1 Copyright 1999, 2000, 2001, 2003, 2004, 2005 Free Software Foundation, Inc.
3 This file is part of the GNU MP Library.
5 The GNU MP Library is free software; you can redistribute it and/or modify
6 it under the terms of the GNU Lesser General Public License as published by
7 the Free Software Foundation; either version 3 of the License, or (at your
8 option) any later version.
10 The GNU MP Library is distributed in the hope that it will be useful, but
11 WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
12 or FITNESS FOR A PARTICULAR PURPOSE. See the GNU Lesser General Public
13 License for more details.
15 You should have received a copy of the GNU Lesser General Public License
16 along with the GNU MP Library. If not, see http://www.gnu.org/licenses/.
20 POWERPC-64 MPN SUBROUTINES
23 This directory contains mpn functions for 64-bit PowerPC chips.
28 mpn/powerpc64 mode-neutral code
29 mpn/powerpc64/mode32 code for mode32
30 mpn/powerpc64/mode64 code for mode64
33 The mode32 and mode64 sub-directories contain code which is for use in the
34 respective chip mode, 32 or 64. The top-level directory is code that's
35 unaffected by the mode.
37 The "adde" instruction is the main difference between mode32 and mode64. It
38 operates on either on a 32-bit or 64-bit quantity according to the chip mode.
39 Other instructions have an operand size in their opcode and hence don't vary.
43 POWER3/PPC630 pipeline information:
45 Decoding is 4-way + branch and issue is 8-way with some out-of-order
51 FXU1 - integer unit 1, handles any simple integer instruction
52 FXU2 - integer unit 2, handles any simple integer instruction
53 FXU3 - integer unit 3, handles integer multiply and divide
54 FPU1 - floating-point unit 1
55 FPU2 - floating-point unit 2
57 Memory: Any two memory operations can issue, but memory subsystem
58 can sustain just one store per cycle. No need for data
59 prefetch; the hardware has very sophisticated prefetch logic.
60 Simple integer: 2 operations (such as add, rl*)
61 Integer multiply: 1 operation every 9th cycle worst case; exact timing depends
62 on 2nd operand's most significant bit position (10 bits per
63 cycle). Multiply unit is not pipelined, only one multiply
64 operation in progress is allowed.
66 Floating-point: Any plain 2 arithmetic instructions (such as fmul, fadd, and
67 fmadd), latency 4 cycles.
68 Floating-point divide:
70 Floating-point square root:
73 POWER3/PPC630 best possible times for the main loops:
74 shift: 1.5 cycles limited by integer unit contention.
75 With 63 special loops, one for each shift count, we could
76 reduce the needed integer instructions to 2, which would
77 reduce the best possible time to 1 cycle.
78 add/sub: 1.5 cycles, limited by ld/st unit contention.
79 mul: 18 cycles (average) unless floating-point operations are used,
80 but that would only help for multiplies of perhaps 10 and more
82 addmul/submul:Same situation as for mul.
85 POWER4/PPC970 and POWER5 pipeline information:
87 This is a very odd pipeline, it is basically a VLIW masquerading as a plain
88 architecture. Its issue rules are not made public, and since it is so weird,
89 it is very hard to figure out any useful information from experimentation.
92 A well-aligned loop with nop's take 3, 4, 6, 7, ... cycles.
93 3 cycles for 0, 1, 2, 3, 4, 5, 6, 7 nop's
94 4 cycles for 8, 9, 10, 11, 12, 13, 14, 15 nop's
95 6 cycles for 16, 17, 18, 19, 20, 21, 22, 23 nop's
96 7 cycles for 24, 25, 26, 27 nop's
97 8 cycles for 28, 29, 30, 31 nop's
98 ... continues regularly
104 FXU1 - integer unit 1, handles any integer instruction
105 FXU2 - integer unit 2, handles any integer instruction
106 FPU1 - floating-point unit 1
107 FPU2 - floating-point unit 2
109 While this is one integer unit less than POWER3/PPC630, the remaining units
110 are more powerful; here they handle multiply and divide.
112 Memory: 2 ld/st. Stores go to the L2 cache, which can sustain just
114 L1 load latency: to gregs 3-4 cycles, to fregs 5-6 cycles.
115 Operations that modify the address register might be split
116 to use also a an integer issue slot.
117 Simple integer: 2 operations every cycle, latency 2.
118 Integer multiply: 2 operations every 6th cycle, latency 7 cycles.
120 Floating-point: Any plain 2 arithmetic instructions (such as fmul, fadd, and
121 fmadd), latency 6 cycles.
122 Floating-point divide:
124 Floating-point square root:
130 *mul_1: Handling one limb using mulld/mulhdu and two limbs using floating-
131 point operations should give performance of about 20 cycles for 3 limbs, or 7
134 We should probably split the single-limb operand in 32-bit chunks, and the
135 multi-limb operand in 16-bit chunks, allowing us to accumulate well in fp
138 Problem is to get 32-bit or 16-bit words to the fp registers. Only 64-bit fp
139 memops copies bits without fiddling with them. We might therefore need to
140 load to integer registers with zero extension, store as 64 bits into temp
141 space, and then load to fp regs. Alternatively, load directly to fp space
142 and add well-chosen constants to get cancelation. (Other part after given by
143 subsequent subtraction.)
145 Possible code mix for load-via-intregs variant:
148 fmadd,fmadd,fmul,fmul
149 fctidz,stfd,ld,fctidz,stfd,ld
152 fmadd,fmadd,fmul,fmul
153 fctidz,stfd,ld,fctidz,stfd,ld
155 srd,sld,add,adde,add,adde