Implement the MVE VADC and VSBC insns. These perform an add-with-carry or subtract-with-carry of the 32-bit elements in each lane of the input vectors, where the carry-out of each add is the carry-in of the next. The initial carry input is either 1 or is from FPSCR.C; the carry out at the end is written back to FPSCR.C.
Signed-off-by: Peter Maydell <peter.mayd...@linaro.org> --- target/arm/helper-mve.h | 3 ++ target/arm/mve.decode | 6 ++++ target/arm/mve_helper.c | 30 +++++++++++++++++ target/arm/translate-mve.c | 69 ++++++++++++++++++++++++++++++++++++++ 4 files changed, 108 insertions(+) diff --git a/target/arm/helper-mve.h b/target/arm/helper-mve.h index cd2cc6252f8..686e5d9a39b 100644 --- a/target/arm/helper-mve.h +++ b/target/arm/helper-mve.h @@ -248,6 +248,9 @@ DEF_HELPER_FLAGS_4(mve_vrhaddub, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) DEF_HELPER_FLAGS_4(mve_vrhadduh, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) DEF_HELPER_FLAGS_4(mve_vrhadduw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr) +DEF_HELPER_FLAGS_5(mve_vadc, TCG_CALL_NO_WG, i32, env, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(mve_vsbc, TCG_CALL_NO_WG, i32, env, ptr, ptr, ptr, i32) + DEF_HELPER_FLAGS_4(mve_vadd_scalarb, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) DEF_HELPER_FLAGS_4(mve_vadd_scalarh, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) DEF_HELPER_FLAGS_4(mve_vadd_scalarw, TCG_CALL_NO_WG, void, env, ptr, ptr, i32) diff --git a/target/arm/mve.decode b/target/arm/mve.decode index 6b969902df0..6a4aae7a1fc 100644 --- a/target/arm/mve.decode +++ b/target/arm/mve.decode @@ -30,6 +30,7 @@ &1op qd qm size &2op qd qm qn size &2scalar qd qn rm size +&vadc qd qm qn i @vldr_vstr ....... . . . . l:1 rn:4 ... ...... imm:7 &vldr_vstr qd=%qd u=0 # Note that both Rn and Qd are 3 bits only (no D bit) @@ -42,6 +43,8 @@ @2op_sz28 .... .... .... .... .... .... .... .... &2op qd=%qd qm=%qm qn=%qn \ size=%size_28 +@vadc .... .... .... .... ... i:1 .... .... .... &vadc qd=%qd qm=%qm qn=%qn + # The _rev suffix indicates that Vn and Vm are reversed. This is # the case for shifts. In the Arm ARM these insns are documented # with the Vm and Vn fields in their usual places, but in the @@ -160,6 +163,9 @@ VQDMULLT 111 . 1110 0 . 11 ... 0 ... 1 1111 . 0 . 0 ... 1 @2op_sz28 VRHADD_S 111 0 1111 0 . .. ... 0 ... 0 0001 . 1 . 0 ... 0 @2op VRHADD_U 111 1 1111 0 . .. ... 0 ... 0 0001 . 1 . 0 ... 0 @2op +VADC 1110 1110 0 . 11 ... 0 ... . 1111 . 0 . 0 ... 0 @vadc +VSBC 1111 1110 0 . 11 ... 0 ... . 1111 . 0 . 0 ... 0 @vadc + # Vector miscellaneous VCLS 1111 1111 1 . 11 .. 00 ... 0 0100 01 . 0 ... 0 @1op diff --git a/target/arm/mve_helper.c b/target/arm/mve_helper.c index c9434479604..e07f12c8389 100644 --- a/target/arm/mve_helper.c +++ b/target/arm/mve_helper.c @@ -580,6 +580,36 @@ DO_2OP_U(vrshlu, DO_VRSHLU) DO_2OP_S(vrhadds, DO_RHADD_S) DO_2OP_U(vrhaddu, DO_RHADD_U) +#define DO_VADC(OP, INV) \ + uint32_t HELPER(glue(mve_, OP))(CPUARMState *env, void *vd, \ + void *vn, void *vm, uint32_t nzcv) \ + { \ + uint32_t *d = vd, *n = vn, *m = vm; \ + uint16_t mask = mve_element_mask(env); \ + unsigned e; \ + int carry = (nzcv & FPCR_C) ? 1 : 0; \ + /* If we do no additions at all the flags are preserved */ \ + bool updates_flags = (mask & 0x1111) != 0; \ + for (e = 0; e < 16 / 4; e++, mask >>= 4) { \ + uint64_t r = (uint64_t)n[H4(e)] + INV(m[H4(e)]) + carry; \ + if (mask & 1) { \ + carry = r >> 32; \ + } \ + uint64_t bytemask = mask_to_bytemask4(mask); \ + d[H4(e)] &= ~bytemask; \ + d[H4(e)] |= (r & bytemask); \ + } \ + mve_advance_vpt(env); \ + if (updates_flags) { \ + nzcv = carry ? FPCR_C : 0; \ + } \ + return nzcv; \ + } + +/* VSBC differs only in inverting op2 before the additiona */ +DO_VADC(vadc, ) +DO_VADC(vsbc, DO_NOT) + static inline int32_t do_sat_bhw(int64_t val, int64_t min, int64_t max, bool *s) { if (val > max) { diff --git a/target/arm/translate-mve.c b/target/arm/translate-mve.c index 9a88583385f..2ed499a6de2 100644 --- a/target/arm/translate-mve.c +++ b/target/arm/translate-mve.c @@ -33,6 +33,7 @@ typedef void MVEGenOneOpFn(TCGv_ptr, TCGv_ptr, TCGv_ptr); typedef void MVEGenTwoOpFn(TCGv_ptr, TCGv_ptr, TCGv_ptr, TCGv_ptr); typedef void MVEGenTwoOpScalarFn(TCGv_ptr, TCGv_ptr, TCGv_ptr, TCGv_i32); typedef void MVEGenDualAccOpFn(TCGv_i64, TCGv_ptr, TCGv_ptr, TCGv_ptr, TCGv_i64); +typedef void MVEGenADCFn(TCGv_i32, TCGv_ptr, TCGv_ptr, TCGv_ptr, TCGv_ptr, TCGv_i32); /* Return the offset of a Qn register (same semantics as aa32_vfp_qreg()) */ static inline long mve_qreg_offset(unsigned reg) @@ -737,3 +738,71 @@ static bool trans_VPST(DisasContext *s, arg_VPST *a) } return true; } + +static bool do_vadc(DisasContext *s, arg_vadc *a, MVEGenADCFn fn, + uint32_t fixed_carry) +{ + /* + * VADC and VSBC: these perform an add-with-carry or subtract-with-carry + * of the 32-bit elements in each lane of the input vectors, where the + * carry-out of each add is the carry-in of the next. The initial carry + * input is either fixed (for the I variant: 0 for VADCI, 1 for VSBCI, + * passed in as fixed_carry) or is from FPSCR.C; the carry out at the + * end is written back to FPSCR.C. + */ + + TCGv_ptr qd, qn, qm; + TCGv_i32 nzcv, fpscr; + + if (!dc_isar_feature(aa32_mve, s)) { + return false; + } + if (a->qd > 7 || a->qn > 7 || a->qm > 7 || !fn) { + return false; + } + if (!mve_eci_check(s)) { + return true; + } + if (!vfp_access_check(s)) { + return true; + } + + /* + * This insn is subject to beat-wise execution. Partial execution + * of an I=1 (initial carry input fixed) insn which does not + * execute the first beat must start with the current FPSCR.NZCV + * value, not the fixed constant input. + */ + if (a->i && !mve_skip_first_beat(s)) { + /* Carry input is 0 (VADCI) or 1 (VSBCI), NZV zeroed */ + nzcv = tcg_const_i32(fixed_carry); + } else { + /* Carry input from existing NZCV flag values */ + nzcv = load_cpu_field(vfp.xregs[ARM_VFP_FPSCR]); + tcg_gen_andi_i32(nzcv, nzcv, FPCR_NZCV_MASK); + } + qd = mve_qreg_ptr(a->qd); + qn = mve_qreg_ptr(a->qn); + qm = mve_qreg_ptr(a->qm); + fn(nzcv, cpu_env, qd, qn, qm, nzcv); + fpscr = load_cpu_field(vfp.xregs[ARM_VFP_FPSCR]); + tcg_gen_andi_i32(fpscr, fpscr, ~FPCR_NZCV_MASK); + tcg_gen_or_i32(fpscr, fpscr, nzcv); + store_cpu_field(fpscr, vfp.xregs[ARM_VFP_FPSCR]); + tcg_temp_free_i32(nzcv); + tcg_temp_free_ptr(qd); + tcg_temp_free_ptr(qn); + tcg_temp_free_ptr(qm); + mve_update_eci(s); + return true; +} + +static bool trans_VADC(DisasContext *s, arg_vadc *a) +{ + return do_vadc(s, a, gen_helper_mve_vadc, 0); +} + +static bool trans_VSBC(DisasContext *s, arg_vadc *a) +{ + return do_vadc(s, a, gen_helper_mve_vsbc, FPCR_C); +} -- 2.20.1