From d4595c8cfd9d34bbb26326c4fb9f1f139442c9e1 Mon Sep 17 00:00:00 2001
From: songkuangshi <kuangshi@kuangshi.info>
Date: Fri, 22 Mar 2019 14:24:05 +0800
Subject: [PATCH] fix avx align error

---
 LightCTR/common/avx.h  | 20 ++++++++++----------
 LightCTR/common/time.h |  2 +-
 build_ring.sh          |  2 +-
 3 files changed, 12 insertions(+), 12 deletions(-)

diff --git a/LightCTR/common/avx.h b/LightCTR/common/avx.h
index 9aef337..6dc467c 100644
--- a/LightCTR/common/avx.h
+++ b/LightCTR/common/avx.h
@@ -22,7 +22,7 @@ inline void avx_vecAdd(const float* x, const float* y, float* res, size_t len) {
     if (len > 7) {
         for (; len > 7; len -= 8) {
             __m256 t = _mm256_add_ps(_mm256_loadu_ps(x), _mm256_loadu_ps(y));
-            _mm256_store_ps(res, t);
+            _mm256_storeu_ps(res, t);
             x += 8;
             y += 8;
             res += 8;
@@ -42,7 +42,7 @@ inline void avx_vecAdd(const float* x, const float const_delta, float* res, size
     if (len > 7) {
         for (; len > 7; len -= 8) {
             __m256 t = _mm256_add_ps(_mm256_loadu_ps(x), delta);
-            _mm256_store_ps(res, t);
+            _mm256_storeu_ps(res, t);
             x += 8;
             res += 8;
         }
@@ -62,7 +62,7 @@ inline void avx_vecScalerAdd(const float* x, const float* y, float* res,
         for (; len > 7; len -= 8) {
             __m256 t = _mm256_add_ps(_mm256_loadu_ps(x),
                                      _mm256_mul_ps(_mm256_loadu_ps(y), _scalar));
-            _mm256_store_ps(res, t);
+            _mm256_storeu_ps(res, t);
             x += 8;
             y += 8;
             res += 8;
@@ -83,7 +83,7 @@ inline void avx_vecScalerAdd(const float* x, const float* y, float* res,
             __m256 t = _mm256_add_ps(_mm256_loadu_ps(x),
                                      _mm256_mul_ps(_mm256_loadu_ps(y),
                                                    _mm256_loadu_ps(y_scalar)));
-            _mm256_store_ps(res, t);
+            _mm256_storeu_ps(res, t);
             x += 8;
             y += 8;
             y_scalar += 8;
@@ -162,7 +162,7 @@ inline void avx_vecSqrt(const float* x, float *res, size_t len) {
     if (len > 7) {
         for (; len > 7; len -= 8) {
             __m256 t = _mm256_sqrt_ps(_mm256_loadu_ps(x));
-            _mm256_store_ps(res, t);
+            _mm256_storeu_ps(res, t);
             x += 8;
             res += 8;
         }
@@ -178,7 +178,7 @@ inline void avx_vecRsqrt(const float* x, float *res, size_t len) {
     if (len > 7) {
         for (; len > 7; len -= 8) {
             __m256 t = _mm256_rsqrt_ps(_mm256_loadu_ps(x));
-            _mm256_store_ps(res, t);
+            _mm256_storeu_ps(res, t);
             x += 8;
             res += 8;
         }
@@ -194,7 +194,7 @@ inline void avx_vecRcp(const float* x, float *res, size_t len) {
     if (len > 7) {
         for (; len > 7; len -= 8) {
             __m256 t = _mm256_rcp_ps(_mm256_loadu_ps(x));
-            _mm256_store_ps(res, t);
+            _mm256_storeu_ps(res, t);
             x += 8;
             res += 8;
         }
@@ -211,7 +211,7 @@ inline void avx_vecScale(const float* x, float *res, size_t len, const float sca
     if (len > 7) {
         for (; len > 7; len -= 8) {
             __m256 t = _mm256_mul_ps(_mm256_loadu_ps(x), _scalar);
-            _mm256_store_ps(res, t);
+            _mm256_storeu_ps(res, t);
             x += 8;
             res += 8;
         }
@@ -227,7 +227,7 @@ inline void avx_vecScale(const float* x, float *res, size_t len, const float* sc
     if (len > 7) {
         for (; len > 7; len -= 8) {
             __m256 t = _mm256_mul_ps(_mm256_loadu_ps(x), _mm256_loadu_ps(scalar));
-            _mm256_store_ps(res, t);
+            _mm256_storeu_ps(res, t);
             x += 8;
             scalar += 8;
             res += 8;
@@ -245,7 +245,7 @@ inline void avx_vecDiv(const float* x, const float* y, float* res, size_t len) {
     if (len > 7) {
         for (; len > 7; len -= 8) {
             __m256 t = _mm256_div_ps(_mm256_loadu_ps(x), _mm256_loadu_ps(y));
-            _mm256_store_ps(res, t);
+            _mm256_storeu_ps(res, t);
             x += 8;
             y += 8;
             res += 8;
diff --git a/LightCTR/common/time.h b/LightCTR/common/time.h
index 1888339..dd15528 100644
--- a/LightCTR/common/time.h
+++ b/LightCTR/common/time.h
@@ -42,7 +42,7 @@ inline time_t __must_inline__ get_now_s(void) {
 
 inline time_t __must_inline__ gettickspan(uint64_t old_tick = get_now_ms()) {
     update_tv();
-    int64_t cur_tick = get_now_ms();
+    uint64_t cur_tick = get_now_ms();
     if (old_tick > cur_tick) {
         return 0;
     }
diff --git a/build_ring.sh b/build_ring.sh
index 2f66b0b..f12146b 100755
--- a/build_ring.sh
+++ b/build_ring.sh
@@ -20,7 +20,7 @@ wait
 echo
 echo
 echo "[Build Success]"
-echo "Please copy different BIN file to corresponding machine, DON'T forget expert LightCTR_PS_NUM, LightCTR_WORKER_NUM and LightCTR_MASTER_ADDR, run Master first"
+echo "Please copy different BIN file to corresponding machine, DON'T forget expert LightCTR_WORKER_NUM and LightCTR_MASTER_ADDR, run Master first"
 echo
 echo "[or] Press any key to run clunster on standalone mode"
 read -n 1