qemu-devel.nongnu.org archive mirror
 help / color / mirror / Atom feed
From: Alexander Monakov <amonakov@ispras.ru>
To: qemu-devel@nongnu.org
Cc: Mikhail Romanov <mmromanov@ispras.ru>,
	Richard Henderson <richard.henderson@linaro.org>,
	Paolo Bonzini <pbonzini@redhat.com>,
	Alexander Monakov <amonakov@ispras.ru>
Subject: [PATCH v3 3/6] util/bufferiszero: remove AVX512 variant
Date: Tue,  6 Feb 2024 23:48:06 +0300	[thread overview]
Message-ID: <20240206204809.9859-4-amonakov@ispras.ru> (raw)
In-Reply-To: <20240206204809.9859-1-amonakov@ispras.ru>

Thanks to early checks in the inline buffer_is_zero wrapper, the SIMD
routines are invoked much more rarely in normal use when most buffers
are non-zero. This makes use of AVX512 unprofitable, as it incurs extra
frequency and voltage transition periods during which the CPU operates
at reduced performance, as described in
https://travisdowns.github.io/blog/2020/01/17/avxfreq1.html

Signed-off-by: Mikhail Romanov <mmromanov@ispras.ru>
Signed-off-by: Alexander Monakov <amonakov@ispras.ru>
---
 util/bufferiszero.c | 36 ++----------------------------------
 1 file changed, 2 insertions(+), 34 deletions(-)

diff --git a/util/bufferiszero.c b/util/bufferiszero.c
index 01050694a6..c037d11d04 100644
--- a/util/bufferiszero.c
+++ b/util/bufferiszero.c
@@ -64,7 +64,7 @@ buffer_is_zero_len_4_plus(const void *buf, size_t len)
     }
 }
 
-#if defined(CONFIG_AVX512F_OPT) || defined(CONFIG_AVX2_OPT) || defined(__SSE2__)
+#if defined(CONFIG_AVX2_OPT) || defined(__SSE2__)
 #include <immintrin.h>
 
 /* Note that each of these vectorized functions require len >= 64.  */
@@ -128,35 +128,6 @@ buffer_zero_avx2(const void *buf, size_t len)
 }
 #endif /* CONFIG_AVX2_OPT */
 
-#ifdef CONFIG_AVX512F_OPT
-static bool __attribute__((target("avx512f")))
-buffer_zero_avx512(const void *buf, size_t len)
-{
-    /* Begin with an unaligned head of 64 bytes.  */
-    __m512i t = _mm512_loadu_si512(buf);
-    __m512i *p = (__m512i *)(((uintptr_t)buf + 5 * 64) & -64);
-    __m512i *e = (__m512i *)(((uintptr_t)buf + len) & -64);
-
-    /* Loop over 64-byte aligned blocks of 256.  */
-    while (p <= e) {
-        __builtin_prefetch(p);
-        if (unlikely(_mm512_test_epi64_mask(t, t))) {
-            return false;
-        }
-        t = p[-4] | p[-3] | p[-2] | p[-1];
-        p += 4;
-    }
-
-    t |= _mm512_loadu_si512(buf + len - 4 * 64);
-    t |= _mm512_loadu_si512(buf + len - 3 * 64);
-    t |= _mm512_loadu_si512(buf + len - 2 * 64);
-    t |= _mm512_loadu_si512(buf + len - 1 * 64);
-
-    return !_mm512_test_epi64_mask(t, t);
-
-}
-#endif /* CONFIG_AVX512F_OPT */
-
 static unsigned __attribute__((noinline))
 select_accel_cpuinfo(unsigned info)
 {
@@ -165,9 +136,6 @@ select_accel_cpuinfo(unsigned info)
         unsigned bit;
         bool (*fn)(const void *, size_t);
     } all[] = {
-#ifdef CONFIG_AVX512F_OPT
-        { CPUINFO_AVX512F, buffer_zero_avx512 },
-#endif
 #ifdef CONFIG_AVX2_OPT
         { CPUINFO_AVX2,    buffer_zero_avx2 },
 #endif
@@ -191,7 +159,7 @@ static unsigned used_accel
     = 0;
 #endif
 
-#if defined(CONFIG_AVX512F_OPT) || defined(CONFIG_AVX2_OPT)
+#if defined(CONFIG_AVX2_OPT)
 static void __attribute__((constructor)) init_accel(void)
 {
     used_accel = select_accel_cpuinfo(cpuinfo_init());
-- 
2.32.0



  parent reply	other threads:[~2024-02-06 20:50 UTC|newest]

Thread overview: 19+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2024-02-06 20:48 [PATCH v3 0/6] Optimize buffer_is_zero Alexander Monakov
2024-02-06 20:48 ` [PATCH v3 1/6] util/bufferiszero: remove SSE4.1 variant Alexander Monakov
2024-02-06 22:24   ` Richard Henderson
2024-02-06 20:48 ` [PATCH v3 2/6] util/bufferiszero: introduce an inline wrapper Alexander Monakov
2024-02-06 22:44   ` Richard Henderson
2024-02-07  7:13     ` Alexander Monakov
2024-02-08 20:07       ` Richard Henderson
2024-02-06 20:48 ` Alexander Monakov [this message]
2024-02-06 22:28   ` [PATCH v3 3/6] util/bufferiszero: remove AVX512 variant Richard Henderson
2024-02-06 23:56   ` Elena Ufimtseva
2024-02-07  6:29     ` Alexander Monakov
2024-02-07 10:38       ` Joao Martins
2024-02-06 20:48 ` [PATCH v3 4/6] util/bufferiszero: remove useless prefetches Alexander Monakov
2024-02-06 22:29   ` Richard Henderson
2024-02-06 20:48 ` [PATCH v3 5/6] util/bufferiszero: optimize SSE2 and AVX2 variants Alexander Monakov
2024-02-06 23:10   ` Richard Henderson
2024-02-06 20:48 ` [PATCH v3 6/6] util/bufferiszero: improve scalar variant Alexander Monakov
2024-02-06 22:34   ` Richard Henderson
2024-02-06 22:46     ` Richard Henderson

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=20240206204809.9859-4-amonakov@ispras.ru \
    --to=amonakov@ispras.ru \
    --cc=mmromanov@ispras.ru \
    --cc=pbonzini@redhat.com \
    --cc=qemu-devel@nongnu.org \
    --cc=richard.henderson@linaro.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;
as well as URLs for NNTP newsgroup(s).