mirror of
https://github.com/guanzhi/GmSSL.git
synced 2026-11-16 17:45:04 +08:00
Merge remote-tracking branch 'origin/master'
# Conflicts: # README.md
This commit is contained in:
@@ -3,6 +3,13 @@
|
||||
.ident "ia64.S, Version 2.1"
|
||||
.ident "IA-64 ISA artwork by Andy Polyakov <appro@fy.chalmers.se>"
|
||||
|
||||
// Copyright 2001-2016 The OpenSSL Project Authors. All Rights Reserved.
|
||||
//
|
||||
// Licensed under the OpenSSL license (the "License"). You may not use
|
||||
// this file except in compliance with the License. You can obtain a copy
|
||||
// in the file LICENSE in the source distribution or at
|
||||
// https://www.openssl.org/source/license.html
|
||||
|
||||
//
|
||||
// ====================================================================
|
||||
// Written by Andy Polyakov <appro@fy.chalmers.se> for the OpenSSL
|
||||
@@ -22,7 +29,7 @@
|
||||
// ports is the same, i.e. 2, while I need 4. In other words, to this
|
||||
// module Itanium2 remains effectively as "wide" as Itanium. Yet it's
|
||||
// essentially different in respect to this module, and a re-tune was
|
||||
// required. Well, because some intruction latencies has changed. Most
|
||||
// required. Well, because some instruction latencies has changed. Most
|
||||
// noticeably those intensively used:
|
||||
//
|
||||
// Itanium Itanium2
|
||||
@@ -363,7 +370,7 @@ bn_mul_words:
|
||||
// The loop therefore spins at the latency of xma minus 1, or in other
|
||||
// words at 6*(n+4) ticks:-( Compare to the "production" loop above
|
||||
// that runs in 2*(n+11) where the low latency problem is worked around
|
||||
// by moving the dependency to one-tick latent interger ALU. Note that
|
||||
// by moving the dependency to one-tick latent integer ALU. Note that
|
||||
// "distance" between ldf8 and xma is not latency of ldf8, but the
|
||||
// *difference* between xma and ldf8 latencies.
|
||||
.L_bn_mul_words_ctop:
|
||||
@@ -422,10 +429,10 @@ bn_mul_add_words:
|
||||
|
||||
// This loop spins in 3*(n+10) ticks on Itanium and in 2*(n+10) on
|
||||
// Itanium 2. Yes, unlike previous versions it scales:-) Previous
|
||||
// version was peforming *all* additions in IALU and was starving
|
||||
// version was performing *all* additions in IALU and was starving
|
||||
// for those even on Itanium 2. In this version one addition is
|
||||
// moved to FPU and is folded with multiplication. This is at cost
|
||||
// of propogating the result from previous call to this subroutine
|
||||
// of propagating the result from previous call to this subroutine
|
||||
// to L2 cache... In other words negligible even for shorter keys.
|
||||
// *Overall* performance improvement [over previous version] varies
|
||||
// from 11 to 22 percent depending on key length.
|
||||
@@ -495,7 +502,7 @@ bn_sqr_words:
|
||||
// scalability. The decision will very likely be reconsidered after the
|
||||
// benchmark program is profiled. I.e. if perfomance gain on Itanium
|
||||
// will appear larger than loss on "wider" IA-64, then the loop should
|
||||
// be explicitely split and the epilogue compressed.
|
||||
// be explicitly split and the epilogue compressed.
|
||||
.L_bn_sqr_words_ctop:
|
||||
{ .mfi; (p16) ldf8 f32=[r33],8
|
||||
(p25) xmpy.lu f42=f41,f41
|
||||
@@ -568,7 +575,7 @@ bn_sqr_comba8:
|
||||
// I've estimated this routine to run in ~120 ticks, but in reality
|
||||
// (i.e. according to ar.itc) it takes ~160 ticks. Are those extra
|
||||
// cycles consumed for instructions fetch? Or did I misinterpret some
|
||||
// clause in Itanium µ-architecture manual? Comments are welcomed and
|
||||
// clause in Itanium µ-architecture manual? Comments are welcomed and
|
||||
// highly appreciated.
|
||||
//
|
||||
// On Itanium 2 it takes ~190 ticks. This is because of stalls on
|
||||
|
||||
Reference in New Issue
Block a user