From: "Emilio G. Cota" <cota@braap.org>
To: QEMU Developers <qemu-devel@nongnu.org>,
MTTCG Devel <mttcg@listserver.greensocs.com>
Cc: "Peter Maydell" <peter.maydell@linaro.org>,
"Peter Crosthwaite" <crosthwaite.peter@gmail.com>,
"Sergey Fedorov" <serge.fdrv@gmail.com>,
"Paolo Bonzini" <pbonzini@redhat.com>,
"Alex Bennée" <alex.bennee@linaro.org>,
"Richard Henderson" <rth@twiddle.net>
Subject: [Qemu-devel] [PATCH 07/10] tb hash: hash phys_pc, pc, and flags with xxhash
Date: Tue, 5 Apr 2016 01:30:50 -0400 [thread overview]
Message-ID: <1459834253-8291-8-git-send-email-cota@braap.org> (raw)
In-Reply-To: <1459834253-8291-1-git-send-email-cota@braap.org>
For some workloads such as arm bootup, tb_phys_hash is performance-critical.
The is due to the high frequency of accesses to the hash table, originated
by (frequent) TLB flushes that wipe out the cpu-private tb_jmp_cache's.
More info:
https://lists.nongnu.org/archive/html/qemu-devel/2016-03/msg05098.html
To dig further into this I modified an arm image booting debian jessie to
immediately shut down after boot. Analysis revealed that quite a bit of time
is unnecessarily spent in tb_phys_hash: the cause is poor hashing that
results in very uneven loading of chains in the hash table's buckets;
the longest observed chain had ~550 elements.
The appended addresses this with two changes:
1) Use xxhash as the hash table's hash function. xxhash is a fast,
high-quality hashing function.
2) Feed the hashing function with not just tb_phys, but also pc and flags.
This improves performance over using just tb_phys for hashing, since that
resulted in some hash buckets having many TB's, while others getting very few;
with these changes, the longest observed chain on a single hash bucket is
brought down from ~550 to ~40.
Tests show that the other element checked for in tb_find_physical,
cs_base, is always a match when tb_phys+pc+flags are a match,
so hashing cs_base is wasteful. It could be that this is an ARM-only
thing, though.
BTW, after this change the hash table should not be called "tb_hash_phys"
anymore; this is addressed later in this series.
This change gives consistent bootup time improvements. I tested two
host machines:
- Intel Xeon E5-2690: 11.6% less time
- Intel i7-4790K: 19.2% less time
Increasing the number of hash buckets yields further improvements. However,
using a larger, fixed number of buckets can degrade performance for other
workloads that do not translate as many blocks (600K+ for debian-jessie arm
bootup). This is dealt with later in this series.
Signed-off-by: Emilio G. Cota <cota@braap.org>
---
cpu-exec.c | 2 +-
include/exec/tb-hash.h | 19 +++++++++++++++++--
translate-all.c | 6 +++---
3 files changed, 21 insertions(+), 6 deletions(-)
diff --git a/cpu-exec.c b/cpu-exec.c
index bbfcbfb..4194a4a 100644
--- a/cpu-exec.c
+++ b/cpu-exec.c
@@ -233,7 +233,7 @@ static TranslationBlock *tb_find_physical(CPUState *cpu,
/* find translated block using physical mappings */
phys_pc = get_page_addr_code(env, pc);
phys_page1 = phys_pc & TARGET_PAGE_MASK;
- h = tb_phys_hash_func(phys_pc);
+ h = tb_hash_func(phys_pc, pc, flags);
ptb1 = &tcg_ctx.tb_ctx.tb_phys_hash[h];
for(;;) {
tb = *ptb1;
diff --git a/include/exec/tb-hash.h b/include/exec/tb-hash.h
index 0f4e8a0..c4942e1 100644
--- a/include/exec/tb-hash.h
+++ b/include/exec/tb-hash.h
@@ -20,6 +20,9 @@
#ifndef EXEC_TB_HASH
#define EXEC_TB_HASH
+#include "exec/exec-all.h"
+#include "qemu/xxhash.h"
+
/* Only the bottom TB_JMP_PAGE_BITS of the jump cache hash bits vary for
addresses on the same page. The top bits are the same. This allows
TLB invalidation to quickly clear a subset of the hash table. */
@@ -43,9 +46,21 @@ static inline unsigned int tb_jmp_cache_hash_func(target_ulong pc)
| (tmp & TB_JMP_ADDR_MASK));
}
-static inline unsigned int tb_phys_hash_func(tb_page_addr_t pc)
+static inline
+uint32_t tb_hash_func(tb_page_addr_t phys_pc, target_ulong pc, uint64_t flags)
{
- return (pc >> 2) & (CODE_GEN_PHYS_HASH_SIZE - 1);
+ struct key {
+ tb_page_addr_t phys_pc;
+ target_ulong pc;
+ uint64_t flags;
+ } QEMU_PACKED k;
+ unsigned int hash;
+
+ k.phys_pc = phys_pc;
+ k.pc = pc;
+ k.flags = flags;
+ hash = qemu_xxh32((uint32_t *)&k, sizeof(k) / sizeof(uint32_t), 1);
+ return hash & (CODE_GEN_PHYS_HASH_SIZE - 1);
}
#endif
diff --git a/translate-all.c b/translate-all.c
index 8329ea6..8e7f9a7 100644
--- a/translate-all.c
+++ b/translate-all.c
@@ -972,7 +972,7 @@ void tb_phys_invalidate(TranslationBlock *tb, tb_page_addr_t page_addr)
/* remove the TB from the hash list */
phys_pc = tb->page_addr[0] + (tb->pc & ~TARGET_PAGE_MASK);
- h = tb_phys_hash_func(phys_pc);
+ h = tb_hash_func(phys_pc, tb->pc, tb->flags);
tb_hash_remove(&tcg_ctx.tb_ctx.tb_phys_hash[h], tb);
/* remove the TB from the page list */
@@ -1474,8 +1474,8 @@ static void tb_link_page(TranslationBlock *tb, tb_page_addr_t phys_pc,
unsigned int h;
TranslationBlock **ptb;
- /* add in the physical hash table */
- h = tb_phys_hash_func(phys_pc);
+ /* add in the hash table */
+ h = tb_hash_func(phys_pc, tb->pc, tb->flags);
ptb = &tcg_ctx.tb_ctx.tb_phys_hash[h];
tb->phys_hash_next = *ptb;
*ptb = tb;
--
2.5.0
next prev parent reply other threads:[~2016-04-05 5:31 UTC|newest]
Thread overview: 62+ messages / expand[flat|nested] mbox.gz Atom feed top
2016-04-05 5:30 [Qemu-devel] [PATCH 00/10] tb hash improvements Emilio G. Cota
2016-04-05 5:30 ` [Qemu-devel] [PATCH 01/10] translate-all: add missing fold of tb_ctx into tcg_ctx Emilio G. Cota
2016-04-05 8:49 ` Paolo Bonzini
2016-04-05 5:30 ` [Qemu-devel] [PATCH 02/10] compiler.h: add QEMU_CACHELINE + QEMU_ALIGN() + QEMU_CACHELINE_ALIGNED Emilio G. Cota
2016-04-05 7:57 ` Peter Maydell
2016-04-05 17:24 ` Emilio G. Cota
2016-04-05 18:01 ` Peter Maydell
2016-04-05 19:13 ` Emilio G. Cota
2016-04-05 8:49 ` Paolo Bonzini
2016-04-05 12:57 ` Lluís Vilanova
2016-04-05 12:58 ` Peter Maydell
2016-04-05 15:29 ` Paolo Bonzini
2016-04-05 16:23 ` Lluís Vilanova
2016-04-05 16:31 ` Richard Henderson
2016-04-05 16:56 ` Peter Maydell
2016-04-05 19:02 ` Lluís Vilanova
2016-04-05 19:15 ` Richard Henderson
2016-04-05 20:09 ` Lluís Vilanova
2016-04-06 11:44 ` Paolo Bonzini
2016-04-06 12:02 ` Laurent Desnogues
2016-04-05 5:30 ` [Qemu-devel] [PATCH 03/10] seqlock: remove optional mutex Emilio G. Cota
2016-04-06 8:38 ` Alex Bennée
2016-04-05 5:30 ` [Qemu-devel] [PATCH 04/10] seqlock: rename write_lock/unlock to write_begin/end Emilio G. Cota
2016-04-06 8:42 ` Alex Bennée
2016-04-05 5:30 ` [Qemu-devel] [PATCH 05/10] include: add spinlock wrapper Emilio G. Cota
2016-04-05 8:51 ` Paolo Bonzini
2016-04-06 15:51 ` Alex Bennée
2016-04-05 5:30 ` [Qemu-devel] [PATCH 06/10] include: add xxhash.h Emilio G. Cota
2016-04-06 11:39 ` Alex Bennée
2016-04-06 22:59 ` Emilio G. Cota
2016-04-05 5:30 ` Emilio G. Cota [this message]
2016-04-05 15:41 ` [Qemu-devel] [PATCH 07/10] tb hash: hash phys_pc, pc, and flags with xxhash Richard Henderson
2016-04-05 15:48 ` Paolo Bonzini
2016-04-05 16:07 ` Richard Henderson
2016-04-05 19:40 ` Emilio G. Cota
2016-04-05 21:08 ` Richard Henderson
2016-04-06 0:52 ` Emilio G. Cota
2016-04-06 11:52 ` Paolo Bonzini
2016-04-06 17:44 ` Emilio G. Cota
2016-04-06 18:23 ` Paolo Bonzini
2016-04-06 18:27 ` Richard Henderson
2016-04-07 0:37 ` Emilio G. Cota
2016-04-07 8:46 ` Paolo Bonzini
2016-04-05 16:33 ` Laurent Desnogues
2016-04-05 17:19 ` Richard Henderson
2016-04-06 6:06 ` Laurent Desnogues
2016-04-06 17:32 ` Emilio G. Cota
2016-04-06 17:42 ` Richard Henderson
2016-04-07 8:12 ` Laurent Desnogues
2016-04-05 5:30 ` [Qemu-devel] [PATCH 08/10] qht: QEMU's fast, resizable and scalable Hash Table Emilio G. Cota
2016-04-05 9:01 ` Paolo Bonzini
2016-04-05 15:50 ` Richard Henderson
2016-04-08 10:27 ` Alex Bennée
2016-04-19 23:03 ` Emilio G. Cota
2016-04-05 5:30 ` [Qemu-devel] [PATCH 09/10] qht: add test program Emilio G. Cota
2016-04-08 10:45 ` Alex Bennée
2016-04-19 23:06 ` Emilio G. Cota
2016-04-20 7:50 ` Alex Bennée
2016-04-05 5:30 ` [Qemu-devel] [PATCH 10/10] tb hash: track translated blocks with qht Emilio G. Cota
2016-04-08 12:39 ` Alex Bennée
2016-04-05 8:47 ` [Qemu-devel] [PATCH 00/10] tb hash improvements Alex Bennée
2016-04-05 9:01 ` Paolo Bonzini
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=1459834253-8291-8-git-send-email-cota@braap.org \
--to=cota@braap.org \
--cc=alex.bennee@linaro.org \
--cc=crosthwaite.peter@gmail.com \
--cc=mttcg@listserver.greensocs.com \
--cc=pbonzini@redhat.com \
--cc=peter.maydell@linaro.org \
--cc=qemu-devel@nongnu.org \
--cc=rth@twiddle.net \
--cc=serge.fdrv@gmail.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).