1 /* Cray PVP/IEEE mpn_mul_basecase.
3 Copyright 2000, 2001 Free Software Foundation, Inc.
5 This file is part of the GNU MP Library.
7 The GNU MP Library is free software; you can redistribute it and/or modify
8 it under the terms of either:
10 * the GNU Lesser General Public License as published by the Free
11 Software Foundation; either version 3 of the License, or (at your
12 option) any later version.
16 * the GNU General Public License as published by the Free Software
17 Foundation; either version 2 of the License, or (at your option) any
20 or both in parallel, as here.
22 The GNU MP Library is distributed in the hope that it will be useful, but
23 WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
24 or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
27 You should have received copies of the GNU General Public License and the
28 GNU Lesser General Public License along with the GNU MP Library. If not,
29 see https://www.gnu.org/licenses/. */
31 /* The most critical loop of this code runs at about 5 cycles/limb on a T90.
32 That is not perfect, mainly due to vector register shortage. */
34 #include <intrinsics.h>
39 mpn_mul_basecase (mp_ptr rp
,
40 mp_srcptr up
, mp_size_t un
,
41 mp_srcptr vp
, mp_size_t vn
)
43 mp_limb_t cy
[un
+ vn
];
45 mp_limb_t a
, b
, r
, s0
, s1
, c0
, c1
;
49 for (i
= 0; i
< un
+ vn
; i
++)
56 for (j
= 0; j
< vn
; j
++)
64 c0
= ((a
& r
) | ((a
| r
) & ~s0
)) >> 63;
68 for (i
= 1; i
< un
; i
++)
71 b
= _int_mult_upper (up
[i
- 1], vl
);
73 c0
= ((a
& b
) | ((a
| b
) & ~s0
)) >> 63;
77 c1
= ((s0
& r
) | ((s0
| r
) & ~s1
)) >> 63;
80 rp
[j
+ un
] = _int_mult_upper (up
[un
- 1], vl
);
85 for (i
= 1; i
< un
+ vn
; i
++)
94 /* If that second loop generated carry, handle that in scalar loop. */
98 for (i
= 1; i
< un
+ vn
; i
++)
101 c0
= (r
< cy
[i
- 1]);
104 c1
= (r
& ~s0
) >> 63;