From: Ingo Molnar <mingo@elte.hu>
To: Linus Torvalds <torvalds@transmeta.com>
Cc: Neil Brown <neilb@cse.unsw.edu.au>, <linux-raid@vger.kernel.org>,
<linux-kernel@vger.kernel.org>
Subject: [patch] raid-xor-2.4.10-A0
Date: Fri, 14 Sep 2001 12:56:43 +0200 (CEST) [thread overview]
Message-ID: <Pine.LNX.4.33.0109141251410.3820-200000@localhost.localdomain> (raw)
[-- Attachment #1: Type: TEXT/PLAIN, Size: 787 bytes --]
another RAID patch against 2.4.10-pre9:
- update the SSE XOR routines to get compiled and used on recent kernels.
- change prefetch code to pollute the cache less, and to prefetch in a
wider window, to give enough time for prefetches to finish.
people with SSE-capable CPUs (PIII, PIV, newer Athlons) should see
something like this in the bootlog:
raid5: measuring checksumming speed
8regs : 1292.400 MB/sec
32regs : 607.600 MB/sec
pIII_sse : 1407.200 MB/sec
pII_mmx : 1600.800 MB/sec
p5_mmx : 1670.000 MB/sec
raid5: using function: pIII_sse (1407.200 MB/sec)
(if present then the SSE code is still picked up exclusively due to its
better cache-properties, even if it's 'cached performance' is lower than
that of MMX routines.)
Ingo
[-- Attachment #2: Type: TEXT/PLAIN, Size: 2542 bytes --]
--- linux/include/asm-i386/xor.h.orig2 Fri Sep 14 12:10:21 2001
+++ linux/include/asm-i386/xor.h Fri Sep 14 12:32:36 2001
@@ -527,8 +527,6 @@
#undef FPU_SAVE
#undef FPU_RESTORE
-#if defined(CONFIG_X86_FXSR) || defined(CONFIG_X86_RUNTIME_FXSR)
-
/*
* Cache avoiding checksumming functions utilizing KNI instructions
* Copyright (C) 1999 Zach Brown (with obvious credit due Ingo)
@@ -559,19 +557,20 @@
: "memory")
#define OFFS(x) "16*("#x")"
-#define PF0(x) " prefetcht0 "OFFS(x)"(%1) ;\n"
-#define LD(x,y) " movaps "OFFS(x)"(%1), %%xmm"#y" ;\n"
-#define ST(x,y) " movaps %%xmm"#y", "OFFS(x)"(%1) ;\n"
-#define PF1(x) " prefetchnta "OFFS(x)"(%2) ;\n"
-#define PF2(x) " prefetchnta "OFFS(x)"(%3) ;\n"
-#define PF3(x) " prefetchnta "OFFS(x)"(%4) ;\n"
-#define PF4(x) " prefetchnta "OFFS(x)"(%5) ;\n"
-#define PF5(x) " prefetchnta "OFFS(x)"(%6) ;\n"
-#define XO1(x,y) " xorps "OFFS(x)"(%2), %%xmm"#y" ;\n"
-#define XO2(x,y) " xorps "OFFS(x)"(%3), %%xmm"#y" ;\n"
-#define XO3(x,y) " xorps "OFFS(x)"(%4), %%xmm"#y" ;\n"
-#define XO4(x,y) " xorps "OFFS(x)"(%5), %%xmm"#y" ;\n"
-#define XO5(x,y) " xorps "OFFS(x)"(%6), %%xmm"#y" ;\n"
+#define PF_OFFS(x) "256+16*("#x")"
+#define PF0(x) " prefetchnta "PF_OFFS(x)"(%1) ;\n"
+#define LD(x,y) " movaps "OFFS(x)"(%1), %%xmm"#y" ;\n"
+#define ST(x,y) " movaps %%xmm"#y", "OFFS(x)"(%1) ;\n"
+#define PF1(x) " prefetchnta "PF_OFFS(x)"(%2) ;\n"
+#define PF2(x) " prefetchnta "PF_OFFS(x)"(%3) ;\n"
+#define PF3(x) " prefetchnta "PF_OFFS(x)"(%4) ;\n"
+#define PF4(x) " prefetchnta "PF_OFFS(x)"(%5) ;\n"
+#define PF5(x) " prefetchnta "PF_OFFS(x)"(%6) ;\n"
+#define XO1(x,y) " xorps "OFFS(x)"(%2), %%xmm"#y" ;\n"
+#define XO2(x,y) " xorps "OFFS(x)"(%3), %%xmm"#y" ;\n"
+#define XO3(x,y) " xorps "OFFS(x)"(%4), %%xmm"#y" ;\n"
+#define XO4(x,y) " xorps "OFFS(x)"(%5), %%xmm"#y" ;\n"
+#define XO5(x,y) " xorps "OFFS(x)"(%6), %%xmm"#y" ;\n"
static void
@@ -849,15 +848,6 @@
deals with a load to a line that is being prefetched. */
#define XOR_SELECT_TEMPLATE(FASTEST) \
(cpu_has_xmm ? &xor_block_pIII_sse : FASTEST)
-
-#else
-
-/* Don't try any SSE2 when FXSR is not enabled, because OSFXSR will not be set
- -AK */
-#define XOR_SSE2
-#define XOR_SELECT_TEMPLATE(FASTEST) (FASTEST)
-
-#endif
/* Also try the generic routines. */
#include <asm-generic/xor.h>
next reply other threads:[~2001-09-15 11:29 UTC|newest]
Thread overview: 2+ messages / expand[flat|nested] mbox.gz Atom feed top
2001-09-14 10:56 Ingo Molnar [this message]
2001-09-14 11:02 ` [patch] raid-xor-2.4.10-A0 Ingo Molnar
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=Pine.LNX.4.33.0109141251410.3820-200000@localhost.localdomain \
--to=mingo@elte.hu \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-raid@vger.kernel.org \
--cc=neilb@cse.unsw.edu.au \
--cc=torvalds@transmeta.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).