This is the mail archive of the
gdb-cvs@sourceware.org
mailing list for the GDB project.
[binutils-gdb] Support 128-bit IEEE floating-point types on Intel and Power
- From: Ulrich Weigand <uweigand at sourceware dot org>
- To: gdb-cvs at sourceware dot org
- Date: 6 Sep 2016 15:35:24 -0000
- Subject: [binutils-gdb] Support 128-bit IEEE floating-point types on Intel and Power
https://sourceware.org/git/gitweb.cgi?p=binutils-gdb.git;h=00d5215ecec4fa0a78dcc37fec9425593753eb66
commit 00d5215ecec4fa0a78dcc37fec9425593753eb66
Author: Ulrich Weigand <ulrich.weigand@de.ibm.com>
Date: Tue Sep 6 17:33:15 2016 +0200
Support 128-bit IEEE floating-point types on Intel and Power
Now that all the prerequisites are in place, this commit finally adds support
for handling the __float128 type on Intel and Power, by providing appropriate
platform-specific versions of the floatformat_for_type callback.
Since at this point we do not yet have any indication in the debug info to
distinguish different floating-point formats of the same length, we simply
use the type name as hint. Types named "__float128" get the IEEE format.
In addition to handling "__float128" itself, we also recognize "_Float128"
and (on Power) "_Float64x", as well as the complex versions of those.
(As pointed out by Joseph Myers, starting with GCC 7, __float128 is just
a typedef for _Float128 -- but it's good to handle this anyway.)
A new test case does some simple verification that the format is decoded
correctly, using both __float128 and "long double" to make sure using both
in the same file still works. Another new test verifies handling of the
_FloatN and _FloatNx types supported by GCC 7, as well as the complex
versions of those types.
Note that this still only supports basic format decoding and encoding.
We do not yet support the GNU extension 'g' suffix for __float128 constants.
In addition, since all *arithmetic* on floating-point values is still
performed in native host "long double" arithmetic, if that format is not
able to encode all target __float128 values, we may get incorrect results.
(To fix this would require implementing fully synthetic target floating-
point arithmetic along the lines of GCC's real.c, presumably using MPFR.)
gdb/ChangeLog:
* i386-tdep.c (i386_floatformat_for_type): New function.
(i386_gdbarch_init): Install it.
* ppc-linux-tdep.c (ppc_floatformat_for_type): New function.
(ppc_linux_init_abi): Install it.
gdb/testsuite/ChangeLog:
* gdb.base/float128.c: New file.
* gdb.base/float128.exp: Likewise.
* gdb.base/floatn.c: Likewise.
* gdb.base/floatn.exp: Likewise.
Signed-off-by: Ulrich Weigand <ulrich.weigand@de.ibm.com>
Diff:
---
gdb/ChangeLog | 7 ++
gdb/i386-tdep.c | 20 ++++++
gdb/ppc-linux-tdep.c | 22 +++++++
gdb/testsuite/ChangeLog | 7 ++
gdb/testsuite/gdb.base/float128.c | 30 +++++++++
gdb/testsuite/gdb.base/float128.exp | 76 ++++++++++++++++++++++
gdb/testsuite/gdb.base/floatn.c | 48 ++++++++++++++
gdb/testsuite/gdb.base/floatn.exp | 124 ++++++++++++++++++++++++++++++++++++
8 files changed, 334 insertions(+)
diff --git a/gdb/ChangeLog b/gdb/ChangeLog
index f269c3c..0df1486 100644
--- a/gdb/ChangeLog
+++ b/gdb/ChangeLog
@@ -1,5 +1,12 @@
2016-09-05 Ulrich Weigand <uweigand@de.ibm.com>
+ * i386-tdep.c (i386_floatformat_for_type): New function.
+ (i386_gdbarch_init): Install it.
+ * ppc-linux-tdep.c (ppc_floatformat_for_type): New function.
+ (ppc_linux_init_abi): Install it.
+
+2016-09-05 Ulrich Weigand <uweigand@de.ibm.com>
+
* gdbarch.sh (floatformat_for_type): New gdbarch callback.
* gdbarch.h, gdbarch.c: Re-generate.
* arch-utils.h (default_floatformat_for_type): New prototype.
diff --git a/gdb/i386-tdep.c b/gdb/i386-tdep.c
index 92987af..9ed24cf 100644
--- a/gdb/i386-tdep.c
+++ b/gdb/i386-tdep.c
@@ -8161,6 +8161,23 @@ i386_fast_tracepoint_valid_at (struct gdbarch *gdbarch, CORE_ADDR addr,
}
}
+/* Return a floating-point format for a floating-point variable of
+ length LEN in bits. If non-NULL, NAME is the name of its type.
+ If no suitable type is found, return NULL. */
+
+const struct floatformat **
+i386_floatformat_for_type (struct gdbarch *gdbarch,
+ const char *name, int len)
+{
+ if (len == 128 && name)
+ if (strcmp (name, "__float128") == 0
+ || strcmp (name, "_Float128") == 0
+ || strcmp (name, "complex _Float128") == 0)
+ return floatformats_ia64_quad;
+
+ return default_floatformat_for_type (gdbarch, name, len);
+}
+
static int
i386_validate_tdesc_p (struct gdbarch_tdep *tdep,
struct tdesc_arch_data *tdesc_data)
@@ -8370,6 +8387,9 @@ i386_gdbarch_init (struct gdbarch_info info, struct gdbarch_list *arches)
alignment. */
set_gdbarch_long_double_bit (gdbarch, 96);
+ /* Support for floating-point data type variants. */
+ set_gdbarch_floatformat_for_type (gdbarch, i386_floatformat_for_type);
+
/* Register numbers of various important registers. */
set_gdbarch_sp_regnum (gdbarch, I386_ESP_REGNUM); /* %esp */
set_gdbarch_pc_regnum (gdbarch, I386_EIP_REGNUM); /* %eip */
diff --git a/gdb/ppc-linux-tdep.c b/gdb/ppc-linux-tdep.c
index cde4f2e..ee158e3 100644
--- a/gdb/ppc-linux-tdep.c
+++ b/gdb/ppc-linux-tdep.c
@@ -1628,6 +1628,25 @@ ppc_init_linux_record_tdep (struct linux_record_tdep *record_tdep,
record_tdep->ioctl_FIOQSIZE = 0x40086680;
}
+/* Return a floating-point format for a floating-point variable of
+ length LEN in bits. If non-NULL, NAME is the name of its type.
+ If no suitable type is found, return NULL. */
+
+const struct floatformat **
+ppc_floatformat_for_type (struct gdbarch *gdbarch,
+ const char *name, int len)
+{
+ if (len == 128 && name)
+ if (strcmp (name, "__float128") == 0
+ || strcmp (name, "_Float128") == 0
+ || strcmp (name, "_Float64x") == 0
+ || strcmp (name, "complex _Float128") == 0
+ || strcmp (name, "complex _Float64x") == 0)
+ return floatformats_ia64_quad;
+
+ return default_floatformat_for_type (gdbarch, name, len);
+}
+
static void
ppc_linux_init_abi (struct gdbarch_info info,
struct gdbarch *gdbarch)
@@ -1651,6 +1670,9 @@ ppc_linux_init_abi (struct gdbarch_info info,
set_gdbarch_long_double_bit (gdbarch, 16 * TARGET_CHAR_BIT);
set_gdbarch_long_double_format (gdbarch, floatformats_ibm_long_double);
+ /* Support for floating-point data type variants. */
+ set_gdbarch_floatformat_for_type (gdbarch, ppc_floatformat_for_type);
+
/* Handle inferior calls during interrupted system calls. */
set_gdbarch_write_pc (gdbarch, ppc_linux_write_pc);
diff --git a/gdb/testsuite/ChangeLog b/gdb/testsuite/ChangeLog
index d15fbbb..16070a3 100644
--- a/gdb/testsuite/ChangeLog
+++ b/gdb/testsuite/ChangeLog
@@ -1,5 +1,12 @@
2016-09-05 Ulrich Weigand <uweigand@de.ibm.com>
+ * gdb.base/float128.c: New file.
+ * gdb.base/float128.exp: Likewise.
+ * gdb.base/floatn.c: Likewise.
+ * gdb.base/floatn.exp: Likewise.
+
+2016-09-05 Ulrich Weigand <uweigand@de.ibm.com>
+
* gdb.cp/hang.exp: Remove reference to TYPE_FLAG_STUB in comment.
2016-09-05 Pedro Alves <palves@redhat.com>
diff --git a/gdb/testsuite/gdb.base/float128.c b/gdb/testsuite/gdb.base/float128.c
new file mode 100644
index 0000000..cc6c250
--- /dev/null
+++ b/gdb/testsuite/gdb.base/float128.c
@@ -0,0 +1,30 @@
+/* This testcase is part of GDB, the GNU debugger.
+
+ Copyright 2016 Free Software Foundation, Inc.
+
+ 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 this program. If not, see <http://www.gnu.org/licenses/>. */
+
+#include <stdio.h>
+#include <stdlib.h>
+
+long double ld;
+__float128 f128;
+
+int main()
+{
+ ld = 1.375l;
+ f128 = 2.375q;
+
+ return 0;
+}
diff --git a/gdb/testsuite/gdb.base/float128.exp b/gdb/testsuite/gdb.base/float128.exp
new file mode 100644
index 0000000..5907a46
--- /dev/null
+++ b/gdb/testsuite/gdb.base/float128.exp
@@ -0,0 +1,76 @@
+# Copyright 2016 Free Software Foundation, Inc.
+
+# 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 this program. If not, see <http://www.gnu.org/licenses/>.
+
+# This file is part of the gdb testsuite. It is intended to test that
+# gdb could correctly handle floating point constant with a suffix.
+
+standard_testfile .c
+
+proc do_compile { {opts {}} } {
+ global srcdir subdir srcfile binfile
+ set ccopts {debug quiet}
+ foreach opt $opts {lappend ccopts "additional_flags=$opt"}
+ gdb_compile "${srcdir}/${subdir}/${srcfile}" "$binfile" executable $ccopts
+}
+
+if { [do_compile] != "" && [do_compile {-mfloat128}] != "" } {
+ untested "compiler can't handle __float128 type?"
+ return -1
+}
+
+clean_restart ${binfile}
+
+if ![runto_main] then {
+ perror "couldn't run to breakpoint"
+ continue
+}
+
+# Run to the breakpoint at return.
+gdb_breakpoint [gdb_get_line_number "return"]
+gdb_continue_to_breakpoint "return"
+
+# Print the original value of ld and f128
+gdb_test "print ld" ".* = 1\\.375.*" "The original value of ld is 1.375"
+gdb_test "print f128" ".* = 2\\.375.*" "The original value of f128 is 2.375"
+
+# Test that gdb could correctly recognize float constant expression with a suffix.
+# FIXME: gdb does not yet recognize the GNU extension 'q' suffix for __float128 constants.
+gdb_test "print ld=-1.375l" ".* = -1\\.375.*" "Try to change ld to -1.375 with 'print ld=-1.375l'"
+gdb_test "print f128=-2.375l" ".* = -2\\.375.*" "Try to change f128 to -2.375 with 'print f128=-2.375l'"
+
+# Test that gdb could handle the above correctly with "set var" command.
+set test "set variable ld=10.375l"
+gdb_test_multiple "set var ld=10.375l" "$test" {
+ -re "$gdb_prompt $" {
+ pass "$test"
+ }
+ -re "Invalid number.*$gdb_prompt $" {
+ fail "$test (do not recognize 10.375l)"
+ }
+}
+
+set test "set variable f128=20.375l"
+gdb_test_multiple "set var f128=20.375l" "$test" {
+ -re "$gdb_prompt $" {
+ pass "$test"
+ }
+ -re "Invalid number.*$gdb_prompt $" {
+ fail "$test (do not recognize 20.375l)"
+ }
+}
+
+gdb_test "print ld" ".* = 10\\.375.*" "The value of ld is changed to 10.375"
+gdb_test "print f128" ".* = 20\\.375.*" "The value of f128 is changed to 20.375"
+
diff --git a/gdb/testsuite/gdb.base/floatn.c b/gdb/testsuite/gdb.base/floatn.c
new file mode 100644
index 0000000..84c5fba
--- /dev/null
+++ b/gdb/testsuite/gdb.base/floatn.c
@@ -0,0 +1,48 @@
+/* This testcase is part of GDB, the GNU debugger.
+
+ Copyright 2016 Free Software Foundation, Inc.
+
+ 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 this program. If not, see <http://www.gnu.org/licenses/>. */
+
+#include <stdio.h>
+#include <stdlib.h>
+
+_Float32 f32;
+_Float64 f64;
+_Float128 f128;
+_Float32x f32x;
+_Float64x f64x;
+
+_Complex _Float32 c32;
+_Complex _Float64 c64;
+_Complex _Float128 c128;
+_Complex _Float32x c32x;
+_Complex _Float64x c64x;
+
+int main()
+{
+ f32 = 1.5f32;
+ f64 = 2.25f64;
+ f128 = 3.375f128;
+ f32x = 10.5f32x;
+ f64x = 20.25f64x;
+
+ c32 = 1.5f32 + 1.0if;
+ c64 = 2.25f64 + 1.0if;
+ c128 = 3.375f128 + 1.0if;
+ c32x = 10.5f32x + 1.0if;
+ c64x = 20.25f64x + 1.0if;
+
+ return 0;
+}
diff --git a/gdb/testsuite/gdb.base/floatn.exp b/gdb/testsuite/gdb.base/floatn.exp
new file mode 100644
index 0000000..0a8b4ff
--- /dev/null
+++ b/gdb/testsuite/gdb.base/floatn.exp
@@ -0,0 +1,124 @@
+# Copyright 2016 Free Software Foundation, Inc.
+
+# 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 this program. If not, see <http://www.gnu.org/licenses/>.
+
+# This file is part of the gdb testsuite. It is intended to test that
+# gdb could correctly handle floating point constant with a suffix.
+
+standard_testfile .c
+
+proc do_compile { {opts {}} } {
+ global srcdir subdir srcfile binfile
+ set ccopts {debug quiet}
+ foreach opt $opts {lappend ccopts "additional_flags=$opt"}
+ gdb_compile "${srcdir}/${subdir}/${srcfile}" "$binfile" executable $ccopts
+}
+
+if { [do_compile] != "" && [do_compile {-mfloat128}] != "" } {
+ untested "compiler can't handle _FloatN/_FloatNx types?"
+ return -1
+}
+
+clean_restart ${binfile}
+
+if ![runto_main] then {
+ perror "couldn't run to breakpoint"
+ continue
+}
+
+# Run to the breakpoint at return.
+gdb_breakpoint [gdb_get_line_number "return"]
+gdb_continue_to_breakpoint "return"
+
+# Print the original values of f32, f64, f128, f32x, f64x.
+gdb_test "print f32" ".* = 1\\.5.*" "The original value of f32 is 1.5"
+gdb_test "print f64" ".* = 2\\.25.*" "The original value of f64 is 2.25"
+gdb_test "print f128" ".* = 3\\.375.*" "The original value of f128 is 3.375"
+gdb_test "print f32x" ".* = 10\\.5.*" "The original value of f32x is 10.5"
+gdb_test "print f64x" ".* = 20\\.25.*" "The original value of f64x is 20.25"
+
+# Test that gdb could correctly recognize float constant expression with a suffix.
+# FIXME: gdb does not yet recognize the suffix for _FloatN/_FloatNx types.
+gdb_test "print f32=-1.5" ".* = -1\\.5.*" "Try to change f32 to -1.5 with 'print f32=-1.5'"
+gdb_test "print f64=-2.25" ".* = -2\\.25.*" "Try to change f64 to -2.25 with 'print f64=-2.25'"
+gdb_test "print f128=-3.375" ".* = -3\\.375.*" "Try to change f128 to -3.375 with 'print f128=-3.375'"
+gdb_test "print f32x=-10.5" ".* = -10\\.5.*" "Try to change f32x to -10.5 with 'print f32=-1.5x'"
+gdb_test "print f64x=-20.25" ".* = -20\\.25.*" "Try to change f64x to -20.25 with 'print f64=-2.25x'"
+
+# Test that gdb could handle the above correctly with "set var" command.
+set test "set variable f32 = 10.5"
+gdb_test_multiple "set var f32=10.5" "$test" {
+ -re "$gdb_prompt $" {
+ pass "$test"
+ }
+ -re "Invalid number.*$gdb_prompt $" {
+ fail "$test (do not recognize 10.5)"
+ }
+}
+
+set test "set variable f64 = 20.25"
+gdb_test_multiple "set var f64=20.25" "$test" {
+ -re "$gdb_prompt $" {
+ pass "$test"
+ }
+ -re "Invalid number.*$gdb_prompt $" {
+ fail "$test (do not recognize 20.25)"
+ }
+}
+
+set test "set variable f128 = 30.375"
+gdb_test_multiple "set var f128=30.375" "$test" {
+ -re "$gdb_prompt $" {
+ pass "$test"
+ }
+ -re "Invalid number.*$gdb_prompt $" {
+ fail "$test (do not recognize 30.375)"
+ }
+}
+
+set test "set variable f32x = 100.5"
+gdb_test_multiple "set var f32x=100.5" "$test" {
+ -re "$gdb_prompt $" {
+ pass "$test"
+ }
+ -re "Invalid number.*$gdb_prompt $" {
+ fail "$test (do not recognize 100.5)"
+ }
+}
+
+set test "set variable f64x = 200.25"
+gdb_test_multiple "set var f64x=200.25" "$test" {
+ -re "$gdb_prompt $" {
+ pass "$test"
+ }
+ -re "Invalid number.*$gdb_prompt $" {
+ fail "$test (do not recognize 200.25)"
+ }
+}
+
+gdb_test "print f32" ".* = 10\\.5.*" "The value of f32 is changed to 10.5"
+gdb_test "print f64" ".* = 20\\.25.*" "The value of f64 is changed to 20.25"
+gdb_test "print f128" ".* = 30\\.375.*" "The value of f128 is changed to 30.375"
+gdb_test "print f32x" ".* = 100\\.5.*" "The value of f32x is changed to 100.5"
+gdb_test "print f64x" ".* = 200\\.25.*" "The value of f64x is changed to 200.25"
+
+# Print the original values of c32, c64, c128, c32x, c64x.
+gdb_test "print c32" ".* = 1\\.5 \\+ 1 \\* I.*" "The original value of c32 is 1.5 + 1 * I"
+gdb_test "print c64" ".* = 2\\.25 \\+ 1 \\* I.*" "The original value of c64 is 2.25 + 1 * I"
+gdb_test "print c128" ".* = 3\\.375 \\+ 1 \\* I.*" "The original value of c128 is 3.375 + 1 * I"
+gdb_test "print c32x" ".* = 10\\.5 \\+ 1 \\* I.*" "The original value of c32x is 10.5 + 1 * I"
+gdb_test "print c64x" ".* = 20\\.25 \\+ 1 \\* I.*" "The original value of c64x is 20.25 + 1 * I"
+
+# FIXME: GDB cannot parse non-trivial complex constants yet.
+