} csky/csky-elf.h"
+ tmake_file="csky/t-csky csky/t-csky-elf"
+ default_use_cxa_atexit=no
+ ;;
+ csky-*-linux*)
+ tm_file="dbxelf.h elfos.h gnu-user.h linux.h glibc-stdint.h ${tm_file} csky/csky-linux-elf.h"
+ tmake_file="${tmake_file} csky/t-csky csky/t-csky-linux"
+
+ if test "x${enable_multilib}" = xyes ; then
+ tm_file="$tm_file ./sysroot-suffix.h"
+ tmake_file="${tmake_file} csky/t-sysroot-suffix"
+ fi
+
+ case ${target} in
+ csky-*-linux-gnu*)
+ tm_defines="$tm_defines DEFAULT_LIBC=LIBC_GLIBC"
+ ;;
+ csky-*-linux-uclibc*)
+ tm_defines="$tm_defines DEFAULT_LIBC=LIBC_UCLIBC"
+ default_use_cxa_atexit=no
+ ;;
+ *)
+ echo "Unknown target $target"
+ exit 1
+ ;;
+ esac
+ ;;
+ *)
+ echo "Unknown target $target"
+ exit 1
+ ;;
+ esac
+ ;;
epiphany-*-elf | epiphany-*-rtems*)
tm_file="${tm_file} dbxelf.h elfos.h"
tmake_file="${tmake_file} epiphany/t-epiphany"
@@ -3831,6 +3895,10 @@ case "${target}" in
fi
;;
+ csky-*-*)
+ supported_defaults="cpu endian float"
+ ;;
+
arm*-*-*)
supported_defaults="arch cpu float tune fpu abi mode tls"
for which in cpu tune arch; do
diff --git a/gcc/configure b/gcc/configure
index 80ac4a3..b7a8e36 100755
--- a/gcc/configure
+++ b/gcc/configure
@@ -27838,7 +27838,7 @@ esac
# ??? Once 2.11 is released, probably need to add first known working
# version to the per-target configury.
case "$cpu_type" in
- aarch64 | alpha | arc | arm | avr | bfin | cris | i386 | m32c | m68k \
+ aarch64 | alpha | arc | arm | avr | bfin | cris | csky | i386 | m32c | m68k \
| microblaze | mips | nios2 | pa | riscv | rs6000 | score | sparc | spu \
| tilegx | tilepro | visium | xstormy16 | xtensa)
insn="nop"
diff --git a/gcc/configure.ac b/gcc/configure.ac
index 4fc851c..65f9c92 100644
--- a/gcc/configure.ac
+++ b/gcc/configure.ac
@@ -4932,7 +4932,7 @@ esac
# ??? Once 2.11 is released, probably need to add first known working
# version to the per-target configury.
case "$cpu_type" in
- aarch64 | alpha | arc | arm | avr | bfin | cris | i386 | m32c | m68k \
+ aarch64 | alpha | arc | arm | avr | bfin | cris | csky | i386 | m32c | m68k \
| microblaze | mips | nios2 | pa | riscv | rs6000 | score | sparc | spu \
| tilegx | tilepro | visium | xstormy16 | xtensa)
insn="nop"
^ permalink raw reply [flat|nested] 30+ messages in thread
* [2/5] C-SKY port: Backend implementation
2018-07-24 4:17 [0/5] C-SKY port Sandra Loosemore
2018-07-24 4:20 ` [1/5] C-SKY port: Configury Sandra Loosemore
@ 2018-07-24 4:21 ` Sandra Loosemore
2018-07-24 15:45 ` Jeff Law
2018-07-24 4:23 ` [3/5] C-SKY port: Documentation Sandra Loosemore
` (3 subsequent siblings)
5 siblings, 1 reply; 30+ messages in thread
From: Sandra Loosemore @ 2018-07-24 4:21 UTC (permalink / raw)
To: gcc-patches; +Cc: Han Mao, Yunhai Shang
[-- Attachment #1: Type: text/plain, Size: 332 bytes --]
2018-07-23 Jojo <jijie_rong@c-sky.com>
Huibin Wang <huibin_wang@c-sky.com>
Sandra Loosemore <sandra@codesourcery.com>
Chung-Lin Tang <cltang@codesourcery.com>
C-SKY port: Backend implementation
gcc/
* config/csky/*: New.
* common/config/csky/*: New.
[-- Attachment #2: csky-gcc-2.patch.gz --]
[-- Type: application/gzip, Size: 91154 bytes --]
^ permalink raw reply [flat|nested] 30+ messages in thread
* [3/5] C-SKY port: Documentation
2018-07-24 4:17 [0/5] C-SKY port Sandra Loosemore
2018-07-24 4:20 ` [1/5] C-SKY port: Configury Sandra Loosemore
2018-07-24 4:21 ` [2/5] C-SKY port: Backend implementation Sandra Loosemore
@ 2018-07-24 4:23 ` Sandra Loosemore
2018-07-24 15:10 ` Jeff Law
2018-07-24 4:25 ` [4/5] C-SKY port: Testsuite Sandra Loosemore
` (2 subsequent siblings)
5 siblings, 1 reply; 30+ messages in thread
From: Sandra Loosemore @ 2018-07-24 4:23 UTC (permalink / raw)
To: gcc-patches; +Cc: Han Mao, Yunhai Shang
[-- Attachment #1: Type: text/plain, Size: 353 bytes --]
2018-07-23 Sandra Loosemore <sandra@codesourcery.com>
C-SKY port: Documentation
gcc/
* doc/extend.texi (C-SKY Function Attributes): New section.
* doc/invoke.texi (Option Summary): Add C-SKY options.
(C-SKY Options): New section.
* doc/md.texi (Machine Constraints): Document C-SKY constraints.
[-- Attachment #2: csky-gcc-3.patch --]
[-- Type: text/x-patch, Size: 10376 bytes --]
diff --git a/gcc/doc/extend.texi b/gcc/doc/extend.texi
index c7745c4..71c1d01 100644
--- a/gcc/doc/extend.texi
+++ b/gcc/doc/extend.texi
@@ -2324,6 +2324,7 @@ GCC plugins may provide their own attributes.
* AVR Function Attributes::
* Blackfin Function Attributes::
* CR16 Function Attributes::
+* C-SKY Function Attributes::
* Epiphany Function Attributes::
* H8/300 Function Attributes::
* IA-64 Function Attributes::
@@ -4145,6 +4146,38 @@ function entry and exit sequences suitable for use in an interrupt handler
when this attribute is present.
@end table
+@node C-SKY Function Attributes
+@subsection C-SKY Function Attributes
+
+These function attributes are supported by the C-SKY back end:
+
+@table @code
+@item interrupt
+@itemx isr
+@cindex @code{interrupt} function attribute, C-SKY
+@cindex @code{isr} function attribute, C-SKY
+Use these attributes to indicate that the specified function
+is an interrupt handler.
+The compiler generates function entry and exit sequences suitable for
+use in an interrupt handler when either of these attributes are present.
+
+Use of these options requires the @option{-mistack} command-line option
+to enable support for the necessary interrupt stack instructions. They
+are ignored with a warning otherwise. @xref{C-SKY Options}.
+
+@item naked
+@cindex @code{naked} function attribute, C-SKY
+This attribute allows the compiler to construct the
+requisite function declaration, while allowing the body of the
+function to be assembly code. The specified function will not have
+prologue/epilogue sequences generated by the compiler. Only basic
+@code{asm} statements can safely be included in naked functions
+(@pxref{Basic Asm}). While using extended @code{asm} or a mixture of
+basic @code{asm} and C code may appear to work, they cannot be
+depended upon to work reliably and are not supported.
+@end table
+
+
@node Epiphany Function Attributes
@subsection Epiphany Function Attributes
diff --git a/gcc/doc/invoke.texi b/gcc/doc/invoke.texi
index e0e59f6..75e147e 100644
--- a/gcc/doc/invoke.texi
+++ b/gcc/doc/invoke.texi
@@ -703,6 +703,16 @@ Objective-C and Objective-C++ Dialects}.
-msim -mint32 -mbit-ops
-mdata-model=@var{model}}
+@emph{C-SKY Options}
+@gccoptlist{-march=@var{arch} -mcpu=@var{cpu} @gol
+-mbig-endian -EB -mlittle-endian -EL @gol
+-mhard-float -msoft-float -mfpu=@var{fpu} -mdouble-float -mfdivdu @gol
+-melrw -mistack -mmp -mcp -mcache -msecurity -mtrust @gol
+-mdsp -medsp -mvdsp @gol
+-mdiv -msmart -mhigh-registers -manchor @gol
+-mpushpop -mmultiple-stld -mconstpool -mstack-size -mccrt @gol
+-mbranch-cost=@var{n} -mcse-cc -msched-prolog}
+
@emph{Darwin Options}
@gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
-arch_only -bind_at_load -bundle -bundle_loader @gol
@@ -14468,6 +14478,7 @@ platform.
* C6X Options::
* CRIS Options::
* CR16 Options::
+* C-SKY Options::
* Darwin Options::
* DEC Alpha Options::
* FR30 Options::
@@ -17581,6 +17592,205 @@ However, @samp{far} is not valid with @option{-mcr16c}, as the
CR16C architecture does not support the far data model.
@end table
+@node C-SKY Options
+@subsection C-SKY Options
+@cindex C-SKY Options
+
+GCC supports these options when compiling for C-SKY V2 processors.
+
+@table @gcctabopt
+
+@item -march=@var{arch}
+@opindex march=
+Specify the C-SKY target architecture. Valid values for @var{arch} are:
+@samp{ck801}, @samp{ck802}, @samp{ck803}, @samp{ck807}, and @samp{ck810}.
+The default is @samp{ck810}.
+
+@item -mcpu=@var{cpu}
+@opindex mcpu=
+Specify the C-SKY target processor. Valid values for @var{cpu} are:
+@samp{ck801}, @samp{ck801t},
+@samp{ck802}, @samp{ck802t}, @samp{ck802j},
+@samp{ck803}, @samp{ck803h}, @samp{ck803t}, @samp{ck803ht},
+@samp{ck803f}, @samp{ck803fh}, @samp{ck803e}, @samp{ck803eh},
+@samp{ck803et}, @samp{ck803eht}, @samp{ck803ef}, @samp{ck803efh},
+@samp{ck803ft}, @samp{ck803eft}, @samp{ck803efht}, @samp{ck803r1},
+@samp{ck803hr1}, @samp{ck803tr1}, @samp{ck803htr1}, @samp{ck803fr1},
+@samp{ck803fhr1}, @samp{ck803er1}, @samp{ck803ehr1}, @samp{ck803etr1},
+@samp{ck803ehtr1}, @samp{ck803efr1}, @samp{ck803efhr1}, @samp{ck803ftr1},
+@samp{ck803eftr1}, @samp{ck803efhtr1},
+@samp{ck803s}, @samp{ck803st}, @samp{ck803se}, @samp{ck803sf},
+@samp{ck803sef}, @samp{ck803seft},
+@samp{ck807e}, @samp{ck807ef}, @samp{ck807}, @samp{ck807f},
+@samp{ck810e}, @samp{ck810et}, @samp{ck810ef}, @samp{ck810eft},
+@samp{ck810}, @samp{ck810v}, @samp{ck810f}, @samp{ck810t}, @samp{ck810fv},
+@samp{ck810tv}, @samp{ck810ft}, and @samp{ck810ftv}.
+
+@item -mbig-endian
+@opindex mbig-endian
+@itemx -EB
+@opindex -EB
+@itemx -mlittle-endian
+@opindex mlittle-endian
+@itemx -EL
+@opindex -EL
+
+Select big- or little-endian code. The default is little-endian.
+
+@item -mhard-float
+@opindex mhard-float
+@itemx -msoft-float
+@opindex msoft-float
+
+Select hardware or software floating-point implementations.
+The default is soft float.
+
+@item -mdouble-float
+@itemx -mno-double-float
+@opindex mdouble-float
+When @option{-mhard-float} is in effect, enable generation of
+double-precision float instructions. This is the default except
+when compiling for CK803.
+
+@item -mfdivdu
+@itemx -mno-fdivdu
+@opindex mfdivdu
+When @option{-mhard-float} is in effect, enable generation of
+@code{frecipd}, @code{fsqrtd}, and @code{fdivd} instructions.
+This is the default except when compiling for CK803.
+
+@item -mfpu=@var{fpu}
+@opindex mfpu=
+Select the floating-point processor. This option can only be used with
+@option{-mhard-float}.
+Values for @var{fpu} are
+@samp{fpv2_sf} (equivalent to @samp{-mno-double-float -mno-fdivdu}),
+@samp{fpv2} (@samp{-mdouble-float -mno-divdu}), and
+@samp{fpv2_divd} (@samp{-mdouble-float -mdivdu}).
+
+@item -melrw
+@itemx -mno-elrw
+@opindex melrw
+Enable the extended @code{lrw} instruction. This option defaults to on
+for CK801 and off otherwise.
+
+@item -mistack
+@itemx -mno-istack
+@opindex mistack
+Enable interrupt stack instructions; the default is off.
+
+The @option{-mistack} option is required to handle the
+@code{interrupt} and @code{isr} function attributes
+(@pxref{C-SKY Function Attributes}).
+
+@item -mmp
+@opindex mmp
+Enable multiprocessor instructions; the default is off.
+
+@item -mcp
+@opindex mcp
+Enable coprocessor instructions; the default is off.
+
+@item -mcache
+@opindex mcache
+Enable coprocessor instructions; the default is off.
+
+@item -msecurity
+@opindex msecurity
+Enable C-SKY security instructions; the default is off.
+
+@item -mtrust
+@opindex mtrust
+Enable C-SKY trust instructions; the default is off.
+
+@item -mdsp
+@opindex mdsp
+@itemx -medsp
+@opindex medsp
+@itemx -mvdsp
+@opindex mvdsp
+Enable C-SKY DSP, Enhanced DSP, or Vector DSP instructions, respectively.
+All of these options default to off.
+
+@item -mdiv
+@itemx -mno-div
+@opindex mdiv
+Generate divide instructions. Default is off.
+
+@item -msmart
+@itemx -mno-smart
+@opindex msmart
+Generate code for Smart Mode, using only registers numbered 0-7 to allow
+use of 16-bit instructions. This option is ignored for CK801 where this
+is the required behavior, and it defaults to on for CK802.
+For other targets, the default is off.
+
+@item -mhigh-registers
+@itemx -mno-high-registers
+@opindex mhigh-registers
+Generate code using the high registers numbered 16-31. This option
+is not supported on CK801, CK802, or CK803, and is enabled by default
+for other processors.
+
+@item -manchor
+@itemx -mno-anchor
+@opindex manchor
+Generate code using global anchor symbol addresses.
+
+@item -mpushpop
+@itemx -mno-pushpop
+@opindex mpushpop
+Generate code using @code{push} and @code{pop} instructions. This option
+defaults to on.
+
+@item -mmultiple-stld
+@itemx -mstm
+@itemx -mno-multiple-stld
+@itemx -mno-stm
+@opindex mmultiple-stld
+Generate code using @code{stm} and @code{ldm} instructions. This option
+isn't supported on CK801 but is enabled by default on other processors.
+
+@item -mconstpool
+@itemx -mno-constpool
+@opindex mconstpool
+Create constant pools in the compiler instead of deferring it to the
+assembler. This option is the default and required for correct code
+generation on CK801 and CK802, and is optional on other processors.
+
+@item -mstack-size
+@item -mno-stack-size
+@opindex mstack-size
+Emit @code{.stack_size} directives for each function in the assembly
+output. This option defaults to off.
+
+@item -mccrt
+@itemx -mno-ccrt
+@opindex mccrt
+Generate code for the C-SKY compiler runtime instead of libgcc. This
+option defaults to off.
+
+@item -mbranch-cost=@var{n}
+@opindex mbranch-cost=
+Set the branch costs to roughly @code{n} instructions. The default is 1.
+
+@item -mcse-cc
+@itemx -mno-cse-cc
+@opindex mcse-cc
+Enable common subexpression elimination optimization on condition code
+expressions. This optimization pass is enabled by default at @option{-O1}
+and higher.
+
+@item -msched-prolog
+@itemx -mno-sched-prolog
+@opindex msched-prolog
+Permit scheduling of function prologue and epilogue sequences. Using
+this option can result in code that is not compliant with the C-SKY V2 ABI
+prologue requirements and that cannot be debugged or backtraced.
+It is disabled by default.
+
+@end table
+
@node Darwin Options
@subsection Darwin Options
@cindex Darwin options
diff --git a/gcc/doc/md.texi b/gcc/doc/md.texi
index 6d15d99..afb2ba8 100644
--- a/gcc/doc/md.texi
+++ b/gcc/doc/md.texi
@@ -2121,6 +2121,42 @@ Check for 64 bits wide constants for add/sub instructions
Floating point constant that is legal for store immediate
@end table
+@item C-SKY---@file{config/csky/constraints.md}
+@table @code
+
+@item a
+The mini registers r0 - r7.
+
+@item b
+The low registers r0 - r15.
+
+@item c
+C register.
+
+@item y
+HI and LO registers.
+
+@item l
+LO register.
+
+@item h
+HI register.
+
+@item v
+Vector registers.
+
+@item z
+Stack pointer register (SP).
+@end table
+
+@ifset INTERNALS
+The C-SKY back end supports a large set of additional constraints
+that are only useful for instruction selection or splitting rather
+than inline asm, such as constraints representing constant integer
+ranges accepted by particular instruction encodings.
+Refer to the source code for details.
+@end ifset
+
@item Epiphany---@file{config/epiphany/constraints.md}
@table @code
@item U16
^ permalink raw reply [flat|nested] 30+ messages in thread
* [4/5] C-SKY port: Testsuite
2018-07-24 4:17 [0/5] C-SKY port Sandra Loosemore
` (2 preceding siblings ...)
2018-07-24 4:23 ` [3/5] C-SKY port: Documentation Sandra Loosemore
@ 2018-07-24 4:25 ` Sandra Loosemore
2018-07-24 15:10 ` Jeff Law
2018-07-24 4:26 ` [5/5] C-SKY port: libgcc Sandra Loosemore
2018-07-24 15:23 ` [0/5] C-SKY port Sandra Loosemore
5 siblings, 1 reply; 30+ messages in thread
From: Sandra Loosemore @ 2018-07-24 4:25 UTC (permalink / raw)
To: gcc-patches; +Cc: Han Mao, Yunhai Shang
[-- Attachment #1: Type: text/plain, Size: 1309 bytes --]
2018-07-23 Sandra Loosemore <sandra@codesourcery.com>
Chung-Lin Tang <cltang@codesourcery.com>
C-SKY port: Testsuite
gcc/testsuite/
* g++.dg/Wno-frame-address.C: Adjust for C-SKY.
* g++.dg/torture/type-generic-1.C: Likewise.
* gcc.c-torture/compile/20000804-1.c: Likewise.
* gcc.c-torture/execute/20101011-1.c: Likewise.
* gcc.c-torture/execute/ieee/mul-subnormal-single-1.x: Likewise.
* gcc.dg/20020312-2.c: Likewise.
* gcc.dg/Wno-frame-address.c: Likewise.
* gcc.dg/c11-true_min-1.c: Likewise.
* gcc.dg/sibcall-10.c: Likewise.
* gcc.dg/sibcall-9.c: Likewise.
* gcc.dg/stack-usage-1.c: Likewise.
* gcc.dg/torture/float32-tg-3.c: Likewise.
* gcc.dg/torture/float32x-tg-3.c: Likewise.
* gcc.dg/torture/float64-tg-3.c: Likewise.
* gcc.dg/torture/float64x-tg-3.c: Likewise.
* gcc.dg/torture/type-generic-1.c: Likewise.
* gcc.target/csky/*: New.
* lib/target-supports.exp (check_profiling_available): Add
csky-*-elf.
(check_effective_target_hard_float): Handle C-SKY targets with
single-precision hard float only.
(check_effective_target_logical_op_short_circuit): Handle C-SKY.
[-- Attachment #2: csky-gcc-4.patch --]
[-- Type: text/x-patch, Size: 23800 bytes --]
diff --git a/gcc/testsuite/g++.dg/Wno-frame-address.C b/gcc/testsuite/g++.dg/Wno-frame-address.C
index a2df034..54a02fe 100644
--- a/gcc/testsuite/g++.dg/Wno-frame-address.C
+++ b/gcc/testsuite/g++.dg/Wno-frame-address.C
@@ -1,5 +1,5 @@
// { dg-do compile }
-// { dg-skip-if "Cannot access arbitrary stack frames." { arm*-*-* hppa*-*-* ia64-*-* } }
+// { dg-skip-if "Cannot access arbitrary stack frames." { arm*-*-* hppa*-*-* ia64-*-* csky*-*-* } }
// { dg-options "-Werror" }
// { dg-additional-options "-mbackchain" { target s390*-*-* } }
diff --git a/gcc/testsuite/g++.dg/torture/type-generic-1.C b/gcc/testsuite/g++.dg/torture/type-generic-1.C
index 4d82592..7708724 100644
--- a/gcc/testsuite/g++.dg/torture/type-generic-1.C
+++ b/gcc/testsuite/g++.dg/torture/type-generic-1.C
@@ -4,6 +4,7 @@
/* { dg-do run } */
/* { dg-add-options ieee } */
/* { dg-skip-if "No Inf/NaN support" { spu-*-* } } */
+/* { dg-skip-if "No subnormal support" { csky-*-* } { "-mhard-float" } } */
#include "../../gcc.dg/tg-tests.h"
diff --git a/gcc/testsuite/gcc.c-torture/compile/20000804-1.c b/gcc/testsuite/gcc.c-torture/compile/20000804-1.c
index 5c6b731..35464c2 100644
--- a/gcc/testsuite/gcc.c-torture/compile/20000804-1.c
+++ b/gcc/testsuite/gcc.c-torture/compile/20000804-1.c
@@ -4,6 +4,7 @@
/* { dg-skip-if "" { { i?86-*-* x86_64-*-* } && { ia32 && { ! nonpic } } } } */
/* { dg-skip-if "No 64-bit registers" { m32c-*-* } } */
/* { dg-skip-if "Not enough 64-bit registers" { pdp11-*-* } { "-O0" } { "" } } */
+/* { dg-xfail-if "Inconsistent constraint on asm" { csky-*-* } { "-O0" } { "" } } */
/* { dg-xfail-if "" { h8300-*-* } } */
/* Copyright (C) 2000, 2003 Free Software Foundation */
diff --git a/gcc/testsuite/gcc.c-torture/execute/20101011-1.c b/gcc/testsuite/gcc.c-torture/execute/20101011-1.c
index dda49a5..f95d900 100644
--- a/gcc/testsuite/gcc.c-torture/execute/20101011-1.c
+++ b/gcc/testsuite/gcc.c-torture/execute/20101011-1.c
@@ -93,6 +93,10 @@ __aeabi_idiv0 (int return_value)
#elif defined (__nvptx__)
/* There isn't even a signal function. */
# define DO_TEST 0
+#elif defined (__csky__)
+ /* This presently doesn't raise SIGFPE even on csky-linux-gnu, much
+ less bare metal. See the implementation of __divsi3 in libgcc. */
+# define DO_TEST 0
#else
# define DO_TEST 1
#endif
diff --git a/gcc/testsuite/gcc.c-torture/execute/ieee/mul-subnormal-single-1.x b/gcc/testsuite/gcc.c-torture/execute/ieee/mul-subnormal-single-1.x
index 16df951..ee40863 100644
--- a/gcc/testsuite/gcc.c-torture/execute/ieee/mul-subnormal-single-1.x
+++ b/gcc/testsuite/gcc.c-torture/execute/ieee/mul-subnormal-single-1.x
@@ -1,3 +1,8 @@
+if {[istarget "csky-*-*"] && [check_effective_target_hard_float]} {
+ # The C-SKY hardware FPU only supports flush-to-zero mode.
+ set torture_execute_xfail "csky-*-*"
+ return 1
+}
if [istarget "epiphany-*-*"] {
# The Epiphany single-precision floating point format does not
# support subnormals.
diff --git a/gcc/testsuite/gcc.dg/20020312-2.c b/gcc/testsuite/gcc.dg/20020312-2.c
index f5929e0..f8be3ce 100644
--- a/gcc/testsuite/gcc.dg/20020312-2.c
+++ b/gcc/testsuite/gcc.dg/20020312-2.c
@@ -111,6 +111,11 @@ extern void abort (void);
/* No pic register. */
#elif defined (__nvptx__)
/* No pic register. */
+#elif defined (__csky__)
+/* Pic register is r28, but some cores only have r0-r15. */
+# if defined (__CK807__) || defined (__CK810__)
+# define PIC_REG "r28"
+# endif
#else
# error "Modify the test for your target."
#endif
diff --git a/gcc/testsuite/gcc.dg/Wno-frame-address.c b/gcc/testsuite/gcc.dg/Wno-frame-address.c
index e6dfe52..9fe4d07 100644
--- a/gcc/testsuite/gcc.dg/Wno-frame-address.c
+++ b/gcc/testsuite/gcc.dg/Wno-frame-address.c
@@ -1,5 +1,5 @@
/* { dg-do compile } */
-/* { dg-skip-if "Cannot access arbitrary stack frames" { arm*-*-* avr-*-* hppa*-*-* ia64-*-* visium-*-* } } */
+/* { dg-skip-if "Cannot access arbitrary stack frames" { arm*-*-* avr-*-* hppa*-*-* ia64-*-* visium-*-* csky-*-* } } */
/* { dg-options "-Werror" } */
/* { dg-additional-options "-mbackchain" { target { s390*-*-* } } } */
diff --git a/gcc/testsuite/gcc.dg/c11-true_min-1.c b/gcc/testsuite/gcc.dg/c11-true_min-1.c
index f666c72..1fbf514 100644
--- a/gcc/testsuite/gcc.dg/c11-true_min-1.c
+++ b/gcc/testsuite/gcc.dg/c11-true_min-1.c
@@ -1,6 +1,7 @@
/* { dg-do run } */
/* { dg-options "-std=c11" } */
/* { dg-xfail-run-if "PR58757 -mieee is required to compare denormals" { alpha*-*-* } } */
+/* { dg-skip-if "No subnormal support" { csky-*-* } { "-mhard-float" } } */
/* Test that the smallest positive value is not 0. This needs to be true
even when denormals are not supported, so we do not pass any flag
diff --git a/gcc/testsuite/gcc.dg/sibcall-10.c b/gcc/testsuite/gcc.dg/sibcall-10.c
index d89909a..54cc604 100644
--- a/gcc/testsuite/gcc.dg/sibcall-10.c
+++ b/gcc/testsuite/gcc.dg/sibcall-10.c
@@ -5,7 +5,7 @@
Copyright (C) 2002 Free Software Foundation Inc.
Contributed by Hans-Peter Nilsson <hp@bitrange.com> */
-/* { dg-do run { xfail { { cris-*-* crisv32-*-* h8300-*-* hppa*64*-*-* m32r-*-* mcore-*-* mn10300-*-* msp430*-*-* nds32*-*-* xstormy16-*-* v850*-*-* vax-*-* xtensa*-*-* } || { arm*-*-* && { ! arm32 } } } } } */
+/* { dg-do run { xfail { { cris-*-* crisv32-*-* csky-*-* h8300-*-* hppa*64*-*-* m32r-*-* mcore-*-* mn10300-*-* msp430*-*-* nds32*-*-* xstormy16-*-* v850*-*-* vax-*-* xtensa*-*-* } || { arm*-*-* && { ! arm32 } } } } } */
/* -mlongcall disables sibcall patterns. */
/* { dg-skip-if "" { powerpc*-*-* } { "-mlongcall" } { "" } } */
/* -msave-restore disables sibcall patterns. */
diff --git a/gcc/testsuite/gcc.dg/sibcall-9.c b/gcc/testsuite/gcc.dg/sibcall-9.c
index 8e30952..fc3bd9d 100644
--- a/gcc/testsuite/gcc.dg/sibcall-9.c
+++ b/gcc/testsuite/gcc.dg/sibcall-9.c
@@ -5,7 +5,7 @@
Copyright (C) 2002 Free Software Foundation Inc.
Contributed by Hans-Peter Nilsson <hp@bitrange.com> */
-/* { dg-do run { xfail { { cris-*-* crisv32-*-* h8300-*-* hppa*64*-*-* m32r-*-* mcore-*-* mn10300-*-* msp430*-*-* nds32*-*-* nvptx-*-* xstormy16-*-* v850*-*-* vax-*-* xtensa*-*-* } || { arm*-*-* && { ! arm32 } } } } } */
+/* { dg-do run { xfail { { cris-*-* crisv32-*-* csky-*-* h8300-*-* hppa*64*-*-* m32r-*-* mcore-*-* mn10300-*-* msp430*-*-* nds32*-*-* nvptx-*-* xstormy16-*-* v850*-*-* vax-*-* xtensa*-*-* } || { arm*-*-* && { ! arm32 } } } } } */
/* -mlongcall disables sibcall patterns. */
/* { dg-skip-if "" { powerpc*-*-* } { "-mlongcall" } { "" } } */
/* -msave-restore disables sibcall patterns. */
diff --git a/gcc/testsuite/gcc.dg/stack-usage-1.c b/gcc/testsuite/gcc.dg/stack-usage-1.c
index 038bd4e..e644014 100644
--- a/gcc/testsuite/gcc.dg/stack-usage-1.c
+++ b/gcc/testsuite/gcc.dg/stack-usage-1.c
@@ -101,6 +101,8 @@
#define SIZE 252
#elif defined (__M32R__)
#define SIZE 252
+#elif defined (__csky__)
+# define SIZE 252
#else
# define SIZE 256
#endif
diff --git a/gcc/testsuite/gcc.dg/torture/float32-tg-3.c b/gcc/testsuite/gcc.dg/torture/float32-tg-3.c
index b07c07a..e478a23 100644
--- a/gcc/testsuite/gcc.dg/torture/float32-tg-3.c
+++ b/gcc/testsuite/gcc.dg/torture/float32-tg-3.c
@@ -4,6 +4,7 @@
/* { dg-add-options float32 } */
/* { dg-add-options ieee } */
/* { dg-require-effective-target float32_runtime } */
+/* { dg-skip-if "No subnormal support" { csky-*-* } { "-mhard-float" } } */
#define WIDTH 32
#define EXT 0
diff --git a/gcc/testsuite/gcc.dg/torture/float32x-tg-3.c b/gcc/testsuite/gcc.dg/torture/float32x-tg-3.c
index 9f9f982..6f55779 100644
--- a/gcc/testsuite/gcc.dg/torture/float32x-tg-3.c
+++ b/gcc/testsuite/gcc.dg/torture/float32x-tg-3.c
@@ -4,6 +4,7 @@
/* { dg-add-options float32x } */
/* { dg-add-options ieee } */
/* { dg-require-effective-target float32x_runtime } */
+/* { dg-skip-if "No subnormal support" { csky-*-* } { "-mhard-float" } } */
#define WIDTH 32
#define EXT 1
diff --git a/gcc/testsuite/gcc.dg/torture/float64-tg-3.c b/gcc/testsuite/gcc.dg/torture/float64-tg-3.c
index a83e781..a1fead7 100644
--- a/gcc/testsuite/gcc.dg/torture/float64-tg-3.c
+++ b/gcc/testsuite/gcc.dg/torture/float64-tg-3.c
@@ -4,6 +4,7 @@
/* { dg-add-options float64 } */
/* { dg-add-options ieee } */
/* { dg-require-effective-target float64_runtime } */
+/* { dg-skip-if "No subnormal support" { csky-*-* } { "-mhard-float" } } */
#define WIDTH 64
#define EXT 0
diff --git a/gcc/testsuite/gcc.dg/torture/float64x-tg-3.c b/gcc/testsuite/gcc.dg/torture/float64x-tg-3.c
index 195c4fd..3cdd933 100644
--- a/gcc/testsuite/gcc.dg/torture/float64x-tg-3.c
+++ b/gcc/testsuite/gcc.dg/torture/float64x-tg-3.c
@@ -4,6 +4,7 @@
/* { dg-add-options float64x } */
/* { dg-add-options ieee } */
/* { dg-require-effective-target float64x_runtime } */
+/* { dg-skip-if "No subnormal support" { csky-*-* } { "-mhard-float" } } */
#define WIDTH 64
#define EXT 1
diff --git a/gcc/testsuite/gcc.dg/torture/type-generic-1.c b/gcc/testsuite/gcc.dg/torture/type-generic-1.c
index ef32b78..a5fa8e8 100644
--- a/gcc/testsuite/gcc.dg/torture/type-generic-1.c
+++ b/gcc/testsuite/gcc.dg/torture/type-generic-1.c
@@ -3,6 +3,7 @@
/* { dg-do run } */
/* { dg-skip-if "No Inf/NaN support" { spu-*-* } } */
+/* { dg-skip-if "No subnormal support" { csky-*-* } { "-mhard-float" } } */
/* { dg-options "-DUNSAFE" { target tic6x*-*-* visium-*-* nvptx-*-* } } */
/* { dg-add-options ieee } */
diff --git a/gcc/testsuite/gcc.target/csky/and1.c b/gcc/testsuite/gcc.target/csky/and1.c
new file mode 100644
index 0000000..14ce11a
--- /dev/null
+++ b/gcc/testsuite/gcc.target/csky/and1.c
@@ -0,0 +1,12 @@
+/* { dg-do compile } */
+/* { dg-csky-options "-O1" } */
+
+/* Test special code generation patterns for bit operators. */
+
+int and1 (int x)
+{
+ return x & 0xfff7ffff;
+}
+
+/* { dg-final { scan-assembler "bclri" } } */
+
diff --git a/gcc/testsuite/gcc.target/csky/and2.c b/gcc/testsuite/gcc.target/csky/and2.c
new file mode 100644
index 0000000..c661199
--- /dev/null
+++ b/gcc/testsuite/gcc.target/csky/and2.c
@@ -0,0 +1,12 @@
+/* { dg-do compile } */
+/* { dg-csky-options "-O1" } */
+
+/* Test special code generation patterns for bit operators. */
+
+int and2 (int x)
+{
+ return x & 0xfff00000;
+}
+
+/* { dg-final { scan-assembler "lsri" } } */
+/* { dg-final { scan-assembler "lsli" } } */
diff --git a/gcc/testsuite/gcc.target/csky/and3a.c b/gcc/testsuite/gcc.target/csky/and3a.c
new file mode 100644
index 0000000..3d706f6
--- /dev/null
+++ b/gcc/testsuite/gcc.target/csky/and3a.c
@@ -0,0 +1,12 @@
+/* { dg-do compile } */
+/* { dg-csky-options "-mcpu=ck801 -O1" } */
+
+/* Test special code generation patterns for bit operators. */
+
+int and3 (int x)
+{
+ return x & 0x000fffff;
+}
+
+/* { dg-final { scan-assembler "lsli" } } */
+/* { dg-final { scan-assembler "lsri" } } */
diff --git a/gcc/testsuite/gcc.target/csky/and3b.c b/gcc/testsuite/gcc.target/csky/and3b.c
new file mode 100644
index 0000000..127207d
--- /dev/null
+++ b/gcc/testsuite/gcc.target/csky/and3b.c
@@ -0,0 +1,11 @@
+/* { dg-do compile } */
+/* { dg-csky-options "-mcpu=ck810f -O1" } */
+
+/* Test special code generation patterns for bit operators. */
+
+int and3 (int x)
+{
+ return x & 0x000fffff;
+}
+
+/* { dg-final { scan-assembler "zext" } } */
diff --git a/gcc/testsuite/gcc.target/csky/ck801-branch.c b/gcc/testsuite/gcc.target/csky/ck801-branch.c
new file mode 100644
index 0000000..95e6962
--- /dev/null
+++ b/gcc/testsuite/gcc.target/csky/ck801-branch.c
@@ -0,0 +1,40 @@
+/* { dg-do compile } */
+/* { dg-csky-options "-mcpu=ck801 -O1 -fno-reorder-blocks" } */
+
+/* Test branch generation on CK801, which cannot rely on assembler
+ branch relaxation because long branches clobber lr. */
+
+#define nop8 asm ("nop\nnop\nnop\nnop\nnop\nnop\nnop\nnop")
+#define nop64 nop8; nop8; nop8; nop8; nop8; nop8; nop8; nop8
+#define nop512 nop64; nop64; nop64; nop64; nop64; nop64; nop64; nop64
+#define nop4k nop512; nop512; nop512; nop512; nop512; nop512; nop512; nop512
+#define nop32k nop4k; nop4k; nop4k; nop4k; nop4k; nop4k; nop4k; nop4k
+
+extern void g (int);
+int f (int x, int y, int z)
+{
+ if (x == 0) // cmpnei; jbt
+ {
+ nop64;
+ x = y;
+ }
+ if (y == 0) // cmpnei; jbf; jbr
+ {
+ nop512;
+ y = z;
+ }
+ if (z == 0) // cmpnei; jbf; bsr
+ {
+ nop32k;
+ z = x;
+ }
+ return x + y + z;
+}
+
+/* { dg-final { scan-assembler "push.*lr" } } */
+/* { dg-final { scan-assembler "pop.*lr" } } */
+/* { dg-final { scan-assembler-times "cmpnei" 3 } } */
+/* { dg-final { scan-assembler-times "jbt" 1 } } */
+/* { dg-final { scan-assembler-times "jbf" 2 } } */
+/* { dg-final { scan-assembler-times "jbr" 1 } } */
+/* { dg-final { scan-assembler-times "bsr" 1 } } */
diff --git a/gcc/testsuite/gcc.target/csky/constpool-1.c b/gcc/testsuite/gcc.target/csky/constpool-1.c
new file mode 100644
index 0000000..5c7cfdc
--- /dev/null
+++ b/gcc/testsuite/gcc.target/csky/constpool-1.c
@@ -0,0 +1,16 @@
+/* { dg-do compile } */
+/* { dg-csky-options "-mcpu=ck801 -O1" } */
+
+/* Make sure that constant pools are emitted by the compiler for ck801.
+ If this is deferred to the assembler, the compiler will compute
+ incorrect branch offsets. */
+
+void f (unsigned int *u, long long int *l, float *f, double *d)
+{
+ *u = 0xdeadbeef;
+ *l = 0xcafef00dc0ffeeULL;
+ *f = 3.14159F;
+ *d = 2.718281828459;
+}
+
+/* { dg-final { scan-assembler-times "\\.long" 6 } } */
diff --git a/gcc/testsuite/gcc.target/csky/constpool-2.c b/gcc/testsuite/gcc.target/csky/constpool-2.c
new file mode 100644
index 0000000..d654420
--- /dev/null
+++ b/gcc/testsuite/gcc.target/csky/constpool-2.c
@@ -0,0 +1,15 @@
+/* { dg-do compile } */
+/* { dg-csky-options "-mcpu=ck810f -O1 -mconstpool" } */
+
+/* Make sure that constant pools are emitted by the compiler when
+ -mconstpool is provided. */
+
+void f (unsigned int *u, long long int *l, float *f, double *d)
+{
+ *u = 0xdeadbeef;
+ *l = 0xcafef00dc0ffeeULL;
+ *f = 3.14159F;
+ *d = 2.718281828459;
+}
+
+/* { dg-final { scan-assembler-times "\\.long" 6 } } */
diff --git a/gcc/testsuite/gcc.target/csky/constpool-3.c b/gcc/testsuite/gcc.target/csky/constpool-3.c
new file mode 100644
index 0000000..e3a6e09
--- /dev/null
+++ b/gcc/testsuite/gcc.target/csky/constpool-3.c
@@ -0,0 +1,15 @@
+/* { dg-do compile } */
+/* { dg-csky-options "-mcpu=ck810f -O1 -mno-constpool" } */
+
+/* Make sure that constant pools are not emitted by the compiler when
+ -mno-constpool is provided. */
+
+void f (unsigned int *u, long long int *l, float *f, double *d)
+{
+ *u = 0xdeadbeef;
+ *l = 0xcafef00dc0ffeeULL;
+ *f = 3.14159F;
+ *d = 2.718281828459;
+}
+
+/* { dg-final { scan-assembler-not "\\.long" } } */
diff --git a/gcc/testsuite/gcc.target/csky/csky.exp b/gcc/testsuite/gcc.target/csky/csky.exp
new file mode 100644
index 0000000..9569490
--- /dev/null
+++ b/gcc/testsuite/gcc.target/csky/csky.exp
@@ -0,0 +1,79 @@
+# GCC testsuite for C-SKY targets.
+# Copyright (C) 2012-2018 Free Software Foundation, Inc.
+# Contributed by C-SKY Microsystems and Mentor Graphics.
+#
+# This program is free software; you can redistribute it and/or modify
+# it under the terms of the GNU General Public License as published by
+# the Free Software Foundation; either version 3 of the License, or
+# (at your option) any later version.
+#
+# This program is distributed in the hope that it will be useful,
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+# GNU General Public License for more details.
+#
+# You should have received a copy of the GNU General Public License
+# along with GCC; see the file COPYING3. If not see
+# <http://www.gnu.org/licenses/>.
+
+# Exit immediately if this isn't a C-SKY target.
+if ![istarget csky*-*-*] then {
+ return
+}
+
+# Load support procs.
+load_lib gcc-dg.exp
+
+# Like dg-options, but treats certain C-SKY-specific options specially:
+#
+# -mcpu=*
+# Select the target cpu. Skip the test if the multilib flags force
+# a different cpu.
+proc dg-csky-options {args} {
+ upvar dg-extra-tool-flags extra_tool_flags
+ upvar dg-do-what do_what
+
+ set multilib_cpu ""
+ set multilib_hf ""
+ set cpu ""
+
+ foreach flag [target_info multilib_flags] {
+ regexp "^-mcpu=(.*)" $flag dummy multilib_cpu
+ regexp "^-mhard-float" $flag multilib_hf
+ }
+
+ set flags [lindex $args 1]
+
+ foreach flag $flags {
+ regexp "^-mcpu=(.*)" $flag dummy cpu
+ }
+
+ if {$cpu == ""} then {
+ set extra_tool_flags $flags
+ } elseif {$multilib_cpu == "" || $multilib_cpu == $cpu} then {
+ if { ($cpu == "ck801" || $cpu == "ck802")
+ && $multilib_hf != "" } then {
+ set do_what [list [lindex $do_what 0] "N" "P"]
+ } else {
+ set extra_tool_flags $flags
+ }
+ } else {
+ set do_what [list [lindex $do_what 0] "N" "P"]
+ }
+}
+
+# If a testcase doesn't have special options, use these.
+global DEFAULT_CFLAGS
+if ![info exists DEFAULT_CFLAGS] then {
+ set DEFAULT_CFLAGS " -ansi -pedantic-errors"
+}
+
+# Initialize `dg'.
+dg-init
+
+# Main loop.
+dg-runtest [lsort [glob -nocomplain $srcdir/$subdir/*.\[cCS\]]] \
+ "" $DEFAULT_CFLAGS
+
+# All done.
+dg-finish
diff --git a/gcc/testsuite/gcc.target/csky/fnargs-1.c b/gcc/testsuite/gcc.target/csky/fnargs-1.c
new file mode 100644
index 0000000..5cc85b6
--- /dev/null
+++ b/gcc/testsuite/gcc.target/csky/fnargs-1.c
@@ -0,0 +1,38 @@
+/* { dg-do run } */
+
+/* Check that a structure argument passed partially in registers and
+ partially on the stack works. */
+
+#include <stdlib.h>
+#include <string.h>
+
+struct s {
+ unsigned int i;
+ double d;
+ char s[16];
+};
+
+/* Note specifically that, since there are 4 argument registers, the
+ value of ss.d is split between the last argument register and the
+ stack. */
+void
+f (struct s *sp, int j, struct s ss, int k)
+{
+ if (sp->i != ss.i
+ || sp->d != ss.d
+ || strcmp (sp->s, ss.s))
+ abort ();
+ if (j != -k)
+ abort ();
+}
+
+int
+main (void)
+{
+ struct s ss;
+ ss.i = 0xdeadbeef;
+ ss.d = 2.71828;
+ strcpy (ss.s, "shazam!");
+ f (&ss, 42, ss, -42);
+ return 0;
+}
diff --git a/gcc/testsuite/gcc.target/csky/fnargs-2.c b/gcc/testsuite/gcc.target/csky/fnargs-2.c
new file mode 100644
index 0000000..d4e1b71
--- /dev/null
+++ b/gcc/testsuite/gcc.target/csky/fnargs-2.c
@@ -0,0 +1,52 @@
+/* { dg-do run } */
+
+/* Check that varargs passed partially in registers and
+ partially on the stack works. */
+
+#include <stdlib.h>
+#include <string.h>
+#include <stdarg.h>
+
+struct s {
+ unsigned int i;
+ double d;
+ char s[16];
+};
+
+/* Note specifically that, as there are 4 argument registers,
+ the value of ss.d is split between the last argument register
+ and the stack. */
+void
+f (struct s *sp, ...)
+{
+ int j, k;
+ unsigned int i;
+ double d;
+ char *s;
+ va_list ap;
+ va_start (ap, sp);
+ j = va_arg (ap, int);
+ i = va_arg (ap, unsigned int);
+ d = va_arg (ap, double);
+ s = va_arg (ap, char *);
+ k = va_arg (ap, int);
+ va_end (ap);
+
+ if (sp->i != i
+ || sp->d != d
+ || strcmp (sp->s, s))
+ abort ();
+ if (j != -k)
+ abort ();
+}
+
+int
+main (void)
+{
+ struct s ss;
+ ss.i = 0xdeadbeef;
+ ss.d = 2.71828;
+ strcpy (ss.s, "shazam!");
+ f (&ss, 42, ss.i, ss.d, ss.s, -42);
+ return 0;
+}
diff --git a/gcc/testsuite/gcc.target/csky/fnargs-3.c b/gcc/testsuite/gcc.target/csky/fnargs-3.c
new file mode 100644
index 0000000..8cf3e5c
--- /dev/null
+++ b/gcc/testsuite/gcc.target/csky/fnargs-3.c
@@ -0,0 +1,31 @@
+/* { dg-do run } */
+
+/* Check that sub-word sized elements of structures passed in in
+ registers are handled correctly with respect to the current endianness. */
+
+#include <stdlib.h>
+#include <string.h>
+
+struct s {
+ short h;
+ char s[8];
+};
+
+void
+f (struct s *sp, struct s ss)
+{
+ if (sp->h != ss.h
+ || strcmp (sp->s, ss.s))
+ abort ();
+}
+
+int
+main (void)
+{
+ struct s ss;
+ ss.h = 42;
+ strcpy (ss.s, "shazam!");
+ f (&ss, ss);
+ return 0;
+}
+
diff --git a/gcc/testsuite/gcc.target/csky/land1.c b/gcc/testsuite/gcc.target/csky/land1.c
new file mode 100644
index 0000000..e5ca51c
--- /dev/null
+++ b/gcc/testsuite/gcc.target/csky/land1.c
@@ -0,0 +1,11 @@
+/* { dg-do compile } */
+/* { dg-csky-options "-O1" } */
+
+/* Test special code generation patterns for bit operators. */
+
+long long int land1 (long long int x)
+{
+ return x & 0xffffffff00000000LL;
+}
+
+/* { dg-final { scan-assembler "movi.*, 0" } } */
diff --git a/gcc/testsuite/gcc.target/csky/land2.c b/gcc/testsuite/gcc.target/csky/land2.c
new file mode 100644
index 0000000..b45e7b6
--- /dev/null
+++ b/gcc/testsuite/gcc.target/csky/land2.c
@@ -0,0 +1,11 @@
+/* { dg-do compile } */
+/* { dg-csky-options "-O1" } */
+
+/* Test special code generation patterns for bit operators. */
+
+long long int land2 (long long int x)
+{
+ return x & 0x00000000ffffffffLL;
+}
+
+/* { dg-final { scan-assembler "movi.*, 0" } } */
diff --git a/gcc/testsuite/gcc.target/csky/naked.c b/gcc/testsuite/gcc.target/csky/naked.c
new file mode 100644
index 0000000..f81984c
--- /dev/null
+++ b/gcc/testsuite/gcc.target/csky/naked.c
@@ -0,0 +1,16 @@
+/* { dg-do compile } */
+/* { dg-final { scan-assembler-not "push" } } */
+/* { dg-final { scan-assembler-not "pop" } } */
+
+/* Check that there is no prologue/epilogue code emitted for a function
+ with the naked attribute. Without the attribute, this function would
+ push/pop lr. */
+
+extern void g (int);
+
+int __attribute__((naked))
+f (int x)
+{
+ g (x);
+ return 42;
+}
diff --git a/gcc/testsuite/gcc.target/csky/or1.c b/gcc/testsuite/gcc.target/csky/or1.c
new file mode 100644
index 0000000..24918bc
--- /dev/null
+++ b/gcc/testsuite/gcc.target/csky/or1.c
@@ -0,0 +1,11 @@
+/* { dg-do compile } */
+/* { dg-csky-options "-O1" } */
+
+/* Test special code generation patterns for bit operators. */
+
+int or1 (int x)
+{
+ return x | 0x00100000;
+}
+
+/* { dg-final { scan-assembler "bseti" } } */
diff --git a/gcc/testsuite/lib/target-supports.exp b/gcc/testsuite/lib/target-supports.exp
index 99613fd..7e81936 100644
--- a/gcc/testsuite/lib/target-supports.exp
+++ b/gcc/testsuite/lib/target-supports.exp
@@ -669,6 +669,7 @@ proc check_profiling_available { test_what } {
|| [istarget bfin-*-*]
|| [istarget cris-*-*]
|| [istarget crisv32-*-*]
+ || [istarget csky-*-elf]
|| [istarget fido-*-elf]
|| [istarget h8300-*-*]
|| [istarget lm32-*-*]
@@ -1228,6 +1229,16 @@ proc check_effective_target_hard_float { } {
# }]
}
+ # The generic test doesn't work for C-SKY because some cores have
+ # hard float for single precision only.
+ if { [istarget csky*-*-*] } {
+ return [check_no_compiler_messages hard_float assembly {
+ #if defined __csky_soft_float__
+ #error __csky_soft_float__
+ #endif
+ }]
+ }
+
# The generic test equates hard_float with "no call for adding doubles".
return [check_no_messages_and_pattern hard_float "!\\(call" rtl-expand {
double a (double b, double c) { return b + c; }
@@ -8804,6 +8815,7 @@ proc check_effective_target_logical_op_short_circuit {} {
|| [istarget arc*-*-*]
|| [istarget avr*-*-*]
|| [istarget crisv32-*-*] || [istarget cris-*-*]
+ || [istarget csky*-*-*]
|| [istarget mmix-*-*]
|| [istarget s390*-*-*]
|| [istarget powerpc*-*-*]
@@ -8822,6 +8834,7 @@ proc check_effective_target_logical_op_short_circuit {} {
proc check_effective_target_branch_cost {} {
if { [ istarget arm*-*-*]
|| [istarget avr*-*-*]
+ || [istarget csky*-*-*]
|| [istarget epiphany*-*-*]
|| [istarget frv*-*-*]
|| [istarget i?86-*-*] || [istarget x86_64-*-*]
^ permalink raw reply [flat|nested] 30+ messages in thread
* [5/5] C-SKY port: libgcc
2018-07-24 4:17 [0/5] C-SKY port Sandra Loosemore
` (3 preceding siblings ...)
2018-07-24 4:25 ` [4/5] C-SKY port: Testsuite Sandra Loosemore
@ 2018-07-24 4:26 ` Sandra Loosemore
2018-07-24 15:12 ` Jeff Law
2018-07-24 18:10 ` Segher Boessenkool
2018-07-24 15:23 ` [0/5] C-SKY port Sandra Loosemore
5 siblings, 2 replies; 30+ messages in thread
From: Sandra Loosemore @ 2018-07-24 4:26 UTC (permalink / raw)
To: gcc-patches; +Cc: Han Mao, Yunhai Shang
[-- Attachment #1: Type: text/plain, Size: 324 bytes --]
2018-07-23 Jojo <jijie_rong@c-sky.com>
Huibin Wang <huibin_wang@c-sky.com>
Sandra Loosemore <sandra@codesourcery.com>
Chung-Lin Tang <cltang@codesourcery.com>
C-SKY port: libgcc
libgcc/
* config.host: Add C-SKY support.
* config/csky/*: New.
[-- Attachment #2: csky-gcc-5.patch --]
[-- Type: text/x-patch, Size: 36068 bytes --]
diff --git a/libgcc/config.host b/libgcc/config.host
index 18cabaf..b2ee0c9 100644
--- a/libgcc/config.host
+++ b/libgcc/config.host
@@ -94,6 +94,9 @@ am33_2.0-*-linux*)
arc*-*-*)
cpu_type=arc
;;
+csky*-*-*)
+ cpu_type=csky
+ ;;
arm*-*-*)
cpu_type=arm
;;
@@ -507,6 +510,15 @@ cris-*-elf)
cris-*-linux* | crisv32-*-linux*)
tmake_file="$tmake_file cris/t-cris t-softfp-sfdf t-softfp cris/t-linux"
;;
+csky-*-elf*)
+ tmake_file="csky/t-csky t-fdpbit"
+ extra_parts="$extra_parts crti.o crtn.o"
+ ;;
+csky-*-linux*)
+ tmake_file="$tmake_file csky/t-csky t-slibgcc-libgcc t-fdpbit csky/t-linux-csky"
+ extra_parts="$extra_parts crti.o crtn.o"
+ md_unwind_header=csky/linux-unwind.h
+ ;;
epiphany-*-elf* | epiphany-*-rtems*)
tmake_file="$tmake_file epiphany/t-epiphany t-fdpbit epiphany/t-custom-eqsf"
extra_parts="$extra_parts crti.o crtint.o crtrunc.o crtm1reg-r43.o crtm1reg-r63.o crtn.o"
diff --git a/libgcc/config/csky/crti.S b/libgcc/config/csky/crti.S
new file mode 100644
index 0000000..3e4beb9
--- /dev/null
+++ b/libgcc/config/csky/crti.S
@@ -0,0 +1,140 @@
+# Define _init and _fini entry points for C-SKY.
+# Copyright (C) 2018 Free Software Foundation, Inc.
+# Contributed by C-SKY Microsystems and Mentor Graphics.
+#
+# This file is free software; you can redistribute it and/or modify it
+# under the terms of the GNU General Public License as published by the
+# Free Software Foundation; either version 3, or (at your option) any
+# later version.
+#
+# This file is distributed in the hope that it will be useful, but
+# WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+# General Public License for more details.
+#
+# Under Section 7 of GPL version 3, you are granted additional
+# permissions described in the GCC Runtime Library Exception, version
+# 3.1, as published by the Free Software Foundation.
+#
+# You should have received a copy of the GNU General Public License and
+# a copy of the GCC Runtime Library Exception along with this program;
+# see the files COPYING3 and COPYING.RUNTIME respectively. If not, see
+# <http://www.gnu.org/licenses/>.
+
+
+# This file just makes a stack frame for the contents of the .fini and
+# .init sections. Users may put any desired instructions in those
+# sections.
+
+ .file "crti.S"
+
+/* We use more complicated versions of this code with GLIBC. */
+#if defined(__gnu_linux__)
+
+#ifndef PREINIT_FUNCTION
+# define PREINIT_FUNCTION __gmon_start__
+#endif
+
+#ifndef PREINIT_FUNCTION_WEAK
+# define PREINIT_FUNCTION_WEAK 1
+#endif
+
+#if PREINIT_FUNCTION_WEAK
+ .global PREINIT_FUNCTION
+ .weak PREINIT_FUNCTION
+ .align 4
+ .type call_weak_fn, %function
+call_weak_fn:
+ // push lr
+ subi sp, 4
+ stw lr, (sp)
+#ifdef __PIC__
+ lrw a2, PREINIT_FUNCTION@GOT
+ addu a2, gb
+ ldw a2, (a2)
+#else
+ lrw a2, PREINIT_FUNCTION
+#endif
+ cmpnei a2, 0
+ bf 1f
+ jsr a2
+1:
+ // pop lr
+ ldw lr, (sp)
+ addi sp, 4
+ rts
+
+ .align 4
+#else
+ .hidden PREINIT_FUNCTION
+#endif /* PREINIT_FUNCTION_WEAK */
+
+ .section .init,"ax",@progbits
+ .align 4
+ .globl _init
+ .type _init, @function
+_init:
+ subi sp, 8
+ stw lr, (sp, 0)
+#ifdef __PIC__
+ // stw gb, (sp, 4)
+ bsr .Lgetpc
+.Lgetpc:
+ lrw gb, .Lgetpc@GOTPC
+ add gb, lr
+#endif
+#if PREINIT_FUNCTION_WEAK
+#ifdef __PIC__
+ lrw a2, call_weak_fn@GOTOFF
+ add a2, gb
+ jsr a2
+#else
+ jsri call_weak_fn
+#endif
+#else /* !PREINIT_FUNCTION_WEAK */
+#ifdef __PIC__
+ lrw a2, PREINIT_FUNCTION@PLT
+ addu a2, gb
+ ldw a2, (a2)
+ jsr a2
+#else
+ jsri PREINIT_FUNCTION
+#endif
+#endif /* PREINIT_FUNCTION_WEAK */
+
+ br 2f
+ .literals
+ .align 4
+2:
+ .section .fini,"ax",@progbits
+ .align 4
+ .globl _fini
+ .type _fini, @function
+_fini:
+ subi sp,8
+ stw lr, (sp, 0)
+ br 2f
+ .literals
+ .align 4
+2:
+
+/* These are the non-GLIBC versions. */
+#else /* !defined(__gnu_linux__) */
+ .section ".init"
+ .global _init
+ .type _init,@function
+ .align 2
+_init:
+ subi sp, 16
+ st.w lr, (sp, 12)
+ mov r0, r0
+
+ .section ".fini"
+ .global _fini
+ .type _fini,@function
+ .align 2
+_fini:
+ subi sp, 16
+ st.w lr, (sp, 12)
+ mov r0, r0
+#endif /* defined(__gnu_linux__) */
diff --git a/libgcc/config/csky/crtn.S b/libgcc/config/csky/crtn.S
new file mode 100644
index 0000000..8bef996
--- /dev/null
+++ b/libgcc/config/csky/crtn.S
@@ -0,0 +1,55 @@
+# Terminate C-SKY .init and .fini sections.
+# Copyright (C) 2018 Free Software Foundation, Inc.
+# Contributed by C-SKY Microsystems and Mentor Graphics.
+#
+# This file is free software; you can redistribute it and/or modify it
+# under the terms of the GNU General Public License as published by the
+# Free Software Foundation; either version 3, or (at your option) any
+# later version.
+#
+# This file is distributed in the hope that it will be useful, but
+# WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+# General Public License for more details.
+#
+# Under Section 7 of GPL version 3, you are granted additional
+# permissions described in the GCC Runtime Library Exception, version
+# 3.1, as published by the Free Software Foundation.
+#
+# You should have received a copy of the GNU General Public License and
+# a copy of the GCC Runtime Library Exception along with this program;
+# see the files COPYING3 and COPYING.RUNTIME respectively. If not, see
+# <http://www.gnu.org/licenses/>.
+
+
+# This file just makes sure that the .fini and .init sections do in
+# fact return. Users may put any desired instructions in those sections.
+# This file is the last thing linked into any executable.
+
+ .file "crtn.S"
+
+# Is this the GLIBC version?
+#if defined(__gnu_linux__)
+ .section .init,"ax",@progbits
+ ldw lr, (sp, 0)
+ addi sp, 8
+ rts
+
+ .section .fini,"ax",@progbits
+ ldw lr, (sp, 0)
+ addi sp, 8
+ rts
+
+#else /* !defined(__gnu_linux__) */
+ .section ".init"
+ ldw lr, (sp, 12)
+ addi sp, 16
+ jmp lr
+
+ .section ".fini"
+ ldw lr, (sp, 12)
+ addi sp, 16
+ jmp lr
+
+# Th-th-th-that is all folks!
+#endif /* defined(__gnu_linux__) */
diff --git a/libgcc/config/csky/lib1funcs.S b/libgcc/config/csky/lib1funcs.S
new file mode 100644
index 0000000..a0a3c73
--- /dev/null
+++ b/libgcc/config/csky/lib1funcs.S
@@ -0,0 +1,675 @@
+/* libgcc routines for C-SKY.
+ Copyright (C) 2018 Free Software Foundation, Inc.
+ Contributed by C-SKY Microsystems and Mentor Graphics.
+
+ This file is part of GCC.
+
+ GCC is free software; you can redistribute it and/or modify it
+ under the terms of the GNU General Public License as published by the
+ Free Software Foundation; either version 3, or (at your option) any
+ later version.
+
+ This file is distributed in the hope that it will be useful, but
+ WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ General Public License for more details.
+
+ Under Section 7 of GPL version 3, you are granted additional
+ permissions described in the GCC Runtime Library Exception, version
+ 3.1, as published by the Free Software Foundation.
+
+ You should have received a copy of the GNU General Public License and
+ a copy of the GCC Runtime Library Exception along with this program;
+ see the files COPYING3 and COPYING.RUNTIME respectively. If not, see
+ <http://www.gnu.org/licenses/>. */
+
+
+/* Use the right prefix for global labels. */
+#define CONCAT1(a, b) CONCAT2(a, b)
+#define CONCAT2(a, b) a ## b
+#define SYM(x) CONCAT1 (__, x)
+
+#ifndef __CSKYBE__
+#define xl r0
+#define xh r1
+#define yl r2
+#define yh r3
+#else
+#define xh r0
+#define xl r1
+#define yh r2
+#define yl r3
+#endif
+
+
+#ifdef __ELF__
+#define TYPE(x) .type SYM (x),@function
+#define SIZE(x) .size SYM (x), . - SYM (x)
+#else
+#define TYPE(x)
+#define SIZE(x)
+#endif
+
+.macro FUNC_START name
+ .text
+ .align 2
+ .globl SYM (\name)
+ TYPE (\name)
+SYM (\name):
+.endm
+
+.macro FUNC_END name
+ SIZE (\name)
+.endm
+
+
+/* Emulate FF1 ("fast find 1") instruction on ck801.
+ Result goes in rx, clobbering ry. */
+#if defined(__CK801__)
+.macro FF1_M rx, ry
+ movi \rx, 32
+10:
+ cmphsi \ry, 1
+ bf 11f
+ subi \rx, \rx, 1
+ lsri \ry, \ry, 1
+ br 10b
+11:
+.endm
+#else
+.macro FF1_M rx, ry
+ ff1 \rx, \ry
+.endm
+#endif
+
+/* Likewise emulate lslc instruction ("logical left shift to C") on CK801. */
+#if defined(__CK801__)
+.macro LSLC_M rx
+ cmpne \rx, \rx
+ addc \rx, \rx
+.endm
+#else
+.macro LSLC_M rx
+ lslc \rx
+.endm
+#endif
+
+/* Emulate the abs instruction. */
+#if defined(__CK802__)
+.macro ABS_M rx
+ btsti \rx, 31
+ bf 10f
+ not \rx
+ addi \rx, 1
+10:
+.endm
+#elif defined(__CK801__)
+.macro ABS_M rx
+ cmplti \rx, 1
+ bf 10f
+ not \rx
+ addi \rx, 1
+10:
+.endm
+#else
+.macro ABS_M rx
+ abs \rx
+.endm
+#endif
+
+/* Emulate the ld.hs ("load signed halfword and extend") instruction
+ on ck801 and ck802. */
+#if defined(__CK801__)
+.macro LDBS_M rx, ry
+ ld.b \rx, (\ry, 0x0)
+ sextb \rx, \rx
+.endm
+#else
+.macro LDBS_M rx, ry
+ ld.bs \rx, (\ry, 0x0)
+.endm
+#endif
+
+#if defined(__CK801__)
+.macro LDHS_M rx, ry
+ ld.h \rx, (\ry, 0x0)
+ sexth \rx, \rx
+.endm
+#else
+.macro LDHS_M rx, ry
+ ld.hs \rx, (\ry, 0x0)
+.endm
+#endif
+
+
+/* Signed and unsigned div/mod/rem functions. */
+
+#ifdef L_udivsi3
+FUNC_START udiv32
+FUNC_START udivsi3
+ cmpnei a1, 0 // look for 0 divisor
+ bt 9f
+ trap 3 // divide by 0
+9:
+ // control iterations, skip across high order 0 bits in dividend
+ cmpnei a0, 0
+ bt 8f
+ jmp lr // 0 dividend quick return
+8:
+ push l0
+ movi a2, 1 // a2 is quotient (1 for a sentinel)
+ mov a3, a0
+ FF1_M l0, a3 // figure distance to skip
+ lsl a2, l0 // move the sentinel along (with 0's behind)
+ lsl a0, l0 // and the low 32 bits of numerator
+
+ // FIXME: Is this correct?
+ mov a3, a1 // looking at divisor
+ FF1_M l0, a3 // I can move 32-l0 more bits to left.
+ addi l0, 1 // ok, one short of that...
+ mov a3, a0
+ lsr a3, l0 // bits that came from low order...
+ not l0 // l0 == "32-n" == LEFT distance
+ addi l0, 33 // this is (32-n)
+ lsl a2,l0 // fixes the high 32 (quotient)
+ lsl a0,l0
+ cmpnei a2,0
+ bf 4f // the sentinel went away...
+
+ // run the remaining bits
+1:
+ LSLC_M a0 // 1 bit left shift of a3-a0
+ addc a3, a3
+ cmphs a3, a1 // upper 32 of dividend >= divisor?
+ bf 2f
+ subu a3, a1 // if yes, subtract divisor
+2:
+ addc a2, a2 // shift by 1 and count subtracts
+ bf 1b // if sentinel falls out of quotient, stop
+
+4:
+ mov a0, a2 // return quotient
+ mov a1, a3 // and piggyback the remainder
+ pop l0
+FUNC_END udiv32
+FUNC_END udivsi3
+#endif
+
+#ifdef L_umodsi3
+FUNC_START urem32
+FUNC_START umodsi3
+ cmpnei a1, 0 // look for 0 divisor
+ bt 9f
+ trap 3 // divide by 0
+9:
+ // control iterations, skip across high order 0 bits in dividend
+ cmpnei a0, 0
+ bt 8f
+ jmp lr // 0 dividend quick return
+8:
+ mov a2, a0
+ FF1_M a3, a2 // figure distance to skip
+ movi a2, 1 // a2 is quotient (1 for a sentinel)
+ lsl a2, a3 // move the sentinel along (with 0's behind)
+ lsl a0, a3 // and the low 32 bits of numerator
+ movi a3, 0
+
+1:
+ LSLC_M a0 // 1 bit left shift of a3-a0
+ addc a3, a3
+ cmphs a3, a1 // upper 32 of dividend >= divisor?
+ bf 2f
+ subu a3, a1 // if yes, subtract divisor
+2:
+ addc a2, a2 // shift by 1 and count subtracts
+ bf 1b // if sentinel falls out of quotient, stop
+
+4:
+ mov a0, a3 // and piggyback the remainder
+ jmp lr
+FUNC_END urem32
+FUNC_END umodsi3
+#endif
+
+
+#ifdef L_divsi3
+FUNC_START div32
+FUNC_START divsi3
+ cmpnei a1, 0 // look for 0 divisor
+ bt 9f
+ trap 3 // divide by 0
+9:
+ // control iterations, skip across high order 0 bits in dividend
+ cmpnei a0, 0
+ bt 8f
+ jmp lr // 0 dividend quick return
+8:
+ push l0, l1
+ mov l1, a0
+ xor l1, a1 // calc sign of quotient
+ ABS_M a0
+ ABS_M a1
+ movi a2, 1 // a2 is quotient (1 for a sentinel)
+ mov a3, a0
+ FF1_M l0, a3 // figure distance to skip
+ lsl a2, l0 // move the sentinel along (with 0's behind)
+ lsl a0, l0 // and the low 32 bits of numerator
+
+ // FIXME: is this correct?
+ mov a3, a1 // looking at divisor
+ FF1_M l0, a3 // I can move 32-l0 more bits to left.
+ addi l0, 1 // ok, one short of that...
+ mov a3, a0
+ lsr a3, l0 // bits that came from low order...
+ not l0 // l0 == "32-n" == LEFT distance
+ addi l0, 33 // this is (32-n)
+ lsl a2,l0 // fixes the high 32 (quotient)
+ lsl a0,l0
+ cmpnei a2,0
+ bf 4f // the sentinel went away...
+
+ // run the remaining bits
+1:
+ LSLC_M a0 // 1 bit left shift of a3-a0
+ addc a3, a3
+ cmphs a3, a1 // upper 32 of dividend >= divisor?
+ bf 2f
+ subu a3, a1 // if yes, subtract divisor
+2:
+ addc a2, a2 // shift by 1 and count subtracts
+ bf 1b // if sentinel falls out of quotient, stop
+
+4:
+ mov a0, a2 // return quotient
+ mov a1, a3 // and piggyback the remainder
+ LSLC_M l1 // after adjusting for sign
+ bf 3f
+ not a0
+ addi a0, 1
+ not a1
+ addi a1, 1
+3:
+ pop l0, l1
+FUNC_END div32
+FUNC_END divsi3
+#endif
+
+#ifdef L_modsi3
+FUNC_START rem32
+FUNC_START modsi3
+ push l0
+ cmpnei a1, 0 // look for 0 divisor
+ bt 9f
+ trap 3 // divide by 0
+9:
+ // control iterations, skip across high order 0 bits in dividend
+ cmpnei a0, 0
+ bt 8f
+ pop l0 // 0 dividend quick return
+8:
+ mov l0, a0
+ ABS_M a0
+ ABS_M a1
+ mov a2, a0
+ FF1_M a3, a2 // figure distance to skip
+ movi a2, 1 // a2 is quotient (1 for a sentinel)
+ lsl a2, a3 // move the sentinel along (with 0's behind)
+ lsl a0, a3 // and the low 32 bits of numerator
+ movi a3, 0
+
+ // run the remaining bits
+1:
+ LSLC_M a0 // 1 bit left shift of a3-a0
+ addc a3, a3
+ cmphs a3, a1 // upper 32 of dividend >= divisor?
+ bf 2f
+ subu a3, a1 // if yes, subtract divisor
+2:
+ addc a2, a2 // shift by 1 and count subtracts
+ bf 1b // if sentinel falls out of quotient, stop
+
+4:
+ mov a0, a3 // and piggyback the remainder
+ LSLC_M l0 // after adjusting for sign
+ bf 3f
+ not a0
+ addi a0, 1
+3:
+ pop l0
+FUNC_END rem32
+FUNC_END modsi3
+#endif
+
+/* Unordered comparisons for single and double float. */
+
+#ifdef L_unordsf2
+FUNC_START unordsf2
+#if defined(__CK801__)
+ subi sp, 4
+ st.w r4, (sp, 0x0)
+ lsli r2, r0, 1
+ lsli r3, r1, 1
+ asri r4, r2, 24
+ not r4
+ cmpnei r4, 0
+ bt 1f
+ lsli r4, r0, 9
+ cmpnei r4, 0
+ bt 3f
+1:
+ asri r4, r3, 24
+ not r4
+ cmpnei r4, 0
+ bt 2f
+ lsli r4, r1, 9
+ cmpnei r4, 0
+ bt 3f
+2:
+ ld.w r4, (sp, 0x0)
+ addi sp, 4
+ movi r0, 0
+ rts
+3:
+ ld.w r4, (sp, 0x0)
+ addi sp, 4
+ movi r0, 1
+ rts
+#elif defined(__CK802__)
+ lsli r2, r0, 1
+ lsli r3, r1, 1
+ asri r2, r2, 24
+ not r13, r2
+ cmpnei r13, 0
+ bt 1f
+ lsli r13, r0, 9
+ cmpnei r13, 0
+ bt 3f
+1:
+ asri r3, r3, 24
+ not r13, r3
+ cmpnei r13, 0
+ bt 2f
+ lsli r13, r1, 9
+ cmpnei r13, 0
+ bt 3f
+2:
+ movi r0, 0
+ rts
+3:
+ movi r0, 1
+ rts
+#else
+ lsli r2, r0, 1
+ lsli r3, r1, 1
+ asri r2, r2, 24
+ not r13, r2
+ bnez r13, 1f
+ lsli r13, r0, 9
+ bnez r13, 3f
+1:
+ asri r3, r3, 24
+ not r13, r3
+ bnez r13, 2f
+ lsli r13, r1, 9
+ bnez r13, 3f
+2:
+ movi r0, 0
+ rts
+3:
+ movi r0, 1
+ rts
+#endif
+FUNC_END unordsf2
+#endif
+
+#ifdef L_unorddf2
+FUNC_START unorddf2
+#if defined(__CK801__)
+ subi sp, 8
+ st.w r4, (sp, 0x0)
+ st.w r5, (sp, 0x4)
+ lsli r4, xh, 1
+ asri r4, r4, 21
+ not r4
+ cmpnei r4, 0
+ bt 1f
+ mov r4, xl
+ lsli r5, xh, 12
+ or r4, r5
+ cmpnei r4, 0
+ bt 3f
+1:
+ lsli r4, yh, 1
+ asri r4, r4, 21
+ not r4
+ cmpnei r4, 0
+ bt 2f
+ mov r4,yl
+ lsli r5, yh, 12
+ or r4, r5
+ cmpnei r4, 0
+ bt 3f
+2:
+ ld.w r4, (sp, 0x0)
+ ld.w r5, (sp, 0x4)
+ addi sp, 8
+ movi r0, 0
+ rts
+3:
+ ld.w r4, (sp, 0x0)
+ ld.w r5, (sp, 0x4)
+ addi sp, 8
+ movi r0, 1
+ rts
+#elif defined(__CK802__)
+ lsli r13, xh, 1
+ asri r13, r13, 21
+ not r13
+ cmpnei r13, 0
+ bt 1f
+ lsli xh, xh, 12
+ or r13, xl, xh
+ cmpnei r13, 0
+ bt 3f
+1:
+ lsli r13, yh, 1
+ asri r13, r13, 21
+ not r13
+ cmpnei r13, 0
+ bt 2f
+ lsli yh, yh, 12
+ or r13, yl, yh
+ cmpnei r13, 0
+ bt 3f
+2:
+ movi r0, 0
+ rts
+3:
+ movi r0, 1
+ rts
+#else
+ lsli r13, xh, 1
+ asri r13, r13, 21
+ not r13
+ bnez r13, 1f
+ lsli xh, xh, 12
+ or r13, xl, xh
+ bnez r13, 3f
+1:
+ lsli r13, yh, 1
+ asri r13, r13, 21
+ not r13
+ bnez r13, 2f
+ lsli yh, yh, 12
+ or r13, yl, yh
+ bnez r13, 3f
+2:
+ movi r0, 0
+ rts
+3:
+ movi r0, 1
+ rts
+#endif
+FUNC_END unorddf2
+#endif
+
+/* When optimizing for size on ck801 and ck802, GCC emits calls to the
+ following helper functions when expanding casesi, instead of emitting
+ the table lookup and jump inline. Note that in these functions the
+ jump is handled by tweaking the value of lr before rts. */
+#ifdef L_csky_case_sqi
+FUNC_START _gnu_csky_case_sqi
+ subi sp, 4
+ st.w a1, (sp, 0x0)
+ mov a1, lr
+ add a1, a1, a0
+ LDBS_M a1, a1
+ lsli a1, a1, 1
+ add lr, lr, a1
+ ld.w a1, (sp, 0x0)
+ addi sp, 4
+ rts
+FUNC_END _gnu_csky_case_sqi
+#endif
+
+#ifdef L_csky_case_uqi
+FUNC_START _gnu_csky_case_uqi
+ subi sp, 4
+ st.w a1, (sp, 0x0)
+ mov a1, lr
+ add a1, a1, a0
+ ld.b a1, (a1, 0x0)
+ lsli a1, a1, 1
+ add lr, lr, a1
+ ld.w a1, (sp, 0x0)
+ addi sp, 4
+ rts
+FUNC_END _gnu_csky_case_uqi
+#endif
+
+#ifdef L_csky_case_shi
+FUNC_START _gnu_csky_case_shi
+ subi sp, 8
+ st.w a0, (sp, 0x4)
+ st.w a1, (sp, 0x0)
+ mov a1, lr
+ lsli a0, a0, 1
+ add a1, a1, a0
+ LDHS_M a1, a1
+ lsli a1, a1, 1
+ add lr, lr, a1
+ ld.w a0, (sp, 0x4)
+ ld.w a1, (sp, 0x0)
+ addi sp, 8
+ rts
+FUNC_END _gnu_csky_case_shi
+#endif
+
+#ifdef L_csky_case_uhi
+FUNC_START _gnu_csky_case_uhi
+ subi sp, 8
+ st.w a0, (sp, 0x4)
+ st.w a1, (sp, 0x0)
+ mov a1, lr
+ lsli a0, a0, 1
+ add a1, a1, a0
+ ld.h a1, (a1, 0x0)
+ lsli a1, a1, 1
+ add lr, lr, a1
+ ld.w a0, (sp, 0x4)
+ ld.w a1, (sp, 0x0)
+ addi sp, 8
+ rts
+FUNC_END _gnu_csky_case_uhi
+#endif
+
+#ifdef L_csky_case_si
+FUNC_START _gnu_csky_case_si
+ subi sp, 8
+ st.w a0, (sp, 0x4)
+ st.w a1, (sp, 0x0)
+ mov a1, lr
+ addi a1, a1, 2 // Align to word.
+ bclri a1, a1, 1
+ mov lr, a1
+ lsli a0, a0, 2
+ add a1, a1, a0
+ ld.w a0, (a1, 0x0)
+ add lr, lr, a0
+ ld.w a0, (sp, 0x4)
+ ld.w a1, (sp, 0x0)
+ addi sp, 8
+ rts
+FUNC_END _gnu_csky_case_si
+#endif
+
+/* GCC expects that {__eq,__ne,__gt,__ge,__le,__lt}{df2,sf2}
+ will behave as __cmpdf2. So, we stub the implementations to
+ jump on to __cmpdf2 and __cmpsf2.
+
+ All of these short-circuit the return path so that __cmp{sd}f2
+ will go directly back to the caller. */
+
+.macro COMPARE_DF_JUMP name
+ .import SYM (cmpdf2)
+FUNC_START \name
+ jmpi SYM (cmpdf2)
+FUNC_END \name
+.endm
+
+#ifdef L_eqdf2
+COMPARE_DF_JUMP eqdf2
+#endif /* L_eqdf2 */
+
+#ifdef L_nedf2
+COMPARE_DF_JUMP nedf2
+#endif /* L_nedf2 */
+
+#ifdef L_gtdf2
+COMPARE_DF_JUMP gtdf2
+#endif /* L_gtdf2 */
+
+#ifdef L_gedf2
+COMPARE_DF_JUMP gedf2
+#endif /* L_gedf2 */
+
+#ifdef L_ltdf2
+COMPARE_DF_JUMP ltdf2
+#endif /* L_ltdf2 */
+
+#ifdef L_ledf2
+COMPARE_DF_JUMP ledf2
+#endif /* L_ledf2 */
+
+/* Single-precision floating point stubs. */
+
+.macro COMPARE_SF_JUMP name
+ .import SYM (cmpsf2)
+FUNC_START \name
+ jmpi SYM (cmpsf2)
+FUNC_END \name
+.endm
+
+#ifdef L_eqsf2
+COMPARE_SF_JUMP eqsf2
+#endif /* L_eqsf2 */
+
+#ifdef L_nesf2
+COMPARE_SF_JUMP nesf2
+#endif /* L_nesf2 */
+
+#ifdef L_gtsf2
+COMPARE_SF_JUMP gtsf2
+#endif /* L_gtsf2 */
+
+#ifdef L_gesf2
+COMPARE_SF_JUMP __gesf2
+#endif /* L_gesf2 */
+
+#ifdef L_ltsf2
+COMPARE_SF_JUMP __ltsf2
+#endif /* L_ltsf2 */
+
+#ifdef L_lesf2
+COMPARE_SF_JUMP lesf2
+#endif /* L_lesf2 */
diff --git a/libgcc/config/csky/linux-atomic.c b/libgcc/config/csky/linux-atomic.c
new file mode 100644
index 0000000..03cf2c0
--- /dev/null
+++ b/libgcc/config/csky/linux-atomic.c
@@ -0,0 +1,299 @@
+/* Linux-specific atomic operations for C-SKY.
+ Copyright (C) 2018 Free Software Foundation, Inc.
+ Contributed by C-SKY Microsystems and Mentor Graphics.
+
+ This file is part of GCC.
+
+ GCC is free software; you can redistribute it and/or modify it under
+ the terms of the GNU General Public License as published by the Free
+ Software Foundation; either version 3, or (at your option) any later
+ version.
+
+ GCC is distributed in the hope that it will be useful, but WITHOUT ANY
+ WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
+ for more details.
+
+ Under Section 7 of GPL version 3, you are granted additional
+ permissions described in the GCC Runtime Library Exception, version
+ 3.1, as published by the Free Software Foundation.
+
+ You should have received a copy of the GNU General Public License and
+ a copy of the GCC Runtime Library Exception along with this program;
+ see the files COPYING3 and COPYING.RUNTIME respectively. If not, see
+ <http://www.gnu.org/licenses/>. */
+
+/* Kernel helper for compare-and-exchange. */
+inline int
+__kernel_cmpxchg (int oldval, int newval, volatile int *ptr)
+{
+ register int _a0 asm ("a0") = oldval;
+ register int _a1 asm ("a1") = newval;
+ register volatile int *_a2 asm ("a2") = ptr;
+ __asm__ __volatile__ ("trap 2\n" \
+ :"+r" (_a0) :"r" (_a1) , "r" (_a2) \
+ : "a3", "memory"); \
+ return _a0;
+}
+
+
+/* Kernel helper for memory barrier. */
+inline void __kernel_dmb (void)
+{
+ asm ("sync":::"memory");
+}
+
+/* Note: we implement byte, short and int versions of atomic operations using
+ the above kernel helpers, but there is no support for "long long" (64-bit)
+ operations as yet. */
+
+#define HIDDEN __attribute__ ((visibility ("hidden")))
+
+#ifdef __CSKYLE__
+#define INVERT_MASK_1 0
+#define INVERT_MASK_2 0
+#else
+#define INVERT_MASK_1 24
+#define INVERT_MASK_2 16
+#endif
+
+#define MASK_1 0xffu
+#define MASK_2 0xffffu
+
+#define FETCH_AND_OP_WORD(OP, PFX_OP, INF_OP) \
+ int HIDDEN \
+ __sync_fetch_and_##OP##_4 (int *ptr, int val) \
+ { \
+ int failure, tmp; \
+ \
+ do \
+ { \
+ tmp = *ptr; \
+ failure = __kernel_cmpxchg (tmp, PFX_OP (tmp INF_OP val), ptr); \
+ } \
+ while (failure != 0); \
+ \
+ return tmp; \
+ }
+
+FETCH_AND_OP_WORD (add, , +)
+FETCH_AND_OP_WORD (sub, , -)
+FETCH_AND_OP_WORD (or, , |)
+FETCH_AND_OP_WORD (and, , &)
+FETCH_AND_OP_WORD (xor, , ^)
+FETCH_AND_OP_WORD (nand, ~, &)
+
+#define NAME_oldval(OP, WIDTH) __sync_fetch_and_##OP##_##WIDTH
+#define NAME_newval(OP, WIDTH) __sync_##OP##_and_fetch_##WIDTH
+
+/* Implement both __sync_<op>_and_fetch and __sync_fetch_and_<op> for
+ subword-sized quantities. */
+
+#define SUBWORD_SYNC_OP(OP, PFX_OP, INF_OP, TYPE, WIDTH, RETURN) \
+ TYPE HIDDEN \
+ NAME##_##RETURN (OP, WIDTH) (TYPE *ptr, TYPE val) \
+ { \
+ int *wordptr = (int *) ((unsigned int) ptr & ~3); \
+ unsigned int mask, shift, oldval, newval; \
+ int failure; \
+ \
+ shift = (((unsigned int) ptr & 3) << 3) ^ INVERT_MASK_##WIDTH; \
+ mask = MASK_##WIDTH << shift; \
+ \
+ do \
+ { \
+ oldval = *wordptr; \
+ newval = ((PFX_OP (((oldval & mask) >> shift) \
+ INF_OP (unsigned int) val)) << shift) & mask; \
+ newval |= oldval & ~mask; \
+ failure = __kernel_cmpxchg (oldval, newval, wordptr); \
+ } \
+ while (failure != 0); \
+ \
+ return (RETURN & mask) >> shift; \
+ }
+
+SUBWORD_SYNC_OP (add, , +, unsigned short, 2, oldval)
+SUBWORD_SYNC_OP (sub, , -, unsigned short, 2, oldval)
+SUBWORD_SYNC_OP (or, , |, unsigned short, 2, oldval)
+SUBWORD_SYNC_OP (and, , &, unsigned short, 2, oldval)
+SUBWORD_SYNC_OP (xor, , ^, unsigned short, 2, oldval)
+SUBWORD_SYNC_OP (nand, ~, &, unsigned short, 2, oldval)
+
+SUBWORD_SYNC_OP (add, , +, unsigned char, 1, oldval)
+SUBWORD_SYNC_OP (sub, , -, unsigned char, 1, oldval)
+SUBWORD_SYNC_OP (or, , |, unsigned char, 1, oldval)
+SUBWORD_SYNC_OP (and, , &, unsigned char, 1, oldval)
+SUBWORD_SYNC_OP (xor, , ^, unsigned char, 1, oldval)
+SUBWORD_SYNC_OP (nand, ~, &, unsigned char, 1, oldval)
+
+#define OP_AND_FETCH_WORD(OP, PFX_OP, INF_OP) \
+ int HIDDEN \
+ __sync_##OP##_and_fetch_4 (int *ptr, int val) \
+ { \
+ int tmp, failure; \
+ \
+ do \
+ { \
+ tmp = *ptr; \
+ failure = __kernel_cmpxchg (tmp, PFX_OP tmp INF_OP val, ptr); \
+ } \
+ while (failure != 0); \
+ \
+ return PFX_OP tmp INF_OP val; \
+ }
+
+OP_AND_FETCH_WORD (add, , +)
+OP_AND_FETCH_WORD (sub, , -)
+OP_AND_FETCH_WORD (or, , |)
+OP_AND_FETCH_WORD (and, , &)
+OP_AND_FETCH_WORD (xor, , ^)
+OP_AND_FETCH_WORD (nand, ~, &)
+
+SUBWORD_SYNC_OP (add, , +, unsigned short, 2, newval)
+SUBWORD_SYNC_OP (sub, , -, unsigned short, 2, newval)
+SUBWORD_SYNC_OP (or, , |, unsigned short, 2, newval)
+SUBWORD_SYNC_OP (and, , &, unsigned short, 2, newval)
+SUBWORD_SYNC_OP (xor, , ^, unsigned short, 2, newval)
+SUBWORD_SYNC_OP (nand, ~, &, unsigned short, 2, newval)
+
+SUBWORD_SYNC_OP (add, , +, unsigned char, 1, newval)
+SUBWORD_SYNC_OP (sub, , -, unsigned char, 1, newval)
+SUBWORD_SYNC_OP (or, , |, unsigned char, 1, newval)
+SUBWORD_SYNC_OP (and, , &, unsigned char, 1, newval)
+SUBWORD_SYNC_OP (xor, , ^, unsigned char, 1, newval)
+SUBWORD_SYNC_OP (nand, ~, &, unsigned char, 1, newval)
+
+int HIDDEN
+__sync_val_compare_and_swap_4 (int *ptr, int oldval, int newval)
+{
+ int actual_oldval, fail;
+
+ while (1)
+ {
+ actual_oldval = *ptr;
+
+ if (oldval != actual_oldval)
+ return actual_oldval;
+
+ fail = __kernel_cmpxchg (actual_oldval, newval, ptr);
+
+ if (!fail)
+ return oldval;
+ }
+}
+
+#define SUBWORD_VAL_CAS(TYPE, WIDTH) \
+ TYPE HIDDEN \
+ __sync_val_compare_and_swap_##WIDTH (TYPE *ptr, TYPE oldval, \
+ TYPE newval) \
+ { \
+ int *wordptr = (int *)((unsigned int) ptr & ~3), fail; \
+ unsigned int mask, shift, actual_oldval, actual_newval; \
+ \
+ shift = (((unsigned int) ptr & 3) << 3) ^ INVERT_MASK_##WIDTH; \
+ mask = MASK_##WIDTH << shift; \
+ \
+ while (1) \
+ { \
+ actual_oldval = *wordptr; \
+ \
+ if (((actual_oldval & mask) >> shift) != (unsigned int) oldval) \
+ return (actual_oldval & mask) >> shift; \
+ \
+ actual_newval = (actual_oldval & ~mask) \
+ | (((unsigned int) newval << shift) & mask); \
+ \
+ fail = __kernel_cmpxchg (actual_oldval, actual_newval, \
+ wordptr); \
+ \
+ if (!fail) \
+ return oldval; \
+ } \
+ }
+
+SUBWORD_VAL_CAS (unsigned short, 2)
+SUBWORD_VAL_CAS (unsigned char, 1)
+
+typedef unsigned char bool;
+
+bool HIDDEN
+__sync_bool_compare_and_swap_4 (int *ptr, int oldval, int newval)
+{
+ int failure = __kernel_cmpxchg (oldval, newval, ptr);
+ return (failure == 0);
+}
+
+#define SUBWORD_BOOL_CAS(TYPE, WIDTH) \
+ bool HIDDEN \
+ __sync_bool_compare_and_swap_##WIDTH (TYPE *ptr, TYPE oldval, \
+ TYPE newval) \
+ { \
+ TYPE actual_oldval \
+ = __sync_val_compare_and_swap_##WIDTH (ptr, oldval, newval); \
+ return (oldval == actual_oldval); \
+ }
+
+SUBWORD_BOOL_CAS (unsigned short, 2)
+SUBWORD_BOOL_CAS (unsigned char, 1)
+
+void HIDDEN
+__sync_synchronize (void)
+{
+ __kernel_dmb ();
+}
+
+int HIDDEN
+__sync_lock_test_and_set_4 (int *ptr, int val)
+{
+ int failure, oldval;
+
+ do
+ {
+ oldval = *ptr;
+ failure = __kernel_cmpxchg (oldval, val, ptr);
+ }
+ while (failure != 0);
+
+ return oldval;
+}
+
+#define SUBWORD_TEST_AND_SET(TYPE, WIDTH) \
+ TYPE HIDDEN \
+ __sync_lock_test_and_set_##WIDTH (TYPE *ptr, TYPE val) \
+ { \
+ int failure; \
+ unsigned int oldval, newval, shift, mask; \
+ int *wordptr = (int *) ((unsigned int) ptr & ~3); \
+ \
+ shift = (((unsigned int) ptr & 3) << 3) ^ INVERT_MASK_##WIDTH; \
+ mask = MASK_##WIDTH << shift; \
+ \
+ do \
+ { \
+ oldval = *wordptr; \
+ newval = ((oldval & ~mask) \
+ | (((unsigned int) val << shift) & mask)); \
+ failure = __kernel_cmpxchg (oldval, newval, wordptr); \
+ } \
+ while (failure != 0); \
+ \
+ return (oldval & mask) >> shift; \
+ }
+
+SUBWORD_TEST_AND_SET (unsigned short, 2)
+SUBWORD_TEST_AND_SET (unsigned char, 1)
+
+#define SYNC_LOCK_RELEASE(TYPE, WIDTH) \
+ void HIDDEN \
+ __sync_lock_release_##WIDTH (TYPE *ptr) \
+ { \
+ /* All writes before this point must be seen before we release \
+ the lock itself. */ \
+ __kernel_dmb (); \
+ *ptr = 0; \
+ }
+
+SYNC_LOCK_RELEASE (int, 4)
+SYNC_LOCK_RELEASE (short, 2)
+SYNC_LOCK_RELEASE (char, 1)
diff --git a/libgcc/config/csky/linux-unwind.h b/libgcc/config/csky/linux-unwind.h
new file mode 100644
index 0000000..24638de
--- /dev/null
+++ b/libgcc/config/csky/linux-unwind.h
@@ -0,0 +1,131 @@
+/* DWARF2 EH unwinding support for C-SKY Linux.
+ Copyright (C) 2018 Free Software Foundation, Inc.
+ Contributed by C-SKY Microsystems and Mentor Graphics.
+
+ This file is part of GCC.
+
+ GCC is free software; you can redistribute it and/or modify
+ it under the terms of the GNU General Public License as published by
+ the Free Software Foundation; either version 3, or (at your option)
+ any later version.
+
+ GCC is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ GNU General Public License for more details.
+
+ Under Section 7 of GPL version 3, you are granted additional
+ permissions described in the GCC Runtime Library Exception, version
+ 3.1, as published by the Free Software Foundation.
+
+ You should have received a copy of the GNU General Public License and
+ a copy of the GCC Runtime Library Exception along with this program;
+ see the files COPYING3 and COPYING.RUNTIME respectively. If not, see
+ <http://www.gnu.org/licenses/>. */
+
+#ifndef inhibit_libc
+
+/* Do code reading to identify a signal frame, and set the frame
+ state data appropriately. See unwind-dw2.c for the structs. */
+
+#include <signal.h>
+#include <asm/unistd.h>
+
+/* The third parameter to the signal handler points to something with
+ this structure defined in asm/ucontext.h, but the name clashes with
+ struct ucontext from sys/ucontext.h so this private copy is used. */
+typedef struct _sig_ucontext {
+ unsigned long uc_flags;
+ struct _sig_ucontext *uc_link;
+ stack_t uc_stack;
+ struct sigcontext uc_mcontext;
+ sigset_t uc_sigmask;
+} _sig_ucontext_t;
+
+#define MD_FALLBACK_FRAME_STATE_FOR csky_fallback_frame_state
+
+static _Unwind_Reason_Code
+csky_fallback_frame_state (struct _Unwind_Context *context,
+ _Unwind_FrameState *fs)
+{
+ u_int16_t *pc = (u_int16_t *) context->ra;
+ struct sigcontext *sc;
+ _Unwind_Ptr new_cfa;
+ int i;
+
+ /* movi r7, __NR_rt_sigreturn; trap 0 */
+ if ((*(pc+0) == 0xea07) && (*(pc+1) == 119)
+ && (*(pc+2) == 0xc000) && (*(pc+3) == 0x2020))
+ {
+ struct sigframe
+ {
+ int sig;
+ int code;
+ struct sigcontext *psc;
+ unsigned long extramask[2]; /* _NSIG_WORDS */
+ struct sigcontext sc;
+ } *_rt = context->cfa;
+ sc = _rt->psc; // &(_rt->sc);
+ }
+ /* movi r7, __NR_rt_sigreturn; trap 0 */
+ else if ((*(pc+0) == 0xea07) && (*(pc+1) == 173)
+ && (*(pc+2) == 0xc000) && (*(pc+3) == 0x2020))
+ {
+ struct rt_sigframe
+ {
+ int sig;
+ struct siginfo *pinfo;
+ void* puc;
+ siginfo_t info;
+ struct ucontext uc;
+ } *_rt = context->cfa;
+ sc = &(_rt->uc.uc_mcontext);
+ }
+ else
+ return _URC_END_OF_STACK;
+
+ new_cfa = (_Unwind_Ptr) sc->sc_usp;
+ fs->regs.cfa_how = CFA_REG_OFFSET;
+ fs->regs.cfa_reg = STACK_POINTER_REGNUM;
+ fs->regs.cfa_offset = new_cfa - (_Unwind_Ptr) context->cfa;
+
+ fs->regs.reg[0].how = REG_SAVED_OFFSET;
+ fs->regs.reg[0].loc.offset = (_Unwind_Ptr)&(sc->sc_a0) - new_cfa;
+
+ fs->regs.reg[1].how = REG_SAVED_OFFSET;
+ fs->regs.reg[1].loc.offset = (_Unwind_Ptr)&(sc->sc_a1) - new_cfa;
+
+ fs->regs.reg[2].how = REG_SAVED_OFFSET;
+ fs->regs.reg[2].loc.offset = (_Unwind_Ptr)&(sc->sc_a2) - new_cfa;
+
+ fs->regs.reg[3].how = REG_SAVED_OFFSET;
+ fs->regs.reg[3].loc.offset = (_Unwind_Ptr)&(sc->sc_a3) - new_cfa;
+
+ for (i = 4; i < 14; i++)
+ {
+ fs->regs.reg[i].how = REG_SAVED_OFFSET;
+ fs->regs.reg[i].loc.offset = ((_Unwind_Ptr)&(sc->sc_regs[i - 4])
+ - new_cfa);
+ }
+
+ for (i = 16; i < 32; i++)
+ {
+ fs->regs.reg[i].how = REG_SAVED_OFFSET;
+ fs->regs.reg[i].loc.offset = ((_Unwind_Ptr)&(sc->sc_exregs[i - 16])
+ - new_cfa);
+ }
+
+ /* FIXME : hi lo ? */
+ fs->regs.reg[15].how = REG_SAVED_OFFSET;
+ fs->regs.reg[15].loc.offset = (_Unwind_Ptr)&(sc->sc_r15) - new_cfa;
+
+ fs->regs.reg[56].how = REG_SAVED_OFFSET;
+ fs->regs.reg[56].loc.offset = (_Unwind_Ptr)&(sc->sc_pc) - new_cfa;
+ fs->retaddr_column = 56;
+ fs->signal_frame = 1;
+
+ return _URC_NO_REASON;
+}
+
+
+#endif
diff --git a/libgcc/config/csky/t-csky b/libgcc/config/csky/t-csky
new file mode 100644
index 0000000..06aa1c1
--- /dev/null
+++ b/libgcc/config/csky/t-csky
@@ -0,0 +1,28 @@
+# Makefile fragment for all C-SKY targets.
+# Copyright (C) 2018 Free Software Foundation, Inc.
+# Contributed by C-SKY Microsystems and Mentor Graphics.
+#
+# This file is part of GCC.
+#
+# GCC is free software; you can redistribute it and/or modify
+# it under the terms of the GNU General Public License as published by
+# the Free Software Foundation; either version 3, or (at your option)
+# any later version.
+#
+# GCC is distributed in the hope that it will be useful,
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+# GNU General Public License for more details.
+#
+# You should have received a copy of the GNU General Public License
+# along with GCC; see the file COPYING3. If not see
+# <http://www.gnu.org/licenses/>.
+
+LIB1ASMSRC = csky/lib1funcs.S
+LIB1ASMFUNCS = _divsi3 _udivsi3 _modsi3 _umodsi3 _unorddf2 _unordsf2 \
+ _csky_case_sqi _csky_case_uqi _csky_case_shi _csky_case_uhi _csky_case_si
+
+LIB2FUNCS_EXCLUDE += _unord_df
+LIB2FUNCS_EXCLUDE += _unord_sf
+
+TARGET_LIBGCC2_CFLAGS=-O3 -DNO_FLOATLIB_FIXUNSDFSI
diff --git a/libgcc/config/csky/t-linux-csky b/libgcc/config/csky/t-linux-csky
new file mode 100644
index 0000000..1f5c4ce
--- /dev/null
+++ b/libgcc/config/csky/t-linux-csky
@@ -0,0 +1,21 @@
+# Makefile fragment for C-SKY targets running Linux.
+# Copyright (C) 2018 Free Software Foundation, Inc.
+# Contributed by C-SKY Microsystems and Mentor Graphics.
+#
+# This file is part of GCC.
+#
+# GCC is free software; you can redistribute it and/or modify
+# it under the terms of the GNU General Public License as published by
+# the Free Software Foundation; either version 3, or (at your option)
+# any later version.
+#
+# GCC is distributed in the hope that it will be useful,
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+# GNU General Public License for more details.
+#
+# You should have received a copy of the GNU General Public License
+# along with GCC; see the file COPYING3. If not see
+# <http://www.gnu.org/licenses/>.
+
+LIB2ADD_ST += $(srcdir)/config/csky/linux-atomic.c
^ permalink raw reply [flat|nested] 30+ messages in thread
* Re: [1/5] C-SKY port: Configury
2018-07-24 4:20 ` [1/5] C-SKY port: Configury Sandra Loosemore
@ 2018-07-24 15:09 ` Jeff Law
0 siblings, 0 replies; 30+ messages in thread
From: Jeff Law @ 2018-07-24 15:09 UTC (permalink / raw)
To: Sandra Loosemore, gcc-patches; +Cc: Han Mao, Yunhai Shang
On 07/23/2018 10:19 PM, Sandra Loosemore wrote:
> 2018-07-23 Jojo <jijie_rong@c-sky.com>
>            Huibin Wang <huibin_wang@c-sky.com>
>            Sandra Loosemore <sandra@codesourcery.com>
>            Chung-Lin Tang <cltang@codesourcery.com>
>            Andrew Jenner <andrew@codesourcery.com>
>
> Â Â Â Â Â Â Â C-SKY port: Configury
>
> Â Â Â Â Â Â Â gcc/
> Â Â Â Â Â Â Â * config.gcc (csky-*-*): New.
> Â Â Â Â Â Â Â * configure.ac: Add csky to targets for dwarf2 debug_line support.
> Â Â Â Â Â Â Â * configure: Regenerated.
>
> Â Â Â Â Â Â Â contrib/
> Â Â Â Â Â Â Â * config-list.mk (LIST): Add csky-elf and csky-linux-gnu.
No concerns here. OK when the kit as a whole is approved.
jeff
^ permalink raw reply [flat|nested] 30+ messages in thread
* Re: [4/5] C-SKY port: Testsuite
2018-07-24 4:25 ` [4/5] C-SKY port: Testsuite Sandra Loosemore
@ 2018-07-24 15:10 ` Jeff Law
0 siblings, 0 replies; 30+ messages in thread
From: Jeff Law @ 2018-07-24 15:10 UTC (permalink / raw)
To: Sandra Loosemore, gcc-patches; +Cc: Han Mao, Yunhai Shang
On 07/23/2018 10:25 PM, Sandra Loosemore wrote:
> 2018-07-23 Sandra Loosemore <sandra@codesourcery.com>
>            Chung-Lin Tang <cltang@codesourcery.com>
>
> Â Â Â Â Â Â Â C-SKY port: Testsuite
>
> Â Â Â Â Â Â Â gcc/testsuite/
> Â Â Â Â Â Â Â * g++.dg/Wno-frame-address.C: Adjust for C-SKY.
> Â Â Â Â Â Â Â * g++.dg/torture/type-generic-1.C: Likewise.
> Â Â Â Â Â Â Â * gcc.c-torture/compile/20000804-1.c: Likewise.
> Â Â Â Â Â Â Â * gcc.c-torture/execute/20101011-1.c: Likewise.
> Â Â Â Â Â Â Â * gcc.c-torture/execute/ieee/mul-subnormal-single-1.x: Likewise.
> Â Â Â Â Â Â Â * gcc.dg/20020312-2.c: Likewise.
> Â Â Â Â Â Â Â * gcc.dg/Wno-frame-address.c: Likewise.
> Â Â Â Â Â Â Â * gcc.dg/c11-true_min-1.c: Likewise.
> Â Â Â Â Â Â Â * gcc.dg/sibcall-10.c: Likewise.
> Â Â Â Â Â Â Â * gcc.dg/sibcall-9.c: Likewise.
> Â Â Â Â Â Â Â * gcc.dg/stack-usage-1.c: Likewise.
> Â Â Â Â Â Â Â * gcc.dg/torture/float32-tg-3.c: Likewise.
> Â Â Â Â Â Â Â * gcc.dg/torture/float32x-tg-3.c: Likewise.
> Â Â Â Â Â Â Â * gcc.dg/torture/float64-tg-3.c: Likewise.
> Â Â Â Â Â Â Â * gcc.dg/torture/float64x-tg-3.c: Likewise.
> Â Â Â Â Â Â Â * gcc.dg/torture/type-generic-1.c: Likewise.
> Â Â Â Â Â Â Â * gcc.target/csky/*: New.
> Â Â Â Â Â Â Â * lib/target-supports.exp (check_profiling_available): Add
> Â Â Â Â Â Â Â csky-*-elf.
> Â Â Â Â Â Â Â (check_effective_target_hard_float): Handle C-SKY targets with
> Â Â Â Â Â Â Â single-precision hard float only.
> Â Â Â Â Â Â Â (check_effective_target_logical_op_short_circuit): Handle C-SKY.
No concerns here. OK when the kit as a whole is approved.
Jeff
^ permalink raw reply [flat|nested] 30+ messages in thread
* Re: [3/5] C-SKY port: Documentation
2018-07-24 4:23 ` [3/5] C-SKY port: Documentation Sandra Loosemore
@ 2018-07-24 15:10 ` Jeff Law
0 siblings, 0 replies; 30+ messages in thread
From: Jeff Law @ 2018-07-24 15:10 UTC (permalink / raw)
To: Sandra Loosemore, gcc-patches; +Cc: Han Mao, Yunhai Shang
On 07/23/2018 10:23 PM, Sandra Loosemore wrote:
> 2018-07-23 Sandra Loosemore <sandra@codesourcery.com>
>
> Â Â Â Â Â Â Â C-SKY port: Documentation
>
> Â Â Â Â Â Â Â gcc/
> Â Â Â Â Â Â Â * doc/extend.texi (C-SKY Function Attributes): New section.
> Â Â Â Â Â Â Â * doc/invoke.texi (Option Summary): Add C-SKY options.
> Â Â Â Â Â Â Â (C-SKY Options): New section.
> Â Â Â Â Â Â Â * doc/md.texi (Machine Constraints): Document C-SKY constraints.
No concerns here. OK when the kit as a whole is approved.
^ permalink raw reply [flat|nested] 30+ messages in thread
* Re: [5/5] C-SKY port: libgcc
2018-07-24 4:26 ` [5/5] C-SKY port: libgcc Sandra Loosemore
@ 2018-07-24 15:12 ` Jeff Law
2018-07-24 18:10 ` Segher Boessenkool
1 sibling, 0 replies; 30+ messages in thread
From: Jeff Law @ 2018-07-24 15:12 UTC (permalink / raw)
To: Sandra Loosemore, gcc-patches; +Cc: Han Mao, Yunhai Shang
On 07/23/2018 10:26 PM, Sandra Loosemore wrote:
> 2018-07-23 Jojo <jijie_rong@c-sky.com>
>            Huibin Wang <huibin_wang@c-sky.com>
>            Sandra Loosemore <sandra@codesourcery.com>
>            Chung-Lin Tang <cltang@codesourcery.com>
>
> Â Â Â Â Â Â Â C-SKY port: libgcc
>
> Â Â Â Â Â Â Â libgcc/
> Â Â Â Â Â Â Â * config.host: Add C-SKY support.
> Â Â Â Â Â Â Â * config/csky/*: New.
Obviously we'll trust you on all that assembly code :-)
OK when the kit as a whole is approved.
Easy stuff is done, onward to the actual port.
jeff
^ permalink raw reply [flat|nested] 30+ messages in thread
* Re: [0/5] C-SKY port
2018-07-24 4:17 [0/5] C-SKY port Sandra Loosemore
` (4 preceding siblings ...)
2018-07-24 4:26 ` [5/5] C-SKY port: libgcc Sandra Loosemore
@ 2018-07-24 15:23 ` Sandra Loosemore
2018-07-26 23:04 ` Joseph Myers
5 siblings, 1 reply; 30+ messages in thread
From: Sandra Loosemore @ 2018-07-24 15:23 UTC (permalink / raw)
To: gcc-patches, Xianmiao Qu; +Cc: Han Mao, Yunhai Shang
On 07/23/2018 10:17 PM, Sandra Loosemore wrote:
> C-SKY proposes Han Mao <han_mao@c-sky.com> and Yunhai Shang
> <yunhai_shang@c-sky.com> as maintainers for this port.
I apologize, I made a mistake here. The correct proposed maintainers
are Xianmiao Qu <xianmiao_qu@c-sky.com> and Yunhai Shang
<yunhai_shang@c-sky.com>.
> We expect that
> C-SKY will also be providing a public link to the processor and ABI
> documentation at some point.
The ABI manual has been posted, but not the ISA documentation yet. (I'd
guess that when it does show up it will be in the same place, though.)
https://github.com/c-sky/csky-doc
-Sandra
^ permalink raw reply [flat|nested] 30+ messages in thread
* Re: [2/5] C-SKY port: Backend implementation
2018-07-24 4:21 ` [2/5] C-SKY port: Backend implementation Sandra Loosemore
@ 2018-07-24 15:45 ` Jeff Law
2018-07-24 18:19 ` Sandra Loosemore
0 siblings, 1 reply; 30+ messages in thread
From: Jeff Law @ 2018-07-24 15:45 UTC (permalink / raw)
To: Sandra Loosemore, gcc-patches; +Cc: Han Mao, Yunhai Shang
On 07/23/2018 10:21 PM, Sandra Loosemore wrote:
> 2018-07-23 Jojo <jijie_rong@c-sky.com>
>            Huibin Wang <huibin_wang@c-sky.com>
>            Sandra Loosemore <sandra@codesourcery.com>
>            Chung-Lin Tang <cltang@codesourcery.com>
>
> Â Â Â Â Â Â Â C-SKY port: Backend implementation
>
> Â Â Â Â Â Â Â gcc/
> Â Â Â Â Â Â Â * config/csky/*: New.
> Â Â Â Â Â Â Â * common/config/csky/*: New.
Let's avoid gratutious whitespace that attempts to line up conditionals.
As an example, look at the predicate csky_load_multiple_operation. I
think just doing a quick pass over the .c, .h and main .md files should
be sufficient here.
I'm not a big fan of more awk code, but I'm not going to object to it :-)
Why does the port have its own little pass for condition code
optimization (cse_cc)? What is it doing that can't be done with our
generic optimizers?
Any thoughts on using the newer function descriptor bits rather than old
style stack trampolines?
I don't see anything terribly concerning in the core of the port. The
amount of support code for minipool is huge and I wonder if some sharing
across the various ports would be possible, but I don't think that
should be a blocking issue for this port.
Can you update the backends.html web page here appropriately for the
c-sky target?
I'd like to take a closer look, but those are the high level comment's
I've got this morning :-)
Jeff
^ permalink raw reply [flat|nested] 30+ messages in thread
* Re: [5/5] C-SKY port: libgcc
2018-07-24 4:26 ` [5/5] C-SKY port: libgcc Sandra Loosemore
2018-07-24 15:12 ` Jeff Law
@ 2018-07-24 18:10 ` Segher Boessenkool
2018-07-24 18:19 ` Sandra Loosemore
1 sibling, 1 reply; 30+ messages in thread
From: Segher Boessenkool @ 2018-07-24 18:10 UTC (permalink / raw)
To: Sandra Loosemore; +Cc: gcc-patches, Han Mao, Yunhai Shang
On Mon, Jul 23, 2018 at 10:26:35PM -0600, Sandra Loosemore wrote:
> diff --git a/libgcc/config.host b/libgcc/config.host
> index 18cabaf..b2ee0c9 100644
> --- a/libgcc/config.host
> +++ b/libgcc/config.host
> @@ -94,6 +94,9 @@ am33_2.0-*-linux*)
> arc*-*-*)
> cpu_type=arc
> ;;
> +csky*-*-*)
> + cpu_type=csky
> + ;;
> arm*-*-*)
> cpu_type=arm
> ;;
This long list was alphabetic before (except x86_64 and tic6x, alas);
let's not make things worse?
Segher
^ permalink raw reply [flat|nested] 30+ messages in thread
* Re: [2/5] C-SKY port: Backend implementation
2018-07-24 15:45 ` Jeff Law
@ 2018-07-24 18:19 ` Sandra Loosemore
2018-07-24 21:25 ` Jeff Law
0 siblings, 1 reply; 30+ messages in thread
From: Sandra Loosemore @ 2018-07-24 18:19 UTC (permalink / raw)
To: Jeff Law, gcc-patches; +Cc: Xianmiao Qu, Yunhai Shang
On 07/24/2018 09:45 AM, Jeff Law wrote:
> On 07/23/2018 10:21 PM, Sandra Loosemore wrote:
>> 2018-07-23 Jojo <jijie_rong@c-sky.com>
>>            Huibin Wang <huibin_wang@c-sky.com>
>>            Sandra Loosemore <sandra@codesourcery.com>
>>            Chung-Lin Tang <cltang@codesourcery.com>
>>
>> Â Â Â Â Â Â Â C-SKY port: Backend implementation
>>
>> Â Â Â Â Â Â Â gcc/
>> Â Â Â Â Â Â Â * config/csky/*: New.
>> Â Â Â Â Â Â Â * common/config/csky/*: New.
>
> Let's avoid gratutious whitespace that attempts to line up conditionals.
> As an example, look at the predicate csky_load_multiple_operation. I
> think just doing a quick pass over the .c, .h and main .md files should
> be sufficient here.
OK, will do.
> I'm not a big fan of more awk code, but I'm not going to object to it :-)
>
> Why does the port have its own little pass for condition code
> optimization (cse_cc)? What is it doing that can't be done with our
> generic optimizers?
This pass was included in the initial patch set we got from C-SKY, and
as it didn't seem to break anything I left it in. Perhaps C-SKY can
provide a testcase that demonstrates why it's still useful in the
current version of GCC; otherwise we can remove this from the initial
port submission and restore it later if some performance analysis shows
it is still worthwhile.
> Any thoughts on using the newer function descriptor bits rather than old
> style stack trampolines?
Has that been committed? I vaguely remembered discussion of a new way
to handle nested functions without using the trampoline interface, but I
couldn't find any documentation in the internals manual.
> I don't see anything terribly concerning in the core of the port. The
> amount of support code for minipool is huge and I wonder if some sharing
> across the various ports would be possible, but I don't think that
> should be a blocking issue for this port.
Yes, that code was clearly copied almost verbatim from the ARM backend.
I left it alone as much as possible to simplify any future attempts at
genericizing it.
> Can you update the backends.html web page here appropriately for the
> c-sky target?
Sure, I can take care of updating that when the port is committed. I
believe the right entry is
"csky b ia"
> I'd like to take a closer look, but those are the high level comment's
> I've got this morning :-)
Thanks. I'll wait a bit for more comments to come in before preparing a
revised patch.
-Sandra
^ permalink raw reply [flat|nested] 30+ messages in thread
* Re: [5/5] C-SKY port: libgcc
2018-07-24 18:10 ` Segher Boessenkool
@ 2018-07-24 18:19 ` Sandra Loosemore
2018-07-24 19:46 ` Segher Boessenkool
0 siblings, 1 reply; 30+ messages in thread
From: Sandra Loosemore @ 2018-07-24 18:19 UTC (permalink / raw)
To: Segher Boessenkool; +Cc: gcc-patches, Han Mao, Yunhai Shang
On 07/24/2018 12:10 PM, Segher Boessenkool wrote:
> On Mon, Jul 23, 2018 at 10:26:35PM -0600, Sandra Loosemore wrote:
>> diff --git a/libgcc/config.host b/libgcc/config.host
>> index 18cabaf..b2ee0c9 100644
>> --- a/libgcc/config.host
>> +++ b/libgcc/config.host
>> @@ -94,6 +94,9 @@ am33_2.0-*-linux*)
>> arc*-*-*)
>> cpu_type=arc
>> ;;
>> +csky*-*-*)
>> + cpu_type=csky
>> + ;;
>> arm*-*-*)
>> cpu_type=arm
>> ;;
>
> This long list was alphabetic before (except x86_64 and tic6x, alas);
> let's not make things worse?
Oops! Good catch on that. I'll take care of it.
-Sandra
^ permalink raw reply [flat|nested] 30+ messages in thread
* Re: [5/5] C-SKY port: libgcc
2018-07-24 18:19 ` Sandra Loosemore
@ 2018-07-24 19:46 ` Segher Boessenkool
0 siblings, 0 replies; 30+ messages in thread
From: Segher Boessenkool @ 2018-07-24 19:46 UTC (permalink / raw)
To: Sandra Loosemore; +Cc: gcc-patches, Han Mao, Yunhai Shang
On Tue, Jul 24, 2018 at 12:19:30PM -0600, Sandra Loosemore wrote:
> On 07/24/2018 12:10 PM, Segher Boessenkool wrote:
> >On Mon, Jul 23, 2018 at 10:26:35PM -0600, Sandra Loosemore wrote:
> >>diff --git a/libgcc/config.host b/libgcc/config.host
> >>index 18cabaf..b2ee0c9 100644
> >>--- a/libgcc/config.host
> >>+++ b/libgcc/config.host
> >>@@ -94,6 +94,9 @@ am33_2.0-*-linux*)
> >> arc*-*-*)
> >> cpu_type=arc
> >> ;;
> >>+csky*-*-*)
> >>+ cpu_type=csky
> >>+ ;;
> >> arm*-*-*)
> >> cpu_type=arm
> >> ;;
> >
> >This long list was alphabetic before (except x86_64 and tic6x, alas);
> >let's not make things worse?
>
> Oops! Good catch on that. I'll take care of it.
Thanks!
Rest looks fine fwiw (I just skimmed it, and I cannot read .gz files
without some effort, so take it for what it's worth: not too much ;-) ).
Segher
^ permalink raw reply [flat|nested] 30+ messages in thread
* Re: [2/5] C-SKY port: Backend implementation
2018-07-24 18:19 ` Sandra Loosemore
@ 2018-07-24 21:25 ` Jeff Law
2018-07-25 0:17 ` Sandra Loosemore
2018-07-26 6:07 ` 瞿仙淼
0 siblings, 2 replies; 30+ messages in thread
From: Jeff Law @ 2018-07-24 21:25 UTC (permalink / raw)
To: Sandra Loosemore, gcc-patches; +Cc: Xianmiao Qu, Yunhai Shang
On 07/24/2018 12:18 PM, Sandra Loosemore wrote:
> On 07/24/2018 09:45 AM, Jeff Law wrote:
>> On 07/23/2018 10:21 PM, Sandra Loosemore wrote:
>>> 2018-07-23 Jojo <jijie_rong@c-sky.com>
>>>             Huibin Wang <huibin_wang@c-sky.com>
>>>             Sandra Loosemore <sandra@codesourcery.com>
>>>             Chung-Lin Tang <cltang@codesourcery.com>
>>>
>>> Â Â Â Â Â Â Â Â C-SKY port: Backend implementation
>>>
>>> Â Â Â Â Â Â Â Â gcc/
>>> Â Â Â Â Â Â Â Â * config/csky/*: New.
>>> Â Â Â Â Â Â Â Â * common/config/csky/*: New.
>>
>> Let's avoid gratutious whitespace that attempts to line up conditionals.
>>   As an example, look at the predicate csky_load_multiple_operation. I
>> think just doing a quick pass over the .c, .h and main .md files should
>> be sufficient here.
>
> OK, will do.
>
>> I'm not a big fan of more awk code, but I'm not going to object to it :-)
>>
>> Why does the port have its own little pass for condition code
>> optimization (cse_cc)? What is it doing that can't be done with our
>> generic optimizers?
>
> This pass was included in the initial patch set we got from C-SKY, and
> as it didn't seem to break anything I left it in. Perhaps C-SKY can
> provide a testcase that demonstrates why it's still useful in the
> current version of GCC; otherwise we can remove this from the initial
> port submission and restore it later if some performance analysis shows
> it is still worthwhile.
FWIW it looks like we model CC setting on just a few insns, (add,
subtract) so I'd be surprised if this little mini pass found much. I'd
definitely like to hear from the csky authors here.
Alternately, you could do add some instrumentation to flag when it
triggers, take a test or two that does, reduce it and we can then look
at the key RTL sequences and see what the pass is really doing.
>
>> Any thoughts on using the newer function descriptor bits rather than old
>> style stack trampolines?
>
> Has that been committed? I vaguely remembered discussion of a new way
> to handle nested functions without using the trampoline interface, but I
> couldn't find any documentation in the internals manual.
It did. See TARGET_CUSTOM_FUNCTION_DESCRIPTORS and the (relatively few)
ports that define it.
>
>> I don't see anything terribly concerning in the core of the port. The
>> amount of support code for minipool is huge and I wonder if some sharing
>> across the various ports would be possible, but I don't think that
>> should be a blocking issue for this port.
>
> Yes, that code was clearly copied almost verbatim from the ARM backend.
> I left it alone as much as possible to simplify any future attempts at
> genericizing it.
Understood -- I'd assumed it was largely copied from ARM, but hadn't
gone back to the ARM bits to verify.
>
>> Can you update the backends.html web page here appropriately for the
>> c-sky target?
>
> Sure, I can take care of updating that when the port is committed. I
> believe the right entry is
>
> "csky                         b  ia"
Yea, that seems right to me.
Jeff
^ permalink raw reply [flat|nested] 30+ messages in thread
* Re: [2/5] C-SKY port: Backend implementation
2018-07-24 21:25 ` Jeff Law
@ 2018-07-25 0:17 ` Sandra Loosemore
2018-07-25 4:50 ` Jeff Law
2018-07-26 6:07 ` 瞿仙淼
1 sibling, 1 reply; 30+ messages in thread
From: Sandra Loosemore @ 2018-07-25 0:17 UTC (permalink / raw)
To: Jeff Law, gcc-patches; +Cc: Xianmiao Qu, Yunhai Shang
On 07/24/2018 03:24 PM, Jeff Law wrote:
>>
>>> Any thoughts on using the newer function descriptor bits rather than old
>>> style stack trampolines?
>>
>> Has that been committed? I vaguely remembered discussion of a new way
>> to handle nested functions without using the trampoline interface, but I
>> couldn't find any documentation in the internals manual.
> It did. See TARGET_CUSTOM_FUNCTION_DESCRIPTORS and the (relatively few)
> ports that define it.
Hmmm, I completely failed to make that connection from the docs -- the
whole description of that hook is pretty gibberishy and I thought it was
something for targets where the ABI already specifies some "standard
calling sequence" using descriptors (C-SKY doesn't), rather than a
generic alternative to executable trampolines. Putting on my doc
maintainer hat briefly, I can see this needs a lot of work. :-(
Anyway, is this required for new ports nowadays? If so, I at least know
what to search for now. At this point I couldn't say whether this would
do anything to fix the situation on ck801 targets where there simply
aren't enough spare registers available to the trampoline to both hold
the static link and do an indirect jump.
-Sandra
^ permalink raw reply [flat|nested] 30+ messages in thread
* Re: [2/5] C-SKY port: Backend implementation
2018-07-25 0:17 ` Sandra Loosemore
@ 2018-07-25 4:50 ` Jeff Law
2018-07-25 13:17 ` Paul Koning
0 siblings, 1 reply; 30+ messages in thread
From: Jeff Law @ 2018-07-25 4:50 UTC (permalink / raw)
To: Sandra Loosemore, gcc-patches; +Cc: Xianmiao Qu, Yunhai Shang
On 07/24/2018 06:17 PM, Sandra Loosemore wrote:
> On 07/24/2018 03:24 PM, Jeff Law wrote:
>>>
>>>> Any thoughts on using the newer function descriptor bits rather than
>>>> old
>>>> style stack trampolines?
>>>
>>> Has that been committed? I vaguely remembered discussion of a new way
>>> to handle nested functions without using the trampoline interface, but I
>>> couldn't find any documentation in the internals manual.
>> It did. See TARGET_CUSTOM_FUNCTION_DESCRIPTORS and the (relatively few)
>> ports that define it.
>
> Hmmm, I completely failed to make that connection from the docs -- the
> whole description of that hook is pretty gibberishy and I thought it was
> something for targets where the ABI already specifies some "standard
> calling sequence" using descriptors (C-SKY doesn't), rather than a
> generic alternative to executable trampolines. Putting on my doc
> maintainer hat briefly, I can see this needs a lot of work. :-(
Most likely :-) So many things to do, so little time.
>
> Anyway, is this required for new ports nowadays? If so, I at least know
> what to search for now. At this point I couldn't say whether this would
> do anything to fix the situation on ck801 targets where there simply
> aren't enough spare registers available to the trampoline to both hold
> the static link and do an indirect jump.
It's not required, but preferred, particularly if the part has an MMU
that can provide no-execute protections on pages in memory. If the
target doesn't have an mmu, then it's of marginal value.
The key advantage it has over the old trampoline implementation is that
stacks can remain non-executable, even for Ada and nested functions.
That's a big win from a security standpoint.
Jeff
^ permalink raw reply [flat|nested] 30+ messages in thread
* Re: [2/5] C-SKY port: Backend implementation
2018-07-25 4:50 ` Jeff Law
@ 2018-07-25 13:17 ` Paul Koning
2018-07-25 14:54 ` Sandra Loosemore
0 siblings, 1 reply; 30+ messages in thread
From: Paul Koning @ 2018-07-25 13:17 UTC (permalink / raw)
To: Jeff Law; +Cc: Sandra Loosemore, gcc-patches, Xianmiao Qu, Yunhai Shang
> On Jul 25, 2018, at 12:50 AM, Jeff Law <law@redhat.com> wrote:
>
>>>> ...
>>> It did. See TARGET_CUSTOM_FUNCTION_DESCRIPTORS and the (relatively few)
>>> ports that define it.
>>
>> Hmmm, I completely failed to make that connection from the docs -- the
>> whole description of that hook is pretty gibberishy and I thought it was
>> something for targets where the ABI already specifies some "standard
>> calling sequence" using descriptors (C-SKY doesn't), rather than a
>> generic alternative to executable trampolines. Putting on my doc
>> maintainer hat briefly, I can see this needs a lot of work. :-(
> Most likely :-) So many things to do, so little time.
>
>
>>
>> Anyway, is this required for new ports nowadays? If so, I at least know
>> what to search for now. At this point I couldn't say whether this would
>> do anything to fix the situation on ck801 targets where there simply
>> aren't enough spare registers available to the trampoline to both hold
>> the static link and do an indirect jump.
> It's not required, but preferred, particularly if the part has an MMU
> that can provide no-execute protections on pages in memory. If the
> target doesn't have an mmu, then it's of marginal value.
>
> The key advantage it has over the old trampoline implementation is that
> stacks can remain non-executable, even for Ada and nested functions.
> That's a big win from a security standpoint.
Non-executable stacks are a very good thing.
That said, I also looked at the target hook documentation and was left without any clue whatsoever. It sure isn't clear what powers of two have to do with descriptors, or what descriptors have to do with support for nested functions.
Can you suggest places to look to get an understanding of this feature? It sounds like the only other option is "Use the source, Luke". Any specific targets that make a good reference implementation for this?
paul
^ permalink raw reply [flat|nested] 30+ messages in thread
* Re: [2/5] C-SKY port: Backend implementation
2018-07-25 13:17 ` Paul Koning
@ 2018-07-25 14:54 ` Sandra Loosemore
0 siblings, 0 replies; 30+ messages in thread
From: Sandra Loosemore @ 2018-07-25 14:54 UTC (permalink / raw)
To: Paul Koning, Jeff Law; +Cc: gcc-patches, Xianmiao Qu, Yunhai Shang
On 07/25/2018 07:16 AM, Paul Koning wrote:
> Non-executable stacks are a very good thing.
>
> That said, I also looked at the target hook documentation and was
> left without any clue whatsoever. It sure isn't clear what powers of
> two have to do with descriptors, or what descriptors have to do with
> support for nested functions.
>
> Can you suggest places to look to get an understanding of this
> feature? It sounds like the only other option is "Use the source,
> Luke". Any specific targets that make a good reference
> implementation for this?
FYI, so far I have found PR ada/67205 and the original patch posting
here, but it looks like "Use the source" is indeed where we are on this.
:-(
https://gcc.gnu.org/ml/gcc-patches/2016-06/msg02016.html
-Sandra
^ permalink raw reply [flat|nested] 30+ messages in thread
* Re: [2/5] C-SKY port: Backend implementation
2018-07-24 21:25 ` Jeff Law
2018-07-25 0:17 ` Sandra Loosemore
@ 2018-07-26 6:07 ` 瞿仙淼
2018-07-28 1:49 ` Sandra Loosemore
2018-08-02 22:28 ` Jeff Law
1 sibling, 2 replies; 30+ messages in thread
From: 瞿仙淼 @ 2018-07-26 6:07 UTC (permalink / raw)
To: Jeff Law; +Cc: Sandra Loosemore, gcc-patches, Yunhai Shang
> 在 2018年7月25日，上午5:24，Jeff Law <law@redhat.com> 写道：
>
> On 07/24/2018 12:18 PM, Sandra Loosemore wrote:
>> On 07/24/2018 09:45 AM, Jeff Law wrote:
>>> On 07/23/2018 10:21 PM, Sandra Loosemore wrote:
>>> I'm not a big fan of more awk code, but I'm not going to object to it :-)
>>>
>>> Why does the port have its own little pass for condition code
>>> optimization (cse_cc)? What is it doing that can't be done with our
>>> generic optimizers?
>>
>> This pass was included in the initial patch set we got from C-SKY, and
>> as it didn't seem to break anything I left it in. Perhaps C-SKY can
>> provide a testcase that demonstrates why it's still useful in the
>> current version of GCC; otherwise we can remove this from the initial
>> port submission and restore it later if some performance analysis shows
>> it is still worthwhile.
> FWIW it looks like we model CC setting on just a few insns, (add,
> subtract) so I'd be surprised if this little mini pass found much. I'd
> definitely like to hear from the csky authors here.
>
> Alternately, you could do add some instrumentation to flag when it
> triggers, take a test or two that does, reduce it and we can then look
> at the key RTL sequences and see what the pass is really doing.
>
I wrote a case to reproduce this problem on C-SKY. C code is as follows:
-----------------------------------------------------------------------
int e1, e2;
void func (int a, int b, int c, int d, int f, int g)
{
e1 = a > b ? f : g;
e2 = a > b ? c : d;
return;
}
-----------------------------------------------------------------------
compile to assembler with option “-O3 -S” :
-----------------------------------------------------------------------
func:
cmplt a1, a0
ld.w t1, (sp, 0)
ld.w t0, (sp, 4)
movt t0, t1
cmplt a1, a0
movt a3, a2
lrw a1, e2
lrw a2, e1
st.w a3, (a1, 0)
st.w t0, (a2, 0)
rts
-----------------------------------------------------------------------
There is an extra “cmplt a1, a0" in the above code without cse_cc. This situation mainly occurs when a relatively short branch jump is converted into a conditional execution instruction. And the CSE pass can not reduce the same conditional comparison instruction . Here is the rtx sequence after “cse2” pass.
-----------------------------------------------------------------------
(insn 28 13 29 2 (set (reg:CC 33 c)
(gt:CC (reg/v:SI 77 [ a ])
(reg/v:SI 78 [ b ]))) func.c:5 1099 {*cmpgtsi}
(nil))
(insn 29 28 30 2 (set (reg/v:SI 82 [ g ])
(if_then_else:SI (eq (reg:CC 33 c)
(const_int 0 [0]))
(reg/v:SI 82 [ g ])
(reg/v:SI 81 [ f ]))) func.c:5 983 {movf}
(expr_list:REG_DEAD (reg/v:SI 81 [ f ])
(expr_list:REG_DEAD (reg:CC 33 c)
(nil))))
(insn 30 29 31 2 (set (reg:CC 33 c)
(gt:CC (reg/v:SI 77 [ a ])
(reg/v:SI 78 [ b ]))) func.c:5 1099 {*cmpgtsi}
(expr_list:REG_DEAD (reg/v:SI 78 [ b ])
(expr_list:REG_DEAD (reg/v:SI 77 [ a ])
(nil))))
(insn 31 30 18 2 (set (reg/v:SI 80 [ d ])
(if_then_else:SI (eq (reg:CC 33 c)
(const_int 0 [0]))
(reg/v:SI 80 [ d ])
(reg/v:SI 79 [ c ]))) func.c:5 983 {movf}
(expr_list:REG_DEAD (reg/v:SI 79 [ c ])
(expr_list:REG_DEAD (reg:CC 33 c)
(nil))))
-----------------------------------------------------------------------
It doesn't seem to check the same conditional comparison instruction .So I wrote this to solve this problem, but I am not sure if this is the best way : )
PS, the same conditional comparison instruction cannot be reduced with the latest version gcc with C-SKY because I just insert the “cse_cc” after “cse1”, when I insert after “cse2”, this problem can be solved very well.
Thanks,
Xianmiao
^ permalink raw reply [flat|nested] 30+ messages in thread
* Re: [0/5] C-SKY port
2018-07-24 15:23 ` [0/5] C-SKY port Sandra Loosemore
@ 2018-07-26 23:04 ` Joseph Myers
2018-07-30 16:59 ` Sandra Loosemore
2018-08-01 14:28 ` 瞿仙淼
0 siblings, 2 replies; 30+ messages in thread
From: Joseph Myers @ 2018-07-26 23:04 UTC (permalink / raw)
To: Sandra Loosemore; +Cc: gcc-patches, Xianmiao Qu, Han Mao, Yunhai Shang
On Tue, 24 Jul 2018, Sandra Loosemore wrote:
> On 07/23/2018 10:17 PM, Sandra Loosemore wrote:
>
> > C-SKY proposes Han Mao <han_mao@c-sky.com> and Yunhai Shang
> > <yunhai_shang@c-sky.com> as maintainers for this port.
>
> I apologize, I made a mistake here. The correct proposed maintainers are
> Xianmiao Qu <xianmiao_qu@c-sky.com> and Yunhai Shang
> <yunhai_shang@c-sky.com>.
I'd like to confirm: will the maintainers be running a bot reporting test
results for trunk for at least one C-SKY configuration to gcc-testresults
(daily or thereabouts), and monitoring results for regressions and
addressing those regressions promptly? We don't yet have such a
requirement, but when an architecture-independent patch causes failures on
a particular architectures, it's much better if the architecture
maintainers raise the issue while the patch is still fresh in the author's
mind (meaning within a few days at most of the commit) than if they only
look at results and track down the patch causing the failure many months
later (e.g. while reviewing test results in preparation for the following
release).
> > We expect that
> > C-SKY will also be providing a public link to the processor and ABI
> > documentation at some point.
>
> The ABI manual has been posted, but not the ISA documentation yet. (I'd guess
> that when it does show up it will be in the same place, though.)
>
> https://github.com/c-sky/csky-doc
Could you provide the proposed GCC website changes for the port
(backends.html, readings.html, news item for index.html)? readings.html,
in particular, would link to the ABI and ISA documentation, while
backends.html gives summary information about the properties of both the
architecture and the GCC port.
--
Joseph S. Myers
joseph@codesourcery.com
^ permalink raw reply [flat|nested] 30+ messages in thread
* Re: [2/5] C-SKY port: Backend implementation
2018-07-26 6:07 ` 瞿仙淼
@ 2018-07-28 1:49 ` Sandra Loosemore
2018-08-02 22:33 ` Jeff Law
2018-08-02 22:28 ` Jeff Law
1 sibling, 1 reply; 30+ messages in thread
From: Sandra Loosemore @ 2018-07-28 1:49 UTC (permalink / raw)
To: 瞿仙淼, Jeff Law; +Cc: gcc-patches, Yunhai Shang
On 07/26/2018 12:06 AM, 瞿仙淼 wrote:
>
> I wrote a case to reproduce this problem on C-SKY. C code is as follows:
> -----------------------------------------------------------------------
> int e1, e2;
>
> void func (int a, int b, int c, int d, int f, int g)
> {
> e1 = a > b ? f : g;
> e2 = a > b ? c : d;
>
> return;
> }
> -----------------------------------------------------------------------
>
> compile to assembler with option “-O3 -S†:
> -----------------------------------------------------------------------
> func:
> cmplt a1, a0
> ld.w t1, (sp, 0)
> ld.w t0, (sp, 4)
> movt t0, t1
> cmplt a1, a0
> movt a3, a2
> lrw a1, e2
> lrw a2, e1
> st.w a3, (a1, 0)
> st.w t0, (a2, 0)
> rts
> -----------------------------------------------------------------------
> There is an extra “cmplt a1, a0" in the above code without cse_cc. This situation mainly occurs when a relatively short branch jump is converted into a conditional execution instruction. And the CSE pass can not reduce the same conditional comparison instruction . Here is the rtx sequence after “cse2†pass.
>
> -----------------------------------------------------------------------
> (insn 28 13 29 2 (set (reg:CC 33 c)
> (gt:CC (reg/v:SI 77 [ a ])
> (reg/v:SI 78 [ b ]))) func.c:5 1099 {*cmpgtsi}
> (nil))
> (insn 29 28 30 2 (set (reg/v:SI 82 [ g ])
> (if_then_else:SI (eq (reg:CC 33 c)
> (const_int 0 [0]))
> (reg/v:SI 82 [ g ])
> (reg/v:SI 81 [ f ]))) func.c:5 983 {movf}
> (expr_list:REG_DEAD (reg/v:SI 81 [ f ])
> (expr_list:REG_DEAD (reg:CC 33 c)
> (nil))))
> (insn 30 29 31 2 (set (reg:CC 33 c)
> (gt:CC (reg/v:SI 77 [ a ])
> (reg/v:SI 78 [ b ]))) func.c:5 1099 {*cmpgtsi}
> (expr_list:REG_DEAD (reg/v:SI 78 [ b ])
> (expr_list:REG_DEAD (reg/v:SI 77 [ a ])
> (nil))))
> (insn 31 30 18 2 (set (reg/v:SI 80 [ d ])
> (if_then_else:SI (eq (reg:CC 33 c)
> (const_int 0 [0]))
> (reg/v:SI 80 [ d ])
> (reg/v:SI 79 [ c ]))) func.c:5 983 {movf}
> (expr_list:REG_DEAD (reg/v:SI 79 [ c ])
> (expr_list:REG_DEAD (reg:CC 33 c)
> (nil))))
> -----------------------------------------------------------------------
>
> It doesn't seem to check the same conditional comparison instruction .So I wrote this to solve this problem, but I am not sure if this is the best way : )
>
> PS, the same conditional comparison instruction cannot be reduced with the latest version gcc with C-SKY because I just insert the “cse_cc†after “cse1â€, when I insert after “cse2â€, this problem can be solved very well.
Thanks, this is very helpful. I've verified this and I'm moving the
pass as you suggest, adding a more detailed comment to the source to
explain what the pass is for, and adding your test case to the
testsuite. This will be included when I resubmit the patches to address
other review comments too.
Jeff, does that adequately address your concerns about whether the pass
is useful?
-Sandra
^ permalink raw reply [flat|nested] 30+ messages in thread
* Re: [0/5] C-SKY port
2018-07-26 23:04 ` Joseph Myers
@ 2018-07-30 16:59 ` Sandra Loosemore
2018-08-01 14:28 ` 瞿仙淼
1 sibling, 0 replies; 30+ messages in thread
From: Sandra Loosemore @ 2018-07-30 16:59 UTC (permalink / raw)
To: Joseph Myers; +Cc: gcc-patches, Xianmiao Qu, Han Mao, Yunhai Shang
[-- Attachment #1: Type: text/plain, Size: 467 bytes --]
On 07/26/2018 05:04 PM, Joseph Myers wrote:
> Could you provide the proposed GCC website changes for the port
> (backends.html, readings.html, news item for index.html)? readings.html,
> in particular, would link to the ABI and ISA documentation, while
> backends.html gives summary information about the properties of both the
> architecture and the GCC port.
The attached patch is a bit drafty, but I think it at least as
placeholders for everything.
-Sandra
[-- Attachment #2: csky-doc.patch --]
[-- Type: text/x-patch, Size: 2807 bytes --]
? htdocs/backends.html.~1.78.~
? htdocs/index.html.~1.1090.~
? htdocs/readings.html.~1.296.~
Index: htdocs/backends.html
===================================================================
RCS file: /cvs/gcc/wwwdocs/htdocs/backends.html,v
retrieving revision 1.78
diff -u -r1.78 backends.html
--- htdocs/backends.html 28 Jul 2018 22:20:20 -0000 1.78
+++ htdocs/backends.html 30 Jul 2018 16:56:17 -0000
@@ -76,6 +76,7 @@
c6x | S CB gi
cr16 | L F C g s
cris | F B c gi s
+csky | b ia
epiphany | C gi s
fr30 | ?? FI B pb mg s
frv | ?? B b i s
Index: htdocs/index.html
===================================================================
RCS file: /cvs/gcc/wwwdocs/htdocs/index.html,v
retrieving revision 1.1090
diff -u -r1.1090 index.html
--- htdocs/index.html 30 Jul 2018 01:38:08 -0000 1.1090
+++ htdocs/index.html 30 Jul 2018 16:56:17 -0000
@@ -52,6 +52,11 @@
<h2 id="news">News</h2>
<dl>
+<dt><span>C-SKY support</span>
+ <span class="date">[2018-xx-xx]</span></dt>
+ <dd>GCC support for C-SKY V2 processors has been added. This back
+ end was contributed by C-SKY Microsystems and Mentor Graphics.</dd>
+
<dt><span><a href="https://gcc.gnu.org/wiki/cauldron2018">GNU Tools Cauldron 2018</a></span>
<span class="date">[2018-07-29]</span></dt>
<dd>Will be held in Manchester, September 7-9 2018</dd>
Index: htdocs/readings.html
===================================================================
RCS file: /cvs/gcc/wwwdocs/htdocs/readings.html,v
retrieving revision 1.296
diff -u -r1.296 readings.html
--- htdocs/readings.html 28 Jul 2018 22:20:20 -0000 1.296
+++ htdocs/readings.html 30 Jul 2018 16:56:17 -0000
@@ -120,6 +120,11 @@
<br /><a href="http://developer.axis.com/">Site with CPU documentation</a>
</li>
+ <li>C-SKY
+ <br />Manufacturer: C-SKY Microsystems
+ <br /><a href="https://github.com/c-sky/csky-doc">C-SKY Documentation</a>
+ </li>
+
<li>Epiphany
<br />Manufacturer: Adapteva
<br /><a href="http://www.adapteva.com/">Manufacturer's website</a> with
Index: htdocs/gcc-9/changes.html
===================================================================
RCS file: /cvs/gcc/wwwdocs/htdocs/gcc-9/changes.html,v
retrieving revision 1.15
diff -u -r1.15 changes.html
--- htdocs/gcc-9/changes.html 25 Jul 2018 21:54:59 -0000 1.15
+++ htdocs/gcc-9/changes.html 30 Jul 2018 16:56:17 -0000
@@ -138,6 +138,13 @@
<!-- <h3 id="avr">AVR</h3> -->
+<h3 id="csky">C-SKY</h3>
+<ul>
+ <li>
+ A new back end targeting C-SKY V2 processors has been contributed to GCC.
+ </li>
+</ul>
+
<!-- <h3 id="hsa">Heterogeneous Systems Architecture</h3> -->
<!-- <h3 id="x86">IA-32/x86-64</h3> -->
^ permalink raw reply [flat|nested] 30+ messages in thread
* Re: [0/5] C-SKY port
2018-07-26 23:04 ` Joseph Myers
2018-07-30 16:59 ` Sandra Loosemore
@ 2018-08-01 14:28 ` 瞿仙淼
1 sibling, 0 replies; 30+ messages in thread
From: 瞿仙淼 @ 2018-08-01 14:28 UTC (permalink / raw)
To: Joseph Myers; +Cc: Sandra Loosemore, gcc-patches, Han Mao, Yunhai Shang
>>> We expect that
>>> C-SKY will also be providing a public link to the processor and ABI
>>> documentation at some point.
>>
>> The ABI manual has been posted, but not the ISA documentation yet. (I'd guess
>> that when it does show up it will be in the same place, though.)
>>
>> https://github.com/c-sky/csky-doc
>
> Could you provide the proposed GCC website changes for the port
> (backends.html, readings.html, news item for index.html)? readings.html,
> in particular, would link to the ABI and ISA documentation, while
> backends.html gives summary information about the properties of both the
> architecture and the GCC port.
>
> --
> Joseph S. Myers
> joseph@codesourcery.com
Hi,
The ISA documentation is now available from https://github.com/c-sky/csky-doc
-Xianmiao
^ permalink raw reply [flat|nested] 30+ messages in thread
* Re: [2/5] C-SKY port: Backend implementation
2018-07-26 6:07 ` 瞿仙淼
2018-07-28 1:49 ` Sandra Loosemore
@ 2018-08-02 22:28 ` Jeff Law
2018-08-03 7:58 ` Yunhai
2018-08-03 16:26 ` Sandra Loosemore
1 sibling, 2 replies; 30+ messages in thread
From: Jeff Law @ 2018-08-02 22:28 UTC (permalink / raw)
To: 瞿仙淼; +Cc: Sandra Loosemore, gcc-patches, Yunhai Shang
On 07/26/2018 12:06 AM, 枚脛脧脡铆碌 wrote:
>
>> 脭脷 2018脛锚7脭脗25脠脮拢卢脡脧脦莽5:24拢卢Jeff Law <law@redhat.com> 脨麓碌脌拢潞
>>
>> On 07/24/2018 12:18 PM, Sandra Loosemore wrote:
>>> On 07/24/2018 09:45 AM, Jeff Law wrote:
>>>> On 07/23/2018 10:21 PM, Sandra Loosemore wrote:
>>>> I'm not a big fan of more awk code, but I'm not going to object to it :-)
>>>>
>>>> Why does the port have its own little pass for condition code
>>>> optimization (cse_cc)? What is it doing that can't be done with our
>>>> generic optimizers?
>>>
>>> This pass was included in the initial patch set we got from C-SKY, and
>>> as it didn't seem to break anything I left it in. Perhaps C-SKY can
>>> provide a testcase that demonstrates why it's still useful in the
>>> current version of GCC; otherwise we can remove this from the initial
>>> port submission and restore it later if some performance analysis shows
>>> it is still worthwhile.
>> FWIW it looks like we model CC setting on just a few insns, (add,
>> subtract) so I'd be surprised if this little mini pass found much. I'd
>> definitely like to hear from the csky authors here.
>>
>> Alternately, you could do add some instrumentation to flag when it
>> triggers, take a test or two that does, reduce it and we can then look
>> at the key RTL sequences and see what the pass is really doing.
>>
>
> I wrote a case to reproduce this problem on C-SKY. C code is as follows:
> -----------------------------------------------------------------------
> int e1, e2;
>
> void func (int a, int b, int c, int d, int f, int g)
> {
> e1 = a > b ? f : g;
> e2 = a > b ? c : d;
>
> return;
> }
> -----------------------------------------------------------------------
>
> compile to assembler with option 隆掳-O3 -S隆卤 :
> -----------------------------------------------------------------------
> func:
> cmplt a1, a0
> ld.w t1, (sp, 0)
> ld.w t0, (sp, 4)
> movt t0, t1
> cmplt a1, a0
> movt a3, a2
> lrw a1, e2
> lrw a2, e1
> st.w a3, (a1, 0)
> st.w t0, (a2, 0)
> rts
> -----------------------------------------------------------------------
> There is an extra 隆掳cmplt a1, a0" in the above code without cse_cc. This situation mainly occurs when a relatively short branch jump is converted into a conditional execution instruction. And the CSE pass can not reduce the same conditional comparison instruction . Here is the rtx sequence after 隆掳cse2隆卤 pass.
>
> -----------------------------------------------------------------------
> (insn 28 13 29 2 (set (reg:CC 33 c)
> (gt:CC (reg/v:SI 77 [ a ])
> (reg/v:SI 78 [ b ]))) func.c:5 1099 {*cmpgtsi}
> (nil))
> (insn 29 28 30 2 (set (reg/v:SI 82 [ g ])
> (if_then_else:SI (eq (reg:CC 33 c)
> (const_int 0 [0]))
> (reg/v:SI 82 [ g ])
> (reg/v:SI 81 [ f ]))) func.c:5 983 {movf}
> (expr_list:REG_DEAD (reg/v:SI 81 [ f ])
> (expr_list:REG_DEAD (reg:CC 33 c)
> (nil))))
> (insn 30 29 31 2 (set (reg:CC 33 c)
> (gt:CC (reg/v:SI 77 [ a ])
> (reg/v:SI 78 [ b ]))) func.c:5 1099 {*cmpgtsi}
> (expr_list:REG_DEAD (reg/v:SI 78 [ b ])
> (expr_list:REG_DEAD (reg/v:SI 77 [ a ])
> (nil))))
> (insn 31 30 18 2 (set (reg/v:SI 80 [ d ])
> (if_then_else:SI (eq (reg:CC 33 c)
> (const_int 0 [0]))
> (reg/v:SI 80 [ d ])
> (reg/v:SI 79 [ c ]))) func.c:5 983 {movf}
> (expr_list:REG_DEAD (reg/v:SI 79 [ c ])
> (expr_list:REG_DEAD (reg:CC 33 c)
> (nil))))
> -----------------------------------------------------------------------
>
> It doesn't seem to check the same conditional comparison instruction .So I wrote this to solve this problem, but I am not sure if this is the best way : )
>
> PS, the same conditional comparison instruction cannot be reduced with the latest version gcc with C-SKY because I just insert the 隆掳cse_cc隆卤 after 隆掳cse1隆卤, when I insert after 隆掳cse2隆卤, this problem can be solved very well.
I think the cse_cc pass is really just working around one or more bugs
in CSE and/or a backend bug. The RTL above clearly shows a common
subexpression that is not eliminated by CSE.
What CSE should be trying to do is changing the second and third
occurrences of (gt:CC (reg 77) (reg 78)) with (reg 33) which would
create nop-sets which would be subsequently deleted. I suspect you do
not have an insn which matches that nop set of the CC register. If you
fix that I suspect CSE will work better and eliminate the need for your
cse_cc pass.
jeff
^ permalink raw reply [flat|nested] 30+ messages in thread
* Re: [2/5] C-SKY port: Backend implementation
2018-07-28 1:49 ` Sandra Loosemore
@ 2018-08-02 22:33 ` Jeff Law
0 siblings, 0 replies; 30+ messages in thread
From: Jeff Law @ 2018-08-02 22:33 UTC (permalink / raw)
To: Sandra Loosemore, 瞿仙淼; +Cc: gcc-patches, Yunhai Shang
On 07/27/2018 07:49 PM, Sandra Loosemore wrote:
> On 07/26/2018 12:06 AM, 瞿仙淼 wrote:
>>
>> I wrote a case to reproduce this problem on C-SKY. C code is as follows:
>> -----------------------------------------------------------------------
>> int e1, e2;
>>
>> void func (int a, int b, int c, int d, int f, int g)
>> {
>> Â Â e1 = a > b ? f : g;
>> Â Â e2 = a > b ? c : d;
>>
>> Â Â return;
>> }
>> -----------------------------------------------------------------------
>>
>> compile to assembler with option “-O3 -S†:
>> -----------------------------------------------------------------------
>> func:
>> Â Â cmplt a1, a0
>>   ld.w t1, (sp, 0)
>>   ld.w t0, (sp, 4)
>>   movt t0, t1
>> Â Â cmplt a1, a0
>>   movt a3, a2
>> Â Â lrw a1, e2
>> Â Â lrw a2, e1
>>   st.w a3, (a1, 0)
>>   st.w t0, (a2, 0)
>> Â Â rts
>> -----------------------------------------------------------------------
>> There is an extra “cmplt a1, a0" in the above code without cse_cc.
>> This situation mainly occurs when a relatively short branch jump is
>> converted into a conditional execution instruction. And the CSE pass
>> can not reduce the same conditional comparison instruction . Here is
>> the rtx sequence after “cse2†pass.
>>
>> -----------------------------------------------------------------------
>> (insn 28 13 29 2 (set (reg:CC 33 c)
>> Â Â Â Â Â Â Â Â (gt:CC (reg/v:SI 77 [ a ])
>> Â Â Â Â Â Â Â Â Â Â Â Â (reg/v:SI 78 [ b ]))) func.c:5 1099 {*cmpgtsi}
>> Â Â Â Â Â (nil))
>> (insn 29 28 30 2 (set (reg/v:SI 82 [ g ])
>> Â Â Â Â Â Â Â Â (if_then_else:SI (eq (reg:CC 33 c)
>> Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â (const_int 0 [0]))
>> Â Â Â Â Â Â Â Â Â Â Â Â (reg/v:SI 82 [ g ])
>> Â Â Â Â Â Â Â Â Â Â Â Â (reg/v:SI 81 [ f ]))) func.c:5 983 {movf}
>> Â Â Â Â Â (expr_list:REG_DEAD (reg/v:SI 81 [ f ])
>> Â Â Â Â Â Â Â Â (expr_list:REG_DEAD (reg:CC 33 c)
>> Â Â Â Â Â Â Â Â Â Â Â Â (nil))))
>> (insn 30 29 31 2 (set (reg:CC 33 c)
>> Â Â Â Â Â Â Â Â (gt:CC (reg/v:SI 77 [ a ])
>> Â Â Â Â Â Â Â Â Â Â Â Â (reg/v:SI 78 [ b ]))) func.c:5 1099 {*cmpgtsi}
>> Â Â Â Â Â (expr_list:REG_DEAD (reg/v:SI 78 [ b ])
>> Â Â Â Â Â Â Â Â (expr_list:REG_DEAD (reg/v:SI 77 [ a ])
>> Â Â Â Â Â Â Â Â Â Â Â Â (nil))))
>> (insn 31 30 18 2 (set (reg/v:SI 80 [ d ])
>> Â Â Â Â Â Â Â Â (if_then_else:SI (eq (reg:CC 33 c)
>> Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â (const_int 0 [0]))
>> Â Â Â Â Â Â Â Â Â Â Â Â (reg/v:SI 80 [ d ])
>> Â Â Â Â Â Â Â Â Â Â Â Â (reg/v:SI 79 [ c ]))) func.c:5 983 {movf}
>> Â Â Â Â Â (expr_list:REG_DEAD (reg/v:SI 79 [ c ])
>> Â Â Â Â Â Â Â Â (expr_list:REG_DEAD (reg:CC 33 c)
>> Â Â Â Â Â Â Â Â Â Â Â Â (nil))))
>> -----------------------------------------------------------------------
>>
>> It doesn't seem to check the same conditional comparison instruction
>> .So I wrote this to solve this problem, but I am not sure if this is
>> the best way : )
>>
>> PS, the same conditional comparison instruction cannot be reduced with
>> the latest version gcc with C-SKY because I just insert the “cse_ccâ€
>> after “cse1â€, when I insert after “cse2â€, this problem can be solved
>> very well.
>
> Thanks, this is very helpful. I've verified this and I'm moving the
> pass as you suggest, adding a more detailed comment to the source to
> explain what the pass is for, and adding your test case to the
> testsuite. This will be included when I resubmit the patches to address
> other review comments too.
>
> Jeff, does that adequately address your concerns about whether the pass
> is useful?
I think the pass is papering over problems elsewhere (see my most other
reply on this thread). I do think it would be useful to take that code
and create a test based on it. I suspect you'll want to verify multiple
GT expressions prior to CSE2 and that after CSE2 you have a single GT
expression. Presumably it'd be in the csky specific test.
jeff
^ permalink raw reply [flat|nested] 30+ messages in thread
* Re: [2/5] C-SKY port: Backend implementation
2018-08-02 22:28 ` Jeff Law
@ 2018-08-03 7:58 ` Yunhai
2018-08-03 16:26 ` Sandra Loosemore
1 sibling, 0 replies; 30+ messages in thread
From: Yunhai @ 2018-08-03 7:58 UTC (permalink / raw)
To: Jeff Law; +Cc: 瞿仙淼, Sandra Loosemore, gcc-patches
> 在 2018年8月3日，06:27，Jeff Law <law@redhat.com> 写道：
>
> On 07/26/2018 12:06 AM, 瞿仙淼 wrote:
>>
>>> 在 2018年7月25日，上午5:24，Jeff Law <law@redhat.com> 写道：
>>>
>>> On 07/24/2018 12:18 PM, Sandra Loosemore wrote:
>>>> On 07/24/2018 09:45 AM, Jeff Law wrote:
>>>>> On 07/23/2018 10:21 PM, Sandra Loosemore wrote:
>>>>> I'm not a big fan of more awk code, but I'm not going to object to it :-)
>>>>>
>>>>> Why does the port have its own little pass for condition code
>>>>> optimization (cse_cc)? What is it doing that can't be done with our
>>>>> generic optimizers?
>>>>
>>>> This pass was included in the initial patch set we got from C-SKY, and
>>>> as it didn't seem to break anything I left it in. Perhaps C-SKY can
>>>> provide a testcase that demonstrates why it's still useful in the
>>>> current version of GCC; otherwise we can remove this from the initial
>>>> port submission and restore it later if some performance analysis shows
>>>> it is still worthwhile.
>>> FWIW it looks like we model CC setting on just a few insns, (add,
>>> subtract) so I'd be surprised if this little mini pass found much. I'd
>>> definitely like to hear from the csky authors here.
>>>
>>> Alternately, you could do add some instrumentation to flag when it
>>> triggers, take a test or two that does, reduce it and we can then look
>>> at the key RTL sequences and see what the pass is really doing.
>>>
>>
>> I wrote a case to reproduce this problem on C-SKY. C code is as follows:
>> -----------------------------------------------------------------------
>> int e1, e2;
>>
>> void func (int a, int b, int c, int d, int f, int g)
>> {
>> e1 = a > b ? f : g;
>> e2 = a > b ? c : d;
>>
>> return;
>> }
>> -----------------------------------------------------------------------
>>
>> compile to assembler with option “-O3 -S” :
>> -----------------------------------------------------------------------
>> func:
>> cmplt a1, a0
>> ld.w t1, (sp, 0)
>> ld.w t0, (sp, 4)
>> movt t0, t1
>> cmplt a1, a0
>> movt a3, a2
>> lrw a1, e2
>> lrw a2, e1
>> st.w a3, (a1, 0)
>> st.w t0, (a2, 0)
>> rts
>> -----------------------------------------------------------------------
>> There is an extra “cmplt a1, a0" in the above code without cse_cc. This situation mainly occurs when a relatively short branch jump is converted into a conditional execution instruction. And the CSE pass can not reduce the same conditional comparison instruction . Here is the rtx sequence after “cse2” pass.
>>
>> -----------------------------------------------------------------------
>> (insn 28 13 29 2 (set (reg:CC 33 c)
>> (gt:CC (reg/v:SI 77 [ a ])
>> (reg/v:SI 78 [ b ]))) func.c:5 1099 {*cmpgtsi}
>> (nil))
>> (insn 29 28 30 2 (set (reg/v:SI 82 [ g ])
>> (if_then_else:SI (eq (reg:CC 33 c)
>> (const_int 0 [0]))
>> (reg/v:SI 82 [ g ])
>> (reg/v:SI 81 [ f ]))) func.c:5 983 {movf}
>> (expr_list:REG_DEAD (reg/v:SI 81 [ f ])
>> (expr_list:REG_DEAD (reg:CC 33 c)
>> (nil))))
>> (insn 30 29 31 2 (set (reg:CC 33 c)
>> (gt:CC (reg/v:SI 77 [ a ])
>> (reg/v:SI 78 [ b ]))) func.c:5 1099 {*cmpgtsi}
>> (expr_list:REG_DEAD (reg/v:SI 78 [ b ])
>> (expr_list:REG_DEAD (reg/v:SI 77 [ a ])
>> (nil))))
>> (insn 31 30 18 2 (set (reg/v:SI 80 [ d ])
>> (if_then_else:SI (eq (reg:CC 33 c)
>> (const_int 0 [0]))
>> (reg/v:SI 80 [ d ])
>> (reg/v:SI 79 [ c ]))) func.c:5 983 {movf}
>> (expr_list:REG_DEAD (reg/v:SI 79 [ c ])
>> (expr_list:REG_DEAD (reg:CC 33 c)
>> (nil))))
>> -----------------------------------------------------------------------
>>
>> It doesn't seem to check the same conditional comparison instruction .So I wrote this to solve this problem, but I am not sure if this is the best way : )
>>
>> PS, the same conditional comparison instruction cannot be reduced with the latest version gcc with C-SKY because I just insert the “cse_cc” after “cse1”, when I insert after “cse2”, this problem can be solved very well.
> I think the cse_cc pass is really just working around one or more bugs
> in CSE and/or a backend bug. The RTL above clearly shows a common
> subexpression that is not eliminated by CSE.
>
> What CSE should be trying to do is changing the second and third
> occurrences of (gt:CC (reg 77) (reg 78)) with (reg 33) which would
> create nop-sets which would be subsequently deleted. I suspect you do
> not have an insn which matches that nop set of the CC register. If you
> fix that I suspect CSE will work better and eliminate the need for your
> cse_cc pass.
Thanks you for your suggestions, we will try this method.
>
> jeff
^ permalink raw reply [flat|nested] 30+ messages in thread
* Re: [2/5] C-SKY port: Backend implementation
2018-08-02 22:28 ` Jeff Law
2018-08-03 7:58 ` Yunhai
@ 2018-08-03 16:26 ` Sandra Loosemore
1 sibling, 0 replies; 30+ messages in thread
From: Sandra Loosemore @ 2018-08-03 16:26 UTC (permalink / raw)
To: Jeff Law, 瞿仙淼; +Cc: gcc-patches, Yunhai Shang
On 08/02/2018 04:27 PM, Jeff Law wrote:
> I think the cse_cc pass is really just working around one or more bugs
> in CSE and/or a backend bug. The RTL above clearly shows a common
> subexpression that is not eliminated by CSE.
>
> What CSE should be trying to do is changing the second and third
> occurrences of (gt:CC (reg 77) (reg 78)) with (reg 33) which would
> create nop-sets which would be subsequently deleted. I suspect you do
> not have an insn which matches that nop set of the CC register. If you
> fix that I suspect CSE will work better and eliminate the need for your
> cse_cc pass.
Thanks for the tip about that! I hacked things up to do as you suggest
and it appears to work. I'll post a new patch set once I've had time
for more testing, probably Monday-ish.
-Sandra
^ permalink raw reply [flat|nested] 30+ messages in thread
end of thread, other threads:[~2018-08-03 16:26 UTC | newest]
Thread overview: 30+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2018-07-24 4:17 [0/5] C-SKY port Sandra Loosemore
2018-07-24 4:20 ` [1/5] C-SKY port: Configury Sandra Loosemore
2018-07-24 15:09 ` Jeff Law
2018-07-24 4:21 ` [2/5] C-SKY port: Backend implementation Sandra Loosemore
2018-07-24 15:45 ` Jeff Law
2018-07-24 18:19 ` Sandra Loosemore
2018-07-24 21:25 ` Jeff Law
2018-07-25 0:17 ` Sandra Loosemore
2018-07-25 4:50 ` Jeff Law
2018-07-25 13:17 ` Paul Koning
2018-07-25 14:54 ` Sandra Loosemore
2018-07-26 6:07 ` 瞿仙淼
2018-07-28 1:49 ` Sandra Loosemore
2018-08-02 22:33 ` Jeff Law
2018-08-02 22:28 ` Jeff Law
2018-08-03 7:58 ` Yunhai
2018-08-03 16:26 ` Sandra Loosemore
2018-07-24 4:23 ` [3/5] C-SKY port: Documentation Sandra Loosemore
2018-07-24 15:10 ` Jeff Law
2018-07-24 4:25 ` [4/5] C-SKY port: Testsuite Sandra Loosemore
2018-07-24 15:10 ` Jeff Law
2018-07-24 4:26 ` [5/5] C-SKY port: libgcc Sandra Loosemore
2018-07-24 15:12 ` Jeff Law
2018-07-24 18:10 ` Segher Boessenkool
2018-07-24 18:19 ` Sandra Loosemore
2018-07-24 19:46 ` Segher Boessenkool
2018-07-24 15:23 ` [0/5] C-SKY port Sandra Loosemore
2018-07-26 23:04 ` Joseph Myers
2018-07-30 16:59 ` Sandra Loosemore
2018-08-01 14:28 ` 瞿仙淼
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).