diff --git a/src/nouveau/mme/mme_fermi_builder.c b/src/nouveau/mme/mme_fermi_builder.c
index d6040759ef3..676c0057c6e 100644
--- a/src/nouveau/mme/mme_fermi_builder.c
+++ b/src/nouveau/mme/mme_fermi_builder.c
@@ -323,6 +323,23 @@ mme_fermi_bfe_to(struct mme_builder *b, struct mme_value dst,
    mme_fermi_bfe(fb, dst, pos, x, mme_zero(), bits);
 }
 
+void
+mme_fermi_umul_32x32_32_to_free_srcs(struct mme_builder *b,
+                                     struct mme_value dst,
+                                     struct mme_value x,
+                                     struct mme_value y)
+{
+   mme_while (b, ine, x, mme_zero()) {
+      struct mme_value lsb = mme_and(b, x, mme_imm(1));
+      mme_if (b, ine, lsb, mme_zero()) {
+         mme_add_to(b, dst, dst, y);
+      }
+      mme_free_reg(b, lsb);
+      mme_srl_to(b, x, x, mme_imm(1u));
+      mme_sll_to(b, y, y, mme_imm(1u));
+   }
+}
+
 static struct mme_value
 mme_fermi_load_imm_to_reg(struct mme_builder *b, struct mme_value data)
 {
@@ -685,6 +702,11 @@ mme_fermi_alu_to(struct mme_builder *b,
          return;
       }
       break;
+   case MME_ALU_OP_MUL:
+      x = mme_mov(b, x);
+      y = mme_mov(b, y);
+      mme_fermi_umul_32x32_32_to_free_srcs(b, dst, x, y);
+      return;
    case MME_ALU_OP_SLL:
       mme_fermi_sll_to(fb, dst, x, y);
       return;
diff --git a/src/nouveau/mme/mme_fermi_builder.h b/src/nouveau/mme/mme_fermi_builder.h
index b4c15ddbccb..0e955e7f090 100644
--- a/src/nouveau/mme/mme_fermi_builder.h
+++ b/src/nouveau/mme/mme_fermi_builder.h
@@ -109,6 +109,12 @@ void
 mme_fermi_bfe_to(struct mme_builder *b, struct mme_value dst,
                  struct mme_value x, struct mme_value pos, uint8_t bits);
 
+void
+mme_fermi_umul_32x32_32_to_free_srcs(struct mme_builder *b,
+                                     struct mme_value dst,
+                                     struct mme_value x,
+                                     struct mme_value y);
+
 void
 mme_fermi_merge_to(struct mme_builder *b, struct mme_value dst,
                    struct mme_value x, struct mme_value y,
diff --git a/src/nouveau/mme/tests/mme_builder_test.cpp b/src/nouveau/mme/tests/mme_builder_test.cpp
index 763a1893c49..0ccc55dd9e8 100644
--- a/src/nouveau/mme/tests/mme_builder_test.cpp
+++ b/src/nouveau/mme/tests/mme_builder_test.cpp
@@ -189,6 +189,48 @@ TEST_F(mme_builder_test, sub_imm)
    }
 }
 
+static const uint32_t mul_cases[] = {
+   0x00000000,
+   0x00000001,
+   0x0000005c,
+   0x00c0ffee,
+   0xffffffff,
+   0x0000ffff,
+   0x00008000,
+   0x0001ffff,
+   0xffff8000,
+   0x00010000,
+   0x00020000,
+   0xfffc0000,
+   0xfffe0000,
+};
+
+TEST_F(mme_builder_test, mul)
+{
+   for (auto sim : sims) {
+      mme_builder b;
+      mme_builder_init(&b, sim->devinfo);
+
+      mme_value x = mme_load(&b);
+      mme_value y = mme_load(&b);
+
+      sim->mme_store_data(&b, 0, mme_mul(&b, x, y));
+
+      auto macro = mme_builder_finish_vec(&b);
+
+      for (uint32_t i = 0; i < ARRAY_SIZE(mul_cases); i++) {
+         for (uint32_t j = 0; j < ARRAY_SIZE(mul_cases); j++) {
+            std::vector<uint32_t> params;
+            params.push_back(mul_cases[i]);
+            params.push_back(mul_cases[j]);
+
+            sim->run_macro(macro, params);
+            ASSERT_EQ(sim->data[0], mul_cases[i] * mul_cases[j]);
+         }
+      }
+   }
+}
+
 TEST_F(mme_builder_test, sll_srl)
 {
    static const uint32_t x = 0xac406fe1;