s390x/tcg: Vector Instruction Support Part 2
v3 cover s390x/tcg: Vector Instruction Support Part 2
This is the second part of vector instruction support for s390x. It is based on the series by Richard:
[PATCH v2 00/29] tcg vector improvements
Part 1: Vector Support Instructions
Part 2: Vector Integer Instructions
Part 3: Vector String Instructions
Part 4: Vector Floating-Point Instructions
The current state can be found at (kept updated):
https://github.com/davidhildenbrand/qemu/tree/vx
With the current state I can boot Linux kernel + user space compiled with SIMD support. This allows to boot distributions compiled exclusively for z13, requiring SIMD support. Also, it is now possible to build a complete kernel using rpmbuild as quite some issues have been sorted out.
In this part, all Vector Integer Instructions introduced with the "Vector Facility" are added. Some instructions part of the Vector Extension Facilities are also added.
v2 -> v3:
- Use new vector helpers tcg_gen_gvec_abs(), tcg_gen_gvec_shl() and friends and tcg_gen_gvec_shlv() and friends.
v1 -> v2:
- Many suggestions and fixes as outlined by Richard H.
-- Mostly avoid oom helpers for 128-bit add/sub variants
-- New vector helpers are not considered for now
David Hildenbrand (40):
s390x/tcg: Implement VECTOR ADD
s390x/tcg: Implement VECTOR ADD COMPUTE CARRY
s390x/tcg: Implement VECTOR ADD WITH CARRY
s390x/tcg: Implement VECTOR ADD WITH CARRY COMPUTE CARRY s390x/tcg: Implement VECTOR AND (WITH COMPLEMENT) s390x/tcg: Implement VECTOR AVERAGE
s390x/tcg: Implement VECTOR AVERAGE LOGICAL
s390x/tcg: Implement VECTOR CHECKSUM
s390x/tcg: Implement VECTOR ELEMENT COMPARE *
s390x/tcg: Implement VECTOR COMPARE *
s390x/tcg: Implement VECTOR COUNT LEADING ZEROS
s390x/tcg: Implement VECTOR COUNT TRAILING ZEROS s390x/tcg: Implement VECTOR EXCLUSIVE OR
s390x/tcg: Implement VECTOR GALOIS FIELD MULTIPLY SUM (AND ACCUMULATE) s390x/tcg: Implement VECTOR LOAD COMPLEMENT
s390x/tcg: Implement VECTOR LOAD POSITIVE
s390x/tcg: Implement VECTOR (MAXIMUM|MINIMUM) (LOGICAL) s390x/tcg: Implement VECTOR MULTIPLY AND ADD *
s390x/tcg: Implement VECTOR MULTIPLY *
s390x/tcg: Implement VECTOR NAND
s390x/tcg: Implement VECTOR NOR
s390x/tcg: Implement VECTOR NOT EXCLUSIVE OR
s390x/tcg: Implement VECTOR OR
s390x/tcg: Implement VECTOR OR WITH COMPLEMENT
s390x/tcg: Implement VECTOR POPULATION COUNT
s390x/tcg: Implement VECTOR ELEMENT ROTATE LEFT LOGICAL s390x/tcg: Implement VECTOR ELEMENT ROTATE AND INSERT UNDER MASK s390x/tcg: Implement VECTOR ELEMENT SHIFT
s390x/tcg: Implement VECTOR SHIFT LEFT (BY BYTE) s390x/tcg: Implement VECTOR SHIFT LEFT DOUBLE BY BYTE s390x/tcg: Implement VECTOR SHIFT RIGHT ARITHMETIC s390x/tcg: Implement VECTOR SHIFT RIGHT LOGICAL * s390x/tcg: Implement VECTOR SUBTRACT
s390x/tcg: Implement VECTOR SUBTRACT COMPUTE BORROW INDICATION s390x/tcg: Implement VECTOR SUBTRACT WITH BORROW INDICATION s390x/tcg: Implement VECTOR SUBTRACT WITH BORROW COMPUTE BORROW
INDICATION
s390x/tcg: Implement VECTOR SUM ACROSS DOUBLEWORD s390x/tcg: Implement VECTOR SUM ACROSS QUADWORD
s390x/tcg: Implement VECTOR SUM ACROSS WORD
s390x/tcg: Implement VECTOR TEST UNDER MASK
target/s390x/Makefile.objs | 2 +- target/s390x/cc_helper.c | 17 + target/s390x/helper.c | 1 + target/s390x/helper.h | 66 ++ target/s390x/insn-data.def | 137 +++ target/s390x/internal.h | 1 + target/s390x/translate.c | 2 + target/s390x/translate_vx.inc.c | 1416 +++++++++++++++++++++++++++++++ target/s390x/vec_int_helper.c | 603 +++++++++++++ 9 files changed, 2244 insertions(+), 1 deletion(-) create mode 100644 target/s390x/vec_int_helper.c
Signature · 1 line
2.20.1
PATCH v3 1/40 s390x/tcg: Implement VECTOR ADD
Introduce two types of fancy new helpers that will be reused a couple of times
1. gen_gvec_fn_3: Call an existing tcg_gen_gvec_X function with 3 parameters, simplifying parameter passing
2. gen_gvec128_3_i64: Call a function that performs 128 bit calculations using two 64 bit values per vector.
Luckily, for VECTOR ADD we already have everything we need.
--- target/s390x/insn-data.def | 5 ++++ target/s390x/translate_vx.inc.c | 52 +++++++++++++++++++++++++++++++++ 2 files changed, 57 insertions(+)
target/s390x/insn-data.def +5 -0
@@ -1054,6 +1054,11 @@ /* VECTOR UNPACK LOGICAL LOW */ F(0xe7d4, VUPLL, VRR_a, V, 0, 0, 0, 0, vup, 0, IF_VEC) +/* === Vector Integer Instructions === */ + +/* VECTOR ADD */ + F(0xe7f3, VA, VRR_c, V, 0, 0, 0, 0, va, 0, IF_VEC) + #ifndef CONFIG_USER_ONLY /* COMPARE AND SWAP AND PURGE */ E(0xb250, CSP, RRE, Z, r1_32u, ra2, r1_P, 0, csp, 0, MO_TEUL, IF_PRIV)
target/s390x/translate_vx.inc.c +52 -0
61 lines not on this page · raw message
Signature · 1 line
2.20.1
PATCH v3 2/40 s390x/tcg: Implement VECTOR ADD COMPUTE CARRY
In reply to David Hildenbrand128-bit handling courtesy of Richard H.
--- target/s390x/insn-data.def | 2 + target/s390x/translate_vx.inc.c | 94 +++++++++++++++++++++++++++++++++ 2 files changed, 96 insertions(+)
target/s390x/insn-data.def +2 -0
@@ -1058,6 +1058,8 @@ /* VECTOR ADD */ F(0xe7f3, VA, VRR_c, V, 0, 0, 0, 0, va, 0, IF_VEC) +/* VECTOR ADD COMPUTE CARRY */ + F(0xe7f1, VACC, VRR_c, V, 0, 0, 0, 0, vacc, 0, IF_VEC) #ifndef CONFIG_USER_ONLY /* COMPARE AND SWAP AND PURGE */
target/s390x/translate_vx.inc.c +94 -0
103 lines not on this page · raw message
Signature · 1 line
2.20.1
Re: [Qemu-devel] [PATCH v3 02/40] s390x/tcg: Implement VECTOR ADD COMPUTE CARRY
Quoted 7 lines · David Hildenbrand
128-bit handling courtesy of Richard H.
Signed-off-by David Hildenbrand <[email protected]>--- target/s390x/insn-data.def | 2 + target/s390x/translate_vx.inc.c | 94 +++++++++++++++++++++++++++++++++ 2 files changed, 96 insertions(+)
Quoted 16 lines
target/s390x/translate_vx.inc.c +16
-0+static void gen_acc (TCGv_i64 d,TCGv_i64 a,TCGv_i64 b,uint8_t es) +{ +const uint8_t msb_bit_nr =NUM_VEC_ELEMENT_BITS (es) -1 ; +TCGv_i64 msb_mask =tcg_const_i64 (dup_const (es,1ull << msb_bit_nr)); +TCGv_i64 t1 =tcg_temp_new_i64 (); +TCGv_i64 t2 =tcg_temp_new_i64 (); +TCGv_i64 t3 =tcg_temp_new_i64 (); + +/* Calculate the carry into the MSB, ignoring the old MSBs */ +tcg_gen_andc_i64 (t1, a, msb_mask); +tcg_gen_andc_i64 (t2, b, msb_mask); +tcg_gen_add_i64 (t1, t1, t2); +/* Calculate the MSB without any carry into it */ +tcg_gen_xor_i64 (t3, a, b); +/* Calculate the carry out of the MSB in the MSB bit position */ +tcg_gen_and_i64 (d, a, b);
target/s390x/translate_vx.inc.c static DisasJumpType op_va(DisasContext *s, DisasOps *o) · line 1017
Quoted 10 lines
target/s390x/translate_vx.inc.c +10
-0+tcg_gen_and_i64 (t1, t1, t3); +tcg_gen_or_i64 (d, d, t1); +/* Isolate and shift the carry into position */ +tcg_gen_and_i64 (d, d, msb_mask); +tcg_gen_shri_i64 (d, d, msb_bit_nr); + +tcg_temp_free_i64 (t1); +tcg_temp_free_i64 (t2); +tcg_temp_free_i64 (t3); +}
...> +static void gen_acc32_i64(TCGv_i64 d, TCGv_i64 a, TCGv_i64 b)
target/s390x/translate_vx.inc.c static DisasJumpType op_va(DisasContext *s, DisasOps *o) · line 1041
Quoted 12 lines
target/s390x/translate_vx.inc.c +12
-0+{ +gen_acc (d, a, b,ES_32 ); +} + +static void gen_acc_i64 (TCGv_i64 d,TCGv_i64 a,TCGv_i64 b) +{ +TCGv_i64 t =tcg_temp_new_i64 (); + +tcg_gen_add_i64 (t, a, b); +tcg_gen_setcond_i64 (TCG_COND_LTU , d, t, b); +tcg_temp_free_i64 (t); +}
As an aside, I think the 32-bit version should use 32-bit ops, as per gen_acc_i64. That would be 4 * 2 operations instead of 2 * 9 over the 128-bit vector.
r~
Re: [Qemu-devel] [PATCH v3 02/40] s390x/tcg: Implement VECTOR ADD COMPUTE CARRY
Quoted 38 lines · Richard Henderson
Quoted 7 lines · David Hildenbrand
128-bit handling courtesy of Richard H.
Signed-off-by David Hildenbrand <[email protected]>--- target/s390x/insn-data.def | 2 + target/s390x/translate_vx.inc.c | 94 +++++++++++++++++++++++++++++++++ 2 files changed, 96 insertions(+)Reviewed-by Richard Henderson <[email protected]>Quoted 16 lines
target/s390x/translate_vx.inc.c +16
-0+static void gen_acc (TCGv_i64 d,TCGv_i64 a,TCGv_i64 b,uint8_t es) +{ +const uint8_t msb_bit_nr =NUM_VEC_ELEMENT_BITS (es) -1 ; +TCGv_i64 msb_mask =tcg_const_i64 (dup_const (es,1ull << msb_bit_nr)); +TCGv_i64 t1 =tcg_temp_new_i64 (); +TCGv_i64 t2 =tcg_temp_new_i64 (); +TCGv_i64 t3 =tcg_temp_new_i64 (); + +/* Calculate the carry into the MSB, ignoring the old MSBs */ +tcg_gen_andc_i64 (t1, a, msb_mask); +tcg_gen_andc_i64 (t2, b, msb_mask); +tcg_gen_add_i64 (t1, t1, t2); +/* Calculate the MSB without any carry into it */ +tcg_gen_xor_i64 (t3, a, b); +/* Calculate the carry out of the MSB in the MSB bit position */ +tcg_gen_and_i64 (d, a, b);target/s390x/translate_vx.inc.c static DisasJumpType op_va(DisasContext *s, DisasOps *o) · line 1017
Quoted 10 lines
target/s390x/translate_vx.inc.c +10
-0+tcg_gen_and_i64 (t1, t1, t3); +tcg_gen_or_i64 (d, d, t1); +/* Isolate and shift the carry into position */ +tcg_gen_and_i64 (d, d, msb_mask); +tcg_gen_shri_i64 (d, d, msb_bit_nr); + +tcg_temp_free_i64 (t1); +tcg_temp_free_i64 (t2); +tcg_temp_free_i64 (t3); +}...> +static void
gen_acc32_i64(TCGv_i64 d, TCGv_i64 a, TCGv_i64 b)
Quoted 16 lines
target/s390x/translate_vx.inc.c static DisasJumpType op_va(DisasContext *s, DisasOps *o) · line 1041
Quoted 12 lines
target/s390x/translate_vx.inc.c +12
-0+{ +gen_acc (d, a, b,ES_32 ); +} + +static void gen_acc_i64 (TCGv_i64 d,TCGv_i64 a,TCGv_i64 b) +{ +TCGv_i64 t =tcg_temp_new_i64 (); + +tcg_gen_add_i64 (t, a, b); +tcg_gen_setcond_i64 (TCG_COND_LTU , d, t, b); +tcg_temp_free_i64 (t); +}As an aside, I think the 32-bit version should use 32-bit ops, as per
gen_acc_i64. That would be 4 * 2 operations instead of 2 * 9 over the 128-bit vector.
Makes sense, thanks!
Quoted 4 lines
r~
Signature · 3 lines
Thanks, David / dhildenb
PATCH v3 3/40 s390x/tcg: Implement VECTOR ADD WITH CARRY
In reply to David HildenbrandOnly slightly ugly, perform two additions. At least it is only supported for 128 bit elements.
Introduce gen_gvec128_4_i64() similar to gen_gvec128_3_i64().
--- target/s390x/insn-data.def | 2 ++ target/s390x/translate_vx.inc.c | 63 +++++++++++++++++++++++++++++++++ 2 files changed, 65 insertions(+)
target/s390x/insn-data.def +2 -0
@@ -1060,6 +1060,8 @@ F(0xe7f3, VA, VRR_c, V, 0, 0, 0, 0, va, 0, IF_VEC) /* VECTOR ADD COMPUTE CARRY */ F(0xe7f1, VACC, VRR_c, V, 0, 0, 0, 0, vacc, 0, IF_VEC) +/* VECTOR ADD WITH CARRY */ + F(0xe7bb, VAC, VRR_d, V, 0, 0, 0, 0, vac, 0, IF_VEC) #ifndef CONFIG_USER_ONLY /* COMPARE AND SWAP AND PURGE */
target/s390x/translate_vx.inc.c +63 -0
72 lines not on this page · raw message
Signature · 1 line
2.20.1
PATCH v3 4/40 s390x/tcg: Implement VECTOR ADD WITH CARRY COMPUTE CARRY
In reply to David HildenbrandSimilar to VECTOR ADD COMPUTE CARRY, however 128-bit handling only. Courtesy of Richard H.
--- target/s390x/insn-data.def | 2 ++ target/s390x/translate_vx.inc.c | 32 ++++++++++++++++++++++++++++++++ 2 files changed, 34 insertions(+)
target/s390x/insn-data.def +2 -0
@@ -1062,6 +1062,8 @@ F(0xe7f1, VACC, VRR_c, V, 0, 0, 0, 0, vacc, 0, IF_VEC) /* VECTOR ADD WITH CARRY */ F(0xe7bb, VAC, VRR_d, V, 0, 0, 0, 0, vac, 0, IF_VEC) +/* VECTOR ADD WITH CARRY COMPUTE CARRY */ + F(0xe7b9, VACCC, VRR_d, V, 0, 0, 0, 0, vaccc, 0, IF_VEC) #ifndef CONFIG_USER_ONLY /* COMPARE AND SWAP AND PURGE */
target/s390x/translate_vx.inc.c +32 -0
@@ -1142,3 +1142,35 @@ static DisasJumpType op_vac(DisasContext *s, DisasOps *o) get_field(s->fields, v4));return DISAS_NEXT ; } + +static void gen_accc2_i64 (TCGv_i64 dl,TCGv_i64 dh,TCGv_i64 al,TCGv_i64 ah, +TCGv_i64 bl,TCGv_i64 bh,TCGv_i64 cl,TCGv_i64 ch) +{ +TCGv_i64 tl =tcg_temp_new_i64 (); +TCGv_i64 th =tcg_temp_new_i64 (); +TCGv_i64 zero =tcg_const_i64 (0 ); + +tcg_gen_andi_i64 (tl, cl,1 ); +tcg_gen_add2_i64 (tl, th, tl, zero, al, zero); +tcg_gen_add2_i64 (tl, th, tl, th, bl, zero); +tcg_gen_add2_i64 (tl, th, th, zero, ah, zero); +tcg_gen_add2_i64 (tl, dl, tl, th, bh, zero); +tcg_gen_mov_i64 (dh, zero); + +tcg_temp_free_i64 (tl); +tcg_temp_free_i64 (th); +tcg_temp_free_i64 (zero); +} + +static DisasJumpType op_vaccc (DisasContext *s,DisasOps *o) +{ +if (get_field (s->fields, m5) !=ES_128 ) { +gen_program_exception (s,PGM_SPECIFICATION ); +return DISAS_NORETURN ; + } + +gen_gvec128_4_i64 (gen_accc2_i64,get_field (s->fields, v1), +get_field (s->fields, v2),get_field (s->fields, v3), +get_field (s->fields, v4)); +return DISAS_NEXT ; +}
Signature · 1 line
2.20.1
Re: [Qemu-devel] [PATCH v3 04/40] s390x/tcg: Implement VECTOR ADD WITH CARRY COMPUTE CARRY
Quoted 8 lines · David Hildenbrand
Similar to VECTOR ADD COMPUTE CARRY, however 128-bit handling only. Courtesy of Richard H.
Signed-off-by David Hildenbrand <[email protected]>--- target/s390x/insn-data.def | 2 ++ target/s390x/translate_vx.inc.c | 32 ++++++++++++++++++++++++++++++++ 2 files changed, 34 insertions(+)
r~
PATCH v3 5/40 s390x/tcg: Implement VECTOR AND (WITH COMPLEMENT)
In reply to David HildenbrandEasy, as we can reuse existing gvec helpers.
--- target/s390x/insn-data.def | 4 ++++ target/s390x/translate_vx.inc.c | 14 ++++++++++++++ 2 files changed, 18 insertions(+)
target/s390x/insn-data.def +4 -0
@@ -1064,6 +1064,10 @@ F(0xe7bb, VAC, VRR_d, V, 0, 0, 0, 0, vac, 0, IF_VEC) /* VECTOR ADD WITH CARRY COMPUTE CARRY */ F(0xe7b9, VACCC, VRR_d, V, 0, 0, 0, 0, vaccc, 0, IF_VEC) +/* VECTOR AND */ + F(0xe768, VN, VRR_c, V, 0, 0, 0, 0, vn, 0, IF_VEC) +/* VECTOR AND WITH COMPLEMENT */ + F(0xe769, VNC, VRR_c, V, 0, 0, 0, 0, vnc, 0, IF_VEC) #ifndef CONFIG_USER_ONLY /* COMPARE AND SWAP AND PURGE */
target/s390x/translate_vx.inc.c +14 -0
@@ -1174,3 +1174,17 @@ static DisasJumpType op_vaccc(DisasContext *s, DisasOps *o) get_field(s->fields, v4));return DISAS_NEXT ; } + +static DisasJumpType op_vn (DisasContext *s,DisasOps *o) +{ +gen_gvec_fn_3 (and,ES_8 ,get_field (s->fields, v1),get_field (s->fields, v2), +get_field (s->fields, v3)); +return DISAS_NEXT ; +} + +static DisasJumpType op_vnc (DisasContext *s,DisasOps *o) +{ +gen_gvec_fn_3 (andc,ES_8 ,get_field (s->fields, v1), +get_field (s->fields, v2),get_field (s->fields, v3)); +return DISAS_NEXT ; +}
Signature · 1 line
2.20.1
PATCH v3 6/40 s390x/tcg: Implement VECTOR AVERAGE
In reply to David HildenbrandHandle 32/64-bit elements via gvec expansion and the 8/16 bits via ool helpers.
--- target/s390x/Makefile.objs | 2 +- target/s390x/helper.h | 4 +++ target/s390x/insn-data.def | 2 ++ target/s390x/translate_vx.inc.c | 64 +++++++++++++++++++++++++++++++++ target/s390x/vec_int_helper.c | 32 +++++++++++++++++ 5 files changed, 103 insertions(+), 1 deletion(-) create mode 100644 target/s390x/vec_int_helper.c
target/s390x/Makefile.objs +1 -1
@@ -1,7 +1,7 @@ obj-y += cpu.o cpu_models.o cpu_features.o gdbstub.o interrupt.o helper.o obj-$(CONFIG_TCG) += translate.o cc_helper.o excp_helper.o fpu_helper.o obj-$(CONFIG_TCG) += int_helper.o mem_helper.o misc_helper.o crypto_helper.o -obj-$(CONFIG_TCG) += vec_helper.o +obj-$(CONFIG_TCG) += vec_helper.o vec_int_helper.o obj-$(CONFIG_SOFTMMU) += machine.o ioinst.o arch_dump.o mmu_helper.o diag.o obj-$(CONFIG_SOFTMMU) += sigp.o obj-$(CONFIG_KVM) += kvm.o
target/s390x/helper.h +4 -0
@@ -145,6 +145,10 @@ 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) +DEF_HELPER_FLAGS_4(gvec_vavg16, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, i32) +#ifndef CONFIG_USER_ONLY DEF_HELPER_3(servc, i32, env, i64, i64)DEF_HELPER_4 (diag, void, env, i32, i32, i32)
target/s390x/insn-data.def +2 -0
@@ -1068,6 +1068,8 @@ F(0xe768, VN, VRR_c, V, 0, 0, 0, 0, vn, 0, IF_VEC) /* VECTOR AND WITH COMPLEMENT */ F(0xe769, VNC, VRR_c, V, 0, 0, 0, 0, vnc, 0, IF_VEC) +/* VECTOR AVERAGE */ + F(0xe7f2, VAVG, VRR_c, V, 0, 0, 0, 0, vavg, 0, IF_VEC) #ifndef CONFIG_USER_ONLY /* COMPARE AND SWAP AND PURGE */
target/s390x/translate_vx.inc.c +64 -0
73 lines not on this page · raw message
target/s390x/vec_int_helper.c new file +32 -0
@@ -0,0 +1,32 @@ +/* +* QEMU TCG support -- s390x vector integer instruction support +* +* Copyright (C) 2019 Red Hat Inc +* +* Authors: +* David Hildenbrand <[email protected]> +* +* This work is licensed under the terms of the GNU GPL, version 2 or later. +* See the COPYING file in the top-level directory. +*/ +#include "qemu/osdep.h" +#include "qemu-common.h" +#include "cpu.h" +#include "vec.h" +#include "exec/helper-proto.h" + +#define DEF_VAVG (BITS ) \ +void HELPER(gvec_vavg##BITS)(void *v1, const void *v2, const void *v3, \ + uint32_t desc) \ +{ \ + int i; \ + \ + for (i = 0; i < (128 / BITS); i++) { \ + const int32_t a = (int##BITS##_t)s390_vec_read_element##BITS(v2, i); \ + const int32_t b = (int##BITS##_t)s390_vec_read_element##BITS(v3, i); \ + \ + s390_vec_write_element##BITS(v1, i, (a + b + 1) >> 1); \ + } \ +} +DEF_VAVG(8) +DEF_VAVG (16 )
Signature · 1 line
2.20.1
PATCH v3 7/40 s390x/tcg: Implement VECTOR AVERAGE LOGICAL
In reply to David HildenbrandSimilar to VECTOR AVERAGE but without sign extension.
--- target/s390x/helper.h | 2 ++ target/s390x/insn-data.def | 2 ++ target/s390x/translate_vx.inc.c | 48 +++++++++++++++++++++++++++++++++ target/s390x/vec_int_helper.c | 16 +++++++++++ 4 files changed, 68 insertions(+)
target/s390x/helper.h +2 -0
@@ -148,6 +148,8 @@ 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) DEF_HELPER_FLAGS_4(gvec_vavg16, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, i32) +DEF_HELPER_FLAGS_4(gvec_vavgl8, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, i32) +DEF_HELPER_FLAGS_4(gvec_vavgl16, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, i32)#ifndef CONFIG_USER_ONLY DEF_HELPER_3 (servc, i32, env, i64, i64)
target/s390x/insn-data.def +2 -0
@@ -1070,6 +1070,8 @@ F(0xe769, VNC, VRR_c, V, 0, 0, 0, 0, vnc, 0, IF_VEC) /* VECTOR AVERAGE */ F(0xe7f2, VAVG, VRR_c, V, 0, 0, 0, 0, vavg, 0, IF_VEC) +/* VECTOR AVERAGE LOGICAL */ + F(0xe7f0, VAVGL, VRR_c, V, 0, 0, 0, 0, vavgl, 0, IF_VEC) #ifndef CONFIG_USER_ONLY /* COMPARE AND SWAP AND PURGE */
target/s390x/translate_vx.inc.c +48 -0
51 lines not on this page · raw message
target/s390x/vec_int_helper.c +16 -0
@@ -30,3 +30,19 @@ void HELPER(gvec_vavg##BITS)(void *v1, const void *v2, const void *v3, \ } DEF_VAVG(8) DEF_VAVG(16) + +#define DEF_VAVGL (BITS ) \ +void HELPER(gvec_vavgl##BITS)(void *v1, const void *v2, const void *v3, \ + uint32_t desc) \ +{ \ + int i; \ + \ + for (i = 0; i < (128 / BITS); i++) { \ + const uint##BITS##_t a = s390_vec_read_element##BITS(v2, i); \ + const uint##BITS##_t b = s390_vec_read_element##BITS(v3, i); \ + \ + s390_vec_write_element##BITS(v1, i, (a + b + 1) >> 1); \ + } \ +} +DEF_VAVGL(8) +DEF_VAVGL (16 )
Signature · 1 line
2.20.1
PATCH v3 8/40 s390x/tcg: Implement VECTOR CHECKSUM
In reply to David HildenbrandTime to introduce read_vec_element_i32 and write_vec_element_i32. Take proper care of properly adding the carry. We can perform both additions including the carry via tcg_gen_add2_i32().
--- target/s390x/insn-data.def | 2 + target/s390x/translate_vx.inc.c | 65 +++++++++++++++++++++++++++++++++ 2 files changed, 67 insertions(+)
target/s390x/insn-data.def +2 -0
@@ -1072,6 +1072,8 @@ F(0xe7f2, VAVG, VRR_c, V, 0, 0, 0, 0, vavg, 0, IF_VEC) /* VECTOR AVERAGE LOGICAL */ F(0xe7f0, VAVGL, VRR_c, V, 0, 0, 0, 0, vavgl, 0, IF_VEC) +/* VECTOR CHECKSUM */ + F(0xe766, VCKSM, VRR_c, V, 0, 0, 0, 0, vcksm, 0, IF_VEC) #ifndef CONFIG_USER_ONLY /* COMPARE AND SWAP AND PURGE */
target/s390x/translate_vx.inc.c +65 -0
80 lines not on this page · raw message
Signature · 1 line
2.20.1
PATCH v3 9/40 s390x/tcg: Implement VECTOR ELEMENT COMPARE *
In reply to David HildenbrandFairly easy to implement, we can make use of the existing CC helpers cmps64 and cmpu64 - we siply have to sign extend the elements.
--- target/s390x/insn-data.def | 4 ++++ target/s390x/translate_vx.inc.c | 20 ++++++++++++++++++++ 2 files changed, 24 insertions(+)
target/s390x/insn-data.def +4 -0
@@ -1074,6 +1074,10 @@ F(0xe7f0, VAVGL, VRR_c, V, 0, 0, 0, 0, vavgl, 0, IF_VEC) /* VECTOR CHECKSUM */ F(0xe766, VCKSM, VRR_c, V, 0, 0, 0, 0, vcksm, 0, IF_VEC) +/* VECTOR ELEMENT COMPARE */ + F(0xe7db, VEC, VRR_a, V, 0, 0, 0, 0, vec, cmps64, IF_VEC) +/* VECTOR ELEMENT COMPARE LOGICAL */ + F(0xe7d9, VECL, VRR_a, V, 0, 0, 0, 0, vec, cmpu64, IF_VEC) #ifndef CONFIG_USER_ONLY /* COMPARE AND SWAP AND PURGE */
target/s390x/translate_vx.inc.c +20 -0
@@ -1365,3 +1365,23 @@ static DisasJumpType op_vcksm(DisasContext *s, DisasOps *o)tcg_temp_free_i32 (sum);return DISAS_NEXT ; } + +static DisasJumpType op_vec (DisasContext *s,DisasOps *o) +{ +uint8_t es =get_field (s->fields, m3); +const uint8_t enr =NUM_VEC_ELEMENTS (es) /2 -1 ; + +if (es >ES_64 ) { +gen_program_exception (s,PGM_SPECIFICATION ); +return DISAS_NORETURN ; + } +if (s->fields->op2 ==0xdb ) { + es |=MO_SIGN ; + } + + o->in1 =tcg_temp_new_i64 (); + o->in2 =tcg_temp_new_i64 (); +read_vec_element_i64 (o->in1,get_field (s->fields, v1), enr, es); +read_vec_element_i64 (o->in2,get_field (s->fields, v2), enr, es); +return DISAS_NEXT ; +}
Signature · 1 line
2.20.1
PATCH v3 10/40 s390x/tcg: Implement VECTOR COMPARE *
In reply to David HildenbrandTo carry out the comparison, we can reuse the existing gvec comparison function. In case the CC is to be computed, save the result vector and compute the CC lazily. The result is a vector consisting of all 1's for elements that matched and 0's for elements that didn't match.
--- target/s390x/cc_helper.c | 17 +++++++++++++++++ target/s390x/helper.c | 1 + target/s390x/insn-data.def | 6 ++++++ target/s390x/internal.h | 1 + target/s390x/translate.c | 1 + target/s390x/translate_vx.inc.c | 28 ++++++++++++++++++++++++++++ 6 files changed, 54 insertions(+)
target/s390x/cc_helper.c +17 -0
@@ -402,6 +402,20 @@ static uint32_t cc_calc_lcbb(uint64_t dst)return dst ==16 ?0 :3 ; } +static uint32_t cc_calc_vc (uint64_t low,uint64_t high) +{ +if (high ==-1ull && low ==-1ull ) { +/* all elements match */ +return 0 ; + }else if (high ==0 && low ==0 ) { +/* no elements match */ +return 3 ; + }else { +/* some elements but not all match */ +return 1 ; + } +} +static uint32_t do_calc_cc (CPUS390XState *env,uint32_t cc_op,uint64_t src,uint64_t dst,uint64_t vr) {
@@ -514,6 +528,9 @@ static uint32_t do_calc_cc(CPUS390XState *env, uint32_t cc_op,case CC_OP_LCBB : r =cc_calc_lcbb (dst);break ; +case CC_OP_VC : + r =cc_calc_vc (src, dst); +break ;case CC_OP_NZ_F32 : r =set_cc_nz_f32 (dst);
target/s390x/helper.c +1 -0
@@ -418,6 +418,7 @@ const char *cc_name(enum cc_op cc_op) [CC_OP_SLA_64] = "CC_OP_SLA_64", [CC_OP_FLOGR] = "CC_OP_FLOGR", [CC_OP_LCBB] = "CC_OP_LCBB", + [CC_OP_VC] = "CC_OP_VC", };return cc_names[cc_op];
target/s390x/insn-data.def +6 -0
@@ -1078,6 +1078,12 @@ F(0xe7db, VEC, VRR_a, V, 0, 0, 0, 0, vec, cmps64, IF_VEC) /* VECTOR ELEMENT COMPARE LOGICAL */ F(0xe7d9, VECL, VRR_a, V, 0, 0, 0, 0, vec, cmpu64, IF_VEC) +/* VECTOR COMPARE EQUAL */ + E(0xe7f8, VCEQ, VRR_b, V, 0, 0, 0, 0, vc, 0, TCG_COND_EQ, IF_VEC) +/* VECTOR COMPARE HIGH */ + E(0xe7fb, VCH, VRR_b, V, 0, 0, 0, 0, vc, 0, TCG_COND_GT, IF_VEC) +/* VECTOR COMPARE HIGH LOGICAL */ + E(0xe7f9, VCHL, VRR_b, V, 0, 0, 0, 0, vc, 0, TCG_COND_GTU, IF_VEC) #ifndef CONFIG_USER_ONLY /* COMPARE AND SWAP AND PURGE */
target/s390x/internal.h +1 -0
@@ -200,6 +200,7 @@ enum cc_op {CC_OP_SLA_64 ,/* Calculate shift left signed (64bit) */ CC_OP_FLOGR ,/* find leftmost one */ CC_OP_LCBB ,/* load count to block boundary */ +CC_OP_VC ,/* vector compare result */ CC_OP_MAX };
target/s390x/translate.c +1 -0
@@ -572,6 +572,7 @@ static void gen_op_calc_cc(DisasContext *s)case CC_OP_SLA_32 :case CC_OP_SLA_64 :case CC_OP_NZ_F128 : +case CC_OP_VC :/* 2 arguments */ gen_helper_calc_cc (cc_op, cpu_env, local_cc_op, cc_src, cc_dst, dummy);break ;
target/s390x/translate_vx.inc.c +28 -0
@@ -1385,3 +1385,31 @@ static DisasJumpType op_vec(DisasContext *s, DisasOps *o)read_vec_element_i64 (o->in2,get_field (s->fields, v2), enr, es);return DISAS_NEXT ; } + +static DisasJumpType op_vc (DisasContext *s,DisasOps *o) +{ +const uint8_t es =get_field (s->fields, m4); +TCGCond cond = s->insn->data; + +if (es >ES_64 ) { +gen_program_exception (s,PGM_SPECIFICATION ); +return DISAS_NORETURN ; + } + +tcg_gen_gvec_cmp (cond, es, +vec_full_reg_offset (get_field (s->fields, v1)), +vec_full_reg_offset (get_field (s->fields, v2)), +vec_full_reg_offset (get_field (s->fields, v3)),16 ,16 ); +if (get_field (s->fields, m5) &0x1 ) { +TCGv_i64 low =tcg_temp_new_i64 (); +TCGv_i64 high =tcg_temp_new_i64 (); + +read_vec_element_i64 (high,get_field (s->fields, v1),0 ,ES_64 ); +read_vec_element_i64 (low,get_field (s->fields, v1),1 ,ES_64 ); +gen_op_update2_cc_i64 (s,CC_OP_VC , low, high); + +tcg_temp_free_i64 (low); +tcg_temp_free_i64 (high); + } +return DISAS_NEXT ; +}
Signature · 1 line
2.20.1
PATCH v3 11/40 s390x/tcg: Implement VECTOR COUNT LEADING ZEROS
In reply to David HildenbrandFor 8/16, use the 32 bit variant and properly subtract the added leading zero bits.
--- target/s390x/helper.h | 2 ++ target/s390x/insn-data.def | 2 ++ target/s390x/translate_vx.inc.c | 31 +++++++++++++++++++++++++++++++ target/s390x/vec_int_helper.c | 14 ++++++++++++++ 4 files changed, 49 insertions(+)
target/s390x/helper.h +2 -0
@@ -150,6 +150,8 @@ DEF_HELPER_FLAGS_4(gvec_vavg8, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, i32) DEF_HELPER_FLAGS_4(gvec_vavg16, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, i32) DEF_HELPER_FLAGS_4(gvec_vavgl8, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, i32) DEF_HELPER_FLAGS_4(gvec_vavgl16, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, i32) +DEF_HELPER_FLAGS_3(gvec_vclz8, TCG_CALL_NO_RWG, void, ptr, cptr, i32) +DEF_HELPER_FLAGS_3(gvec_vclz16, TCG_CALL_NO_RWG, void, ptr, cptr, i32)#ifndef CONFIG_USER_ONLY DEF_HELPER_3 (servc, i32, env, i64, i64)
target/s390x/insn-data.def +2 -0
@@ -1084,6 +1084,8 @@ E(0xe7fb, VCH, VRR_b, V, 0, 0, 0, 0, vc, 0, TCG_COND_GT, IF_VEC) /* VECTOR COMPARE HIGH LOGICAL */ E(0xe7f9, VCHL, VRR_b, V, 0, 0, 0, 0, vc, 0, TCG_COND_GTU, IF_VEC) +/* VECTOR COUNT LEADING ZEROS */ + F(0xe753, VCLZ, VRR_a, V, 0, 0, 0, 0, vclz, 0, IF_VEC) #ifndef CONFIG_USER_ONLY /* COMPARE AND SWAP AND PURGE */
target/s390x/translate_vx.inc.c +31 -0
@@ -182,6 +182,9 @@ static void get_vec_element_ptr_i64(TCGv_ptr ptr, uint8_t reg, TCGv_i64 enr,tcg_temp_free_i64 (tmp); } +#define gen_gvec_2 (v1, v2, gen) \ + tcg_gen_gvec_2(vec_full_reg_offset(v1), vec_full_reg_offset(v2), \ + 16, 16, gen)#define gen_gvec_3 (v1, v2, v3, gen) \ tcg_gen_gvec_3(vec_full_reg_offset(v1), vec_full_reg_offset(v2), \ vec_full_reg_offset(v3), 16, 16, gen)
@@ -1413,3 +1416,31 @@ static DisasJumpType op_vc(DisasContext *s, DisasOps *o) }return DISAS_NEXT ; } + +static void gen_clz_i32 (TCGv_i32 d,TCGv_i32 a) +{ +tcg_gen_clzi_i32 (d, a,32 ); +} + +static void gen_clz_i64 (TCGv_i64 d,TCGv_i64 a) +{ +tcg_gen_clzi_i64 (d, a,64 ); +} + +static DisasJumpType op_vclz (DisasContext *s,DisasOps *o) +{ +const uint8_t es =get_field (s->fields, m3); +static const GVecGen2 g[4 ] = { + { .fno = gen_helper_gvec_vclz8, }, + { .fno = gen_helper_gvec_vclz16, }, + { .fni4 = gen_clz_i32, }, + { .fni8 = gen_clz_i64, }, + }; + +if (es >ES_64 ) { +gen_program_exception (s,PGM_SPECIFICATION ); +return DISAS_NORETURN ; + } +gen_gvec_2 (get_field (s->fields, v1),get_field (s->fields, v2), &g[es]); +return DISAS_NEXT ; +}
target/s390x/vec_int_helper.c +14 -0
@@ -46,3 +46,17 @@ void HELPER(gvec_vavgl##BITS)(void *v1, const void *v2, const void *v3, \ } DEF_VAVGL(8) DEF_VAVGL(16) + +#define DEF_VCLZ (BITS ) \ +void HELPER(gvec_vclz##BITS)(void *v1, const void *v2, uint32_t desc) \ +{ \ + int i; \ + \ + for (i = 0; i < (128 / BITS); i++) { \ + const uint##BITS##_t a = s390_vec_read_element##BITS(v2, i); \ + \ + s390_vec_write_element##BITS(v1, i, clz32(a) - 32 + BITS); \ + } \ +} +DEF_VCLZ(8) +DEF_VCLZ (16 )
Signature · 1 line
2.20.1
PATCH v3 12/40 s390x/tcg: Implement VECTOR COUNT TRAILING ZEROS
In reply to David HildenbrandImplement it similar to VECTOR COUNT LEADING ZEROS.
--- target/s390x/helper.h | 2 ++ target/s390x/insn-data.def | 2 ++ target/s390x/translate_vx.inc.c | 28 ++++++++++++++++++++++++++++ target/s390x/vec_int_helper.c | 14 ++++++++++++++ 4 files changed, 46 insertions(+)
target/s390x/helper.h +2 -0
@@ -152,6 +152,8 @@ DEF_HELPER_FLAGS_4(gvec_vavgl8, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, i32) DEF_HELPER_FLAGS_4(gvec_vavgl16, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, i32) DEF_HELPER_FLAGS_3(gvec_vclz8, TCG_CALL_NO_RWG, void, ptr, cptr, i32) DEF_HELPER_FLAGS_3(gvec_vclz16, TCG_CALL_NO_RWG, void, ptr, cptr, i32) +DEF_HELPER_FLAGS_3(gvec_vctz8, TCG_CALL_NO_RWG, void, ptr, cptr, i32) +DEF_HELPER_FLAGS_3(gvec_vctz16, TCG_CALL_NO_RWG, void, ptr, cptr, i32)#ifndef CONFIG_USER_ONLY DEF_HELPER_3 (servc, i32, env, i64, i64)
target/s390x/insn-data.def +2 -0
@@ -1086,6 +1086,8 @@ E(0xe7f9, VCHL, VRR_b, V, 0, 0, 0, 0, vc, 0, TCG_COND_GTU, IF_VEC) /* VECTOR COUNT LEADING ZEROS */ F(0xe753, VCLZ, VRR_a, V, 0, 0, 0, 0, vclz, 0, IF_VEC) +/* VECTOR COUNT TRAILING ZEROS */ + F(0xe752, VCTZ, VRR_a, V, 0, 0, 0, 0, vctz, 0, IF_VEC) #ifndef CONFIG_USER_ONLY /* COMPARE AND SWAP AND PURGE */
target/s390x/translate_vx.inc.c +28 -0
@@ -1444,3 +1444,31 @@ static DisasJumpType op_vclz(DisasContext *s, DisasOps *o)gen_gvec_2 (get_field (s->fields, v1),get_field (s->fields, v2), &g[es]);return DISAS_NEXT ; } + +static void gen_ctz_i32 (TCGv_i32 d,TCGv_i32 a) +{ +tcg_gen_ctzi_i32 (d, a,32 ); +} + +static void gen_ctz_i64 (TCGv_i64 d,TCGv_i64 a) +{ +tcg_gen_ctzi_i64 (d, a,64 ); +} + +static DisasJumpType op_vctz (DisasContext *s,DisasOps *o) +{ +const uint8_t es =get_field (s->fields, m3); +static const GVecGen2 g[4 ] = { + { .fno = gen_helper_gvec_vctz8, }, + { .fno = gen_helper_gvec_vctz16, }, + { .fni4 = gen_ctz_i32, }, + { .fni8 = gen_ctz_i64, }, + }; + +if (es >ES_64 ) { +gen_program_exception (s,PGM_SPECIFICATION ); +return DISAS_NORETURN ; + } +gen_gvec_2 (get_field (s->fields, v1),get_field (s->fields, v2), &g[es]); +return DISAS_NEXT ; +}
target/s390x/vec_int_helper.c +14 -0
@@ -60,3 +60,17 @@ void HELPER(gvec_vclz##BITS)(void *v1, const void *v2, uint32_t desc) \ } DEF_VCLZ(8) DEF_VCLZ(16) + +#define DEF_VCTZ (BITS ) \ +void HELPER(gvec_vctz##BITS)(void *v1, const void *v2, uint32_t desc) \ +{ \ + int i; \ + \ + for (i = 0; i < (128 / BITS); i++) { \ + const uint##BITS##_t a = s390_vec_read_element##BITS(v2, i); \ + \ + s390_vec_write_element##BITS(v1, i, a ? ctz32(a) : BITS); \ + } \ +} +DEF_VCTZ(8) +DEF_VCTZ (16 )
Signature · 1 line
2.20.1
Re: [Qemu-devel] [PATCH v3 12/40] s390x/tcg: Implement VECTOR COUNT TRAILING ZEROS
Quoted 9 lines · David Hildenbrand
Implement it similar to VECTOR COUNT LEADING ZEROS.
Signed-off-by David Hildenbrand <[email protected]>--- target/s390x/helper.h | 2 ++ target/s390x/insn-data.def | 2 ++ target/s390x/translate_vx.inc.c | 28 ++++++++++++++++++++++++++++ target/s390x/vec_int_helper.c | 14 ++++++++++++++ 4 files changed, 46 insertions(+)
r~
PATCH v3 13/40 s390x/tcg: Implement VECTOR EXCLUSIVE OR
In reply to David HildenbrandEasy, we can reuse an existing gvec helper.
--- target/s390x/insn-data.def | 2 ++ target/s390x/translate_vx.inc.c | 7 +++++++ 2 files changed, 9 insertions(+)
target/s390x/insn-data.def +2 -0
@@ -1088,6 +1088,8 @@ F(0xe753, VCLZ, VRR_a, V, 0, 0, 0, 0, vclz, 0, IF_VEC) /* VECTOR COUNT TRAILING ZEROS */ F(0xe752, VCTZ, VRR_a, V, 0, 0, 0, 0, vctz, 0, IF_VEC) +/* VECTOR EXCLUSIVE OR */ + F(0xe76d, VX, VRR_c, V, 0, 0, 0, 0, vx, 0, IF_VEC) #ifndef CONFIG_USER_ONLY /* COMPARE AND SWAP AND PURGE */
target/s390x/translate_vx.inc.c +7 -0
@@ -1472,3 +1472,10 @@ static DisasJumpType op_vctz(DisasContext *s, DisasOps *o)gen_gvec_2 (get_field (s->fields, v1),get_field (s->fields, v2), &g[es]);return DISAS_NEXT ; } + +static DisasJumpType op_vx (DisasContext *s,DisasOps *o) +{ +gen_gvec_fn_3 (xor,ES_8 ,get_field (s->fields, v1),get_field (s->fields, v2), +get_field (s->fields, v3)); +return DISAS_NEXT ; +}
Signature · 1 line
2.20.1
PATCH v3 14/40 s390x/tcg: Implement VECTOR GALOIS FIELD MULTIPLY SUM (AND ACCUMULATE)
In reply to David HildenbrandA galois field multiplication in field 2 is like binary multiplication, however instead of doing ordinary binary additions, xor's are performed. So no carries are considered.
Implement all variants via helpers. s390_vec_sar() and s390_vec_shr() will be reused later on.
--- target/s390x/helper.h | 8 ++ target/s390x/insn-data.def | 4 + target/s390x/translate_vx.inc.c | 38 ++++++++ target/s390x/vec_int_helper.c | 167 ++++++++++++++++++++++++++++++++ 4 files changed, 217 insertions(+)
target/s390x/helper.h +8 -0
@@ -154,6 +154,14 @@ DEF_HELPER_FLAGS_3(gvec_vclz8, TCG_CALL_NO_RWG, void, ptr, cptr, i32) DEF_HELPER_FLAGS_3(gvec_vclz16, TCG_CALL_NO_RWG, void, ptr, cptr, i32) DEF_HELPER_FLAGS_3(gvec_vctz8, TCG_CALL_NO_RWG, void, ptr, cptr, i32) DEF_HELPER_FLAGS_3(gvec_vctz16, TCG_CALL_NO_RWG, void, ptr, cptr, i32) +DEF_HELPER_FLAGS_4(gvec_vgfm8, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, i32) +DEF_HELPER_FLAGS_4(gvec_vgfm16, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, i32) +DEF_HELPER_FLAGS_4(gvec_vgfm32, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, i32) +DEF_HELPER_FLAGS_4(gvec_vgfm64, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, i32) +DEF_HELPER_FLAGS_5(gvec_vgfma8, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, cptr, i32) +DEF_HELPER_FLAGS_5(gvec_vgfma16, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, cptr, i32) +DEF_HELPER_FLAGS_5(gvec_vgfma32, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, cptr, i32) +DEF_HELPER_FLAGS_5(gvec_vgfma64, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, cptr, i32)#ifndef CONFIG_USER_ONLY DEF_HELPER_3 (servc, i32, env, i64, i64)
target/s390x/insn-data.def +4 -0
@@ -1090,6 +1090,10 @@ F(0xe752, VCTZ, VRR_a, V, 0, 0, 0, 0, vctz, 0, IF_VEC) /* VECTOR EXCLUSIVE OR */ F(0xe76d, VX, VRR_c, V, 0, 0, 0, 0, vx, 0, IF_VEC) +/* VECTOR GALOIS FIELD MULTIPLY SUM */ + F(0xe7b4, VGFM, VRR_c, V, 0, 0, 0, 0, vgfm, 0, IF_VEC) +/* VECTOR GALOIS FIELD MULTIPLY SUM AND ACCUMULATE */ + F(0xe7bc, VGFMA, VRR_d, V, 0, 0, 0, 0, vgfma, 0, IF_VEC) #ifndef CONFIG_USER_ONLY /* COMPARE AND SWAP AND PURGE */
target/s390x/translate_vx.inc.c +38 -0
@@ -1479,3 +1479,41 @@ static DisasJumpType op_vx(DisasContext *s, DisasOps *o) get_field(s->fields, v3));return DISAS_NEXT ; } + +static DisasJumpType op_vgfm (DisasContext *s,DisasOps *o) +{ +const uint8_t es =get_field (s->fields, m4); +static const GVecGen3 g[4 ] = { + { .fno = gen_helper_gvec_vgfm8, }, + { .fno = gen_helper_gvec_vgfm16, }, + { .fno = gen_helper_gvec_vgfm32, }, + { .fno = gen_helper_gvec_vgfm64, }, + }; + +if (es >ES_64 ) { +gen_program_exception (s,PGM_SPECIFICATION ); +return DISAS_NORETURN ; + } +gen_gvec_3 (get_field (s->fields, v1),get_field (s->fields, v2), +get_field (s->fields, v3), &g[es]); +return DISAS_NEXT ; +} + +static DisasJumpType op_vgfma (DisasContext *s,DisasOps *o) +{ +const uint8_t es =get_field (s->fields, m5); +static const GVecGen4 g[4 ] = { + { .fno = gen_helper_gvec_vgfma8, }, + { .fno = gen_helper_gvec_vgfma16, }, + { .fno = gen_helper_gvec_vgfma32, }, + { .fno = gen_helper_gvec_vgfma64, }, + }; + +if (es >ES_64 ) { +gen_program_exception (s,PGM_SPECIFICATION ); +return DISAS_NORETURN ; + } +gen_gvec_4 (get_field (s->fields, v1),get_field (s->fields, v2), +get_field (s->fields, v3),get_field (s->fields, v4), &g[es]); +return DISAS_NEXT ; +}
target/s390x/vec_int_helper.c +167 -0
176 lines not on this page · raw message
Signature · 1 line
2.20.1
PATCH v3 15/40 s390x/tcg: Implement VECTOR LOAD COMPLEMENT
In reply to David HildenbrandWe can reuse an existing gvec helper for negating the values.
--- target/s390x/insn-data.def | 2 ++ target/s390x/translate_vx.inc.c | 16 ++++++++++++++++ 2 files changed, 18 insertions(+)
target/s390x/insn-data.def +2 -0
@@ -1094,6 +1094,8 @@ F(0xe7b4, VGFM, VRR_c, V, 0, 0, 0, 0, vgfm, 0, IF_VEC) /* VECTOR GALOIS FIELD MULTIPLY SUM AND ACCUMULATE */ F(0xe7bc, VGFMA, VRR_d, V, 0, 0, 0, 0, vgfma, 0, IF_VEC) +/* VECTOR LOAD COMPLEMENT */ + F(0xe7de, VLC, VRR_a, V, 0, 0, 0, 0, vlc, 0, IF_VEC) #ifndef CONFIG_USER_ONLY /* COMPARE AND SWAP AND PURGE */
target/s390x/translate_vx.inc.c +16 -0
@@ -209,6 +209,9 @@ static void get_vec_element_ptr_i64(TCGv_ptr ptr, uint8_t reg, TCGv_i64 enr, 16)#define gen_gvec_dup64i (v1, c) \ tcg_gen_gvec_dup64i(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)#define gen_gvec_fn_3 (fn, es, v1, v2, v3) \ tcg_gen_gvec_##fn(es, vec_full_reg_offset(v1), vec_full_reg_offset(v2), \ vec_full_reg_offset(v3), 16, 16)
@@ -1517,3 +1520,16 @@ static DisasJumpType op_vgfma(DisasContext *s, DisasOps *o) get_field(s->fields, v3), get_field(s->fields, v4), &g[es]);return DISAS_NEXT ; } + +static DisasJumpType op_vlc (DisasContext *s,DisasOps *o) +{ +const uint8_t es =get_field (s->fields, m3); + +if (es >ES_64 ) { +gen_program_exception (s,PGM_SPECIFICATION ); +return DISAS_NORETURN ; + } + +gen_gvec_fn_2 (neg, es,get_field (s->fields, v1),get_field (s->fields, v2)); +return DISAS_NEXT ; +}
Signature · 1 line
2.20.1
PATCH v3 16/40 s390x/tcg: Implement VECTOR LOAD POSITIVE
In reply to David HildenbrandSimilar to VECTOR LOAD COMPLEMENT.
--- target/s390x/insn-data.def | 2 ++ target/s390x/translate_vx.inc.c | 13 +++++++++++++ 2 files changed, 15 insertions(+)
target/s390x/insn-data.def +2 -0
@@ -1096,6 +1096,8 @@ F(0xe7bc, VGFMA, VRR_d, V, 0, 0, 0, 0, vgfma, 0, IF_VEC) /* VECTOR LOAD COMPLEMENT */ F(0xe7de, VLC, VRR_a, V, 0, 0, 0, 0, vlc, 0, IF_VEC) +/* VECTOR LOAD POSITIVE */ + F(0xe7df, VLP, VRR_a, V, 0, 0, 0, 0, vlp, 0, IF_VEC) #ifndef CONFIG_USER_ONLY /* COMPARE AND SWAP AND PURGE */
target/s390x/translate_vx.inc.c +13 -0
@@ -1533,3 +1533,16 @@ static DisasJumpType op_vlc(DisasContext *s, DisasOps *o)gen_gvec_fn_2 (neg, es,get_field (s->fields, v1),get_field (s->fields, v2));return DISAS_NEXT ; } + +static DisasJumpType op_vlp (DisasContext *s,DisasOps *o) +{ +const uint8_t es =get_field (s->fields, m3); + +if (es >ES_64 ) { +gen_program_exception (s,PGM_SPECIFICATION ); +return DISAS_NORETURN ; + } + +gen_gvec_fn_2 (abs, es,get_field (s->fields, v1),get_field (s->fields, v2)); +return DISAS_NEXT ; +}
Signature · 1 line
2.20.1
PATCH v3 17/40 s390x/tcg: Implement VECTOR (MAXIMUM|MINIMUM) (LOGICAL)
In reply to David HildenbrandLuckily, we already have gvec helpers for all four cases.
--- target/s390x/insn-data.def | 8 ++++++++ target/s390x/translate_vx.inc.c | 31 +++++++++++++++++++++++++++++++ 2 files changed, 39 insertions(+)
target/s390x/insn-data.def +8 -0
@@ -1098,6 +1098,14 @@ F(0xe7de, VLC, VRR_a, V, 0, 0, 0, 0, vlc, 0, IF_VEC) /* VECTOR LOAD POSITIVE */ F(0xe7df, VLP, VRR_a, V, 0, 0, 0, 0, vlp, 0, IF_VEC) +/* VECTOR MAXIMUM */ + F(0xe7ff, VMX, VRR_c, V, 0, 0, 0, 0, vmx, 0, IF_VEC) +/* VECTOR MAXIMUM LOGICAL */ + F(0xe7fd, VMXL, VRR_c, V, 0, 0, 0, 0, vmx, 0, IF_VEC) +/* VECTOR MINIMUM */ + F(0xe7fe, VMN, VRR_c, V, 0, 0, 0, 0, vmx, 0, IF_VEC) +/* VECTOR MINIMUM LOGICAL */ + F(0xe7fc, VMNL, VRR_c, V, 0, 0, 0, 0, vmx, 0, IF_VEC) #ifndef CONFIG_USER_ONLY /* COMPARE AND SWAP AND PURGE */
target/s390x/translate_vx.inc.c +31 -0
@@ -1546,3 +1546,34 @@ static DisasJumpType op_vlp(DisasContext *s, DisasOps *o)gen_gvec_fn_2 (abs, es,get_field (s->fields, v1),get_field (s->fields, v2));return DISAS_NEXT ; } + +static DisasJumpType op_vmx (DisasContext *s,DisasOps *o) +{ +const uint8_t v1 =get_field (s->fields, v1); +const uint8_t v2 =get_field (s->fields, v2); +const uint8_t v3 =get_field (s->fields, v3); +const uint8_t es =get_field (s->fields, m4); + +if (es >ES_64 ) { +gen_program_exception (s,PGM_SPECIFICATION ); +return DISAS_NORETURN ; + } + +switch (s->fields->op2) { +case 0xff : +gen_gvec_fn_3 (smax, es, v1, v2, v3); +break ; +case 0xfd : +gen_gvec_fn_3 (umax, es, v1, v2, v3); +break ; +case 0xfe : +gen_gvec_fn_3 (smin, es, v1, v2, v3); +break ; +case 0xfc : +gen_gvec_fn_3 (umin, es, v1, v2, v3); +break ; +default : +g_assert_not_reached (); + } +return DISAS_NEXT ; +}
Signature · 1 line
2.20.1
PATCH v3 18/40 s390x/tcg: Implement VECTOR MULTIPLY AND ADD *
In reply to David HildenbrandQuite some variants to handle. At least handle some 32-bit element variants via gvec expansion (we could also handle 16/32-bit variants for ODD and EVEN easily via gvec expansion, but let's keep it simple for now).
--- target/s390x/helper.h | 18 +++++ target/s390x/insn-data.def | 14 ++++ target/s390x/translate_vx.inc.c | 122 +++++++++++++++++++++++++++++++ target/s390x/vec_int_helper.c | 123 ++++++++++++++++++++++++++++++++ 4 files changed, 277 insertions(+)
target/s390x/helper.h +18 -0
@@ -162,6 +162,24 @@ DEF_HELPER_FLAGS_5(gvec_vgfma8, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, cptr, i3 DEF_HELPER_FLAGS_5(gvec_vgfma16, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, cptr, i32) DEF_HELPER_FLAGS_5(gvec_vgfma32, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, cptr, i32) DEF_HELPER_FLAGS_5(gvec_vgfma64, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, cptr, i32) +DEF_HELPER_FLAGS_5(gvec_vmal8, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, cptr, i32) +DEF_HELPER_FLAGS_5(gvec_vmal16, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, cptr, i32) +DEF_HELPER_FLAGS_5(gvec_vmah8, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, cptr, i32) +DEF_HELPER_FLAGS_5(gvec_vmah16, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, cptr, i32) +DEF_HELPER_FLAGS_5(gvec_vmalh8, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, cptr, i32) +DEF_HELPER_FLAGS_5(gvec_vmalh16, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, cptr, i32) +DEF_HELPER_FLAGS_5(gvec_vmae8, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, cptr, i32) +DEF_HELPER_FLAGS_5(gvec_vmae16, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, cptr, i32) +DEF_HELPER_FLAGS_5(gvec_vmae32, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, cptr, i32) +DEF_HELPER_FLAGS_5(gvec_vmale8, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, cptr, i32) +DEF_HELPER_FLAGS_5(gvec_vmale16, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, cptr, i32) +DEF_HELPER_FLAGS_5(gvec_vmale32, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, cptr, i32) +DEF_HELPER_FLAGS_5(gvec_vmao8, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, cptr, i32) +DEF_HELPER_FLAGS_5(gvec_vmao16, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, cptr, i32) +DEF_HELPER_FLAGS_5(gvec_vmao32, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, cptr, i32) +DEF_HELPER_FLAGS_5(gvec_vmalo8, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, cptr, i32) +DEF_HELPER_FLAGS_5(gvec_vmalo16, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, cptr, i32) +DEF_HELPER_FLAGS_5(gvec_vmalo32, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, cptr, i32)#ifndef CONFIG_USER_ONLY DEF_HELPER_3 (servc, i32, env, i64, i64)
target/s390x/insn-data.def +14 -0
@@ -1106,6 +1106,20 @@ F(0xe7fe, VMN, VRR_c, V, 0, 0, 0, 0, vmx, 0, IF_VEC) /* VECTOR MINIMUM LOGICAL */ F(0xe7fc, VMNL, VRR_c, V, 0, 0, 0, 0, vmx, 0, IF_VEC) +/* VECTOR MULTIPLY AND ADD LOW */ + F(0xe7aa, VMAL, VRR_d, V, 0, 0, 0, 0, vma, 0, IF_VEC) +/* VECTOR MULTIPLY AND ADD HIGH */ + F(0xe7ab, VMAH, VRR_d, V, 0, 0, 0, 0, vma, 0, IF_VEC) +/* VECTOR MULTIPLY AND ADD LOGICAL HIGH */ + F(0xe7a9, VMALH, VRR_d, V, 0, 0, 0, 0, vma, 0, IF_VEC) +/* VECTOR MULTIPLY AND ADD EVEN */ + F(0xe7ae, VMAE, VRR_d, V, 0, 0, 0, 0, vma, 0, IF_VEC) +/* VECTOR MULTIPLY AND ADD LOGICAL EVEN */ + F(0xe7ac, VMALE, VRR_d, V, 0, 0, 0, 0, vma, 0, IF_VEC) +/* VECTOR MULTIPLY AND ADD ODD */ + F(0xe7af, VMAO, VRR_d, V, 0, 0, 0, 0, vma, 0, IF_VEC) +/* VECTOR MULTIPLY AND ADD LOGICAL ODD */ + F(0xe7ad, VMALO, VRR_d, V, 0, 0, 0, 0, vma, 0, IF_VEC) #ifndef CONFIG_USER_ONLY /* COMPARE AND SWAP AND PURGE */
target/s390x/translate_vx.inc.c +122 -0
125 lines not on this page · raw message
target/s390x/vec_int_helper.c +123 -0
126 lines not on this page · raw message
Signature · 1 line
2.20.1
PATCH v3 19/40 s390x/tcg: Implement VECTOR MULTIPLY *
In reply to David HildenbrandYet another set of variants. Implement it similar to VECTOR MULTIPLY AND ADD *. At least for one variant we have a gvec helper we can reuse.
--- target/s390x/helper.h | 16 +++++ target/s390x/insn-data.def | 14 +++++ target/s390x/translate_vx.inc.c | 88 ++++++++++++++++++++++++++++ target/s390x/vec_int_helper.c | 100 ++++++++++++++++++++++++++++++++ 4 files changed, 218 insertions(+)
target/s390x/helper.h +16 -0
@@ -180,6 +180,22 @@ DEF_HELPER_FLAGS_5(gvec_vmao32, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, cptr, i3 DEF_HELPER_FLAGS_5(gvec_vmalo8, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, cptr, i32) DEF_HELPER_FLAGS_5(gvec_vmalo16, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, cptr, i32) DEF_HELPER_FLAGS_5(gvec_vmalo32, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, cptr, i32) +DEF_HELPER_FLAGS_4(gvec_vmh8, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, i32) +DEF_HELPER_FLAGS_4(gvec_vmh16, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, i32) +DEF_HELPER_FLAGS_4(gvec_vmlh8, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, i32) +DEF_HELPER_FLAGS_4(gvec_vmlh16, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, i32) +DEF_HELPER_FLAGS_4(gvec_vme8, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, i32) +DEF_HELPER_FLAGS_4(gvec_vme16, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, i32) +DEF_HELPER_FLAGS_4(gvec_vme32, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, i32) +DEF_HELPER_FLAGS_4(gvec_vmle8, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, i32) +DEF_HELPER_FLAGS_4(gvec_vmle16, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, i32) +DEF_HELPER_FLAGS_4(gvec_vmle32, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, i32) +DEF_HELPER_FLAGS_4(gvec_vmo8, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, i32) +DEF_HELPER_FLAGS_4(gvec_vmo16, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, i32) +DEF_HELPER_FLAGS_4(gvec_vmo32, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, i32) +DEF_HELPER_FLAGS_4(gvec_vmlo8, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, i32) +DEF_HELPER_FLAGS_4(gvec_vmlo16, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, i32) +DEF_HELPER_FLAGS_4(gvec_vmlo32, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, i32)#ifndef CONFIG_USER_ONLY DEF_HELPER_3 (servc, i32, env, i64, i64)
target/s390x/insn-data.def +14 -0
@@ -1120,6 +1120,20 @@ F(0xe7af, VMAO, VRR_d, V, 0, 0, 0, 0, vma, 0, IF_VEC) /* VECTOR MULTIPLY AND ADD LOGICAL ODD */ F(0xe7ad, VMALO, VRR_d, V, 0, 0, 0, 0, vma, 0, IF_VEC) +/* VECTOR MULTIPLY HIGH */ + F(0xe7a3, VMH, VRR_c, V, 0, 0, 0, 0, vm, 0, IF_VEC) +/* VECTOR MULTIPLY LOGICAL HIGH */ + F(0xe7a1, VMLH, VRR_c, V, 0, 0, 0, 0, vm, 0, IF_VEC) +/* VECTOR MULTIPLY LOW */ + F(0xe7a2, VML, VRR_c, V, 0, 0, 0, 0, vm, 0, IF_VEC) +/* VECTOR MULTIPLY EVEN */ + F(0xe7a6, VME, VRR_c, V, 0, 0, 0, 0, vm, 0, IF_VEC) +/* VECTOR MULTIPLY LOGICAL EVEN */ + F(0xe7a4, VMLE, VRR_c, V, 0, 0, 0, 0, vm, 0, IF_VEC) +/* VECTOR MULTIPLY ODD */ + F(0xe7a7, VMO, VRR_c, V, 0, 0, 0, 0, vm, 0, IF_VEC) +/* VECTOR MULTIPLY LOGICAL ODD */ + F(0xe7a5, VMLO, VRR_c, V, 0, 0, 0, 0, vm, 0, IF_VEC) #ifndef CONFIG_USER_ONLY /* COMPARE AND SWAP AND PURGE */
target/s390x/translate_vx.inc.c +88 -0
91 lines not on this page · raw message
target/s390x/vec_int_helper.c +100 -0
103 lines not on this page · raw message
Signature · 1 line
2.20.1
PATCH v3 20/40 s390x/tcg: Implement VECTOR NAND
In reply to David HildenbrandPart of vector enhancements facility 1, but easy to implement.
--- target/s390x/insn-data.def | 2 ++ target/s390x/translate.c | 1 + target/s390x/translate_vx.inc.c | 7 +++++++ 3 files changed, 10 insertions(+)
target/s390x/insn-data.def +2 -0
@@ -1134,6 +1134,8 @@ F(0xe7a7, VMO, VRR_c, V, 0, 0, 0, 0, vm, 0, IF_VEC) /* VECTOR MULTIPLY LOGICAL ODD */ F(0xe7a5, VMLO, VRR_c, V, 0, 0, 0, 0, vm, 0, IF_VEC) +/* VECTOR NAND */ + F(0xe76e, VNN, VRR_c, VE, 0, 0, 0, 0, vnn, 0, IF_VEC) #ifndef CONFIG_USER_ONLY /* COMPARE AND SWAP AND PURGE */
target/s390x/translate.c +1 -0
@@ -6093,6 +6093,7 @@ 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 */ static const DisasInsn insn_info[] = { #include "insn-data.def"
target/s390x/translate_vx.inc.c +7 -0
@@ -1787,3 +1787,10 @@ static DisasJumpType op_vm(DisasContext *s, DisasOps *o) get_field(s->fields, v3), fn);return DISAS_NEXT ; } + +static DisasJumpType op_vnn (DisasContext *s,DisasOps *o) +{ +gen_gvec_fn_3 (nand,ES_8 ,get_field (s->fields, v1), +get_field (s->fields, v2),get_field (s->fields, v3)); +return DISAS_NEXT ; +}
Signature · 1 line
2.20.1
PATCH v3 21/40 s390x/tcg: Implement VECTOR NOR
In reply to David Hildenbrand--- target/s390x/insn-data.def | 2 ++ target/s390x/translate_vx.inc.c | 7 +++++++ 2 files changed, 9 insertions(+)
target/s390x/insn-data.def +2 -0
@@ -1136,6 +1136,8 @@ F(0xe7a5, VMLO, VRR_c, V, 0, 0, 0, 0, vm, 0, IF_VEC) /* VECTOR NAND */ F(0xe76e, VNN, VRR_c, VE, 0, 0, 0, 0, vnn, 0, IF_VEC) +/* VECTOR NOR */ + F(0xe76b, VNO, VRR_c, V, 0, 0, 0, 0, vno, 0, IF_VEC) #ifndef CONFIG_USER_ONLY /* COMPARE AND SWAP AND PURGE */
target/s390x/translate_vx.inc.c +7 -0
@@ -1794,3 +1794,10 @@ static DisasJumpType op_vnn(DisasContext *s, DisasOps *o) get_field(s->fields, v2), get_field(s->fields, v3));return DISAS_NEXT ; } + +static DisasJumpType op_vno (DisasContext *s,DisasOps *o) +{ +gen_gvec_fn_3 (nor,ES_8 ,get_field (s->fields, v1),get_field (s->fields, v2), +get_field (s->fields, v3)); +return DISAS_NEXT ; +}
Signature · 1 line
2.20.1
PATCH v3 22/40 s390x/tcg: Implement VECTOR NOT EXCLUSIVE OR
In reply to David HildenbrandAgain, part of vector enhancement facility 1. The operation corresponds to an bitwise equality check.
--- target/s390x/insn-data.def | 2 ++ target/s390x/translate_vx.inc.c | 7 +++++++ 2 files changed, 9 insertions(+)
target/s390x/insn-data.def +2 -0
@@ -1138,6 +1138,8 @@ F(0xe76e, VNN, VRR_c, VE, 0, 0, 0, 0, vnn, 0, IF_VEC) /* VECTOR NOR */ F(0xe76b, VNO, VRR_c, V, 0, 0, 0, 0, vno, 0, IF_VEC) +/* VECTOR NOT EXCLUSIVE OR */ + F(0xe76c, VNX, VRR_c, VE, 0, 0, 0, 0, vnx, 0, IF_VEC) #ifndef CONFIG_USER_ONLY /* COMPARE AND SWAP AND PURGE */
target/s390x/translate_vx.inc.c +7 -0
@@ -1801,3 +1801,10 @@ static DisasJumpType op_vno(DisasContext *s, DisasOps *o) get_field(s->fields, v3));return DISAS_NEXT ; } + +static DisasJumpType op_vnx (DisasContext *s,DisasOps *o) +{ +gen_gvec_fn_3 (eqv,ES_8 ,get_field (s->fields, v1),get_field (s->fields, v2), +get_field (s->fields, v3)); +return DISAS_NEXT ; +}
Signature · 1 line
2.20.1
PATCH v3 23/40 s390x/tcg: Implement VECTOR OR
In reply to David HildenbrandReuse a gvec helper.
--- target/s390x/insn-data.def | 2 ++ target/s390x/translate_vx.inc.c | 7 +++++++ 2 files changed, 9 insertions(+)
target/s390x/insn-data.def +2 -0
@@ -1140,6 +1140,8 @@ F(0xe76b, VNO, VRR_c, V, 0, 0, 0, 0, vno, 0, IF_VEC) /* VECTOR NOT EXCLUSIVE OR */ F(0xe76c, VNX, VRR_c, VE, 0, 0, 0, 0, vnx, 0, IF_VEC) +/* VECTOR OR */ + F(0xe76a, VO, VRR_c, V, 0, 0, 0, 0, vo, 0, IF_VEC) #ifndef CONFIG_USER_ONLY /* COMPARE AND SWAP AND PURGE */
target/s390x/translate_vx.inc.c +7 -0
@@ -1808,3 +1808,10 @@ static DisasJumpType op_vnx(DisasContext *s, DisasOps *o) get_field(s->fields, v3));return DISAS_NEXT ; } + +static DisasJumpType op_vo (DisasContext *s,DisasOps *o) +{ +gen_gvec_fn_3 (or,ES_8 ,get_field (s->fields, v1),get_field (s->fields, v2), +get_field (s->fields, v3)); +return DISAS_NEXT ; +}
Signature · 1 line
2.20.1
PATCH v3 24/40 s390x/tcg: Implement VECTOR OR WITH COMPLEMENT
In reply to David HildenbrandAgain, vector enhancements facility 1 material.
--- target/s390x/insn-data.def | 2 ++ target/s390x/translate_vx.inc.c | 7 +++++++ 2 files changed, 9 insertions(+)
target/s390x/insn-data.def +2 -0
@@ -1142,6 +1142,8 @@ F(0xe76c, VNX, VRR_c, VE, 0, 0, 0, 0, vnx, 0, IF_VEC) /* VECTOR OR */ F(0xe76a, VO, VRR_c, V, 0, 0, 0, 0, vo, 0, IF_VEC) +/* VECTOR OR WITH COMPLEMENT */ + F(0xe76f, VOC, VRR_c, VE, 0, 0, 0, 0, voc, 0, IF_VEC) #ifndef CONFIG_USER_ONLY /* COMPARE AND SWAP AND PURGE */
target/s390x/translate_vx.inc.c +7 -0
@@ -1815,3 +1815,10 @@ static DisasJumpType op_vo(DisasContext *s, DisasOps *o) get_field(s->fields, v3));return DISAS_NEXT ; } + +static DisasJumpType op_voc (DisasContext *s,DisasOps *o) +{ +gen_gvec_fn_3 (orc,ES_8 ,get_field (s->fields, v1),get_field (s->fields, v2), +get_field (s->fields, v3)); +return DISAS_NEXT ; +}
Signature · 1 line
2.20.1
PATCH v3 25/40 s390x/tcg: Implement VECTOR POPULATION COUNT
In reply to David HildenbrandSimilar to VECTOR COUNT TRAILING ZEROES.
--- target/s390x/helper.h | 2 ++ target/s390x/insn-data.def | 2 ++ target/s390x/translate_vx.inc.c | 19 +++++++++++++++++++ target/s390x/vec_int_helper.c | 14 ++++++++++++++ 4 files changed, 37 insertions(+)
target/s390x/helper.h +2 -0
@@ -196,6 +196,8 @@ DEF_HELPER_FLAGS_4(gvec_vmo32, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, i32) DEF_HELPER_FLAGS_4(gvec_vmlo8, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, i32) DEF_HELPER_FLAGS_4(gvec_vmlo16, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, i32) DEF_HELPER_FLAGS_4(gvec_vmlo32, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, i32) +DEF_HELPER_FLAGS_3(gvec_vpopct8, TCG_CALL_NO_RWG, void, ptr, cptr, i32) +DEF_HELPER_FLAGS_3(gvec_vpopct16, TCG_CALL_NO_RWG, void, ptr, cptr, i32)#ifndef CONFIG_USER_ONLY DEF_HELPER_3 (servc, i32, env, i64, i64)
target/s390x/insn-data.def +2 -0
@@ -1144,6 +1144,8 @@ F(0xe76a, VO, VRR_c, V, 0, 0, 0, 0, vo, 0, IF_VEC) /* VECTOR OR WITH COMPLEMENT */ F(0xe76f, VOC, VRR_c, VE, 0, 0, 0, 0, voc, 0, IF_VEC) +/* VECTOR POPULATION COUNT */ + F(0xe750, VPOPCT, VRR_a, V, 0, 0, 0, 0, vpopct, 0, IF_VEC) #ifndef CONFIG_USER_ONLY /* COMPARE AND SWAP AND PURGE */
target/s390x/translate_vx.inc.c +19 -0
@@ -1822,3 +1822,22 @@ static DisasJumpType op_voc(DisasContext *s, DisasOps *o) get_field(s->fields, v3));return DISAS_NEXT ; } + +static DisasJumpType op_vpopct (DisasContext *s,DisasOps *o) +{ +const uint8_t es =get_field (s->fields, m3); +static const GVecGen2 g[4 ] = { + { .fno = gen_helper_gvec_vpopct8, }, + { .fno = gen_helper_gvec_vpopct16, }, + { .fni4 = tcg_gen_ctpop_i32, }, + { .fni8 = tcg_gen_ctpop_i64, }, + }; + +if (es >ES_64 || (es !=ES_8 && !s390_has_feat (S390_FEAT_VECTOR_ENH ))) { +gen_program_exception (s,PGM_SPECIFICATION ); +return DISAS_NORETURN ; + } + +gen_gvec_2 (get_field (s->fields, v1),get_field (s->fields, v2), &g[es]); +return DISAS_NEXT ; +}
target/s390x/vec_int_helper.c +14 -0
@@ -464,3 +464,17 @@ void HELPER(gvec_vmlo##BITS)(void *v1, const void *v2, const void *v3, \ DEF_VMLO(8, 16) DEF_VMLO(16, 32) DEF_VMLO(32, 64) + +#define DEF_VPOPCT (BITS ) \ +void HELPER(gvec_vpopct##BITS)(void *v1, const void *v2, uint32_t desc) \ +{ \ + int i; \ + \ + for (i = 0; i < (128 / BITS); i++) { \ + const uint##BITS##_t a = s390_vec_read_element##BITS(v2, i); \ + \ + s390_vec_write_element##BITS(v1, i, ctpop32(a)); \ + } \ +} +DEF_VPOPCT(8) +DEF_VPOPCT (16 )
Signature · 1 line
2.20.1
PATCH v3 26/40 s390x/tcg: Implement VECTOR ELEMENT ROTATE LEFT LOGICAL
In reply to David HildenbrandTake care of properly taking the modulo of the count. We might later want to come back and create a variant of VERLL where the base register is 0, resulting in an immediate.
--- target/s390x/helper.h | 4 +++ target/s390x/insn-data.def | 3 ++ target/s390x/translate_vx.inc.c | 60 +++++++++++++++++++++++++++++++++ target/s390x/vec_int_helper.c | 31 +++++++++++++++++ 4 files changed, 98 insertions(+)
target/s390x/helper.h +4 -0
@@ -198,6 +198,10 @@ DEF_HELPER_FLAGS_4(gvec_vmlo16, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, i32) DEF_HELPER_FLAGS_4(gvec_vmlo32, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, i32) DEF_HELPER_FLAGS_3(gvec_vpopct8, TCG_CALL_NO_RWG, void, ptr, cptr, i32) DEF_HELPER_FLAGS_3(gvec_vpopct16, TCG_CALL_NO_RWG, void, ptr, cptr, i32) +DEF_HELPER_FLAGS_4(gvec_verllv8, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, i32) +DEF_HELPER_FLAGS_4(gvec_verllv16, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, i32) +DEF_HELPER_FLAGS_4(gvec_verll8, TCG_CALL_NO_RWG, void, ptr, cptr, i64, i32) +DEF_HELPER_FLAGS_4(gvec_verll16, TCG_CALL_NO_RWG, void, ptr, cptr, i64, i32)#ifndef CONFIG_USER_ONLY DEF_HELPER_3 (servc, i32, env, i64, i64)
target/s390x/insn-data.def +3 -0
@@ -1146,6 +1146,9 @@ F(0xe76f, VOC, VRR_c, VE, 0, 0, 0, 0, voc, 0, IF_VEC) /* VECTOR POPULATION COUNT */ F(0xe750, VPOPCT, VRR_a, V, 0, 0, 0, 0, vpopct, 0, IF_VEC) +/* VECTOR ELEMENT ROTATE LEFT LOGICAL */ + F(0xe773, VERLLV, VRR_c, V, 0, 0, 0, 0, verllv, 0, IF_VEC) + F(0xe733, VERLL, VRS_a, V, la2, 0, 0, 0, verll, 0, IF_VEC) #ifndef CONFIG_USER_ONLY /* COMPARE AND SWAP AND PURGE */
target/s390x/translate_vx.inc.c +60 -0
69 lines not on this page · raw message
target/s390x/vec_int_helper.c +31 -0
@@ -478,3 +478,34 @@ void HELPER(gvec_vpopct##BITS)(void *v1, const void *v2, uint32_t desc) \ } DEF_VPOPCT(8) DEF_VPOPCT(16) + +#define DEF_VERLLV (BITS ) \ +void HELPER(gvec_verllv##BITS)(void *v1, const void *v2, const void *v3, \ + uint32_t desc) \ +{ \ + int i; \ + \ + for (i = 0; i < (128 / BITS); i++) { \ + const uint##BITS##_t a = s390_vec_read_element##BITS(v2, i); \ + const uint##BITS##_t b = s390_vec_read_element##BITS(v3, i); \ + \ + s390_vec_write_element##BITS(v1, i, rol##BITS(a, b)); \ + } \ +} +DEF_VERLLV(8) +DEF_VERLLV(16) + +#define DEF_VERLL (BITS ) \ +void HELPER(gvec_verll##BITS)(void *v1, const void *v2, uint64_t count, \ + uint32_t desc) \ +{ \ + int i; \ + \ + for (i = 0; i < (128 / BITS); i++) { \ + const uint##BITS##_t a = s390_vec_read_element##BITS(v2, i); \ + \ + s390_vec_write_element##BITS(v1, i, rol##BITS(a, count)); \ + } \ +} +DEF_VERLL(8) +DEF_VERLL (16 )
Signature · 1 line
2.20.1
PATCH v3 27/40 s390x/tcg: Implement VECTOR ELEMENT ROTATE AND INSERT UNDER MASK
In reply to David HildenbrandUse the new vector expansion for GVecGen3i.
--- target/s390x/helper.h | 2 ++ target/s390x/insn-data.def | 2 ++ target/s390x/translate_vx.inc.c | 51 +++++++++++++++++++++++++++++++++ target/s390x/vec_int_helper.c | 20 +++++++++++++ 4 files changed, 75 insertions(+)
target/s390x/helper.h +2 -0
@@ -202,6 +202,8 @@ DEF_HELPER_FLAGS_4(gvec_verllv8, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, i32) DEF_HELPER_FLAGS_4(gvec_verllv16, TCG_CALL_NO_RWG, void, ptr, cptr, cptr, i32) DEF_HELPER_FLAGS_4(gvec_verll8, TCG_CALL_NO_RWG, void, ptr, cptr, i64, i32) DEF_HELPER_FLAGS_4(gvec_verll16, TCG_CALL_NO_RWG, void, ptr, cptr, i64, 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)#ifndef CONFIG_USER_ONLY DEF_HELPER_3 (servc, i32, env, i64, i64)
target/s390x/insn-data.def +2 -0
@@ -1149,6 +1149,8 @@ /* VECTOR ELEMENT ROTATE LEFT LOGICAL */ F(0xe773, VERLLV, VRR_c, V, 0, 0, 0, 0, verllv, 0, IF_VEC) F(0xe733, VERLL, VRS_a, V, la2, 0, 0, 0, verll, 0, IF_VEC) +/* VECTOR ELEMENT ROTATE AND INSERT UNDER MASK */ + F(0xe772, VERIM, VRI_d, V, 0, 0, 0, 0, verim, 0, IF_VEC) #ifndef CONFIG_USER_ONLY /* COMPARE AND SWAP AND PURGE */
target/s390x/translate_vx.inc.c +51 -0
60 lines not on this page · raw message
target/s390x/vec_int_helper.c +20 -0
@@ -14,6 +14,7 @@#include "cpu.h" #include "vec.h" #include "exec/helper-proto.h" +#include "tcg/tcg-gvec-desc.h" static bool s390_vec_is_zero (const S390Vector *v) {
@@ -509,3 +510,22 @@ void HELPER(gvec_verll##BITS)(void *v1, const void *v2, uint64_t count, \ } DEF_VERLL(8) DEF_VERLL(16) + +#define DEF_VERIM (BITS ) \ +void HELPER(gvec_verim##BITS)(void *v1, const void *v2, const void *v3, \ + uint32_t desc) \ +{ \ + const uint8_t count = simd_data(desc); \ + int i; \ + \ + for (i = 0; i < (128 / BITS); i++) { \ + const uint##BITS##_t a = s390_vec_read_element##BITS(v1, i); \ + const uint##BITS##_t b = s390_vec_read_element##BITS(v2, i); \ + const uint##BITS##_t mask = s390_vec_read_element##BITS(v3, i); \ + const uint##BITS##_t d = (a & ~mask) | (rol##BITS(b, count) & mask); \ + \ + s390_vec_write_element##BITS(v1, i, d); \ + } \ +} +DEF_VERIM(8) +DEF_VERIM (16 )
Signature · 1 line
2.20.1
Re: [Qemu-devel] [PATCH v3 27/40] s390x/tcg: Implement VECTOR ELEMENT ROTATE AND INSERT UNDER MASK
Quoted 9 lines · David Hildenbrand
Use the new vector expansion for GVecGen3i.
Signed-off-by David Hildenbrand <[email protected]>--- target/s390x/helper.h | 2 ++ target/s390x/insn-data.def | 2 ++ target/s390x/translate_vx.inc.c | 51 +++++++++++++++++++++++++++++++++ target/s390x/vec_int_helper.c | 20 +++++++++++++ 4 files changed, 75 insertions(+)
r~
PATCH v3 28/40 s390x/tcg: Implement VECTOR ELEMENT SHIFT
In reply to David HildenbrandWe can use all the fancy new vector helpers implemented by Richard.
One important thing to take care of is always to properly mask of unused bits from the shift count.
--- target/s390x/insn-data.def | 9 ++++ target/s390x/translate_vx.inc.c | 84 +++++++++++++++++++++++++++++++++ 2 files changed, 93 insertions(+)
target/s390x/insn-data.def +9 -0
@@ -1151,6 +1151,15 @@ F(0xe733, VERLL, VRS_a, V, la2, 0, 0, 0, verll, 0, IF_VEC) /* VECTOR ELEMENT ROTATE AND INSERT UNDER MASK */ F(0xe772, VERIM, VRI_d, V, 0, 0, 0, 0, verim, 0, IF_VEC) +/* VECTOR ELEMENT SHIFT LEFT */ + F(0xe770, VESLV, VRR_c, V, 0, 0, 0, 0, vesv, 0, IF_VEC) + F(0xe730, VESL, VRS_a, V, la2, 0, 0, 0, ves, 0, IF_VEC) +/* VECTOR ELEMENT SHIFT RIGHT ARITHMETIC */ + F(0xe77a, VESRAV, VRR_c, V, 0, 0, 0, 0, vesv, 0, IF_VEC) + F(0xe73a, VESRA, VRS_a, V, la2, 0, 0, 0, ves, 0, IF_VEC) +/* VECTOR ELEMENT SHIFT RIGHT LOGICAL */ + F(0xe778, VESRLV, VRR_c, V, 0, 0, 0, 0, vesv, 0, IF_VEC) + F(0xe738, VESRL, VRS_a, V, la2, 0, 0, 0, ves, 0, IF_VEC) #ifndef CONFIG_USER_ONLY /* COMPARE AND SWAP AND PURGE */
target/s390x/translate_vx.inc.c +84 -0
93 lines not on this page · raw message
Signature · 1 line
2.20.1
PATCH v3 29/40 s390x/tcg: Implement VECTOR SHIFT LEFT (BY BYTE)
In reply to David HildenbrandWe can reuse the existing 128-bit shift utility function.
--- target/s390x/helper.h | 1 + target/s390x/insn-data.def | 4 ++++ target/s390x/translate_vx.inc.c | 20 ++++++++++++++++++++ target/s390x/vec_int_helper.c | 6 ++++++ 4 files changed, 31 insertions(+)
target/s390x/helper.h +1 -0
@@ -204,6 +204,7 @@ DEF_HELPER_FLAGS_4(gvec_verll8, TCG_CALL_NO_RWG, void, ptr, cptr, i64, i32) DEF_HELPER_FLAGS_4(gvec_verll16, TCG_CALL_NO_RWG, void, ptr, cptr, i64, 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)#ifndef CONFIG_USER_ONLY DEF_HELPER_3 (servc, i32, env, i64, i64)
target/s390x/insn-data.def +4 -0
@@ -1160,6 +1160,10 @@ /* VECTOR ELEMENT SHIFT RIGHT LOGICAL */ F(0xe778, VESRLV, VRR_c, V, 0, 0, 0, 0, vesv, 0, IF_VEC) F(0xe738, VESRL, VRS_a, V, la2, 0, 0, 0, ves, 0, IF_VEC) +/* VECTOR SHIFT LEFT */ + 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) #ifndef CONFIG_USER_ONLY /* COMPARE AND SWAP AND PURGE */
target/s390x/translate_vx.inc.c +20 -0
@@ -188,6 +188,9 @@ static void get_vec_element_ptr_i64(TCGv_ptr ptr, uint8_t reg, TCGv_i64 enr,#define gen_gvec_2s (v1, v2, c, gen) \ tcg_gen_gvec_2s(vec_full_reg_offset(v1), vec_full_reg_offset(v2), \ 16, 16, c, gen) +#define gen_gvec_2i_ool (v1, v2, c, data, fn) \ + tcg_gen_gvec_2i_ool(vec_full_reg_offset(v1), vec_full_reg_offset(v2), \ + c, 16, 16, data, fn)#define gen_gvec_3 (v1, v2, v3, gen) \ tcg_gen_gvec_3(vec_full_reg_offset(v1), vec_full_reg_offset(v2), \ vec_full_reg_offset(v3), 16, 16, gen)
@@ -2036,3 +2039,20 @@ static DisasJumpType op_ves(DisasContext *s, DisasOps *o) }return DISAS_NEXT ; } + +static DisasJumpType op_vsl (DisasContext *s,DisasOps *o) +{ +TCGv_i64 shift =tcg_temp_new_i64 (); + +read_vec_element_i64 (shift,get_field (s->fields, v3),7 ,ES_8 ); +if (s->fields->op2 ==0x74 ) { +tcg_gen_andi_i64 (shift, shift,0x7 ); + }else { +tcg_gen_andi_i64 (shift, shift,0x78 ); + } + +gen_gvec_2i_ool (get_field (s->fields, v1),get_field (s->fields, v2), + shift,0 , gen_helper_gvec_vsl); +tcg_temp_free_i64 (shift); +return DISAS_NEXT ; +}
target/s390x/vec_int_helper.c +6 -0
@@ -529,3 +529,9 @@ void HELPER(gvec_verim##BITS)(void *v1, const void *v2, const void *v3, \ } DEF_VERIM(8) DEF_VERIM(16) + +void HELPER (gvec_vsl )(void *v1,const void *v2,uint64_t count, +uint32_t desc) +{ +s390_vec_shl (v1, v2, count); +}
Signature · 1 line
2.20.1
PATCH v3 30/40 s390x/tcg: Implement VECTOR SHIFT LEFT DOUBLE BY BYTE
In reply to David HildenbrandInline expansion courtesy of Richard H.
--- target/s390x/insn-data.def | 2 ++ target/s390x/translate_vx.inc.c | 29 +++++++++++++++++++++++++++++ 2 files changed, 31 insertions(+)
target/s390x/insn-data.def +2 -0
@@ -1164,6 +1164,8 @@ 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 BYTE */ + F(0xe777, VSLDB, VRI_d, V, 0, 0, 0, 0, vsldb, 0, IF_VEC) #ifndef CONFIG_USER_ONLY /* COMPARE AND SWAP AND PURGE */
target/s390x/translate_vx.inc.c +29 -0
@@ -2056,3 +2056,32 @@ static DisasJumpType op_vsl(DisasContext *s, DisasOps *o)tcg_temp_free_i64 (shift);return DISAS_NEXT ; } + +static DisasJumpType op_vsldb (DisasContext *s,DisasOps *o) +{ +const uint8_t i4 =get_field (s->fields, i4) &0xf ; +const int left_shift = (i4 &7 ) *8 ; +const int right_shift =64 - left_shift; +TCGv_i64 t0 =tcg_temp_new_i64 (); +TCGv_i64 t1 =tcg_temp_new_i64 (); +TCGv_i64 t2 =tcg_temp_new_i64 (); + +if ((i4 &8 ) ==0 ) { +read_vec_element_i64 (t0,get_field (s->fields, v2),0 ,ES_64 ); +read_vec_element_i64 (t1,get_field (s->fields, v2),1 ,ES_64 ); +read_vec_element_i64 (t2,get_field (s->fields, v3),0 ,ES_64 ); + }else { +read_vec_element_i64 (t0,get_field (s->fields, v2),1 ,ES_64 ); +read_vec_element_i64 (t1,get_field (s->fields, v3),0 ,ES_64 ); +read_vec_element_i64 (t2,get_field (s->fields, v3),1 ,ES_64 ); + } +tcg_gen_extract2_i64 (t0, t1, t0, right_shift); +tcg_gen_extract2_i64 (t1, t2, t1, right_shift); +write_vec_element_i64 (t0,get_field (s->fields, v1),0 ,ES_64 ); +write_vec_element_i64 (t1,get_field (s->fields, v1),1 ,ES_64 ); + +tcg_temp_free (t0); +tcg_temp_free (t1); +tcg_temp_free (t2); +return DISAS_NEXT ; +}
Signature · 1 line
2.20.1
Re: [Qemu-devel] [PATCH v3 30/40] s390x/tcg: Implement VECTOR SHIFT LEFT DOUBLE BY BYTE
Quoted 7 lines · David Hildenbrand
Inline expansion courtesy of Richard H.
Signed-off-by David Hildenbrand <[email protected]>--- target/s390x/insn-data.def | 2 ++ target/s390x/translate_vx.inc.c | 29 +++++++++++++++++++++++++++++ 2 files changed, 31 insertions(+)
r~
PATCH v3 31/40 s390x/tcg: Implement VECTOR SHIFT RIGHT ARITHMETIC
In reply to David HildenbrandSimilar to VECTOR SHIFT LEFT ARITHMETIC. Add s390_vec_sar() similar to s390_vec_shr().
--- target/s390x/helper.h | 1 + target/s390x/insn-data.def | 4 ++++ target/s390x/translate_vx.inc.c | 17 +++++++++++++++++ target/s390x/vec_int_helper.c | 26 ++++++++++++++++++++++++++ 4 files changed, 48 insertions(+)
target/s390x/helper.h +1 -0
@@ -205,6 +205,7 @@ DEF_HELPER_FLAGS_4(gvec_verll16, TCG_CALL_NO_RWG, void, ptr, cptr, i64, 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_vsra, TCG_CALL_NO_RWG, void, ptr, cptr, i64, i32)#ifndef CONFIG_USER_ONLY DEF_HELPER_3 (servc, i32, env, i64, i64)
target/s390x/insn-data.def +4 -0
@@ -1166,6 +1166,10 @@ F(0xe775, VSLB, VRR_c, V, 0, 0, 0, 0, vsl, 0, IF_VEC) /* VECTOR SHIFT LEFT DOUBLE BY BYTE */ F(0xe777, VSLDB, VRI_d, V, 0, 0, 0, 0, vsldb, 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) #ifndef CONFIG_USER_ONLY /* COMPARE AND SWAP AND PURGE */
target/s390x/translate_vx.inc.c +17 -0
@@ -2085,3 +2085,20 @@ static DisasJumpType op_vsldb(DisasContext *s, DisasOps *o)tcg_temp_free (t2);return DISAS_NEXT ; } + +static DisasJumpType op_vsra (DisasContext *s,DisasOps *o) +{ +TCGv_i64 shift =tcg_temp_new_i64 (); + +read_vec_element_i64 (shift,get_field (s->fields, v3),7 ,ES_8 ); +if (s->fields->op2 ==0x7e ) { +tcg_gen_andi_i64 (shift, shift,0x7 ); + }else { +tcg_gen_andi_i64 (shift, shift,0x78 ); + } + +gen_gvec_2i_ool (get_field (s->fields, v1),get_field (s->fields, v2), + shift,0 , gen_helper_gvec_vsra); +tcg_temp_free_i64 (shift); +return DISAS_NEXT ; +}
target/s390x/vec_int_helper.c +26 -0
@@ -49,6 +49,26 @@ static void s390_vec_shl(S390Vector *d, const S390Vector *a, uint64_t count) } } +static void s390_vec_sar (S390Vector *d,const S390Vector *a,uint64_t count) +{ +uint64_t tmp; + +if (count ==0 ) { + d->doubleword[0 ] = a->doubleword[0 ]; + d->doubleword[1 ] = a->doubleword[1 ]; + }else if (count ==64 ) { + d->doubleword[1 ] = a->doubleword[0 ]; + d->doubleword[0 ] =0 ; + }else if (count <64 ) { + tmp = a->doubleword[1 ] >> count; + d->doubleword[1 ] =deposit64 (tmp,64 - count, count, a->doubleword[0 ]); + d->doubleword[0 ] = (int64_t )a->doubleword[0 ] >> count; + }else { + d->doubleword[1 ] = (int64_t )a->doubleword[0 ] >> (count -64 ); + d->doubleword[0 ] =0 ; + } +} +static void s390_vec_shr (S390Vector *d,const S390Vector *a,uint64_t count) {uint64_t tmp;
@@ -535,3 +555,9 @@ void HELPER(gvec_vsl)(void *v1, const void *v2, uint64_t count, {s390_vec_shl (v1, v2, count); } + +void HELPER (gvec_vsra )(void *v1,const void *v2,uint64_t count, +uint32_t desc) +{ +s390_vec_sar (v1, v2, count); +}
Signature · 1 line
2.20.1
PATCH v3 32/40 s390x/tcg: Implement VECTOR SHIFT RIGHT LOGICAL *
In reply to David HildenbrandSimilar to VECTOR SHIFT RIGHT ARITHMETICAL.
--- target/s390x/helper.h | 1 + target/s390x/insn-data.def | 4 ++++ target/s390x/translate_vx.inc.c | 17 +++++++++++++++++ target/s390x/vec_int_helper.c | 6 ++++++ 4 files changed, 28 insertions(+)
target/s390x/helper.h +1 -0
@@ -206,6 +206,7 @@ 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_vsra, TCG_CALL_NO_RWG, void, ptr, cptr, i64, i32) +DEF_HELPER_FLAGS_4(gvec_vsrl, TCG_CALL_NO_RWG, void, ptr, cptr, i64, i32)#ifndef CONFIG_USER_ONLY DEF_HELPER_3 (servc, i32, env, i64, i64)
target/s390x/insn-data.def +4 -0
@@ -1170,6 +1170,10 @@ 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 LOGICAL */ + F(0xe77c, VSRL, VRR_c, V, 0, 0, 0, 0, vsrl, 0, IF_VEC) +/* VECTOR SHIFT RIGHT LOGICAL BY BYTE */ + F(0xe77d, VSRLB, VRR_c, V, 0, 0, 0, 0, vsrl, 0, IF_VEC) #ifndef CONFIG_USER_ONLY /* COMPARE AND SWAP AND PURGE */
target/s390x/translate_vx.inc.c +17 -0
@@ -2102,3 +2102,20 @@ static DisasJumpType op_vsra(DisasContext *s, DisasOps *o)tcg_temp_free_i64 (shift);return DISAS_NEXT ; } + +static DisasJumpType op_vsrl (DisasContext *s,DisasOps *o) +{ +TCGv_i64 shift =tcg_temp_new_i64 (); + +read_vec_element_i64 (shift,get_field (s->fields, v3),7 ,ES_8 ); +if (s->fields->op2 ==0x7c ) { +tcg_gen_andi_i64 (shift, shift,0x7 ); + }else { +tcg_gen_andi_i64 (shift, shift,0x78 ); + } + +gen_gvec_2i_ool (get_field (s->fields, v1),get_field (s->fields, v2), + shift,0 , gen_helper_gvec_vsrl); +tcg_temp_free_i64 (shift); +return DISAS_NEXT ; +}
target/s390x/vec_int_helper.c +6 -0
@@ -561,3 +561,9 @@ void HELPER(gvec_vsra)(void *v1, const void *v2, uint64_t count, {s390_vec_sar (v1, v2, count); } + +void HELPER (gvec_vsrl )(void *v1,const void *v2,uint64_t count, +uint32_t desc) +{ +s390_vec_shr (v1, v2, count); +}
Signature · 1 line
2.20.1
PATCH v3 33/40 s390x/tcg: Implement VECTOR SUBTRACT
In reply to David HildenbrandWe can use tcg_gen_sub2_i64() to do 128-bit subtraction and otherwise existing gvec helpers.
--- target/s390x/insn-data.def | 2 ++ target/s390x/translate_vx.inc.c | 17 +++++++++++++++++ 2 files changed, 19 insertions(+)
target/s390x/insn-data.def +2 -0
@@ -1174,6 +1174,8 @@ F(0xe77c, VSRL, VRR_c, V, 0, 0, 0, 0, vsrl, 0, IF_VEC) /* VECTOR SHIFT RIGHT LOGICAL BY BYTE */ F(0xe77d, VSRLB, VRR_c, V, 0, 0, 0, 0, vsrl, 0, IF_VEC) +/* VECTOR SUBTRACT */ + F(0xe7f7, VS, VRR_c, V, 0, 0, 0, 0, vs, 0, IF_VEC) #ifndef CONFIG_USER_ONLY /* COMPARE AND SWAP AND PURGE */
target/s390x/translate_vx.inc.c +17 -0
@@ -2119,3 +2119,20 @@ static DisasJumpType op_vsrl(DisasContext *s, DisasOps *o)tcg_temp_free_i64 (shift);return DISAS_NEXT ; } + +static DisasJumpType op_vs (DisasContext *s,DisasOps *o) +{ +const uint8_t es =get_field (s->fields, m4); + +if (es >ES_128 ) { +gen_program_exception (s,PGM_SPECIFICATION ); +return DISAS_NORETURN ; + }else if (es ==ES_128 ) { +gen_gvec128_3_i64 (tcg_gen_sub2_i64,get_field (s->fields, v1), +get_field (s->fields, v2),get_field (s->fields, v3)); +return DISAS_NEXT ; + } +gen_gvec_fn_3 (sub, es,get_field (s->fields, v1),get_field (s->fields, v2), +get_field (s->fields, v3)); +return DISAS_NEXT ; +}
Signature · 1 line
2.20.1
PATCH v3 34/40 s390x/tcg: Implement VECTOR SUBTRACT COMPUTE BORROW INDICATION
In reply to David HildenbrandLet's keep it simple for now and handle 8/16 bit elements via helpers. Especially for 8/16, we could come up with some bit tricks.
--- target/s390x/helper.h | 2 ++ target/s390x/insn-data.def | 2 ++ target/s390x/translate_vx.inc.c | 52 +++++++++++++++++++++++++++++++++ target/s390x/vec_int_helper.c | 16 ++++++++++ 4 files changed, 72 insertions(+)
target/s390x/helper.h +2 -0
@@ -207,6 +207,8 @@ 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_vsra, TCG_CALL_NO_RWG, void, ptr, cptr, i64, i32) DEF_HELPER_FLAGS_4(gvec_vsrl, TCG_CALL_NO_RWG, void, ptr, cptr, i64, 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)#ifndef CONFIG_USER_ONLY DEF_HELPER_3 (servc, i32, env, i64, i64)
target/s390x/insn-data.def +2 -0
@@ -1176,6 +1176,8 @@ F(0xe77d, VSRLB, VRR_c, V, 0, 0, 0, 0, vsrl, 0, IF_VEC) /* VECTOR SUBTRACT */ F(0xe7f7, VS, VRR_c, V, 0, 0, 0, 0, vs, 0, IF_VEC) +/* VECTOR SUBTRACT COMPUTE BORROW INDICATION */ + F(0xe7f5, VSCBI, VRR_c, V, 0, 0, 0, 0, vscbi, 0, IF_VEC) #ifndef CONFIG_USER_ONLY /* COMPARE AND SWAP AND PURGE */
target/s390x/translate_vx.inc.c +52 -0
55 lines not on this page · raw message
target/s390x/vec_int_helper.c +16 -0
@@ -567,3 +567,19 @@ void HELPER(gvec_vsrl)(void *v1, const void *v2, uint64_t count, {s390_vec_shr (v1, v2, count); } + +#define DEF_VSCBI (BITS ) \ +void HELPER(gvec_vscbi##BITS)(void *v1, const void *v2, const void *v3, \ + uint32_t desc) \ +{ \ + int i; \ + \ + for (i = 0; i < (128 / BITS); i++) { \ + const uint##BITS##_t a = s390_vec_read_element##BITS(v2, i); \ + const uint##BITS##_t b = s390_vec_read_element##BITS(v3, i); \ + \ + s390_vec_write_element##BITS(v1, i, a < b); \ + } \ +} +DEF_VSCBI(8) +DEF_VSCBI (16 )
Signature · 1 line
2.20.1
Re: [Qemu-devel] [PATCH v3 34/40] s390x/tcg: Implement VECTOR SUBTRACT COMPUTE BORROW INDICATION
Quoted 10 lines · David Hildenbrand
Let's keep it simple for now and handle 8/16 bit elements via helpers. Especially for 8/16, we could come up with some bit tricks.
Signed-off-by David Hildenbrand <[email protected]>--- target/s390x/helper.h | 2 ++ target/s390x/insn-data.def | 2 ++ target/s390x/translate_vx.inc.c | 52 +++++++++++++++++++++++++++++++++ target/s390x/vec_int_helper.c | 16 ++++++++++ 4 files changed, 72 insertions(+)
r~
PATCH v3 35/40 s390x/tcg: Implement VECTOR SUBTRACT WITH BORROW INDICATION
In reply to David HildenbrandFairly easy as only 128-bit handling is required. Simply perform the subtraction and then subtract the borrow.
--- target/s390x/insn-data.def | 2 ++ target/s390x/translate_vx.inc.c | 26 ++++++++++++++++++++++++++ 2 files changed, 28 insertions(+)
target/s390x/insn-data.def +2 -0
@@ -1178,6 +1178,8 @@ F(0xe7f7, VS, VRR_c, V, 0, 0, 0, 0, vs, 0, IF_VEC) /* VECTOR SUBTRACT COMPUTE BORROW INDICATION */ F(0xe7f5, VSCBI, VRR_c, V, 0, 0, 0, 0, vscbi, 0, IF_VEC) +/* VECTOR SUBTRACT WITH BORROW INDICATION */ + F(0xe7bf, VSBI, VRR_d, V, 0, 0, 0, 0, vsbi, 0, IF_VEC) #ifndef CONFIG_USER_ONLY /* COMPARE AND SWAP AND PURGE */
target/s390x/translate_vx.inc.c +26 -0
@@ -2188,3 +2188,29 @@ static DisasJumpType op_vscbi(DisasContext *s, DisasOps *o) get_field(s->fields, v3), &g[es]);return DISAS_NEXT ; } + +static void gen_sbi2_i64 (TCGv_i64 dl,TCGv_i64 dh,TCGv_i64 al,TCGv_i64 ah, +TCGv_i64 bl,TCGv_i64 bh,TCGv_i64 cl,TCGv_i64 ch) +{ +TCGv_i64 tl =tcg_temp_new_i64 (); +TCGv_i64 zero =tcg_const_i64 (0 ); + +tcg_gen_andi_i64 (tl, cl,1 ); +tcg_gen_sub2_i64 (dl, dh, al, ah, bl, bh); +tcg_gen_sub2_i64 (dl, dh, dl, dh, tl, zero); +tcg_temp_free_i64 (tl); +tcg_temp_free_i64 (zero); +} + +static DisasJumpType op_vsbi (DisasContext *s,DisasOps *o) +{ +if (get_field (s->fields, m5) !=ES_128 ) { +gen_program_exception (s,PGM_SPECIFICATION ); +return DISAS_NORETURN ; + } + +gen_gvec128_4_i64 (gen_sbi2_i64,get_field (s->fields, v1), +get_field (s->fields, v2),get_field (s->fields, v3), +get_field (s->fields, v4)); +return DISAS_NEXT ; +}
Signature · 1 line
2.20.1
PATCH v3 36/40 s390x/tcg: Implement VECTOR SUBTRACT WITH BORROW COMPUTE BORROW INDICATION
In reply to David HildenbrandMostly courtesy of Richard H.
--- target/s390x/insn-data.def | 2 ++ target/s390x/translate_vx.inc.c | 34 +++++++++++++++++++++++++++++++++ 2 files changed, 36 insertions(+)
target/s390x/insn-data.def +2 -0
@@ -1180,6 +1180,8 @@ F(0xe7f5, VSCBI, VRR_c, V, 0, 0, 0, 0, vscbi, 0, IF_VEC) /* VECTOR SUBTRACT WITH BORROW INDICATION */ F(0xe7bf, VSBI, VRR_d, V, 0, 0, 0, 0, vsbi, 0, IF_VEC) +/* VECTOR SUBTRACT WITH BORROW COMPUTE BORROW INDICATION */ + F(0xe7bd, VSBCBI, VRR_d, V, 0, 0, 0, 0, vsbcbi, 0, IF_VEC) #ifndef CONFIG_USER_ONLY /* COMPARE AND SWAP AND PURGE */
target/s390x/translate_vx.inc.c +34 -0
@@ -2214,3 +2214,37 @@ static DisasJumpType op_vsbi(DisasContext *s, DisasOps *o) get_field(s->fields, v4));return DISAS_NEXT ; } + +static void gen_sbcbi2_i64 (TCGv_i64 dl,TCGv_i64 dh,TCGv_i64 al,TCGv_i64 ah, +TCGv_i64 bl,TCGv_i64 bh,TCGv_i64 cl,TCGv_i64 ch) +{ +TCGv_i64 th =tcg_temp_new_i64 (); +TCGv_i64 tl =tcg_temp_new_i64 (); +TCGv_i64 zero =tcg_const_i64 (0 ); + +tcg_gen_andi_i64 (tl, cl,1 ); +tcg_gen_sub2_i64 (tl, th, al, zero, tl, zero); +tcg_gen_sub2_i64 (tl, th, tl, th, bl, zero); +tcg_gen_andi_i64 (th, th,1 ); +tcg_gen_sub2_i64 (tl, th, ah, zero, th, zero); +tcg_gen_sub2_i64 (tl, th, tl, th, bh, zero); +tcg_gen_andi_i64 (dl, th,1 ); +tcg_gen_mov_i64 (dh, zero); + +tcg_temp_free_i64 (tl); +tcg_temp_free_i64 (th); +tcg_temp_free_i64 (zero); +} + +static DisasJumpType op_vsbcbi (DisasContext *s,DisasOps *o) +{ +if (get_field (s->fields, m5) !=ES_128 ) { +gen_program_exception (s,PGM_SPECIFICATION ); +return DISAS_NORETURN ; + } + +gen_gvec128_4_i64 (gen_sbcbi2_i64,get_field (s->fields, v1), +get_field (s->fields, v2),get_field (s->fields, v3), +get_field (s->fields, v4)); +return DISAS_NEXT ; +}
Signature · 1 line
2.20.1
Re: [Qemu-devel] [PATCH v3 36/40] s390x/tcg: Implement VECTOR SUBTRACT WITH BORROW COMPUTE BORROW INDICATION
Quoted 7 lines · David Hildenbrand
Mostly courtesy of Richard H.
Signed-off-by David Hildenbrand <[email protected]>--- target/s390x/insn-data.def | 2 ++ target/s390x/translate_vx.inc.c | 34 +++++++++++++++++++++++++++++++++ 2 files changed, 36 insertions(+)
r~
PATCH v3 37/40 s390x/tcg: Implement VECTOR SUM ACROSS DOUBLEWORD
In reply to David HildenbrandPerform the calculations without a helper. Only 16 bit or 32 bit values have to be added.
--- target/s390x/insn-data.def | 2 ++ target/s390x/translate_vx.inc.c | 29 +++++++++++++++++++++++++++++ 2 files changed, 31 insertions(+)
target/s390x/insn-data.def +2 -0
@@ -1182,6 +1182,8 @@ F(0xe7bf, VSBI, VRR_d, V, 0, 0, 0, 0, vsbi, 0, IF_VEC) /* VECTOR SUBTRACT WITH BORROW COMPUTE BORROW INDICATION */ F(0xe7bd, VSBCBI, VRR_d, V, 0, 0, 0, 0, vsbcbi, 0, IF_VEC) +/* VECTOR SUM ACROSS DOUBLEWORD */ + F(0xe765, VSUMG, VRR_c, V, 0, 0, 0, 0, vsumg, 0, IF_VEC) #ifndef CONFIG_USER_ONLY /* COMPARE AND SWAP AND PURGE */
target/s390x/translate_vx.inc.c +29 -0
@@ -2248,3 +2248,32 @@ static DisasJumpType op_vsbcbi(DisasContext *s, DisasOps *o) get_field(s->fields, v4));return DISAS_NEXT ; } + +static DisasJumpType op_vsumg (DisasContext *s,DisasOps *o) +{ +const uint8_t es =get_field (s->fields, m4); +TCGv_i64 sum, tmp; +uint8_t dst_idx; + +if (es ==ES_8 || es >ES_32 ) { +gen_program_exception (s,PGM_SPECIFICATION ); +return DISAS_NORETURN ; + } + + sum =tcg_temp_new_i64 (); + tmp =tcg_temp_new_i64 (); +for (dst_idx =0 ; dst_idx <2 ; dst_idx++) { +uint8_t idx = dst_idx *NUM_VEC_ELEMENTS (es) /2 ; +const uint8_t max_idx = idx +NUM_VEC_ELEMENTS (es) /2 -1 ; + +read_vec_element_i64 (sum,get_field (s->fields, v3), max_idx, es); +for (; idx <= max_idx; idx++) { +read_vec_element_i64 (tmp,get_field (s->fields, v2), idx, es); +tcg_gen_add_i64 (sum, sum, tmp); + } +write_vec_element_i64 (sum,get_field (s->fields, v1), dst_idx,ES_64 ); + } +tcg_temp_free_i64 (sum); +tcg_temp_free_i64 (tmp); +return DISAS_NEXT ; +}
Signature · 1 line
2.20.1
PATCH v3 38/40 s390x/tcg: Implement VECTOR SUM ACROSS QUADWORD
In reply to David HildenbrandSimilar to VECTOR SUM ACROSS DOUBLEWORD, however without a loop and using 128-bit calculations.
--- target/s390x/insn-data.def | 2 ++ target/s390x/translate_vx.inc.c | 32 ++++++++++++++++++++++++++++++++ 2 files changed, 34 insertions(+)
target/s390x/insn-data.def +2 -0
@@ -1184,6 +1184,8 @@ F(0xe7bd, VSBCBI, VRR_d, V, 0, 0, 0, 0, vsbcbi, 0, IF_VEC) /* VECTOR SUM ACROSS DOUBLEWORD */ F(0xe765, VSUMG, VRR_c, V, 0, 0, 0, 0, vsumg, 0, IF_VEC) +/* VECTOR SUM ACROSS QUADWORD */ + F(0xe767, VSUMQ, VRR_c, V, 0, 0, 0, 0, vsumq, 0, IF_VEC) #ifndef CONFIG_USER_ONLY /* COMPARE AND SWAP AND PURGE */
target/s390x/translate_vx.inc.c +32 -0
@@ -2277,3 +2277,35 @@ static DisasJumpType op_vsumg(DisasContext *s, DisasOps *o)tcg_temp_free_i64 (tmp);return DISAS_NEXT ; } + +static DisasJumpType op_vsumq (DisasContext *s,DisasOps *o) +{ +const uint8_t es =get_field (s->fields, m4); +const uint8_t max_idx =NUM_VEC_ELEMENTS (es) -1 ; +TCGv_i64 sumh, suml, zero, tmpl; +uint8_t idx; + +if (es <ES_32 || es >ES_64 ) { +gen_program_exception (s,PGM_SPECIFICATION ); +return DISAS_NORETURN ; + } + + sumh =tcg_const_i64 (0 ); + suml =tcg_temp_new_i64 (); + zero =tcg_const_i64 (0 ); + tmpl =tcg_temp_new_i64 (); + +read_vec_element_i64 (suml,get_field (s->fields, v3), max_idx, es); +for (idx =0 ; idx <= max_idx; idx++) { +read_vec_element_i64 (tmpl,get_field (s->fields, v2), idx, es); +tcg_gen_add2_i64 (suml, sumh, suml, sumh, tmpl, zero); + } +write_vec_element_i64 (sumh,get_field (s->fields, v1),0 ,ES_64 ); +write_vec_element_i64 (suml,get_field (s->fields, v1),1 ,ES_64 ); + +tcg_temp_free_i64 (sumh); +tcg_temp_free_i64 (suml); +tcg_temp_free_i64 (zero); +tcg_temp_free_i64 (tmpl); +return DISAS_NEXT ; +}
Signature · 1 line
2.20.1
PATCH v3 39/40 s390x/tcg: Implement VECTOR SUM ACROSS WORD
In reply to David HildenbrandSimilar to VECTOR SUM ACROSS DOUBLEWORD.
--- target/s390x/insn-data.def | 2 ++ target/s390x/translate_vx.inc.c | 29 +++++++++++++++++++++++++++++ 2 files changed, 31 insertions(+)
target/s390x/insn-data.def +2 -0
@@ -1186,6 +1186,8 @@ F(0xe765, VSUMG, VRR_c, V, 0, 0, 0, 0, vsumg, 0, IF_VEC) /* VECTOR SUM ACROSS QUADWORD */ F(0xe767, VSUMQ, VRR_c, V, 0, 0, 0, 0, vsumq, 0, IF_VEC) +/* VECTOR SUM ACROSS WORD */ + F(0xe764, VSUM, VRR_c, V, 0, 0, 0, 0, vsum, 0, IF_VEC) #ifndef CONFIG_USER_ONLY /* COMPARE AND SWAP AND PURGE */
target/s390x/translate_vx.inc.c +29 -0
@@ -2309,3 +2309,32 @@ static DisasJumpType op_vsumq(DisasContext *s, DisasOps *o)tcg_temp_free_i64 (tmpl);return DISAS_NEXT ; } + +static DisasJumpType op_vsum (DisasContext *s,DisasOps *o) +{ +const uint8_t es =get_field (s->fields, m4); +TCGv_i32 sum, tmp; +uint8_t dst_idx; + +if (es >ES_16 ) { +gen_program_exception (s,PGM_SPECIFICATION ); +return DISAS_NORETURN ; + } + + sum =tcg_temp_new_i32 (); + tmp =tcg_temp_new_i32 (); +for (dst_idx =0 ; dst_idx <4 ; dst_idx++) { +uint8_t idx = dst_idx *NUM_VEC_ELEMENTS (es) /4 ; +const uint8_t max_idx = idx +NUM_VEC_ELEMENTS (es) /4 -1 ; + +read_vec_element_i32 (sum,get_field (s->fields, v3), max_idx, es); +for (; idx <= max_idx; idx++) { +read_vec_element_i32 (tmp,get_field (s->fields, v2), idx, es); +tcg_gen_add_i32 (sum, sum, tmp); + } +write_vec_element_i32 (sum,get_field (s->fields, v1), dst_idx,ES_32 ); + } +tcg_temp_free_i32 (sum); +tcg_temp_free_i32 (tmp); +return DISAS_NEXT ; +}
Signature · 1 line
2.20.1
PATCH v3 40/40 s390x/tcg: Implement VECTOR TEST UNDER MASK
In reply to David HildenbrandLet's return the cc value directly via cpu_env. Unfortunately there isn't a simple way to calculate the value lazily - one would have to calculate and store e.g. the population count of the mask and the result so it can be evaluated in a cc helper.
But as VTM only sets the cc, we can assume the value will be needed soon either way.
--- target/s390x/helper.h | 1 + target/s390x/insn-data.def | 2 ++ target/s390x/translate_vx.inc.c | 11 +++++++++++ target/s390x/vec_int_helper.c | 18 ++++++++++++++++++ 4 files changed, 32 insertions(+)
target/s390x/helper.h +1 -0
@@ -209,6 +209,7 @@ DEF_HELPER_FLAGS_4(gvec_vsra, TCG_CALL_NO_RWG, void, ptr, cptr, i64, i32) DEF_HELPER_FLAGS_4(gvec_vsrl, TCG_CALL_NO_RWG, void, ptr, cptr, i64, 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)#ifndef CONFIG_USER_ONLY DEF_HELPER_3 (servc, i32, env, i64, i64)
target/s390x/insn-data.def +2 -0
@@ -1188,6 +1188,8 @@ F(0xe767, VSUMQ, VRR_c, V, 0, 0, 0, 0, vsumq, 0, IF_VEC) /* VECTOR SUM ACROSS WORD */ F(0xe764, VSUM, VRR_c, V, 0, 0, 0, 0, vsum, 0, IF_VEC) +/* VECTOR TEST UNDER MASK */ + F(0xe7d8, VTM, VRR_a, V, 0, 0, 0, 0, vtm, 0, IF_VEC) #ifndef CONFIG_USER_ONLY /* COMPARE AND SWAP AND PURGE */
target/s390x/translate_vx.inc.c +11 -0
@@ -191,6 +191,9 @@ static void get_vec_element_ptr_i64(TCGv_ptr ptr, uint8_t reg, TCGv_i64 enr,#define gen_gvec_2i_ool (v1, v2, c, data, fn) \ tcg_gen_gvec_2i_ool(vec_full_reg_offset(v1), vec_full_reg_offset(v2), \ c, 16, 16, data, fn) +#define gen_gvec_2_ptr (v1, v2, ptr, data, fn) \ + tcg_gen_gvec_2_ptr(vec_full_reg_offset(v1), vec_full_reg_offset(v2), \ + ptr, 16, 16, data, fn)#define gen_gvec_3 (v1, v2, v3, gen) \ tcg_gen_gvec_3(vec_full_reg_offset(v1), vec_full_reg_offset(v2), \ vec_full_reg_offset(v3), 16, 16, gen)
@@ -2338,3 +2341,11 @@ static DisasJumpType op_vsum(DisasContext *s, DisasOps *o)tcg_temp_free_i32 (tmp);return DISAS_NEXT ; } + +static DisasJumpType op_vtm (DisasContext *s,DisasOps *o) +{ +gen_gvec_2_ptr (get_field (s->fields, v1),get_field (s->fields, v2), + cpu_env,0 , gen_helper_gvec_vtm); +set_cc_static (s); +return DISAS_NEXT ; +}
target/s390x/vec_int_helper.c +18 -0
@@ -583,3 +583,21 @@ void HELPER(gvec_vscbi##BITS)(void *v1, const void *v2, const void *v3, \ } DEF_VSCBI(8) DEF_VSCBI(16) + +void HELPER (gvec_vtm )(void *v1,const void *v2,CPUS390XState *env, +uint32_t desc) +{ +S390Vector tmp; + +s390_vec_and (&tmp, v1, v2); +if (s390_vec_is_zero (&tmp)) { +/* Selected bits all zeros; or all mask bits zero */ + env->cc_op =0 ; + }else if (s390_vec_equal (&tmp, v2)) { +/* Selected bits all ones */ + env->cc_op =3 ; + }else { +/* Selected bits a mix of zeros and ones */ + env->cc_op =1 ; + } +}
Signature · 1 line
2.20.1
Re: [Qemu-devel] [PATCH v3 40/40] s390x/tcg: Implement VECTOR TEST UNDER MASK
Quoted 85 lines · David Hildenbrand
Let's return the cc value directly via
cpu_env. Unfortunately there isn't a simple way to calculate the value lazily - one would have to calculate and store e.g. the population count of the mask and the result so it can be evaluated in a cc helper.But as VTM only sets the cc, we can assume the value will be needed soon either way.
Reviewed-by Richard Henderson <[email protected]>Signed-off-by David Hildenbrand <[email protected]>--- target/s390x/helper.h | 1 + target/s390x/insn-data.def | 2 ++ target/s390x/translate_vx.inc.c | 11 +++++++++++ target/s390x/vec_int_helper.c | 18 ++++++++++++++++++ 4 files changed, 32 insertions(+)target/s390x/helper.h +1
-0@@ -209,6 +209,7 @@ DEF_HELPER_FLAGS_4(gvec_vsra, TCG_CALL_NO_RWG, void, ptr, cptr, i64, i32) DEF_HELPER_FLAGS_4(gvec_vsrl, TCG_CALL_NO_RWG, void, ptr, cptr, i64, 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)#ifndef CONFIG_USER_ONLY DEF_HELPER_3 (servc, i32, env, i64, i64)target/s390x/insn-data.def +2
-0@@ -1188,6 +1188,8 @@ F(0xe767, VSUMQ, VRR_c, V, 0, 0, 0, 0, vsumq, 0, IF_VEC) /* VECTOR SUM ACROSS WORD */ F(0xe764, VSUM, VRR_c, V, 0, 0, 0, 0, vsum, 0, IF_VEC) +/* VECTOR TEST UNDER MASK */ + F(0xe7d8, VTM, VRR_a, V, 0, 0, 0, 0, vtm, 0, IF_VEC) #ifndef CONFIG_USER_ONLY /* COMPARE AND SWAP AND PURGE */target/s390x/translate_vx.inc.c +11
-0@@ -191,6 +191,9 @@ static void get_vec_element_ptr_i64(TCGv_ptr ptr, uint8_t reg, TCGv_i64 enr,#define gen_gvec_2i_ool (v1, v2, c, data, fn) \ tcg_gen_gvec_2i_ool(vec_full_reg_offset(v1), vec_full_reg_offset(v2), \ c, 16, 16, data, fn) +#define gen_gvec_2_ptr (v1, v2, ptr, data, fn) \ + tcg_gen_gvec_2_ptr(vec_full_reg_offset(v1), vec_full_reg_offset(v2), \ + ptr, 16, 16, data, fn)#define gen_gvec_3 (v1, v2, v3, gen) \ tcg_gen_gvec_3(vec_full_reg_offset(v1), vec_full_reg_offset(v2), \ vec_full_reg_offset(v3), 16, 16, gen)@@ -2338,3 +2341,11 @@ static DisasJumpType op_vsum(DisasContext *s, DisasOps *o)tcg_temp_free_i32 (tmp);return DISAS_NEXT ; } + +static DisasJumpType op_vtm (DisasContext *s,DisasOps *o) +{ +gen_gvec_2_ptr (get_field (s->fields, v1),get_field (s->fields, v2), + cpu_env,0 , gen_helper_gvec_vtm); +set_cc_static (s); +return DISAS_NEXT ; +}target/s390x/vec_int_helper.c +8
-0@@ -583,3 +583,21 @@ void HELPER(gvec_vscbi##BITS)(void *v1, const void *v2, const void *v3, \ } DEF_VSCBI(8) DEF_VSCBI(16) + +void HELPER (gvec_vtm )(void *v1,const void *v2,CPUS390XState *env, +uint32_t desc) +{ +S390Vector tmp; + +s390_vec_and (&tmp, v1, v2); +if (s390_vec_is_zero (&tmp)) {
target/s390x/vec_int_helper.c void HELPER(gvec_vscbi##BITS)(void *v1, const void *v2, const void *v3, \ · line 603
Quoted 10 lines
target/s390x/vec_int_helper.c +10
-0+/* Selected bits all zeros; or all mask bits zero */ + env->cc_op =0 ; + }else if (s390_vec_equal (&tmp, v2)) { +/* Selected bits all ones */ + env->cc_op =3 ; + }else { +/* Selected bits a mix of zeros and ones */ + env->cc_op =1 ; + } +}
This hunk got lost in follow-up patches:
target/s390x/vec_int_helper.c +0 -0
@@ -28,6 +28,19 @@ static void s390_vec_xor(S390Vector *res, const
S390Vector *a,
res->doubleword[1] = a->doubleword[1] ^ b->doubleword[1];
}
+static void s390_vec_and(S390Vector *res, const S390Vector *a,
+ const S390Vector *b)
+{
+ res->doubleword[0] = a->doubleword[0] & b->doubleword[0];
+ res->doubleword[1] = a->doubleword[1] & b->doubleword[1];
+}
+
+static bool s390_vec_equal(const S390Vector *a, const S390Vector *b) +{
+ return a->doubleword[0] == b->doubleword[0] &&
+ a->doubleword[1] == b->doubleword[1];
+}
+
static void s390_vec_shl(S390Vector *d, const S390Vector *a, uint64_t
count)
{
uint64_t tmp;Signature · 3 lines
Thanks, David / dhildenb