Reorganisation of the source tree
[ghc-hetmet.git] / rts / gmp / mpn / alpha / ev5 / sub_n.asm
diff --git a/rts/gmp/mpn/alpha/ev5/sub_n.asm b/rts/gmp/mpn/alpha/ev5/sub_n.asm
new file mode 100644 (file)
index 0000000..5248a2a
--- /dev/null
@@ -0,0 +1,143 @@
+dnl  Alpha EV5 __gmpn_sub_n -- Subtract two limb vectors of the same length > 0
+dnl  and store difference in a third limb vector.
+
+dnl  Copyright (C) 1995, 1999, 2000 Free Software Foundation, Inc.
+
+dnl  This file is part of the GNU MP Library.
+
+dnl  The GNU MP Library is free software; you can redistribute it and/or modify
+dnl  it under the terms of the GNU Lesser General Public License as published by
+dnl  the Free Software Foundation; either version 2.1 of the License, or (at your
+dnl  option) any later version.
+
+dnl  The GNU MP Library is distributed in the hope that it will be useful, but
+dnl  WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
+dnl  or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU Lesser General Public
+dnl  License for more details.
+
+dnl  You should have received a copy of the GNU Lesser General Public License
+dnl  along with the GNU MP Library; see the file COPYING.LIB.  If not, write to
+dnl  the Free Software Foundation, Inc., 59 Temple Place - Suite 330, Boston,
+dnl  MA 02111-1307, USA.
+
+include(`../config.m4')
+
+dnl  INPUT PARAMETERS
+dnl  res_ptr   r16
+dnl  s1_ptr    r17
+dnl  s2_ptr    r18
+dnl  size      r19
+
+ASM_START()
+PROLOGUE(mpn_sub_n)
+       bis     r31,r31,r25             C clear cy
+       subq    r19,4,r19               C decr loop cnt
+       blt     r19,$Lend2              C if less than 4 limbs, goto 2nd loop
+C Start software pipeline for 1st loop
+       ldq     r0,0(r18)
+       ldq     r4,0(r17)
+       ldq     r1,8(r18)
+       ldq     r5,8(r17)
+       addq    r17,32,r17              C update s1_ptr
+       ldq     r2,16(r18)
+       subq    r4,r0,r20               C 1st main subtract
+       ldq     r3,24(r18)
+       subq    r19,4,r19               C decr loop cnt
+       ldq     r6,-16(r17)
+       cmpult  r4,r0,r25               C compute cy from last subtract
+       ldq     r7,-8(r17)
+       subq    r5,r1,r28               C 2nd main subtract
+       addq    r18,32,r18              C update s2_ptr
+       subq    r28,r25,r21             C 2nd carry subtract
+       cmpult  r5,r1,r8                C compute cy from last subtract
+       blt     r19,$Lend1              C if less than 4 limbs remain, jump
+C 1st loop handles groups of 4 limbs in a software pipeline
+       ALIGN(16)
+$Loop: cmpult  r28,r25,r25             C compute cy from last subtract
+       ldq     r0,0(r18)
+       bis     r8,r25,r25              C combine cy from the two subtracts
+       ldq     r1,8(r18)
+       subq    r6,r2,r28               C 3rd main subtract
+       ldq     r4,0(r17)
+       subq    r28,r25,r22             C 3rd carry subtract
+       ldq     r5,8(r17)
+       cmpult  r6,r2,r8                C compute cy from last subtract
+       cmpult  r28,r25,r25             C compute cy from last subtract
+       stq     r20,0(r16)
+       bis     r8,r25,r25              C combine cy from the two subtracts
+       stq     r21,8(r16)
+       subq    r7,r3,r28               C 4th main subtract
+       subq    r28,r25,r23             C 4th carry subtract
+       cmpult  r7,r3,r8                C compute cy from last subtract
+       cmpult  r28,r25,r25             C compute cy from last subtract
+               addq    r17,32,r17              C update s1_ptr
+       bis     r8,r25,r25              C combine cy from the two subtracts
+               addq    r16,32,r16              C update res_ptr
+       subq    r4,r0,r28               C 1st main subtract
+       ldq     r2,16(r18)
+       subq    r28,r25,r20             C 1st carry subtract
+       ldq     r3,24(r18)
+       cmpult  r4,r0,r8                C compute cy from last subtract
+       ldq     r6,-16(r17)
+       cmpult  r28,r25,r25             C compute cy from last subtract
+       ldq     r7,-8(r17)
+       bis     r8,r25,r25              C combine cy from the two subtracts
+       subq    r19,4,r19               C decr loop cnt
+       stq     r22,-16(r16)
+       subq    r5,r1,r28               C 2nd main subtract
+       stq     r23,-8(r16)
+       subq    r28,r25,r21             C 2nd carry subtract
+               addq    r18,32,r18              C update s2_ptr
+       cmpult  r5,r1,r8                C compute cy from last subtract
+       bge     r19,$Loop
+C Finish software pipeline for 1st loop
+$Lend1:        cmpult  r28,r25,r25             C compute cy from last subtract
+       bis     r8,r25,r25              C combine cy from the two subtracts
+       subq    r6,r2,r28               C cy add
+       subq    r28,r25,r22             C 3rd main subtract
+       cmpult  r6,r2,r8                C compute cy from last subtract
+       cmpult  r28,r25,r25             C compute cy from last subtract
+       stq     r20,0(r16)
+       bis     r8,r25,r25              C combine cy from the two subtracts
+       stq     r21,8(r16)
+       subq    r7,r3,r28               C cy add
+       subq    r28,r25,r23             C 4th main subtract
+       cmpult  r7,r3,r8                C compute cy from last subtract
+       cmpult  r28,r25,r25             C compute cy from last subtract
+       bis     r8,r25,r25              C combine cy from the two subtracts
+       addq    r16,32,r16              C update res_ptr
+       stq     r22,-16(r16)
+       stq     r23,-8(r16)
+$Lend2:        addq    r19,4,r19               C restore loop cnt
+       beq     r19,$Lret
+C Start software pipeline for 2nd loop
+       ldq     r0,0(r18)
+       ldq     r4,0(r17)
+       subq    r19,1,r19
+       beq     r19,$Lend0
+C 2nd loop handles remaining 1-3 limbs
+       ALIGN(16)
+$Loop0:        subq    r4,r0,r28               C main subtract
+       cmpult  r4,r0,r8                C compute cy from last subtract
+       ldq     r0,8(r18)
+       ldq     r4,8(r17)
+       subq    r28,r25,r20             C carry subtract
+       addq    r18,8,r18
+       addq    r17,8,r17
+       stq     r20,0(r16)
+       cmpult  r28,r25,r25             C compute cy from last subtract
+       subq    r19,1,r19               C decr loop cnt
+       bis     r8,r25,r25              C combine cy from the two subtracts
+       addq    r16,8,r16
+       bne     r19,$Loop0
+$Lend0:        subq    r4,r0,r28               C main subtract
+       subq    r28,r25,r20             C carry subtract
+       cmpult  r4,r0,r8                C compute cy from last subtract
+       cmpult  r28,r25,r25             C compute cy from last subtract
+       stq     r20,0(r16)
+       bis     r8,r25,r25              C combine cy from the two subtracts
+
+$Lret: bis     r25,r31,r0              C return cy
+       ret     r31,(r26),1
+EPILOGUE(mpn_sub_n)
+ASM_END()