qemu/target/riscv/vcrypto_helper.c

Ignoring revisions in .git-blame-ignore-revs. Click here to bypass and see the normal blame view.

1003 lines
33 KiB
C
Raw Permalink Normal View History

/*
* RISC-V Vector Crypto Extension Helpers for QEMU.
*
* Copyright (C) 2023 SiFive, Inc.
* Written by Codethink Ltd and SiFive.
*
* This program is free software; you can redistribute it and/or modify it
* under the terms and conditions of the GNU General Public License,
* version 2 or later, as published by the Free Software Foundation.
*
* This program is distributed in the hope it will be useful, but WITHOUT
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
* FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
* more details.
*
* You should have received a copy of the GNU General Public License along with
* this program. If not, see <http://www.gnu.org/licenses/>.
*/
#include "qemu/osdep.h"
#include "qemu/host-utils.h"
#include "qemu/bitops.h"
#include "qemu/bswap.h"
#include "cpu.h"
#include "crypto/aes.h"
#include "crypto/aes-round.h"
#include "crypto/sm4.h"
#include "exec/memop.h"
#include "exec/exec-all.h"
#include "exec/helper-proto.h"
#include "internals.h"
#include "vector_internals.h"
static uint64_t clmul64(uint64_t y, uint64_t x)
{
uint64_t result = 0;
for (int j = 63; j >= 0; j--) {
if ((y >> j) & 1) {
result ^= (x << j);
}
}
return result;
}
static uint64_t clmulh64(uint64_t y, uint64_t x)
{
uint64_t result = 0;
for (int j = 63; j >= 1; j--) {
if ((y >> j) & 1) {
result ^= (x >> (64 - j));
}
}
return result;
}
RVVCALL(OPIVV2, vclmul_vv, OP_UUU_D, H8, H8, H8, clmul64)
GEN_VEXT_VV(vclmul_vv, 8)
RVVCALL(OPIVX2, vclmul_vx, OP_UUU_D, H8, H8, clmul64)
GEN_VEXT_VX(vclmul_vx, 8)
RVVCALL(OPIVV2, vclmulh_vv, OP_UUU_D, H8, H8, H8, clmulh64)
GEN_VEXT_VV(vclmulh_vv, 8)
RVVCALL(OPIVX2, vclmulh_vx, OP_UUU_D, H8, H8, clmulh64)
GEN_VEXT_VX(vclmulh_vx, 8)
RVVCALL(OPIVV2, vror_vv_b, OP_UUU_B, H1, H1, H1, ror8)
RVVCALL(OPIVV2, vror_vv_h, OP_UUU_H, H2, H2, H2, ror16)
RVVCALL(OPIVV2, vror_vv_w, OP_UUU_W, H4, H4, H4, ror32)
RVVCALL(OPIVV2, vror_vv_d, OP_UUU_D, H8, H8, H8, ror64)
GEN_VEXT_VV(vror_vv_b, 1)
GEN_VEXT_VV(vror_vv_h, 2)
GEN_VEXT_VV(vror_vv_w, 4)
GEN_VEXT_VV(vror_vv_d, 8)
RVVCALL(OPIVX2, vror_vx_b, OP_UUU_B, H1, H1, ror8)
RVVCALL(OPIVX2, vror_vx_h, OP_UUU_H, H2, H2, ror16)
RVVCALL(OPIVX2, vror_vx_w, OP_UUU_W, H4, H4, ror32)
RVVCALL(OPIVX2, vror_vx_d, OP_UUU_D, H8, H8, ror64)
GEN_VEXT_VX(vror_vx_b, 1)
GEN_VEXT_VX(vror_vx_h, 2)
GEN_VEXT_VX(vror_vx_w, 4)
GEN_VEXT_VX(vror_vx_d, 8)
RVVCALL(OPIVV2, vrol_vv_b, OP_UUU_B, H1, H1, H1, rol8)
RVVCALL(OPIVV2, vrol_vv_h, OP_UUU_H, H2, H2, H2, rol16)
RVVCALL(OPIVV2, vrol_vv_w, OP_UUU_W, H4, H4, H4, rol32)
RVVCALL(OPIVV2, vrol_vv_d, OP_UUU_D, H8, H8, H8, rol64)
GEN_VEXT_VV(vrol_vv_b, 1)
GEN_VEXT_VV(vrol_vv_h, 2)
GEN_VEXT_VV(vrol_vv_w, 4)
GEN_VEXT_VV(vrol_vv_d, 8)
RVVCALL(OPIVX2, vrol_vx_b, OP_UUU_B, H1, H1, rol8)
RVVCALL(OPIVX2, vrol_vx_h, OP_UUU_H, H2, H2, rol16)
RVVCALL(OPIVX2, vrol_vx_w, OP_UUU_W, H4, H4, rol32)
RVVCALL(OPIVX2, vrol_vx_d, OP_UUU_D, H8, H8, rol64)
GEN_VEXT_VX(vrol_vx_b, 1)
GEN_VEXT_VX(vrol_vx_h, 2)
GEN_VEXT_VX(vrol_vx_w, 4)
GEN_VEXT_VX(vrol_vx_d, 8)
static uint64_t brev8(uint64_t val)
{
val = ((val & 0x5555555555555555ull) << 1) |
((val & 0xAAAAAAAAAAAAAAAAull) >> 1);
val = ((val & 0x3333333333333333ull) << 2) |
((val & 0xCCCCCCCCCCCCCCCCull) >> 2);
val = ((val & 0x0F0F0F0F0F0F0F0Full) << 4) |
((val & 0xF0F0F0F0F0F0F0F0ull) >> 4);
return val;
}
RVVCALL(OPIVV1, vbrev8_v_b, OP_UU_B, H1, H1, brev8)
RVVCALL(OPIVV1, vbrev8_v_h, OP_UU_H, H2, H2, brev8)
RVVCALL(OPIVV1, vbrev8_v_w, OP_UU_W, H4, H4, brev8)
RVVCALL(OPIVV1, vbrev8_v_d, OP_UU_D, H8, H8, brev8)
GEN_VEXT_V(vbrev8_v_b, 1)
GEN_VEXT_V(vbrev8_v_h, 2)
GEN_VEXT_V(vbrev8_v_w, 4)
GEN_VEXT_V(vbrev8_v_d, 8)
#define DO_IDENTITY(a) (a)
RVVCALL(OPIVV1, vrev8_v_b, OP_UU_B, H1, H1, DO_IDENTITY)
RVVCALL(OPIVV1, vrev8_v_h, OP_UU_H, H2, H2, bswap16)
RVVCALL(OPIVV1, vrev8_v_w, OP_UU_W, H4, H4, bswap32)
RVVCALL(OPIVV1, vrev8_v_d, OP_UU_D, H8, H8, bswap64)
GEN_VEXT_V(vrev8_v_b, 1)
GEN_VEXT_V(vrev8_v_h, 2)
GEN_VEXT_V(vrev8_v_w, 4)
GEN_VEXT_V(vrev8_v_d, 8)
#define DO_ANDN(a, b) ((a) & ~(b))
RVVCALL(OPIVV2, vandn_vv_b, OP_UUU_B, H1, H1, H1, DO_ANDN)
RVVCALL(OPIVV2, vandn_vv_h, OP_UUU_H, H2, H2, H2, DO_ANDN)
RVVCALL(OPIVV2, vandn_vv_w, OP_UUU_W, H4, H4, H4, DO_ANDN)
RVVCALL(OPIVV2, vandn_vv_d, OP_UUU_D, H8, H8, H8, DO_ANDN)
GEN_VEXT_VV(vandn_vv_b, 1)
GEN_VEXT_VV(vandn_vv_h, 2)
GEN_VEXT_VV(vandn_vv_w, 4)
GEN_VEXT_VV(vandn_vv_d, 8)
RVVCALL(OPIVX2, vandn_vx_b, OP_UUU_B, H1, H1, DO_ANDN)
RVVCALL(OPIVX2, vandn_vx_h, OP_UUU_H, H2, H2, DO_ANDN)
RVVCALL(OPIVX2, vandn_vx_w, OP_UUU_W, H4, H4, DO_ANDN)
RVVCALL(OPIVX2, vandn_vx_d, OP_UUU_D, H8, H8, DO_ANDN)
GEN_VEXT_VX(vandn_vx_b, 1)
GEN_VEXT_VX(vandn_vx_h, 2)
GEN_VEXT_VX(vandn_vx_w, 4)
GEN_VEXT_VX(vandn_vx_d, 8)
RVVCALL(OPIVV1, vbrev_v_b, OP_UU_B, H1, H1, revbit8)
RVVCALL(OPIVV1, vbrev_v_h, OP_UU_H, H2, H2, revbit16)
RVVCALL(OPIVV1, vbrev_v_w, OP_UU_W, H4, H4, revbit32)
RVVCALL(OPIVV1, vbrev_v_d, OP_UU_D, H8, H8, revbit64)
GEN_VEXT_V(vbrev_v_b, 1)
GEN_VEXT_V(vbrev_v_h, 2)
GEN_VEXT_V(vbrev_v_w, 4)
GEN_VEXT_V(vbrev_v_d, 8)
RVVCALL(OPIVV1, vclz_v_b, OP_UU_B, H1, H1, clz8)
RVVCALL(OPIVV1, vclz_v_h, OP_UU_H, H2, H2, clz16)
RVVCALL(OPIVV1, vclz_v_w, OP_UU_W, H4, H4, clz32)
RVVCALL(OPIVV1, vclz_v_d, OP_UU_D, H8, H8, clz64)
GEN_VEXT_V(vclz_v_b, 1)
GEN_VEXT_V(vclz_v_h, 2)
GEN_VEXT_V(vclz_v_w, 4)
GEN_VEXT_V(vclz_v_d, 8)
RVVCALL(OPIVV1, vctz_v_b, OP_UU_B, H1, H1, ctz8)
RVVCALL(OPIVV1, vctz_v_h, OP_UU_H, H2, H2, ctz16)
RVVCALL(OPIVV1, vctz_v_w, OP_UU_W, H4, H4, ctz32)
RVVCALL(OPIVV1, vctz_v_d, OP_UU_D, H8, H8, ctz64)
GEN_VEXT_V(vctz_v_b, 1)
GEN_VEXT_V(vctz_v_h, 2)
GEN_VEXT_V(vctz_v_w, 4)
GEN_VEXT_V(vctz_v_d, 8)
RVVCALL(OPIVV1, vcpop_v_b, OP_UU_B, H1, H1, ctpop8)
RVVCALL(OPIVV1, vcpop_v_h, OP_UU_H, H2, H2, ctpop16)
RVVCALL(OPIVV1, vcpop_v_w, OP_UU_W, H4, H4, ctpop32)
RVVCALL(OPIVV1, vcpop_v_d, OP_UU_D, H8, H8, ctpop64)
GEN_VEXT_V(vcpop_v_b, 1)
GEN_VEXT_V(vcpop_v_h, 2)
GEN_VEXT_V(vcpop_v_w, 4)
GEN_VEXT_V(vcpop_v_d, 8)
#define DO_SLL(N, M) (N << (M & (sizeof(N) * 8 - 1)))
RVVCALL(OPIVV2, vwsll_vv_b, WOP_UUU_B, H2, H1, H1, DO_SLL)
RVVCALL(OPIVV2, vwsll_vv_h, WOP_UUU_H, H4, H2, H2, DO_SLL)
RVVCALL(OPIVV2, vwsll_vv_w, WOP_UUU_W, H8, H4, H4, DO_SLL)
GEN_VEXT_VV(vwsll_vv_b, 2)
GEN_VEXT_VV(vwsll_vv_h, 4)
GEN_VEXT_VV(vwsll_vv_w, 8)
RVVCALL(OPIVX2, vwsll_vx_b, WOP_UUU_B, H2, H1, DO_SLL)
RVVCALL(OPIVX2, vwsll_vx_h, WOP_UUU_H, H4, H2, DO_SLL)
RVVCALL(OPIVX2, vwsll_vx_w, WOP_UUU_W, H8, H4, DO_SLL)
GEN_VEXT_VX(vwsll_vx_b, 2)
GEN_VEXT_VX(vwsll_vx_h, 4)
GEN_VEXT_VX(vwsll_vx_w, 8)
void HELPER(egs_check)(uint32_t egs, CPURISCVState *env)
{
uint32_t vl = env->vl;
uint32_t vstart = env->vstart;
if (vl % egs != 0 || vstart % egs != 0) {
riscv_raise_exception(env, RISCV_EXCP_ILLEGAL_INST, GETPC());
}
}
static inline void xor_round_key(AESState *round_state, AESState *round_key)
{
round_state->v = round_state->v ^ round_key->v;
}
#define GEN_ZVKNED_HELPER_VV(NAME, ...) \
void HELPER(NAME)(void *vd, void *vs2, CPURISCVState *env, \
uint32_t desc) \
{ \
uint32_t vl = env->vl; \
uint32_t total_elems = vext_get_total_elems(env, desc, 4); \
uint32_t vta = vext_vta(desc); \
\
target/riscv/vector_helpers: do early exit when vstart >= vl We're going to make changes that will required each helper to be responsible for the 'vstart' management, i.e. we will relieve the 'vstart < vl' assumption that helpers have today. Helpers are usually able to deal with vstart >= vl, i.e. doing nothing aside from setting vstart = 0 at the end, but the tail update functions will update the tail regardless of vstart being valid or not. Unifying the tail update process in a single function that would handle the vstart >= vl case isn't trivial (see [1] for more info). This patch takes a blunt approach: do an early exit in every single vector helper if vstart >= vl, unless the helper is guarded with vstart_eq_zero in the translation. For those cases the helper is ready to deal with cases where vl might be zero, i.e. throwing exceptions based on it like vcpop_m() and first_m(). Helpers that weren't changed: - vcpop_m(), vfirst_m(), vmsetm(), GEN_VEXT_VIOTA_M(): these are guarded directly with vstart_eq_zero; - GEN_VEXT_VCOMPRESS_VM(): guarded with vcompress_vm_check() that checks vstart_eq_zero; - GEN_VEXT_RED(): guarded with either reduction_check() or reduction_widen_check(), both check vstart_eq_zero; - GEN_VEXT_FRED(): guarded with either freduction_check() or freduction_widen_check(), both check vstart_eq_zero. Another exception is vext_ldst_whole(), who operates on effective vector length regardless of the current settings in vtype and vl. [1] https://lore.kernel.org/qemu-riscv/1590234b-0291-432a-a0fa-c5a6876097bc@linux.alibaba.com/ Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com> Acked-by: Alistair Francis <alistair.francis@wdc.com> Message-ID: <20240314175704.478276-7-dbarboza@ventanamicro.com> Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
2024-03-14 20:57:00 +03:00
VSTART_CHECK_EARLY_EXIT(env); \
\
for (uint32_t i = env->vstart / 4; i < env->vl / 4; i++) { \
AESState round_key; \
round_key.d[0] = *((uint64_t *)vs2 + H8(i * 2 + 0)); \
round_key.d[1] = *((uint64_t *)vs2 + H8(i * 2 + 1)); \
AESState round_state; \
round_state.d[0] = *((uint64_t *)vd + H8(i * 2 + 0)); \
round_state.d[1] = *((uint64_t *)vd + H8(i * 2 + 1)); \
__VA_ARGS__; \
*((uint64_t *)vd + H8(i * 2 + 0)) = round_state.d[0]; \
*((uint64_t *)vd + H8(i * 2 + 1)) = round_state.d[1]; \
} \
env->vstart = 0; \
/* set tail elements to 1s */ \
vext_set_elems_1s(vd, vta, vl * 4, total_elems * 4); \
}
#define GEN_ZVKNED_HELPER_VS(NAME, ...) \
void HELPER(NAME)(void *vd, void *vs2, CPURISCVState *env, \
uint32_t desc) \
{ \
uint32_t vl = env->vl; \
uint32_t total_elems = vext_get_total_elems(env, desc, 4); \
uint32_t vta = vext_vta(desc); \
\
target/riscv/vector_helpers: do early exit when vstart >= vl We're going to make changes that will required each helper to be responsible for the 'vstart' management, i.e. we will relieve the 'vstart < vl' assumption that helpers have today. Helpers are usually able to deal with vstart >= vl, i.e. doing nothing aside from setting vstart = 0 at the end, but the tail update functions will update the tail regardless of vstart being valid or not. Unifying the tail update process in a single function that would handle the vstart >= vl case isn't trivial (see [1] for more info). This patch takes a blunt approach: do an early exit in every single vector helper if vstart >= vl, unless the helper is guarded with vstart_eq_zero in the translation. For those cases the helper is ready to deal with cases where vl might be zero, i.e. throwing exceptions based on it like vcpop_m() and first_m(). Helpers that weren't changed: - vcpop_m(), vfirst_m(), vmsetm(), GEN_VEXT_VIOTA_M(): these are guarded directly with vstart_eq_zero; - GEN_VEXT_VCOMPRESS_VM(): guarded with vcompress_vm_check() that checks vstart_eq_zero; - GEN_VEXT_RED(): guarded with either reduction_check() or reduction_widen_check(), both check vstart_eq_zero; - GEN_VEXT_FRED(): guarded with either freduction_check() or freduction_widen_check(), both check vstart_eq_zero. Another exception is vext_ldst_whole(), who operates on effective vector length regardless of the current settings in vtype and vl. [1] https://lore.kernel.org/qemu-riscv/1590234b-0291-432a-a0fa-c5a6876097bc@linux.alibaba.com/ Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com> Acked-by: Alistair Francis <alistair.francis@wdc.com> Message-ID: <20240314175704.478276-7-dbarboza@ventanamicro.com> Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
2024-03-14 20:57:00 +03:00
VSTART_CHECK_EARLY_EXIT(env); \
\
for (uint32_t i = env->vstart / 4; i < env->vl / 4; i++) { \
AESState round_key; \
round_key.d[0] = *((uint64_t *)vs2 + H8(0)); \
round_key.d[1] = *((uint64_t *)vs2 + H8(1)); \
AESState round_state; \
round_state.d[0] = *((uint64_t *)vd + H8(i * 2 + 0)); \
round_state.d[1] = *((uint64_t *)vd + H8(i * 2 + 1)); \
__VA_ARGS__; \
*((uint64_t *)vd + H8(i * 2 + 0)) = round_state.d[0]; \
*((uint64_t *)vd + H8(i * 2 + 1)) = round_state.d[1]; \
} \
env->vstart = 0; \
/* set tail elements to 1s */ \
vext_set_elems_1s(vd, vta, vl * 4, total_elems * 4); \
}
GEN_ZVKNED_HELPER_VV(vaesef_vv, aesenc_SB_SR_AK(&round_state,
&round_state,
&round_key,
false);)
GEN_ZVKNED_HELPER_VS(vaesef_vs, aesenc_SB_SR_AK(&round_state,
&round_state,
&round_key,
false);)
GEN_ZVKNED_HELPER_VV(vaesdf_vv, aesdec_ISB_ISR_AK(&round_state,
&round_state,
&round_key,
false);)
GEN_ZVKNED_HELPER_VS(vaesdf_vs, aesdec_ISB_ISR_AK(&round_state,
&round_state,
&round_key,
false);)
GEN_ZVKNED_HELPER_VV(vaesem_vv, aesenc_SB_SR_MC_AK(&round_state,
&round_state,
&round_key,
false);)
GEN_ZVKNED_HELPER_VS(vaesem_vs, aesenc_SB_SR_MC_AK(&round_state,
&round_state,
&round_key,
false);)
GEN_ZVKNED_HELPER_VV(vaesdm_vv, aesdec_ISB_ISR_AK_IMC(&round_state,
&round_state,
&round_key,
false);)
GEN_ZVKNED_HELPER_VS(vaesdm_vs, aesdec_ISB_ISR_AK_IMC(&round_state,
&round_state,
&round_key,
false);)
GEN_ZVKNED_HELPER_VS(vaesz_vs, xor_round_key(&round_state, &round_key);)
void HELPER(vaeskf1_vi)(void *vd_vptr, void *vs2_vptr, uint32_t uimm,
CPURISCVState *env, uint32_t desc)
{
uint32_t *vd = vd_vptr;
uint32_t *vs2 = vs2_vptr;
uint32_t vl = env->vl;
uint32_t total_elems = vext_get_total_elems(env, desc, 4);
uint32_t vta = vext_vta(desc);
target/riscv/vector_helpers: do early exit when vstart >= vl We're going to make changes that will required each helper to be responsible for the 'vstart' management, i.e. we will relieve the 'vstart < vl' assumption that helpers have today. Helpers are usually able to deal with vstart >= vl, i.e. doing nothing aside from setting vstart = 0 at the end, but the tail update functions will update the tail regardless of vstart being valid or not. Unifying the tail update process in a single function that would handle the vstart >= vl case isn't trivial (see [1] for more info). This patch takes a blunt approach: do an early exit in every single vector helper if vstart >= vl, unless the helper is guarded with vstart_eq_zero in the translation. For those cases the helper is ready to deal with cases where vl might be zero, i.e. throwing exceptions based on it like vcpop_m() and first_m(). Helpers that weren't changed: - vcpop_m(), vfirst_m(), vmsetm(), GEN_VEXT_VIOTA_M(): these are guarded directly with vstart_eq_zero; - GEN_VEXT_VCOMPRESS_VM(): guarded with vcompress_vm_check() that checks vstart_eq_zero; - GEN_VEXT_RED(): guarded with either reduction_check() or reduction_widen_check(), both check vstart_eq_zero; - GEN_VEXT_FRED(): guarded with either freduction_check() or freduction_widen_check(), both check vstart_eq_zero. Another exception is vext_ldst_whole(), who operates on effective vector length regardless of the current settings in vtype and vl. [1] https://lore.kernel.org/qemu-riscv/1590234b-0291-432a-a0fa-c5a6876097bc@linux.alibaba.com/ Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com> Acked-by: Alistair Francis <alistair.francis@wdc.com> Message-ID: <20240314175704.478276-7-dbarboza@ventanamicro.com> Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
2024-03-14 20:57:00 +03:00
VSTART_CHECK_EARLY_EXIT(env);
uimm &= 0b1111;
if (uimm > 10 || uimm == 0) {
uimm ^= 0b1000;
}
for (uint32_t i = env->vstart / 4; i < env->vl / 4; i++) {
uint32_t rk[8], tmp;
static const uint32_t rcon[] = {
0x00000001, 0x00000002, 0x00000004, 0x00000008, 0x00000010,
0x00000020, 0x00000040, 0x00000080, 0x0000001B, 0x00000036,
};
rk[0] = vs2[i * 4 + H4(0)];
rk[1] = vs2[i * 4 + H4(1)];
rk[2] = vs2[i * 4 + H4(2)];
rk[3] = vs2[i * 4 + H4(3)];
tmp = ror32(rk[3], 8);
rk[4] = rk[0] ^ (((uint32_t)AES_sbox[(tmp >> 24) & 0xff] << 24) |
((uint32_t)AES_sbox[(tmp >> 16) & 0xff] << 16) |
((uint32_t)AES_sbox[(tmp >> 8) & 0xff] << 8) |
((uint32_t)AES_sbox[(tmp >> 0) & 0xff] << 0))
^ rcon[uimm - 1];
rk[5] = rk[1] ^ rk[4];
rk[6] = rk[2] ^ rk[5];
rk[7] = rk[3] ^ rk[6];
vd[i * 4 + H4(0)] = rk[4];
vd[i * 4 + H4(1)] = rk[5];
vd[i * 4 + H4(2)] = rk[6];
vd[i * 4 + H4(3)] = rk[7];
}
env->vstart = 0;
/* set tail elements to 1s */
vext_set_elems_1s(vd, vta, vl * 4, total_elems * 4);
}
void HELPER(vaeskf2_vi)(void *vd_vptr, void *vs2_vptr, uint32_t uimm,
CPURISCVState *env, uint32_t desc)
{
uint32_t *vd = vd_vptr;
uint32_t *vs2 = vs2_vptr;
uint32_t vl = env->vl;
uint32_t total_elems = vext_get_total_elems(env, desc, 4);
uint32_t vta = vext_vta(desc);
target/riscv/vector_helpers: do early exit when vstart >= vl We're going to make changes that will required each helper to be responsible for the 'vstart' management, i.e. we will relieve the 'vstart < vl' assumption that helpers have today. Helpers are usually able to deal with vstart >= vl, i.e. doing nothing aside from setting vstart = 0 at the end, but the tail update functions will update the tail regardless of vstart being valid or not. Unifying the tail update process in a single function that would handle the vstart >= vl case isn't trivial (see [1] for more info). This patch takes a blunt approach: do an early exit in every single vector helper if vstart >= vl, unless the helper is guarded with vstart_eq_zero in the translation. For those cases the helper is ready to deal with cases where vl might be zero, i.e. throwing exceptions based on it like vcpop_m() and first_m(). Helpers that weren't changed: - vcpop_m(), vfirst_m(), vmsetm(), GEN_VEXT_VIOTA_M(): these are guarded directly with vstart_eq_zero; - GEN_VEXT_VCOMPRESS_VM(): guarded with vcompress_vm_check() that checks vstart_eq_zero; - GEN_VEXT_RED(): guarded with either reduction_check() or reduction_widen_check(), both check vstart_eq_zero; - GEN_VEXT_FRED(): guarded with either freduction_check() or freduction_widen_check(), both check vstart_eq_zero. Another exception is vext_ldst_whole(), who operates on effective vector length regardless of the current settings in vtype and vl. [1] https://lore.kernel.org/qemu-riscv/1590234b-0291-432a-a0fa-c5a6876097bc@linux.alibaba.com/ Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com> Acked-by: Alistair Francis <alistair.francis@wdc.com> Message-ID: <20240314175704.478276-7-dbarboza@ventanamicro.com> Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
2024-03-14 20:57:00 +03:00
VSTART_CHECK_EARLY_EXIT(env);
uimm &= 0b1111;
if (uimm > 14 || uimm < 2) {
uimm ^= 0b1000;
}
for (uint32_t i = env->vstart / 4; i < env->vl / 4; i++) {
uint32_t rk[12], tmp;
static const uint32_t rcon[] = {
0x00000001, 0x00000002, 0x00000004, 0x00000008, 0x00000010,
0x00000020, 0x00000040, 0x00000080, 0x0000001B, 0x00000036,
};
rk[0] = vd[i * 4 + H4(0)];
rk[1] = vd[i * 4 + H4(1)];
rk[2] = vd[i * 4 + H4(2)];
rk[3] = vd[i * 4 + H4(3)];
rk[4] = vs2[i * 4 + H4(0)];
rk[5] = vs2[i * 4 + H4(1)];
rk[6] = vs2[i * 4 + H4(2)];
rk[7] = vs2[i * 4 + H4(3)];
if (uimm % 2 == 0) {
tmp = ror32(rk[7], 8);
rk[8] = rk[0] ^ (((uint32_t)AES_sbox[(tmp >> 24) & 0xff] << 24) |
((uint32_t)AES_sbox[(tmp >> 16) & 0xff] << 16) |
((uint32_t)AES_sbox[(tmp >> 8) & 0xff] << 8) |
((uint32_t)AES_sbox[(tmp >> 0) & 0xff] << 0))
^ rcon[(uimm - 1) / 2];
} else {
rk[8] = rk[0] ^ (((uint32_t)AES_sbox[(rk[7] >> 24) & 0xff] << 24) |
((uint32_t)AES_sbox[(rk[7] >> 16) & 0xff] << 16) |
((uint32_t)AES_sbox[(rk[7] >> 8) & 0xff] << 8) |
((uint32_t)AES_sbox[(rk[7] >> 0) & 0xff] << 0));
}
rk[9] = rk[1] ^ rk[8];
rk[10] = rk[2] ^ rk[9];
rk[11] = rk[3] ^ rk[10];
vd[i * 4 + H4(0)] = rk[8];
vd[i * 4 + H4(1)] = rk[9];
vd[i * 4 + H4(2)] = rk[10];
vd[i * 4 + H4(3)] = rk[11];
}
env->vstart = 0;
/* set tail elements to 1s */
vext_set_elems_1s(vd, vta, vl * 4, total_elems * 4);
}
static inline uint32_t sig0_sha256(uint32_t x)
{
return ror32(x, 7) ^ ror32(x, 18) ^ (x >> 3);
}
static inline uint32_t sig1_sha256(uint32_t x)
{
return ror32(x, 17) ^ ror32(x, 19) ^ (x >> 10);
}
static inline uint64_t sig0_sha512(uint64_t x)
{
return ror64(x, 1) ^ ror64(x, 8) ^ (x >> 7);
}
static inline uint64_t sig1_sha512(uint64_t x)
{
return ror64(x, 19) ^ ror64(x, 61) ^ (x >> 6);
}
static inline void vsha2ms_e32(uint32_t *vd, uint32_t *vs1, uint32_t *vs2)
{
uint32_t res[4];
res[0] = sig1_sha256(vs1[H4(2)]) + vs2[H4(1)] + sig0_sha256(vd[H4(1)]) +
vd[H4(0)];
res[1] = sig1_sha256(vs1[H4(3)]) + vs2[H4(2)] + sig0_sha256(vd[H4(2)]) +
vd[H4(1)];
res[2] =
sig1_sha256(res[0]) + vs2[H4(3)] + sig0_sha256(vd[H4(3)]) + vd[H4(2)];
res[3] =
sig1_sha256(res[1]) + vs1[H4(0)] + sig0_sha256(vs2[H4(0)]) + vd[H4(3)];
vd[H4(3)] = res[3];
vd[H4(2)] = res[2];
vd[H4(1)] = res[1];
vd[H4(0)] = res[0];
}
static inline void vsha2ms_e64(uint64_t *vd, uint64_t *vs1, uint64_t *vs2)
{
uint64_t res[4];
res[0] = sig1_sha512(vs1[2]) + vs2[1] + sig0_sha512(vd[1]) + vd[0];
res[1] = sig1_sha512(vs1[3]) + vs2[2] + sig0_sha512(vd[2]) + vd[1];
res[2] = sig1_sha512(res[0]) + vs2[3] + sig0_sha512(vd[3]) + vd[2];
res[3] = sig1_sha512(res[1]) + vs1[0] + sig0_sha512(vs2[0]) + vd[3];
vd[3] = res[3];
vd[2] = res[2];
vd[1] = res[1];
vd[0] = res[0];
}
void HELPER(vsha2ms_vv)(void *vd, void *vs1, void *vs2, CPURISCVState *env,
uint32_t desc)
{
uint32_t sew = FIELD_EX64(env->vtype, VTYPE, VSEW);
uint32_t esz = sew == MO_32 ? 4 : 8;
uint32_t total_elems;
uint32_t vta = vext_vta(desc);
target/riscv/vector_helpers: do early exit when vstart >= vl We're going to make changes that will required each helper to be responsible for the 'vstart' management, i.e. we will relieve the 'vstart < vl' assumption that helpers have today. Helpers are usually able to deal with vstart >= vl, i.e. doing nothing aside from setting vstart = 0 at the end, but the tail update functions will update the tail regardless of vstart being valid or not. Unifying the tail update process in a single function that would handle the vstart >= vl case isn't trivial (see [1] for more info). This patch takes a blunt approach: do an early exit in every single vector helper if vstart >= vl, unless the helper is guarded with vstart_eq_zero in the translation. For those cases the helper is ready to deal with cases where vl might be zero, i.e. throwing exceptions based on it like vcpop_m() and first_m(). Helpers that weren't changed: - vcpop_m(), vfirst_m(), vmsetm(), GEN_VEXT_VIOTA_M(): these are guarded directly with vstart_eq_zero; - GEN_VEXT_VCOMPRESS_VM(): guarded with vcompress_vm_check() that checks vstart_eq_zero; - GEN_VEXT_RED(): guarded with either reduction_check() or reduction_widen_check(), both check vstart_eq_zero; - GEN_VEXT_FRED(): guarded with either freduction_check() or freduction_widen_check(), both check vstart_eq_zero. Another exception is vext_ldst_whole(), who operates on effective vector length regardless of the current settings in vtype and vl. [1] https://lore.kernel.org/qemu-riscv/1590234b-0291-432a-a0fa-c5a6876097bc@linux.alibaba.com/ Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com> Acked-by: Alistair Francis <alistair.francis@wdc.com> Message-ID: <20240314175704.478276-7-dbarboza@ventanamicro.com> Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
2024-03-14 20:57:00 +03:00
VSTART_CHECK_EARLY_EXIT(env);
for (uint32_t i = env->vstart / 4; i < env->vl / 4; i++) {
if (sew == MO_32) {
vsha2ms_e32(((uint32_t *)vd) + i * 4, ((uint32_t *)vs1) + i * 4,
((uint32_t *)vs2) + i * 4);
} else {
/* If not 32 then SEW should be 64 */
vsha2ms_e64(((uint64_t *)vd) + i * 4, ((uint64_t *)vs1) + i * 4,
((uint64_t *)vs2) + i * 4);
}
}
/* set tail elements to 1s */
total_elems = vext_get_total_elems(env, desc, esz);
vext_set_elems_1s(vd, vta, env->vl * esz, total_elems * esz);
env->vstart = 0;
}
static inline uint64_t sum0_64(uint64_t x)
{
return ror64(x, 28) ^ ror64(x, 34) ^ ror64(x, 39);
}
static inline uint32_t sum0_32(uint32_t x)
{
return ror32(x, 2) ^ ror32(x, 13) ^ ror32(x, 22);
}
static inline uint64_t sum1_64(uint64_t x)
{
return ror64(x, 14) ^ ror64(x, 18) ^ ror64(x, 41);
}
static inline uint32_t sum1_32(uint32_t x)
{
return ror32(x, 6) ^ ror32(x, 11) ^ ror32(x, 25);
}
#define ch(x, y, z) ((x & y) ^ ((~x) & z))
#define maj(x, y, z) ((x & y) ^ (x & z) ^ (y & z))
static void vsha2c_64(uint64_t *vs2, uint64_t *vd, uint64_t *vs1)
{
uint64_t a = vs2[3], b = vs2[2], e = vs2[1], f = vs2[0];
uint64_t c = vd[3], d = vd[2], g = vd[1], h = vd[0];
uint64_t W0 = vs1[0], W1 = vs1[1];
uint64_t T1 = h + sum1_64(e) + ch(e, f, g) + W0;
uint64_t T2 = sum0_64(a) + maj(a, b, c);
h = g;
g = f;
f = e;
e = d + T1;
d = c;
c = b;
b = a;
a = T1 + T2;
T1 = h + sum1_64(e) + ch(e, f, g) + W1;
T2 = sum0_64(a) + maj(a, b, c);
h = g;
g = f;
f = e;
e = d + T1;
d = c;
c = b;
b = a;
a = T1 + T2;
vd[0] = f;
vd[1] = e;
vd[2] = b;
vd[3] = a;
}
static void vsha2c_32(uint32_t *vs2, uint32_t *vd, uint32_t *vs1)
{
uint32_t a = vs2[H4(3)], b = vs2[H4(2)], e = vs2[H4(1)], f = vs2[H4(0)];
uint32_t c = vd[H4(3)], d = vd[H4(2)], g = vd[H4(1)], h = vd[H4(0)];
uint32_t W0 = vs1[H4(0)], W1 = vs1[H4(1)];
uint32_t T1 = h + sum1_32(e) + ch(e, f, g) + W0;
uint32_t T2 = sum0_32(a) + maj(a, b, c);
h = g;
g = f;
f = e;
e = d + T1;
d = c;
c = b;
b = a;
a = T1 + T2;
T1 = h + sum1_32(e) + ch(e, f, g) + W1;
T2 = sum0_32(a) + maj(a, b, c);
h = g;
g = f;
f = e;
e = d + T1;
d = c;
c = b;
b = a;
a = T1 + T2;
vd[H4(0)] = f;
vd[H4(1)] = e;
vd[H4(2)] = b;
vd[H4(3)] = a;
}
void HELPER(vsha2ch32_vv)(void *vd, void *vs1, void *vs2, CPURISCVState *env,
uint32_t desc)
{
const uint32_t esz = 4;
uint32_t total_elems;
uint32_t vta = vext_vta(desc);
target/riscv/vector_helpers: do early exit when vstart >= vl We're going to make changes that will required each helper to be responsible for the 'vstart' management, i.e. we will relieve the 'vstart < vl' assumption that helpers have today. Helpers are usually able to deal with vstart >= vl, i.e. doing nothing aside from setting vstart = 0 at the end, but the tail update functions will update the tail regardless of vstart being valid or not. Unifying the tail update process in a single function that would handle the vstart >= vl case isn't trivial (see [1] for more info). This patch takes a blunt approach: do an early exit in every single vector helper if vstart >= vl, unless the helper is guarded with vstart_eq_zero in the translation. For those cases the helper is ready to deal with cases where vl might be zero, i.e. throwing exceptions based on it like vcpop_m() and first_m(). Helpers that weren't changed: - vcpop_m(), vfirst_m(), vmsetm(), GEN_VEXT_VIOTA_M(): these are guarded directly with vstart_eq_zero; - GEN_VEXT_VCOMPRESS_VM(): guarded with vcompress_vm_check() that checks vstart_eq_zero; - GEN_VEXT_RED(): guarded with either reduction_check() or reduction_widen_check(), both check vstart_eq_zero; - GEN_VEXT_FRED(): guarded with either freduction_check() or freduction_widen_check(), both check vstart_eq_zero. Another exception is vext_ldst_whole(), who operates on effective vector length regardless of the current settings in vtype and vl. [1] https://lore.kernel.org/qemu-riscv/1590234b-0291-432a-a0fa-c5a6876097bc@linux.alibaba.com/ Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com> Acked-by: Alistair Francis <alistair.francis@wdc.com> Message-ID: <20240314175704.478276-7-dbarboza@ventanamicro.com> Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
2024-03-14 20:57:00 +03:00
VSTART_CHECK_EARLY_EXIT(env);
for (uint32_t i = env->vstart / 4; i < env->vl / 4; i++) {
vsha2c_32(((uint32_t *)vs2) + 4 * i, ((uint32_t *)vd) + 4 * i,
((uint32_t *)vs1) + 4 * i + 2);
}
/* set tail elements to 1s */
total_elems = vext_get_total_elems(env, desc, esz);
vext_set_elems_1s(vd, vta, env->vl * esz, total_elems * esz);
env->vstart = 0;
}
void HELPER(vsha2ch64_vv)(void *vd, void *vs1, void *vs2, CPURISCVState *env,
uint32_t desc)
{
const uint32_t esz = 8;
uint32_t total_elems;
uint32_t vta = vext_vta(desc);
target/riscv/vector_helpers: do early exit when vstart >= vl We're going to make changes that will required each helper to be responsible for the 'vstart' management, i.e. we will relieve the 'vstart < vl' assumption that helpers have today. Helpers are usually able to deal with vstart >= vl, i.e. doing nothing aside from setting vstart = 0 at the end, but the tail update functions will update the tail regardless of vstart being valid or not. Unifying the tail update process in a single function that would handle the vstart >= vl case isn't trivial (see [1] for more info). This patch takes a blunt approach: do an early exit in every single vector helper if vstart >= vl, unless the helper is guarded with vstart_eq_zero in the translation. For those cases the helper is ready to deal with cases where vl might be zero, i.e. throwing exceptions based on it like vcpop_m() and first_m(). Helpers that weren't changed: - vcpop_m(), vfirst_m(), vmsetm(), GEN_VEXT_VIOTA_M(): these are guarded directly with vstart_eq_zero; - GEN_VEXT_VCOMPRESS_VM(): guarded with vcompress_vm_check() that checks vstart_eq_zero; - GEN_VEXT_RED(): guarded with either reduction_check() or reduction_widen_check(), both check vstart_eq_zero; - GEN_VEXT_FRED(): guarded with either freduction_check() or freduction_widen_check(), both check vstart_eq_zero. Another exception is vext_ldst_whole(), who operates on effective vector length regardless of the current settings in vtype and vl. [1] https://lore.kernel.org/qemu-riscv/1590234b-0291-432a-a0fa-c5a6876097bc@linux.alibaba.com/ Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com> Acked-by: Alistair Francis <alistair.francis@wdc.com> Message-ID: <20240314175704.478276-7-dbarboza@ventanamicro.com> Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
2024-03-14 20:57:00 +03:00
VSTART_CHECK_EARLY_EXIT(env);
for (uint32_t i = env->vstart / 4; i < env->vl / 4; i++) {
vsha2c_64(((uint64_t *)vs2) + 4 * i, ((uint64_t *)vd) + 4 * i,
((uint64_t *)vs1) + 4 * i + 2);
}
/* set tail elements to 1s */
total_elems = vext_get_total_elems(env, desc, esz);
vext_set_elems_1s(vd, vta, env->vl * esz, total_elems * esz);
env->vstart = 0;
}
void HELPER(vsha2cl32_vv)(void *vd, void *vs1, void *vs2, CPURISCVState *env,
uint32_t desc)
{
const uint32_t esz = 4;
uint32_t total_elems;
uint32_t vta = vext_vta(desc);
target/riscv/vector_helpers: do early exit when vstart >= vl We're going to make changes that will required each helper to be responsible for the 'vstart' management, i.e. we will relieve the 'vstart < vl' assumption that helpers have today. Helpers are usually able to deal with vstart >= vl, i.e. doing nothing aside from setting vstart = 0 at the end, but the tail update functions will update the tail regardless of vstart being valid or not. Unifying the tail update process in a single function that would handle the vstart >= vl case isn't trivial (see [1] for more info). This patch takes a blunt approach: do an early exit in every single vector helper if vstart >= vl, unless the helper is guarded with vstart_eq_zero in the translation. For those cases the helper is ready to deal with cases where vl might be zero, i.e. throwing exceptions based on it like vcpop_m() and first_m(). Helpers that weren't changed: - vcpop_m(), vfirst_m(), vmsetm(), GEN_VEXT_VIOTA_M(): these are guarded directly with vstart_eq_zero; - GEN_VEXT_VCOMPRESS_VM(): guarded with vcompress_vm_check() that checks vstart_eq_zero; - GEN_VEXT_RED(): guarded with either reduction_check() or reduction_widen_check(), both check vstart_eq_zero; - GEN_VEXT_FRED(): guarded with either freduction_check() or freduction_widen_check(), both check vstart_eq_zero. Another exception is vext_ldst_whole(), who operates on effective vector length regardless of the current settings in vtype and vl. [1] https://lore.kernel.org/qemu-riscv/1590234b-0291-432a-a0fa-c5a6876097bc@linux.alibaba.com/ Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com> Acked-by: Alistair Francis <alistair.francis@wdc.com> Message-ID: <20240314175704.478276-7-dbarboza@ventanamicro.com> Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
2024-03-14 20:57:00 +03:00
VSTART_CHECK_EARLY_EXIT(env);
for (uint32_t i = env->vstart / 4; i < env->vl / 4; i++) {
vsha2c_32(((uint32_t *)vs2) + 4 * i, ((uint32_t *)vd) + 4 * i,
(((uint32_t *)vs1) + 4 * i));
}
/* set tail elements to 1s */
total_elems = vext_get_total_elems(env, desc, esz);
vext_set_elems_1s(vd, vta, env->vl * esz, total_elems * esz);
env->vstart = 0;
}
void HELPER(vsha2cl64_vv)(void *vd, void *vs1, void *vs2, CPURISCVState *env,
uint32_t desc)
{
uint32_t esz = 8;
uint32_t total_elems;
uint32_t vta = vext_vta(desc);
target/riscv/vector_helpers: do early exit when vstart >= vl We're going to make changes that will required each helper to be responsible for the 'vstart' management, i.e. we will relieve the 'vstart < vl' assumption that helpers have today. Helpers are usually able to deal with vstart >= vl, i.e. doing nothing aside from setting vstart = 0 at the end, but the tail update functions will update the tail regardless of vstart being valid or not. Unifying the tail update process in a single function that would handle the vstart >= vl case isn't trivial (see [1] for more info). This patch takes a blunt approach: do an early exit in every single vector helper if vstart >= vl, unless the helper is guarded with vstart_eq_zero in the translation. For those cases the helper is ready to deal with cases where vl might be zero, i.e. throwing exceptions based on it like vcpop_m() and first_m(). Helpers that weren't changed: - vcpop_m(), vfirst_m(), vmsetm(), GEN_VEXT_VIOTA_M(): these are guarded directly with vstart_eq_zero; - GEN_VEXT_VCOMPRESS_VM(): guarded with vcompress_vm_check() that checks vstart_eq_zero; - GEN_VEXT_RED(): guarded with either reduction_check() or reduction_widen_check(), both check vstart_eq_zero; - GEN_VEXT_FRED(): guarded with either freduction_check() or freduction_widen_check(), both check vstart_eq_zero. Another exception is vext_ldst_whole(), who operates on effective vector length regardless of the current settings in vtype and vl. [1] https://lore.kernel.org/qemu-riscv/1590234b-0291-432a-a0fa-c5a6876097bc@linux.alibaba.com/ Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com> Acked-by: Alistair Francis <alistair.francis@wdc.com> Message-ID: <20240314175704.478276-7-dbarboza@ventanamicro.com> Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
2024-03-14 20:57:00 +03:00
VSTART_CHECK_EARLY_EXIT(env);
for (uint32_t i = env->vstart / 4; i < env->vl / 4; i++) {
vsha2c_64(((uint64_t *)vs2) + 4 * i, ((uint64_t *)vd) + 4 * i,
(((uint64_t *)vs1) + 4 * i));
}
/* set tail elements to 1s */
total_elems = vext_get_total_elems(env, desc, esz);
vext_set_elems_1s(vd, vta, env->vl * esz, total_elems * esz);
env->vstart = 0;
}
static inline uint32_t p1(uint32_t x)
{
return x ^ rol32(x, 15) ^ rol32(x, 23);
}
static inline uint32_t zvksh_w(uint32_t m16, uint32_t m9, uint32_t m3,
uint32_t m13, uint32_t m6)
{
return p1(m16 ^ m9 ^ rol32(m3, 15)) ^ rol32(m13, 7) ^ m6;
}
void HELPER(vsm3me_vv)(void *vd_vptr, void *vs1_vptr, void *vs2_vptr,
CPURISCVState *env, uint32_t desc)
{
uint32_t esz = memop_size(FIELD_EX64(env->vtype, VTYPE, VSEW));
uint32_t total_elems = vext_get_total_elems(env, desc, esz);
uint32_t vta = vext_vta(desc);
uint32_t *vd = vd_vptr;
uint32_t *vs1 = vs1_vptr;
uint32_t *vs2 = vs2_vptr;
target/riscv/vector_helpers: do early exit when vstart >= vl We're going to make changes that will required each helper to be responsible for the 'vstart' management, i.e. we will relieve the 'vstart < vl' assumption that helpers have today. Helpers are usually able to deal with vstart >= vl, i.e. doing nothing aside from setting vstart = 0 at the end, but the tail update functions will update the tail regardless of vstart being valid or not. Unifying the tail update process in a single function that would handle the vstart >= vl case isn't trivial (see [1] for more info). This patch takes a blunt approach: do an early exit in every single vector helper if vstart >= vl, unless the helper is guarded with vstart_eq_zero in the translation. For those cases the helper is ready to deal with cases where vl might be zero, i.e. throwing exceptions based on it like vcpop_m() and first_m(). Helpers that weren't changed: - vcpop_m(), vfirst_m(), vmsetm(), GEN_VEXT_VIOTA_M(): these are guarded directly with vstart_eq_zero; - GEN_VEXT_VCOMPRESS_VM(): guarded with vcompress_vm_check() that checks vstart_eq_zero; - GEN_VEXT_RED(): guarded with either reduction_check() or reduction_widen_check(), both check vstart_eq_zero; - GEN_VEXT_FRED(): guarded with either freduction_check() or freduction_widen_check(), both check vstart_eq_zero. Another exception is vext_ldst_whole(), who operates on effective vector length regardless of the current settings in vtype and vl. [1] https://lore.kernel.org/qemu-riscv/1590234b-0291-432a-a0fa-c5a6876097bc@linux.alibaba.com/ Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com> Acked-by: Alistair Francis <alistair.francis@wdc.com> Message-ID: <20240314175704.478276-7-dbarboza@ventanamicro.com> Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
2024-03-14 20:57:00 +03:00
VSTART_CHECK_EARLY_EXIT(env);
for (int i = env->vstart / 8; i < env->vl / 8; i++) {
uint32_t w[24];
for (int j = 0; j < 8; j++) {
w[j] = bswap32(vs1[H4((i * 8) + j)]);
w[j + 8] = bswap32(vs2[H4((i * 8) + j)]);
}
for (int j = 0; j < 8; j++) {
w[j + 16] =
zvksh_w(w[j], w[j + 7], w[j + 13], w[j + 3], w[j + 10]);
}
for (int j = 0; j < 8; j++) {
vd[(i * 8) + j] = bswap32(w[H4(j + 16)]);
}
}
vext_set_elems_1s(vd_vptr, vta, env->vl * esz, total_elems * esz);
env->vstart = 0;
}
static inline uint32_t ff1(uint32_t x, uint32_t y, uint32_t z)
{
return x ^ y ^ z;
}
static inline uint32_t ff2(uint32_t x, uint32_t y, uint32_t z)
{
return (x & y) | (x & z) | (y & z);
}
static inline uint32_t ff_j(uint32_t x, uint32_t y, uint32_t z, uint32_t j)
{
return (j <= 15) ? ff1(x, y, z) : ff2(x, y, z);
}
static inline uint32_t gg1(uint32_t x, uint32_t y, uint32_t z)
{
return x ^ y ^ z;
}
static inline uint32_t gg2(uint32_t x, uint32_t y, uint32_t z)
{
return (x & y) | (~x & z);
}
static inline uint32_t gg_j(uint32_t x, uint32_t y, uint32_t z, uint32_t j)
{
return (j <= 15) ? gg1(x, y, z) : gg2(x, y, z);
}
static inline uint32_t t_j(uint32_t j)
{
return (j <= 15) ? 0x79cc4519 : 0x7a879d8a;
}
static inline uint32_t p_0(uint32_t x)
{
return x ^ rol32(x, 9) ^ rol32(x, 17);
}
static void sm3c(uint32_t *vd, uint32_t *vs1, uint32_t *vs2, uint32_t uimm)
{
uint32_t x0, x1;
uint32_t j;
uint32_t ss1, ss2, tt1, tt2;
x0 = vs2[0] ^ vs2[4];
x1 = vs2[1] ^ vs2[5];
j = 2 * uimm;
ss1 = rol32(rol32(vs1[0], 12) + vs1[4] + rol32(t_j(j), j % 32), 7);
ss2 = ss1 ^ rol32(vs1[0], 12);
tt1 = ff_j(vs1[0], vs1[1], vs1[2], j) + vs1[3] + ss2 + x0;
tt2 = gg_j(vs1[4], vs1[5], vs1[6], j) + vs1[7] + ss1 + vs2[0];
vs1[3] = vs1[2];
vd[3] = rol32(vs1[1], 9);
vs1[1] = vs1[0];
vd[1] = tt1;
vs1[7] = vs1[6];
vd[7] = rol32(vs1[5], 19);
vs1[5] = vs1[4];
vd[5] = p_0(tt2);
j = 2 * uimm + 1;
ss1 = rol32(rol32(vd[1], 12) + vd[5] + rol32(t_j(j), j % 32), 7);
ss2 = ss1 ^ rol32(vd[1], 12);
tt1 = ff_j(vd[1], vs1[1], vd[3], j) + vs1[3] + ss2 + x1;
tt2 = gg_j(vd[5], vs1[5], vd[7], j) + vs1[7] + ss1 + vs2[1];
vd[2] = rol32(vs1[1], 9);
vd[0] = tt1;
vd[6] = rol32(vs1[5], 19);
vd[4] = p_0(tt2);
}
void HELPER(vsm3c_vi)(void *vd_vptr, void *vs2_vptr, uint32_t uimm,
CPURISCVState *env, uint32_t desc)
{
uint32_t esz = memop_size(FIELD_EX64(env->vtype, VTYPE, VSEW));
uint32_t total_elems = vext_get_total_elems(env, desc, esz);
uint32_t vta = vext_vta(desc);
uint32_t *vd = vd_vptr;
uint32_t *vs2 = vs2_vptr;
uint32_t v1[8], v2[8], v3[8];
target/riscv/vector_helpers: do early exit when vstart >= vl We're going to make changes that will required each helper to be responsible for the 'vstart' management, i.e. we will relieve the 'vstart < vl' assumption that helpers have today. Helpers are usually able to deal with vstart >= vl, i.e. doing nothing aside from setting vstart = 0 at the end, but the tail update functions will update the tail regardless of vstart being valid or not. Unifying the tail update process in a single function that would handle the vstart >= vl case isn't trivial (see [1] for more info). This patch takes a blunt approach: do an early exit in every single vector helper if vstart >= vl, unless the helper is guarded with vstart_eq_zero in the translation. For those cases the helper is ready to deal with cases where vl might be zero, i.e. throwing exceptions based on it like vcpop_m() and first_m(). Helpers that weren't changed: - vcpop_m(), vfirst_m(), vmsetm(), GEN_VEXT_VIOTA_M(): these are guarded directly with vstart_eq_zero; - GEN_VEXT_VCOMPRESS_VM(): guarded with vcompress_vm_check() that checks vstart_eq_zero; - GEN_VEXT_RED(): guarded with either reduction_check() or reduction_widen_check(), both check vstart_eq_zero; - GEN_VEXT_FRED(): guarded with either freduction_check() or freduction_widen_check(), both check vstart_eq_zero. Another exception is vext_ldst_whole(), who operates on effective vector length regardless of the current settings in vtype and vl. [1] https://lore.kernel.org/qemu-riscv/1590234b-0291-432a-a0fa-c5a6876097bc@linux.alibaba.com/ Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com> Acked-by: Alistair Francis <alistair.francis@wdc.com> Message-ID: <20240314175704.478276-7-dbarboza@ventanamicro.com> Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
2024-03-14 20:57:00 +03:00
VSTART_CHECK_EARLY_EXIT(env);
for (int i = env->vstart / 8; i < env->vl / 8; i++) {
for (int k = 0; k < 8; k++) {
v2[k] = bswap32(vd[H4(i * 8 + k)]);
v3[k] = bswap32(vs2[H4(i * 8 + k)]);
}
sm3c(v1, v2, v3, uimm);
for (int k = 0; k < 8; k++) {
vd[i * 8 + k] = bswap32(v1[H4(k)]);
}
}
vext_set_elems_1s(vd_vptr, vta, env->vl * esz, total_elems * esz);
env->vstart = 0;
}
void HELPER(vghsh_vv)(void *vd_vptr, void *vs1_vptr, void *vs2_vptr,
CPURISCVState *env, uint32_t desc)
{
uint64_t *vd = vd_vptr;
uint64_t *vs1 = vs1_vptr;
uint64_t *vs2 = vs2_vptr;
uint32_t vta = vext_vta(desc);
uint32_t total_elems = vext_get_total_elems(env, desc, 4);
target/riscv/vector_helpers: do early exit when vstart >= vl We're going to make changes that will required each helper to be responsible for the 'vstart' management, i.e. we will relieve the 'vstart < vl' assumption that helpers have today. Helpers are usually able to deal with vstart >= vl, i.e. doing nothing aside from setting vstart = 0 at the end, but the tail update functions will update the tail regardless of vstart being valid or not. Unifying the tail update process in a single function that would handle the vstart >= vl case isn't trivial (see [1] for more info). This patch takes a blunt approach: do an early exit in every single vector helper if vstart >= vl, unless the helper is guarded with vstart_eq_zero in the translation. For those cases the helper is ready to deal with cases where vl might be zero, i.e. throwing exceptions based on it like vcpop_m() and first_m(). Helpers that weren't changed: - vcpop_m(), vfirst_m(), vmsetm(), GEN_VEXT_VIOTA_M(): these are guarded directly with vstart_eq_zero; - GEN_VEXT_VCOMPRESS_VM(): guarded with vcompress_vm_check() that checks vstart_eq_zero; - GEN_VEXT_RED(): guarded with either reduction_check() or reduction_widen_check(), both check vstart_eq_zero; - GEN_VEXT_FRED(): guarded with either freduction_check() or freduction_widen_check(), both check vstart_eq_zero. Another exception is vext_ldst_whole(), who operates on effective vector length regardless of the current settings in vtype and vl. [1] https://lore.kernel.org/qemu-riscv/1590234b-0291-432a-a0fa-c5a6876097bc@linux.alibaba.com/ Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com> Acked-by: Alistair Francis <alistair.francis@wdc.com> Message-ID: <20240314175704.478276-7-dbarboza@ventanamicro.com> Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
2024-03-14 20:57:00 +03:00
VSTART_CHECK_EARLY_EXIT(env);
for (uint32_t i = env->vstart / 4; i < env->vl / 4; i++) {
uint64_t Y[2] = {vd[i * 2 + 0], vd[i * 2 + 1]};
uint64_t H[2] = {brev8(vs2[i * 2 + 0]), brev8(vs2[i * 2 + 1])};
uint64_t X[2] = {vs1[i * 2 + 0], vs1[i * 2 + 1]};
uint64_t Z[2] = {0, 0};
uint64_t S[2] = {brev8(Y[0] ^ X[0]), brev8(Y[1] ^ X[1])};
for (int j = 0; j < 128; j++) {
if ((S[j / 64] >> (j % 64)) & 1) {
Z[0] ^= H[0];
Z[1] ^= H[1];
}
bool reduce = ((H[1] >> 63) & 1);
H[1] = H[1] << 1 | H[0] >> 63;
H[0] = H[0] << 1;
if (reduce) {
H[0] ^= 0x87;
}
}
vd[i * 2 + 0] = brev8(Z[0]);
vd[i * 2 + 1] = brev8(Z[1]);
}
/* set tail elements to 1s */
vext_set_elems_1s(vd, vta, env->vl * 4, total_elems * 4);
env->vstart = 0;
}
void HELPER(vgmul_vv)(void *vd_vptr, void *vs2_vptr, CPURISCVState *env,
uint32_t desc)
{
uint64_t *vd = vd_vptr;
uint64_t *vs2 = vs2_vptr;
uint32_t vta = vext_vta(desc);
uint32_t total_elems = vext_get_total_elems(env, desc, 4);
target/riscv/vector_helpers: do early exit when vstart >= vl We're going to make changes that will required each helper to be responsible for the 'vstart' management, i.e. we will relieve the 'vstart < vl' assumption that helpers have today. Helpers are usually able to deal with vstart >= vl, i.e. doing nothing aside from setting vstart = 0 at the end, but the tail update functions will update the tail regardless of vstart being valid or not. Unifying the tail update process in a single function that would handle the vstart >= vl case isn't trivial (see [1] for more info). This patch takes a blunt approach: do an early exit in every single vector helper if vstart >= vl, unless the helper is guarded with vstart_eq_zero in the translation. For those cases the helper is ready to deal with cases where vl might be zero, i.e. throwing exceptions based on it like vcpop_m() and first_m(). Helpers that weren't changed: - vcpop_m(), vfirst_m(), vmsetm(), GEN_VEXT_VIOTA_M(): these are guarded directly with vstart_eq_zero; - GEN_VEXT_VCOMPRESS_VM(): guarded with vcompress_vm_check() that checks vstart_eq_zero; - GEN_VEXT_RED(): guarded with either reduction_check() or reduction_widen_check(), both check vstart_eq_zero; - GEN_VEXT_FRED(): guarded with either freduction_check() or freduction_widen_check(), both check vstart_eq_zero. Another exception is vext_ldst_whole(), who operates on effective vector length regardless of the current settings in vtype and vl. [1] https://lore.kernel.org/qemu-riscv/1590234b-0291-432a-a0fa-c5a6876097bc@linux.alibaba.com/ Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com> Acked-by: Alistair Francis <alistair.francis@wdc.com> Message-ID: <20240314175704.478276-7-dbarboza@ventanamicro.com> Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
2024-03-14 20:57:00 +03:00
VSTART_CHECK_EARLY_EXIT(env);
for (uint32_t i = env->vstart / 4; i < env->vl / 4; i++) {
uint64_t Y[2] = {brev8(vd[i * 2 + 0]), brev8(vd[i * 2 + 1])};
uint64_t H[2] = {brev8(vs2[i * 2 + 0]), brev8(vs2[i * 2 + 1])};
uint64_t Z[2] = {0, 0};
for (int j = 0; j < 128; j++) {
if ((Y[j / 64] >> (j % 64)) & 1) {
Z[0] ^= H[0];
Z[1] ^= H[1];
}
bool reduce = ((H[1] >> 63) & 1);
H[1] = H[1] << 1 | H[0] >> 63;
H[0] = H[0] << 1;
if (reduce) {
H[0] ^= 0x87;
}
}
vd[i * 2 + 0] = brev8(Z[0]);
vd[i * 2 + 1] = brev8(Z[1]);
}
/* set tail elements to 1s */
vext_set_elems_1s(vd, vta, env->vl * 4, total_elems * 4);
env->vstart = 0;
}
void HELPER(vsm4k_vi)(void *vd, void *vs2, uint32_t uimm5, CPURISCVState *env,
uint32_t desc)
{
const uint32_t egs = 4;
uint32_t rnd = uimm5 & 0x7;
uint32_t group_start = env->vstart / egs;
uint32_t group_end = env->vl / egs;
uint32_t esz = sizeof(uint32_t);
uint32_t total_elems = vext_get_total_elems(env, desc, esz);
target/riscv/vector_helpers: do early exit when vstart >= vl We're going to make changes that will required each helper to be responsible for the 'vstart' management, i.e. we will relieve the 'vstart < vl' assumption that helpers have today. Helpers are usually able to deal with vstart >= vl, i.e. doing nothing aside from setting vstart = 0 at the end, but the tail update functions will update the tail regardless of vstart being valid or not. Unifying the tail update process in a single function that would handle the vstart >= vl case isn't trivial (see [1] for more info). This patch takes a blunt approach: do an early exit in every single vector helper if vstart >= vl, unless the helper is guarded with vstart_eq_zero in the translation. For those cases the helper is ready to deal with cases where vl might be zero, i.e. throwing exceptions based on it like vcpop_m() and first_m(). Helpers that weren't changed: - vcpop_m(), vfirst_m(), vmsetm(), GEN_VEXT_VIOTA_M(): these are guarded directly with vstart_eq_zero; - GEN_VEXT_VCOMPRESS_VM(): guarded with vcompress_vm_check() that checks vstart_eq_zero; - GEN_VEXT_RED(): guarded with either reduction_check() or reduction_widen_check(), both check vstart_eq_zero; - GEN_VEXT_FRED(): guarded with either freduction_check() or freduction_widen_check(), both check vstart_eq_zero. Another exception is vext_ldst_whole(), who operates on effective vector length regardless of the current settings in vtype and vl. [1] https://lore.kernel.org/qemu-riscv/1590234b-0291-432a-a0fa-c5a6876097bc@linux.alibaba.com/ Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com> Acked-by: Alistair Francis <alistair.francis@wdc.com> Message-ID: <20240314175704.478276-7-dbarboza@ventanamicro.com> Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
2024-03-14 20:57:00 +03:00
VSTART_CHECK_EARLY_EXIT(env);
for (uint32_t i = group_start; i < group_end; ++i) {
uint32_t vstart = i * egs;
uint32_t vend = (i + 1) * egs;
uint32_t rk[4] = {0};
uint32_t tmp[8] = {0};
for (uint32_t j = vstart; j < vend; ++j) {
rk[j - vstart] = *((uint32_t *)vs2 + H4(j));
}
for (uint32_t j = 0; j < egs; ++j) {
tmp[j] = rk[j];
}
for (uint32_t j = 0; j < egs; ++j) {
uint32_t b, s;
b = tmp[j + 1] ^ tmp[j + 2] ^ tmp[j + 3] ^ sm4_ck[rnd * 4 + j];
s = sm4_subword(b);
tmp[j + 4] = tmp[j] ^ (s ^ rol32(s, 13) ^ rol32(s, 23));
}
for (uint32_t j = vstart; j < vend; ++j) {
*((uint32_t *)vd + H4(j)) = tmp[egs + (j - vstart)];
}
}
env->vstart = 0;
/* set tail elements to 1s */
vext_set_elems_1s(vd, vext_vta(desc), env->vl * esz, total_elems * esz);
}
static void do_sm4_round(uint32_t *rk, uint32_t *buf)
{
const uint32_t egs = 4;
uint32_t s, b;
for (uint32_t j = egs; j < egs * 2; ++j) {
b = buf[j - 3] ^ buf[j - 2] ^ buf[j - 1] ^ rk[j - 4];
s = sm4_subword(b);
buf[j] = buf[j - 4] ^ (s ^ rol32(s, 2) ^ rol32(s, 10) ^ rol32(s, 18) ^
rol32(s, 24));
}
}
void HELPER(vsm4r_vv)(void *vd, void *vs2, CPURISCVState *env, uint32_t desc)
{
const uint32_t egs = 4;
uint32_t group_start = env->vstart / egs;
uint32_t group_end = env->vl / egs;
uint32_t esz = sizeof(uint32_t);
uint32_t total_elems = vext_get_total_elems(env, desc, esz);
target/riscv/vector_helpers: do early exit when vstart >= vl We're going to make changes that will required each helper to be responsible for the 'vstart' management, i.e. we will relieve the 'vstart < vl' assumption that helpers have today. Helpers are usually able to deal with vstart >= vl, i.e. doing nothing aside from setting vstart = 0 at the end, but the tail update functions will update the tail regardless of vstart being valid or not. Unifying the tail update process in a single function that would handle the vstart >= vl case isn't trivial (see [1] for more info). This patch takes a blunt approach: do an early exit in every single vector helper if vstart >= vl, unless the helper is guarded with vstart_eq_zero in the translation. For those cases the helper is ready to deal with cases where vl might be zero, i.e. throwing exceptions based on it like vcpop_m() and first_m(). Helpers that weren't changed: - vcpop_m(), vfirst_m(), vmsetm(), GEN_VEXT_VIOTA_M(): these are guarded directly with vstart_eq_zero; - GEN_VEXT_VCOMPRESS_VM(): guarded with vcompress_vm_check() that checks vstart_eq_zero; - GEN_VEXT_RED(): guarded with either reduction_check() or reduction_widen_check(), both check vstart_eq_zero; - GEN_VEXT_FRED(): guarded with either freduction_check() or freduction_widen_check(), both check vstart_eq_zero. Another exception is vext_ldst_whole(), who operates on effective vector length regardless of the current settings in vtype and vl. [1] https://lore.kernel.org/qemu-riscv/1590234b-0291-432a-a0fa-c5a6876097bc@linux.alibaba.com/ Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com> Acked-by: Alistair Francis <alistair.francis@wdc.com> Message-ID: <20240314175704.478276-7-dbarboza@ventanamicro.com> Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
2024-03-14 20:57:00 +03:00
VSTART_CHECK_EARLY_EXIT(env);
for (uint32_t i = group_start; i < group_end; ++i) {
uint32_t vstart = i * egs;
uint32_t vend = (i + 1) * egs;
uint32_t rk[4] = {0};
uint32_t tmp[8] = {0};
for (uint32_t j = vstart; j < vend; ++j) {
rk[j - vstart] = *((uint32_t *)vs2 + H4(j));
}
for (uint32_t j = vstart; j < vend; ++j) {
tmp[j - vstart] = *((uint32_t *)vd + H4(j));
}
do_sm4_round(rk, tmp);
for (uint32_t j = vstart; j < vend; ++j) {
*((uint32_t *)vd + H4(j)) = tmp[egs + (j - vstart)];
}
}
env->vstart = 0;
/* set tail elements to 1s */
vext_set_elems_1s(vd, vext_vta(desc), env->vl * esz, total_elems * esz);
}
void HELPER(vsm4r_vs)(void *vd, void *vs2, CPURISCVState *env, uint32_t desc)
{
const uint32_t egs = 4;
uint32_t group_start = env->vstart / egs;
uint32_t group_end = env->vl / egs;
uint32_t esz = sizeof(uint32_t);
uint32_t total_elems = vext_get_total_elems(env, desc, esz);
target/riscv/vector_helpers: do early exit when vstart >= vl We're going to make changes that will required each helper to be responsible for the 'vstart' management, i.e. we will relieve the 'vstart < vl' assumption that helpers have today. Helpers are usually able to deal with vstart >= vl, i.e. doing nothing aside from setting vstart = 0 at the end, but the tail update functions will update the tail regardless of vstart being valid or not. Unifying the tail update process in a single function that would handle the vstart >= vl case isn't trivial (see [1] for more info). This patch takes a blunt approach: do an early exit in every single vector helper if vstart >= vl, unless the helper is guarded with vstart_eq_zero in the translation. For those cases the helper is ready to deal with cases where vl might be zero, i.e. throwing exceptions based on it like vcpop_m() and first_m(). Helpers that weren't changed: - vcpop_m(), vfirst_m(), vmsetm(), GEN_VEXT_VIOTA_M(): these are guarded directly with vstart_eq_zero; - GEN_VEXT_VCOMPRESS_VM(): guarded with vcompress_vm_check() that checks vstart_eq_zero; - GEN_VEXT_RED(): guarded with either reduction_check() or reduction_widen_check(), both check vstart_eq_zero; - GEN_VEXT_FRED(): guarded with either freduction_check() or freduction_widen_check(), both check vstart_eq_zero. Another exception is vext_ldst_whole(), who operates on effective vector length regardless of the current settings in vtype and vl. [1] https://lore.kernel.org/qemu-riscv/1590234b-0291-432a-a0fa-c5a6876097bc@linux.alibaba.com/ Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com> Acked-by: Alistair Francis <alistair.francis@wdc.com> Message-ID: <20240314175704.478276-7-dbarboza@ventanamicro.com> Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
2024-03-14 20:57:00 +03:00
VSTART_CHECK_EARLY_EXIT(env);
for (uint32_t i = group_start; i < group_end; ++i) {
uint32_t vstart = i * egs;
uint32_t vend = (i + 1) * egs;
uint32_t rk[4] = {0};
uint32_t tmp[8] = {0};
for (uint32_t j = 0; j < egs; ++j) {
rk[j] = *((uint32_t *)vs2 + H4(j));
}
for (uint32_t j = vstart; j < vend; ++j) {
tmp[j - vstart] = *((uint32_t *)vd + H4(j));
}
do_sm4_round(rk, tmp);
for (uint32_t j = vstart; j < vend; ++j) {
*((uint32_t *)vd + H4(j)) = tmp[egs + (j - vstart)];
}
}
env->vstart = 0;
/* set tail elements to 1s */
vext_set_elems_1s(vd, vext_vta(desc), env->vl * esz, total_elems * esz);
}