s390x/tcg: Implement Vector-Enhancements Facility 2
v2 cover s390x/tcg: Implement Vector-Enhancements Facility 2
Implement Vector-Enhancements Facility 2 for s390x
resolves: https://gitlab.com/qemu-project/qemu/-/issues/738
implements:
VECTOR LOAD ELEMENTS REVERSED (VLER)
VECTOR LOAD BYTE REVERSED ELEMENTS (VLBR)
VECTOR LOAD BYTE REVERSED ELEMENT (VLEBRH, VLEBRF, VLEBRG)
VECTOR LOAD BYTE REVERSED ELEMENT AND ZERO (VLLEBRZ) VECTOR LOAD BYTE REVERSED ELEMENT AND REPLICATE (VLBRREP)
VECTOR STORE ELEMENTS REVERSED (VSTER)
VECTOR STORE BYTE REVERSED ELEMENTS (VSTBR)
VECTOR STORE BYTE REVERSED ELEMENTS (VSTEBRH, VSTEBRF, VSTEBRG)
VECTOR SHIFT LEFT DOUBLE BY BIT (VSLD)
VECTOR SHIFT RIGHT DOUBLE BY BIT (VSRD)
VECTOR STRING SEARCH (VSTRS)
modifies:
VECTOR FP CONVERT FROM FIXED (VCFPS)
VECTOR FP CONVERT FROM LOGICAL (VCFPL)
VECTOR FP CONVERT TO FIXED (VCSFP)
VECTOR FP CONVERT TO LOGICAL (VCLFP)
VECTOR SHIFT LEFT (VSL)
VECTOR SHIFT RIGHT ARITHMETIC (VSRA)
VECTOR SHIFT RIGHT LOGICAL (VSRL)
David Miller (7):
target/s390x: vxeh2: vector convert short/32b target/s390x: vxeh2: vector string search target/s390x: vxeh2: vector shift {double by bit, left, right
{logical,arithmetic}} target/s390x: vxeh2: vector {load,store} elements reversed target/s390x: vxeh2: vector {load,store} reversed elements [and
{zero,replicate}]
target/s390x: add S390_FEAT_VECTOR_ENH2 to cpu max tests/tcg/s390x: Tests for Vector Enhancements Facility 2
include/qemu/bitops.h | 25 ++ target/s390x/gen-features.c | 2 + target/s390x/helper.h | 12 +- target/s390x/tcg/insn-data.def | 28 +- target/s390x/tcg/translate.c | 3 +- target/s390x/tcg/translate_vx.c.inc | 369 ++++++++++++++++++++++++--- target/s390x/tcg/vec.h | 24 ++ target/s390x/tcg/vec_fpu_helper.c | 31 +++ target/s390x/tcg/vec_helper.c | 31 +++ target/s390x/tcg/vec_int_helper.c | 58 +++++ target/s390x/tcg/vec_string_helper.c | 65 +++++ tests/tcg/s390x/Makefile.target | 8 + tests/tcg/s390x/vxeh2_vcvt.c | 97 +++++++ tests/tcg/s390x/vxeh2_vlstr.c | 146 +++++++++++ tests/tcg/s390x/vxeh2_vs.c | 91 +++++++ 15 files changed, 947 insertions(+), 43 deletions(-) create mode 100644 tests/tcg/s390x/vxeh2_vcvt.c create mode 100644 tests/tcg/s390x/vxeh2_vlstr.c create mode 100644 tests/tcg/s390x/vxeh2_vs.c
Signature · 1 line
2.34.1
PATCH v2 2/7 target/s390x: vxeh2: vector string search
--- target/s390x/helper.h | 1 + target/s390x/tcg/insn-data.def | 2 + target/s390x/tcg/translate.c | 3 +- target/s390x/tcg/translate_vx.c.inc | 17 ++++++++ target/s390x/tcg/vec_string_helper.c | 65 ++++++++++++++++++++++++++++ 5 files changed, 87 insertions(+), 1 deletion(-)
target/s390x/helper.h +1 -0
@@ -246,6 +246,7 @@ DEF_HELPER_6(gvec_vstrc_cc32, void, ptr, cptr, cptr, cptr, env, i32) DEF_HELPER_6(gvec_vstrc_cc_rt8, void, ptr, cptr, cptr, cptr, env, i32) DEF_HELPER_6(gvec_vstrc_cc_rt16, void, ptr, cptr, cptr, cptr, env, i32) DEF_HELPER_6(gvec_vstrc_cc_rt32, void, ptr, cptr, cptr, cptr, env, i32) +DEF_HELPER_6 (vstrs, void, ptr, cptr, cptr, ptr, env, i32) /* === Vector Floating-Point Instructions */DEF_HELPER_FLAGS_5 (gvec_vfa32,TCG_CALL_NO_WG , void, ptr, cptr, cptr, env, i32)
target/s390x/tcg/insn-data.def +2 -0
@@ -1246,6 +1246,8 @@ F(0xe75c, VISTR, VRR_a, V, 0, 0, 0, 0, vistr, 0, IF_VEC) /* VECTOR STRING RANGE COMPARE */ F(0xe78a, VSTRC, VRR_d, V, 0, 0, 0, 0, vstrc, 0, IF_VEC) +/* VECTOR STRING SEARCH */ + F(0xe78b, VSTRS, VRR_d, VE2, 0, 0, 0, 0, vstrs, 0, IF_VEC) /* === Vector Floating-Point Instructions */
target/s390x/tcg/translate.c +2 -1
@@ -6222,7 +6222,8 @@ enum DisasInsnEnum {#define FAC_PCI S390_FEAT_ZPCI/* z/PCI facility */ #define FAC_AIS S390_FEAT_ADAPTER_INT_SUPPRESSION#define FAC_V S390_FEAT_VECTOR/* vector facility */ -#define FAC_VE S390_FEAT_VECTOR_ENH/* vector enhancements facility 1 */ +#define FAC_VE S390_FEAT_VECTOR_ENH/* vector enhancements facility 1 */ +#define FAC_VE2 S390_FEAT_VECTOR_ENH2/* vector enhancements facility 2 */ #define FAC_MIE2 S390_FEAT_MISC_INSTRUCTION_EXT2/* miscellaneous-instruction-extensions facility 2 */ #define FAC_MIE3 S390_FEAT_MISC_INSTRUCTION_EXT3/* miscellaneous-instruction-extensions facility 3 */
target/s390x/tcg/translate_vx.c.inc +17 -0
@@ -2497,6 +2497,23 @@ static DisasJumpType op_vstrc(DisasContext *s, DisasOps *o) return DISAS_NEXT; } +static DisasJumpType op_vstrs(DisasContext *s, DisasOps *o) +{ + const uint8_t es = get_field(s, m5); + const uint32_t D = get_field(s, m6); + + if (es > ES_32) { + gen_program_exception(s, PGM_SPECIFICATION); + return DISAS_NORETURN; + } + gen_gvec_4_ptr(get_field(s, v1), get_field(s, v2), + get_field(s, v3), get_field(s, v4), + cpu_env, (D << 16) | es, gen_helper_vstrs); + + set_cc_static(s); + return DISAS_NEXT; +} + static DisasJumpType op_vfa(DisasContext *s, DisasOps *o) { const uint8_t fpf = get_field(s, m4);
target/s390x/tcg/vec_string_helper.c +65 -0
68 lines not on this page · raw message
Signature · 1 line
2.34.1
PATCH v2 1/7 target/s390x: vxeh2: vector convert short/32b
In reply to David Miller--- target/s390x/helper.h | 4 +++ target/s390x/tcg/translate_vx.c.inc | 44 ++++++++++++++++++++++++++--- target/s390x/tcg/vec_fpu_helper.c | 31 ++++++++++++++++++++ 3 files changed, 75 insertions(+), 4 deletions(-)
target/s390x/helper.h +4 -0
@@ -275,6 +275,10 @@ DEF_HELPER_FLAGS_5(gvec_vfche64, TCG_CALL_NO_WG, void, ptr, cptr, cptr, env, i32 DEF_HELPER_5(gvec_vfche64_cc, void, ptr, cptr, cptr, env, i32) DEF_HELPER_FLAGS_5(gvec_vfche128, TCG_CALL_NO_WG, void, ptr, cptr, cptr, env, i32) DEF_HELPER_5(gvec_vfche128_cc, void, ptr, cptr, cptr, env, i32) +DEF_HELPER_FLAGS_4(gvec_vcdg32, TCG_CALL_NO_WG, void, ptr, cptr, env, i32) +DEF_HELPER_FLAGS_4(gvec_vcdlg32, TCG_CALL_NO_WG, void, ptr, cptr, env, i32) +DEF_HELPER_FLAGS_4(gvec_vcgd32, TCG_CALL_NO_WG, void, ptr, cptr, env, i32) +DEF_HELPER_FLAGS_4(gvec_vclgd32, TCG_CALL_NO_WG, void, ptr, cptr, env, i32) DEF_HELPER_FLAGS_4(gvec_vcdg64, TCG_CALL_NO_WG, void, ptr, cptr, env, i32) DEF_HELPER_FLAGS_4(gvec_vcdlg64, TCG_CALL_NO_WG, void, ptr, cptr, env, i32)DEF_HELPER_FLAGS_4 (gvec_vcgd64,TCG_CALL_NO_WG , void, ptr, cptr, env, i32)
target/s390x/tcg/translate_vx.c.inc +40 -4
63 lines not on this page · raw message
target/s390x/tcg/vec_fpu_helper.c +31 -0
@@ -176,6 +176,30 @@ static void vop128_2(S390Vector *v1, const S390Vector *v2, CPUS390XState *env, *v1 = tmp; } +static float32 vcdg32 (float32 a,float_status *s) +{ +return int32_to_float32 (a, s); +} + +static float32 vcdlg32 (float32 a,float_status *s) +{ +return uint32_to_float32 (a, s); +} + +static float32 vcgd32 (float32 a,float_status *s) +{ +const float32 tmp =float32_to_int32 (a, s); + +return float32_is_any_nan (a) ?INT32_MIN : tmp; +} + +static float32 vclgd32 (float32 a,float_status *s) +{ +const float32 tmp =float32_to_uint32 (a, s); + +return float32_is_any_nan (a) ?0 : tmp; +} +static float64 vcdg64 (float64 a,float_status *s) {return int64_to_float64 (a, s);
@@ -211,6 +235,9 @@ void HELPER(gvec_##NAME##BITS)(void *v1, const void *v2, CPUS390XState *env, \ vop##BITS##_2(v1, v2, env, se, XxC, erm, FN, GETPC()); \ } +#define DEF_GVEC_VOP2_32 (NAME ) \ +DEF_GVEC_VOP2_FN(NAME, NAME##32, 32) +#define DEF_GVEC_VOP2_64 (NAME ) \ DEF_GVEC_VOP2_FN(NAME, NAME##64, 64)
@@ -219,6 +246,10 @@ DEF_GVEC_VOP2_FN(NAME, float32_##OP, 32) \ DEF_GVEC_VOP2_FN(NAME, float64_##OP, 64) \ DEF_GVEC_VOP2_FN(NAME, float128_##OP, 128) +DEF_GVEC_VOP2_32(vcdg) +DEF_GVEC_VOP2_32(vcdlg) +DEF_GVEC_VOP2_32(vcgd) +DEF_GVEC_VOP2_32(vclgd) DEF_GVEC_VOP2_64(vcdg) DEF_GVEC_VOP2_64(vcdlg) DEF_GVEC_VOP2_64(vcgd)
Signature · 1 line
2.34.1
PATCH v2 3/7 target/s390x: vxeh2: vector shift {double by bit, left, right {logical, arithmetic}}
In reply to David Miller--- include/qemu/bitops.h | 25 +++++++ target/s390x/helper.h | 3 + target/s390x/tcg/insn-data.def | 6 +- target/s390x/tcg/translate_vx.c.inc | 108 ++++++++++++++++++---------- target/s390x/tcg/vec_int_helper.c | 58 +++++++++++++++ 5 files changed, 163 insertions(+), 37 deletions(-)
include/qemu/bitops.h +25 -0
@@ -445,6 +445,31 @@ static inline int64_t sextract64(uint64_t value, int start, int length)*/ return ((int64_t )(value << (64 - length - start))) >> (64 - length); } +/** +* deposit8: +* @value: initial value to insert bit field into +* @start: the lowest bit in the bit field (numbered from 0) +* @length: the length of the bit field +* @fieldval: the value to insert into the bit field +* +* Deposit @fieldval into the 8 bit @value at the bit field specified +* by the @start and @length parameters, and return the modified +* @value. Bits of @value outside the bit field are not modified. +* Bits of @fieldval above the least significant @length bits are +* ignored. The bit field must lie entirely within the 8 bit byte. +* It is valid to request that all 8 bits are modified (ie @length +* 8 and @start 0). +* +* Returns: the modified @value. +*/ +static inline uint8_t deposit8 (uint8_t value,int start,int length, +uint8_t fieldval) +{ +uint8_t mask; +assert (start >=0 && length >0 && length <=8 - start); + mask = (~0ULL >> (8 - length)) << start; +return (value & ~mask) | ((fieldval << start) & mask); +}/** * deposit32:
target/s390x/helper.h +3 -0
@@ -203,8 +203,11 @@ DEF_HELPER_FLAGS_3(gvec_vpopct16, TCG_CALL_NO_RWG, void, ptr, cptr, i32) DEF_HELPER_FLAGS_4(gvec_verim8, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, i32) DEF_HELPER_FLAGS_4(gvec_verim16, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, i32) DEF_HELPER_FLAGS_4(gvec_vsl, TCG_CALL_NO_RWG, void, ptr, cptr, i64, i32) +DEF_HELPER_FLAGS_4(gvec_vsl_ve2, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, i32) DEF_HELPER_FLAGS_4(gvec_vsra, TCG_CALL_NO_RWG, void, ptr, cptr, i64, i32) +DEF_HELPER_FLAGS_4(gvec_vsra_ve2, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, i32) DEF_HELPER_FLAGS_4(gvec_vsrl, TCG_CALL_NO_RWG, void, ptr, cptr, i64, i32) +DEF_HELPER_FLAGS_4(gvec_vsrl_ve2, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, i32) DEF_HELPER_FLAGS_4(gvec_vscbi8, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, i32) DEF_HELPER_FLAGS_4(gvec_vscbi16, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, i32)DEF_HELPER_4 (gvec_vtm, void, ptr, cptr, env, i32)
target/s390x/tcg/insn-data.def +5 -1
@@ -1207,12 +1207,16 @@ F(0xe774, VSL, VRR_c, V, 0, 0, 0, 0, vsl, 0, IF_VEC) /* VECTOR SHIFT LEFT BY BYTE */ F(0xe775, VSLB, VRR_c, V, 0, 0, 0, 0, vsl, 0, IF_VEC) +/* VECTOR SHIFT LEFT DOUBLE BY BIT */ + F(0xe786, VSLD, VRI_d, VE2, 0, 0, 0, 0, vsld, 0, IF_VEC) /* VECTOR SHIFT LEFT DOUBLE BY BYTE */ - F(0xe777, VSLDB, VRI_d, V, 0, 0, 0, 0, vsldb, 0, IF_VEC) + F(0xe777, VSLDB, VRI_d, V, 0, 0, 0, 0, vsld, 0, IF_VEC) /* VECTOR SHIFT RIGHT ARITHMETIC */ F(0xe77e, VSRA, VRR_c, V, 0, 0, 0, 0, vsra, 0, IF_VEC) /* VECTOR SHIFT RIGHT ARITHMETIC BY BYTE */ F(0xe77f, VSRAB, VRR_c, V, 0, 0, 0, 0, vsra, 0, IF_VEC) +/* VECTOR SHIFT RIGHT DOUBLE BY BIT */ + F(0xe787, VSRD, VRI_d, VE2, 0, 0, 0, 0, vsrd, 0, IF_VEC) /* VECTOR SHIFT RIGHT LOGICAL */ F(0xe77c, VSRL, VRR_c, V, 0, 0, 0, 0, vsrl, 0, IF_VEC) /* VECTOR SHIFT RIGHT LOGICAL BY BYTE */
target/s390x/tcg/translate_vx.c.inc +72 -36
143 lines not on this page · raw message
target/s390x/tcg/vec_int_helper.c +58 -0
76 lines not on this page · raw message
Signature · 1 line
2.34.1
PATCH v2 4/7 target/s390x: vxeh2: vector {load, store} elements reversed
In reply to David Miller--- target/s390x/helper.h | 4 ++- target/s390x/tcg/insn-data.def | 4 +++ target/s390x/tcg/translate_vx.c.inc | 39 +++++++++++++++++++++++++++++ target/s390x/tcg/vec.h | 24 ++++++++++++++++++ target/s390x/tcg/vec_helper.c | 31 +++++++++++++++++++++++ 5 files changed, 101 insertions(+), 1 deletion(-)
target/s390x/helper.h +3 -1
@@ -129,6 +129,9 @@ DEF_HELPER_FLAGS_3(probe_write_access, TCG_CALL_NO_WG, void, env, i64, i64)/* === Vector Support Instructions === */ DEF_HELPER_FLAGS_4(gvec_vbperm, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, i32) DEF_HELPER_FLAGS_4(vll, TCG_CALL_NO_WG, void, env, ptr, i64, i64) +DEF_HELPER_FLAGS_4(vler, TCG_CALL_NO_WG, void, env, ptr, i64, i64) +DEF_HELPER_FLAGS_4(vster, TCG_CALL_NO_WG, void, env, ptr, i64, i64) +DEF_HELPER_FLAGS_4(vstl, TCG_CALL_NO_WG, void, env, cptr, i64, i64) DEF_HELPER_FLAGS_4(gvec_vpk16, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, i32) DEF_HELPER_FLAGS_4(gvec_vpk32, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, i32)DEF_HELPER_FLAGS_4 (gvec_vpk64,TCG_CALL_NO_RWG , void, ptr, cptr, cptr, i32)
@@ -145,7 +148,6 @@ DEF_HELPER_5(gvec_vpkls_cc16, void, ptr, cptr, cptr, env, i32) DEF_HELPER_5(gvec_vpkls_cc32, void, ptr, cptr, cptr, env, i32) DEF_HELPER_5(gvec_vpkls_cc64, void, ptr, cptr, cptr, env, i32)DEF_HELPER_FLAGS_5 (gvec_vperm,TCG_CALL_NO_RWG , void, ptr, cptr, cptr, cptr, i32) -DEF_HELPER_FLAGS_4 (vstl,TCG_CALL_NO_WG , void, env, cptr, i64, i64) /* === Vector Integer Instructions === */DEF_HELPER_FLAGS_4 (gvec_vavg8,TCG_CALL_NO_RWG , void, ptr, cptr, cptr, i32)
target/s390x/tcg/insn-data.def +4 -0
@@ -1037,6 +1037,8 @@ E(0xe741, VLEIH, VRI_a, V, 0, 0, 0, 0, vlei, 0, ES_16, IF_VEC) E(0xe743, VLEIF, VRI_a, V, 0, 0, 0, 0, vlei, 0, ES_32, IF_VEC) E(0xe742, VLEIG, VRI_a, V, 0, 0, 0, 0, vlei, 0, ES_64, IF_VEC) +/* VECTOR LOAD ELEMENTS REVERSED */ + F(0xe607, VLER, VRX, VE2, la2, 0, 0, 0, vler, 0, IF_VEC) /* VECTOR LOAD GR FROM VR ELEMENT */ F(0xe721, VLGV, VRS_c, V, la2, 0, r1, 0, vlgv, 0, IF_VEC) /* VECTOR LOAD LOGICAL ELEMENT AND ZERO */
@@ -1082,6 +1084,8 @@ E(0xe709, VSTEH, VRX, V, la2, 0, 0, 0, vste, 0, ES_16, IF_VEC) E(0xe70b, VSTEF, VRX, V, la2, 0, 0, 0, vste, 0, ES_32, IF_VEC) E(0xe70a, VSTEG, VRX, V, la2, 0, 0, 0, vste, 0, ES_64, IF_VEC) +/* VECTOR STORE ELEMENTS REVERSED */ + F(0xe60f, VSTER, VRX, VE2, la2, 0, 0, 0, vster, 0, IF_VEC) /* VECTOR STORE MULTIPLE */ F(0xe73e, VSTM, VRS_a, V, la2, 0, 0, 0, vstm, 0, IF_VEC) /* VECTOR STORE WITH LENGTH */
target/s390x/tcg/translate_vx.c.inc +39 -0
@@ -492,6 +492,26 @@ static DisasJumpType op_vlei(DisasContext *s, DisasOps *o) return DISAS_NEXT; } +static DisasJumpType op_vler(DisasContext *s, DisasOps *o) +{ + const uint8_t es = get_field(s, m3); + + if (es < ES_16 || es > ES_64) { + gen_program_exception(s, PGM_SPECIFICATION); + return DISAS_NORETURN; + } + + TCGv_ptr a0 = tcg_temp_new_ptr(); + TCGv_i64 tes = tcg_const_i64(es & 0xf); + + tcg_gen_addi_ptr(a0, cpu_env, vec_full_reg_offset(get_field(s, v1))); + gen_helper_vler(cpu_env, a0, o->addr1, tes); + + tcg_temp_free_i64(tes); + tcg_temp_free_ptr(a0); + return DISAS_NEXT; +} + static DisasJumpType op_vlgv(DisasContext *s, DisasOps *o) { const uint8_t es = get_field(s, m4);
@@ -976,6 +996,25 @@ static DisasJumpType op_vste(DisasContext *s, DisasOps *o) return DISAS_NEXT; } +static DisasJumpType op_vster(DisasContext *s, DisasOps *o) +{ + const uint8_t es = get_field(s, m3); + if (es < ES_16 || es > ES_64) { + gen_program_exception(s, PGM_SPECIFICATION); + return DISAS_NORETURN; + } + + TCGv_ptr a0 = tcg_temp_new_ptr(); + TCGv_i64 tes = tcg_const_i64(es); + + tcg_gen_addi_ptr(a0, cpu_env, vec_full_reg_offset(get_field(s, v1))); + gen_helper_vster(cpu_env, a0, o->addr1, tes); + + tcg_temp_free_i64(tes); + tcg_temp_free_ptr(a0); + return DISAS_NEXT; +} + static DisasJumpType op_vstm(DisasContext *s, DisasOps *o) { const uint8_t v3 = get_field(s, v3);
target/s390x/tcg/vec.h +24 -0
@@ -138,4 +138,28 @@ static inline void s390_vec_write_element(S390Vector *v, uint8_t enr, } } +static inline void s390_vec_reverse (uint64_t *vdst,uint64_t *vsrc,uint8_t es) +{ +const uint64_t l = vsrc[0 ], h = vsrc[1 ]; + +switch (es) { +case MO_8 : + vdst[1 ] =bswap64 (l); + vdst[0 ] =bswap64 (h); +break ; +case MO_16 : + vdst[1 ] =hswap64 (l); + vdst[0 ] =hswap64 (h); +break ; +case MO_32 : + vdst[1 ] =wswap64 (l); + vdst[0 ] =wswap64 (h); +break ; +case MO_64 : + vdst[1 ] = l; + vdst[0 ] = h; +break ; + } +} +#endif /* S390X_VEC_H */
target/s390x/tcg/vec_helper.c +31 -0
@@ -212,3 +212,34 @@ void HELPER(vstl)(CPUS390XState *env, const void *v1, uint64_t addr, *(S390Vector *)v1 = tmp; } } + +void HELPER (vler )(CPUS390XState *env,void *v1,uint64_t addr,uint64_t es) +{ +uint64_t t0, t1; + t0 =cpu_ldq_data_ra (env, addr,GETPC ()); + addr =wrap_address (env, addr +8 ); + t1 =cpu_ldq_data_ra (env, addr,GETPC ()); + +if (MO_64 == es) { +s390_vec_write_element64 (v1,1 , t0); +s390_vec_write_element64 (v1,0 , t1); + }else { +S390Vector tmp = {}; +s390_vec_write_element64 (&tmp,0 , t0); +s390_vec_write_element64 (&tmp,1 , t1); +s390_vec_reverse (v1, &tmp, (uint8_t )es); + } +} + +void HELPER (vster )(CPUS390XState *env,void *v1,uint64_t addr,uint64_t es) +{ +S390Vector tmp = {}; +/* Probe write access before actually modifying memory */ +probe_write_access (env, addr,16 ,GETPC ()); + +s390_vec_reverse (&tmp, v1, (uint8_t )es); + +cpu_stq_data_ra (env, addr, tmp.doubleword[0 ],GETPC ()); + addr =wrap_address (env, addr +8 ); +cpu_stq_data_ra (env, addr, tmp.doubleword[1 ],GETPC ()); +}
Signature · 1 line
2.34.1
PATCH v2 6/7 target/s390x: add S390_FEAT_VECTOR_ENH2 to cpu max
In reply to David Miller--- target/s390x/gen-features.c | 2 ++ 1 file changed, 2 insertions(+)
target/s390x/gen-features.c +2 -0
@@ -740,7 +740,9 @@ static uint16_t qemu_V6_2[] = {static uint16_t qemu_LATEST[] = {S390_FEAT_MISC_INSTRUCTION_EXT3 , +S390_FEAT_VECTOR_ENH2 , }; +/* add all new definitions before this point */ static uint16_t qemu_MAX[] = {/* generates a dependency warning, leave it out for now */
Signature · 1 line
2.34.1
PATCH v2 5/7 target/s390x: vxeh2: vector {load, store} reversed elements [and {zero, replicate}]
In reply to David Miller--- target/s390x/tcg/insn-data.def | 16 +++ target/s390x/tcg/translate_vx.c.inc | 161 ++++++++++++++++++++++++++++ 2 files changed, 177 insertions(+)
target/s390x/tcg/insn-data.def +16 -0
@@ -1027,6 +1027,16 @@ F(0xe756, VLR, VRR_a, V, 0, 0, 0, 0, vlr, 0, IF_VEC) /* VECTOR LOAD AND REPLICATE */ F(0xe705, VLREP, VRX, V, la2, 0, 0, 0, vlrep, 0, IF_VEC) +/* VECTOR LOAD BYTE REVERSED ELEMENTS */ + F(0xe601, VLEBRH, VRX, VE2, la2, 0, 0, 0, vlebr, 0, IF_VEC) + F(0xe603, VLEBRF, VRX, VE2, la2, 0, 0, 0, vlebr, 0, IF_VEC) + F(0xe602, VLEBRG, VRX, VE2, la2, 0, 0, 0, vlebr, 0, IF_VEC) +/* VECTOR LOAD BYTE REVERSED ELEMENT AND REPLOCATE */ + F(0xe605, VLBRREP, VRX, VE2, la2, 0, 0, 0, vlbrrep, 0, IF_VEC) +/* VECTOR LOAD BYTE REVERSED ELEMENT AND ZERO */ + F(0xe604, VLLEBRZ, VRX, VE2, la2, 0, 0, 0, vllebrz, 0, IF_VEC) +/* VECTOR LOAD BYTE REVERSED ELEMENTS */ + F(0xe606, VLBR, VRX, VE2, la2, 0, 0, 0, vlbr, 0, IF_VEC) /* VECTOR LOAD ELEMENT */ E(0xe700, VLEB, VRX, V, la2, 0, 0, 0, vle, 0, ES_8, IF_VEC) E(0xe701, VLEH, VRX, V, la2, 0, 0, 0, vle, 0, ES_16, IF_VEC)
@@ -1079,6 +1089,12 @@ F(0xe75f, VSEG, VRR_a, V, 0, 0, 0, 0, vseg, 0, IF_VEC) /* VECTOR STORE */ F(0xe70e, VST, VRX, V, la2, 0, 0, 0, vst, 0, IF_VEC) +/* VECTOR STORE BYTE REVERSED ELEMENT */ + F(0xe609, VSTEBRH, VRX, VE2, la2, 0, 0, 0, vsteb, 0, IF_VEC) + F(0xe60b, VSTEBRF, VRX, VE2, la2, 0, 0, 0, vsteb, 0, IF_VEC) + F(0xe60a, VSTEBRG, VRX, VE2, la2, 0, 0, 0, vsteb, 0, IF_VEC) +/* VECTOR STORE BYTE REVERSED ELEMENTS */ + F(0xe60e, VSTBR, VRX, VE2, la2, 0, 0, 0, vstbr, 0, IF_VEC) /* VECTOR STORE ELEMENT */ E(0xe708, VSTEB, VRX, V, la2, 0, 0, 0, vste, 0, ES_8, IF_VEC) E(0xe709, VSTEH, VRX, V, la2, 0, 0, 0, vste, 0, ES_16, IF_VEC)
target/s390x/tcg/translate_vx.c.inc +161 -0
173 lines not on this page · raw message
Signature · 1 line
2.34.1
PATCH v2 7/7 tests/tcg/s390x: Tests for Vector Enhancements Facility 2
In reply to David Miller* tests/tcg/s390x/vxeh2_vcvt.c : vector convert
* tests/tcg/s390x/vxeh2_vs.c : vector shift
* tests/tcg/s390x/vxeh2_vlstr.c : vector load/store reversed
--- tests/tcg/s390x/Makefile.target | 8 ++ tests/tcg/s390x/vxeh2_vcvt.c | 97 +++++++++++++++++++++ tests/tcg/s390x/vxeh2_vlstr.c | 146 ++++++++++++++++++++++++++++++++ tests/tcg/s390x/vxeh2_vs.c | 91 ++++++++++++++++++++ 4 files changed, 342 insertions(+) create mode 100644 tests/tcg/s390x/vxeh2_vcvt.c create mode 100644 tests/tcg/s390x/vxeh2_vlstr.c create mode 100644 tests/tcg/s390x/vxeh2_vs.c
tests/tcg/s390x/Makefile.target +8 -0
@@ -16,6 +16,14 @@ TESTS+=shift TESTS+=trap TESTS+=signals-s390x +VECTOR_TESTS=vxeh2_vs +VECTOR_TESTS+=vxeh2_vcvt +VECTOR_TESTS+=vxeh2_vlstr + +TESTS+=$(VECTOR_TESTS) + +$(VECTOR_TESTS): CFLAGS+=-march=z15 -O2 + ifneq ($(HAVE_GDB_BIN),) GDB_SCRIPT=$(SRC_PATH)/tests/guest-debug/run-test.py
tests/tcg/s390x/vxeh2_vcvt.c new file +97 -0
97 lines not on this page · raw message
tests/tcg/s390x/vxeh2_vlstr.c new file +146 -0
146 lines not on this page · raw message
tests/tcg/s390x/vxeh2_vs.c new file +91 -0
91 lines not on this page · raw message
Signature · 1 line
2.34.1
Re: [PATCH v2 1/7] target/s390x: vxeh2: vector convert short/32b
In reply to David MillerQuoted 6 lines · David Miller
Signed-off-by David Miller<[email protected]>--- target/s390x/helper.h | 4 +++ target/s390x/tcg/translate_vx.c.inc | 44 ++++++++++++++++++++++++++--- target/s390x/tcg/vec_fpu_helper.c | 31 ++++++++++++++++++++ 3 files changed, 75 insertions(+), 4 deletions(-)
r~
Re: [PATCH v2 2/7] target/s390x: vxeh2: vector string search
In reply to David Millertarget/s390x/tcg/translate_vx.c.inc static DisasJumpType op_vstrc(DisasContext *s, DisasOps *o) · line 2508
Quoted 9 lines · David Miller
target/s390x/tcg/translate_vx.c.inc +9
-0+static DisasJumpType op_vstrs(DisasContext *s, DisasOps *o) +{ + const uint8_t es = get_field(s, m5); + const uint32_t D = get_field(s, m6); + + if (es > ES_32) { + gen_program_exception(s, PGM_SPECIFICATION); + return DISAS_NORETURN; + }
Missed a check for bits of m6 other than zs must be zero.
target/s390x/tcg/translate_vx.c.inc static DisasJumpType op_vstrc(DisasContext *s, DisasOps *o) · line 2511
Quoted 3 lines
target/s390x/tcg/translate_vx.c.inc +3
-0+ gen_gvec_4_ptr(get_field(s, v1), get_field(s, v2), + get_field(s, v3), get_field(s, v4), + cpu_env, (D << 16) | es, gen_helper_vstrs);
Why << 16?
target/s390x/tcg/vec_string_helper.c void HELPER(gvec_vstrc_cc_rt##BITS)(void *v1, const void *v2, const void *v3, \ · line 477
Quoted 3 lines
target/s390x/tcg/vec_string_helper.c +3
-0+void HELPER (vstrs )(void *v1,const void *v2,const void *v3,void *v4, +CPUS390XState *env,uint32_t desc) { +const bool zs = (desc >>16 );
If this is meant to recover D, it won't. The value that you passed above is had via simd_data(desc). After that, you could get D via >> 16.
Quoted 1 line
+ const
uint8_tes = desc & 16;
I think you clearly meant & 15 here. Or maybe & 3, since ES <= 2? However!
I think you shouldn't actually pass these values via simd_data(). I think you should have 6 helper functions, which call this as an inline function passing ES and ZS as immediate arguments. This will allow the compiler to fold away all of the multiple checks vs ES and ZS.
target/s390x/tcg/vec_string_helper.c void HELPER(gvec_vstrc_cc_rt##BITS)(void *v1, const void *v2, const void *v3, \ · line 488
Quoted 8 lines
target/s390x/tcg/vec_string_helper.c +8
-0+uint32_t str_len =0 , eos =0 ; +uint32_t i =0 , j =0 , k =0 , cc =0 ; +uint32_t substr_len = ((uint8_t *)v4)[H1 (7 )] &31 ; + +for (i =0 ; i <16 ; i += char_size) { +if (0 == es && !((uint8_t *)v3)[H1 (i >> es)]) {break ; } +if (1 == es && !((uint16_t *)v3)[H2 (i >> es)]) {break ; } +if (2 == es && !((uint32_t *)v3)[H4 (i >> es)]) {break ; }
Instead of iterating to 16 by char_size and dividing by es, you should iterate to element_count by 1. You should use s390_vec_read_element here, and elsewhere.
target/s390x/tcg/vec_string_helper.c void HELPER(gvec_vstrc_cc_rt##BITS)(void *v1, const void *v2, const void *v3, \ · line 492
Quoted 4 lines
target/s390x/tcg/vec_string_helper.c +4
-0+ } +if (i < substr_len) { + substr_len = i; + }
This bounding of substr_len is only done when ZS is true.
target/s390x/tcg/vec_string_helper.c void HELPER(gvec_vstrc_cc_rt##BITS)(void *v1, const void *v2, const void *v3, \ · line 536
Quoted 2 lines
target/s390x/tcg/vec_string_helper.c +2
-0+ ((uint64_t *)v1)[0 ] = ((uint64_t *)v1)[1 ] =0 ; + ((uint8_t *)v1)[H1 (7 )] = k;
s390_vec_write_element64(v1, 0, k);s390_vec_write_element64(v1, 1, 0);
r~
Re: [PATCH v2 3/7] target/s390x: vxeh2: vector shift {double by bit, left, right {logical,arithmetic}}
In reply to David MillerQuoted 17 lines · David Miller
}include/qemu/bitops.h +15
-0+/** +* deposit8: +* @value: initial value to insert bit field into +* @start: the lowest bit in the bit field (numbered from 0) +* @length: the length of the bit field +* @fieldval: the value to insert into the bit field +* +* Deposit @fieldval into the 8 bit @value at the bit field specified +* by the @start and @length parameters, and return the modified +* @value. Bits of @value outside the bit field are not modified. +* Bits of @fieldval above the least significant @length bits are +* ignored. The bit field must lie entirely within the 8 bit byte. +* It is valid to request that all 8 bits are modified (ie @length +* 8 and @start 0). +*
include/qemu/bitops.h static inline int64_t sextract64(uint64_t value, int start, int length) · line 472
Quoted 10 lines
include/qemu/bitops.h +10
-0+* Returns: the modified @value. +*/ +static inline uint8_t deposit8 (uint8_t value,int start,int length, +uint8_t fieldval) +{ +uint8_t mask; +assert (start >=0 && length >0 && length <=8 - start); + mask = (~0ULL >> (8 - length)) << start; +return (value & ~mask) | ((fieldval << start) & mask); +}
(1) must be a separate patch.
(2) watch the whitespace at the top.
Given we have extract8 already, this is indeed missing. But I'm surprised you'd need this...
Also, this is still doing too much.
Changes to existing instructions should not be mixed with new instructions.
target/s390x/tcg/translate_vx.c.inc static DisasJumpType op_ves(DisasContext *s, DisasOps *o) · line 2023
Quoted 8 lines
static DisasJumpType
op_vsl(DisasContext *s, DisasOps *o){target/s390x/tcg/translate_vx.c.inc +1
-5- TCGv_i64 shift = tcg_temp_new_i64(); - - read_vec_element_i64(shift, get_field(s, v3), 7, ES_8); - if (s->fields.op2 == 0x74) { - tcg_gen_andi_i64(shift, shift, 0x7); + const bool B = 0x75 == s->fields.op2;
I really don't like testing opcodes after the fact. This is the job for insn-data.def. Either pass in data with the DATA element of F(), or use a helper function.
Quoted 18 lines
target/s390x/tcg/translate_vx.c.inc +10
-3+static DisasJumpType op_vsld(DisasContext *s, DisasOps *o) { - const uint8_t i4 = get_field(s, i4) & 0xf; - const int left_shift = (i4 & 7) * 8; - const int right_shift = 64 - left_shift; + const uint8_t mask = (0x86 == s->fields.op2) ? 7 : 15; + const uint8_t mul = (0x86 == s->fields.op2) ? 1 : 8; + const uint8_t i4 = get_field(s, i4); + const int shift = 64 - (i4 & 7) * mul; + + if (i4 & ~mask) { + gen_program_exception(s, PGM_SPECIFICATION); + return DISAS_NORETURN; + } TCGv_i64 t0 = tcg_temp_new_i64(); TCGv_i64 t1 = tcg_temp_new_i64(); TCGv_i64 t2 = tcg_temp_new_i64();
target/s390x/tcg/translate_vx.c.inc static DisasJumpType op_vsldb(DisasContext *s, DisasOps *o) · line 2064
Quoted 7 lines
target/s390x/tcg/translate_vx.c.inc +2
-2@@ -2053,8 +2060,8 @@ static DisasJumpType op_vsldb(DisasContext *s, DisasOps *o) read_vec_element_i64(t1, get_field(s, v3), 0, ES_64); read_vec_element_i64(t2, get_field(s, v3), 1, ES_64); } - tcg_gen_extract2_i64(t0, t1, t0, right_shift); - tcg_gen_extract2_i64(t1, t2, t1, right_shift); + tcg_gen_extract2_i64(t0, t1, t0, shift); + tcg_gen_extract2_i64(t1, t2, t1, shift);
The renaming of right_shift to shift is probably misleading, since extract2 *always* performs a right-shift.
target/s390x/tcg/translate_vx.c.inc static DisasJumpType op_vsldb(DisasContext *s, DisasOps *o) · line 2091
Quoted 2 lines
target/s390x/tcg/translate_vx.c.inc +2
-0+ tcg_gen_extract2_i64(t0, t1, t0, left_shift); + tcg_gen_extract2_i64(t1, t2, t1, left_shift);
Which makes this bit from op_vsrd actively misleading (though the code appears to be correct, its just the variable name that's wrong).
Quoted 5 lines
target/s390x/tcg/vec_int_helper.c +5
-0+void HELPER (gvec_vsl_ve2 )(void *v1,const void *v2,const void *v3, +uint32_t desc) +{ +uint8_t i, v; +S390Vector tmp = {};
target/s390x/tcg/vec_int_helper.c void HELPER(gvec_vsl)(void *v1, const void *v2, uint64_t count, · line 557
Quoted 10 lines
target/s390x/tcg/vec_int_helper.c +10
-0+for (i =0 ; i <16 ; i++) { +const uint8_t shift =s390_vec_read_element8 (v3, i) &7 ; + v =s390_vec_read_element8 (v2, i); + +if (shift) { + v <<= shift; +if (i <15 ) { + v |=extract8 (s390_vec_read_element8 (v2, i +1 ), +8 - shift, shift); + }
Possibly better as
if (shift) {
uint16_t tmp = (uint16_t)v << 8;
if (i < 15) {
tmp |= s390_vec_read_element8(v2, i + 1);
}
tmp <<= shift;
v = tmp >> 8;
}Similarly for the right shifts.
I wonder if it's worth checking that the values are identical, so that we can use the original vsl implementation, using double-word shifts. E.g.
uint64_t v3_0 = s390_vec_read_element64(v3, 0);
uint64_t v3_1 = s390_vec_read_element64(v3, 1);
uint64_t sh_0 = dup_const(MO_8, v3_0 & 7);
uint64_t sh_m = dup_const(MO_8, 7);
if ((v3_0 & sh_m) == sh_0 && (v3_1 & sh_m) == sh_0) {
helper_gvec_vsrl(v1, v2, v3, desc);
return;
}r~
Re: [PATCH v2 6/7] target/s390x: add S390_FEAT_VECTOR_ENH2 to cpu max
In reply to David MillerQuoted 4 lines · David Miller
Signed-off-by David Miller<[email protected]>--- target/s390x/gen-features.c | 2 ++ 1 file changed, 2 insertions(+)
r~
Re: [PATCH v2 1/7] target/s390x: vxeh2: vector convert short/32b
In reply to David MillerQuoted 7 lines · David Miller
Signed-off-by David Miller <[email protected]>--- target/s390x/helper.h | 4 +++ target/s390x/tcg/translate_vx.c.inc | 44 ++++++++++++++++++++++++++--- target/s390x/tcg/vec_fpu_helper.c | 31 ++++++++++++++++++++ 3 files changed, 75 insertions(+), 4 deletions(-)
Signature · 3 lines
Thanks, David / dhildenb