diff options
-rw-r--r-- | include/tcg/tcg-op-gvec.h | 13 | ||||
-rw-r--r-- | include/tcg/tcg-op.h | 8 | ||||
-rw-r--r-- | target/arm/translate-a64.c | 10 | ||||
-rw-r--r-- | target/arm/translate-sve.c | 12 | ||||
-rw-r--r-- | target/arm/translate.c | 9 | ||||
-rw-r--r-- | target/ppc/translate/vmx-impl.inc.c | 32 | ||||
-rw-r--r-- | target/ppc/translate/vsx-impl.inc.c | 2 | ||||
-rw-r--r-- | target/s390x/translate_vx.inc.c | 41 | ||||
-rw-r--r-- | tcg/tcg-op-gvec.c | 164 | ||||
-rw-r--r-- | tcg/tcg-op.c | 16 |
10 files changed, 167 insertions, 140 deletions
diff --git a/include/tcg/tcg-op-gvec.h b/include/tcg/tcg-op-gvec.h index 74534e2480..cea6497341 100644 --- a/include/tcg/tcg-op-gvec.h +++ b/include/tcg/tcg-op-gvec.h @@ -109,6 +109,8 @@ typedef struct { uint8_t vece; /* Prefer i64 to v64. */ bool prefer_i64; + /* Load dest as a 2nd source operand. */ + bool load_dest; } GVecGen2; typedef struct { @@ -313,15 +315,18 @@ void tcg_gen_gvec_ors(unsigned vece, uint32_t dofs, uint32_t aofs, void tcg_gen_gvec_dup_mem(unsigned vece, uint32_t dofs, uint32_t aofs, uint32_t s, uint32_t m); +void tcg_gen_gvec_dup_imm(unsigned vece, uint32_t dofs, uint32_t s, + uint32_t m, uint64_t imm); void tcg_gen_gvec_dup_i32(unsigned vece, uint32_t dofs, uint32_t s, uint32_t m, TCGv_i32); void tcg_gen_gvec_dup_i64(unsigned vece, uint32_t dofs, uint32_t s, uint32_t m, TCGv_i64); -void tcg_gen_gvec_dup8i(uint32_t dofs, uint32_t s, uint32_t m, uint8_t x); -void tcg_gen_gvec_dup16i(uint32_t dofs, uint32_t s, uint32_t m, uint16_t x); -void tcg_gen_gvec_dup32i(uint32_t dofs, uint32_t s, uint32_t m, uint32_t x); -void tcg_gen_gvec_dup64i(uint32_t dofs, uint32_t s, uint32_t m, uint64_t x); +#if TARGET_LONG_BITS == 64 +# define tcg_gen_gvec_dup_tl tcg_gen_gvec_dup_i64 +#else +# define tcg_gen_gvec_dup_tl tcg_gen_gvec_dup_i32 +#endif void tcg_gen_gvec_shli(unsigned vece, uint32_t dofs, uint32_t aofs, int64_t shift, uint32_t oprsz, uint32_t maxsz); diff --git a/include/tcg/tcg-op.h b/include/tcg/tcg-op.h index 230db6e022..e3399d6a5e 100644 --- a/include/tcg/tcg-op.h +++ b/include/tcg/tcg-op.h @@ -297,9 +297,9 @@ void tcg_gen_ctzi_i32(TCGv_i32 ret, TCGv_i32 arg1, uint32_t arg2); void tcg_gen_clrsb_i32(TCGv_i32 ret, TCGv_i32 arg); void tcg_gen_ctpop_i32(TCGv_i32 a1, TCGv_i32 a2); void tcg_gen_rotl_i32(TCGv_i32 ret, TCGv_i32 arg1, TCGv_i32 arg2); -void tcg_gen_rotli_i32(TCGv_i32 ret, TCGv_i32 arg1, unsigned arg2); +void tcg_gen_rotli_i32(TCGv_i32 ret, TCGv_i32 arg1, int32_t arg2); void tcg_gen_rotr_i32(TCGv_i32 ret, TCGv_i32 arg1, TCGv_i32 arg2); -void tcg_gen_rotri_i32(TCGv_i32 ret, TCGv_i32 arg1, unsigned arg2); +void tcg_gen_rotri_i32(TCGv_i32 ret, TCGv_i32 arg1, int32_t arg2); void tcg_gen_deposit_i32(TCGv_i32 ret, TCGv_i32 arg1, TCGv_i32 arg2, unsigned int ofs, unsigned int len); void tcg_gen_deposit_z_i32(TCGv_i32 ret, TCGv_i32 arg, @@ -493,9 +493,9 @@ void tcg_gen_ctzi_i64(TCGv_i64 ret, TCGv_i64 arg1, uint64_t arg2); void tcg_gen_clrsb_i64(TCGv_i64 ret, TCGv_i64 arg); void tcg_gen_ctpop_i64(TCGv_i64 a1, TCGv_i64 a2); void tcg_gen_rotl_i64(TCGv_i64 ret, TCGv_i64 arg1, TCGv_i64 arg2); -void tcg_gen_rotli_i64(TCGv_i64 ret, TCGv_i64 arg1, unsigned arg2); +void tcg_gen_rotli_i64(TCGv_i64 ret, TCGv_i64 arg1, int64_t arg2); void tcg_gen_rotr_i64(TCGv_i64 ret, TCGv_i64 arg1, TCGv_i64 arg2); -void tcg_gen_rotri_i64(TCGv_i64 ret, TCGv_i64 arg1, unsigned arg2); +void tcg_gen_rotri_i64(TCGv_i64 ret, TCGv_i64 arg1, int64_t arg2); void tcg_gen_deposit_i64(TCGv_i64 ret, TCGv_i64 arg1, TCGv_i64 arg2, unsigned int ofs, unsigned int len); void tcg_gen_deposit_z_i64(TCGv_i64 ret, TCGv_i64 arg, diff --git a/target/arm/translate-a64.c b/target/arm/translate-a64.c index a896f9c4b8..62e5729904 100644 --- a/target/arm/translate-a64.c +++ b/target/arm/translate-a64.c @@ -502,7 +502,7 @@ static void clear_vec_high(DisasContext *s, bool is_q, int rd) tcg_temp_free_i64(tcg_zero); } if (vsz > 16) { - tcg_gen_gvec_dup8i(ofs + 16, vsz - 16, vsz - 16, 0); + tcg_gen_gvec_dup_imm(MO_64, ofs + 16, vsz - 16, vsz - 16, 0); } } @@ -7785,8 +7785,8 @@ static void disas_simd_mod_imm(DisasContext *s, uint32_t insn) if (!((cmode & 0x9) == 0x1 || (cmode & 0xd) == 0x9)) { /* MOVI or MVNI, with MVNI negation handled above. */ - tcg_gen_gvec_dup64i(vec_full_reg_offset(s, rd), is_q ? 16 : 8, - vec_full_reg_size(s), imm); + tcg_gen_gvec_dup_imm(MO_64, vec_full_reg_offset(s, rd), is_q ? 16 : 8, + vec_full_reg_size(s), imm); } else { /* ORR or BIC, with BIC negation to AND handled above. */ if (is_neg) { @@ -10214,8 +10214,8 @@ static void handle_vec_simd_shri(DisasContext *s, bool is_q, bool is_u, if (is_u) { if (shift == 8 << size) { /* Shift count the same size as element size produces zero. */ - tcg_gen_gvec_dup8i(vec_full_reg_offset(s, rd), - is_q ? 16 : 8, vec_full_reg_size(s), 0); + tcg_gen_gvec_dup_imm(size, vec_full_reg_offset(s, rd), + is_q ? 16 : 8, vec_full_reg_size(s), 0); } else { gen_gvec_fn2i(s, is_q, rd, rn, shift, tcg_gen_gvec_shri, size); } diff --git a/target/arm/translate-sve.c b/target/arm/translate-sve.c index b35bad245e..6c8bda4e4c 100644 --- a/target/arm/translate-sve.c +++ b/target/arm/translate-sve.c @@ -177,7 +177,7 @@ static bool do_mov_z(DisasContext *s, int rd, int rn) static void do_dupi_z(DisasContext *s, int rd, uint64_t word) { unsigned vsz = vec_full_reg_size(s); - tcg_gen_gvec_dup64i(vec_full_reg_offset(s, rd), vsz, vsz, word); + tcg_gen_gvec_dup_imm(MO_64, vec_full_reg_offset(s, rd), vsz, vsz, word); } /* Invoke a vector expander on two Pregs. */ @@ -1453,7 +1453,7 @@ static bool do_predset(DisasContext *s, int esz, int rd, int pat, bool setflag) unsigned oprsz = size_for_gvec(setsz / 8); if (oprsz * 8 == setsz) { - tcg_gen_gvec_dup64i(ofs, oprsz, maxsz, word); + tcg_gen_gvec_dup_imm(MO_64, ofs, oprsz, maxsz, word); goto done; } } @@ -2044,7 +2044,7 @@ static bool trans_DUP_x(DisasContext *s, arg_DUP_x *a) unsigned nofs = vec_reg_offset(s, a->rn, index, esz); tcg_gen_gvec_dup_mem(esz, dofs, nofs, vsz, vsz); } else { - tcg_gen_gvec_dup64i(dofs, vsz, vsz, 0); + tcg_gen_gvec_dup_imm(esz, dofs, vsz, vsz, 0); } } return true; @@ -3260,9 +3260,7 @@ static bool trans_FDUP(DisasContext *s, arg_FDUP *a) /* Decode the VFP immediate. */ imm = vfp_expand_imm(a->esz, a->imm); - imm = dup_const(a->esz, imm); - - tcg_gen_gvec_dup64i(dofs, vsz, vsz, imm); + tcg_gen_gvec_dup_imm(a->esz, dofs, vsz, vsz, imm); } return true; } @@ -3276,7 +3274,7 @@ static bool trans_DUP_i(DisasContext *s, arg_DUP_i *a) unsigned vsz = vec_full_reg_size(s); int dofs = vec_full_reg_offset(s, a->rd); - tcg_gen_gvec_dup64i(dofs, vsz, vsz, dup_const(a->esz, a->imm)); + tcg_gen_gvec_dup_imm(a->esz, dofs, vsz, vsz, a->imm); } return true; } diff --git a/target/arm/translate.c b/target/arm/translate.c index 025747c0bd..74fac1d09c 100644 --- a/target/arm/translate.c +++ b/target/arm/translate.c @@ -5209,7 +5209,8 @@ static int disas_neon_data_insn(DisasContext *s, uint32_t insn) MIN(shift, (8 << size) - 1), vec_size, vec_size); } else if (shift >= 8 << size) { - tcg_gen_gvec_dup8i(rd_ofs, vec_size, vec_size, 0); + tcg_gen_gvec_dup_imm(MO_8, rd_ofs, vec_size, + vec_size, 0); } else { tcg_gen_gvec_shri(size, rd_ofs, rm_ofs, shift, vec_size, vec_size); @@ -5260,7 +5261,8 @@ static int disas_neon_data_insn(DisasContext *s, uint32_t insn) * architecturally valid and results in zero. */ if (shift >= 8 << size) { - tcg_gen_gvec_dup8i(rd_ofs, vec_size, vec_size, 0); + tcg_gen_gvec_dup_imm(size, rd_ofs, + vec_size, vec_size, 0); } else { tcg_gen_gvec_shli(size, rd_ofs, rm_ofs, shift, vec_size, vec_size); @@ -5606,7 +5608,8 @@ static int disas_neon_data_insn(DisasContext *s, uint32_t insn) } tcg_temp_free_i64(t64); } else { - tcg_gen_gvec_dup32i(reg_ofs, vec_size, vec_size, imm); + tcg_gen_gvec_dup_imm(MO_32, reg_ofs, vec_size, + vec_size, imm); } } } diff --git a/target/ppc/translate/vmx-impl.inc.c b/target/ppc/translate/vmx-impl.inc.c index 81d5a7a341..403ed3a01c 100644 --- a/target/ppc/translate/vmx-impl.inc.c +++ b/target/ppc/translate/vmx-impl.inc.c @@ -1035,21 +1035,25 @@ GEN_VXRFORM_DUAL(vcmpbfp, PPC_ALTIVEC, PPC_NONE, \ GEN_VXRFORM_DUAL(vcmpgtfp, PPC_ALTIVEC, PPC_NONE, \ vcmpgtud, PPC_NONE, PPC2_ALTIVEC_207) -#define GEN_VXFORM_DUPI(name, tcg_op, opc2, opc3) \ -static void glue(gen_, name)(DisasContext *ctx) \ - { \ - int simm; \ - if (unlikely(!ctx->altivec_enabled)) { \ - gen_exception(ctx, POWERPC_EXCP_VPU); \ - return; \ - } \ - simm = SIMM5(ctx->opcode); \ - tcg_op(avr_full_offset(rD(ctx->opcode)), 16, 16, simm); \ +static void gen_vsplti(DisasContext *ctx, int vece) +{ + int simm; + + if (unlikely(!ctx->altivec_enabled)) { + gen_exception(ctx, POWERPC_EXCP_VPU); + return; } -GEN_VXFORM_DUPI(vspltisb, tcg_gen_gvec_dup8i, 6, 12); -GEN_VXFORM_DUPI(vspltish, tcg_gen_gvec_dup16i, 6, 13); -GEN_VXFORM_DUPI(vspltisw, tcg_gen_gvec_dup32i, 6, 14); + simm = SIMM5(ctx->opcode); + tcg_gen_gvec_dup_imm(vece, avr_full_offset(rD(ctx->opcode)), 16, 16, simm); +} + +#define GEN_VXFORM_VSPLTI(name, vece, opc2, opc3) \ +static void glue(gen_, name)(DisasContext *ctx) { gen_vsplti(ctx, vece); } + +GEN_VXFORM_VSPLTI(vspltisb, MO_8, 6, 12); +GEN_VXFORM_VSPLTI(vspltish, MO_16, 6, 13); +GEN_VXFORM_VSPLTI(vspltisw, MO_32, 6, 14); #define GEN_VXFORM_NOA(name, opc2, opc3) \ static void glue(gen_, name)(DisasContext *ctx) \ @@ -1559,7 +1563,7 @@ GEN_VXFORM_DUAL(vsldoi, PPC_ALTIVEC, PPC_NONE, #undef GEN_VXRFORM_DUAL #undef GEN_VXRFORM1 #undef GEN_VXRFORM -#undef GEN_VXFORM_DUPI +#undef GEN_VXFORM_VSPLTI #undef GEN_VXFORM_NOA #undef GEN_VXFORM_UIMM #undef GEN_VAFORM_PAIRED diff --git a/target/ppc/translate/vsx-impl.inc.c b/target/ppc/translate/vsx-impl.inc.c index 8287e272f5..b518de46db 100644 --- a/target/ppc/translate/vsx-impl.inc.c +++ b/target/ppc/translate/vsx-impl.inc.c @@ -1579,7 +1579,7 @@ static void gen_xxspltib(DisasContext *ctx) return; } } - tcg_gen_gvec_dup8i(vsr_full_offset(rt), 16, 16, uim8); + tcg_gen_gvec_dup_imm(MO_8, vsr_full_offset(rt), 16, 16, uim8); } static void gen_xxsldwi(DisasContext *ctx) diff --git a/target/s390x/translate_vx.inc.c b/target/s390x/translate_vx.inc.c index 24558cce80..12347f8a03 100644 --- a/target/s390x/translate_vx.inc.c +++ b/target/s390x/translate_vx.inc.c @@ -231,8 +231,8 @@ static void get_vec_element_ptr_i64(TCGv_ptr ptr, uint8_t reg, TCGv_i64 enr, #define gen_gvec_mov(v1, v2) \ tcg_gen_gvec_mov(0, vec_full_reg_offset(v1), vec_full_reg_offset(v2), 16, \ 16) -#define gen_gvec_dup64i(v1, c) \ - tcg_gen_gvec_dup64i(vec_full_reg_offset(v1), 16, 16, c) +#define gen_gvec_dup_imm(es, v1, c) \ + tcg_gen_gvec_dup_imm(es, vec_full_reg_offset(v1), 16, 16, c); #define gen_gvec_fn_2(fn, es, v1, v2) \ tcg_gen_gvec_##fn(es, vec_full_reg_offset(v1), vec_full_reg_offset(v2), \ 16, 16) @@ -316,31 +316,6 @@ static void gen_gvec128_4_i64(gen_gvec128_4_i64_fn fn, uint8_t d, uint8_t a, tcg_temp_free_i64(cl); } -static void gen_gvec_dupi(uint8_t es, uint8_t reg, uint64_t c) -{ - switch (es) { - case ES_8: - tcg_gen_gvec_dup8i(vec_full_reg_offset(reg), 16, 16, c); - break; - case ES_16: - tcg_gen_gvec_dup16i(vec_full_reg_offset(reg), 16, 16, c); - break; - case ES_32: - tcg_gen_gvec_dup32i(vec_full_reg_offset(reg), 16, 16, c); - break; - case ES_64: - gen_gvec_dup64i(reg, c); - break; - default: - g_assert_not_reached(); - } -} - -static void zero_vec(uint8_t reg) -{ - tcg_gen_gvec_dup8i(vec_full_reg_offset(reg), 16, 16, 0); -} - static void gen_addi2_i64(TCGv_i64 dl, TCGv_i64 dh, TCGv_i64 al, TCGv_i64 ah, uint64_t b) { @@ -396,8 +371,8 @@ static DisasJumpType op_vgbm(DisasContext *s, DisasOps *o) * Masks for both 64 bit elements of the vector are the same. * Trust tcg to produce a good constant loading. */ - gen_gvec_dup64i(get_field(s, v1), - generate_byte_mask(i2 & 0xff)); + gen_gvec_dup_imm(ES_64, get_field(s, v1), + generate_byte_mask(i2 & 0xff)); } else { TCGv_i64 t = tcg_temp_new_i64(); @@ -432,7 +407,7 @@ static DisasJumpType op_vgm(DisasContext *s, DisasOps *o) } } - gen_gvec_dupi(es, get_field(s, v1), mask); + gen_gvec_dup_imm(es, get_field(s, v1), mask); return DISAS_NEXT; } @@ -585,7 +560,7 @@ static DisasJumpType op_vllez(DisasContext *s, DisasOps *o) t = tcg_temp_new_i64(); tcg_gen_qemu_ld_i64(t, o->addr1, get_mem_index(s), MO_TE | es); - zero_vec(get_field(s, v1)); + gen_gvec_dup_imm(es, get_field(s, v1), 0); write_vec_element_i64(t, get_field(s, v1), enr, es); tcg_temp_free_i64(t); return DISAS_NEXT; @@ -892,7 +867,7 @@ static DisasJumpType op_vrepi(DisasContext *s, DisasOps *o) return DISAS_NORETURN; } - gen_gvec_dupi(es, get_field(s, v1), data); + gen_gvec_dup_imm(es, get_field(s, v1), data); return DISAS_NEXT; } @@ -1372,7 +1347,7 @@ static DisasJumpType op_vcksm(DisasContext *s, DisasOps *o) read_vec_element_i32(tmp, get_field(s, v2), i, ES_32); tcg_gen_add2_i32(tmp, sum, sum, sum, tmp, tmp); } - zero_vec(get_field(s, v1)); + gen_gvec_dup_imm(ES_32, get_field(s, v1), 0); write_vec_element_i32(sum, get_field(s, v1), 1, ES_32); tcg_temp_free_i32(tmp); diff --git a/tcg/tcg-op-gvec.c b/tcg/tcg-op-gvec.c index 327d9588e0..049a55e700 100644 --- a/tcg/tcg-op-gvec.c +++ b/tcg/tcg-op-gvec.c @@ -326,11 +326,34 @@ void tcg_gen_gvec_5_ptr(uint32_t dofs, uint32_t aofs, uint32_t bofs, in units of LNSZ. This limits the expansion of inline code. */ static inline bool check_size_impl(uint32_t oprsz, uint32_t lnsz) { - if (oprsz % lnsz == 0) { - uint32_t lnct = oprsz / lnsz; - return lnct >= 1 && lnct <= MAX_UNROLL; + uint32_t q, r; + + if (oprsz < lnsz) { + return false; + } + + q = oprsz / lnsz; + r = oprsz % lnsz; + tcg_debug_assert((r & 7) == 0); + + if (lnsz < 16) { + /* For sizes below 16, accept no remainder. */ + if (r != 0) { + return false; + } + } else { + /* + * Recall that ARM SVE allows vector sizes that are not a + * power of 2, but always a multiple of 16. The intent is + * that e.g. size == 80 would be expanded with 2x32 + 1x16. + * In addition, expand_clr needs to handle a multiple of 8. + * Thus we can handle the tail with one more operation per + * diminishing power of 2. + */ + q += ctpop32(r); } - return false; + + return q <= MAX_UNROLL; } static void expand_clr(uint32_t dofs, uint32_t maxsz); @@ -402,22 +425,31 @@ static void gen_dup_i64(unsigned vece, TCGv_i64 out, TCGv_i64 in) static TCGType choose_vector_type(const TCGOpcode *list, unsigned vece, uint32_t size, bool prefer_i64) { - if (TCG_TARGET_HAS_v256 && check_size_impl(size, 32)) { - /* - * Recall that ARM SVE allows vector sizes that are not a - * power of 2, but always a multiple of 16. The intent is - * that e.g. size == 80 would be expanded with 2x32 + 1x16. - * It is hard to imagine a case in which v256 is supported - * but v128 is not, but check anyway. - */ - if (tcg_can_emit_vecop_list(list, TCG_TYPE_V256, vece) - && (size % 32 == 0 - || tcg_can_emit_vecop_list(list, TCG_TYPE_V128, vece))) { - return TCG_TYPE_V256; - } - } - if (TCG_TARGET_HAS_v128 && check_size_impl(size, 16) - && tcg_can_emit_vecop_list(list, TCG_TYPE_V128, vece)) { + /* + * Recall that ARM SVE allows vector sizes that are not a + * power of 2, but always a multiple of 16. The intent is + * that e.g. size == 80 would be expanded with 2x32 + 1x16. + * It is hard to imagine a case in which v256 is supported + * but v128 is not, but check anyway. + * In addition, expand_clr needs to handle a multiple of 8. + */ + if (TCG_TARGET_HAS_v256 && + check_size_impl(size, 32) && + tcg_can_emit_vecop_list(list, TCG_TYPE_V256, vece) && + (!(size & 16) || + (TCG_TARGET_HAS_v128 && + tcg_can_emit_vecop_list(list, TCG_TYPE_V128, vece))) && + (!(size & 8) || + (TCG_TARGET_HAS_v64 && + tcg_can_emit_vecop_list(list, TCG_TYPE_V64, vece)))) { + return TCG_TYPE_V256; + } + if (TCG_TARGET_HAS_v128 && + check_size_impl(size, 16) && + tcg_can_emit_vecop_list(list, TCG_TYPE_V128, vece) && + (!(size & 8) || + (TCG_TARGET_HAS_v64 && + tcg_can_emit_vecop_list(list, TCG_TYPE_V64, vece)))) { return TCG_TYPE_V128; } if (TCG_TARGET_HAS_v64 && !prefer_i64 && check_size_impl(size, 8) @@ -432,6 +464,18 @@ static void do_dup_store(TCGType type, uint32_t dofs, uint32_t oprsz, { uint32_t i = 0; + tcg_debug_assert(oprsz >= 8); + + /* + * This may be expand_clr for the tail of an operation, e.g. + * oprsz == 8 && maxsz == 64. The first 8 bytes of this store + * are misaligned wrt the maximum vector size, so do that first. + */ + if (dofs & 8) { + tcg_gen_stl_vec(t_vec, cpu_env, dofs + i, TCG_TYPE_V64); + i += 8; + } + switch (type) { case TCG_TYPE_V256: /* @@ -619,17 +663,22 @@ static void expand_clr(uint32_t dofs, uint32_t maxsz) /* Expand OPSZ bytes worth of two-operand operations using i32 elements. */ static void expand_2_i32(uint32_t dofs, uint32_t aofs, uint32_t oprsz, - void (*fni)(TCGv_i32, TCGv_i32)) + bool load_dest, void (*fni)(TCGv_i32, TCGv_i32)) { TCGv_i32 t0 = tcg_temp_new_i32(); + TCGv_i32 t1 = tcg_temp_new_i32(); uint32_t i; for (i = 0; i < oprsz; i += 4) { tcg_gen_ld_i32(t0, cpu_env, aofs + i); - fni(t0, t0); - tcg_gen_st_i32(t0, cpu_env, dofs + i); + if (load_dest) { + tcg_gen_ld_i32(t1, cpu_env, dofs + i); + } + fni(t1, t0); + tcg_gen_st_i32(t1, cpu_env, dofs + i); } tcg_temp_free_i32(t0); + tcg_temp_free_i32(t1); } static void expand_2i_i32(uint32_t dofs, uint32_t aofs, uint32_t oprsz, @@ -749,17 +798,22 @@ static void expand_4_i32(uint32_t dofs, uint32_t aofs, uint32_t bofs, /* Expand OPSZ bytes worth of two-operand operations using i64 elements. */ static void expand_2_i64(uint32_t dofs, uint32_t aofs, uint32_t oprsz, - void (*fni)(TCGv_i64, TCGv_i64)) + bool load_dest, void (*fni)(TCGv_i64, TCGv_i64)) { TCGv_i64 t0 = tcg_temp_new_i64(); + TCGv_i64 t1 = tcg_temp_new_i64(); uint32_t i; for (i = 0; i < oprsz; i += 8) { tcg_gen_ld_i64(t0, cpu_env, aofs + i); - fni(t0, t0); - tcg_gen_st_i64(t0, cpu_env, dofs + i); + if (load_dest) { + tcg_gen_ld_i64(t1, cpu_env, dofs + i); + } + fni(t1, t0); + tcg_gen_st_i64(t1, cpu_env, dofs + i); } tcg_temp_free_i64(t0); + tcg_temp_free_i64(t1); } static void expand_2i_i64(uint32_t dofs, uint32_t aofs, uint32_t oprsz, @@ -880,17 +934,23 @@ static void expand_4_i64(uint32_t dofs, uint32_t aofs, uint32_t bofs, /* Expand OPSZ bytes worth of two-operand operations using host vectors. */ static void expand_2_vec(unsigned vece, uint32_t dofs, uint32_t aofs, uint32_t oprsz, uint32_t tysz, TCGType type, + bool load_dest, void (*fni)(unsigned, TCGv_vec, TCGv_vec)) { TCGv_vec t0 = tcg_temp_new_vec(type); + TCGv_vec t1 = tcg_temp_new_vec(type); uint32_t i; for (i = 0; i < oprsz; i += tysz) { tcg_gen_ld_vec(t0, cpu_env, aofs + i); - fni(vece, t0, t0); - tcg_gen_st_vec(t0, cpu_env, dofs + i); + if (load_dest) { + tcg_gen_ld_vec(t1, cpu_env, dofs + i); + } + fni(vece, t1, t0); + tcg_gen_st_vec(t1, cpu_env, dofs + i); } tcg_temp_free_vec(t0); + tcg_temp_free_vec(t1); } /* Expand OPSZ bytes worth of two-vector operands and an immediate operand @@ -1044,7 +1104,8 @@ void tcg_gen_gvec_2(uint32_t dofs, uint32_t aofs, * that e.g. size == 80 would be expanded with 2x32 + 1x16. */ some = QEMU_ALIGN_DOWN(oprsz, 32); - expand_2_vec(g->vece, dofs, aofs, some, 32, TCG_TYPE_V256, g->fniv); + expand_2_vec(g->vece, dofs, aofs, some, 32, TCG_TYPE_V256, + g->load_dest, g->fniv); if (some == oprsz) { break; } @@ -1054,17 +1115,19 @@ void tcg_gen_gvec_2(uint32_t dofs, uint32_t aofs, maxsz -= some; /* fallthru */ case TCG_TYPE_V128: - expand_2_vec(g->vece, dofs, aofs, oprsz, 16, TCG_TYPE_V128, g->fniv); + expand_2_vec(g->vece, dofs, aofs, oprsz, 16, TCG_TYPE_V128, + g->load_dest, g->fniv); break; case TCG_TYPE_V64: - expand_2_vec(g->vece, dofs, aofs, oprsz, 8, TCG_TYPE_V64, g->fniv); + expand_2_vec(g->vece, dofs, aofs, oprsz, 8, TCG_TYPE_V64, + g->load_dest, g->fniv); break; case 0: if (g->fni8 && check_size_impl(oprsz, 8)) { - expand_2_i64(dofs, aofs, oprsz, g->fni8); + expand_2_i64(dofs, aofs, oprsz, g->load_dest, g->fni8); } else if (g->fni4 && check_size_impl(oprsz, 4)) { - expand_2_i32(dofs, aofs, oprsz, g->fni4); + expand_2_i32(dofs, aofs, oprsz, g->load_dest, g->fni4); } else { assert(g->fno != NULL); tcg_gen_gvec_2_ool(dofs, aofs, oprsz, maxsz, g->data, g->fno); @@ -1541,32 +1604,11 @@ void tcg_gen_gvec_dup_mem(unsigned vece, uint32_t dofs, uint32_t aofs, } } -void tcg_gen_gvec_dup64i(uint32_t dofs, uint32_t oprsz, - uint32_t maxsz, uint64_t x) -{ - check_size_align(oprsz, maxsz, dofs); - do_dup(MO_64, dofs, oprsz, maxsz, NULL, NULL, x); -} - -void tcg_gen_gvec_dup32i(uint32_t dofs, uint32_t oprsz, - uint32_t maxsz, uint32_t x) -{ - check_size_align(oprsz, maxsz, dofs); - do_dup(MO_32, dofs, oprsz, maxsz, NULL, NULL, x); -} - -void tcg_gen_gvec_dup16i(uint32_t dofs, uint32_t oprsz, - uint32_t maxsz, uint16_t x) -{ - check_size_align(oprsz, maxsz, dofs); - do_dup(MO_16, dofs, oprsz, maxsz, NULL, NULL, x); -} - -void tcg_gen_gvec_dup8i(uint32_t dofs, uint32_t oprsz, - uint32_t maxsz, uint8_t x) +void tcg_gen_gvec_dup_imm(unsigned vece, uint32_t dofs, uint32_t oprsz, + uint32_t maxsz, uint64_t x) { check_size_align(oprsz, maxsz, dofs); - do_dup(MO_8, dofs, oprsz, maxsz, NULL, NULL, x); + do_dup(vece, dofs, oprsz, maxsz, NULL, NULL, x); } void tcg_gen_gvec_not(unsigned vece, uint32_t dofs, uint32_t aofs, @@ -2319,7 +2361,7 @@ void tcg_gen_gvec_xor(unsigned vece, uint32_t dofs, uint32_t aofs, }; if (aofs == bofs) { - tcg_gen_gvec_dup8i(dofs, oprsz, maxsz, 0); + tcg_gen_gvec_dup_imm(MO_64, dofs, oprsz, maxsz, 0); } else { tcg_gen_gvec_3(dofs, aofs, bofs, oprsz, maxsz, &g); } @@ -2336,7 +2378,7 @@ void tcg_gen_gvec_andc(unsigned vece, uint32_t dofs, uint32_t aofs, }; if (aofs == bofs) { - tcg_gen_gvec_dup8i(dofs, oprsz, maxsz, 0); + tcg_gen_gvec_dup_imm(MO_64, dofs, oprsz, maxsz, 0); } else { tcg_gen_gvec_3(dofs, aofs, bofs, oprsz, maxsz, &g); } @@ -2353,7 +2395,7 @@ void tcg_gen_gvec_orc(unsigned vece, uint32_t dofs, uint32_t aofs, }; if (aofs == bofs) { - tcg_gen_gvec_dup8i(dofs, oprsz, maxsz, -1); + tcg_gen_gvec_dup_imm(MO_64, dofs, oprsz, maxsz, -1); } else { tcg_gen_gvec_3(dofs, aofs, bofs, oprsz, maxsz, &g); } @@ -2404,7 +2446,7 @@ void tcg_gen_gvec_eqv(unsigned vece, uint32_t dofs, uint32_t aofs, }; if (aofs == bofs) { - tcg_gen_gvec_dup8i(dofs, oprsz, maxsz, -1); + tcg_gen_gvec_dup_imm(MO_64, dofs, oprsz, maxsz, -1); } else { tcg_gen_gvec_3(dofs, aofs, bofs, oprsz, maxsz, &g); } diff --git a/tcg/tcg-op.c b/tcg/tcg-op.c index e2e25ebf7d..e60b74fb82 100644 --- a/tcg/tcg-op.c +++ b/tcg/tcg-op.c @@ -540,9 +540,9 @@ void tcg_gen_rotl_i32(TCGv_i32 ret, TCGv_i32 arg1, TCGv_i32 arg2) } } -void tcg_gen_rotli_i32(TCGv_i32 ret, TCGv_i32 arg1, unsigned arg2) +void tcg_gen_rotli_i32(TCGv_i32 ret, TCGv_i32 arg1, int32_t arg2) { - tcg_debug_assert(arg2 < 32); + tcg_debug_assert(arg2 >= 0 && arg2 < 32); /* some cases can be optimized here */ if (arg2 == 0) { tcg_gen_mov_i32(ret, arg1); @@ -580,9 +580,9 @@ void tcg_gen_rotr_i32(TCGv_i32 ret, TCGv_i32 arg1, TCGv_i32 arg2) } } -void tcg_gen_rotri_i32(TCGv_i32 ret, TCGv_i32 arg1, unsigned arg2) +void tcg_gen_rotri_i32(TCGv_i32 ret, TCGv_i32 arg1, int32_t arg2) { - tcg_debug_assert(arg2 < 32); + tcg_debug_assert(arg2 >= 0 && arg2 < 32); /* some cases can be optimized here */ if (arg2 == 0) { tcg_gen_mov_i32(ret, arg1); @@ -1962,9 +1962,9 @@ void tcg_gen_rotl_i64(TCGv_i64 ret, TCGv_i64 arg1, TCGv_i64 arg2) } } -void tcg_gen_rotli_i64(TCGv_i64 ret, TCGv_i64 arg1, unsigned arg2) +void tcg_gen_rotli_i64(TCGv_i64 ret, TCGv_i64 arg1, int64_t arg2) { - tcg_debug_assert(arg2 < 64); + tcg_debug_assert(arg2 >= 0 && arg2 < 64); /* some cases can be optimized here */ if (arg2 == 0) { tcg_gen_mov_i64(ret, arg1); @@ -2001,9 +2001,9 @@ void tcg_gen_rotr_i64(TCGv_i64 ret, TCGv_i64 arg1, TCGv_i64 arg2) } } -void tcg_gen_rotri_i64(TCGv_i64 ret, TCGv_i64 arg1, unsigned arg2) +void tcg_gen_rotri_i64(TCGv_i64 ret, TCGv_i64 arg1, int64_t arg2) { - tcg_debug_assert(arg2 < 64); + tcg_debug_assert(arg2 >= 0 && arg2 < 64); /* some cases can be optimized here */ if (arg2 == 0) { tcg_gen_mov_i64(ret, arg1); |