diff mbox series

[27/65] target/riscv: Add single-width integer multiply-add instructions for XTheadVector

Message ID 20240412073735.76413-28-eric.huang@linux.alibaba.com
State New
Headers show
Series target/riscv: Support XTheadVector extension | expand

Commit Message

Huang Tao April 12, 2024, 7:36 a.m. UTC
The instructions have the same function as RVV1.0. Overall there are only
general differences between XTheadVector and RVV1.0.

Signed-off-by: Huang Tao <eric.huang@linux.alibaba.com>
---
 target/riscv/helper.h                         | 33 +++++++
 .../riscv/insn_trans/trans_xtheadvector.c.inc | 18 ++--
 target/riscv/xtheadvector_helper.c            | 87 +++++++++++++++++++
 3 files changed, 130 insertions(+), 8 deletions(-)
diff mbox series

Patch

diff --git a/target/riscv/helper.h b/target/riscv/helper.h
index 93e6a3f33d..a6abb48b55 100644
--- a/target/riscv/helper.h
+++ b/target/riscv/helper.h
@@ -1813,3 +1813,36 @@  DEF_HELPER_6(th_vwmulu_vx_w, void, ptr, ptr, tl, ptr, env, i32)
 DEF_HELPER_6(th_vwmulsu_vx_b, void, ptr, ptr, tl, ptr, env, i32)
 DEF_HELPER_6(th_vwmulsu_vx_h, void, ptr, ptr, tl, ptr, env, i32)
 DEF_HELPER_6(th_vwmulsu_vx_w, void, ptr, ptr, tl, ptr, env, i32)
+
+DEF_HELPER_6(th_vmacc_vv_b, void, ptr, ptr, ptr, ptr, env, i32)
+DEF_HELPER_6(th_vmacc_vv_h, void, ptr, ptr, ptr, ptr, env, i32)
+DEF_HELPER_6(th_vmacc_vv_w, void, ptr, ptr, ptr, ptr, env, i32)
+DEF_HELPER_6(th_vmacc_vv_d, void, ptr, ptr, ptr, ptr, env, i32)
+DEF_HELPER_6(th_vnmsac_vv_b, void, ptr, ptr, ptr, ptr, env, i32)
+DEF_HELPER_6(th_vnmsac_vv_h, void, ptr, ptr, ptr, ptr, env, i32)
+DEF_HELPER_6(th_vnmsac_vv_w, void, ptr, ptr, ptr, ptr, env, i32)
+DEF_HELPER_6(th_vnmsac_vv_d, void, ptr, ptr, ptr, ptr, env, i32)
+DEF_HELPER_6(th_vmadd_vv_b, void, ptr, ptr, ptr, ptr, env, i32)
+DEF_HELPER_6(th_vmadd_vv_h, void, ptr, ptr, ptr, ptr, env, i32)
+DEF_HELPER_6(th_vmadd_vv_w, void, ptr, ptr, ptr, ptr, env, i32)
+DEF_HELPER_6(th_vmadd_vv_d, void, ptr, ptr, ptr, ptr, env, i32)
+DEF_HELPER_6(th_vnmsub_vv_b, void, ptr, ptr, ptr, ptr, env, i32)
+DEF_HELPER_6(th_vnmsub_vv_h, void, ptr, ptr, ptr, ptr, env, i32)
+DEF_HELPER_6(th_vnmsub_vv_w, void, ptr, ptr, ptr, ptr, env, i32)
+DEF_HELPER_6(th_vnmsub_vv_d, void, ptr, ptr, ptr, ptr, env, i32)
+DEF_HELPER_6(th_vmacc_vx_b, void, ptr, ptr, tl, ptr, env, i32)
+DEF_HELPER_6(th_vmacc_vx_h, void, ptr, ptr, tl, ptr, env, i32)
+DEF_HELPER_6(th_vmacc_vx_w, void, ptr, ptr, tl, ptr, env, i32)
+DEF_HELPER_6(th_vmacc_vx_d, void, ptr, ptr, tl, ptr, env, i32)
+DEF_HELPER_6(th_vnmsac_vx_b, void, ptr, ptr, tl, ptr, env, i32)
+DEF_HELPER_6(th_vnmsac_vx_h, void, ptr, ptr, tl, ptr, env, i32)
+DEF_HELPER_6(th_vnmsac_vx_w, void, ptr, ptr, tl, ptr, env, i32)
+DEF_HELPER_6(th_vnmsac_vx_d, void, ptr, ptr, tl, ptr, env, i32)
+DEF_HELPER_6(th_vmadd_vx_b, void, ptr, ptr, tl, ptr, env, i32)
+DEF_HELPER_6(th_vmadd_vx_h, void, ptr, ptr, tl, ptr, env, i32)
+DEF_HELPER_6(th_vmadd_vx_w, void, ptr, ptr, tl, ptr, env, i32)
+DEF_HELPER_6(th_vmadd_vx_d, void, ptr, ptr, tl, ptr, env, i32)
+DEF_HELPER_6(th_vnmsub_vx_b, void, ptr, ptr, tl, ptr, env, i32)
+DEF_HELPER_6(th_vnmsub_vx_h, void, ptr, ptr, tl, ptr, env, i32)
+DEF_HELPER_6(th_vnmsub_vx_w, void, ptr, ptr, tl, ptr, env, i32)
+DEF_HELPER_6(th_vnmsub_vx_d, void, ptr, ptr, tl, ptr, env, i32)
diff --git a/target/riscv/insn_trans/trans_xtheadvector.c.inc b/target/riscv/insn_trans/trans_xtheadvector.c.inc
index 681e967078..d84edd90ca 100644
--- a/target/riscv/insn_trans/trans_xtheadvector.c.inc
+++ b/target/riscv/insn_trans/trans_xtheadvector.c.inc
@@ -1557,20 +1557,22 @@  GEN_OPIVX_WIDEN_TRANS_TH(th_vwmul_vx, opivx_widen_check_th)
 GEN_OPIVX_WIDEN_TRANS_TH(th_vwmulu_vx, opivx_widen_check_th)
 GEN_OPIVX_WIDEN_TRANS_TH(th_vwmulsu_vx, opivx_widen_check_th)
 
+/* Vector Single-Width Integer Multiply-Add Instructions */
+GEN_OPIVV_TRANS_TH(th_vmacc_vv, opivv_check_th)
+GEN_OPIVV_TRANS_TH(th_vnmsac_vv, opivv_check_th)
+GEN_OPIVV_TRANS_TH(th_vmadd_vv, opivv_check_th)
+GEN_OPIVV_TRANS_TH(th_vnmsub_vv, opivv_check_th)
+GEN_OPIVX_TRANS_TH(th_vmacc_vx, opivx_check_th)
+GEN_OPIVX_TRANS_TH(th_vnmsac_vx, opivx_check_th)
+GEN_OPIVX_TRANS_TH(th_vmadd_vx, opivx_check_th)
+GEN_OPIVX_TRANS_TH(th_vnmsub_vx, opivx_check_th)
+
 #define TH_TRANS_STUB(NAME)                                \
 static bool trans_##NAME(DisasContext *s, arg_##NAME *a)   \
 {                                                          \
     return require_xtheadvector(s);                        \
 }
 
-TH_TRANS_STUB(th_vmacc_vv)
-TH_TRANS_STUB(th_vmacc_vx)
-TH_TRANS_STUB(th_vnmsac_vv)
-TH_TRANS_STUB(th_vnmsac_vx)
-TH_TRANS_STUB(th_vmadd_vv)
-TH_TRANS_STUB(th_vmadd_vx)
-TH_TRANS_STUB(th_vnmsub_vv)
-TH_TRANS_STUB(th_vnmsub_vx)
 TH_TRANS_STUB(th_vwmaccu_vv)
 TH_TRANS_STUB(th_vwmaccu_vx)
 TH_TRANS_STUB(th_vwmacc_vv)
diff --git a/target/riscv/xtheadvector_helper.c b/target/riscv/xtheadvector_helper.c
index b5b1e55452..ccf6eb8a43 100644
--- a/target/riscv/xtheadvector_helper.c
+++ b/target/riscv/xtheadvector_helper.c
@@ -1791,3 +1791,90 @@  GEN_TH_VX(th_vwmulu_vx_w, 4, 8, clearq_th)
 GEN_TH_VX(th_vwmulsu_vx_b, 1, 2, clearh_th)
 GEN_TH_VX(th_vwmulsu_vx_h, 2, 4, clearl_th)
 GEN_TH_VX(th_vwmulsu_vx_w, 4, 8, clearq_th)
+
+/* Vector Single-Width Integer Multiply-Add Instructions */
+#define TH_OPIVV3(NAME, TD, T1, T2, TX1, TX2, HD, HS1, HS2, OP)   \
+static void do_##NAME(void *vd, void *vs1, void *vs2, int i)       \
+{                                                                  \
+    TX1 s1 = *((T1 *)vs1 + HS1(i));                                \
+    TX2 s2 = *((T2 *)vs2 + HS2(i));                                \
+    TD d = *((TD *)vd + HD(i));                                    \
+    *((TD *)vd + HD(i)) = OP(s2, s1, d);                           \
+}
+#define TH_MACC(N, M, D) (M * N + D)
+#define TH_NMSAC(N, M, D) (-(M * N) + D)
+#define TH_MADD(N, M, D) (M * D + N)
+#define TH_NMSUB(N, M, D) (-(M * D) + N)
+THCALL(TH_OPIVV3, th_vmacc_vv_b, OP_SSS_B, H1, H1, H1, TH_MACC)
+THCALL(TH_OPIVV3, th_vmacc_vv_h, OP_SSS_H, H2, H2, H2, TH_MACC)
+THCALL(TH_OPIVV3, th_vmacc_vv_w, OP_SSS_W, H4, H4, H4, TH_MACC)
+THCALL(TH_OPIVV3, th_vmacc_vv_d, OP_SSS_D, H8, H8, H8, TH_MACC)
+THCALL(TH_OPIVV3, th_vnmsac_vv_b, OP_SSS_B, H1, H1, H1, TH_NMSAC)
+THCALL(TH_OPIVV3, th_vnmsac_vv_h, OP_SSS_H, H2, H2, H2, TH_NMSAC)
+THCALL(TH_OPIVV3, th_vnmsac_vv_w, OP_SSS_W, H4, H4, H4, TH_NMSAC)
+THCALL(TH_OPIVV3, th_vnmsac_vv_d, OP_SSS_D, H8, H8, H8, TH_NMSAC)
+THCALL(TH_OPIVV3, th_vmadd_vv_b, OP_SSS_B, H1, H1, H1, TH_MADD)
+THCALL(TH_OPIVV3, th_vmadd_vv_h, OP_SSS_H, H2, H2, H2, TH_MADD)
+THCALL(TH_OPIVV3, th_vmadd_vv_w, OP_SSS_W, H4, H4, H4, TH_MADD)
+THCALL(TH_OPIVV3, th_vmadd_vv_d, OP_SSS_D, H8, H8, H8, TH_MADD)
+THCALL(TH_OPIVV3, th_vnmsub_vv_b, OP_SSS_B, H1, H1, H1, TH_NMSUB)
+THCALL(TH_OPIVV3, th_vnmsub_vv_h, OP_SSS_H, H2, H2, H2, TH_NMSUB)
+THCALL(TH_OPIVV3, th_vnmsub_vv_w, OP_SSS_W, H4, H4, H4, TH_NMSUB)
+THCALL(TH_OPIVV3, th_vnmsub_vv_d, OP_SSS_D, H8, H8, H8, TH_NMSUB)
+GEN_TH_VV(th_vmacc_vv_b, 1, 1, clearb_th)
+GEN_TH_VV(th_vmacc_vv_h, 2, 2, clearh_th)
+GEN_TH_VV(th_vmacc_vv_w, 4, 4, clearl_th)
+GEN_TH_VV(th_vmacc_vv_d, 8, 8, clearq_th)
+GEN_TH_VV(th_vnmsac_vv_b, 1, 1, clearb_th)
+GEN_TH_VV(th_vnmsac_vv_h, 2, 2, clearh_th)
+GEN_TH_VV(th_vnmsac_vv_w, 4, 4, clearl_th)
+GEN_TH_VV(th_vnmsac_vv_d, 8, 8, clearq_th)
+GEN_TH_VV(th_vmadd_vv_b, 1, 1, clearb_th)
+GEN_TH_VV(th_vmadd_vv_h, 2, 2, clearh_th)
+GEN_TH_VV(th_vmadd_vv_w, 4, 4, clearl_th)
+GEN_TH_VV(th_vmadd_vv_d, 8, 8, clearq_th)
+GEN_TH_VV(th_vnmsub_vv_b, 1, 1, clearb_th)
+GEN_TH_VV(th_vnmsub_vv_h, 2, 2, clearh_th)
+GEN_TH_VV(th_vnmsub_vv_w, 4, 4, clearl_th)
+GEN_TH_VV(th_vnmsub_vv_d, 8, 8, clearq_th)
+
+#define TH_OPIVX3(NAME, TD, T1, T2, TX1, TX2, HD, HS2, OP)             \
+static void do_##NAME(void *vd, target_long s1, void *vs2, int i)   \
+{                                                                   \
+    TX2 s2 = *((T2 *)vs2 + HS2(i));                                 \
+    TD d = *((TD *)vd + HD(i));                                     \
+    *((TD *)vd + HD(i)) = OP(s2, (TX1)(T1)s1, d);                   \
+}
+
+THCALL(TH_OPIVX3, th_vmacc_vx_b, OP_SSS_B, H1, H1, TH_MACC)
+THCALL(TH_OPIVX3, th_vmacc_vx_h, OP_SSS_H, H2, H2, TH_MACC)
+THCALL(TH_OPIVX3, th_vmacc_vx_w, OP_SSS_W, H4, H4, TH_MACC)
+THCALL(TH_OPIVX3, th_vmacc_vx_d, OP_SSS_D, H8, H8, TH_MACC)
+THCALL(TH_OPIVX3, th_vnmsac_vx_b, OP_SSS_B, H1, H1, TH_NMSAC)
+THCALL(TH_OPIVX3, th_vnmsac_vx_h, OP_SSS_H, H2, H2, TH_NMSAC)
+THCALL(TH_OPIVX3, th_vnmsac_vx_w, OP_SSS_W, H4, H4, TH_NMSAC)
+THCALL(TH_OPIVX3, th_vnmsac_vx_d, OP_SSS_D, H8, H8, TH_NMSAC)
+THCALL(TH_OPIVX3, th_vmadd_vx_b, OP_SSS_B, H1, H1, TH_MADD)
+THCALL(TH_OPIVX3, th_vmadd_vx_h, OP_SSS_H, H2, H2, TH_MADD)
+THCALL(TH_OPIVX3, th_vmadd_vx_w, OP_SSS_W, H4, H4, TH_MADD)
+THCALL(TH_OPIVX3, th_vmadd_vx_d, OP_SSS_D, H8, H8, TH_MADD)
+THCALL(TH_OPIVX3, th_vnmsub_vx_b, OP_SSS_B, H1, H1, TH_NMSUB)
+THCALL(TH_OPIVX3, th_vnmsub_vx_h, OP_SSS_H, H2, H2, TH_NMSUB)
+THCALL(TH_OPIVX3, th_vnmsub_vx_w, OP_SSS_W, H4, H4, TH_NMSUB)
+THCALL(TH_OPIVX3, th_vnmsub_vx_d, OP_SSS_D, H8, H8, TH_NMSUB)
+GEN_TH_VX(th_vmacc_vx_b, 1, 1, clearb_th)
+GEN_TH_VX(th_vmacc_vx_h, 2, 2, clearh_th)
+GEN_TH_VX(th_vmacc_vx_w, 4, 4, clearl_th)
+GEN_TH_VX(th_vmacc_vx_d, 8, 8, clearq_th)
+GEN_TH_VX(th_vnmsac_vx_b, 1, 1, clearb_th)
+GEN_TH_VX(th_vnmsac_vx_h, 2, 2, clearh_th)
+GEN_TH_VX(th_vnmsac_vx_w, 4, 4, clearl_th)
+GEN_TH_VX(th_vnmsac_vx_d, 8, 8, clearq_th)
+GEN_TH_VX(th_vmadd_vx_b, 1, 1, clearb_th)
+GEN_TH_VX(th_vmadd_vx_h, 2, 2, clearh_th)
+GEN_TH_VX(th_vmadd_vx_w, 4, 4, clearl_th)
+GEN_TH_VX(th_vmadd_vx_d, 8, 8, clearq_th)
+GEN_TH_VX(th_vnmsub_vx_b, 1, 1, clearb_th)
+GEN_TH_VX(th_vnmsub_vx_h, 2, 2, clearh_th)
+GEN_TH_VX(th_vnmsub_vx_w, 4, 4, clearl_th)
+GEN_TH_VX(th_vnmsub_vx_d, 8, 8, clearq_th)