[PATCH v6 0/2] This patch updates AVX512 support for xbzrle

[PATCH v6 0/2] This patch updates AVX512 support for xbzrle

[ling xu]

This patch updates code of avx512 support for xbzrle_encode_buffer
function. We modified code of algorithm and check of avx512. Besides, we provide benchmark in xbzrle-bench.c for performance comparison.

Signed-off-by: ling xu <ling1.xu@intel.com>
Co-authored-by: Zhou Zhao <zhou.zhao@intel.com>
Co-authored-by: Jun Jin <jun.i.jin@intel.com>

ling xu (2):
  Update AVX512 support for xbzrle_encode_buffer
  Unit test code and benchmark code

 meson.build                |  16 ++
 meson_options.txt          |   2 +
 migration/ram.c            |  34 ++-
 migration/xbzrle.c         | 124 ++++++++++
 migration/xbzrle.h         |   4 +
 tests/bench/meson.build    |   4 +
 tests/bench/xbzrle-bench.c | 465 +++++++++++++++++++++++++++++++++++++
 tests/unit/test-xbzrle.c   |  39 +++-
 8 files changed, 680 insertions(+), 8 deletions(-)
 create mode 100644 tests/bench/xbzrle-bench.c

-- 
2.25.1

[PATCH v6 1/2] Update AVX512 support for xbzrle_encode_buffer

[ling xu]

This commit updates code of avx512 support for xbzrle_encode_buffer
function to accelerate xbzrle encoding speed. Runtime check of avx512
support and benchmark for this feature are added. Compared with C
version of xbzrle_encode_buffer function, avx512 version can achieve
50%-70% performance improvement on benchmarking. In addition, if dirty
data is randomly located in 4K page, the avx512 version can achieve
almost 140% performance gain.

Signed-off-by: ling xu <ling1.xu@intel.com>
Co-authored-by: Zhou Zhao <zhou.zhao@intel.com>
Co-authored-by: Jun Jin <jun.i.jin@intel.com>
---
 meson.build        |  16 ++++++
 meson_options.txt  |   2 +
 migration/ram.c    |  34 +++++++++++--
 migration/xbzrle.c | 124 +++++++++++++++++++++++++++++++++++++++++++++
 migration/xbzrle.h |   4 ++
 5 files changed, 177 insertions(+), 3 deletions(-)

diff --git a/meson.build b/meson.build
index 20fddbd707..5d4b82d7f3 100644
--- a/meson.build
+++ b/meson.build
@@ -2264,6 +2264,22 @@ config_host_data.set('CONFIG_AVX512F_OPT', get_option('avx512f') \
     int main(int argc, char *argv[]) { return bar(argv[0]); }
   '''), error_message: 'AVX512F not available').allowed())
 
+config_host_data.set('CONFIG_AVX512BW_OPT', get_option('avx512bw') \
+  .require(have_cpuid_h, error_message: 'cpuid.h not available, cannot enable AVX512BW') \
+  .require(cc.links('''
+    #pragma GCC push_options
+    #pragma GCC target("avx512bw")
+    #include <cpuid.h>
+    #include <immintrin.h>
+    static int bar(void *a) {
+
+      __m512i *x = a;
+      __m512i res= _mm512_abs_epi8(*x);
+      return res[1];
+    }
+    int main(int argc, char *argv[]) { return bar(argv[0]); }
+  '''), error_message: 'AVX512BW not available').allowed())
+
 have_pvrdma = get_option('pvrdma') \
   .require(rdma.found(), error_message: 'PVRDMA requires OpenFabrics libraries') \
   .require(cc.compiles(gnu_source_prefix + '''
diff --git a/meson_options.txt b/meson_options.txt
index e58e158396..07194bf680 100644
--- a/meson_options.txt
+++ b/meson_options.txt
@@ -104,6 +104,8 @@ option('avx2', type: 'feature', value: 'auto',
        description: 'AVX2 optimizations')
 option('avx512f', type: 'feature', value: 'disabled',
        description: 'AVX512F optimizations')
+option('avx512bw', type: 'feature', value: 'auto',
+       description: 'AVX512BW optimizations')
 option('keyring', type: 'feature', value: 'auto',
        description: 'Linux keyring support')
 
diff --git a/migration/ram.c b/migration/ram.c
index dc1de9ddbc..ff4c15c9c3 100644
--- a/migration/ram.c
+++ b/migration/ram.c
@@ -83,6 +83,34 @@
 /* 0x80 is reserved in migration.h start with 0x100 next */
 #define RAM_SAVE_FLAG_COMPRESS_PAGE    0x100
 
+int (*xbzrle_encode_buffer_func)(uint8_t *, uint8_t *, int,
+     uint8_t *, int) = xbzrle_encode_buffer;
+#if defined(CONFIG_AVX512BW_OPT)
+#include "qemu/cpuid.h"
+static void __attribute__((constructor)) init_cpu_flag(void)
+{
+    unsigned max = __get_cpuid_max(0, NULL);
+    int a, b, c, d;
+    if (max >= 1) {
+        __cpuid(1, a, b, c, d);
+         /* We must check that AVX is not just available, but usable.  */
+        if ((c & bit_OSXSAVE) && (c & bit_AVX) && max >= 7) {
+            int bv;
+            __asm("xgetbv" : "=a"(bv), "=d"(d) : "c"(0));
+            __cpuid_count(7, 0, a, b, c, d);
+           /* 0xe6:
+            *  XCR0[7:5] = 111b (OPMASK state, upper 256-bit of ZMM0-ZMM15
+            *                    and ZMM16-ZMM31 state are enabled by OS)
+            *  XCR0[2:1] = 11b (XMM state and YMM state are enabled by OS)
+            */
+            if ((bv & 0xe6) == 0xe6 && (b & bit_AVX512BW)) {
+                xbzrle_encode_buffer_func = xbzrle_encode_buffer_avx512;
+            }
+        }
+    }
+}
+#endif
+
 XBZRLECacheStats xbzrle_counters;
 
 /* struct contains XBZRLE cache and a static page
@@ -802,9 +830,9 @@ static int save_xbzrle_page(RAMState *rs, uint8_t **current_data,
     memcpy(XBZRLE.current_buf, *current_data, TARGET_PAGE_SIZE);
 
     /* XBZRLE encoding (if there is no overflow) */
-    encoded_len = xbzrle_encode_buffer(prev_cached_page, XBZRLE.current_buf,
-                                       TARGET_PAGE_SIZE, XBZRLE.encoded_buf,
-                                       TARGET_PAGE_SIZE);
+    encoded_len = xbzrle_encode_buffer_func(prev_cached_page, XBZRLE.current_buf,
+                                            TARGET_PAGE_SIZE, XBZRLE.encoded_buf,
+                                            TARGET_PAGE_SIZE);
 
     /*
      * Update the cache contents, so that it corresponds to the data
diff --git a/migration/xbzrle.c b/migration/xbzrle.c
index 1ba482ded9..05366e86c0 100644
--- a/migration/xbzrle.c
+++ b/migration/xbzrle.c
@@ -174,3 +174,127 @@ int xbzrle_decode_buffer(uint8_t *src, int slen, uint8_t *dst, int dlen)
 
     return d;
 }
+
+#if defined(CONFIG_AVX512BW_OPT)
+#pragma GCC push_options
+#pragma GCC target("avx512bw")
+#include <immintrin.h>
+int xbzrle_encode_buffer_avx512(uint8_t *old_buf, uint8_t *new_buf, int slen,
+                             uint8_t *dst, int dlen)
+{
+    uint32_t zrun_len = 0, nzrun_len = 0;
+    int d = 0, i = 0, num = 0;
+    uint8_t *nzrun_start = NULL;
+    /* add 1 to include residual part in main loop */
+    uint32_t count512s = (slen >> 6) + 1;
+    /* countResidual is tail of data, i.e., countResidual = slen % 64 */
+    uint32_t count_residual = slen & 0b111111;
+    bool never_same = true;
+    uint64_t mask_residual = 1;
+    mask_residual <<= count_residual;
+    mask_residual -= 1;
+    __m512i r = _mm512_set1_epi32(0);
+
+    while (count512s) {
+        if (d + 2 > dlen) {
+            return -1;
+        }
+
+        int bytes_to_check = 64;
+        uint64_t mask = 0xffffffffffffffff;
+        if (count512s == 1) {
+            bytes_to_check = count_residual;
+            mask = mask_residual;
+        }
+        __m512i old_data = _mm512_mask_loadu_epi8(r,
+                                                  mask, old_buf + i);
+        __m512i new_data = _mm512_mask_loadu_epi8(r,
+                                                  mask, new_buf + i);
+        uint64_t comp = _mm512_cmpeq_epi8_mask(old_data, new_data);
+        count512s--;
+
+        bool is_same = (comp & 0x1);
+        while (bytes_to_check) {
+            if (is_same) {
+                if (nzrun_len) {
+                    d += uleb128_encode_small(dst + d, nzrun_len);
+                    if (d + nzrun_len > dlen) {
+                        return -1;
+                    }
+                    nzrun_start = new_buf + i - nzrun_len;
+                    memcpy(dst + d, nzrun_start, nzrun_len);
+                    d += nzrun_len;
+                    nzrun_len = 0;
+                }
+                /* 64 data at a time for speed */
+                if (count512s && (comp == 0xffffffffffffffff)) {
+                    i += 64;
+                    zrun_len += 64;
+                    break;
+                }
+                never_same = false;
+                num = __builtin_ctzll(~comp);
+                num = (num < bytes_to_check) ? num : bytes_to_check;
+                zrun_len += num;
+                bytes_to_check -= num;
+                comp >>= num;
+                i += num;
+                if (bytes_to_check) {
+                    /* still has different data after same data */
+                    d += uleb128_encode_small(dst + d, zrun_len);
+                    zrun_len = 0;
+                } else {
+                    break;
+                }
+            }
+            if (never_same || zrun_len) {
+                /*
+                 * never_same only acts if
+                 * data begins with diff in first count512s
+                 */
+                d += uleb128_encode_small(dst + d, zrun_len);
+                zrun_len = 0;
+                never_same = false;
+            }
+            /* has diff, 64 data at a time for speed */
+            if ((bytes_to_check == 64) && (comp == 0x0)) {
+                i += 64;
+                nzrun_len += 64;
+                break;
+            }
+            num = __builtin_ctzll(comp);
+            num = (num < bytes_to_check) ? num : bytes_to_check;
+            nzrun_len += num;
+            bytes_to_check -= num;
+            comp >>= num;
+            i += num;
+            if (bytes_to_check) {
+                /* mask like 111000 */
+                d += uleb128_encode_small(dst + d, nzrun_len);
+                /* overflow */
+                if (d + nzrun_len > dlen) {
+                    return -1;
+                }
+                nzrun_start = new_buf + i - nzrun_len;
+                memcpy(dst + d, nzrun_start, nzrun_len);
+                d += nzrun_len;
+                nzrun_len = 0;
+                is_same = true;
+            }
+        }
+    }
+
+    if (nzrun_len != 0) {
+        d += uleb128_encode_small(dst + d, nzrun_len);
+        /* overflow */
+        if (d + nzrun_len > dlen) {
+            return -1;
+        }
+        nzrun_start = new_buf + i - nzrun_len;
+        memcpy(dst + d, nzrun_start, nzrun_len);
+        d += nzrun_len;
+    }
+    return d;
+}
+#pragma GCC pop_options
+#endif
diff --git a/migration/xbzrle.h b/migration/xbzrle.h
index a0db507b9c..6feb49160a 100644
--- a/migration/xbzrle.h
+++ b/migration/xbzrle.h
@@ -18,4 +18,8 @@ int xbzrle_encode_buffer(uint8_t *old_buf, uint8_t *new_buf, int slen,
                          uint8_t *dst, int dlen);
 
 int xbzrle_decode_buffer(uint8_t *src, int slen, uint8_t *dst, int dlen);
+#if defined(CONFIG_AVX512BW_OPT)
+int xbzrle_encode_buffer_avx512(uint8_t *old_buf, uint8_t *new_buf, int slen,
+                                uint8_t *dst, int dlen);
+#endif
 #endif
-- 
2.25.1

[PATCH v6 2/2] Unit test code and benchmark code

[ling xu]

Unit test code is in test-xbzrle.c, and benchmark code is in xbzrle-bench.c
for performance benchmarking.

Signed-off-by: ling xu <ling1.xu@intel.com>
Co-authored-by: Zhou Zhao <zhou.zhao@intel.com>
Co-authored-by: Jun Jin <jun.i.jin@intel.com>
---
 tests/bench/meson.build    |   4 +
 tests/bench/xbzrle-bench.c | 465 +++++++++++++++++++++++++++++++++++++
 tests/unit/test-xbzrle.c   |  39 +++-
 3 files changed, 503 insertions(+), 5 deletions(-)
 create mode 100644 tests/bench/xbzrle-bench.c

diff --git a/tests/bench/meson.build b/tests/bench/meson.build
index 279a8fcc33..daefead58d 100644
--- a/tests/bench/meson.build
+++ b/tests/bench/meson.build
@@ -3,6 +3,10 @@ qht_bench = executable('qht-bench',
                        sources: 'qht-bench.c',
                        dependencies: [qemuutil])
 
+xbzrle_bench = executable('xbzrle-bench',
+                       sources: 'xbzrle-bench.c',
+                       dependencies: [qemuutil,migration])
+
 executable('atomic_add-bench',
            sources: files('atomic_add-bench.c'),
            dependencies: [qemuutil],
diff --git a/tests/bench/xbzrle-bench.c b/tests/bench/xbzrle-bench.c
new file mode 100644
index 0000000000..d71397e6f4
--- /dev/null
+++ b/tests/bench/xbzrle-bench.c
@@ -0,0 +1,465 @@
+/*
+ * Xor Based Zero Run Length Encoding unit tests.
+ *
+ * Copyright 2013 Red Hat, Inc. and/or its affiliates
+ *
+ * Authors:
+ *  Orit Wasserman  <owasserm@redhat.com>
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2 or later.
+ * See the COPYING file in the top-level directory.
+ *
+ */
+#include "qemu/osdep.h"
+#include "qemu/cutils.h"
+#include "../migration/xbzrle.h"
+
+#define XBZRLE_PAGE_SIZE 4096
+
+#if defined(CONFIG_AVX512BW_OPT)
+static bool is_cpu_support_avx512bw;
+#include "qemu/cpuid.h"
+static void __attribute__((constructor)) init_cpu_flag(void)
+{
+    unsigned max = __get_cpuid_max(0, NULL);
+    int a, b, c, d;
+    is_cpu_support_avx512bw = false;
+    if (max >= 1) {
+        __cpuid(1, a, b, c, d);
+         /* We must check that AVX is not just available, but usable.  */
+        if ((c & bit_OSXSAVE) && (c & bit_AVX) && max >= 7) {
+            int bv;
+            __asm("xgetbv" : "=a"(bv), "=d"(d) : "c"(0));
+            __cpuid_count(7, 0, a, b, c, d);
+           /* 0xe6:
+            *  XCR0[7:5] = 111b (OPMASK state, upper 256-bit of ZMM0-ZMM15
+            *                    and ZMM16-ZMM31 state are enabled by OS)
+            *  XCR0[2:1] = 11b (XMM state and YMM state are enabled by OS)
+            */
+            if ((bv & 0xe6) == 0xe6 && (b & bit_AVX512BW)) {
+                is_cpu_support_avx512bw = true;
+            }
+        }
+    }
+    return ;
+}
+#endif
+
+struct ResTime {
+    float t_raw;
+    float t_512;
+};
+
+static void encode_decode_zero(struct ResTime *res)
+{
+    uint8_t *buffer = g_malloc0(XBZRLE_PAGE_SIZE);
+    uint8_t *compressed = g_malloc0(XBZRLE_PAGE_SIZE);
+    uint8_t *buffer512 = g_malloc0(XBZRLE_PAGE_SIZE);
+    uint8_t *compressed512 = g_malloc0(XBZRLE_PAGE_SIZE);
+    int i = 0;
+    int dlen = 0, dlen512 = 0;
+    int diff_len = g_test_rand_int_range(0, XBZRLE_PAGE_SIZE - 1006);
+
+    for (i = diff_len; i > 0; i--) {
+        buffer[1000 + i] = i;
+        buffer512[1000 + i] = i;
+    }
+
+    buffer[1000 + diff_len + 3] = 103;
+    buffer[1000 + diff_len + 5] = 105;
+
+    buffer512[1000 + diff_len + 3] = 103;
+    buffer512[1000 + diff_len + 5] = 105;
+
+    /* encode zero page */
+    time_t t_start, t_end, t_start512, t_end512;
+    t_start = clock();
+    dlen = xbzrle_encode_buffer(buffer, buffer, XBZRLE_PAGE_SIZE, compressed,
+                       XBZRLE_PAGE_SIZE);
+    t_end = clock();
+    float time_val = difftime(t_end, t_start);
+    g_assert(dlen == 0);
+
+    t_start512 = clock();
+    dlen512 = xbzrle_encode_buffer_avx512(buffer512, buffer512, XBZRLE_PAGE_SIZE,
+                                       compressed512, XBZRLE_PAGE_SIZE);
+    t_end512 = clock();
+    float time_val512 = difftime(t_end512, t_start512);
+    g_assert(dlen512 == 0);
+
+    res->t_raw = time_val;
+    res->t_512 = time_val512;
+
+    g_free(buffer);
+    g_free(compressed);
+    g_free(buffer512);
+    g_free(compressed512);
+
+}
+
+static void test_encode_decode_zero_avx512(void)
+{
+    int i;
+    float time_raw = 0.0, time_512 = 0.0;
+    struct ResTime res;
+    for (i = 0; i < 10000; i++) {
+        encode_decode_zero(&res);
+        time_raw += res.t_raw;
+        time_512 += res.t_512;
+    }
+    printf("Zero test:\n");
+    printf("Raw xbzrle_encode time is %f ms\n", time_raw);
+    printf("512 xbzrle_encode time is %f ms\n", time_512);
+}
+
+static void encode_decode_unchanged(struct ResTime *res)
+{
+    uint8_t *compressed = g_malloc0(XBZRLE_PAGE_SIZE);
+    uint8_t *test = g_malloc0(XBZRLE_PAGE_SIZE);
+    uint8_t *compressed512 = g_malloc0(XBZRLE_PAGE_SIZE);
+    uint8_t *test512 = g_malloc0(XBZRLE_PAGE_SIZE);
+    int i = 0;
+    int dlen = 0, dlen512 = 0;
+    int diff_len = g_test_rand_int_range(0, XBZRLE_PAGE_SIZE - 1006);
+
+    for (i = diff_len; i > 0; i--) {
+        test[1000 + i] = i + 4;
+        test512[1000 + i] = i + 4;
+    }
+
+    test[1000 + diff_len + 3] = 107;
+    test[1000 + diff_len + 5] = 109;
+
+    test512[1000 + diff_len + 3] = 107;
+    test512[1000 + diff_len + 5] = 109;
+
+    /* test unchanged buffer */
+    time_t t_start, t_end, t_start512, t_end512;
+    t_start = clock();
+    dlen = xbzrle_encode_buffer(test, test, XBZRLE_PAGE_SIZE, compressed,
+                                XBZRLE_PAGE_SIZE);
+    t_end = clock();
+    float time_val = difftime(t_end, t_start);
+    g_assert(dlen == 0);
+
+    t_start512 = clock();
+    dlen512 = xbzrle_encode_buffer_avx512(test512, test512, XBZRLE_PAGE_SIZE,
+                                       compressed512, XBZRLE_PAGE_SIZE);
+    t_end512 = clock();
+    float time_val512 = difftime(t_end512, t_start512);
+    g_assert(dlen512 == 0);
+
+    res->t_raw = time_val;
+    res->t_512 = time_val512;
+
+    g_free(test);
+    g_free(compressed);
+    g_free(test512);
+    g_free(compressed512);
+
+}
+
+static void test_encode_decode_unchanged_avx512(void)
+{
+    int i;
+    float time_raw = 0.0, time_512 = 0.0;
+    struct ResTime res;
+    for (i = 0; i < 10000; i++) {
+        encode_decode_unchanged(&res);
+        time_raw += res.t_raw;
+        time_512 += res.t_512;
+    }
+    printf("Unchanged test:\n");
+    printf("Raw xbzrle_encode time is %f ms\n", time_raw);
+    printf("512 xbzrle_encode time is %f ms\n", time_512);
+}
+
+static void encode_decode_1_byte(struct ResTime *res)
+{
+    uint8_t *buffer = g_malloc0(XBZRLE_PAGE_SIZE);
+    uint8_t *test = g_malloc0(XBZRLE_PAGE_SIZE);
+    uint8_t *compressed = g_malloc(XBZRLE_PAGE_SIZE);
+    uint8_t *buffer512 = g_malloc0(XBZRLE_PAGE_SIZE);
+    uint8_t *test512 = g_malloc0(XBZRLE_PAGE_SIZE);
+    uint8_t *compressed512 = g_malloc(XBZRLE_PAGE_SIZE);
+    int dlen = 0, rc = 0, dlen512 = 0, rc512 = 0;
+    uint8_t buf[2];
+    uint8_t buf512[2];
+
+    test[XBZRLE_PAGE_SIZE - 1] = 1;
+    test512[XBZRLE_PAGE_SIZE - 1] = 1;
+
+    time_t t_start, t_end, t_start512, t_end512;
+    t_start = clock();
+    dlen = xbzrle_encode_buffer(buffer, test, XBZRLE_PAGE_SIZE, compressed,
+                       XBZRLE_PAGE_SIZE);
+    t_end = clock();
+    float time_val = difftime(t_end, t_start);
+    g_assert(dlen == (uleb128_encode_small(&buf[0], 4095) + 2));
+
+    rc = xbzrle_decode_buffer(compressed, dlen, buffer, XBZRLE_PAGE_SIZE);
+    g_assert(rc == XBZRLE_PAGE_SIZE);
+    g_assert(memcmp(test, buffer, XBZRLE_PAGE_SIZE) == 0);
+
+    t_start512 = clock();
+    dlen512 = xbzrle_encode_buffer_avx512(buffer512, test512, XBZRLE_PAGE_SIZE,
+                                       compressed512, XBZRLE_PAGE_SIZE);
+    t_end512 = clock();
+    float time_val512 = difftime(t_end512, t_start512);
+    g_assert(dlen512 == (uleb128_encode_small(&buf512[0], 4095) + 2));
+
+    rc512 = xbzrle_decode_buffer(compressed512, dlen512, buffer512,
+                                 XBZRLE_PAGE_SIZE);
+    g_assert(rc512 == XBZRLE_PAGE_SIZE);
+    g_assert(memcmp(test512, buffer512, XBZRLE_PAGE_SIZE) == 0);
+
+    res->t_raw = time_val;
+    res->t_512 = time_val512;
+
+    g_free(buffer);
+    g_free(compressed);
+    g_free(test);
+    g_free(buffer512);
+    g_free(compressed512);
+    g_free(test512);
+
+}
+
+static void test_encode_decode_1_byte_avx512(void)
+{
+    int i;
+    float time_raw = 0.0, time_512 = 0.0;
+    struct ResTime res;
+    for (i = 0; i < 10000; i++) {
+        encode_decode_1_byte(&res);
+        time_raw += res.t_raw;
+        time_512 += res.t_512;
+    }
+    printf("1 byte test:\n");
+    printf("Raw xbzrle_encode time is %f ms\n", time_raw);
+    printf("512 xbzrle_encode time is %f ms\n", time_512);
+}
+
+static void encode_decode_overflow(struct ResTime *res)
+{
+    uint8_t *compressed = g_malloc0(XBZRLE_PAGE_SIZE);
+    uint8_t *test = g_malloc0(XBZRLE_PAGE_SIZE);
+    uint8_t *buffer = g_malloc0(XBZRLE_PAGE_SIZE);
+    uint8_t *compressed512 = g_malloc0(XBZRLE_PAGE_SIZE);
+    uint8_t *test512 = g_malloc0(XBZRLE_PAGE_SIZE);
+    uint8_t *buffer512 = g_malloc0(XBZRLE_PAGE_SIZE);
+    int i = 0, rc = 0, rc512 = 0;
+
+    for (i = 0; i < XBZRLE_PAGE_SIZE / 2 - 1; i++) {
+        test[i * 2] = 1;
+        test512[i * 2] = 1;
+    }
+
+    /* encode overflow */
+    time_t t_start, t_end, t_start512, t_end512;
+    t_start = clock();
+    rc = xbzrle_encode_buffer(buffer, test, XBZRLE_PAGE_SIZE, compressed,
+                              XBZRLE_PAGE_SIZE);
+    t_end = clock();
+    float time_val = difftime(t_end, t_start);
+    g_assert(rc == -1);
+
+    t_start512 = clock();
+    rc512 = xbzrle_encode_buffer_avx512(buffer512, test512, XBZRLE_PAGE_SIZE,
+                                     compressed512, XBZRLE_PAGE_SIZE);
+    t_end512 = clock();
+    float time_val512 = difftime(t_end512, t_start512);
+    g_assert(rc512 == -1);
+
+    res->t_raw = time_val;
+    res->t_512 = time_val512;
+
+    g_free(buffer);
+    g_free(compressed);
+    g_free(test);
+    g_free(buffer512);
+    g_free(compressed512);
+    g_free(test512);
+
+}
+
+static void test_encode_decode_overflow_avx512(void)
+{
+    int i;
+    float time_raw = 0.0, time_512 = 0.0;
+    struct ResTime res;
+    for (i = 0; i < 10000; i++) {
+        encode_decode_overflow(&res);
+        time_raw += res.t_raw;
+        time_512 += res.t_512;
+    }
+    printf("Overflow test:\n");
+    printf("Raw xbzrle_encode time is %f ms\n", time_raw);
+    printf("512 xbzrle_encode time is %f ms\n", time_512);
+}
+
+static void encode_decode_range_avx512(struct ResTime *res)
+{
+    uint8_t *buffer = g_malloc0(XBZRLE_PAGE_SIZE);
+    uint8_t *compressed = g_malloc(XBZRLE_PAGE_SIZE);
+    uint8_t *test = g_malloc0(XBZRLE_PAGE_SIZE);
+    uint8_t *buffer512 = g_malloc0(XBZRLE_PAGE_SIZE);
+    uint8_t *compressed512 = g_malloc(XBZRLE_PAGE_SIZE);
+    uint8_t *test512 = g_malloc0(XBZRLE_PAGE_SIZE);
+    int i = 0, rc = 0, rc512 = 0;
+    int dlen = 0, dlen512 = 0;
+
+    int diff_len = g_test_rand_int_range(0, XBZRLE_PAGE_SIZE - 1006);
+
+    for (i = diff_len; i > 0; i--) {
+        buffer[1000 + i] = i;
+        test[1000 + i] = i + 4;
+        buffer512[1000 + i] = i;
+        test512[1000 + i] = i + 4;
+    }
+
+    buffer[1000 + diff_len + 3] = 103;
+    test[1000 + diff_len + 3] = 107;
+
+    buffer[1000 + diff_len + 5] = 105;
+    test[1000 + diff_len + 5] = 109;
+
+    buffer512[1000 + diff_len + 3] = 103;
+    test512[1000 + diff_len + 3] = 107;
+
+    buffer512[1000 + diff_len + 5] = 105;
+    test512[1000 + diff_len + 5] = 109;
+
+    /* test encode/decode */
+    time_t t_start, t_end, t_start512, t_end512;
+    t_start = clock();
+    dlen = xbzrle_encode_buffer(test, buffer, XBZRLE_PAGE_SIZE, compressed,
+                                XBZRLE_PAGE_SIZE);
+    t_end = clock();
+    float time_val = difftime(t_end, t_start);
+    rc = xbzrle_decode_buffer(compressed, dlen, test, XBZRLE_PAGE_SIZE);
+    g_assert(rc < XBZRLE_PAGE_SIZE);
+    g_assert(memcmp(test, buffer, XBZRLE_PAGE_SIZE) == 0);
+
+    t_start512 = clock();
+    dlen512 = xbzrle_encode_buffer_avx512(test512, buffer512, XBZRLE_PAGE_SIZE,
+                                       compressed512, XBZRLE_PAGE_SIZE);
+    t_end512 = clock();
+    float time_val512 = difftime(t_end512, t_start512);
+    rc512 = xbzrle_decode_buffer(compressed512, dlen512, test512, XBZRLE_PAGE_SIZE);
+    g_assert(rc512 < XBZRLE_PAGE_SIZE);
+    g_assert(memcmp(test512, buffer512, XBZRLE_PAGE_SIZE) == 0);
+
+    res->t_raw = time_val;
+    res->t_512 = time_val512;
+
+    g_free(buffer);
+    g_free(compressed);
+    g_free(test);
+    g_free(buffer512);
+    g_free(compressed512);
+    g_free(test512);
+
+}
+
+static void test_encode_decode_avx512(void)
+{
+    int i;
+    float time_raw = 0.0, time_512 = 0.0;
+    struct ResTime res;
+    for (i = 0; i < 10000; i++) {
+        encode_decode_range_avx512(&res);
+        time_raw += res.t_raw;
+        time_512 += res.t_512;
+    }
+    printf("Encode decode test:\n");
+    printf("Raw xbzrle_encode time is %f ms\n", time_raw);
+    printf("512 xbzrle_encode time is %f ms\n", time_512);
+}
+
+static void encode_decode_random(struct ResTime *res)
+{
+    uint8_t *buffer = g_malloc0(XBZRLE_PAGE_SIZE);
+    uint8_t *compressed = g_malloc(XBZRLE_PAGE_SIZE);
+    uint8_t *test = g_malloc0(XBZRLE_PAGE_SIZE);
+    uint8_t *buffer512 = g_malloc0(XBZRLE_PAGE_SIZE);
+    uint8_t *compressed512 = g_malloc(XBZRLE_PAGE_SIZE);
+    uint8_t *test512 = g_malloc0(XBZRLE_PAGE_SIZE);
+    int i = 0, rc = 0, rc512 = 0;
+    int dlen = 0, dlen512 = 0;
+
+    int diff_len = g_test_rand_int_range(0, XBZRLE_PAGE_SIZE - 1);
+    /* store the index of diff */
+    int dirty_index[diff_len];
+    for (int j = 0; j < diff_len; j++) {
+        dirty_index[j] = g_test_rand_int_range(0, XBZRLE_PAGE_SIZE - 1);
+    }
+    for (i = diff_len - 1; i >= 0; i--) {
+        buffer[dirty_index[i]] = i;
+        test[dirty_index[i]] = i + 4;
+        buffer512[dirty_index[i]] = i;
+        test512[dirty_index[i]] = i + 4;
+    }
+
+    time_t t_start, t_end, t_start512, t_end512;
+    t_start = clock();
+    dlen = xbzrle_encode_buffer(test, buffer, XBZRLE_PAGE_SIZE, compressed,
+                                XBZRLE_PAGE_SIZE);
+    t_end = clock();
+    float time_val = difftime(t_end, t_start);
+    rc = xbzrle_decode_buffer(compressed, dlen, test, XBZRLE_PAGE_SIZE);
+    g_assert(rc < XBZRLE_PAGE_SIZE);
+
+    t_start512 = clock();
+    dlen512 = xbzrle_encode_buffer_avx512(test512, buffer512, XBZRLE_PAGE_SIZE,
+                                       compressed512, XBZRLE_PAGE_SIZE);
+    t_end512 = clock();
+    float time_val512 = difftime(t_end512, t_start512);
+    rc512 = xbzrle_decode_buffer(compressed512, dlen512, test512, XBZRLE_PAGE_SIZE);
+    g_assert(rc512 < XBZRLE_PAGE_SIZE);
+
+    res->t_raw = time_val;
+    res->t_512 = time_val512;
+
+    g_free(buffer);
+    g_free(compressed);
+    g_free(test);
+    g_free(buffer512);
+    g_free(compressed512);
+    g_free(test512);
+
+}
+
+static void test_encode_decode_random_avx512(void)
+{
+    int i;
+    float time_raw = 0.0, time_512 = 0.0;
+    struct ResTime res;
+    for (i = 0; i < 10000; i++) {
+        encode_decode_random(&res);
+        time_raw += res.t_raw;
+        time_512 += res.t_512;
+    }
+    printf("Random test:\n");
+    printf("Raw xbzrle_encode time is %f ms\n", time_raw);
+    printf("512 xbzrle_encode time is %f ms\n", time_512);
+}
+
+int main(int argc, char **argv)
+{
+    g_test_init(&argc, &argv, NULL);
+    g_test_rand_int();
+    #if defined(CONFIG_AVX512BW_OPT)
+    if (likely(is_cpu_support_avx512bw)) {
+        g_test_add_func("/xbzrle/encode_decode_zero", test_encode_decode_zero_avx512);
+        g_test_add_func("/xbzrle/encode_decode_unchanged",
+                        test_encode_decode_unchanged_avx512);
+        g_test_add_func("/xbzrle/encode_decode_1_byte", test_encode_decode_1_byte_avx512);
+        g_test_add_func("/xbzrle/encode_decode_overflow",
+                        test_encode_decode_overflow_avx512);
+        g_test_add_func("/xbzrle/encode_decode", test_encode_decode_avx512);
+        g_test_add_func("/xbzrle/encode_decode_random", test_encode_decode_random_avx512);
+    }
+    #endif
+    return g_test_run();
+}
diff --git a/tests/unit/test-xbzrle.c b/tests/unit/test-xbzrle.c
index ef951b6e54..547046d093 100644
--- a/tests/unit/test-xbzrle.c
+++ b/tests/unit/test-xbzrle.c
@@ -16,6 +16,35 @@
 
 #define XBZRLE_PAGE_SIZE 4096
 
+int (*xbzrle_encode_buffer_func)(uint8_t *, uint8_t *, int,
+     uint8_t *, int) = xbzrle_encode_buffer;
+#if defined(CONFIG_AVX512BW_OPT)
+#include "qemu/cpuid.h"
+static void __attribute__((constructor)) init_cpu_flag(void)
+{
+    unsigned max = __get_cpuid_max(0, NULL);
+    int a, b, c, d;
+    if (max >= 1) {
+        __cpuid(1, a, b, c, d);
+         /* We must check that AVX is not just available, but usable.  */
+        if ((c & bit_OSXSAVE) && (c & bit_AVX) && max >= 7) {
+            int bv;
+            __asm("xgetbv" : "=a"(bv), "=d"(d) : "c"(0));
+            __cpuid_count(7, 0, a, b, c, d);
+           /* 0xe6:
+            *  XCR0[7:5] = 111b (OPMASK state, upper 256-bit of ZMM0-ZMM15
+            *                    and ZMM16-ZMM31 state are enabled by OS)
+            *  XCR0[2:1] = 11b (XMM state and YMM state are enabled by OS)
+            */
+            if ((bv & 0xe6) == 0xe6 && (b & bit_AVX512BW)) {
+                xbzrle_encode_buffer_func = xbzrle_encode_buffer_avx512;
+            }
+        }
+    }
+    return ;
+}
+#endif
+
 static void test_uleb(void)
 {
     uint32_t i, val;
@@ -54,7 +83,7 @@ static void test_encode_decode_zero(void)
     buffer[1000 + diff_len + 5] = 105;
 
     /* encode zero page */
-    dlen = xbzrle_encode_buffer(buffer, buffer, XBZRLE_PAGE_SIZE, compressed,
+    dlen = xbzrle_encode_buffer_func(buffer, buffer, XBZRLE_PAGE_SIZE, compressed,
                        XBZRLE_PAGE_SIZE);
     g_assert(dlen == 0);
 
@@ -78,7 +107,7 @@ static void test_encode_decode_unchanged(void)
     test[1000 + diff_len + 5] = 109;
 
     /* test unchanged buffer */
-    dlen = xbzrle_encode_buffer(test, test, XBZRLE_PAGE_SIZE, compressed,
+    dlen = xbzrle_encode_buffer_func(test, test, XBZRLE_PAGE_SIZE, compressed,
                                 XBZRLE_PAGE_SIZE);
     g_assert(dlen == 0);
 
@@ -96,7 +125,7 @@ static void test_encode_decode_1_byte(void)
 
     test[XBZRLE_PAGE_SIZE - 1] = 1;
 
-    dlen = xbzrle_encode_buffer(buffer, test, XBZRLE_PAGE_SIZE, compressed,
+    dlen = xbzrle_encode_buffer_func(buffer, test, XBZRLE_PAGE_SIZE, compressed,
                        XBZRLE_PAGE_SIZE);
     g_assert(dlen == (uleb128_encode_small(&buf[0], 4095) + 2));
 
@@ -121,7 +150,7 @@ static void test_encode_decode_overflow(void)
     }
 
     /* encode overflow */
-    rc = xbzrle_encode_buffer(buffer, test, XBZRLE_PAGE_SIZE, compressed,
+    rc = xbzrle_encode_buffer_func(buffer, test, XBZRLE_PAGE_SIZE, compressed,
                               XBZRLE_PAGE_SIZE);
     g_assert(rc == -1);
 
@@ -152,7 +181,7 @@ static void encode_decode_range(void)
     test[1000 + diff_len + 5] = 109;
 
     /* test encode/decode */
-    dlen = xbzrle_encode_buffer(test, buffer, XBZRLE_PAGE_SIZE, compressed,
+    dlen = xbzrle_encode_buffer_func(test, buffer, XBZRLE_PAGE_SIZE, compressed,
                                 XBZRLE_PAGE_SIZE);
 
     rc = xbzrle_decode_buffer(compressed, dlen, test, XBZRLE_PAGE_SIZE);
-- 
2.25.1

RE: [PATCH v6 1/2] Update AVX512 support for xbzrle_encode_buffer

[Xu, Ling1]

Hi, All,
     This is a "ping" email~. 
     It seems that my patch has been ignored. So I "ping" this patchset. 
     Link for the patch: https://lore.kernel.org/qemu-devel/20220826095719.2887535-2-ling1.xu@intel.com/

Best Regards
Ling

-----Original Message-----
From: Xu, Ling1 <ling1.xu@intel.com> 
Sent: Friday, August 26, 2022 5:57 PM
To: qemu-devel@nongnu.org
Cc: quintela@redhat.com; dgilbert@redhat.com; Xu, Ling1 <ling1.xu@intel.com>; Zhao, Zhou <zhou.zhao@intel.com>; Jin, Jun I <jun.i.jin@intel.com>
Subject: [PATCH v6 1/2] Update AVX512 support for xbzrle_encode_buffer

This commit updates code of avx512 support for xbzrle_encode_buffer function to accelerate xbzrle encoding speed. Runtime check of avx512 support and benchmark for this feature are added. Compared with C version of xbzrle_encode_buffer function, avx512 version can achieve 50%-70% performance improvement on benchmarking. In addition, if dirty data is randomly located in 4K page, the avx512 version can achieve almost 140% performance gain.

Signed-off-by: ling xu <ling1.xu@intel.com>
Co-authored-by: Zhou Zhao <zhou.zhao@intel.com>
Co-authored-by: Jun Jin <jun.i.jin@intel.com>
---
 meson.build        |  16 ++++++
 meson_options.txt  |   2 +
 migration/ram.c    |  34 +++++++++++--
 migration/xbzrle.c | 124 +++++++++++++++++++++++++++++++++++++++++++++
 migration/xbzrle.h |   4 ++
 5 files changed, 177 insertions(+), 3 deletions(-)

diff --git a/meson.build b/meson.build
index 20fddbd707..5d4b82d7f3 100644
--- a/meson.build
+++ b/meson.build
@@ -2264,6 +2264,22 @@ config_host_data.set('CONFIG_AVX512F_OPT', get_option('avx512f') \
     int main(int argc, char *argv[]) { return bar(argv[0]); }
   '''), error_message: 'AVX512F not available').allowed())
 
+config_host_data.set('CONFIG_AVX512BW_OPT', get_option('avx512bw') \
+  .require(have_cpuid_h, error_message: 'cpuid.h not available, cannot 
+enable AVX512BW') \
+  .require(cc.links('''
+    #pragma GCC push_options
+    #pragma GCC target("avx512bw")
+    #include <cpuid.h>
+    #include <immintrin.h>
+    static int bar(void *a) {
+
+      __m512i *x = a;
+      __m512i res= _mm512_abs_epi8(*x);
+      return res[1];
+    }
+    int main(int argc, char *argv[]) { return bar(argv[0]); }  '''), 
+ error_message: 'AVX512BW not available').allowed())
+
 have_pvrdma = get_option('pvrdma') \
   .require(rdma.found(), error_message: 'PVRDMA requires OpenFabrics libraries') \
   .require(cc.compiles(gnu_source_prefix + '''
diff --git a/meson_options.txt b/meson_options.txt index e58e158396..07194bf680 100644
--- a/meson_options.txt
+++ b/meson_options.txt
@@ -104,6 +104,8 @@ option('avx2', type: 'feature', value: 'auto',
        description: 'AVX2 optimizations')  option('avx512f', type: 'feature', value: 'disabled',
        description: 'AVX512F optimizations')
+option('avx512bw', type: 'feature', value: 'auto',
+       description: 'AVX512BW optimizations')
 option('keyring', type: 'feature', value: 'auto',
        description: 'Linux keyring support')
 
diff --git a/migration/ram.c b/migration/ram.c index dc1de9ddbc..ff4c15c9c3 100644
--- a/migration/ram.c
+++ b/migration/ram.c
@@ -83,6 +83,34 @@
 /* 0x80 is reserved in migration.h start with 0x100 next */
 #define RAM_SAVE_FLAG_COMPRESS_PAGE    0x100
 
+int (*xbzrle_encode_buffer_func)(uint8_t *, uint8_t *, int,
+     uint8_t *, int) = xbzrle_encode_buffer; #if 
+defined(CONFIG_AVX512BW_OPT) #include "qemu/cpuid.h"
+static void __attribute__((constructor)) init_cpu_flag(void) {
+    unsigned max = __get_cpuid_max(0, NULL);
+    int a, b, c, d;
+    if (max >= 1) {
+        __cpuid(1, a, b, c, d);
+         /* We must check that AVX is not just available, but usable.  */
+        if ((c & bit_OSXSAVE) && (c & bit_AVX) && max >= 7) {
+            int bv;
+            __asm("xgetbv" : "=a"(bv), "=d"(d) : "c"(0));
+            __cpuid_count(7, 0, a, b, c, d);
+           /* 0xe6:
+            *  XCR0[7:5] = 111b (OPMASK state, upper 256-bit of ZMM0-ZMM15
+            *                    and ZMM16-ZMM31 state are enabled by OS)
+            *  XCR0[2:1] = 11b (XMM state and YMM state are enabled by OS)
+            */
+            if ((bv & 0xe6) == 0xe6 && (b & bit_AVX512BW)) {
+                xbzrle_encode_buffer_func = xbzrle_encode_buffer_avx512;
+            }
+        }
+    }
+}
+#endif
+
 XBZRLECacheStats xbzrle_counters;
 
 /* struct contains XBZRLE cache and a static page @@ -802,9 +830,9 @@ static int save_xbzrle_page(RAMState *rs, uint8_t **current_data,
     memcpy(XBZRLE.current_buf, *current_data, TARGET_PAGE_SIZE);
 
     /* XBZRLE encoding (if there is no overflow) */
-    encoded_len = xbzrle_encode_buffer(prev_cached_page, XBZRLE.current_buf,
-                                       TARGET_PAGE_SIZE, XBZRLE.encoded_buf,
-                                       TARGET_PAGE_SIZE);
+    encoded_len = xbzrle_encode_buffer_func(prev_cached_page, XBZRLE.current_buf,
+                                            TARGET_PAGE_SIZE, XBZRLE.encoded_buf,
+                                            TARGET_PAGE_SIZE);
 
     /*
      * Update the cache contents, so that it corresponds to the data diff --git a/migration/xbzrle.c b/migration/xbzrle.c index 1ba482ded9..05366e86c0 100644
--- a/migration/xbzrle.c
+++ b/migration/xbzrle.c
@@ -174,3 +174,127 @@ int xbzrle_decode_buffer(uint8_t *src, int slen, uint8_t *dst, int dlen)
 
     return d;
 }
+
+#if defined(CONFIG_AVX512BW_OPT)
+#pragma GCC push_options
+#pragma GCC target("avx512bw")
+#include <immintrin.h>
+int xbzrle_encode_buffer_avx512(uint8_t *old_buf, uint8_t *new_buf, int slen,
+                             uint8_t *dst, int dlen) {
+    uint32_t zrun_len = 0, nzrun_len = 0;
+    int d = 0, i = 0, num = 0;
+    uint8_t *nzrun_start = NULL;
+    /* add 1 to include residual part in main loop */
+    uint32_t count512s = (slen >> 6) + 1;
+    /* countResidual is tail of data, i.e., countResidual = slen % 64 */
+    uint32_t count_residual = slen & 0b111111;
+    bool never_same = true;
+    uint64_t mask_residual = 1;
+    mask_residual <<= count_residual;
+    mask_residual -= 1;
+    __m512i r = _mm512_set1_epi32(0);
+
+    while (count512s) {
+        if (d + 2 > dlen) {
+            return -1;
+        }
+
+        int bytes_to_check = 64;
+        uint64_t mask = 0xffffffffffffffff;
+        if (count512s == 1) {
+            bytes_to_check = count_residual;
+            mask = mask_residual;
+        }
+        __m512i old_data = _mm512_mask_loadu_epi8(r,
+                                                  mask, old_buf + i);
+        __m512i new_data = _mm512_mask_loadu_epi8(r,
+                                                  mask, new_buf + i);
+        uint64_t comp = _mm512_cmpeq_epi8_mask(old_data, new_data);
+        count512s--;
+
+        bool is_same = (comp & 0x1);
+        while (bytes_to_check) {
+            if (is_same) {
+                if (nzrun_len) {
+                    d += uleb128_encode_small(dst + d, nzrun_len);
+                    if (d + nzrun_len > dlen) {
+                        return -1;
+                    }
+                    nzrun_start = new_buf + i - nzrun_len;
+                    memcpy(dst + d, nzrun_start, nzrun_len);
+                    d += nzrun_len;
+                    nzrun_len = 0;
+                }
+                /* 64 data at a time for speed */
+                if (count512s && (comp == 0xffffffffffffffff)) {
+                    i += 64;
+                    zrun_len += 64;
+                    break;
+                }
+                never_same = false;
+                num = __builtin_ctzll(~comp);
+                num = (num < bytes_to_check) ? num : bytes_to_check;
+                zrun_len += num;
+                bytes_to_check -= num;
+                comp >>= num;
+                i += num;
+                if (bytes_to_check) {
+                    /* still has different data after same data */
+                    d += uleb128_encode_small(dst + d, zrun_len);
+                    zrun_len = 0;
+                } else {
+                    break;
+                }
+            }
+            if (never_same || zrun_len) {
+                /*
+                 * never_same only acts if
+                 * data begins with diff in first count512s
+                 */
+                d += uleb128_encode_small(dst + d, zrun_len);
+                zrun_len = 0;
+                never_same = false;
+            }
+            /* has diff, 64 data at a time for speed */
+            if ((bytes_to_check == 64) && (comp == 0x0)) {
+                i += 64;
+                nzrun_len += 64;
+                break;
+            }
+            num = __builtin_ctzll(comp);
+            num = (num < bytes_to_check) ? num : bytes_to_check;
+            nzrun_len += num;
+            bytes_to_check -= num;
+            comp >>= num;
+            i += num;
+            if (bytes_to_check) {
+                /* mask like 111000 */
+                d += uleb128_encode_small(dst + d, nzrun_len);
+                /* overflow */
+                if (d + nzrun_len > dlen) {
+                    return -1;
+                }
+                nzrun_start = new_buf + i - nzrun_len;
+                memcpy(dst + d, nzrun_start, nzrun_len);
+                d += nzrun_len;
+                nzrun_len = 0;
+                is_same = true;
+            }
+        }
+    }
+
+    if (nzrun_len != 0) {
+        d += uleb128_encode_small(dst + d, nzrun_len);
+        /* overflow */
+        if (d + nzrun_len > dlen) {
+            return -1;
+        }
+        nzrun_start = new_buf + i - nzrun_len;
+        memcpy(dst + d, nzrun_start, nzrun_len);
+        d += nzrun_len;
+    }
+    return d;
+}
+#pragma GCC pop_options
+#endif
diff --git a/migration/xbzrle.h b/migration/xbzrle.h index a0db507b9c..6feb49160a 100644
--- a/migration/xbzrle.h
+++ b/migration/xbzrle.h
@@ -18,4 +18,8 @@ int xbzrle_encode_buffer(uint8_t *old_buf, uint8_t *new_buf, int slen,
                          uint8_t *dst, int dlen);
 
 int xbzrle_decode_buffer(uint8_t *src, int slen, uint8_t *dst, int dlen);
+#if defined(CONFIG_AVX512BW_OPT)
+int xbzrle_encode_buffer_avx512(uint8_t *old_buf, uint8_t *new_buf, int slen,
+                                uint8_t *dst, int dlen); #endif
 #endif
--
2.25.1

RE: [PATCH v6 1/2] Update AVX512 support for xbzrle_encode_buffer

[Xu, Ling1]

Hi, All,
     This is a "ping" email~. 
     It seems that the newest version of our patch has been ignored. So I "ping" this patchset again. 
     All comments and suggestions have been revised and updated in this V6 version patch, and link for the patch is below:
     https://lore.kernel.org/qemu-devel/20220826095719.2887535-2-ling1.xu@intel.com/
     Looking forward to hearing from you!

Best Regards
Ling

-----Original Message-----
From: Xu, Ling1 <ling1.xu@intel.com> 
Sent: Friday, August 26, 2022 5:57 PM
To: qemu-devel@nongnu.org
Cc: quintela@redhat.com; dgilbert@redhat.com; Xu, Ling1 <ling1.xu@intel.com>; Zhao, Zhou <zhou.zhao@intel.com>; Jin, Jun I <jun.i.jin@intel.com>
Subject: [PATCH v6 1/2] Update AVX512 support for xbzrle_encode_buffer

This commit updates code of avx512 support for xbzrle_encode_buffer function to accelerate xbzrle encoding speed. Runtime check of avx512 support and benchmark for this feature are added. Compared with C version of xbzrle_encode_buffer function, avx512 version can achieve 50%-70% performance improvement on benchmarking. In addition, if dirty data is randomly located in 4K page, the avx512 version can achieve almost 140% performance gain.

Signed-off-by: ling xu <ling1.xu@intel.com>
Co-authored-by: Zhou Zhao <zhou.zhao@intel.com>
Co-authored-by: Jun Jin <jun.i.jin@intel.com>
---
 meson.build        |  16 ++++++
 meson_options.txt  |   2 +
 migration/ram.c    |  34 +++++++++++--
 migration/xbzrle.c | 124 +++++++++++++++++++++++++++++++++++++++++++++
 migration/xbzrle.h |   4 ++
 5 files changed, 177 insertions(+), 3 deletions(-)

diff --git a/meson.build b/meson.build
index 20fddbd707..5d4b82d7f3 100644
--- a/meson.build
+++ b/meson.build
@@ -2264,6 +2264,22 @@ config_host_data.set('CONFIG_AVX512F_OPT', get_option('avx512f') \
     int main(int argc, char *argv[]) { return bar(argv[0]); }
   '''), error_message: 'AVX512F not available').allowed())
 
+config_host_data.set('CONFIG_AVX512BW_OPT', get_option('avx512bw') \
+  .require(have_cpuid_h, error_message: 'cpuid.h not available, cannot 
+enable AVX512BW') \
+  .require(cc.links('''
+    #pragma GCC push_options
+    #pragma GCC target("avx512bw")
+    #include <cpuid.h>
+    #include <immintrin.h>
+    static int bar(void *a) {
+
+      __m512i *x = a;
+      __m512i res= _mm512_abs_epi8(*x);
+      return res[1];
+    }
+    int main(int argc, char *argv[]) { return bar(argv[0]); }  '''), 
+ error_message: 'AVX512BW not available').allowed())
+
 have_pvrdma = get_option('pvrdma') \
   .require(rdma.found(), error_message: 'PVRDMA requires OpenFabrics libraries') \
   .require(cc.compiles(gnu_source_prefix + '''
diff --git a/meson_options.txt b/meson_options.txt index e58e158396..07194bf680 100644
--- a/meson_options.txt
+++ b/meson_options.txt
@@ -104,6 +104,8 @@ option('avx2', type: 'feature', value: 'auto',
        description: 'AVX2 optimizations')  option('avx512f', type: 'feature', value: 'disabled',
        description: 'AVX512F optimizations')
+option('avx512bw', type: 'feature', value: 'auto',
+       description: 'AVX512BW optimizations')
 option('keyring', type: 'feature', value: 'auto',
        description: 'Linux keyring support')
 
diff --git a/migration/ram.c b/migration/ram.c index dc1de9ddbc..ff4c15c9c3 100644
--- a/migration/ram.c
+++ b/migration/ram.c
@@ -83,6 +83,34 @@
 /* 0x80 is reserved in migration.h start with 0x100 next */
 #define RAM_SAVE_FLAG_COMPRESS_PAGE    0x100
 
+int (*xbzrle_encode_buffer_func)(uint8_t *, uint8_t *, int,
+     uint8_t *, int) = xbzrle_encode_buffer; #if 
+defined(CONFIG_AVX512BW_OPT) #include "qemu/cpuid.h"
+static void __attribute__((constructor)) init_cpu_flag(void) {
+    unsigned max = __get_cpuid_max(0, NULL);
+    int a, b, c, d;
+    if (max >= 1) {
+        __cpuid(1, a, b, c, d);
+         /* We must check that AVX is not just available, but usable.  */
+        if ((c & bit_OSXSAVE) && (c & bit_AVX) && max >= 7) {
+            int bv;
+            __asm("xgetbv" : "=a"(bv), "=d"(d) : "c"(0));
+            __cpuid_count(7, 0, a, b, c, d);
+           /* 0xe6:
+            *  XCR0[7:5] = 111b (OPMASK state, upper 256-bit of ZMM0-ZMM15
+            *                    and ZMM16-ZMM31 state are enabled by OS)
+            *  XCR0[2:1] = 11b (XMM state and YMM state are enabled by OS)
+            */
+            if ((bv & 0xe6) == 0xe6 && (b & bit_AVX512BW)) {
+                xbzrle_encode_buffer_func = xbzrle_encode_buffer_avx512;
+            }
+        }
+    }
+}
+#endif
+
 XBZRLECacheStats xbzrle_counters;
 
 /* struct contains XBZRLE cache and a static page @@ -802,9 +830,9 @@ static int save_xbzrle_page(RAMState *rs, uint8_t **current_data,
     memcpy(XBZRLE.current_buf, *current_data, TARGET_PAGE_SIZE);
 
     /* XBZRLE encoding (if there is no overflow) */
-    encoded_len = xbzrle_encode_buffer(prev_cached_page, XBZRLE.current_buf,
-                                       TARGET_PAGE_SIZE, XBZRLE.encoded_buf,
-                                       TARGET_PAGE_SIZE);
+    encoded_len = xbzrle_encode_buffer_func(prev_cached_page, XBZRLE.current_buf,
+                                            TARGET_PAGE_SIZE, XBZRLE.encoded_buf,
+                                            TARGET_PAGE_SIZE);
 
     /*
      * Update the cache contents, so that it corresponds to the data diff --git a/migration/xbzrle.c b/migration/xbzrle.c index 1ba482ded9..05366e86c0 100644
--- a/migration/xbzrle.c
+++ b/migration/xbzrle.c
@@ -174,3 +174,127 @@ int xbzrle_decode_buffer(uint8_t *src, int slen, uint8_t *dst, int dlen)
 
     return d;
 }
+
+#if defined(CONFIG_AVX512BW_OPT)
+#pragma GCC push_options
+#pragma GCC target("avx512bw")
+#include <immintrin.h>
+int xbzrle_encode_buffer_avx512(uint8_t *old_buf, uint8_t *new_buf, int slen,
+                             uint8_t *dst, int dlen) {
+    uint32_t zrun_len = 0, nzrun_len = 0;
+    int d = 0, i = 0, num = 0;
+    uint8_t *nzrun_start = NULL;
+    /* add 1 to include residual part in main loop */
+    uint32_t count512s = (slen >> 6) + 1;
+    /* countResidual is tail of data, i.e., countResidual = slen % 64 */
+    uint32_t count_residual = slen & 0b111111;
+    bool never_same = true;
+    uint64_t mask_residual = 1;
+    mask_residual <<= count_residual;
+    mask_residual -= 1;
+    __m512i r = _mm512_set1_epi32(0);
+
+    while (count512s) {
+        if (d + 2 > dlen) {
+            return -1;
+        }
+
+        int bytes_to_check = 64;
+        uint64_t mask = 0xffffffffffffffff;
+        if (count512s == 1) {
+            bytes_to_check = count_residual;
+            mask = mask_residual;
+        }
+        __m512i old_data = _mm512_mask_loadu_epi8(r,
+                                                  mask, old_buf + i);
+        __m512i new_data = _mm512_mask_loadu_epi8(r,
+                                                  mask, new_buf + i);
+        uint64_t comp = _mm512_cmpeq_epi8_mask(old_data, new_data);
+        count512s--;
+
+        bool is_same = (comp & 0x1);
+        while (bytes_to_check) {
+            if (is_same) {
+                if (nzrun_len) {
+                    d += uleb128_encode_small(dst + d, nzrun_len);
+                    if (d + nzrun_len > dlen) {
+                        return -1;
+                    }
+                    nzrun_start = new_buf + i - nzrun_len;
+                    memcpy(dst + d, nzrun_start, nzrun_len);
+                    d += nzrun_len;
+                    nzrun_len = 0;
+                }
+                /* 64 data at a time for speed */
+                if (count512s && (comp == 0xffffffffffffffff)) {
+                    i += 64;
+                    zrun_len += 64;
+                    break;
+                }
+                never_same = false;
+                num = __builtin_ctzll(~comp);
+                num = (num < bytes_to_check) ? num : bytes_to_check;
+                zrun_len += num;
+                bytes_to_check -= num;
+                comp >>= num;
+                i += num;
+                if (bytes_to_check) {
+                    /* still has different data after same data */
+                    d += uleb128_encode_small(dst + d, zrun_len);
+                    zrun_len = 0;
+                } else {
+                    break;
+                }
+            }
+            if (never_same || zrun_len) {
+                /*
+                 * never_same only acts if
+                 * data begins with diff in first count512s
+                 */
+                d += uleb128_encode_small(dst + d, zrun_len);
+                zrun_len = 0;
+                never_same = false;
+            }
+            /* has diff, 64 data at a time for speed */
+            if ((bytes_to_check == 64) && (comp == 0x0)) {
+                i += 64;
+                nzrun_len += 64;
+                break;
+            }
+            num = __builtin_ctzll(comp);
+            num = (num < bytes_to_check) ? num : bytes_to_check;
+            nzrun_len += num;
+            bytes_to_check -= num;
+            comp >>= num;
+            i += num;
+            if (bytes_to_check) {
+                /* mask like 111000 */
+                d += uleb128_encode_small(dst + d, nzrun_len);
+                /* overflow */
+                if (d + nzrun_len > dlen) {
+                    return -1;
+                }
+                nzrun_start = new_buf + i - nzrun_len;
+                memcpy(dst + d, nzrun_start, nzrun_len);
+                d += nzrun_len;
+                nzrun_len = 0;
+                is_same = true;
+            }
+        }
+    }
+
+    if (nzrun_len != 0) {
+        d += uleb128_encode_small(dst + d, nzrun_len);
+        /* overflow */
+        if (d + nzrun_len > dlen) {
+            return -1;
+        }
+        nzrun_start = new_buf + i - nzrun_len;
+        memcpy(dst + d, nzrun_start, nzrun_len);
+        d += nzrun_len;
+    }
+    return d;
+}
+#pragma GCC pop_options
+#endif
diff --git a/migration/xbzrle.h b/migration/xbzrle.h index a0db507b9c..6feb49160a 100644
--- a/migration/xbzrle.h
+++ b/migration/xbzrle.h
@@ -18,4 +18,8 @@ int xbzrle_encode_buffer(uint8_t *old_buf, uint8_t *new_buf, int slen,
                          uint8_t *dst, int dlen);
 
 int xbzrle_decode_buffer(uint8_t *src, int slen, uint8_t *dst, int dlen);
+#if defined(CONFIG_AVX512BW_OPT)
+int xbzrle_encode_buffer_avx512(uint8_t *old_buf, uint8_t *new_buf, int slen,
+                                uint8_t *dst, int dlen); #endif
 #endif
--
2.25.1

Re: [PATCH v6 1/2] Update AVX512 support for xbzrle_encode_buffer

[Juan Quintela]

ling xu <ling1.xu@intel.com> wrote:
> This commit updates code of avx512 support for xbzrle_encode_buffer
> function to accelerate xbzrle encoding speed. Runtime check of avx512
> support and benchmark for this feature are added. Compared with C
> version of xbzrle_encode_buffer function, avx512 version can achieve
> 50%-70% performance improvement on benchmarking. In addition, if dirty
> data is randomly located in 4K page, the avx512 version can achieve
> almost 140% performance gain.
>
> Signed-off-by: ling xu <ling1.xu@intel.com>
> Co-authored-by: Zhou Zhao <zhou.zhao@intel.com>
> Co-authored-by: Jun Jin <jun.i.jin@intel.com>

Reviewed-by: Juan Quintela <quintela@redhat.com>

queued.

Re: [PATCH v6 2/2] Unit test code and benchmark code

[Juan Quintela]

ling xu <ling1.xu@intel.com> wrote:
> Unit test code is in test-xbzrle.c, and benchmark code is in xbzrle-bench.c
> for performance benchmarking.
>
> Signed-off-by: ling xu <ling1.xu@intel.com>
> Co-authored-by: Zhou Zhao <zhou.zhao@intel.com>
> Co-authored-by: Jun Jin <jun.i.jin@intel.com>

Reviewed-by: Juan Quintela <quintela@redhat.com>

queued.