NetCube-Systems-Austria/kumquat-buildroot
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
abc292a625
kumquat-buildroot/package/gcc/4.9.4/870-xtensa-add-mauto-litpools-option.patch

288 lines
10 KiB
Diff
Raw Normal View History

gcc: backport mauto-litpools xtensa option With support from assembler this option allows compiling huge functions, where single literal pool at the beginning of a function may not be reachable by L32R instructions at its end. Currently assembler --auto-litpools option cannot deal with literals used from multiple locations separated by more than 256 KBytes of code. Don't turn constants into literals, instead use MOVI instruction to load them into registers and let the assembler turn them into literals as necessary. Signed-off-by: Max Filippov <jcmvbkbc@gmail.com> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
2015-08-13 00:20:01 +02:00
From 6d852ffb43b111a39162135c95249e749c4e285b Mon Sep 17 00:00:00 2001
From: Max Filippov <jcmvbkbc@gmail.com>
Date: Thu, 6 Aug 2015 01:16:02 +0300
Subject: [PATCH] xtensa: add -mauto-litpools option
With support from assembler this option allows compiling huge functions,
where single literal pool at the beginning of a function may not be
reachable by L32R instructions at its end.
Currently assembler --auto-litpools option cannot deal with literals
used from multiple locations separated by more than 256 KBytes of code.
Don't turn constants into literals, instead use MOVI instruction to load
them into registers and let the assembler turn them into literals as
necessary.
2015-08-12 Max Filippov <jcmvbkbc@gmail.com>
gcc/
* config/xtensa/constraints.md (define_constraint "Y"): New
constraint.
* config/xtensa/elf.h (ASM_SPEC): Add m(no-)auto-litpools.
* config/xtensa/linux.h (ASM_SPEC): Likewise.
* config/xtensa/predicates.md (move_operand): Match constants
and symbols in the presence of TARGET_AUTO_LITPOOLS.
* config/xtensa/xtensa.c (xtensa_valid_move): Don't allow
immediate references to TLS data.
(xtensa_emit_move_sequence): Don't force constants to memory in
the presence of TARGET_AUTO_LITPOOLS.
(print_operand): Add 'y' format, same as default, but capable of
printing SF mode constants as well.
* config/xtensa/xtensa.md (movsi_internal, movhi_internal)
(movsf_internal): Add movi pattern that loads literal.
(movsf, movdf): Don't force constants to memory in the presence
of TARGET_AUTO_LITPOOLS.
(movdf_internal): Add 'Y' constraint.
* config/xtensa/xtensa.opt (mauto-litpools): New option.
Signed-off-by: Max Filippov <jcmvbkbc@gmail.com>
---
Backported from: r226828
Changes to ChangeLogs and documentation are dropped.
gcc/config/xtensa/constraints.md | 5 +++++
gcc/config/xtensa/elf.h | 4 +++-
gcc/config/xtensa/linux.h | 4 +++-
gcc/config/xtensa/predicates.md | 3 ++-
gcc/config/xtensa/xtensa.c | 19 ++++++++++++++++++-
gcc/config/xtensa/xtensa.md | 35 +++++++++++++++++++----------------
gcc/config/xtensa/xtensa.opt | 4 ++++
7 files changed, 54 insertions(+), 20 deletions(-)
gcc: bump 4.9 series to 4.9.4 Two patches are removed, as they have been upstreamed: - 130-fix_build_with_gcc-6.patch (svn commit 233721, Git commit 8c3fa311caa86f61b4e28d1563d1110b44340fb2) - 920-libgcc-remove-unistd-header.patch (svn commit 226092, Git commit e940d7953f06af11d09229a29ecbcc1ba25b378d) All other patches have simply been refreshed, with no manual edit needed. A build+runtime test has been done with an ARM, Cortex-A8, EABIhf, musl configuration, booted under Qemu. Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
2016-08-03 23:48:24 +02:00
Index: b/gcc/config/xtensa/constraints.md
===================================================================
gcc: backport mauto-litpools xtensa option With support from assembler this option allows compiling huge functions, where single literal pool at the beginning of a function may not be reachable by L32R instructions at its end. Currently assembler --auto-litpools option cannot deal with literals used from multiple locations separated by more than 256 KBytes of code. Don't turn constants into literals, instead use MOVI instruction to load them into registers and let the assembler turn them into literals as necessary. Signed-off-by: Max Filippov <jcmvbkbc@gmail.com> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
2015-08-13 00:20:01 +02:00
--- a/gcc/config/xtensa/constraints.md
+++ b/gcc/config/xtensa/constraints.md
@@ -111,6 +111,11 @@
(and (match_code "const_int")
(match_test "xtensa_mask_immediate (ival)")))
+(define_constraint "Y"
+ "A constant that can be used in relaxed MOVI instructions."
+ (and (match_code "const_int,const_double,const,symbol_ref,label_ref")
+ (match_test "TARGET_AUTO_LITPOOLS")))
+
;; Memory constraints. Do not use define_memory_constraint here. Doing so
;; causes reload to force some constants into the constant pool, but since
;; the Xtensa constant pool can only be accessed with L32R instructions, it
gcc: bump 4.9 series to 4.9.4 Two patches are removed, as they have been upstreamed: - 130-fix_build_with_gcc-6.patch (svn commit 233721, Git commit 8c3fa311caa86f61b4e28d1563d1110b44340fb2) - 920-libgcc-remove-unistd-header.patch (svn commit 226092, Git commit e940d7953f06af11d09229a29ecbcc1ba25b378d) All other patches have simply been refreshed, with no manual edit needed. A build+runtime test has been done with an ARM, Cortex-A8, EABIhf, musl configuration, booted under Qemu. Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
2016-08-03 23:48:24 +02:00
Index: b/gcc/config/xtensa/elf.h
===================================================================
gcc: backport mauto-litpools xtensa option With support from assembler this option allows compiling huge functions, where single literal pool at the beginning of a function may not be reachable by L32R instructions at its end. Currently assembler --auto-litpools option cannot deal with literals used from multiple locations separated by more than 256 KBytes of code. Don't turn constants into literals, instead use MOVI instruction to load them into registers and let the assembler turn them into literals as necessary. Signed-off-by: Max Filippov <jcmvbkbc@gmail.com> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
2015-08-13 00:20:01 +02:00
--- a/gcc/config/xtensa/elf.h
+++ b/gcc/config/xtensa/elf.h
gcc: bump 4.9 series to 4.9.4 Two patches are removed, as they have been upstreamed: - 130-fix_build_with_gcc-6.patch (svn commit 233721, Git commit 8c3fa311caa86f61b4e28d1563d1110b44340fb2) - 920-libgcc-remove-unistd-header.patch (svn commit 226092, Git commit e940d7953f06af11d09229a29ecbcc1ba25b378d) All other patches have simply been refreshed, with no manual edit needed. A build+runtime test has been done with an ARM, Cortex-A8, EABIhf, musl configuration, booted under Qemu. Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
2016-08-03 23:48:24 +02:00
@@ -48,7 +48,9 @@
gcc: backport mauto-litpools xtensa option With support from assembler this option allows compiling huge functions, where single literal pool at the beginning of a function may not be reachable by L32R instructions at its end. Currently assembler --auto-litpools option cannot deal with literals used from multiple locations separated by more than 256 KBytes of code. Don't turn constants into literals, instead use MOVI instruction to load them into registers and let the assembler turn them into literals as necessary. Signed-off-by: Max Filippov <jcmvbkbc@gmail.com> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
2015-08-13 00:20:01 +02:00
%{mtarget-align:--target-align} \
%{mno-target-align:--no-target-align} \
%{mlongcalls:--longcalls} \
- %{mno-longcalls:--no-longcalls}"
+ %{mno-longcalls:--no-longcalls} \
+ %{mauto-litpools:--auto-litpools} \
+ %{mno-auto-litpools:--no-auto-litpools}"
#undef LIB_SPEC
#define LIB_SPEC "-lc -lsim -lc -lhandlers-sim -lhal"
gcc: bump 4.9 series to 4.9.4 Two patches are removed, as they have been upstreamed: - 130-fix_build_with_gcc-6.patch (svn commit 233721, Git commit 8c3fa311caa86f61b4e28d1563d1110b44340fb2) - 920-libgcc-remove-unistd-header.patch (svn commit 226092, Git commit e940d7953f06af11d09229a29ecbcc1ba25b378d) All other patches have simply been refreshed, with no manual edit needed. A build+runtime test has been done with an ARM, Cortex-A8, EABIhf, musl configuration, booted under Qemu. Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
2016-08-03 23:48:24 +02:00
Index: b/gcc/config/xtensa/linux.h
===================================================================
gcc: backport mauto-litpools xtensa option With support from assembler this option allows compiling huge functions, where single literal pool at the beginning of a function may not be reachable by L32R instructions at its end. Currently assembler --auto-litpools option cannot deal with literals used from multiple locations separated by more than 256 KBytes of code. Don't turn constants into literals, instead use MOVI instruction to load them into registers and let the assembler turn them into literals as necessary. Signed-off-by: Max Filippov <jcmvbkbc@gmail.com> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
2015-08-13 00:20:01 +02:00
--- a/gcc/config/xtensa/linux.h
+++ b/gcc/config/xtensa/linux.h
gcc: bump 4.9 series to 4.9.4 Two patches are removed, as they have been upstreamed: - 130-fix_build_with_gcc-6.patch (svn commit 233721, Git commit 8c3fa311caa86f61b4e28d1563d1110b44340fb2) - 920-libgcc-remove-unistd-header.patch (svn commit 226092, Git commit e940d7953f06af11d09229a29ecbcc1ba25b378d) All other patches have simply been refreshed, with no manual edit needed. A build+runtime test has been done with an ARM, Cortex-A8, EABIhf, musl configuration, booted under Qemu. Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
2016-08-03 23:48:24 +02:00
@@ -42,7 +42,9 @@
gcc: backport mauto-litpools xtensa option With support from assembler this option allows compiling huge functions, where single literal pool at the beginning of a function may not be reachable by L32R instructions at its end. Currently assembler --auto-litpools option cannot deal with literals used from multiple locations separated by more than 256 KBytes of code. Don't turn constants into literals, instead use MOVI instruction to load them into registers and let the assembler turn them into literals as necessary. Signed-off-by: Max Filippov <jcmvbkbc@gmail.com> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
2015-08-13 00:20:01 +02:00
%{mtarget-align:--target-align} \
%{mno-target-align:--no-target-align} \
%{mlongcalls:--longcalls} \
- %{mno-longcalls:--no-longcalls}"
+ %{mno-longcalls:--no-longcalls} \
+ %{mauto-litpools:--auto-litpools} \
+ %{mno-auto-litpools:--no-auto-litpools}"
#define GLIBC_DYNAMIC_LINKER "/lib/ld.so.1"
gcc: bump 4.9 series to 4.9.4 Two patches are removed, as they have been upstreamed: - 130-fix_build_with_gcc-6.patch (svn commit 233721, Git commit 8c3fa311caa86f61b4e28d1563d1110b44340fb2) - 920-libgcc-remove-unistd-header.patch (svn commit 226092, Git commit e940d7953f06af11d09229a29ecbcc1ba25b378d) All other patches have simply been refreshed, with no manual edit needed. A build+runtime test has been done with an ARM, Cortex-A8, EABIhf, musl configuration, booted under Qemu. Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
2016-08-03 23:48:24 +02:00
Index: b/gcc/config/xtensa/predicates.md
===================================================================
gcc: backport mauto-litpools xtensa option With support from assembler this option allows compiling huge functions, where single literal pool at the beginning of a function may not be reachable by L32R instructions at its end. Currently assembler --auto-litpools option cannot deal with literals used from multiple locations separated by more than 256 KBytes of code. Don't turn constants into literals, instead use MOVI instruction to load them into registers and let the assembler turn them into literals as necessary. Signed-off-by: Max Filippov <jcmvbkbc@gmail.com> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
2015-08-13 00:20:01 +02:00
--- a/gcc/config/xtensa/predicates.md
+++ b/gcc/config/xtensa/predicates.md
@@ -142,7 +142,8 @@
(match_test "GET_MODE_CLASS (mode) == MODE_INT
&& xtensa_simm12b (INTVAL (op))"))
(and (match_code "const_int,const_double,const,symbol_ref,label_ref")
- (match_test "TARGET_CONST16 && CONSTANT_P (op)
+ (match_test "(TARGET_CONST16 || TARGET_AUTO_LITPOOLS)
+ && CONSTANT_P (op)
&& GET_MODE_SIZE (mode) % UNITS_PER_WORD == 0")))))
;; Accept the floating point constant 1 in the appropriate mode.
gcc: bump 4.9 series to 4.9.4 Two patches are removed, as they have been upstreamed: - 130-fix_build_with_gcc-6.patch (svn commit 233721, Git commit 8c3fa311caa86f61b4e28d1563d1110b44340fb2) - 920-libgcc-remove-unistd-header.patch (svn commit 226092, Git commit e940d7953f06af11d09229a29ecbcc1ba25b378d) All other patches have simply been refreshed, with no manual edit needed. A build+runtime test has been done with an ARM, Cortex-A8, EABIhf, musl configuration, booted under Qemu. Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
2016-08-03 23:48:24 +02:00
Index: b/gcc/config/xtensa/xtensa.c
===================================================================
gcc: backport mauto-litpools xtensa option With support from assembler this option allows compiling huge functions, where single literal pool at the beginning of a function may not be reachable by L32R instructions at its end. Currently assembler --auto-litpools option cannot deal with literals used from multiple locations separated by more than 256 KBytes of code. Don't turn constants into literals, instead use MOVI instruction to load them into registers and let the assembler turn them into literals as necessary. Signed-off-by: Max Filippov <jcmvbkbc@gmail.com> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
2015-08-13 00:20:01 +02:00
--- a/gcc/config/xtensa/xtensa.c
+++ b/gcc/config/xtensa/xtensa.c
gcc: bump 4.9 series to 4.9.4 Two patches are removed, as they have been upstreamed: - 130-fix_build_with_gcc-6.patch (svn commit 233721, Git commit 8c3fa311caa86f61b4e28d1563d1110b44340fb2) - 920-libgcc-remove-unistd-header.patch (svn commit 226092, Git commit e940d7953f06af11d09229a29ecbcc1ba25b378d) All other patches have simply been refreshed, with no manual edit needed. A build+runtime test has been done with an ARM, Cortex-A8, EABIhf, musl configuration, booted under Qemu. Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
2016-08-03 23:48:24 +02:00
@@ -477,6 +477,9 @@
gcc: backport mauto-litpools xtensa option With support from assembler this option allows compiling huge functions, where single literal pool at the beginning of a function may not be reachable by L32R instructions at its end. Currently assembler --auto-litpools option cannot deal with literals used from multiple locations separated by more than 256 KBytes of code. Don't turn constants into literals, instead use MOVI instruction to load them into registers and let the assembler turn them into literals as necessary. Signed-off-by: Max Filippov <jcmvbkbc@gmail.com> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
2015-08-13 00:20:01 +02:00
{
int dst_regnum = xt_true_regnum (operands[0]);
+ if (xtensa_tls_referenced_p (operands[1]))
+ return FALSE;
+
/* The stack pointer can only be assigned with a MOVSP opcode. */
if (dst_regnum == STACK_POINTER_REGNUM)
gcc: bump 4.9 series to 4.9.4 Two patches are removed, as they have been upstreamed: - 130-fix_build_with_gcc-6.patch (svn commit 233721, Git commit 8c3fa311caa86f61b4e28d1563d1110b44340fb2) - 920-libgcc-remove-unistd-header.patch (svn commit 226092, Git commit e940d7953f06af11d09229a29ecbcc1ba25b378d) All other patches have simply been refreshed, with no manual edit needed. A build+runtime test has been done with an ARM, Cortex-A8, EABIhf, musl configuration, booted under Qemu. Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
2016-08-03 23:48:24 +02:00
return (mode == SImode
@@ -1044,7 +1047,7 @@
gcc: backport mauto-litpools xtensa option With support from assembler this option allows compiling huge functions, where single literal pool at the beginning of a function may not be reachable by L32R instructions at its end. Currently assembler --auto-litpools option cannot deal with literals used from multiple locations separated by more than 256 KBytes of code. Don't turn constants into literals, instead use MOVI instruction to load them into registers and let the assembler turn them into literals as necessary. Signed-off-by: Max Filippov <jcmvbkbc@gmail.com> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
2015-08-13 00:20:01 +02:00
return 1;
}
- if (! TARGET_CONST16)
+ if (! TARGET_AUTO_LITPOOLS && ! TARGET_CONST16)
{
src = force_const_mem (SImode, src);
operands[1] = src;
gcc: bump 4.9 series to 4.9.4 Two patches are removed, as they have been upstreamed: - 130-fix_build_with_gcc-6.patch (svn commit 233721, Git commit 8c3fa311caa86f61b4e28d1563d1110b44340fb2) - 920-libgcc-remove-unistd-header.patch (svn commit 226092, Git commit e940d7953f06af11d09229a29ecbcc1ba25b378d) All other patches have simply been refreshed, with no manual edit needed. A build+runtime test has been done with an ARM, Cortex-A8, EABIhf, musl configuration, booted under Qemu. Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
2016-08-03 23:48:24 +02:00
@@ -2428,6 +2431,20 @@
gcc: backport mauto-litpools xtensa option With support from assembler this option allows compiling huge functions, where single literal pool at the beginning of a function may not be reachable by L32R instructions at its end. Currently assembler --auto-litpools option cannot deal with literals used from multiple locations separated by more than 256 KBytes of code. Don't turn constants into literals, instead use MOVI instruction to load them into registers and let the assembler turn them into literals as necessary. Signed-off-by: Max Filippov <jcmvbkbc@gmail.com> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
2015-08-13 00:20:01 +02:00
}
break;
+ case 'y':
+ if (GET_CODE (x) == CONST_DOUBLE &&
+ GET_MODE (x) == SFmode)
+ {
+ REAL_VALUE_TYPE r;
+ long l;
+ REAL_VALUE_FROM_CONST_DOUBLE (r, x);
+ REAL_VALUE_TO_TARGET_SINGLE (r, l);
+ fprintf (file, "0x%08lx", l);
+ break;
+ }
+
+ /* fall through */
+
default:
if (GET_CODE (x) == REG || GET_CODE (x) == SUBREG)
fprintf (file, "%s", reg_names[xt_true_regnum (x)]);
gcc: bump 4.9 series to 4.9.4 Two patches are removed, as they have been upstreamed: - 130-fix_build_with_gcc-6.patch (svn commit 233721, Git commit 8c3fa311caa86f61b4e28d1563d1110b44340fb2) - 920-libgcc-remove-unistd-header.patch (svn commit 226092, Git commit e940d7953f06af11d09229a29ecbcc1ba25b378d) All other patches have simply been refreshed, with no manual edit needed. A build+runtime test has been done with an ARM, Cortex-A8, EABIhf, musl configuration, booted under Qemu. Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
2016-08-03 23:48:24 +02:00
Index: b/gcc/config/xtensa/xtensa.md
===================================================================
gcc: backport mauto-litpools xtensa option With support from assembler this option allows compiling huge functions, where single literal pool at the beginning of a function may not be reachable by L32R instructions at its end. Currently assembler --auto-litpools option cannot deal with literals used from multiple locations separated by more than 256 KBytes of code. Don't turn constants into literals, instead use MOVI instruction to load them into registers and let the assembler turn them into literals as necessary. Signed-off-by: Max Filippov <jcmvbkbc@gmail.com> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
2015-08-13 00:20:01 +02:00
--- a/gcc/config/xtensa/xtensa.md
+++ b/gcc/config/xtensa/xtensa.md
gcc: bump 4.9 series to 4.9.4 Two patches are removed, as they have been upstreamed: - 130-fix_build_with_gcc-6.patch (svn commit 233721, Git commit 8c3fa311caa86f61b4e28d1563d1110b44340fb2) - 920-libgcc-remove-unistd-header.patch (svn commit 226092, Git commit e940d7953f06af11d09229a29ecbcc1ba25b378d) All other patches have simply been refreshed, with no manual edit needed. A build+runtime test has been done with an ARM, Cortex-A8, EABIhf, musl configuration, booted under Qemu. Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
2016-08-03 23:48:24 +02:00
@@ -799,8 +799,8 @@
gcc: backport mauto-litpools xtensa option With support from assembler this option allows compiling huge functions, where single literal pool at the beginning of a function may not be reachable by L32R instructions at its end. Currently assembler --auto-litpools option cannot deal with literals used from multiple locations separated by more than 256 KBytes of code. Don't turn constants into literals, instead use MOVI instruction to load them into registers and let the assembler turn them into literals as necessary. Signed-off-by: Max Filippov <jcmvbkbc@gmail.com> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
2015-08-13 00:20:01 +02:00
})
(define_insn "movsi_internal"
- [(set (match_operand:SI 0 "nonimmed_operand" "=D,D,D,D,R,R,a,q,a,W,a,a,U,*a,*A")
- (match_operand:SI 1 "move_operand" "M,D,d,R,D,d,r,r,I,i,T,U,r,*A,*r"))]
+ [(set (match_operand:SI 0 "nonimmed_operand" "=D,D,D,D,R,R,a,q,a,a,W,a,a,U,*a,*A")
+ (match_operand:SI 1 "move_operand" "M,D,d,R,D,d,r,r,I,Y,i,T,U,r,*A,*r"))]
"xtensa_valid_move (SImode, operands)"
"@
movi.n\t%0, %x1
gcc: bump 4.9 series to 4.9.4 Two patches are removed, as they have been upstreamed: - 130-fix_build_with_gcc-6.patch (svn commit 233721, Git commit 8c3fa311caa86f61b4e28d1563d1110b44340fb2) - 920-libgcc-remove-unistd-header.patch (svn commit 226092, Git commit e940d7953f06af11d09229a29ecbcc1ba25b378d) All other patches have simply been refreshed, with no manual edit needed. A build+runtime test has been done with an ARM, Cortex-A8, EABIhf, musl configuration, booted under Qemu. Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
2016-08-03 23:48:24 +02:00
@@ -812,15 +812,16 @@
gcc: backport mauto-litpools xtensa option With support from assembler this option allows compiling huge functions, where single literal pool at the beginning of a function may not be reachable by L32R instructions at its end. Currently assembler --auto-litpools option cannot deal with literals used from multiple locations separated by more than 256 KBytes of code. Don't turn constants into literals, instead use MOVI instruction to load them into registers and let the assembler turn them into literals as necessary. Signed-off-by: Max Filippov <jcmvbkbc@gmail.com> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
2015-08-13 00:20:01 +02:00
mov\t%0, %1
movsp\t%0, %1
movi\t%0, %x1
+ movi\t%0, %1
const16\t%0, %t1\;const16\t%0, %b1
%v1l32r\t%0, %1
%v1l32i\t%0, %1
%v0s32i\t%1, %0
rsr\t%0, ACCLO
wsr\t%1, ACCLO"
- [(set_attr "type" "move,move,move,load,store,store,move,move,move,move,load,load,store,rsr,wsr")
+ [(set_attr "type" "move,move,move,load,store,store,move,move,move,move,move,load,load,store,rsr,wsr")
(set_attr "mode" "SI")
- (set_attr "length" "2,2,2,2,2,2,3,3,3,6,3,3,3,3,3")])
+ (set_attr "length" "2,2,2,2,2,2,3,3,3,3,6,3,3,3,3,3")])
;; 16-bit Integer moves
gcc: bump 4.9 series to 4.9.4 Two patches are removed, as they have been upstreamed: - 130-fix_build_with_gcc-6.patch (svn commit 233721, Git commit 8c3fa311caa86f61b4e28d1563d1110b44340fb2) - 920-libgcc-remove-unistd-header.patch (svn commit 226092, Git commit e940d7953f06af11d09229a29ecbcc1ba25b378d) All other patches have simply been refreshed, with no manual edit needed. A build+runtime test has been done with an ARM, Cortex-A8, EABIhf, musl configuration, booted under Qemu. Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
2016-08-03 23:48:24 +02:00
@@ -834,21 +835,22 @@
gcc: backport mauto-litpools xtensa option With support from assembler this option allows compiling huge functions, where single literal pool at the beginning of a function may not be reachable by L32R instructions at its end. Currently assembler --auto-litpools option cannot deal with literals used from multiple locations separated by more than 256 KBytes of code. Don't turn constants into literals, instead use MOVI instruction to load them into registers and let the assembler turn them into literals as necessary. Signed-off-by: Max Filippov <jcmvbkbc@gmail.com> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
2015-08-13 00:20:01 +02:00
})
(define_insn "movhi_internal"
- [(set (match_operand:HI 0 "nonimmed_operand" "=D,D,a,a,a,U,*a,*A")
- (match_operand:HI 1 "move_operand" "M,d,r,I,U,r,*A,*r"))]
+ [(set (match_operand:HI 0 "nonimmed_operand" "=D,D,a,a,a,a,U,*a,*A")
+ (match_operand:HI 1 "move_operand" "M,d,r,I,Y,U,r,*A,*r"))]
"xtensa_valid_move (HImode, operands)"
"@
movi.n\t%0, %x1
mov.n\t%0, %1
mov\t%0, %1
movi\t%0, %x1
+ movi\t%0, %1
%v1l16ui\t%0, %1
%v0s16i\t%1, %0
rsr\t%0, ACCLO
wsr\t%1, ACCLO"
- [(set_attr "type" "move,move,move,move,load,store,rsr,wsr")
+ [(set_attr "type" "move,move,move,move,move,load,store,rsr,wsr")
(set_attr "mode" "HI")
- (set_attr "length" "2,2,3,3,3,3,3,3")])
+ (set_attr "length" "2,2,3,3,3,3,3,3,3")])
;; 8-bit Integer moves
gcc: bump 4.9 series to 4.9.4 Two patches are removed, as they have been upstreamed: - 130-fix_build_with_gcc-6.patch (svn commit 233721, Git commit 8c3fa311caa86f61b4e28d1563d1110b44340fb2) - 920-libgcc-remove-unistd-header.patch (svn commit 226092, Git commit e940d7953f06af11d09229a29ecbcc1ba25b378d) All other patches have simply been refreshed, with no manual edit needed. A build+runtime test has been done with an ARM, Cortex-A8, EABIhf, musl configuration, booted under Qemu. Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
2016-08-03 23:48:24 +02:00
@@ -919,7 +921,7 @@
gcc: backport mauto-litpools xtensa option With support from assembler this option allows compiling huge functions, where single literal pool at the beginning of a function may not be reachable by L32R instructions at its end. Currently assembler --auto-litpools option cannot deal with literals used from multiple locations separated by more than 256 KBytes of code. Don't turn constants into literals, instead use MOVI instruction to load them into registers and let the assembler turn them into literals as necessary. Signed-off-by: Max Filippov <jcmvbkbc@gmail.com> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
2015-08-13 00:20:01 +02:00
(match_operand:SF 1 "general_operand" ""))]
""
{
- if (!TARGET_CONST16 && CONSTANT_P (operands[1]))
+ if (!TARGET_CONST16 && !TARGET_AUTO_LITPOOLS && CONSTANT_P (operands[1]))
operands[1] = force_const_mem (SFmode, operands[1]);
if ((!register_operand (operands[0], SFmode)
gcc: bump 4.9 series to 4.9.4 Two patches are removed, as they have been upstreamed: - 130-fix_build_with_gcc-6.patch (svn commit 233721, Git commit 8c3fa311caa86f61b4e28d1563d1110b44340fb2) - 920-libgcc-remove-unistd-header.patch (svn commit 226092, Git commit e940d7953f06af11d09229a29ecbcc1ba25b378d) All other patches have simply been refreshed, with no manual edit needed. A build+runtime test has been done with an ARM, Cortex-A8, EABIhf, musl configuration, booted under Qemu. Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
2016-08-03 23:48:24 +02:00
@@ -934,8 +936,8 @@
gcc: backport mauto-litpools xtensa option With support from assembler this option allows compiling huge functions, where single literal pool at the beginning of a function may not be reachable by L32R instructions at its end. Currently assembler --auto-litpools option cannot deal with literals used from multiple locations separated by more than 256 KBytes of code. Don't turn constants into literals, instead use MOVI instruction to load them into registers and let the assembler turn them into literals as necessary. Signed-off-by: Max Filippov <jcmvbkbc@gmail.com> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
2015-08-13 00:20:01 +02:00
})
(define_insn "movsf_internal"
- [(set (match_operand:SF 0 "nonimmed_operand" "=f,f,U,D,D,R,a,f,a,W,a,a,U")
- (match_operand:SF 1 "move_operand" "f,U,f,d,R,d,r,r,f,iF,T,U,r"))]
+ [(set (match_operand:SF 0 "nonimmed_operand" "=f,f,U,D,D,R,a,f,a,a,W,a,a,U")
+ (match_operand:SF 1 "move_operand" "f,U,f,d,R,d,r,r,f,Y,iF,T,U,r"))]
"((register_operand (operands[0], SFmode)
|| register_operand (operands[1], SFmode))
&& !(FP_REG_P (xt_true_regnum (operands[0]))
gcc: bump 4.9 series to 4.9.4 Two patches are removed, as they have been upstreamed: - 130-fix_build_with_gcc-6.patch (svn commit 233721, Git commit 8c3fa311caa86f61b4e28d1563d1110b44340fb2) - 920-libgcc-remove-unistd-header.patch (svn commit 226092, Git commit e940d7953f06af11d09229a29ecbcc1ba25b378d) All other patches have simply been refreshed, with no manual edit needed. A build+runtime test has been done with an ARM, Cortex-A8, EABIhf, musl configuration, booted under Qemu. Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
2016-08-03 23:48:24 +02:00
@@ -950,13 +952,14 @@
gcc: backport mauto-litpools xtensa option With support from assembler this option allows compiling huge functions, where single literal pool at the beginning of a function may not be reachable by L32R instructions at its end. Currently assembler --auto-litpools option cannot deal with literals used from multiple locations separated by more than 256 KBytes of code. Don't turn constants into literals, instead use MOVI instruction to load them into registers and let the assembler turn them into literals as necessary. Signed-off-by: Max Filippov <jcmvbkbc@gmail.com> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
2015-08-13 00:20:01 +02:00
mov\t%0, %1
wfr\t%0, %1
rfr\t%0, %1
+ movi\t%0, %y1
const16\t%0, %t1\;const16\t%0, %b1
%v1l32r\t%0, %1
%v1l32i\t%0, %1
%v0s32i\t%1, %0"
- [(set_attr "type" "farith,fload,fstore,move,load,store,move,farith,farith,move,load,load,store")
+ [(set_attr "type" "farith,fload,fstore,move,load,store,move,farith,farith,move,move,load,load,store")
(set_attr "mode" "SF")
- (set_attr "length" "3,3,3,2,2,2,3,3,3,6,3,3,3")])
+ (set_attr "length" "3,3,3,2,2,2,3,3,3,3,6,3,3,3")])
(define_insn "*lsiu"
[(set (match_operand:SF 0 "register_operand" "=f")
gcc: bump 4.9 series to 4.9.4 Two patches are removed, as they have been upstreamed: - 130-fix_build_with_gcc-6.patch (svn commit 233721, Git commit 8c3fa311caa86f61b4e28d1563d1110b44340fb2) - 920-libgcc-remove-unistd-header.patch (svn commit 226092, Git commit e940d7953f06af11d09229a29ecbcc1ba25b378d) All other patches have simply been refreshed, with no manual edit needed. A build+runtime test has been done with an ARM, Cortex-A8, EABIhf, musl configuration, booted under Qemu. Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
2016-08-03 23:48:24 +02:00
@@ -997,7 +1000,7 @@
gcc: backport mauto-litpools xtensa option With support from assembler this option allows compiling huge functions, where single literal pool at the beginning of a function may not be reachable by L32R instructions at its end. Currently assembler --auto-litpools option cannot deal with literals used from multiple locations separated by more than 256 KBytes of code. Don't turn constants into literals, instead use MOVI instruction to load them into registers and let the assembler turn them into literals as necessary. Signed-off-by: Max Filippov <jcmvbkbc@gmail.com> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
2015-08-13 00:20:01 +02:00
(match_operand:DF 1 "general_operand" ""))]
""
{
- if (CONSTANT_P (operands[1]) && !TARGET_CONST16)
+ if (CONSTANT_P (operands[1]) && !TARGET_CONST16 && !TARGET_AUTO_LITPOOLS)
operands[1] = force_const_mem (DFmode, operands[1]);
if (!register_operand (operands[0], DFmode)
gcc: bump 4.9 series to 4.9.4 Two patches are removed, as they have been upstreamed: - 130-fix_build_with_gcc-6.patch (svn commit 233721, Git commit 8c3fa311caa86f61b4e28d1563d1110b44340fb2) - 920-libgcc-remove-unistd-header.patch (svn commit 226092, Git commit e940d7953f06af11d09229a29ecbcc1ba25b378d) All other patches have simply been refreshed, with no manual edit needed. A build+runtime test has been done with an ARM, Cortex-A8, EABIhf, musl configuration, booted under Qemu. Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
2016-08-03 23:48:24 +02:00
@@ -1008,8 +1011,8 @@
gcc: backport mauto-litpools xtensa option With support from assembler this option allows compiling huge functions, where single literal pool at the beginning of a function may not be reachable by L32R instructions at its end. Currently assembler --auto-litpools option cannot deal with literals used from multiple locations separated by more than 256 KBytes of code. Don't turn constants into literals, instead use MOVI instruction to load them into registers and let the assembler turn them into literals as necessary. Signed-off-by: Max Filippov <jcmvbkbc@gmail.com> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
2015-08-13 00:20:01 +02:00
})
(define_insn_and_split "movdf_internal"
- [(set (match_operand:DF 0 "nonimmed_operand" "=a,W,a,a,U")
- (match_operand:DF 1 "move_operand" "r,iF,T,U,r"))]
+ [(set (match_operand:DF 0 "nonimmed_operand" "=a,a,W,a,a,U")
+ (match_operand:DF 1 "move_operand" "r,Y,iF,T,U,r"))]
"register_operand (operands[0], DFmode)
|| register_operand (operands[1], DFmode)"
"#"
gcc: bump 4.9 series to 4.9.4 Two patches are removed, as they have been upstreamed: - 130-fix_build_with_gcc-6.patch (svn commit 233721, Git commit 8c3fa311caa86f61b4e28d1563d1110b44340fb2) - 920-libgcc-remove-unistd-header.patch (svn commit 226092, Git commit e940d7953f06af11d09229a29ecbcc1ba25b378d) All other patches have simply been refreshed, with no manual edit needed. A build+runtime test has been done with an ARM, Cortex-A8, EABIhf, musl configuration, booted under Qemu. Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
2016-08-03 23:48:24 +02:00
Index: b/gcc/config/xtensa/xtensa.opt
===================================================================
gcc: backport mauto-litpools xtensa option With support from assembler this option allows compiling huge functions, where single literal pool at the beginning of a function may not be reachable by L32R instructions at its end. Currently assembler --auto-litpools option cannot deal with literals used from multiple locations separated by more than 256 KBytes of code. Don't turn constants into literals, instead use MOVI instruction to load them into registers and let the assembler turn them into literals as necessary. Signed-off-by: Max Filippov <jcmvbkbc@gmail.com> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
2015-08-13 00:20:01 +02:00
--- a/gcc/config/xtensa/xtensa.opt
+++ b/gcc/config/xtensa/xtensa.opt
gcc: bump 4.9 series to 4.9.4 Two patches are removed, as they have been upstreamed: - 130-fix_build_with_gcc-6.patch (svn commit 233721, Git commit 8c3fa311caa86f61b4e28d1563d1110b44340fb2) - 920-libgcc-remove-unistd-header.patch (svn commit 226092, Git commit e940d7953f06af11d09229a29ecbcc1ba25b378d) All other patches have simply been refreshed, with no manual edit needed. A build+runtime test has been done with an ARM, Cortex-A8, EABIhf, musl configuration, booted under Qemu. Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
2016-08-03 23:48:24 +02:00
@@ -38,6 +38,10 @@
gcc: backport mauto-litpools xtensa option With support from assembler this option allows compiling huge functions, where single literal pool at the beginning of a function may not be reachable by L32R instructions at its end. Currently assembler --auto-litpools option cannot deal with literals used from multiple locations separated by more than 256 KBytes of code. Don't turn constants into literals, instead use MOVI instruction to load them into registers and let the assembler turn them into literals as necessary. Signed-off-by: Max Filippov <jcmvbkbc@gmail.com> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
2015-08-13 00:20:01 +02:00
Target
Intersperse literal pools with code in the text section
+mauto-litpools
+Target Report Mask(AUTO_LITPOOLS)
+Relax literals in assembler and place them automatically in the text section
+
mserialize-volatile
Target Report Mask(SERIALIZE_VOLATILE)
-mno-serialize-volatile Do not serialize volatile memory references with MEMW instructions
Reference in New Issue Copy Permalink