1 Copyright 2000-2005 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 either:
8 * the GNU Lesser General Public License as published by the Free
9 Software Foundation; either version 3 of the License, or (at your
10 option) any later version.
14 * the GNU General Public License as published by the Free Software
15 Foundation; either version 2 of the License, or (at your option) any
18 or both in parallel, as here.
20 The GNU MP Library is distributed in the hope that it will be useful, but
21 WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
22 or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
25 You should have received copies of the GNU General Public License and the
26 GNU Lesser General Public License along with the GNU MP Library. If not,
27 see https://www.gnu.org/licenses/.
34 This directory contains mpn functions for the IA-64 architecture.
39 mpn/ia64 itanium-2, and generic ia64
41 The code here has been optimized primarily for Itanium 2. Very few Itanium 1
42 chips were ever sold, and Itanium 2 is more powerful, so the latter is what
49 The IA-64 ISA keeps instructions three and three in 128 bit bundles.
50 Programmers/compilers need to put explicit breaks `;;' when there are WAW or
51 RAW dependencies, with some notable exceptions. Such "breaks" are typically
52 at the end of a bundle, but can be put between operations within some bundle
55 The Itanium 1 and Itanium 2 implementations can under ideal conditions
56 execute two bundles per cycle. The Itanium 1 allows 4 of these instructions
57 to do integer operations, while the Itanium 2 allows all 6 to be integer
60 Taken cloop branches seem to insert a bubble into the pipeline most of the
63 Loads to the fp registers bypass the L1 cache and thus get extremely long
64 latencies, 9 cycles on the Itanium 1 and 6 cycles on the Itanium 2.
66 The software pipeline stuff using br.ctop instruction causes delays, since
67 many issue slots are taken up by instructions with zero predicates, and
68 since many extra instructions are needed to set things up. These features
69 are clearly designed for code density, not speed.
71 Misc pipeline limitations (Itanium 1):
72 * The getf.sig instruction can only execute in M0.
73 * At most four integer instructions/cycle.
74 * Nops take up resources like any plain instructions.
76 Misc pipeline limitations (Itanium 2):
77 * The getf.sig instruction can only execute in M0.
78 * Nops take up resources like any plain instructions.
83 .align pads with nops in a text segment, but gas 2.14 and earlier
84 incorrectly byte-swaps its nop bundle in big endian mode (eg. hpux), making
85 it come out as break instructions. We use the ALIGN() macro in
86 mpn/ia64/ia64-defs.m4 when it might be executed across. That macro
87 suppresses any .align if the problem is detected by configure. Lack of
88 alignment might hurt performance but will at least be correct.
90 foo:: to create a global symbol is not accepted by gas. Use separate
91 ".global foo" and "foo:" instead.
93 .global is the standard global directive. gas accepts .globl, but hpux "as"
96 .proc / .endp generates the appropriate .type and .size information for ELF,
97 so the latter directives don't need to be given explicitly.
99 .pred.rel "mutex"... is standard for annotating predicate register
100 relationships. gas also accepts .pred.rel.mutex, but hpux "as" doesn't.
102 .pred directives can't be put on a line with a label, like
103 ".Lfoo: .pred ...", the HP assembler on HP-UX 11.23 rejects that.
104 gas is happy with it, and past versions of HP had seemed ok.
106 // is the standard comment sequence, but we prefer "C" since it inhibits m4
107 macro expansion. See comments in ia64-defs.m4.
114 r1: global pointer (gp)
116 r12: stack pointer (sp)
117 r13: thread pointer (tp)
118 Caller-saves: r8-r11 r14-r31 f6-f15 f32-f127
119 Caller-saves but rotating: r32-
122 ================================================================
123 mpn_add_n, mpn_sub_n:
125 The current code runs at 1.25 c/l on Itanium 2.
127 ================================================================
130 The current code runs at 2 c/l on Itanium 2.
132 Using a blocked approach, working off of 4 separate places in the operands,
133 one could make use of the xma accumulation, and approach 1 c/l.
140 ================================================================
143 The current code runs at 2 c/l on Itanium 2.
145 It seems possible to use a blocked approach, as with mpn_mul_1. We should
146 read rp[] to integer registers, allowing for just one getf.sig per cycle.
156 These 10 instructions can be scheduled to approach 1.667 cycles, and with
157 the 4 cycle latency of xma, this means we need at least 3 blocks. Using
158 ldfp8 we could approach 1.583 c/l.
160 ================================================================
163 The current code runs at 2.25 c/l on Itanium 2. Getting to 2 c/l requires
164 ldfp8 with all alignment headache that implies.
166 ================================================================
169 For best speed, we need to give up using mpn_addmul_2 as the main multiply
170 building block, and instead take multiple v limbs per loop. For the Itanium
171 1, we need to take about 8 limbs at a time for full speed. For the Itanium
172 2, something like mpn_addmul_4 should be enough.
174 The add+cmp+cmp+add we use on the other codes is optimal for shortening
175 recurrencies (1 cycle) but the sequence takes up 4 execution slots. When
176 recurrency depth is not critical, a more standard 3-cycle add+cmp+add is
179 /* First load the 8 values from v */
180 ldfp8 v0, v1 = [r35], 16;;
181 ldfp8 v2, v3 = [r35], 16;;
182 ldfp8 v4, v5 = [r35], 16;;
183 ldfp8 v6, v7 = [r35], 16;;
185 /* In the inner loop, get a new U limb and store a result limb. */
187 Loop: ldf8 u0 = [r33], 8
189 xma.l lp0 = v0, u0, hp0
190 xma.hu hp0 = v0, u0, hp0
191 xma.l lp1 = v1, u0, hp1
192 xma.hu hp1 = v1, u0, hp1
193 xma.l lp2 = v2, u0, hp2
194 xma.hu hp2 = v2, u0, hp2
195 xma.l lp3 = v3, u0, hp3
196 xma.hu hp3 = v3, u0, hp3
197 xma.l lp4 = v4, u0, hp4
198 xma.hu hp4 = v4, u0, hp4
199 xma.l lp5 = v5, u0, hp5
200 xma.hu hp5 = v5, u0, hp5
201 xma.l lp6 = v6, u0, hp6
202 xma.hu hp6 = v6, u0, hp6
203 xma.l lp7 = v7, u0, hp7
204 xma.hu hp7 = v7, u0, hp7
212 add+cmp+add xx, l0, r0
213 add+cmp+add acc0, acc1, l1
214 add+cmp+add acc1, acc2, l2
215 add+cmp+add acc2, acc3, l3
216 add+cmp+add acc3, acc4, l4
217 add+cmp+add acc4, acc5, l5
218 add+cmp+add acc5, acc6, l6
223 49 insn at max 6 insn/cycle: 8.167 cycles/limb8
224 11 memops at max 2 memops/cycle: 5.5 cycles/limb8
225 16 fpops at max 2 fpops/cycle: 8 cycles/limb8
226 21 intops at max 4 intops/cycle: 5.25 cycles/limb8
227 11+21 memops+intops at max 4/cycle 8 cycles/limb8
229 ================================================================
230 mpn_lshift, mpn_rshift
232 The current code runs at 1 cycle/limb on Itanium 2.
234 Using 63 separate loops, we could use the double-word shrp instruction.
235 That instruction has a plain single-cycle latency. We need 63 loops since
236 this instruction only accept immediate count. That would lead to a somewhat
237 silly code size, but the speed would be 0.75 c/l on Itanium 2 (by using shrp
238 each cycle plus shl/shr going down I1 for a further limb every second
241 ================================================================
244 The current code runs at 0.5 c/l on Itanium 2. But that is just for L1
245 cache hit. The 4-way unrolled loop takes just 2 cycles, and thus load-use
246 scheduling isn't great. It might be best to actually use modulo scheduled
247 loops, since that will allow us to do better load-use scheduling without too
250 Depending on size or operand alignment, we get 1 c/l or 0.5 c/l on Itanium
251 2, according to tune/speed. Cache bank conflicts?
257 Intel Itanium Architecture Software Developer's Manual, volumes 1 to 3,
258 Intel document 245317-004, 245318-004, 245319-004 October 2002. Volume 1
259 includes an Itanium optimization guide.
261 Intel Itanium Processor-specific Application Binary Interface (ABI), Intel
262 document 245370-003, May 2001. Describes C type sizes, dynamic linking,
265 Intel Itanium Architecture Assembly Language Reference Guide, Intel document
266 248801-004, 2000-2002. Describes assembly instruction syntax and other
269 Itanium Software Conventions and Runtime Architecture Guide, Intel document
270 245358-003, May 2001. Describes calling conventions, including stack
271 unwinding requirements.
273 Intel Itanium Processor Reference Manual for Software Optimization, Intel
274 document 245473-003, November 2001.
276 Intel Itanium-2 Processor Reference Manual for Software Development and
277 Optimization, Intel document 251110-003, May 2004.
279 All the above documents can be found online at
281 http://developer.intel.com/design/itanium/manuals.htm