target/riscv: vector single-width averaging add and subtract

Signed-off-by: LIU Zhiwei <zhiwei_liu@c-sky.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20200701152549.1218-26-zhiwei_liu@c-sky.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

1 files changed, 100 insertions, 0 deletions

diff --git a/target/riscv/vector_helper.c b/target/riscv/vector_helper.c
index 1277aa1c10..0b2119b6cc 100644
--- a/target/riscv/vector_helper.c
+++ b/target/riscv/vector_helper.c
@@ -2497,3 +2497,103 @@ GEN_VEXT_VX_RM(vssub_vx_b, 1, 1, clearb)
 GEN_VEXT_VX_RM(vssub_vx_h, 2, 2, clearh)
 GEN_VEXT_VX_RM(vssub_vx_w, 4, 4, clearl)
 GEN_VEXT_VX_RM(vssub_vx_d, 8, 8, clearq)
+
+/* Vector Single-Width Averaging Add and Subtract */
+static inline uint8_t get_round(int vxrm, uint64_t v, uint8_t shift)
+{
+    uint8_t d = extract64(v, shift, 1);
+    uint8_t d1;
+    uint64_t D1, D2;
+
+    if (shift == 0 || shift > 64) {
+        return 0;
+    }
+
+    d1 = extract64(v, shift - 1, 1);
+    D1 = extract64(v, 0, shift);
+    if (vxrm == 0) { /* round-to-nearest-up (add +0.5 LSB) */
+        return d1;
+    } else if (vxrm == 1) { /* round-to-nearest-even */
+        if (shift > 1) {
+            D2 = extract64(v, 0, shift - 1);
+            return d1 & ((D2 != 0) | d);
+        } else {
+            return d1 & d;
+        }
+    } else if (vxrm == 3) { /* round-to-odd (OR bits into LSB, aka "jam") */
+        return !d & (D1 != 0);
+    }
+    return 0; /* round-down (truncate) */
+}
+
+static inline int32_t aadd32(CPURISCVState *env, int vxrm, int32_t a, int32_t b)
+{
+    int64_t res = (int64_t)a + b;
+    uint8_t round = get_round(vxrm, res, 1);
+
+    return (res >> 1) + round;
+}
+
+static inline int64_t aadd64(CPURISCVState *env, int vxrm, int64_t a, int64_t b)
+{
+    int64_t res = a + b;
+    uint8_t round = get_round(vxrm, res, 1);
+    int64_t over = (res ^ a) & (res ^ b) & INT64_MIN;
+
+    /* With signed overflow, bit 64 is inverse of bit 63. */
+    return ((res >> 1) ^ over) + round;
+}
+
+RVVCALL(OPIVV2_RM, vaadd_vv_b, OP_SSS_B, H1, H1, H1, aadd32)
+RVVCALL(OPIVV2_RM, vaadd_vv_h, OP_SSS_H, H2, H2, H2, aadd32)
+RVVCALL(OPIVV2_RM, vaadd_vv_w, OP_SSS_W, H4, H4, H4, aadd32)
+RVVCALL(OPIVV2_RM, vaadd_vv_d, OP_SSS_D, H8, H8, H8, aadd64)
+GEN_VEXT_VV_RM(vaadd_vv_b, 1, 1, clearb)
+GEN_VEXT_VV_RM(vaadd_vv_h, 2, 2, clearh)
+GEN_VEXT_VV_RM(vaadd_vv_w, 4, 4, clearl)
+GEN_VEXT_VV_RM(vaadd_vv_d, 8, 8, clearq)
+
+RVVCALL(OPIVX2_RM, vaadd_vx_b, OP_SSS_B, H1, H1, aadd32)
+RVVCALL(OPIVX2_RM, vaadd_vx_h, OP_SSS_H, H2, H2, aadd32)
+RVVCALL(OPIVX2_RM, vaadd_vx_w, OP_SSS_W, H4, H4, aadd32)
+RVVCALL(OPIVX2_RM, vaadd_vx_d, OP_SSS_D, H8, H8, aadd64)
+GEN_VEXT_VX_RM(vaadd_vx_b, 1, 1, clearb)
+GEN_VEXT_VX_RM(vaadd_vx_h, 2, 2, clearh)
+GEN_VEXT_VX_RM(vaadd_vx_w, 4, 4, clearl)
+GEN_VEXT_VX_RM(vaadd_vx_d, 8, 8, clearq)
+
+static inline int32_t asub32(CPURISCVState *env, int vxrm, int32_t a, int32_t b)
+{
+    int64_t res = (int64_t)a - b;
+    uint8_t round = get_round(vxrm, res, 1);
+
+    return (res >> 1) + round;
+}
+
+static inline int64_t asub64(CPURISCVState *env, int vxrm, int64_t a, int64_t b)
+{
+    int64_t res = (int64_t)a - b;
+    uint8_t round = get_round(vxrm, res, 1);
+    int64_t over = (res ^ a) & (a ^ b) & INT64_MIN;
+
+    /* With signed overflow, bit 64 is inverse of bit 63. */
+    return ((res >> 1) ^ over) + round;
+}
+
+RVVCALL(OPIVV2_RM, vasub_vv_b, OP_SSS_B, H1, H1, H1, asub32)
+RVVCALL(OPIVV2_RM, vasub_vv_h, OP_SSS_H, H2, H2, H2, asub32)
+RVVCALL(OPIVV2_RM, vasub_vv_w, OP_SSS_W, H4, H4, H4, asub32)
+RVVCALL(OPIVV2_RM, vasub_vv_d, OP_SSS_D, H8, H8, H8, asub64)
+GEN_VEXT_VV_RM(vasub_vv_b, 1, 1, clearb)
+GEN_VEXT_VV_RM(vasub_vv_h, 2, 2, clearh)
+GEN_VEXT_VV_RM(vasub_vv_w, 4, 4, clearl)
+GEN_VEXT_VV_RM(vasub_vv_d, 8, 8, clearq)
+
+RVVCALL(OPIVX2_RM, vasub_vx_b, OP_SSS_B, H1, H1, asub32)
+RVVCALL(OPIVX2_RM, vasub_vx_h, OP_SSS_H, H2, H2, asub32)
+RVVCALL(OPIVX2_RM, vasub_vx_w, OP_SSS_W, H4, H4, asub32)
+RVVCALL(OPIVX2_RM, vasub_vx_d, OP_SSS_D, H8, H8, asub64)
+GEN_VEXT_VX_RM(vasub_vx_b, 1, 1, clearb)
+GEN_VEXT_VX_RM(vasub_vx_h, 2, 2, clearh)
+GEN_VEXT_VX_RM(vasub_vx_w, 4, 4, clearl)
+GEN_VEXT_VX_RM(vasub_vx_d, 8, 8, clearq)