diff --git a/src/cpu/x64/gemm/f32/jit_avx512_core_gemm_smalln_tn_f32_kern.cpp b/src/cpu/x64/gemm/f32/jit_avx512_core_gemm_smalln_tn_f32_kern.cpp
index a79c6eed0d5..b5bd7cf7000 100644
--- a/src/cpu/x64/gemm/f32/jit_avx512_core_gemm_smalln_tn_f32_kern.cpp
+++ b/src/cpu/x64/gemm/f32/jit_avx512_core_gemm_smalln_tn_f32_kern.cpp
@@ -589,7 +589,7 @@ struct xbyak_gemm_smalln_tn : public jit_generator {
             int endval = (MROW < 5) ? MROW : 5;
             for (int ii = 8; ii < 8 + endval; ii++) {
                 // Storing A values in zmm_reg[8-12]
-                vmovups(zmm_reg[ii] | (krem ? k_rem : k0), ptr[AO2]);
+                vmovups(zmm_reg[ii] | (krem ? k_rem : k0) | T_z, ptr[AO2]);
                 add(AO2, LDA);
             }
             for (int ii = 0; ii < endval; ii++) {
@@ -614,7 +614,7 @@ struct xbyak_gemm_smalln_tn : public jit_generator {
                     ? 8
                     : MROW; // Do not process more than 8 rows here.
             for (int ii = 0; ii < MROW2; ii++) {
-                vmovups(zmm_reg[ii] | (krem ? k_rem : k0), ptr[AO2]);
+                vmovups(zmm_reg[ii] | (krem ? k_rem : k0) | T_z, ptr[AO2]);
                 add(AO2, LDA);
             }
             for (int ii = 0; ii < MROW2; ii++) {
@@ -627,7 +627,7 @@ struct xbyak_gemm_smalln_tn : public jit_generator {
             if (MROW > 8) {
                 vmovaps(zmm_reg[0], zmm_reg[15]);
                 for (int ii = 8; ii < MROW; ii++) {
-                    vmovups(zmm_reg[ii] | (krem ? k_rem : k0), ptr[AO2]);
+                    vmovups(zmm_reg[ii] | (krem ? k_rem : k0) | T_z, ptr[AO2]);
                     add(AO2, LDA);
                 }
                 for (int ii = 8; ii < MROW; ii++)