qemu-devel.nongnu.org archive mirror
 help / color / mirror / Atom feed
From: Robert Hoo <robert.hu@linux.intel.com>
To: qemu-devel@nongnu.org, pbonzini@redhat.com, richard.henderson@linaro.org
Cc: robert.hu@intel.com, Robert Hoo <robert.hu@linux.intel.com>
Subject: [PATCH 2/2] util/bufferiszero: improve avx2 accelerator
Date: Wed, 25 Mar 2020 14:50:21 +0800	[thread overview]
Message-ID: <1585119021-46593-2-git-send-email-robert.hu@linux.intel.com> (raw)
In-Reply-To: <1585119021-46593-1-git-send-email-robert.hu@linux.intel.com>

By increasing avx2 length_to_accel to 128, we can simplify its logic and reduce a
branch.

The authorship of this patch actually belongs to Richard Henderson <richard.henderson@linaro.org>,
I just fix a boudary case on his original patch.

Suggested-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Robert Hoo <robert.hu@linux.intel.com>
---
 util/bufferiszero.c | 26 +++++++++-----------------
 1 file changed, 9 insertions(+), 17 deletions(-)

diff --git a/util/bufferiszero.c b/util/bufferiszero.c
index b801253..695bb4c 100644
--- a/util/bufferiszero.c
+++ b/util/bufferiszero.c
@@ -158,27 +158,19 @@ buffer_zero_avx2(const void *buf, size_t len)
     __m256i *p = (__m256i *)(((uintptr_t)buf + 5 * 32) & -32);
     __m256i *e = (__m256i *)(((uintptr_t)buf + len) & -32);
 
-    if (likely(p <= e)) {
-        /* Loop over 32-byte aligned blocks of 128.  */
-        do {
-            __builtin_prefetch(p);
-            if (unlikely(!_mm256_testz_si256(t, t))) {
-                return false;
-            }
-            t = p[-4] | p[-3] | p[-2] | p[-1];
-            p += 4;
-        } while (p <= e);
-    } else {
-        t |= _mm256_loadu_si256(buf + 32);
-        if (len <= 128) {
-            goto last2;
+    /* Loop over 32-byte aligned blocks of 128.  */
+    while (p <= e) {
+        __builtin_prefetch(p);
+        if (unlikely(!_mm256_testz_si256(t, t))) {
+            return false;
         }
-    }
+        t = p[-4] | p[-3] | p[-2] | p[-1];
+        p += 4;
+    } ;
 
     /* Finish the last block of 128 unaligned.  */
     t |= _mm256_loadu_si256(buf + len - 4 * 32);
     t |= _mm256_loadu_si256(buf + len - 3 * 32);
- last2:
     t |= _mm256_loadu_si256(buf + len - 2 * 32);
     t |= _mm256_loadu_si256(buf + len - 1 * 32);
 
@@ -263,7 +255,7 @@ static void init_accel(unsigned cache)
     }
     if (cache & CACHE_AVX2) {
         fn = buffer_zero_avx2;
-        length_to_accel = 64;
+        length_to_accel = 128;
     }
 #endif
 #ifdef CONFIG_AVX512F_OPT
-- 
1.8.3.1



  reply	other threads:[~2020-03-25  6:51 UTC|newest]

Thread overview: 7+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-03-25  6:50 [PATCH 1/2] util/bufferiszero: assign length_to_accel value for each accelerator case Robert Hoo
2020-03-25  6:50 ` Robert Hoo [this message]
2020-03-25 12:54   ` [PATCH 2/2] util/bufferiszero: improve avx2 accelerator Eric Blake
2020-03-26  2:09     ` Hu, Robert
2020-03-26  9:43       ` Paolo Bonzini
2020-03-26 13:26       ` Eric Blake
2020-03-26 13:51         ` Robert Hoo

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=1585119021-46593-2-git-send-email-robert.hu@linux.intel.com \
    --to=robert.hu@linux.intel.com \
    --cc=pbonzini@redhat.com \
    --cc=qemu-devel@nongnu.org \
    --cc=richard.henderson@linaro.org \
    --cc=robert.hu@intel.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).