diff --git a/be/src/util/simd/vstring_function.h b/be/src/util/simd/vstring_function.h
index a1db67325e74c15..5da2a67e34e5100 100644
--- a/be/src/util/simd/vstring_function.h
+++ b/be/src/util/simd/vstring_function.h
@@ -108,21 +108,24 @@ class VStringFunctions {
             return end;
         }
         const auto* p = end;
-#if defined(__SSE2__) || defined(__aarch64__)
+#if defined(__AVX2__)
         if constexpr (trim_single) {
+            constexpr auto AVX2_BYTES = sizeof(__m256i);
+            const auto ch = remove_str.data[0];
             const auto size = end - begin;
-            const auto SSE2_BYTES = sizeof(__m128i);
-            const auto* const sse2_begin = end - (size & ~(SSE2_BYTES - 1));
-            const auto spaces = _mm_set1_epi8(remove_str.data[0]);
-            for (p = end - SSE2_BYTES; p >= sse2_begin; p -= SSE2_BYTES) {
-                uint32_t masks =
-                        _mm_movemask_epi8(_mm_cmpeq_epi8(_mm_loadu_si128((__m128i*)p), spaces));
-                int pos = __builtin_clz(~(masks << SSE2_BYTES));
-                if (pos < SSE2_BYTES) {
-                    return p + SSE2_BYTES - pos;
+            const auto* const avx2_begin = end - size / AVX2_BYTES * AVX2_BYTES;
+            const auto spaces = _mm256_set1_epi8(ch);
+            for (p = end - AVX2_BYTES; p >= avx2_begin; p -= AVX2_BYTES) {
+                uint32_t masks = _mm256_movemask_epi8(
+                        _mm256_cmpeq_epi8(_mm256_loadu_si256((__m256i*)p), spaces));
+                if ((~masks)) {
+                    break;
                 }
             }
-            p += SSE2_BYTES;
+            p += AVX2_BYTES;
+            for (; (p - 1) >= begin && *(p - 1) == ch; p--) {
+            }
+            return p;
         }
 #endif
         const auto remove_size = remove_str.size;
@@ -144,20 +147,23 @@ class VStringFunctions {
             return begin;
         }
         const auto* p = begin;
-#if defined(__SSE2__) || defined(__aarch64__)
+#if defined(__AVX2__)
         if constexpr (trim_single) {
+            constexpr auto AVX2_BYTES = sizeof(__m256i);
+            const auto ch = remove_str.data[0];
             const auto size = end - begin;
-            const auto SSE2_BYTES = sizeof(__m128i);
-            const auto* const sse2_end = begin + (size & ~(SSE2_BYTES - 1));
-            const auto spaces = _mm_set1_epi8(remove_str.data[0]);
-            for (; p < sse2_end; p += SSE2_BYTES) {
-                uint32_t masks =
-                        _mm_movemask_epi8(_mm_cmpeq_epi8(_mm_loadu_si128((__m128i*)p), spaces));
-                int pos = __builtin_ctz((1U << SSE2_BYTES) | ~masks);
-                if (pos < SSE2_BYTES) {
-                    return p + pos;
+            const auto* const avx2_end = begin + size / AVX2_BYTES * AVX2_BYTES;
+            const auto spaces = _mm256_set1_epi8(ch);
+            for (; p < avx2_end; p += AVX2_BYTES) {
+                uint32_t masks = _mm256_movemask_epi8(
+                        _mm256_cmpeq_epi8(_mm256_loadu_si256((__m256i*)p), spaces));
+                if ((~masks)) {
+                    break;
                 }
             }
+            for (; p < end && *p == ch; ++p) {
+            }
+            return p;
         }
 #endif
 
diff --git a/regression-test/suites/correctness/test_trim_new_parameters.groovy b/regression-test/suites/correctness/test_trim_new_parameters.groovy
index 3209eb7aae743de..17ac4a0c65eae5b 100644
--- a/regression-test/suites/correctness/test_trim_new_parameters.groovy
+++ b/regression-test/suites/correctness/test_trim_new_parameters.groovy
@@ -67,4 +67,7 @@ suite("test_trim_new_parameters") {
 
     rtrim = sql "select rtrim('bcTTTabcabc','abc')"
     assertEquals(rtrim[0][0], 'bcTTT')   
+
+    trim_one = sql "select trim('aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaabaaaaaaaaaaabcTTTabcabcaaaaaaaaaaaaaaaaaaaaaaaaaabaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa','a')"
+    assertEquals(trim_one[0][0], 'baaaaaaaaaaabcTTTabcabcaaaaaaaaaaaaaaaaaaaaaaaaaab')  
 }