[PATCH 0/2] Add AVX512-FP16 support.

[PATCH 0/2] Add AVX512-FP16 support.

[Felix Willgerodt]

New CPUs will add support for the FP16 data type in AVX512:
https://software.intel.com/content/www/us/en/develop/download/intel-avx512-fp16-architecture-specification.html

This comes with ABI changes that allow _Float16 and _Float16 _Complex as
function arguments and return values.

This series adds support for the register view and ABI changes.

Regards,
Felix

Felix Willgerodt (2):
  Add half support for AVX512 register view.
  gdb: Enable finish command and inferior calls for _Float16 on amd64
    and i386.

 gdb/amd64-tdep.c                              |  12 +-
 gdb/features/i386/32bit-sse.c                 |   5 +
 gdb/features/i386/32bit-sse.xml               |   2 +
 gdb/features/i386/64bit-avx512.c              |   5 +
 gdb/features/i386/64bit-avx512.xml            |   2 +
 gdb/features/i386/64bit-sse.c                 |   5 +
 gdb/features/i386/64bit-sse.xml               |   2 +
 gdb/i386-tdep.c                               |  15 +-
 gdb/testsuite/gdb.arch/x86-avx512fp16-abi.c   |  38 ++++
 gdb/testsuite/gdb.arch/x86-avx512fp16-abi.exp |  63 +++++++
 gdb/testsuite/gdb.arch/x86-avx512fp16.c       | 164 ++++++++++++++++++
 gdb/testsuite/gdb.arch/x86-avx512fp16.exp     |  68 ++++++++
 gdb/testsuite/lib/gdb.exp                     |  51 ++++++
 13 files changed, 425 insertions(+), 7 deletions(-)
 create mode 100644 gdb/testsuite/gdb.arch/x86-avx512fp16-abi.c
 create mode 100644 gdb/testsuite/gdb.arch/x86-avx512fp16-abi.exp
 create mode 100644 gdb/testsuite/gdb.arch/x86-avx512fp16.c
 create mode 100644 gdb/testsuite/gdb.arch/x86-avx512fp16.exp

-- 
2.25.4

Intel Deutschland GmbH
Registered Address: Am Campeon 10, 85579 Neubiberg, Germany
Tel: +49 89 99 8853-0, www.intel.de <http://www.intel.de>
Managing Directors: Christin Eisenschmid, Sharon Heck, Tiffany Doon Silva  
Chairperson of the Supervisory Board: Nicole Lau
Registered Office: Munich
Commercial Register: Amtsgericht Muenchen HRB 186928

[PATCH 1/2] Add half support for AVX512 register view.

[Felix Willgerodt]

This adds support for the half datatype, FP16, to the AVX512 register printing.

gdb/ChangeLog:
2020-07-21  Felix Willgerodt  <Felix.Willgerodt@intel.com>

	* i386-tdep.c (i386_zmm_type) <v32_half>: New field.
	(i386_ymm_type) <v16_half>: New field.
	(i386_gdbarch_init): Add set_gdbarch_half_format.
	* features/i386/64bit-avx512.xml: Add half type.
	* features/i386/64bit-avx512.c: Regenerated.
	* features/i386/64bit-sse.xml: Add half type.
	* features/i386/64bit-sse.c: Regenerated.

gdb/testsuite/ChangeLog:
2021-07-21  Felix Willgerodt  <Felix.Willgerodt@intel.com>

	* gdb.arch/x86-avx512fp16.c: New file.
	* gdb.arch/x86-avx512fp16.exp: New file.
	* lib/gdb.exp (skip_avx512fp16_tests): New function.
---
 gdb/features/i386/32bit-sse.c             |   5 +
 gdb/features/i386/32bit-sse.xml           |   2 +
 gdb/features/i386/64bit-avx512.c          |   5 +
 gdb/features/i386/64bit-avx512.xml        |   2 +
 gdb/features/i386/64bit-sse.c             |   5 +
 gdb/features/i386/64bit-sse.xml           |   2 +
 gdb/i386-tdep.c                           |   6 +
 gdb/testsuite/gdb.arch/x86-avx512fp16.c   | 164 ++++++++++++++++++++++
 gdb/testsuite/gdb.arch/x86-avx512fp16.exp |  68 +++++++++
 gdb/testsuite/lib/gdb.exp                 |  51 +++++++
 10 files changed, 310 insertions(+)
 create mode 100644 gdb/testsuite/gdb.arch/x86-avx512fp16.c
 create mode 100644 gdb/testsuite/gdb.arch/x86-avx512fp16.exp

diff --git a/gdb/features/i386/32bit-sse.c b/gdb/features/i386/32bit-sse.c
index e5eaf8feb59..5554849a6ca 100644
--- a/gdb/features/i386/32bit-sse.c
+++ b/gdb/features/i386/32bit-sse.c
@@ -10,6 +10,9 @@ create_feature_i386_32bit_sse (struct target_desc *result, long regnum)
 
   feature = tdesc_create_feature (result, "org.gnu.gdb.i386.sse");
   tdesc_type *element_type;
+  element_type = tdesc_named_type (feature, "ieee_half");
+  tdesc_create_vector (feature, "v8h", element_type, 8);
+
   element_type = tdesc_named_type (feature, "ieee_single");
   tdesc_create_vector (feature, "v4f", element_type, 4);
 
@@ -31,6 +34,8 @@ create_feature_i386_32bit_sse (struct target_desc *result, long regnum)
   tdesc_type_with_fields *type_with_fields;
   type_with_fields = tdesc_create_union (feature, "vec128");
   tdesc_type *field_type;
+  field_type = tdesc_named_type (feature, "v8h");
+  tdesc_add_field (type_with_fields, "v8_half", field_type);
   field_type = tdesc_named_type (feature, "v4f");
   tdesc_add_field (type_with_fields, "v4_float", field_type);
   field_type = tdesc_named_type (feature, "v2d");
diff --git a/gdb/features/i386/32bit-sse.xml b/gdb/features/i386/32bit-sse.xml
index 7bf3e0f3828..78adc04786b 100644
--- a/gdb/features/i386/32bit-sse.xml
+++ b/gdb/features/i386/32bit-sse.xml
@@ -7,6 +7,7 @@
 
 <!DOCTYPE feature SYSTEM "gdb-target.dtd">
 <feature name="org.gnu.gdb.i386.sse">
+  <vector id="v8h" type="ieee_half" count="8"/>
   <vector id="v4f" type="ieee_single" count="4"/>
   <vector id="v2d" type="ieee_double" count="2"/>
   <vector id="v16i8" type="int8" count="16"/>
@@ -14,6 +15,7 @@
   <vector id="v4i32" type="int32" count="4"/>
   <vector id="v2i64" type="int64" count="2"/>
   <union id="vec128">
+    <field name="v8_half" type="v8h"/>
     <field name="v4_float" type="v4f"/>
     <field name="v2_double" type="v2d"/>
     <field name="v16_int8" type="v16i8"/>
diff --git a/gdb/features/i386/64bit-avx512.c b/gdb/features/i386/64bit-avx512.c
index 1bd49dcd9d5..9ceadd89cda 100644
--- a/gdb/features/i386/64bit-avx512.c
+++ b/gdb/features/i386/64bit-avx512.c
@@ -13,6 +13,9 @@ create_feature_i386_64bit_avx512 (struct target_desc *result, long regnum)
   element_type = tdesc_named_type (feature, "bfloat16");
   tdesc_create_vector (feature, "v8bf16", element_type, 8);
 
+  element_type = tdesc_named_type (feature, "ieee_half");
+  tdesc_create_vector (feature, "v8h", element_type, 8);
+
   element_type = tdesc_named_type (feature, "ieee_single");
   tdesc_create_vector (feature, "v4f", element_type, 4);
 
@@ -36,6 +39,8 @@ create_feature_i386_64bit_avx512 (struct target_desc *result, long regnum)
   tdesc_type *field_type;
   field_type = tdesc_named_type (feature, "v8bf16");
   tdesc_add_field (type_with_fields, "v8_bfloat16", field_type);
+  field_type = tdesc_named_type (feature, "v8h");
+  tdesc_add_field (type_with_fields, "v8_half", field_type);
   field_type = tdesc_named_type (feature, "v4f");
   tdesc_add_field (type_with_fields, "v4_float", field_type);
   field_type = tdesc_named_type (feature, "v2d");
diff --git a/gdb/features/i386/64bit-avx512.xml b/gdb/features/i386/64bit-avx512.xml
index a4c7f02d6d8..b7fe4d5015b 100644
--- a/gdb/features/i386/64bit-avx512.xml
+++ b/gdb/features/i386/64bit-avx512.xml
@@ -8,6 +8,7 @@
 <!DOCTYPE feature SYSTEM "gdb-target.dtd">
 <feature name="org.gnu.gdb.i386.avx512">
   <vector id="v8bf16" type="bfloat16" count="8"/>
+  <vector id="v8h" type="ieee_half" count="8"/>
   <vector id="v4f" type="ieee_single" count="4"/>
   <vector id="v2d" type="ieee_double" count="2"/>
   <vector id="v16i8" type="int8" count="16"/>
@@ -16,6 +17,7 @@
   <vector id="v2i64" type="int64" count="2"/>
   <union id="vec128">
     <field name="v8_bfloat16" type="v8bf16"/>
+    <field name="v8_half" type="v8h"/>
     <field name="v4_float" type="v4f"/>
     <field name="v2_double" type="v2d"/>
     <field name="v16_int8" type="v16i8"/>
diff --git a/gdb/features/i386/64bit-sse.c b/gdb/features/i386/64bit-sse.c
index 645e3141337..ffb7a45ec1a 100644
--- a/gdb/features/i386/64bit-sse.c
+++ b/gdb/features/i386/64bit-sse.c
@@ -13,6 +13,9 @@ create_feature_i386_64bit_sse (struct target_desc *result, long regnum)
   element_type = tdesc_named_type (feature, "bfloat16");
   tdesc_create_vector (feature, "v8bf16", element_type, 8);
 
+  element_type = tdesc_named_type (feature, "ieee_half");
+  tdesc_create_vector (feature, "v8h", element_type, 8);
+
   element_type = tdesc_named_type (feature, "ieee_single");
   tdesc_create_vector (feature, "v4f", element_type, 4);
 
@@ -36,6 +39,8 @@ create_feature_i386_64bit_sse (struct target_desc *result, long regnum)
   tdesc_type *field_type;
   field_type = tdesc_named_type (feature, "v8bf16");
   tdesc_add_field (type_with_fields, "v8_bfloat16", field_type);
+  field_type = tdesc_named_type (feature, "v8h");
+  tdesc_add_field (type_with_fields, "v8_half", field_type);
   field_type = tdesc_named_type (feature, "v4f");
   tdesc_add_field (type_with_fields, "v4_float", field_type);
   field_type = tdesc_named_type (feature, "v2d");
diff --git a/gdb/features/i386/64bit-sse.xml b/gdb/features/i386/64bit-sse.xml
index f23d4d5cb4e..9ae2310af33 100644
--- a/gdb/features/i386/64bit-sse.xml
+++ b/gdb/features/i386/64bit-sse.xml
@@ -8,6 +8,7 @@
 <!DOCTYPE feature SYSTEM "gdb-target.dtd">
 <feature name="org.gnu.gdb.i386.sse">
   <vector id="v8bf16" type="bfloat16" count="8"/>
+  <vector id="v8h" type="ieee_half" count="8"/>
   <vector id="v4f" type="ieee_single" count="4"/>
   <vector id="v2d" type="ieee_double" count="2"/>
   <vector id="v16i8" type="int8" count="16"/>
@@ -16,6 +17,7 @@
   <vector id="v2i64" type="int64" count="2"/>
   <union id="vec128">
     <field name="v8_bfloat16" type="v8bf16"/>
+    <field name="v8_half" type="v8h"/>
     <field name="v4_float" type="v4f"/>
     <field name="v2_double" type="v2d"/>
     <field name="v16_int8" type="v16i8"/>
diff --git a/gdb/i386-tdep.c b/gdb/i386-tdep.c
index e436437c683..c2835a2458d 100644
--- a/gdb/i386-tdep.c
+++ b/gdb/i386-tdep.c
@@ -3117,6 +3117,7 @@ i386_zmm_type (struct gdbarch *gdbarch)
 	int8_t v64_int8[64];
 	double v8_double[8];
 	float v16_float[16];
+	float16_t v32_half[32];
 	bfloat16_t v32_bfloat16[32];
       };
 #endif
@@ -3127,6 +3128,8 @@ i386_zmm_type (struct gdbarch *gdbarch)
 			       "__gdb_builtin_type_vec512i", TYPE_CODE_UNION);
       append_composite_type_field (t, "v32_bfloat16",
 				   init_vector_type (bt->builtin_bfloat16, 32));
+      append_composite_type_field (t, "v32_half",
+				   init_vector_type (bt->builtin_half, 32));
       append_composite_type_field (t, "v16_float",
 				   init_vector_type (bt->builtin_float, 16));
       append_composite_type_field (t, "v8_double",
@@ -3173,6 +3176,7 @@ i386_ymm_type (struct gdbarch *gdbarch)
 	int8_t v32_int8[32];
 	double v4_double[4];
 	float v8_float[8];
+	float16_t v16_half[16];
 	bfloat16_t v16_bfloat16[16];
       };
 #endif
@@ -3183,6 +3187,8 @@ i386_ymm_type (struct gdbarch *gdbarch)
 			       "__gdb_builtin_type_vec256i", TYPE_CODE_UNION);
       append_composite_type_field (t, "v16_bfloat16",
 				   init_vector_type (bt->builtin_bfloat16, 16));
+      append_composite_type_field (t, "v16_half",
+				   init_vector_type (bt->builtin_half, 16));
       append_composite_type_field (t, "v8_float",
 				   init_vector_type (bt->builtin_float, 8));
       append_composite_type_field (t, "v4_double",
diff --git a/gdb/testsuite/gdb.arch/x86-avx512fp16.c b/gdb/testsuite/gdb.arch/x86-avx512fp16.c
new file mode 100644
index 00000000000..ebc72957bb3
--- /dev/null
+++ b/gdb/testsuite/gdb.arch/x86-avx512fp16.c
@@ -0,0 +1,164 @@
+/* Test program for bfloat16 of AVX 512 registers.
+
+   Copyright 2021 Free Software Foundation, Inc.
+
+   This file is part of GDB.
+
+   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/>.  */
+
+typedef struct
+{
+  float f[4];
+} v4sd_t;
+
+typedef struct
+{
+  float f[8];
+} v8sd_t;
+
+typedef struct
+{
+  float f[16];
+} v16sd_t;
+
+v4sd_t xmm_data[] =
+{
+  { {  0.0,  0.125,  0.25,  0.375 } },
+  { {  0.5,  0.625,  0.75,  0.875 } },
+  { {  1.0,  1.125,  1.25,  1.375 } },
+  { {  1.5,  1.625,  1.75,  1.875 } },
+  { {  2.0,  2.125,  2.25,  2.375 } },
+  { {  2.5,  2.625,  2.75,  2.875 } },
+  { {  3.0,  3.125,  3.25,  3.375 } },
+  { {  3.5,  3.625,  3.75,  3.875 } },
+};
+
+v8sd_t ymm_data[] =
+{
+  { {  8.0,  8.25,  8.5,  8.75,  9.0,  9.25,  9.5,  9.75 } },
+  { { 10.0, 10.25, 10.5, 10.75, 11.0, 11.25, 11.5, 11.75 } },
+  { { 12.0, 12.25, 12.5, 12.75, 13.0, 13.25, 13.5, 13.75 } },
+  { { 14.0, 14.25, 14.5, 14.75, 15.0, 15.25, 15.5, 15.75 } },
+  { { 16.0, 16.25, 16.5, 16.75, 17.0, 17.25, 17.5, 17.75 } },
+  { { 18.0, 18.25, 18.5, 18.75, 19.0, 19.25, 19.5, 19.75 } },
+  { { 20.0, 20.25, 20.5, 20.75, 21.0, 21.25, 21.5, 21.75 } },
+  { { 22.0, 22.25, 22.5, 22.75, 23.0, 23.25, 23.5, 23.75 } },
+};
+
+v16sd_t zmm_data[] =
+{
+  { { 20.0,  20.5,  21.0,  21.5,  22.0,  22.5,  23.0,  23.5,  24.0,  24.5,
+      25.0,  25.5,  26.0,  26.5,  27.0,  27.5 } },
+  { { 28.0,  28.5,  29.0,  29.5,  30.0,  30.5,  31.0,  31.5,  32.0,  32.5,
+      33.0,  33.5,  34.0,  34.5,  35.0,  35.5 } },
+  { { 36.0,  36.5,  37.0,  37.5,  38.0,  38.5,  39.0,  39.5,  40.0,  40.5,
+      41.0,  41.5,  42.0,  42.5,  43.0,  43.5 } },
+  { { 44.0,  44.5,  45.0,  45.5,  46.0,  46.5,  47.0,  47.5,  48.0,  48.5,
+      49.0,  49.5,  50.0,  50.5,  51.0,  51.5 } },
+  { { 52.0,  52.5,  53.0,  53.5,  54.0,  54.5,  55.0,  55.5,  56.0,  56.5,
+      57.0,  57.5,  58.0,  58.5,  59.0,  59.5 } },
+  { { 60.0,  60.5,  61.0,  61.5,  62.0,  62.5,  63.0,  63.5,  64.0,  64.5,
+      65.0,  65.5,  66.0,  66.5,  67.0,  67.5 } },
+  { { 68.0,  68.5,  69.0,  69.5,  70.0,  70.5,  71.0,  71.5,  72.0,  72.5,
+      73.0,  73.5,  74.0,  74.5,  75.0,  75.5 } },
+  { { 76.0,  76.5,  77.0,  77.5,  78.0,  78.5,  79.0,  79.5,  80.0,  80.5,
+      81.0,  81.5,  82.0,  82.5,  83.0,  83.5 } },
+};
+
+void
+move_data_to_xmm_reg (void)
+{
+  asm ("vmovups 0(%0), %%xmm0 \n\t"
+       "vmovups 16(%0), %%xmm1 \n\t"
+       "vmovups 32(%0), %%xmm2 \n\t"
+       "vmovups 48(%0), %%xmm3 \n\t"
+       "vmovups 64(%0), %%xmm4 \n\t"
+       "vmovups 80(%0), %%xmm5 \n\t"
+       "vmovups 96(%0), %%xmm6 \n\t"
+       "vmovups 112(%0), %%xmm7 \n\t"
+       : /* no output operands  */
+       : "r" (xmm_data));
+}
+
+void
+move_data_to_ymm_reg (void)
+{
+  asm ("vmovups 0(%0), %%ymm0 \n\t"
+       "vmovups 32(%0), %%ymm1 \n\t"
+       "vmovups 64(%0), %%ymm2 \n\t"
+       "vmovups 96(%0), %%ymm3 \n\t"
+       "vmovups 128(%0), %%ymm4 \n\t"
+       "vmovups 160(%0), %%ymm5 \n\t"
+       "vmovups 192(%0), %%ymm6 \n\t"
+       "vmovups 224(%0), %%ymm7 \n\t"
+       : /* no output operands  */
+       : "r" (ymm_data));
+}
+
+void
+move_data_to_zmm_reg (void)
+{
+  asm ("vmovups 0(%0), %%zmm0 \n\t"
+       "vmovups 64(%0), %%zmm1 \n\t"
+       "vmovups 128(%0), %%zmm2 \n\t"
+       "vmovups 192(%0), %%zmm3 \n\t"
+       "vmovups 256(%0), %%zmm4 \n\t"
+       "vmovups 320(%0), %%zmm5 \n\t"
+       "vmovups 384(%0), %%zmm6 \n\t"
+       "vmovups 448(%0), %%zmm7 \n\t"
+       : /* no output operands  */
+       : "r" (zmm_data));
+}
+
+void
+convert_xmm_from_float_to_half (void)
+{
+  asm("vcvtps2phx %xmm1, %xmm0");
+  asm("vcvtps2phx %xmm7, %xmm6");
+}
+
+void
+convert_ymm_from_float_to_half (void)
+{
+  asm("vcvtps2phx %ymm1, %xmm0");
+  asm("vcvtps2phx %ymm7, %xmm6");
+}
+
+void
+convert_zmm_from_float_to_half (void)
+{
+  asm("vcvtps2phx %zmm1, %ymm0");
+  asm("vcvtps2phx %zmm7, %ymm6");
+}
+
+int
+main ()
+{
+  /* Move initial values from array to registers and read from XMM regs.  */
+  move_data_to_xmm_reg ();
+  convert_xmm_from_float_to_half ();
+  asm ("nop"); /* first breakpoint here  */
+
+  /* Move initial values from array to registers and read from YMM regs.  */
+  move_data_to_ymm_reg ();
+  convert_ymm_from_float_to_half ();
+  asm ("nop"); /* second breakpoint here  */
+
+  /* Move initial values from array to registers and read from ZMM regs.  */
+  move_data_to_zmm_reg ();
+  convert_zmm_from_float_to_half ();
+  asm ("nop"); /* third breakpoint here  */
+
+  return 0;
+}
diff --git a/gdb/testsuite/gdb.arch/x86-avx512fp16.exp b/gdb/testsuite/gdb.arch/x86-avx512fp16.exp
new file mode 100644
index 00000000000..22575cc2fe8
--- /dev/null
+++ b/gdb/testsuite/gdb.arch/x86-avx512fp16.exp
@@ -0,0 +1,68 @@
+# Copyright 2021 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/>.
+
+# Test fp16 support in AVX512 registers.
+
+if { [skip_avx512fp16_tests] } {
+    unsupported "target does not support AVX512fp16"
+    return -1
+}
+
+standard_testfile
+
+if { [prepare_for_testing "failed to prepare" ${testfile} ${srcfile}] } {
+    return -1
+}
+
+if { ![runto_main] } {
+    unsupported "could not run to main"
+    return -1
+}
+
+# Test xmm.
+set line1 [gdb_get_line_number "first breakpoint here"]
+gdb_breakpoint $line1
+gdb_continue_to_breakpoint "line1" ".*$srcfile:$line1.*"
+
+gdb_test "print \$xmm0.v8_half" \
+    "= \\{0.5, 0.625, 0.75, 0.875, 0, 0, 0, 0\\}"
+
+gdb_test "print \$xmm6.v8_half" \
+    "= \\{3.5, 3.625, 3.75, 3.875, 0, 0, 0, 0\\}"
+
+# Test ymm.
+set line2 [gdb_get_line_number "second breakpoint here"]
+gdb_breakpoint $line2
+gdb_continue_to_breakpoint "line2" ".*$srcfile:$line2.*"
+
+gdb_test "print \$ymm0.v16_half\[1\]" "= 10.25"
+gdb_test "print \$ymm6.v16_half\[1\]" "= 22.25"
+
+# Test zmm.
+set line3 [gdb_get_line_number "third breakpoint here"]
+gdb_breakpoint $line3
+gdb_continue_to_breakpoint "line3" ".*$srcfile:$line3.*"
+
+gdb_test "print \$zmm0.v32_half\[1\]" "= 28.5"
+gdb_test "print \$zmm6.v32_half\[1\]" "= 76.5"
+
+# Test setting of half values.
+gdb_test_no_output "set var \$xmm0.v8_half\[0\] = 32.25"
+gdb_test_no_output "set var \$ymm3.v16_half\[1\] = 33.5"
+gdb_test_no_output "set var \$zmm7.v32_half\[2\] = 22.75"
+
+gdb_test "p \$xmm0.v8_half\[0\]" "= 32.25"
+gdb_test "p \$ymm3.v16_half\[1\]" "= 33.5"
+gdb_test "p \$zmm7.v32_half\[2\]" "= 22.75"
diff --git a/gdb/testsuite/lib/gdb.exp b/gdb/testsuite/lib/gdb.exp
index 2f95989440a..b38603aaded 100644
--- a/gdb/testsuite/lib/gdb.exp
+++ b/gdb/testsuite/lib/gdb.exp
@@ -3384,6 +3384,57 @@ gdb_caching_proc skip_avx512bf16_tests {
     return $skip_avx512bf16_tests
 }
 
+# Run a test on the target to see if it supports avx512fp16.  Return 0 if so,
+# 1 if it does not.  Based on 'check_vmx_hw_available' from the GCC testsuite.
+
+gdb_caching_proc skip_avx512fp16_tests {
+    global srcdir subdir gdb_prompt inferior_exited_re
+
+    set me "skip_avx512fp16_tests"
+    if { ![istarget "i?86-*-*"] && ![istarget "x86_64-*-*"] } {
+        verbose "$me:  target does not support avx512fp16, returning 1" 2
+        return 1
+    }
+
+    # Compile a test program.
+    set src {
+        int main() {
+            asm volatile ("vcvtps2phx %xmm1, %xmm0");
+            return 0;
+        }
+    }
+    if {![gdb_simple_compile $me $src executable]} {
+        return 1
+    }
+
+    # No error message, compilation succeeded so now run it via gdb.
+
+    gdb_exit
+    gdb_start
+    gdb_reinitialize_dir $srcdir/$subdir
+    gdb_load "$obj"
+    gdb_run_cmd
+    gdb_expect {
+        -re ".*Illegal instruction.*${gdb_prompt} $" {
+            verbose -log "$me:  avx512fp16 hardware not detected."
+            set skip_avx512fp16_tests 1
+        }
+        -re ".*$inferior_exited_re normally.*${gdb_prompt} $" {
+            verbose -log "$me:  avx512fp16 hardware detected."
+            set skip_avx512fp16_tests 0
+        }
+        default {
+            warning "\n$me:  default case taken."
+            set skip_avx512fp16_tests 1
+        }
+    }
+    gdb_exit
+    remote_file build delete $obj
+
+    verbose "$me:  returning $skip_avx512fp16_tests" 2
+    return $skip_avx512fp16_tests
+}
+
 # Run a test on the target to see if it supports btrace hardware.  Return 0 if so,
 # 1 if it does not.  Based on 'check_vmx_hw_available' from the GCC testsuite.
 
-- 
2.25.4

Intel Deutschland GmbH
Registered Address: Am Campeon 10, 85579 Neubiberg, Germany
Tel: +49 89 99 8853-0, www.intel.de <http://www.intel.de>
Managing Directors: Christin Eisenschmid, Sharon Heck, Tiffany Doon Silva  
Chairperson of the Supervisory Board: Nicole Lau
Registered Office: Munich
Commercial Register: Amtsgericht Muenchen HRB 186928

[PATCH 2/2] gdb: Enable finish command and inferior calls for _Float16 on amd64 and i386.

[Felix Willgerodt]

Values of type _Float16 and _Float16 _Complex can now be used on CPUs with
AVX512-FP16 support. Return values of those types are located in XMM0.
Compiler support for gcc and clang is in progress, see e.g.:
https://gcc.gnu.org/pipermail/gcc-patches/2021-July/574117.html

gdb/ChangeLog:
2021-07-21  Felix Willgerodt  <Felix.Willgerodt@intel.com>

	* amd64-tdep.c (amd64_classify): Classify _Float16 and
	_Float16 _Complex as AMD64_SSE.
	* i386-tdep.c (i386_extract_return_value): Read _Float16 and
	_Float16 _Complex from xmm0.

gdb/testsuite/ChangeLog:
2021-07-21  Felix Willgerodt  <Felix.Willgerodt@intel.com>

	* gdb.arch/x86-avx512fp16-abi.c: New file.
	* gdb.arch/x86-avx512fp16-abi.exp: New file.
---
 gdb/amd64-tdep.c                              | 12 ++--
 gdb/i386-tdep.c                               |  9 ++-
 gdb/testsuite/gdb.arch/x86-avx512fp16-abi.c   | 38 +++++++++++
 gdb/testsuite/gdb.arch/x86-avx512fp16-abi.exp | 63 +++++++++++++++++++
 4 files changed, 115 insertions(+), 7 deletions(-)
 create mode 100644 gdb/testsuite/gdb.arch/x86-avx512fp16-abi.c
 create mode 100644 gdb/testsuite/gdb.arch/x86-avx512fp16-abi.exp

diff --git a/gdb/amd64-tdep.c b/gdb/amd64-tdep.c
index c028e1bd0f6..129f07e598d 100644
--- a/gdb/amd64-tdep.c
+++ b/gdb/amd64-tdep.c
@@ -728,10 +728,10 @@ amd64_classify (struct type *type, enum amd64_reg_class theclass[2])
       && (len == 1 || len == 2 || len == 4 || len == 8))
     theclass[0] = AMD64_INTEGER;
 
-  /* Arguments of types float, double, _Decimal32, _Decimal64 and __m64
-     are in class SSE.  */
+  /* Arguments of types _Float16, float, double, _Decimal32, _Decimal64 and
+     __m64 are in class SSE.  */
   else if ((code == TYPE_CODE_FLT || code == TYPE_CODE_DECFLOAT)
-	   && (len == 4 || len == 8))
+	   && (len == 2 || len == 4 || len == 8))
     /* FIXME: __m64 .  */
     theclass[0] = AMD64_SSE;
 
@@ -749,8 +749,8 @@ amd64_classify (struct type *type, enum amd64_reg_class theclass[2])
     /* Class X87 and X87UP.  */
     theclass[0] = AMD64_X87, theclass[1] = AMD64_X87UP;
 
-  /* Arguments of complex T where T is one of the types float or
-     double get treated as if they are implemented as:
+  /* Arguments of complex T - where T is one of the types _Float16, float or
+     double - get treated as if they are implemented as:
 
      struct complexT {
        T real;
@@ -758,7 +758,7 @@ amd64_classify (struct type *type, enum amd64_reg_class theclass[2])
      };
 
   */
-  else if (code == TYPE_CODE_COMPLEX && len == 8)
+  else if (code == TYPE_CODE_COMPLEX && (len == 8 || len == 4))
     theclass[0] = AMD64_SSE;
   else if (code == TYPE_CODE_COMPLEX && len == 16)
     theclass[0] = theclass[1] = AMD64_SSE;
diff --git a/gdb/i386-tdep.c b/gdb/i386-tdep.c
index c2835a2458d..6c7f7f7592d 100644
--- a/gdb/i386-tdep.c
+++ b/gdb/i386-tdep.c
@@ -2818,7 +2818,14 @@ i386_extract_return_value (struct gdbarch *gdbarch, struct type *type,
   int len = TYPE_LENGTH (type);
   gdb_byte buf[I386_MAX_REGISTER_SIZE];
 
-  if (type->code () == TYPE_CODE_FLT)
+  /* _FLoat16 and _Float16 _Complex values are returned via xmm0.  */
+  if (((type->code () == TYPE_CODE_FLT) && len == 2)
+      || ((type->code () == TYPE_CODE_COMPLEX) && len == 4))
+    {
+	regcache->raw_read (I387_XMM0_REGNUM (tdep), valbuf);
+	return;
+    }
+  else if (type->code () == TYPE_CODE_FLT)
     {
       if (tdep->st0_regnum < 0)
 	{
diff --git a/gdb/testsuite/gdb.arch/x86-avx512fp16-abi.c b/gdb/testsuite/gdb.arch/x86-avx512fp16-abi.c
new file mode 100644
index 00000000000..c5b188ea546
--- /dev/null
+++ b/gdb/testsuite/gdb.arch/x86-avx512fp16-abi.c
@@ -0,0 +1,38 @@
+/* Test program for bfloat16 of AVX 512 registers.
+
+   Copyright 2021 Free Software Foundation, Inc.
+
+   This file is part of GDB.
+
+   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 <complex.h>
+
+_Float16
+square (_Float16 num) {
+  return num * num; /* BP1.  */
+}
+
+_Float16 _Complex
+plus (_Float16 _Complex num) {
+  return num + (2.5 + 0.5I); /* BP2.  */
+}
+
+int
+main ()
+{
+  _Float16 a = square (1.25);
+  _Float16 _Complex b = 6.25 + I;
+  _Float16 _Complex ret = plus (b); /* BP3.  */
+  return 0;
+}
diff --git a/gdb/testsuite/gdb.arch/x86-avx512fp16-abi.exp b/gdb/testsuite/gdb.arch/x86-avx512fp16-abi.exp
new file mode 100644
index 00000000000..61cd189a3dd
--- /dev/null
+++ b/gdb/testsuite/gdb.arch/x86-avx512fp16-abi.exp
@@ -0,0 +1,63 @@
+# Copyright 2021 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/>.
+
+# Test support for _Float16 parameters and return values.
+
+if { [skip_avx512fp16_tests] } {
+    unsupported "target does not support AVX512fp16"
+    return -1
+}
+
+standard_testfile
+
+if { [prepare_for_testing "failed to prepare" ${testfile} ${srcfile} \
+     {debug additional_flags="-mavx512fp16"}] } {
+    return -1
+}
+
+if { ![runto_main] } {
+    unsupported "could not run to main"
+    return -1
+}
+
+gdb_test "p square(2.2)" "= 4\\.8359"
+
+set line1 [gdb_get_line_number "BP1"]
+set line2 [gdb_get_line_number "BP2"]
+set line3 [gdb_get_line_number "BP3"]
+gdb_breakpoint $line1
+gdb_breakpoint $line3
+
+gdb_continue_to_breakpoint "line1" ".*$srcfile:$line1.*"
+
+with_test_prefix "real" {
+    gdb_test "p num" "= 1\\.25"
+    gdb_test "ptype num" "type = _Float16"
+    gdb_test "finish" "Value returned is.*= 1\\.5625"
+}
+
+gdb_continue_to_breakpoint "line3" ".*$srcfile:$line3.*"
+gdb_test "p plus(b)" "= 8\\.75 \\+ 1\\.5i"
+
+gdb_breakpoint $line2
+gdb_continue_to_breakpoint "line2" ".*$srcfile:$line2.*"
+
+with_test_prefix "complex" {
+    gdb_test "p num" "= 6\\.25 \\+ 1i"
+    gdb_test "ptype num" "type = complex _Float16"
+    gdb_test "finish" "Value returned is.*= 8\\.75 \\+ 1\\.5i"
+}
+
+gdb_continue_to_end
-- 
2.25.4

Intel Deutschland GmbH
Registered Address: Am Campeon 10, 85579 Neubiberg, Germany
Tel: +49 89 99 8853-0, www.intel.de <http://www.intel.de>
Managing Directors: Christin Eisenschmid, Sharon Heck, Tiffany Doon Silva  
Chairperson of the Supervisory Board: Nicole Lau
Registered Office: Munich
Commercial Register: Amtsgericht Muenchen HRB 186928

Re: [PATCH 2/2] gdb: Enable finish command and inferior calls for _Float16 on amd64 and i386.

[Luis Machado]

Hi,

On 7/29/21 5:38 AM, Felix Willgerodt via Gdb-patches wrote:
> Values of type _Float16 and _Float16 _Complex can now be used on CPUs with
> AVX512-FP16 support. Return values of those types are located in XMM0.
> Compiler support for gcc and clang is in progress, see e.g.:
> https://gcc.gnu.org/pipermail/gcc-patches/2021-July/574117.html

Is _Float16 equivalent to fp16? If so, the presence of yet another name 
for a fp16 type is a bit unfortunate.

> 
> gdb/ChangeLog:
> 2021-07-21  Felix Willgerodt  <Felix.Willgerodt@intel.com>
> 
> 	* amd64-tdep.c (amd64_classify): Classify _Float16 and
> 	_Float16 _Complex as AMD64_SSE.
> 	* i386-tdep.c (i386_extract_return_value): Read _Float16 and
> 	_Float16 _Complex from xmm0.
> 
> gdb/testsuite/ChangeLog:
> 2021-07-21  Felix Willgerodt  <Felix.Willgerodt@intel.com>
> 
> 	* gdb.arch/x86-avx512fp16-abi.c: New file.
> 	* gdb.arch/x86-avx512fp16-abi.exp: New file.
> ---
>   gdb/amd64-tdep.c                              | 12 ++--
>   gdb/i386-tdep.c                               |  9 ++-
>   gdb/testsuite/gdb.arch/x86-avx512fp16-abi.c   | 38 +++++++++++
>   gdb/testsuite/gdb.arch/x86-avx512fp16-abi.exp | 63 +++++++++++++++++++
>   4 files changed, 115 insertions(+), 7 deletions(-)
>   create mode 100644 gdb/testsuite/gdb.arch/x86-avx512fp16-abi.c
>   create mode 100644 gdb/testsuite/gdb.arch/x86-avx512fp16-abi.exp
> 
> diff --git a/gdb/amd64-tdep.c b/gdb/amd64-tdep.c
> index c028e1bd0f6..129f07e598d 100644
> --- a/gdb/amd64-tdep.c
> +++ b/gdb/amd64-tdep.c
> @@ -728,10 +728,10 @@ amd64_classify (struct type *type, enum amd64_reg_class theclass[2])
>         && (len == 1 || len == 2 || len == 4 || len == 8))
>       theclass[0] = AMD64_INTEGER;
>   
> -  /* Arguments of types float, double, _Decimal32, _Decimal64 and __m64
> -     are in class SSE.  */
> +  /* Arguments of types _Float16, float, double, _Decimal32, _Decimal64 and
> +     __m64 are in class SSE.  */
>     else if ((code == TYPE_CODE_FLT || code == TYPE_CODE_DECFLOAT)
> -	   && (len == 4 || len == 8))
> +	   && (len == 2 || len == 4 || len == 8))
>       /* FIXME: __m64 .  */
>       theclass[0] = AMD64_SSE;
>   
> @@ -749,8 +749,8 @@ amd64_classify (struct type *type, enum amd64_reg_class theclass[2])
>       /* Class X87 and X87UP.  */
>       theclass[0] = AMD64_X87, theclass[1] = AMD64_X87UP;
>   
> -  /* Arguments of complex T where T is one of the types float or
> -     double get treated as if they are implemented as:
> +  /* Arguments of complex T - where T is one of the types _Float16, float or
> +     double - get treated as if they are implemented as:
>   
>        struct complexT {
>          T real;
> @@ -758,7 +758,7 @@ amd64_classify (struct type *type, enum amd64_reg_class theclass[2])
>        };
>   
>     */
> -  else if (code == TYPE_CODE_COMPLEX && len == 8)
> +  else if (code == TYPE_CODE_COMPLEX && (len == 8 || len == 4))
>       theclass[0] = AMD64_SSE;
>     else if (code == TYPE_CODE_COMPLEX && len == 16)
>       theclass[0] = theclass[1] = AMD64_SSE;
> diff --git a/gdb/i386-tdep.c b/gdb/i386-tdep.c
> index c2835a2458d..6c7f7f7592d 100644
> --- a/gdb/i386-tdep.c
> +++ b/gdb/i386-tdep.c
> @@ -2818,7 +2818,14 @@ i386_extract_return_value (struct gdbarch *gdbarch, struct type *type,
>     int len = TYPE_LENGTH (type);
>     gdb_byte buf[I386_MAX_REGISTER_SIZE];
>   
> -  if (type->code () == TYPE_CODE_FLT)
> +  /* _FLoat16 and _Float16 _Complex values are returned via xmm0.  */

_Float16

> +  if (((type->code () == TYPE_CODE_FLT) && len == 2)
> +      || ((type->code () == TYPE_CODE_COMPLEX) && len == 4))
> +    {
> +	regcache->raw_read (I387_XMM0_REGNUM (tdep), valbuf);
> +	return;
> +    }
> +  else if (type->code () == TYPE_CODE_FLT)
>       {
>         if (tdep->st0_regnum < 0)
>   	{
> diff --git a/gdb/testsuite/gdb.arch/x86-avx512fp16-abi.c b/gdb/testsuite/gdb.arch/x86-avx512fp16-abi.c
> new file mode 100644
> index 00000000000..c5b188ea546
> --- /dev/null
> +++ b/gdb/testsuite/gdb.arch/x86-avx512fp16-abi.c
> @@ -0,0 +1,38 @@
> +/* Test program for bfloat16 of AVX 512 registers.

float16 and not bfloat16, right?

> +
> +   Copyright 2021 Free Software Foundation, Inc.
> +
> +   This file is part of GDB.
> +
> +   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 <complex.h>
> +
> +_Float16
> +square (_Float16 num) {
> +  return num * num; /* BP1.  */
> +}
> +
> +_Float16 _Complex
> +plus (_Float16 _Complex num) {
> +  return num + (2.5 + 0.5I); /* BP2.  */
> +}
> +
> +int
> +main ()
> +{
> +  _Float16 a = square (1.25);
> +  _Float16 _Complex b = 6.25 + I;
> +  _Float16 _Complex ret = plus (b); /* BP3.  */
> +  return 0;
> +}
> diff --git a/gdb/testsuite/gdb.arch/x86-avx512fp16-abi.exp b/gdb/testsuite/gdb.arch/x86-avx512fp16-abi.exp
> new file mode 100644
> index 00000000000..61cd189a3dd
> --- /dev/null
> +++ b/gdb/testsuite/gdb.arch/x86-avx512fp16-abi.exp
> @@ -0,0 +1,63 @@
> +# Copyright 2021 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/>.
> +
> +# Test support for _Float16 parameters and return values.
> +
> +if { [skip_avx512fp16_tests] } {
> +    unsupported "target does not support AVX512fp16"
> +    return -1
> +}
> +
> +standard_testfile
> +
> +if { [prepare_for_testing "failed to prepare" ${testfile} ${srcfile} \
> +     {debug additional_flags="-mavx512fp16"}] } {
> +    return -1
> +}
> +
> +if { ![runto_main] } {
> +    unsupported "could not run to main"
> +    return -1
> +}
> +
> +gdb_test "p square(2.2)" "= 4\\.8359"
> +
> +set line1 [gdb_get_line_number "BP1"]
> +set line2 [gdb_get_line_number "BP2"]
> +set line3 [gdb_get_line_number "BP3"]
> +gdb_breakpoint $line1
> +gdb_breakpoint $line3
> +
> +gdb_continue_to_breakpoint "line1" ".*$srcfile:$line1.*"
> +
> +with_test_prefix "real" {
> +    gdb_test "p num" "= 1\\.25"
> +    gdb_test "ptype num" "type = _Float16"
> +    gdb_test "finish" "Value returned is.*= 1\\.5625"
> +}
> +
> +gdb_continue_to_breakpoint "line3" ".*$srcfile:$line3.*"
> +gdb_test "p plus(b)" "= 8\\.75 \\+ 1\\.5i"
> +
> +gdb_breakpoint $line2
> +gdb_continue_to_breakpoint "line2" ".*$srcfile:$line2.*"
> +
> +with_test_prefix "complex" {
> +    gdb_test "p num" "= 6\\.25 \\+ 1i"
> +    gdb_test "ptype num" "type = complex _Float16"
> +    gdb_test "finish" "Value returned is.*= 8\\.75 \\+ 1\\.5i"
> +}
> +
> +gdb_continue_to_end
> 

RE: [PATCH 2/2] gdb: Enable finish command and inferior calls for _Float16 on amd64 and i386.

[Willgerodt, Felix]

> Is _Float16 equivalent to fp16? If so, the presence of yet another name for a
> fp16 type is a bit unfortunate.
> 

They are the same usually, but on ARM the story is apparently a bit more complex:
https://gcc.gnu.org/onlinedocs/gcc/Half-Precision.html

But to cite the link:
" The __fp16 type may only be used as an argument to intrinsics defined in
<arm_fp16.h>, or as a storage format."

And:
"It is recommended that portable code use the _Float16 type defined by
ISO/IEC TS 18661-3:2015."

Therefore I don't think it is that unfortunate and not really negotiable by now.
Both GCC and clang know about both already.

> > -  if (type->code () == TYPE_CODE_FLT)
> > +  /* _FLoat16 and _Float16 _Complex values are returned via xmm0.  */
> 
> _Float16
>

Thanks for spotting this. I fixed it locally.

> > +/* Test program for bfloat16 of AVX 512 registers.
> 
> float16 and not bfloat16, right?
> 

Thanks. I reworded it to say:
"Test program for _Float16 parameters and return values."

Regards,
Felix
Intel Deutschland GmbH
Registered Address: Am Campeon 10, 85579 Neubiberg, Germany
Tel: +49 89 99 8853-0, www.intel.de <http://www.intel.de>
Managing Directors: Christin Eisenschmid, Sharon Heck, Tiffany Doon Silva  
Chairperson of the Supervisory Board: Nicole Lau
Registered Office: Munich
Commercial Register: Amtsgericht Muenchen HRB 186928

Re: [PATCH 2/2] gdb: Enable finish command and inferior calls for _Float16 on amd64 and i386.

[Luis Machado]

On 7/30/21 4:52 AM, Willgerodt, Felix wrote:
>> Is _Float16 equivalent to fp16? If so, the presence of yet another name for a
>> fp16 type is a bit unfortunate.
>>
> 
> They are the same usually, but on ARM the story is apparently a bit more complex:
> https://gcc.gnu.org/onlinedocs/gcc/Half-Precision.html

It is. :-)

> 
> But to cite the link:
> " The __fp16 type may only be used as an argument to intrinsics defined in
> <arm_fp16.h>, or as a storage format."
> 
> And:
> "It is recommended that portable code use the _Float16 type defined by
> ISO/IEC TS 18661-3:2015."
> 
> Therefore I don't think it is that unfortunate and not really negotiable by now.
> Both GCC and clang know about both already.

The unfortunate bit is the lack of enough DWARF information to make the 
distinction between different float formats of the same size. If the 
compilers are OK with the naming and already support it, it's not a big 
deal then.

> 
>>> -  if (type->code () == TYPE_CODE_FLT)
>>> +  /* _FLoat16 and _Float16 _Complex values are returned via xmm0.  */
>>
>> _Float16
>>
> 
> Thanks for spotting this. I fixed it locally.
> 
>>> +/* Test program for bfloat16 of AVX 512 registers.
>>
>> float16 and not bfloat16, right?
>>
> 
> Thanks. I reworded it to say:
> "Test program for _Float16 parameters and return values."
> 
> Regards,
> Felix
> Intel Deutschland GmbH
> Registered Address: Am Campeon 10, 85579 Neubiberg, Germany
> Tel: +49 89 99 8853-0, www.intel.de <http://www.intel.de>
> Managing Directors: Christin Eisenschmid, Sharon Heck, Tiffany Doon Silva
> Chairperson of the Supervisory Board: Nicole Lau
> Registered Office: Munich
> Commercial Register: Amtsgericht Muenchen HRB 186928
> 

Re: [PATCH 1/2] Add half support for AVX512 register view.

[Simon Marchi]

> diff --git a/gdb/features/i386/32bit-sse.c b/gdb/features/i386/32bit-sse.c
> index e5eaf8feb59..5554849a6ca 100644
> --- a/gdb/features/i386/32bit-sse.c
> +++ b/gdb/features/i386/32bit-sse.c
> @@ -10,6 +10,9 @@ create_feature_i386_32bit_sse (struct target_desc *result, long regnum)
>  
>    feature = tdesc_create_feature (result, "org.gnu.gdb.i386.sse");
>    tdesc_type *element_type;
> +  element_type = tdesc_named_type (feature, "ieee_half");

ieee_half isn't listed on this page:

https://sourceware.org/gdb/current/onlinedocs/gdb/Predefined-Target-Types.html

Should it?

Otherwise, LGTM.

Simon

Re: [PATCH 2/2] gdb: Enable finish command and inferior calls for _Float16 on amd64 and i386.

[Simon Marchi]

On 2021-07-30 8:04 a.m., Luis Machado via Gdb-patches wrote:
> On 7/30/21 4:52 AM, Willgerodt, Felix wrote:
>>> Is _Float16 equivalent to fp16? If so, the presence of yet another name for a
>>> fp16 type is a bit unfortunate.
>>>
>>
>> They are the same usually, but on ARM the story is apparently a bit more complex:
>> https://gcc.gnu.org/onlinedocs/gcc/Half-Precision.html
> 
> It is. :-)
> 
>>
>> But to cite the link:
>> " The __fp16 type may only be used as an argument to intrinsics defined in
>> <arm_fp16.h>, or as a storage format."
>>
>> And:
>> "It is recommended that portable code use the _Float16 type defined by
>> ISO/IEC TS 18661-3:2015."
>>
>> Therefore I don't think it is that unfortunate and not really negotiable by now.
>> Both GCC and clang know about both already.
> 
> The unfortunate bit is the lack of enough DWARF information to make the distinction between different float formats of the same size. If the compilers are OK with the naming and already support it, it's not a big deal then.

I didn't find anything else to change in addition to what Luis already
reported, so the patch is OK what that fixed.

Simon

RE: [PATCH 1/2] Add half support for AVX512 register view.

[Willgerodt, Felix]

> ieee_half isn't listed on this page:
> 
> https://sourceware.org/gdb/current/onlinedocs/gdb/Predefined-Target-
> Types.html
> 
> Should it?

It probably should. It already existed before my patches though, for both ARM and
x86. Is it ok if I post a separate patch to add it (and bfloat16 as well I guess)?

Thanks,
Felix
Intel Deutschland GmbH
Registered Address: Am Campeon 10, 85579 Neubiberg, Germany
Tel: +49 89 99 8853-0, www.intel.de <http://www.intel.de>
Managing Directors: Christin Eisenschmid, Sharon Heck, Tiffany Doon Silva  
Chairperson of the Supervisory Board: Nicole Lau
Registered Office: Munich
Commercial Register: Amtsgericht Muenchen HRB 186928

Re: [PATCH 1/2] Add half support for AVX512 register view.

[Simon Marchi]

On 2021-09-03 8:55 a.m., Willgerodt, Felix wrote:
>> ieee_half isn't listed on this page:
>>
>> https://sourceware.org/gdb/current/onlinedocs/gdb/Predefined-Target-
>> Types.html
>>
>> Should it?
> 
> It probably should. It already existed before my patches though, for both ARM and
> x86. Is it ok if I post a separate patch to add it (and bfloat16 as well I guess)?

Yes, no problem, thanks in advance!

Simon

RE: [PATCH 2/2] gdb: Enable finish command and inferior calls for _Float16 on amd64 and i386.

[Willgerodt, Felix]

> -----Original Message-----
> From: Simon Marchi <simon.marchi@polymtl.ca>
> Sent: Freitag, 3. September 2021 14:37
> To: Luis Machado <luis.machado@linaro.org>; Willgerodt, Felix
> <felix.willgerodt@intel.com>; gdb-patches@sourceware.org
> Subject: Re: [PATCH 2/2] gdb: Enable finish command and inferior calls for
> _Float16 on amd64 and i386.
> 
> On 2021-07-30 8:04 a.m., Luis Machado via Gdb-patches wrote:
> > On 7/30/21 4:52 AM, Willgerodt, Felix wrote:
> >>> Is _Float16 equivalent to fp16? If so, the presence of yet another name for a
> >>> fp16 type is a bit unfortunate.
> >>>
> >>
> >> They are the same usually, but on ARM the story is apparently a bit more
> complex:
> >> https://gcc.gnu.org/onlinedocs/gcc/Half-Precision.html
> >
> > It is. :-)
> >
> >>
> >> But to cite the link:
> >> " The __fp16 type may only be used as an argument to intrinsics defined in
> >> <arm_fp16.h>, or as a storage format."
> >>
> >> And:
> >> "It is recommended that portable code use the _Float16 type defined by
> >> ISO/IEC TS 18661-3:2015."
> >>
> >> Therefore I don't think it is that unfortunate and not really negotiable by
> now.
> >> Both GCC and clang know about both already.
> >
> > The unfortunate bit is the lack of enough DWARF information to make the
> distinction between different float formats of the same size. If the compilers are
> OK with the naming and already support it, it's not a big deal then.
> 
> I didn't find anything else to change in addition to what Luis already
> reported, so the patch is OK what that fixed.
> 
> Simon

Thanks! I pushed both patches. I will try to post the documentation patch
(comment from patch 1) hopefully later today or next week.

Regards,
Felix
Intel Deutschland GmbH
Registered Address: Am Campeon 10, 85579 Neubiberg, Germany
Tel: +49 89 99 8853-0, www.intel.de <http://www.intel.de>
Managing Directors: Christin Eisenschmid, Sharon Heck, Tiffany Doon Silva  
Chairperson of the Supervisory Board: Nicole Lau
Registered Office: Munich
Commercial Register: Amtsgericht Muenchen HRB 186928