[PATCH] Handle half-float in 'x' command

[PATCH] Handle half-float in 'x' command

[Tom Tromey]

Using 'x/hf' should print bytes as float16, but instead it currently
prints as an integer.  I tracked this down to a missing case in
float_type_from_length.

Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=30161
---
 gdb/printcmd.c                            |  4 +++-
 gdb/testsuite/gdb.dwarf2/dw2-bfloat16.c   | 26 +++++++++++++++++++++++
 gdb/testsuite/gdb.dwarf2/dw2-bfloat16.exp | 19 ++++++++++++++++-
 3 files changed, 47 insertions(+), 2 deletions(-)
 create mode 100644 gdb/testsuite/gdb.dwarf2/dw2-bfloat16.c

diff --git a/gdb/printcmd.c b/gdb/printcmd.c
index 0d3bc292d4e..6fd69818db9 100644
--- a/gdb/printcmd.c
+++ b/gdb/printcmd.c
@@ -349,7 +349,9 @@ float_type_from_length (struct type *type)
   struct gdbarch *gdbarch = type->arch ();
   const struct builtin_type *builtin = builtin_type (gdbarch);
 
-  if (type->length () == builtin->builtin_float->length ())
+  if (type->length () == builtin->builtin_half->length ())
+    type = builtin->builtin_half;
+  else if (type->length () == builtin->builtin_float->length ())
     type = builtin->builtin_float;
   else if (type->length () == builtin->builtin_double->length ())
     type = builtin->builtin_double;
diff --git a/gdb/testsuite/gdb.dwarf2/dw2-bfloat16.c b/gdb/testsuite/gdb.dwarf2/dw2-bfloat16.c
new file mode 100644
index 00000000000..30dcd38bf06
--- /dev/null
+++ b/gdb/testsuite/gdb.dwarf2/dw2-bfloat16.c
@@ -0,0 +1,26 @@
+/* This testcase is part of GDB, the GNU debugger.
+
+   Copyright 2023 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/>.  */
+
+/* Storage for the 'x' test.  */
+unsigned char storage[2];
+
+int
+main()
+{
+  asm ("main_label: .globl main_label");
+  return 0;
+}
diff --git a/gdb/testsuite/gdb.dwarf2/dw2-bfloat16.exp b/gdb/testsuite/gdb.dwarf2/dw2-bfloat16.exp
index 79058cafe7f..242d1197103 100644
--- a/gdb/testsuite/gdb.dwarf2/dw2-bfloat16.exp
+++ b/gdb/testsuite/gdb.dwarf2/dw2-bfloat16.exp
@@ -21,7 +21,7 @@ load_lib dwarf.exp
 # This test can only be run on targets which support DWARF-2 and use gas.
 require dwarf2_support
 
-standard_testfile main.c -dw.S
+standard_testfile .c -dw.S
 
 # Make some DWARF for the test.
 # 0x4049 is the equivalent of 3.141 for bfloat16.
@@ -78,3 +78,20 @@ if ![runto_main] {
 # Make sure we can print both types correctly.
 gdb_test "print bf16_1" " = 3.141"
 gdb_test "print fp16_1" " = 3.1406"
+
+set bytes {0x48 0x42}
+set endian [get_endianness]
+if {$endian == "big"} {
+    set bytes [lreverse $bytes]
+}
+
+gdb_test_no_output "set var \$s = (unsigned char *) &storage" \
+    "set convenience variable"
+gdb_test "print/x \$s\[0\] = [lindex $bytes 0]" \
+    " = $hex" "set first element"
+gdb_test "print/x \$s\[1\] = [lindex $bytes 1]" \
+    " = $hex" "set second element"
+gdb_test "print *(__fp16 *) \$s" " = 3\\.1406" \
+    "print float16 value"
+gdb_test "x/hf \$s" "$hex <storage>:\[ \t\]*3\\.1406" \
+    "examine float16 value"
-- 
2.39.1

Re: [PATCH] Handle half-float in 'x' command

[Simon Marchi]

On 2/24/23 11:38, Tom Tromey via Gdb-patches wrote:
> Using 'x/hf' should print bytes as float16, but instead it currently
> prints as an integer.  I tracked this down to a missing case in
> float_type_from_length.

I looked at other uses of builtin_type::builtin_float, to see if there
would be other spots with a similar problem.  Maybe
dwarf2_init_complex_target_type?  However, after googling a bit, a
"half-precision complex" doesn't seem to exist anywhere, so I guess it's
fine.

> 
> Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=30161
> ---
>  gdb/printcmd.c                            |  4 +++-
>  gdb/testsuite/gdb.dwarf2/dw2-bfloat16.c   | 26 +++++++++++++++++++++++
>  gdb/testsuite/gdb.dwarf2/dw2-bfloat16.exp | 19 ++++++++++++++++-
>  3 files changed, 47 insertions(+), 2 deletions(-)
>  create mode 100644 gdb/testsuite/gdb.dwarf2/dw2-bfloat16.c
> 
> diff --git a/gdb/printcmd.c b/gdb/printcmd.c
> index 0d3bc292d4e..6fd69818db9 100644
> --- a/gdb/printcmd.c
> +++ b/gdb/printcmd.c
> @@ -349,7 +349,9 @@ float_type_from_length (struct type *type)
>    struct gdbarch *gdbarch = type->arch ();
>    const struct builtin_type *builtin = builtin_type (gdbarch);
>  
> -  if (type->length () == builtin->builtin_float->length ())
> +  if (type->length () == builtin->builtin_half->length ())
> +    type = builtin->builtin_half;
> +  else if (type->length () == builtin->builtin_float->length ())
>      type = builtin->builtin_float;
>    else if (type->length () == builtin->builtin_double->length ())
>      type = builtin->builtin_double;
> diff --git a/gdb/testsuite/gdb.dwarf2/dw2-bfloat16.c b/gdb/testsuite/gdb.dwarf2/dw2-bfloat16.c
> new file mode 100644
> index 00000000000..30dcd38bf06
> --- /dev/null
> +++ b/gdb/testsuite/gdb.dwarf2/dw2-bfloat16.c
> @@ -0,0 +1,26 @@
> +/* This testcase is part of GDB, the GNU debugger.
> +
> +   Copyright 2023 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/>.  */
> +
> +/* Storage for the 'x' test.  */
> +unsigned char storage[2];
> +
> +int
> +main()
> +{
> +  asm ("main_label: .globl main_label");
> +  return 0;
> +}
> diff --git a/gdb/testsuite/gdb.dwarf2/dw2-bfloat16.exp b/gdb/testsuite/gdb.dwarf2/dw2-bfloat16.exp
> index 79058cafe7f..242d1197103 100644
> --- a/gdb/testsuite/gdb.dwarf2/dw2-bfloat16.exp
> +++ b/gdb/testsuite/gdb.dwarf2/dw2-bfloat16.exp
> @@ -21,7 +21,7 @@ load_lib dwarf.exp
>  # This test can only be run on targets which support DWARF-2 and use gas.
>  require dwarf2_support
>  
> -standard_testfile main.c -dw.S
> +standard_testfile .c -dw.S
>  
>  # Make some DWARF for the test.
>  # 0x4049 is the equivalent of 3.141 for bfloat16.
> @@ -78,3 +78,20 @@ if ![runto_main] {
>  # Make sure we can print both types correctly.
>  gdb_test "print bf16_1" " = 3.141"
>  gdb_test "print fp16_1" " = 3.1406"
> +
> +set bytes {0x48 0x42}
> +set endian [get_endianness]
> +if {$endian == "big"} {
> +    set bytes [lreverse $bytes]
> +}
> +
> +gdb_test_no_output "set var \$s = (unsigned char *) &storage" \
> +    "set convenience variable"
> +gdb_test "print/x \$s\[0\] = [lindex $bytes 0]" \
> +    " = $hex" "set first element"
> +gdb_test "print/x \$s\[1\] = [lindex $bytes 1]" \
> +    " = $hex" "set second element"
> +gdb_test "print *(__fp16 *) \$s" " = 3\\.1406" \
> +    "print float16 value"
> +gdb_test "x/hf \$s" "$hex <storage>:\[ \t\]*3\\.1406" \
> +    "examine float16 value"

Can you add a comment before the section you added, to indicate what it
intends to test?

Otherwise, that LGTM:

Approved-By: Simon Marchi <simon.marchi@efficios.com>

Simon

Re: [PATCH] Handle half-float in 'x' command

[Tom Tromey]

>>>>> "Simon" == Simon Marchi <simark@simark.ca> writes:

Simon> Can you add a comment before the section you added, to indicate what it
Simon> intends to test?

Simon> Otherwise, that LGTM:

Simon> Approved-By: Simon Marchi <simon.marchi@efficios.com>

Simon> Simon

Tom

Re: [PATCH] Handle half-float in 'x' command

[Tom Tromey]

Simon> Can you add a comment before the section you added, to indicate what it
Simon> intends to test?

I mean to say, I did it.  Thanks for the review.

Tom

Re: [PATCH] Handle half-float in 'x' command

[Simon Marchi]

On 3/2/23 10:50, Tom Tromey wrote:
> Simon> Can you add a comment before the section you added, to indicate what it
> Simon> intends to test?
> 
> I mean to say, I did it.  Thanks for the review.

Oops, this causes:

    p/f *(short *)s^M
    $41 = 1.7345e-05^M
    (gdb) FAIL: gdb.base/long_long.exp: p/f *(short *)s

    x/2hf h^M
    0x555555558060 <h>:     1.7345e-05      -0.028046^M
    (gdb) FAIL: gdb.base/long_long.exp: x/2hf h

I think it's just the test outcomes that need to be updated.

It goes through that code:

https://gitlab.com/gnutools/binutils-gdb/-/blob/70728e1d396475e8e630bfdd3fb8e8c8211bdbbd/gdb/printcmd.c#L455-464


  /* Printing a non-float type as 'f' will interpret the data as if it were
     of a floating-point type of the same length, if that exists.  Otherwise,
     the data is printed as integer.  */
  char format = options->format;
  if (format == 'f' && type->code () != TYPE_CODE_FLT)
    {
      type = float_type_from_length (type);
      if (type->code () != TYPE_CODE_FLT)
	format = 0;
    }

Before, we didn't find a 2-byte float format, so it was printed as
integers, and now we do, so it's printed as a float.  It seems
consistent with the comment in the code.

Simon