XTheadVector adc/sbc instructions diff from RVV1.0 in the following points:
1. Different mask reg layout.
2. Different tail elements process policy.
3. Different check policy.
4. When vm = 1, RVV1.0 vmadc and vmsbc perform the computation without
carry-in/borrow-in. While XTheadVector does not have this kind of situation.
Signed-off-by: Huang Tao <eric.huang@linux.alibaba.com>
---
target/riscv/helper.h | 33 ++++
.../riscv/insn_trans/trans_xtheadvector.c.inc | 139 +++++++++++++-
target/riscv/xtheadvector_helper.c | 173 ++++++++++++++++++
3 files changed, 335 insertions(+), 10 deletions(-)
diff --git a/target/riscv/helper.h b/target/riscv/helper.h
index 3906f17079..25fb8f81c7 100644
--- a/target/riscv/helper.h
+++ b/target/riscv/helper.h
@@ -1542,3 +1542,36 @@ DEF_HELPER_6(th_vwadd_wx_w, void, ptr, ptr, tl, ptr, env, i32)
DEF_HELPER_6(th_vwsub_wx_b, void, ptr, ptr, tl, ptr, env, i32)
DEF_HELPER_6(th_vwsub_wx_h, void, ptr, ptr, tl, ptr, env, i32)
DEF_HELPER_6(th_vwsub_wx_w, void, ptr, ptr, tl, ptr, env, i32)
+
+DEF_HELPER_6(th_vadc_vvm_b, void, ptr, ptr, ptr, ptr, env, i32)
+DEF_HELPER_6(th_vadc_vvm_h, void, ptr, ptr, ptr, ptr, env, i32)
+DEF_HELPER_6(th_vadc_vvm_w, void, ptr, ptr, ptr, ptr, env, i32)
+DEF_HELPER_6(th_vadc_vvm_d, void, ptr, ptr, ptr, ptr, env, i32)
+DEF_HELPER_6(th_vsbc_vvm_b, void, ptr, ptr, ptr, ptr, env, i32)
+DEF_HELPER_6(th_vsbc_vvm_h, void, ptr, ptr, ptr, ptr, env, i32)
+DEF_HELPER_6(th_vsbc_vvm_w, void, ptr, ptr, ptr, ptr, env, i32)
+DEF_HELPER_6(th_vsbc_vvm_d, void, ptr, ptr, ptr, ptr, env, i32)
+DEF_HELPER_6(th_vmadc_vvm_b, void, ptr, ptr, ptr, ptr, env, i32)
+DEF_HELPER_6(th_vmadc_vvm_h, void, ptr, ptr, ptr, ptr, env, i32)
+DEF_HELPER_6(th_vmadc_vvm_w, void, ptr, ptr, ptr, ptr, env, i32)
+DEF_HELPER_6(th_vmadc_vvm_d, void, ptr, ptr, ptr, ptr, env, i32)
+DEF_HELPER_6(th_vmsbc_vvm_b, void, ptr, ptr, ptr, ptr, env, i32)
+DEF_HELPER_6(th_vmsbc_vvm_h, void, ptr, ptr, ptr, ptr, env, i32)
+DEF_HELPER_6(th_vmsbc_vvm_w, void, ptr, ptr, ptr, ptr, env, i32)
+DEF_HELPER_6(th_vmsbc_vvm_d, void, ptr, ptr, ptr, ptr, env, i32)
+DEF_HELPER_6(th_vadc_vxm_b, void, ptr, ptr, tl, ptr, env, i32)
+DEF_HELPER_6(th_vadc_vxm_h, void, ptr, ptr, tl, ptr, env, i32)
+DEF_HELPER_6(th_vadc_vxm_w, void, ptr, ptr, tl, ptr, env, i32)
+DEF_HELPER_6(th_vadc_vxm_d, void, ptr, ptr, tl, ptr, env, i32)
+DEF_HELPER_6(th_vsbc_vxm_b, void, ptr, ptr, tl, ptr, env, i32)
+DEF_HELPER_6(th_vsbc_vxm_h, void, ptr, ptr, tl, ptr, env, i32)
+DEF_HELPER_6(th_vsbc_vxm_w, void, ptr, ptr, tl, ptr, env, i32)
+DEF_HELPER_6(th_vsbc_vxm_d, void, ptr, ptr, tl, ptr, env, i32)
+DEF_HELPER_6(th_vmadc_vxm_b, void, ptr, ptr, tl, ptr, env, i32)
+DEF_HELPER_6(th_vmadc_vxm_h, void, ptr, ptr, tl, ptr, env, i32)
+DEF_HELPER_6(th_vmadc_vxm_w, void, ptr, ptr, tl, ptr, env, i32)
+DEF_HELPER_6(th_vmadc_vxm_d, void, ptr, ptr, tl, ptr, env, i32)
+DEF_HELPER_6(th_vmsbc_vxm_b, void, ptr, ptr, tl, ptr, env, i32)
+DEF_HELPER_6(th_vmsbc_vxm_h, void, ptr, ptr, tl, ptr, env, i32)
+DEF_HELPER_6(th_vmsbc_vxm_w, void, ptr, ptr, tl, ptr, env, i32)
+DEF_HELPER_6(th_vmsbc_vxm_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 f6aea9deff..a9e20a6dcb 100644
--- a/target/riscv/insn_trans/trans_xtheadvector.c.inc
+++ b/target/riscv/insn_trans/trans_xtheadvector.c.inc
@@ -1185,22 +1185,141 @@ GEN_OPIWX_WIDEN_TRANS_TH(th_vwadd_wx)
GEN_OPIWX_WIDEN_TRANS_TH(th_vwsubu_wx)
GEN_OPIWX_WIDEN_TRANS_TH(th_vwsub_wx)
+/* Vector Integer Add-with-Carry / Subtract-with-Borrow Instructions */
+
+/*
+ * This function is almost the copy of opivv_trans, except:
+ * 1) XTheadVector using different data encoding, add MLEN,
+ * delete VTA and VMA.
+ */
+static bool opivv_trans_th(uint32_t vd, uint32_t vs1, uint32_t vs2, uint32_t vm,
+ gen_helper_gvec_4_ptr *fn, DisasContext *s)
+{
+ uint32_t data = 0;
+
+ data = FIELD_DP32(data, VDATA_TH, MLEN, s->mlen);
+ data = FIELD_DP32(data, VDATA_TH, VM, vm);
+ data = FIELD_DP32(data, VDATA_TH, LMUL, s->lmul);
+ tcg_gen_gvec_4_ptr(vreg_ofs(s, vd), vreg_ofs(s, 0),
+ vreg_ofs(s, vs1), vreg_ofs(s, vs2),
+ tcg_env, s->cfg_ptr->vlenb,
+ s->cfg_ptr->vlenb, data, fn);
+ finalize_rvv_inst(s);
+ return true;
+}
+
+/* OPIVV without GVEC IR */
+#define GEN_OPIVV_TRANS_TH(NAME, CHECK) \
+static bool trans_##NAME(DisasContext *s, arg_rmrr *a) \
+{ \
+ if (CHECK(s, a)) { \
+ static gen_helper_gvec_4_ptr * const fns[4] = { \
+ gen_helper_##NAME##_b, gen_helper_##NAME##_h, \
+ gen_helper_##NAME##_w, gen_helper_##NAME##_d, \
+ }; \
+ return opivv_trans_th(a->rd, a->rs1, a->rs2, a->vm, \
+ fns[s->sew], s); \
+ } \
+ return false; \
+}
+
+/*
+ * For vadc and vsbc, an illegal instruction exception is raised if the
+ * destination vector register is v0 and LMUL > 1.
+ */
+static bool opivv_vadc_check_th(DisasContext *s, arg_rmrr *a)
+{
+ return (require_xtheadvector(s) &&
+ vext_check_isa_ill(s) &&
+ th_check_reg(s, a->rd, false) &&
+ th_check_reg(s, a->rs2, false) &&
+ th_check_reg(s, a->rs1, false) &&
+ ((a->rd != 0) || (s->lmul == 0)));
+}
+
+GEN_OPIVV_TRANS_TH(th_vadc_vvm, opivv_vadc_check_th)
+GEN_OPIVV_TRANS_TH(th_vsbc_vvm, opivv_vadc_check_th)
+
+/*
+ * For vmadc and vmsbc, an illegal instruction exception is raised if the
+ * destination vector register overlaps a source vector register group.
+ */
+static bool opivv_vmadc_check_th(DisasContext *s, arg_rmrr *a)
+{
+ return (require_xtheadvector(s) &&
+ vext_check_isa_ill(s) &&
+ th_check_reg(s, a->rs2, false) &&
+ th_check_reg(s, a->rs1, false) &&
+ th_check_overlap_group(a->rd, 1, a->rs1, 1 << s->lmul) &&
+ th_check_overlap_group(a->rd, 1, a->rs2, 1 << s->lmul));
+}
+
+GEN_OPIVV_TRANS_TH(th_vmadc_vvm, opivv_vmadc_check_th)
+GEN_OPIVV_TRANS_TH(th_vmsbc_vvm, opivv_vmadc_check_th)
+
+static bool opivx_vadc_check_th(DisasContext *s, arg_rmrr *a)
+{
+ return (require_xtheadvector(s) &&
+ vext_check_isa_ill(s) &&
+ th_check_reg(s, a->rd, false) &&
+ th_check_reg(s, a->rs2, false) &&
+ ((a->rd != 0) || (s->lmul == 0)));
+}
+
+/* OPIVX without GVEC IR */
+#define GEN_OPIVX_TRANS_TH(NAME, CHECK) \
+static bool trans_##NAME(DisasContext *s, arg_rmrr *a) \
+{ \
+ if (CHECK(s, a)) { \
+ static gen_helper_opivx * const fns[4] = { \
+ gen_helper_##NAME##_b, gen_helper_##NAME##_h, \
+ gen_helper_##NAME##_w, gen_helper_##NAME##_d, \
+ }; \
+ \
+ return opivx_trans_th(a->rd, a->rs1, a->rs2, a->vm, \
+ fns[s->sew], s); \
+ } \
+ return false; \
+}
+
+GEN_OPIVX_TRANS_TH(th_vadc_vxm, opivx_vadc_check_th)
+GEN_OPIVX_TRANS_TH(th_vsbc_vxm, opivx_vadc_check_th)
+
+static bool opivx_vmadc_check_th(DisasContext *s, arg_rmrr *a)
+{
+ return (require_xtheadvector(s) &&
+ vext_check_isa_ill(s) &&
+ th_check_reg(s, a->rs2, false) &&
+ th_check_overlap_group(a->rd, 1, a->rs2, 1 << s->lmul));
+}
+
+GEN_OPIVX_TRANS_TH(th_vmadc_vxm, opivx_vmadc_check_th)
+GEN_OPIVX_TRANS_TH(th_vmsbc_vxm, opivx_vmadc_check_th)
+
+/* OPIVI without GVEC IR */
+#define GEN_OPIVI_TRANS_TH(NAME, ZX, OPIVX, CHECK) \
+static bool trans_##NAME(DisasContext *s, arg_rmrr *a) \
+{ \
+ if (CHECK(s, a)) { \
+ static gen_helper_opivx * const fns[4] = { \
+ gen_helper_##OPIVX##_b, gen_helper_##OPIVX##_h, \
+ gen_helper_##OPIVX##_w, gen_helper_##OPIVX##_d, \
+ }; \
+ return opivi_trans_th(a->rd, a->rs1, a->rs2, a->vm, \
+ fns[s->sew], s, ZX); \
+ } \
+ return false; \
+}
+
+GEN_OPIVI_TRANS_TH(th_vadc_vim, IMM_SX, th_vadc_vxm, opivx_vadc_check_th)
+GEN_OPIVI_TRANS_TH(th_vmadc_vim, IMM_SX, th_vmadc_vxm, opivx_vmadc_check_th)
+
#define TH_TRANS_STUB(NAME) \
static bool trans_##NAME(DisasContext *s, arg_##NAME *a) \
{ \
return require_xtheadvector(s); \
}
-TH_TRANS_STUB(th_vadc_vvm)
-TH_TRANS_STUB(th_vadc_vxm)
-TH_TRANS_STUB(th_vadc_vim)
-TH_TRANS_STUB(th_vmadc_vvm)
-TH_TRANS_STUB(th_vmadc_vxm)
-TH_TRANS_STUB(th_vmadc_vim)
-TH_TRANS_STUB(th_vsbc_vvm)
-TH_TRANS_STUB(th_vsbc_vxm)
-TH_TRANS_STUB(th_vmsbc_vvm)
-TH_TRANS_STUB(th_vmsbc_vxm)
TH_TRANS_STUB(th_vand_vv)
TH_TRANS_STUB(th_vand_vx)
TH_TRANS_STUB(th_vand_vi)
diff --git a/target/riscv/xtheadvector_helper.c b/target/riscv/xtheadvector_helper.c
index 5ebdb5a375..e5058d09f6 100644
--- a/target/riscv/xtheadvector_helper.c
+++ b/target/riscv/xtheadvector_helper.c
@@ -134,6 +134,15 @@ static inline int th_elem_mask(void *v0, int mlen, int index)
return (((uint64_t *)v0)[idx] >> pos) & 1;
}
+static inline void th_set_elem_mask(void *v0, int mlen, int index,
+ uint8_t value)
+{
+ int idx = (index * mlen) / 64;
+ int pos = (index * mlen) % 64;
+ uint64_t old = ((uint64_t *)v0)[idx];
+ ((uint64_t *)v0)[idx] = deposit64(old, pos, mlen, value);
+}
+
/* elements operations for load and store */
typedef void th_ldst_elem_fn(CPURISCVState *env, abi_ptr addr,
uint32_t idx, void *vd, uintptr_t retaddr);
@@ -1050,3 +1059,167 @@ GEN_TH_VX(th_vwadd_wx_w, 4, 8, clearq_th)
GEN_TH_VX(th_vwsub_wx_b, 1, 2, clearh_th)
GEN_TH_VX(th_vwsub_wx_h, 2, 4, clearl_th)
GEN_TH_VX(th_vwsub_wx_w, 4, 8, clearq_th)
+
+/* Vector Integer Add-with-Carry / Subtract-with-Borrow Instructions */
+
+#define TH_VADC(N, M, C) (N + M + C)
+#define TH_VSBC(N, M, C) (N - M - C)
+/*
+ * This function is almost the copy of GEN_VEXT_VADC_VVM, except:
+ * 1) different mask layout
+ * 2) different data encoding
+ * 3) different tail elements process policy
+ */
+#define GEN_TH_VADC_VVM(NAME, ETYPE, H, DO_OP, CLEAR_FN) \
+void HELPER(NAME)(void *vd, void *v0, void *vs1, void *vs2, \
+ CPURISCVState *env, uint32_t desc) \
+{ \
+ uint32_t mlen = th_mlen(desc); \
+ uint32_t vl = env->vl; \
+ uint32_t esz = sizeof(ETYPE); \
+ uint32_t vlmax = th_maxsz(desc) / esz; \
+ uint32_t i; \
+ \
+ VSTART_CHECK_EARLY_EXIT(env); \
+ for (i = env->vstart; i < vl; i++) { \
+ ETYPE s1 = *((ETYPE *)vs1 + H(i)); \
+ ETYPE s2 = *((ETYPE *)vs2 + H(i)); \
+ uint8_t carry = th_elem_mask(v0, mlen, i); \
+ \
+ *((ETYPE *)vd + H(i)) = DO_OP(s2, s1, carry); \
+ } \
+ env->vstart = 0; \
+ CLEAR_FN(vd, vl, vl * esz, vlmax * esz); \
+}
+
+GEN_TH_VADC_VVM(th_vadc_vvm_b, uint8_t, H1, TH_VADC, clearb_th)
+GEN_TH_VADC_VVM(th_vadc_vvm_h, uint16_t, H2, TH_VADC, clearh_th)
+GEN_TH_VADC_VVM(th_vadc_vvm_w, uint32_t, H4, TH_VADC, clearl_th)
+GEN_TH_VADC_VVM(th_vadc_vvm_d, uint64_t, H8, TH_VADC, clearq_th)
+
+GEN_TH_VADC_VVM(th_vsbc_vvm_b, uint8_t, H1, TH_VSBC, clearb_th)
+GEN_TH_VADC_VVM(th_vsbc_vvm_h, uint16_t, H2, TH_VSBC, clearh_th)
+GEN_TH_VADC_VVM(th_vsbc_vvm_w, uint32_t, H4, TH_VSBC, clearl_th)
+GEN_TH_VADC_VVM(th_vsbc_vvm_d, uint64_t, H8, TH_VSBC, clearq_th)
+/*
+ * This function is almost the copy of GEN_VEXT_VADC_VXM, except:
+ * 1) different mask layout
+ * 2) different data encoding
+ * 3) different tail elements process policy
+ */
+#define GEN_TH_VADC_VXM(NAME, ETYPE, H, DO_OP, CLEAR_FN) \
+void HELPER(NAME)(void *vd, void *v0, target_ulong s1, void *vs2, \
+ CPURISCVState *env, uint32_t desc) \
+{ \
+ uint32_t mlen = th_mlen(desc); \
+ uint32_t vl = env->vl; \
+ uint32_t esz = sizeof(ETYPE); \
+ uint32_t vlmax = th_maxsz(desc) / esz; \
+ uint32_t i; \
+ \
+ VSTART_CHECK_EARLY_EXIT(env); \
+ for (i = env->vstart; i < vl; i++) { \
+ ETYPE s2 = *((ETYPE *)vs2 + H(i)); \
+ uint8_t carry = th_elem_mask(v0, mlen, i); \
+ \
+ *((ETYPE *)vd + H(i)) = DO_OP(s2, (ETYPE)(target_long)s1, carry);\
+ } \
+ env->vstart = 0; \
+ CLEAR_FN(vd, vl, vl * esz, vlmax * esz); \
+}
+
+GEN_TH_VADC_VXM(th_vadc_vxm_b, uint8_t, H1, TH_VADC, clearb_th)
+GEN_TH_VADC_VXM(th_vadc_vxm_h, uint16_t, H2, TH_VADC, clearh_th)
+GEN_TH_VADC_VXM(th_vadc_vxm_w, uint32_t, H4, TH_VADC, clearl_th)
+GEN_TH_VADC_VXM(th_vadc_vxm_d, uint64_t, H8, TH_VADC, clearq_th)
+
+GEN_TH_VADC_VXM(th_vsbc_vxm_b, uint8_t, H1, TH_VSBC, clearb_th)
+GEN_TH_VADC_VXM(th_vsbc_vxm_h, uint16_t, H2, TH_VSBC, clearh_th)
+GEN_TH_VADC_VXM(th_vsbc_vxm_w, uint32_t, H4, TH_VSBC, clearl_th)
+GEN_TH_VADC_VXM(th_vsbc_vxm_d, uint64_t, H8, TH_VSBC, clearq_th)
+
+#define TH_MADC(N, M, C) (C ? (__typeof(N))(N + M + 1) <= N : \
+ (__typeof(N))(N + M) < N)
+#define TH_MSBC(N, M, C) (C ? N <= M : N < M)
+/*
+ * This function is almost the copy of GEN_VEXT_VMADC_VVM, except:
+ * 1) different mask layout
+ * 2) different data encoding
+ * 3) different tail elements process policy
+ * 4) When vm = 1, RVV1.0 vmadc and vmsbc perform the computation
+ * without carry-in/borrow-in. While XTheadVector does not have
+ * this kind of situation.
+ */
+#define GEN_TH_VMADC_VVM(NAME, ETYPE, H, DO_OP) \
+void HELPER(NAME)(void *vd, void *v0, void *vs1, void *vs2, \
+ CPURISCVState *env, uint32_t desc) \
+{ \
+ uint32_t mlen = th_mlen(desc); \
+ uint32_t vl = env->vl; \
+ uint32_t vlmax = th_maxsz(desc) / sizeof(ETYPE); \
+ uint32_t i; \
+ \
+ VSTART_CHECK_EARLY_EXIT(env); \
+ for (i = env->vstart; i < vl; i++) { \
+ ETYPE s1 = *((ETYPE *)vs1 + H(i)); \
+ ETYPE s2 = *((ETYPE *)vs2 + H(i)); \
+ uint8_t carry = th_elem_mask(v0, mlen, i); \
+ \
+ th_set_elem_mask(vd, mlen, i, DO_OP(s2, s1, carry)); \
+ } \
+ env->vstart = 0; \
+ for (; i < vlmax; i++) { \
+ th_set_elem_mask(vd, mlen, i, 0); \
+ } \
+}
+
+GEN_TH_VMADC_VVM(th_vmadc_vvm_b, uint8_t, H1, TH_MADC)
+GEN_TH_VMADC_VVM(th_vmadc_vvm_h, uint16_t, H2, TH_MADC)
+GEN_TH_VMADC_VVM(th_vmadc_vvm_w, uint32_t, H4, TH_MADC)
+GEN_TH_VMADC_VVM(th_vmadc_vvm_d, uint64_t, H8, TH_MADC)
+
+GEN_TH_VMADC_VVM(th_vmsbc_vvm_b, uint8_t, H1, TH_MSBC)
+GEN_TH_VMADC_VVM(th_vmsbc_vvm_h, uint16_t, H2, TH_MSBC)
+GEN_TH_VMADC_VVM(th_vmsbc_vvm_w, uint32_t, H4, TH_MSBC)
+GEN_TH_VMADC_VVM(th_vmsbc_vvm_d, uint64_t, H8, TH_MSBC)
+/*
+ * This function is almost the copy of GEN_VEXT_VMADC_VXM, except:
+ * 1) different mask layout
+ * 2) different data encoding
+ * 3) different tail elements process policy
+ * 4) When vm = 1, RVV1.0 vmadc and vmsbc perform the computation
+ * without carry-in/borrow-in. While XTheadVector does not have
+ * this kind of situation.
+ */
+#define GEN_TH_VMADC_VXM(NAME, ETYPE, H, DO_OP) \
+void HELPER(NAME)(void *vd, void *v0, target_ulong s1, \
+ void *vs2, CPURISCVState *env, uint32_t desc) \
+{ \
+ uint32_t mlen = th_mlen(desc); \
+ uint32_t vl = env->vl; \
+ uint32_t vlmax = th_maxsz(desc) / sizeof(ETYPE); \
+ uint32_t i; \
+ \
+ VSTART_CHECK_EARLY_EXIT(env); \
+ for (i = env->vstart; i < vl; i++) { \
+ ETYPE s2 = *((ETYPE *)vs2 + H(i)); \
+ uint8_t carry = th_elem_mask(v0, mlen, i); \
+ \
+ th_set_elem_mask(vd, mlen, i, \
+ DO_OP(s2, (ETYPE)(target_long)s1, carry)); \
+ } \
+ env->vstart = 0; \
+ for (; i < vlmax; i++) { \
+ th_set_elem_mask(vd, mlen, i, 0); \
+ } \
+}
+
+GEN_TH_VMADC_VXM(th_vmadc_vxm_b, uint8_t, H1, TH_MADC)
+GEN_TH_VMADC_VXM(th_vmadc_vxm_h, uint16_t, H2, TH_MADC)
+GEN_TH_VMADC_VXM(th_vmadc_vxm_w, uint32_t, H4, TH_MADC)
+GEN_TH_VMADC_VXM(th_vmadc_vxm_d, uint64_t, H8, TH_MADC)
+
+GEN_TH_VMADC_VXM(th_vmsbc_vxm_b, uint8_t, H1, TH_MSBC)
+GEN_TH_VMADC_VXM(th_vmsbc_vxm_h, uint16_t, H2, TH_MSBC)
+GEN_TH_VMADC_VXM(th_vmsbc_vxm_w, uint32_t, H4, TH_MSBC)
+GEN_TH_VMADC_VXM(th_vmsbc_vxm_d, uint64_t, H8, TH_MSBC)
--
2.44.0