From: Uros Bizjak <ubizjak@gmail.com>
To: x86@kernel.org, linux-kernel@vger.kernel.org
Cc: Uros Bizjak <ubizjak@gmail.com>,
Thomas Gleixner <tglx@linutronix.de>,
Ingo Molnar <mingo@kernel.org>, Borislav Petkov <bp@alien8.de>,
Dave Hansen <dave.hansen@linux.intel.com>,
"H. Peter Anvin" <hpa@zytor.com>
Subject: [PATCH -tip 1/2] x86/hweight: Fix false output register dependency of POPCNT insn
Date: Tue, 25 Mar 2025 17:48:37 +0100 [thread overview]
Message-ID: <20250325164854.199420-1-ubizjak@gmail.com> (raw)
On Sandy/Ivy Bridge and later Intel processors, the POPCNT instruction
appears to have a false dependency on the destination register. Even
though the instruction only writes to it, the instruction will wait
until destination is ready before executing. This false dependency
was fixed for Cannon Lake (and later) processors.
Fix false dependency by clearing the destination register first.
The x86_64 defconfig object size increases by 779 bytes:
text data bss dec hex filename
27341418 4643015 814852 32799285 1f47a35 vmlinux-old.o
27342197 4643015 814852 32800064 1f47d40 vmlinux-new.o
Signed-off-by: Uros Bizjak <ubizjak@gmail.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
---
arch/x86/include/asm/arch_hweight.h | 23 +++++++++++++++++------
1 file changed, 17 insertions(+), 6 deletions(-)
diff --git a/arch/x86/include/asm/arch_hweight.h b/arch/x86/include/asm/arch_hweight.h
index cbc6157f0b4b..aa0b3bd309fc 100644
--- a/arch/x86/include/asm/arch_hweight.h
+++ b/arch/x86/include/asm/arch_hweight.h
@@ -4,12 +4,21 @@
#include <asm/cpufeatures.h>
+/*
+ * On Sandy/Ivy Bridge and later Intel processors, the POPCNT instruction
+ * appears to have a false dependency on the destination register. Even
+ * though the instruction only writes to it, the instruction will wait
+ * until destination is ready before executing. This false dependency
+ * was fixed for Cannon Lake (and later) processors.
+ */
+#define ASM_FORCE_CLR "xorl %k[cnt], %k[cnt]\n\t"
+
#ifdef CONFIG_64BIT
#define REG_IN "D"
-#define REG_OUT "a"
+#define ASM_CLR ASM_FORCE_CLR
#else
#define REG_IN "a"
-#define REG_OUT "a"
+#define ASM_CLR
#endif
static __always_inline unsigned int __arch_hweight32(unsigned int w)
@@ -18,8 +27,9 @@ static __always_inline unsigned int __arch_hweight32(unsigned int w)
asm_inline (ALTERNATIVE(ANNOTATE_IGNORE_ALTERNATIVE
"call __sw_hweight32",
- "popcntl %[val], %[cnt]", X86_FEATURE_POPCNT)
- : [cnt] "=" REG_OUT (res), ASM_CALL_CONSTRAINT
+ ASM_CLR "popcntl %[val], %[cnt]",
+ X86_FEATURE_POPCNT)
+ : [cnt] "=a" (res), ASM_CALL_CONSTRAINT
: [val] REG_IN (w));
return res;
@@ -48,8 +58,9 @@ static __always_inline unsigned long __arch_hweight64(__u64 w)
asm_inline (ALTERNATIVE(ANNOTATE_IGNORE_ALTERNATIVE
"call __sw_hweight64",
- "popcntq %[val], %[cnt]", X86_FEATURE_POPCNT)
- : [cnt] "=" REG_OUT (res), ASM_CALL_CONSTRAINT
+ ASM_CLR "popcntq %[val], %[cnt]",
+ X86_FEATURE_POPCNT)
+ : [cnt] "=a" (res), ASM_CALL_CONSTRAINT
: [val] REG_IN (w));
return res;
--
2.42.0
next reply other threads:[~2025-03-25 16:49 UTC|newest]
Thread overview: 25+ messages / expand[flat|nested] mbox.gz Atom feed top
2025-03-25 16:48 Uros Bizjak [this message]
2025-03-25 16:48 ` [PATCH -tip 2/2] x86/hweight: Use POPCNT when available with X86_NATIVE_CPU option Uros Bizjak
2025-03-25 17:11 ` Borislav Petkov
2025-03-30 15:15 ` Uros Bizjak
2025-03-30 17:31 ` Borislav Petkov
2025-03-30 18:47 ` Ingo Molnar
2025-03-30 19:06 ` Borislav Petkov
2025-03-30 19:20 ` Ingo Molnar
2025-03-30 19:28 ` Borislav Petkov
2025-03-25 21:56 ` Ingo Molnar
2025-03-29 9:19 ` Uros Bizjak
2025-03-29 11:00 ` David Laight
2025-03-30 7:49 ` Uros Bizjak
2025-03-30 18:02 ` David Laight
2025-03-29 23:10 ` H. Peter Anvin
2025-03-30 6:54 ` Uros Bizjak
2025-03-30 9:56 ` Ingo Molnar
2025-03-30 16:07 ` Uros Bizjak
2025-03-30 18:15 ` David Laight
2025-03-30 22:44 ` H. Peter Anvin
2025-03-30 18:54 ` Ingo Molnar
2025-03-25 17:09 ` [PATCH -tip 1/2] x86/hweight: Fix false output register dependency of POPCNT insn Borislav Petkov
2025-03-25 17:17 ` Uros Bizjak
2025-03-25 17:44 ` Borislav Petkov
2025-03-25 21:50 ` 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=20250325164854.199420-1-ubizjak@gmail.com \
--to=ubizjak@gmail.com \
--cc=bp@alien8.de \
--cc=dave.hansen@linux.intel.com \
--cc=hpa@zytor.com \
--cc=linux-kernel@vger.kernel.org \
--cc=mingo@kernel.org \
--cc=tglx@linutronix.de \
--cc=x86@kernel.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