[Gmp-commit] /home/hgfiles/gmp: 5 new changesets
mercurial at gmplib.org
mercurial at gmplib.org
Tue Mar 23 13:32:20 CET 2010
details: /home/hgfiles/gmp/rev/a7c16af206d8
changeset: 13524:a7c16af206d8
user: Torbjorn Granlund <tege at gmplib.org>
date: Tue Mar 23 10:28:45 2010 +0100
description:
Generalise ia64 add+lsh functions.
details: /home/hgfiles/gmp/rev/3d1fecb2a105
changeset: 13525:3d1fecb2a105
user: Torbjorn Granlund <tege at gmplib.org>
date: Tue Mar 23 10:30:14 2010 +0100
description:
Misc cleanups.
details: /home/hgfiles/gmp/rev/a4ea14d9b3cd
changeset: 13526:a4ea14d9b3cd
user: Torbjorn Granlund <tege at gmplib.org>
date: Tue Mar 23 10:33:56 2010 +0100
description:
Clean up obsolete comments.
details: /home/hgfiles/gmp/rev/581f85c5ff21
changeset: 13527:581f85c5ff21
user: Torbjorn Granlund <tege at gmplib.org>
date: Tue Mar 23 10:35:06 2010 +0100
description:
Remove a pointless mutex decl.
details: /home/hgfiles/gmp/rev/bbd2c2155a37
changeset: 13528:bbd2c2155a37
user: Torbjorn Granlund <tege at gmplib.org>
date: Tue Mar 23 10:37:50 2010 +0100
description:
Use mpn_invert_limb instead of div insn.
diffstat:
ChangeLog | 13 +-
mpn/ia64/aorslsh1_n.asm | 284 +--------------------------------
mpn/ia64/aorslsh2_n.asm | 48 +++++
mpn/ia64/aorslshC_n.asm | 309 ++++++++++++++++++++++++++++++++++++
mpn/ia64/mul_1.asm | 1 -
mpn/powerpc64/mode64/aorslshC_n.asm | 30 +-
mpn/x86_64/core2/divrem_1.asm | 8 +-
mpn/x86_64/divrem_1.asm | 2 +-
mpn/x86_64/divrem_2.asm | 41 ++--
9 files changed, 412 insertions(+), 324 deletions(-)
diffs (truncated from 877 to 300 lines):
diff -r 4366f13f8b34 -r bbd2c2155a37 ChangeLog
--- a/ChangeLog Mon Mar 22 15:13:36 2010 +0100
+++ b/ChangeLog Tue Mar 23 10:37:50 2010 +0100
@@ -1,3 +1,12 @@
+2010-03-23 Torbjorn Granlund <tege at gmplib.org>
+
+ * mpn/x86_64/divrem_2.asm: Use mpn_invert_limb instead of div insn.
+
+ * mpn/ia64/aorslshC_n.asm: New file, generalised from last iteration of
+ aorslsh1_n.asm.
+ * mpn/ia64/aorslsh1_n.asm: Use aorslshC_n.asm.
+ * mpn/ia64/aorslsh1_n.asm: New file, use aorslshC_n.asm.
+
2010-03-20 Torbjorn Granlund <tege at gmplib.org>
* mpn/powerpc64/mode64/invert_limb.asm: Rewrite to exploit cancellation
@@ -6,10 +15,10 @@
2010-03-20 Marco Bodrato <bodrato at mail.dm.unipi.it>
* mpn/generic/toom_interpolate_8pts.c: Use mpn_sublsh2_n.
-
+
2010-03-20 Torbjorn Granlund <tege at gmplib.org>
- * mpn/powerpc64/mode64/aorslshC_n.asm: New file, generlised from
+ * mpn/powerpc64/mode64/aorslshC_n.asm: New file, generalised from
last iteration of aorslsh1_n.asm.
* mpn/powerpc64/mode64/aorslsh1_n.asm: Use aorslshC_n.asm.
* mpn/powerpc64/mode64/aorslsh1_n.asm: New file, use aorslshC_n.asm.
diff -r 4366f13f8b34 -r bbd2c2155a37 mpn/ia64/aorslsh1_n.asm
--- a/mpn/ia64/aorslsh1_n.asm Mon Mar 22 15:13:36 2010 +0100
+++ b/mpn/ia64/aorslsh1_n.asm Tue Mar 23 10:37:50 2010 +0100
@@ -23,14 +23,9 @@
C Itanium: 3.0
C Itanium 2: 1.5
-C TODO
-C * Use shladd in feed-in code (for mpn_addlsh1_n).
-C INPUT PARAMETERS
-define(`rp',`r32')
-define(`up',`r33')
-define(`vp',`r34')
-define(`n',`r35')
+define(LSH, 1)
+define(RSH, 63)
ifdef(`OPERATION_addlsh1_n',`
define(ADDSUB, add)
@@ -47,280 +42,7 @@
define(func, mpn_sublsh1_n)
')
-C Some useful aliases for registers we use
-define(`u0',`r14') define(`u1',`r15') define(`u2',`r16') define(`u3',`r17')
-define(`v0',`r18') define(`v1',`r19') define(`v2',`r20') define(`v3',`r21')
-define(`w0',`r22') define(`w1',`r23') define(`w2',`r24') define(`w3',`r25')
-define(`s0',`r26') define(`s1',`r27') define(`s2',`r28') define(`s3',`r29')
-define(`x0',`r30') define(`x1',`r31') define(`x2',`r30') define(`x3',`r31')
MULFUNC_PROLOGUE(mpn_addlsh1_n mpn_sublsh1_n)
-ASM_START()
-PROLOGUE(func)
- .prologue
- .save ar.lc, r2
- .body
-ifdef(`HAVE_ABI_32',`
- addp4 rp = 0, rp C M I
- addp4 up = 0, up C M I
- addp4 vp = 0, vp C M I
- zxt4 n = n C I
- ;;
-')
- {.mmi; ld8 r11 = [vp], 8 C M01
- ld8 r10 = [up], 8 C M01
- mov.i r2 = ar.lc C I0
-}{.mmi; and r14 = 3, n C M I
- cmp.lt p15, p0 = 4, n C M I
- add n = -4, n C M I
- ;;
-}{.mmi; cmp.eq p6, p0 = 1, r14 C M I
- cmp.eq p7, p0 = 2, r14 C M I
- cmp.eq p8, p0 = 3, r14 C M I
-}{.bbb
- (p6) br.dptk .Lb01 C B
- (p7) br.dptk .Lb10 C B
- (p8) br.dptk .Lb11 C B
-}
-
-.Lb00: ld8 v0 = [vp], 8 C M01
- ld8 u0 = [up], 8 C M01
- shr.u n = n, 2 C I0
- ;;
- ld8 v1 = [vp], 8 C M01
- ld8 u1 = [up], 8 C M01
- add x3 = r11, r11 C M I
- ;;
- ld8 v2 = [vp], 8 C M01
- ld8 u2 = [up], 8 C M01
- ADDSUB w3 = r10, x3 C M I
- (p15) br.dpnt .grt4 C B
- ;;
- shrp x0 = v0, r11, 63 C I0
- cmp.PRED p8, p0 = w3, r10 C M I
- ;;
- shrp x1 = v1, v0, 63 C I0
- ADDSUB w0 = u0, x0 C M I
- ;;
- cmp.PRED p6, p0 = w0, u0 C M I
- ADDSUB w1 = u1, x1 C M I
- br .Lcj4 C B
-
-.grt4: ld8 v3 = [vp], 8 C M01
- shrp x0 = v0, r11, 63 C I0
- cmp.PRED p8, p0 = w3, r10 C M I
- add n = -1, n
- ;;
- ld8 u3 = [up], 8 C M01
- mov.i ar.lc = n C I0
- shrp x1 = v1, v0, 63 C I0
- ld8 v0 = [vp], 8 C M01
- ADDSUB w0 = u0, x0 C M I
- ;;
- cmp.PRED p6, p0 = w0, u0 C M I
- ld8 u0 = [up], 8 C M01
- ADDSUB w1 = u1, x1 C M I
- br .LL00 C B
-
-.Lb01: add x2 = r11, r11 C M I
- shr.u n = n, 2 C I0
- (p15) br.dpnt .grt1 C B
- ;;
- ADDSUB w2 = r10, x2 C M I
- shr.u r8 = r11, 63 C retval I0
- ;;
- cmp.PRED p6, p0 = w2, r10 C M I
- ;;
- st8 [rp] = w2, 8 C M23
- (p6) add r8 = 1, r8 C M I
- br.ret.sptk.many b0 C B
-
-.grt1: ld8 v3 = [vp], 8 C M01
- ld8 u3 = [up], 8 C M01
- mov.i ar.lc = n C FIXME swap with next I0
- ;;
- ld8 v0 = [vp], 8 C M01
- ld8 u0 = [up], 8 C M01
- ADDSUB w2 = r10, x2
- ;;
- ld8 v1 = [vp], 8 C M01
- ld8 u1 = [up], 8 C M01
- shrp x3 = v3, r11, 63 C I0
- ;;
- ld8 v2 = [vp], 8 C M01
- ld8 u2 = [up], 8 C M01
- cmp.PRED p6, p0 = w2, r10 C M I
- ADDSUB w3 = u3, x3 C M I
- br.cloop.dptk .grt5 C B
- ;;
- shrp x0 = v0, v3, 63 C I0
- cmp.PRED p8, p0 = w3, u3 C M I
- br .Lcj5 C B
-
-.grt5: shrp x0 = v0, v3, 63 C I0
- ld8 v3 = [vp], 8 C M01
- cmp.PRED p8, p0 = w3, u3 C M I
- br .LL01 C B
-
-.Lb10: ld8 v2 = [vp], 8 C M01
- ld8 u2 = [up], 8 C M01
- shr.u n = n, 2 C I0
- add x1 = r11, r11 C M I
- (p15) br.dpnt .grt2 C B
- ;;
- ADDSUB w1 = r10, x1 C M I
- shrp x2 = v2, r11, 63 C I0
- ;;
- cmp.PRED p8, p0 = w1, r10 C M I
- ADDSUB w2 = u2, x2 C M I
- shr.u r8 = v2, 63 C retval I0
- ;;
- cmp.PRED p6, p0 = w2, u2 C M I
- br .Lcj2 C B
-
-.grt2: ld8 v3 = [vp], 8 C M01
- ld8 u3 = [up], 8 C M01
- mov.i ar.lc = n C I0
- ;;
- ld8 v0 = [vp], 8 C M01
- ld8 u0 = [up], 8 C M01
- ADDSUB w1 = r10, x1 C M I
- ;;
- ld8 v1 = [vp], 8 C M01
- shrp x2 = v2, r11, 63 C I0
- cmp.PRED p8, p0 = w1, r10 C M I
- ;;
- ld8 u1 = [up], 8 C M01
- shrp x3 = v3, v2, 63 C I0
- ld8 v2 = [vp], 8 C M01
- ADDSUB w2 = u2, x2 C M I
- ;;
- cmp.PRED p6, p0 = w2, u2 C M I
- ld8 u2 = [up], 8 C M01
- ADDSUB w3 = u3, x3 C M I
- br.cloop.dpnt .Loop C B
- br .Lskip C B
-
-.Lb11: ld8 v1 = [vp], 8 C M01
- ld8 u1 = [up], 8 C M01
- shr.u n = n, 2 C I0
- add x0 = r11, r11 C M I
- ;;
- ld8 v2 = [vp], 8 C M01
- ld8 u2 = [up], 8 C M01
- (p15) br.dpnt .grt3 C B
- ;;
-
- shrp x1 = v1, r11, 63 C I0
- ADDSUB w0 = r10, x0 C M I
- ;;
- cmp.PRED p6, p0 = w0, r10 C M I
- ADDSUB w1 = u1, x1 C M I
- ;;
- shrp x2 = v2, v1, 63 C I0
- cmp.PRED p8, p0 = w1, u1 C M I
- br .Lcj3 C B
-
-.grt3: ld8 v3 = [vp], 8 C M01
- ld8 u3 = [up], 8 C M01
- mov.i ar.lc = n C I0
- shrp x1 = v1, r11, 63 C I0
- ADDSUB w0 = r10, x0 C M I
- ;;
- ld8 v0 = [vp], 8 C M01
- cmp.PRED p6, p0 = w0, r10 C M I
- ld8 u0 = [up], 8 C M01
- ADDSUB w1 = u1, x1 C M I
- ;;
- shrp x2 = v2, v1, 63 C I0
- ld8 v1 = [vp], 8 C M01
- cmp.PRED p8, p0 = w1, u1 C M I
- br .LL11 C B
-
-
-C *** MAIN LOOP START ***
- ALIGN(32)
-.Loop: st8 [rp] = w1, 8 C M23
- shrp x0 = v0, v3, 63 C I0
- (p8) cmp.eq.or p6, p0 = LIM, w2 C M I
- (p8) add w2 = INCR, w2 C M I
- ld8 v3 = [vp], 8 C M01
- cmp.PRED p8, p0 = w3, u3 C M I
- ;;
-.LL01: ld8 u3 = [up], 8 C M01
- shrp x1 = v1, v0, 63 C I0
- (p6) cmp.eq.or p8, p0 = LIM, w3 C M I
- (p6) add w3 = INCR, w3 C M I
- ld8 v0 = [vp], 8 C M01
- ADDSUB w0 = u0, x0 C M I
- ;;
- st8 [rp] = w2, 8 C M23
- cmp.PRED p6, p0 = w0, u0 C M I
- nop.b 0
- ld8 u0 = [up], 8 C M01
- ADDSUB w1 = u1, x1 C M I
- nop.b 0
- ;;
-.LL00: st8 [rp] = w3, 8 C M23
- shrp x2 = v2, v1, 63 C I0
- (p8) cmp.eq.or p6, p0 = LIM, w0 C M I
- (p8) add w0 = INCR, w0 C M I
- ld8 v1 = [vp], 8 C M01
- cmp.PRED p8, p0 = w1, u1 C M I
- ;;
-.LL11: ld8 u1 = [up], 8 C M01
- shrp x3 = v3, v2, 63 C I0
- (p6) cmp.eq.or p8, p0 = LIM, w1 C M I
- (p6) add w1 = INCR, w1 C M I
- ld8 v2 = [vp], 8 C M01
- ADDSUB w2 = u2, x2 C M I
- ;;
- st8 [rp] = w0, 8 C M23
- cmp.PRED p6, p0 = w2, u2 C M I
- nop.b 0
- ld8 u2 = [up], 8 C M01
- ADDSUB w3 = u3, x3 C M I
- br.cloop.dptk .Loop C B
- ;;
-C *** MAIN LOOP END ***
-
-.Lskip: st8 [rp] = w1, 8 C M23
- shrp x0 = v0, v3, 63 C I0
- (p8) cmp.eq.or p6, p0 = LIM, w2 C M I
- (p8) add w2 = INCR, w2 C M I
- cmp.PRED p8, p0 = w3, u3 C M I
- ;;
-.Lcj5: shrp x1 = v1, v0, 63 C I0
- (p6) cmp.eq.or p8, p0 = LIM, w3 C M I
- (p6) add w3 = INCR, w3 C M I
More information about the gmp-commit
mailing list