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.hu...@linux.alibaba.com> --- target/riscv/helper.h | 9 ++ .../riscv/insn_trans/trans_xtheadvector.c.inc | 85 ++++++++++++++++++- target/riscv/xtheadvector_helper.c | 64 ++++++++++++++ 3 files changed, 155 insertions(+), 3 deletions(-) diff --git a/target/riscv/helper.h b/target/riscv/helper.h index 6ce0bcbba7..b650e299cf 100644 --- a/target/riscv/helper.h +++ b/target/riscv/helper.h @@ -2333,3 +2333,12 @@ DEF_HELPER_6(th_vslide1down_vx_b, void, ptr, ptr, tl, ptr, env, i32) DEF_HELPER_6(th_vslide1down_vx_h, void, ptr, ptr, tl, ptr, env, i32) DEF_HELPER_6(th_vslide1down_vx_w, void, ptr, ptr, tl, ptr, env, i32) DEF_HELPER_6(th_vslide1down_vx_d, void, ptr, ptr, tl, ptr, env, i32) + +DEF_HELPER_6(th_vrgather_vv_b, void, ptr, ptr, ptr, ptr, env, i32) +DEF_HELPER_6(th_vrgather_vv_h, void, ptr, ptr, ptr, ptr, env, i32) +DEF_HELPER_6(th_vrgather_vv_w, void, ptr, ptr, ptr, ptr, env, i32) +DEF_HELPER_6(th_vrgather_vv_d, void, ptr, ptr, ptr, ptr, env, i32) +DEF_HELPER_6(th_vrgather_vx_b, void, ptr, ptr, tl, ptr, env, i32) +DEF_HELPER_6(th_vrgather_vx_h, void, ptr, ptr, tl, ptr, env, i32) +DEF_HELPER_6(th_vrgather_vx_w, void, ptr, ptr, tl, ptr, env, i32) +DEF_HELPER_6(th_vrgather_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 46cfc51690..f6da1ff384 100644 --- a/target/riscv/insn_trans/trans_xtheadvector.c.inc +++ b/target/riscv/insn_trans/trans_xtheadvector.c.inc @@ -2816,13 +2816,92 @@ GEN_OPIVX_TRANS_TH(th_vslidedown_vx, opivx_check_th) GEN_OPIVX_TRANS_TH(th_vslide1down_vx, opivx_check_th) GEN_OPIVI_TRANS_TH(th_vslidedown_vi, IMM_ZX, th_vslidedown_vx, opivx_check_th) +/* Vector Register Gather Instruction */ +static bool vrgather_vv_check_th(DisasContext *s, arg_rmrr *a) +{ + return (require_xtheadvector(s) && + vext_check_isa_ill(s) && + th_check_overlap_mask(s, a->rd, a->vm, true) && + th_check_reg(s, a->rd, false) && + th_check_reg(s, a->rs1, false) && + th_check_reg(s, a->rs2, false) && + (a->rd != a->rs2) && (a->rd != a->rs1)); +} + +GEN_OPIVV_TRANS_TH(th_vrgather_vv, vrgather_vv_check_th) + +static bool vrgather_vx_check_th(DisasContext *s, arg_rmrr *a) +{ + return (require_xtheadvector(s) && + vext_check_isa_ill(s) && + th_check_overlap_mask(s, a->rd, a->vm, true) && + th_check_reg(s, a->rd, false) && + th_check_reg(s, a->rs2, false) && + (a->rd != a->rs2)); +} + +/* vrgather.vx vd, vs2, rs1, vm # vd[i] = (x[rs1] >= VLMAX) ? 0 : vs2[rs1] */ +static bool trans_th_vrgather_vx(DisasContext *s, arg_rmrr *a) +{ + if (!vrgather_vx_check_th(s, a)) { + return false; + } + + if (a->vm && s->vl_eq_vlmax) { + int vlmax = (s->cfg_ptr->vlenb << 3) / s->mlen; + TCGv_i64 dest = tcg_temp_new_i64(); + + if (a->rs1 == 0) { + th_element_loadi(s, dest, a->rs2, 0); + } else { + th_element_loadx(s, dest, a->rs2, cpu_gpr[a->rs1], vlmax); + } + + tcg_gen_gvec_dup_i64(s->sew, vreg_ofs(s, a->rd), + MAXSZ(s), MAXSZ(s), dest); + finalize_rvv_inst(s); + } else { + static gen_helper_opivx * const fns[4] = { + gen_helper_th_vrgather_vx_b, gen_helper_th_vrgather_vx_h, + gen_helper_th_vrgather_vx_w, gen_helper_th_vrgather_vx_d + }; + return opivx_trans_th(a->rd, a->rs1, a->rs2, a->vm, fns[s->sew], s); + } + return true; +} + +/* vrgather.vi vd, vs2, imm, vm # vd[i] = (imm >= VLMAX) ? 0 : vs2[imm] */ +static bool trans_th_vrgather_vi(DisasContext *s, arg_rmrr *a) +{ + if (!vrgather_vx_check_th(s, a)) { + return false; + } + + if (a->vm && s->vl_eq_vlmax) { + if (a->rs1 >= (s->cfg_ptr->vlenb << 3) / s->mlen) { + tcg_gen_gvec_dup_imm(MO_64, vreg_ofs(s, a->rd), + MAXSZ(s), MAXSZ(s), 0); + } else { + tcg_gen_gvec_dup_mem(s->sew, vreg_ofs(s, a->rd), + endian_ofs(s, a->rs2, a->rs1), + MAXSZ(s), MAXSZ(s)); + } + finalize_rvv_inst(s); + } else { + static gen_helper_opivx * const fns[4] = { + gen_helper_th_vrgather_vx_b, gen_helper_th_vrgather_vx_h, + gen_helper_th_vrgather_vx_w, gen_helper_th_vrgather_vx_d + }; + return opivi_trans_th(a->rd, a->rs1, a->rs2, a->vm, fns[s->sew], + s, IMM_ZX); + } + return true; +} + #define TH_TRANS_STUB(NAME) \ static bool trans_##NAME(DisasContext *s, arg_##NAME *a) \ { \ return require_xtheadvector(s); \ } -TH_TRANS_STUB(th_vrgather_vv) -TH_TRANS_STUB(th_vrgather_vx) -TH_TRANS_STUB(th_vrgather_vi) TH_TRANS_STUB(th_vcompress_vm) diff --git a/target/riscv/xtheadvector_helper.c b/target/riscv/xtheadvector_helper.c index 73a15eb070..2598824bb3 100644 --- a/target/riscv/xtheadvector_helper.c +++ b/target/riscv/xtheadvector_helper.c @@ -3801,3 +3801,67 @@ GEN_TH_VSLIDE1DOWN_VX(th_vslide1down_vx_b, uint8_t, H1, clearb_th) GEN_TH_VSLIDE1DOWN_VX(th_vslide1down_vx_h, uint16_t, H2, clearh_th) GEN_TH_VSLIDE1DOWN_VX(th_vslide1down_vx_w, uint32_t, H4, clearl_th) GEN_TH_VSLIDE1DOWN_VX(th_vslide1down_vx_d, uint64_t, H8, clearq_th) + +/* Vector Register Gather Instruction */ +#define GEN_TH_VRGATHER_VV(NAME, ETYPE, H, 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 vlmax = (env_archcpu(env)->cfg.vlenb << 3) / mlen; \ + uint32_t vm = th_vm(desc); \ + uint32_t vl = env->vl; \ + uint32_t index, i; \ + \ + VSTART_CHECK_EARLY_EXIT(env); \ + for (i = env->vstart; i < vl; i++) { \ + if (!vm && !th_elem_mask(v0, mlen, i)) { \ + continue; \ + } \ + index = *((ETYPE *)vs1 + H(i)); \ + if (index >= vlmax) { \ + *((ETYPE *)vd + H(i)) = 0; \ + } else { \ + *((ETYPE *)vd + H(i)) = *((ETYPE *)vs2 + H(index)); \ + } \ + } \ + env->vstart = 0; \ + CLEAR_FN(vd, vl, vl * sizeof(ETYPE), vlmax * sizeof(ETYPE)); \ +} + +/* vd[i] = (vs1[i] >= VLMAX) ? 0 : vs2[vs1[i]]; */ +GEN_TH_VRGATHER_VV(th_vrgather_vv_b, uint8_t, H1, clearb_th) +GEN_TH_VRGATHER_VV(th_vrgather_vv_h, uint16_t, H2, clearh_th) +GEN_TH_VRGATHER_VV(th_vrgather_vv_w, uint32_t, H4, clearl_th) +GEN_TH_VRGATHER_VV(th_vrgather_vv_d, uint64_t, H8, clearq_th) + +#define GEN_TH_VRGATHER_VX(NAME, ETYPE, H, 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 vlmax = (env_archcpu(env)->cfg.vlenb << 3) / mlen; \ + uint32_t vm = th_vm(desc); \ + uint32_t vl = env->vl; \ + uint32_t index = s1, i; \ + \ + VSTART_CHECK_EARLY_EXIT(env); \ + for (i = env->vstart; i < vl; i++) { \ + if (!vm && !th_elem_mask(v0, mlen, i)) { \ + continue; \ + } \ + if (index >= vlmax) { \ + *((ETYPE *)vd + H(i)) = 0; \ + } else { \ + *((ETYPE *)vd + H(i)) = *((ETYPE *)vs2 + H(index)); \ + } \ + } \ + env->vstart = 0; \ + CLEAR_FN(vd, vl, vl * sizeof(ETYPE), vlmax * sizeof(ETYPE)); \ +} + +/* vd[i] = (x[rs1] >= VLMAX) ? 0 : vs2[rs1] */ +GEN_TH_VRGATHER_VX(th_vrgather_vx_b, uint8_t, H1, clearb_th) +GEN_TH_VRGATHER_VX(th_vrgather_vx_h, uint16_t, H2, clearh_th) +GEN_TH_VRGATHER_VX(th_vrgather_vx_w, uint32_t, H4, clearl_th) +GEN_TH_VRGATHER_VX(th_vrgather_vx_d, uint64_t, H8, clearq_th) -- 2.44.0