From: Peter Zijlstra <peterz@infradead.org>
To: Andrew Morton <akpm@linux-foundation.org>
Cc: mingo@redhat.com, hpa@zytor.com, eranian@google.com,
linux-kernel@vger.kernel.org, torvalds@linux-foundation.org,
tglx@linutronix.de, mingo@elte.hu,
linux-tip-commits@vger.kernel.org
Subject: Re: [tip:perf/core] bitops: Provide compile time HWEIGHT{8,16,32,64}
Date: Fri, 29 Jan 2010 12:03:43 +0100 [thread overview]
Message-ID: <1264763023.4283.2213.camel@laptop> (raw)
In-Reply-To: <20100129020128.716af8fb.akpm@linux-foundation.org>
On Fri, 2010-01-29 at 02:01 -0800, Andrew Morton wrote:
> On Fri, 29 Jan 2010 09:28:04 GMT tip-bot for Peter Zijlstra <a.p.zijlstra@chello.nl> wrote:
>
> > +#define HWEIGHT8(w) \
> > + ( (!!((w) & (1ULL << 0))) + \
> > + (!!((w) & (1ULL << 1))) + \
> > + (!!((w) & (1ULL << 2))) + \
> > + (!!((w) & (1ULL << 3))) + \
> > + (!!((w) & (1ULL << 4))) + \
> > + (!!((w) & (1ULL << 5))) + \
> > + (!!((w) & (1ULL << 6))) + \
> > + (!!((w) & (1ULL << 7))) )
> > +
> > +#define HWEIGHT16(w) (HWEIGHT8(w) + HWEIGHT8(w >> 8))
> > +#define HWEIGHT32(w) (HWEIGHT16(w) + HWEIGHT16(w >> 16))
> > +#define HWEIGHT64(w) (HWEIGHT32(w) + HWEIGHT32(w >> 32))
>
> Would be nice if it had a comment explaining why it exists. If people
> accidentally use this with non-constant arguments, the generated code
> will be pretty ghastly.
*sigh* and here I though it being placed right next to hweight_long()
which uses the arch hweightN() would be clue enough.
If people are so clueless, who says they'll read a comment.. but sure I
guess I can add one.
> Or add some barf-if-not-__constant_p() thing, perhaps.
I've actually sneaked one non-constant usage in, but since its in an
init path I didn't care to fix that, but I guess here goes:
---
Subject: bitops: Dummyify the compile-time hweight versions
Because it seems allowed to not think and write kernel code.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
---
Index: linux-2.6/arch/x86/kernel/cpu/perf_event.c
===================================================================
--- linux-2.6.orig/arch/x86/kernel/cpu/perf_event.c
+++ linux-2.6/arch/x86/kernel/cpu/perf_event.c
@@ -93,13 +93,16 @@ struct cpu_hw_events {
struct perf_event *event_list[X86_PMC_IDX_MAX]; /* in enabled order */
};
-#define EVENT_CONSTRAINT(c, n, m) { \
+#define __EVENT_CONSTRAINT(c, n, m, w) {\
{ .idxmsk64[0] = (n) }, \
.code = (c), \
.cmask = (m), \
- .weight = HWEIGHT64((u64)(n)), \
+ .weight = (w), \
}
+#define EVENT_CONSTRAINT(c, n, m) \
+ __EVENT_CONSTRAINT(c, n, m, HWEIGHT(n))
+
#define INTEL_EVENT_CONSTRAINT(c, n) \
EVENT_CONSTRAINT(c, n, INTEL_ARCH_EVTSEL_MASK)
@@ -2646,7 +2649,8 @@ void __init init_hw_perf_events(void)
register_die_notifier(&perf_event_nmi_notifier);
unconstrained = (struct event_constraint)
- EVENT_CONSTRAINT(0, (1ULL << x86_pmu.num_events) - 1, 0);
+ __EVENT_CONSTRAINT(0, (1ULL << x86_pmu.num_events) - 1,
+ 0, x86_pmu.num_events);
pr_info("... version: %d\n", x86_pmu.version);
pr_info("... bit width: %d\n", x86_pmu.event_bits);
Index: linux-2.6/include/linux/bitops.h
===================================================================
--- linux-2.6.orig/include/linux/bitops.h
+++ linux-2.6/include/linux/bitops.h
@@ -45,19 +45,29 @@ static inline unsigned long hweight_long
return sizeof(w) == 4 ? hweight32(w) : hweight64(w);
}
-#define HWEIGHT8(w) \
- ( (!!((w) & (1ULL << 0))) + \
- (!!((w) & (1ULL << 1))) + \
- (!!((w) & (1ULL << 2))) + \
- (!!((w) & (1ULL << 3))) + \
- (!!((w) & (1ULL << 4))) + \
- (!!((w) & (1ULL << 5))) + \
- (!!((w) & (1ULL << 6))) + \
+/*
+ * Clearly slow versions of the hweightN() functions, their benefit is
+ * of course compile time evaluation of constant arguments.
+ */
+#define HWEIGHT8(w) \
+ ( BUILD_BUG_ON_ZERO(!__builtin_constant_p(w)) + \
+ (!!((w) & (1ULL << 0))) + \
+ (!!((w) & (1ULL << 1))) + \
+ (!!((w) & (1ULL << 2))) + \
+ (!!((w) & (1ULL << 3))) + \
+ (!!((w) & (1ULL << 4))) + \
+ (!!((w) & (1ULL << 5))) + \
+ (!!((w) & (1ULL << 6))) + \
(!!((w) & (1ULL << 7))) )
-#define HWEIGHT16(w) (HWEIGHT8(w) + HWEIGHT8(w >> 8))
-#define HWEIGHT32(w) (HWEIGHT16(w) + HWEIGHT16(w >> 16))
-#define HWEIGHT64(w) (HWEIGHT32(w) + HWEIGHT32(w >> 32))
+#define HWEIGHT16(w) (HWEIGHT8(w) + HWEIGHT8((w) >> 8))
+#define HWEIGHT32(w) (HWEIGHT16(w) + HWEIGHT16((w) >> 16))
+#define HWEIGHT64(w) (HWEIGHT32(w) + HWEIGHT32((w) >> 32))
+
+/*
+ * For us lazy bastards
+ */
+#define HWEIGHT(w) HWEIGHT64((u64)(w))
/**
* rol32 - rotate a 32-bit value left
next prev parent reply other threads:[~2010-01-29 11:04 UTC|newest]
Thread overview: 120+ messages / expand[flat|nested] mbox.gz Atom feed top
2010-01-22 15:50 [PATCH 00/10] perf/x86 queue Peter Zijlstra
2010-01-22 15:50 ` [PATCH 01/10] perf_events: improve x86 event scheduling (v5) Peter Zijlstra
2010-01-22 15:50 ` [PATCH 02/10] perf_events: Add fast-path to the rescheduling code Peter Zijlstra
2010-01-22 15:50 ` [PATCH 03/10] perf_event: x86: Allocate the fake_cpuc Peter Zijlstra
2010-01-29 9:27 ` [tip:perf/core] " tip-bot for Peter Zijlstra
2010-01-22 15:50 ` [PATCH 04/10] perf_event: x86: Fixup weight tying issue Peter Zijlstra
2010-01-29 9:27 ` [tip:perf/core] perf_event: x86: Fixup constraints typing issue tip-bot for Peter Zijlstra
2010-01-22 15:50 ` [PATCH 05/10] perf_event: x86: Clean up some of the u64/long bitmask casting Peter Zijlstra
2010-01-29 9:27 ` [tip:perf/core] " tip-bot for Peter Zijlstra
2010-01-22 15:50 ` [PATCH 06/10] perf_event: x86: Reduce some overly long lines with some MACROs Peter Zijlstra
2010-01-29 9:27 ` [tip:perf/core] " tip-bot for Peter Zijlstra
2010-01-22 15:50 ` [PATCH 07/10] bitops: Provide compile time HWEIGHT{8,16,32,64} Peter Zijlstra
2010-01-29 9:28 ` [tip:perf/core] " tip-bot for Peter Zijlstra
2010-01-29 10:01 ` Andrew Morton
2010-01-29 10:04 ` Ingo Molnar
2010-01-29 10:13 ` Andrew Morton
2010-01-29 11:03 ` Peter Zijlstra [this message]
2010-01-29 16:24 ` Linus Torvalds
2010-01-29 22:50 ` H. Peter Anvin
2010-01-30 16:28 ` Peter Zijlstra
2010-02-01 12:43 ` Peter Zijlstra
2010-02-01 19:06 ` H. Peter Anvin
2010-04-06 23:03 ` [tip:core/hweight] bitops: Optimize hweight() by making use of compile-time evaluation tip-bot for Peter Zijlstra
2010-01-29 10:32 ` [PATCH 07/10] bitops: Provide compile time HWEIGHT{8,16,32,64} John Kacur
2010-01-29 11:05 ` Peter Zijlstra
2010-01-29 11:13 ` John Kacur
2010-01-30 0:09 ` H. Peter Anvin
2010-01-30 7:34 ` Ingo Molnar
2010-01-22 15:50 ` [PATCH 08/10] perf_event: Optimize the constraint searching bits Peter Zijlstra
2010-01-22 16:08 ` Stephane Eranian
2010-01-22 16:22 ` Peter Zijlstra
2010-01-22 16:28 ` Stephane Eranian
2010-01-29 9:28 ` [tip:perf/core] perf_event: x86: " tip-bot for Peter Zijlstra
2010-01-22 15:50 ` [PATCH 09/10] perf_event: x86: Optimize constraint weight computation Peter Zijlstra
2010-01-29 9:28 ` [tip:perf/core] " tip-bot for Peter Zijlstra
2010-01-22 15:50 ` [PATCH 10/10] perf_event: Optimize the fast path a little more Peter Zijlstra
2010-01-29 9:28 ` [tip:perf/core] perf_event: x86: " tip-bot for Peter Zijlstra
-- strict thread matches above, loose matches on Subject: below --
2010-01-30 9:45 [PATCH 0/5] [RESEND] FMODE_NONOTIFY and FMODE_NEG_OFFSET bits Wu Fengguang
2010-01-30 9:45 ` [PATCH 1/5] fanotify: fix FMODE_NONOTIFY bit number Wu Fengguang
2010-02-01 20:44 ` Andrew Morton
2010-01-30 9:45 ` [PATCH 2/5] bitops: compile time optimization for hweight_long(CONSTANT) Wu Fengguang
2010-02-01 20:48 ` Andrew Morton
2010-02-03 13:39 ` Wu Fengguang
2010-02-03 15:08 ` Andrew Morton
2010-02-03 15:15 ` Peter Zijlstra
2010-02-03 15:42 ` Andrew Morton
2010-02-03 15:47 ` Peter Zijlstra
2010-02-03 17:11 ` H. Peter Anvin
2010-02-03 18:14 ` Borislav Petkov
2010-02-03 18:47 ` Peter Zijlstra
2010-02-03 19:49 ` H. Peter Anvin
2010-02-04 15:10 ` Borislav Petkov
2010-02-04 15:13 ` Peter Zijlstra
2010-02-04 15:54 ` Borislav Petkov
2010-02-04 16:04 ` Peter Zijlstra
2010-02-05 12:11 ` Borislav Petkov
2010-02-05 12:14 ` Peter Zijlstra
2010-02-05 21:54 ` H. Peter Anvin
2010-02-06 9:36 ` Borislav Petkov
2010-02-07 1:55 ` H. Peter Anvin
2010-02-08 9:28 ` Borislav Petkov
2010-02-08 9:35 ` H. Peter Anvin
2010-02-08 9:59 ` Borislav Petkov
2010-02-11 17:24 ` Borislav Petkov
2010-02-11 17:33 ` H. Peter Anvin
2010-02-12 17:06 ` Borislav Petkov
2010-02-12 17:28 ` H. Peter Anvin
2010-02-12 17:47 ` Borislav Petkov
2010-02-12 19:05 ` H. Peter Anvin
2010-02-17 13:57 ` Michal Marek
2010-02-17 17:20 ` Borislav Petkov
2010-02-17 17:31 ` Michal Marek
2010-02-17 17:34 ` Borislav Petkov
2010-02-17 17:39 ` Michal Marek
2010-02-18 6:19 ` Borislav Petkov
2010-02-19 14:22 ` [PATCH] x86: Add optimized popcnt variants Borislav Petkov
2010-02-19 16:06 ` H. Peter Anvin
2010-02-19 16:45 ` Borislav Petkov
2010-02-19 16:53 ` H. Peter Anvin
2010-02-22 14:17 ` Borislav Petkov
2010-02-22 17:21 ` H. Peter Anvin
2010-02-22 18:49 ` Borislav Petkov
2010-02-22 19:55 ` H. Peter Anvin
2010-02-23 6:37 ` Borislav Petkov
2010-02-23 15:58 ` Borislav Petkov
2010-02-23 17:34 ` H. Peter Anvin
2010-02-23 17:54 ` Borislav Petkov
2010-02-23 18:17 ` H. Peter Anvin
2010-02-23 19:06 ` Borislav Petkov
2010-02-26 5:27 ` H. Peter Anvin
2010-02-26 7:47 ` Borislav Petkov
2010-02-26 17:48 ` H. Peter Anvin
2010-02-27 8:28 ` Borislav Petkov
2010-02-27 20:00 ` H. Peter Anvin
2010-03-09 15:36 ` Borislav Petkov
2010-03-09 15:50 ` Peter Zijlstra
2010-03-09 16:23 ` Borislav Petkov
2010-03-09 16:32 ` Peter Zijlstra
2010-03-09 17:32 ` Borislav Petkov
2010-03-09 17:37 ` Peter Zijlstra
2010-03-18 11:17 ` Borislav Petkov
2010-03-18 11:19 ` [PATCH 1/2] bitops: Optimize hweight() by making use of compile-time evaluation Borislav Petkov
2010-03-18 11:20 ` [PATCH 2/2] x86: Add optimized popcnt variants Borislav Petkov
2010-04-06 23:04 ` [tip:core/hweight] " tip-bot for Borislav Petkov
2010-04-07 7:02 ` Borislav Petkov
2010-02-18 10:51 ` [PATCH 2/5] bitops: compile time optimization for hweight_long(CONSTANT) Peter Zijlstra
2010-02-18 11:51 ` Borislav Petkov
2010-02-14 10:12 ` Peter Zijlstra
2010-02-14 11:24 ` Borislav Petkov
2010-02-14 12:23 ` Peter Zijlstra
2010-02-14 14:19 ` Borislav Petkov
2010-02-14 18:36 ` H. Peter Anvin
2010-02-14 20:28 ` Borislav Petkov
2010-02-14 22:13 ` H. Peter Anvin
2010-02-04 15:16 ` H. Peter Anvin
2010-02-04 15:39 ` Brian Gerst
2010-02-03 17:10 ` H. Peter Anvin
2010-01-30 9:45 ` [PATCH 3/5] vfs: O_* bit numbers uniqueness check Wu Fengguang
2010-01-30 9:45 ` [PATCH 4/5] vfs: introduce FMODE_NEG_OFFSET for allowing negative f_pos Wu Fengguang
2010-01-30 9:45 ` [PATCH 5/5] devmem: dont allow seek to last page Wu Fengguang
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=1264763023.4283.2213.camel@laptop \
--to=peterz@infradead.org \
--cc=akpm@linux-foundation.org \
--cc=eranian@google.com \
--cc=hpa@zytor.com \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-tip-commits@vger.kernel.org \
--cc=mingo@elte.hu \
--cc=mingo@redhat.com \
--cc=tglx@linutronix.de \
--cc=torvalds@linux-foundation.org \
/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