Skip to content

Commit 8ca1398

Browse files
kc9judkc9jud
committed
Merge branch 'us05-fp-interop' into us05-fp-interop
2 parents d4187a6 + fc23168 commit 8ca1398

File tree

1 file changed

+156
-0
lines changed

1 file changed

+156
-0
lines changed

‎proposals/fp_interop/25-189r1.txt‎

Lines changed: 156 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,156 @@
1+
To: J3 J3/##-###
2+
From: Patrick Fasano
3+
Subject: Requirements and specifications for interoperability with C
4+
interchange and extended types
5+
Date: 2025-Oct-XX
6+
7+
References:
8+
ISO/IEC 9899:2024 Programming languages -- C ("C23")
9+
(working draft N3220)
10+
WG5/N2249 Fortran 202Y Work Items
11+
23-176 Add extended floating-point types defined in ISO C23
12+
to ISO_C_BINDING
13+
ISO/IEC 60559:2020 Floating-Point arithmetic
14+
IEEE 754-2019 IEEE Standard for Floating-Point Arithmetic
15+
16+
17+
1 Background
18+
============
19+
20+
bg01. The ISO C23 specification (ISO/IEC 9899:2024) added support for
21+
interchange and extended floating-point types (C23 Annex H). The
22+
current Fortran 202Y work list WG5/N2249 includes adding
23+
interoperability with these types as accepted work item US05. At
24+
meeting #230 in June 2023, J3 accepted paper 23-176 as
25+
background/use-case for additional interchange types.
26+
27+
bg03. This paper contains formal requirements and specifications for
28+
Fortran 202Y proposal US05, "Add extended floating-point types
29+
defined in ISO C23 to ISO_C_BINDING".
30+
31+
2 Requirements
32+
==============
33+
34+
rq01. The C23 standard adds a large number of new types, importing
35+
significant portions of ISO/IEC 60559 (a.k.a IEEE 754) into its
36+
specification. The most important of these for Fortran
37+
interoperability with C are the '_Float16', '_Float32',
38+
'_Float64', and '_Float128' types (and their related complex
39+
types). Requirements rq02-r05 deal with these new types. Many of
40+
the other types, covered by rq07-rq17, are marked as optional and
41+
could be dropped independently while keeping the rest of this
42+
paper. These optional requirements primarily serve to maintain
43+
parity with Annex H of C23.
44+
45+
rq02. The ISO_C_BINDING module should provide integer named constants
46+
for KIND parameters corresponding to the C interchange floating
47+
types '_Float16', '_Float32', '_Float64', and '_Float128'.
48+
49+
rq03. If the companion processor defines '__STDC_IEC_60559_BFP__' and
50+
'__STDC_IEC_60559_TYPES__', then the KIND parameters corresponding
51+
to '_Float32' and '_Float64' should equal those corresponding to
52+
'float' and 'double', respectively.
53+
54+
rq05. The ISO_C_BINDING module should provide integer named constants
55+
for KIND parameters corresponding to the C interchange complex
56+
types '_Float16_Complex', '_Float32_Complex', '_Float64_Complex',
57+
and '_Float128_Complex'. The values of the named constants should
58+
equal those of the respective constants for the interchange
59+
floating types.
60+
61+
rq07. The ISO_C_BINDING module should provide integer named constants
62+
for KIND parameters corresponding to the C extended floating types
63+
'_Float32x', '_Float64x', and '_Float128x'.
64+
65+
rq09. The ISO_C_BINDING module should provide integer named constants
66+
for KIND parameters corresponding to the C extended complex types
67+
'_Float32x_Complex', '_Float64x_Complex', and
68+
'_Float128x_Complex'. The values of the named constants should
69+
match those of the respective constants for the extended floating
70+
types.
71+
72+
rq11. The ISO_C_BINDING module should provide integer named constants
73+
for KIND parameters corresponding to the C interchange decimal
74+
floating types '_Decimal32', '_Decimal64', and '_Decimal128'.
75+
76+
3 Specifications
77+
================
78+
79+
sp01. For all integer named constants to be added to the ISO_C_BINDING
80+
module, the value shall be a valid real kind type parameter or
81+
shall be -1 if the companion processor’s type does not have a
82+
precision equal to the precision of any of the Fortran processor’s
83+
real kinds, −2 if the companion processor’s type does not have a
84+
range equal to the range of any of the Fortran processor’s real
85+
kinds, −3 if the companion processor's type has neither the
86+
precision nor range of any of the Fortran processor’s real kinds,
87+
−4 if there is no interoperating Fortran processor kind for other
88+
reasons, and -5 if the companion processor does not define the
89+
corresponding C type.
90+
91+
sp02. To the ISO_C_BINDING module, add the following integer named
92+
constants for real KIND parameters:
93+
94+
+----------------+--------------------+
95+
| Named constant | C interchange type |
96+
+----------------+--------------------+
97+
| C_FLOAT16 | _Float16 |
98+
| C_FLOAT32 | _Float32 |
99+
| C_FLOAT64 | _Float64 |
100+
| C_FLOAT128 | _Float128 |
101+
+----------------+--------------------+
102+
103+
If the companion processor defines '__STDC_IEC_60559_BFP__' and
104+
'__STDC_IEC_60559_TYPES__', then the constants 'C_FLOAT32' and
105+
'C_FLOAT64' should have the same values as 'C_FLOAT' and
106+
'C_DOUBLE', respectively (C2023 H.2.4.2).
107+
108+
sp05. To the ISO_C_BINDING module, add the following integer named
109+
constants for real KIND parameters:
110+
111+
+--------------------+--------------------+
112+
| Named constant | C interchange type |
113+
+--------------------+--------------------+
114+
| C_FLOAT16_COMPLEX | _Float16_Complex |
115+
| C_FLOAT32_COMPLEX | _Float32_Complex |
116+
| C_FLOAT64_COMPLEX | _Float64_Complex |
117+
| C_FLOAT128_COMPLEX | _Float128_Complex |
118+
+--------------------+--------------------+
119+
120+
The values of each of these named constants shall be the same as
121+
the respective constants defined in sp01.
122+
123+
sp07. To the ISO_C_BINDING module, add the following integer named
124+
constants for real KIND parameters:
125+
126+
+----------------+-----------------+
127+
| Named constant | C extended type |
128+
+----------------+-----------------+
129+
| C_FLOAT32X | _Float32x |
130+
| C_FLOAT64X | _Float64x |
131+
| C_FLOAT128X | _Float128x |
132+
+----------------+-----------------+
133+
134+
sp09. To the ISO_C_BINDING module, add the following integer named
135+
constants for real KIND parameters:
136+
137+
+---------------------+--------------------+
138+
| Named constant | C extended type |
139+
+---------------------+--------------------+
140+
| C_FLOAT32X_COMPLEX | _Float32x_Complex |
141+
| C_FLOAT64X_COMPLEX | _Float64x_Complex |
142+
| C_FLOAT128X_COMPLEX | _Float128x_Complex |
143+
+---------------------+--------------------+
144+
145+
sp11. To the ISO_C_BINDING module, add the following integer named
146+
constants for real KIND parameters:
147+
148+
+----------------+--------------------+
149+
| Named constant | C interchange type |
150+
+----------------+--------------------+
151+
| C_DECIMAL32 | _Decimal32 |
152+
| C_DECIMAL64 | _Decimal64 |
153+
| C_DECIMAL128 | _Decimal128 |
154+
+----------------+--------------------+
155+
156+
=== END ===

0 commit comments

Comments
 (0)