[Patch][binutils][aarch64] .bfloat16 directive for AArch64 [7/10]

Hi,

This patch is part of a series that adds support for Armv8.6-A
(Matrix Multiply and BFloat16 extensions) to binutils.

This patch implements the '.bfloat' directive for the AArch64 backend.

The syntax for the directive is:
  .bfloat16 <0-n numbers>
e.g.
  .bfloat16 12.0
  .bfloat16 0.123, 1.0, NaN, 5

This is implemented by utilizing the ieee_atof_detail function in order
to encode the slightly
different bfloat16 format.

Added testcases to verify the correct encoding for various bfloat16
values (NaN, Infinity (+ & -), normals, subnormals etc...).

Cross compiled and tested on aarch64-none-elf and aarch64-none-linux-gnu
with no issues.

Committed on behalf of Mihail Ionescu.

gas/ChangeLog:

2019-10-29  Mihail Ionescu  <mihail.ionescu@arm.com>
2019-10-29  Barnaby Wilks  <barnaby.wilks@arm.com>

	* config/tc-aarch64.c (md_atof): Add encoding for the bfloat16 format.
	* testsuite/gas/aarch64/bfloat16-directive-le.d: New test.
	* testsuite/gas/aarch64/bfloat16-directive-be.d: New test.
	* testsuite/gas/aarch64/bfloat16-directive.s: New test.

Is it ok for trunk?

Regards,
Mihail

1 files changed, 49 insertions, 0 deletions

diff --git a/gas/config/tc-aarch64.c b/gas/config/tc-aarch64.c
index 527d859..eea863b 100644
--- a/gas/config/tc-aarch64.c
+++ b/gas/config/tc-aarch64.c
@@ -635,6 +635,54 @@ my_get_expression (expressionS * ep, char **str, int prefix_mode,
 const char *
 md_atof (int type, char *litP, int *sizeP)
 {
+  /* If this is a bfloat16 type, then parse it slightly differently -
+     as it does not follow the IEEE standard exactly.  */
+  if (type == 'b')
+    {
+      char * t;
+      LITTLENUM_TYPE words[MAX_LITTLENUMS];
+      FLONUM_TYPE generic_float;
+
+      t = atof_ieee_detail (input_line_pointer, 1, 8, words, &generic_float);
+
+      if (t)
+	input_line_pointer = t;
+      else
+	return _("invalid floating point number");
+
+      switch (generic_float.sign)
+	{
+	/* Is +Inf.  */
+	case 'P':
+	  words[0] = 0x7f80;
+	  break;
+
+	/* Is -Inf.  */
+	case 'N':
+	  words[0] = 0xff80;
+	  break;
+
+	/* Is NaN.  */
+	/* bfloat16 has two types of NaN - quiet and signalling.
+	   Quiet NaN has bit[6] == 1 && faction != 0, whereas
+	   signalling Nan's have bit[0] == 0 && fraction != 0.
+	   Chose this specific encoding as it is the same form
+	   as used by other IEEE 754 encodings in GAS.  */
+	case 0:
+	  words[0] = 0x7fff;
+	  break;
+
+	default:
+	  break;
+	}
+
+      *sizeP = 2;
+
+      md_number_to_chars (litP, (valueT) words[0], sizeof (LITTLENUM_TYPE));
+
+      return NULL;
+    }
+
   return ieee_md_atof (type, litP, sizeP, target_big_endian);
 }
 
@@ -2107,6 +2155,7 @@ const pseudo_typeS md_pseudo_table[] = {
   {"variant_pcs", s_variant_pcs, 0},
 #endif
   {"float16", float_cons, 'h'},
+  {"bfloat16", float_cons, 'b'},
   {0, 0, 0}
 };