mirror of
https://github.com/yuzu-emu/unicorn.git
synced 2024-12-23 00:45:40 +00:00
target/arm: Implement SVE dot product (vectors)
Backports commit d730ecaae77ac696515207a5ef99509240fc792b from qemu
This commit is contained in:
parent
01a7224cdb
commit
2f6d555473
|
@ -1189,6 +1189,8 @@
|
|||
#define helper_gvec_sar16i helper_gvec_sar16i_aarch64
|
||||
#define helper_gvec_sar32i helper_gvec_sar32i_aarch64
|
||||
#define helper_gvec_sar64i helper_gvec_sar64i_aarch64
|
||||
#define helper_gvec_sdot_b helper_gvec_sdot_b_aarch64
|
||||
#define helper_gvec_sdot_h helper_gvec_sdot_h_aarch64
|
||||
#define helper_gvec_shl8i helper_gvec_shl8i_aarch64
|
||||
#define helper_gvec_shl16i helper_gvec_shl16i_aarch64
|
||||
#define helper_gvec_shl32i helper_gvec_shl32i_aarch64
|
||||
|
@ -1213,6 +1215,8 @@
|
|||
#define helper_gvec_sssub16 helper_gvec_sssub16_aarch64
|
||||
#define helper_gvec_sssub32 helper_gvec_sssub32_aarch64
|
||||
#define helper_gvec_sssub64 helper_gvec_sssub64_aarch64
|
||||
#define helper_gvec_udot_b helper_gvec_udot_b_aarch64
|
||||
#define helper_gvec_udot_h helper_gvec_udot_h_aarch64
|
||||
#define helper_gvec_usadd8 helper_gvec_usadd8_aarch64
|
||||
#define helper_gvec_usadd16 helper_gvec_usadd16_aarch64
|
||||
#define helper_gvec_usadd32 helper_gvec_usadd32_aarch64
|
||||
|
|
|
@ -1189,6 +1189,8 @@
|
|||
#define helper_gvec_sar16i helper_gvec_sar16i_aarch64eb
|
||||
#define helper_gvec_sar32i helper_gvec_sar32i_aarch64eb
|
||||
#define helper_gvec_sar64i helper_gvec_sar64i_aarch64eb
|
||||
#define helper_gvec_sdot_b helper_gvec_sdot_b_aarch64eb
|
||||
#define helper_gvec_sdot_h helper_gvec_sdot_h_aarch64eb
|
||||
#define helper_gvec_shl8i helper_gvec_shl8i_aarch64eb
|
||||
#define helper_gvec_shl16i helper_gvec_shl16i_aarch64eb
|
||||
#define helper_gvec_shl32i helper_gvec_shl32i_aarch64eb
|
||||
|
@ -1213,6 +1215,8 @@
|
|||
#define helper_gvec_sssub16 helper_gvec_sssub16_aarch64eb
|
||||
#define helper_gvec_sssub32 helper_gvec_sssub32_aarch64eb
|
||||
#define helper_gvec_sssub64 helper_gvec_sssub64_aarch64eb
|
||||
#define helper_gvec_udot_b helper_gvec_udot_b_aarch64eb
|
||||
#define helper_gvec_udot_h helper_gvec_udot_h_aarch64eb
|
||||
#define helper_gvec_usadd8 helper_gvec_usadd8_aarch64eb
|
||||
#define helper_gvec_usadd16 helper_gvec_usadd16_aarch64eb
|
||||
#define helper_gvec_usadd32 helper_gvec_usadd32_aarch64eb
|
||||
|
|
|
@ -1189,6 +1189,8 @@
|
|||
#define helper_gvec_sar16i helper_gvec_sar16i_arm
|
||||
#define helper_gvec_sar32i helper_gvec_sar32i_arm
|
||||
#define helper_gvec_sar64i helper_gvec_sar64i_arm
|
||||
#define helper_gvec_sdot_b helper_gvec_sdot_b_arm
|
||||
#define helper_gvec_sdot_h helper_gvec_sdot_h_arm
|
||||
#define helper_gvec_shl8i helper_gvec_shl8i_arm
|
||||
#define helper_gvec_shl16i helper_gvec_shl16i_arm
|
||||
#define helper_gvec_shl32i helper_gvec_shl32i_arm
|
||||
|
@ -1213,6 +1215,8 @@
|
|||
#define helper_gvec_sssub16 helper_gvec_sssub16_arm
|
||||
#define helper_gvec_sssub32 helper_gvec_sssub32_arm
|
||||
#define helper_gvec_sssub64 helper_gvec_sssub64_arm
|
||||
#define helper_gvec_udot_b helper_gvec_udot_b_arm
|
||||
#define helper_gvec_udot_h helper_gvec_udot_h_arm
|
||||
#define helper_gvec_usadd8 helper_gvec_usadd8_arm
|
||||
#define helper_gvec_usadd16 helper_gvec_usadd16_arm
|
||||
#define helper_gvec_usadd32 helper_gvec_usadd32_arm
|
||||
|
|
|
@ -1189,6 +1189,8 @@
|
|||
#define helper_gvec_sar16i helper_gvec_sar16i_armeb
|
||||
#define helper_gvec_sar32i helper_gvec_sar32i_armeb
|
||||
#define helper_gvec_sar64i helper_gvec_sar64i_armeb
|
||||
#define helper_gvec_sdot_b helper_gvec_sdot_b_armeb
|
||||
#define helper_gvec_sdot_h helper_gvec_sdot_h_armeb
|
||||
#define helper_gvec_shl8i helper_gvec_shl8i_armeb
|
||||
#define helper_gvec_shl16i helper_gvec_shl16i_armeb
|
||||
#define helper_gvec_shl32i helper_gvec_shl32i_armeb
|
||||
|
@ -1213,6 +1215,8 @@
|
|||
#define helper_gvec_sssub16 helper_gvec_sssub16_armeb
|
||||
#define helper_gvec_sssub32 helper_gvec_sssub32_armeb
|
||||
#define helper_gvec_sssub64 helper_gvec_sssub64_armeb
|
||||
#define helper_gvec_udot_b helper_gvec_udot_b_armeb
|
||||
#define helper_gvec_udot_h helper_gvec_udot_h_armeb
|
||||
#define helper_gvec_usadd8 helper_gvec_usadd8_armeb
|
||||
#define helper_gvec_usadd16 helper_gvec_usadd16_armeb
|
||||
#define helper_gvec_usadd32 helper_gvec_usadd32_armeb
|
||||
|
|
|
@ -1195,6 +1195,8 @@ symbols = (
|
|||
'helper_gvec_sar16i',
|
||||
'helper_gvec_sar32i',
|
||||
'helper_gvec_sar64i',
|
||||
'helper_gvec_sdot_b',
|
||||
'helper_gvec_sdot_h',
|
||||
'helper_gvec_shl8i',
|
||||
'helper_gvec_shl16i',
|
||||
'helper_gvec_shl32i',
|
||||
|
@ -1219,6 +1221,8 @@ symbols = (
|
|||
'helper_gvec_sssub16',
|
||||
'helper_gvec_sssub32',
|
||||
'helper_gvec_sssub64',
|
||||
'helper_gvec_udot_b',
|
||||
'helper_gvec_udot_h',
|
||||
'helper_gvec_usadd8',
|
||||
'helper_gvec_usadd16',
|
||||
'helper_gvec_usadd32',
|
||||
|
|
|
@ -1189,6 +1189,8 @@
|
|||
#define helper_gvec_sar16i helper_gvec_sar16i_m68k
|
||||
#define helper_gvec_sar32i helper_gvec_sar32i_m68k
|
||||
#define helper_gvec_sar64i helper_gvec_sar64i_m68k
|
||||
#define helper_gvec_sdot_b helper_gvec_sdot_b_m68k
|
||||
#define helper_gvec_sdot_h helper_gvec_sdot_h_m68k
|
||||
#define helper_gvec_shl8i helper_gvec_shl8i_m68k
|
||||
#define helper_gvec_shl16i helper_gvec_shl16i_m68k
|
||||
#define helper_gvec_shl32i helper_gvec_shl32i_m68k
|
||||
|
@ -1213,6 +1215,8 @@
|
|||
#define helper_gvec_sssub16 helper_gvec_sssub16_m68k
|
||||
#define helper_gvec_sssub32 helper_gvec_sssub32_m68k
|
||||
#define helper_gvec_sssub64 helper_gvec_sssub64_m68k
|
||||
#define helper_gvec_udot_b helper_gvec_udot_b_m68k
|
||||
#define helper_gvec_udot_h helper_gvec_udot_h_m68k
|
||||
#define helper_gvec_usadd8 helper_gvec_usadd8_m68k
|
||||
#define helper_gvec_usadd16 helper_gvec_usadd16_m68k
|
||||
#define helper_gvec_usadd32 helper_gvec_usadd32_m68k
|
||||
|
|
|
@ -1189,6 +1189,8 @@
|
|||
#define helper_gvec_sar16i helper_gvec_sar16i_mips
|
||||
#define helper_gvec_sar32i helper_gvec_sar32i_mips
|
||||
#define helper_gvec_sar64i helper_gvec_sar64i_mips
|
||||
#define helper_gvec_sdot_b helper_gvec_sdot_b_mips
|
||||
#define helper_gvec_sdot_h helper_gvec_sdot_h_mips
|
||||
#define helper_gvec_shl8i helper_gvec_shl8i_mips
|
||||
#define helper_gvec_shl16i helper_gvec_shl16i_mips
|
||||
#define helper_gvec_shl32i helper_gvec_shl32i_mips
|
||||
|
@ -1213,6 +1215,8 @@
|
|||
#define helper_gvec_sssub16 helper_gvec_sssub16_mips
|
||||
#define helper_gvec_sssub32 helper_gvec_sssub32_mips
|
||||
#define helper_gvec_sssub64 helper_gvec_sssub64_mips
|
||||
#define helper_gvec_udot_b helper_gvec_udot_b_mips
|
||||
#define helper_gvec_udot_h helper_gvec_udot_h_mips
|
||||
#define helper_gvec_usadd8 helper_gvec_usadd8_mips
|
||||
#define helper_gvec_usadd16 helper_gvec_usadd16_mips
|
||||
#define helper_gvec_usadd32 helper_gvec_usadd32_mips
|
||||
|
|
|
@ -1189,6 +1189,8 @@
|
|||
#define helper_gvec_sar16i helper_gvec_sar16i_mips64
|
||||
#define helper_gvec_sar32i helper_gvec_sar32i_mips64
|
||||
#define helper_gvec_sar64i helper_gvec_sar64i_mips64
|
||||
#define helper_gvec_sdot_b helper_gvec_sdot_b_mips64
|
||||
#define helper_gvec_sdot_h helper_gvec_sdot_h_mips64
|
||||
#define helper_gvec_shl8i helper_gvec_shl8i_mips64
|
||||
#define helper_gvec_shl16i helper_gvec_shl16i_mips64
|
||||
#define helper_gvec_shl32i helper_gvec_shl32i_mips64
|
||||
|
@ -1213,6 +1215,8 @@
|
|||
#define helper_gvec_sssub16 helper_gvec_sssub16_mips64
|
||||
#define helper_gvec_sssub32 helper_gvec_sssub32_mips64
|
||||
#define helper_gvec_sssub64 helper_gvec_sssub64_mips64
|
||||
#define helper_gvec_udot_b helper_gvec_udot_b_mips64
|
||||
#define helper_gvec_udot_h helper_gvec_udot_h_mips64
|
||||
#define helper_gvec_usadd8 helper_gvec_usadd8_mips64
|
||||
#define helper_gvec_usadd16 helper_gvec_usadd16_mips64
|
||||
#define helper_gvec_usadd32 helper_gvec_usadd32_mips64
|
||||
|
|
|
@ -1189,6 +1189,8 @@
|
|||
#define helper_gvec_sar16i helper_gvec_sar16i_mips64el
|
||||
#define helper_gvec_sar32i helper_gvec_sar32i_mips64el
|
||||
#define helper_gvec_sar64i helper_gvec_sar64i_mips64el
|
||||
#define helper_gvec_sdot_b helper_gvec_sdot_b_mips64el
|
||||
#define helper_gvec_sdot_h helper_gvec_sdot_h_mips64el
|
||||
#define helper_gvec_shl8i helper_gvec_shl8i_mips64el
|
||||
#define helper_gvec_shl16i helper_gvec_shl16i_mips64el
|
||||
#define helper_gvec_shl32i helper_gvec_shl32i_mips64el
|
||||
|
@ -1213,6 +1215,8 @@
|
|||
#define helper_gvec_sssub16 helper_gvec_sssub16_mips64el
|
||||
#define helper_gvec_sssub32 helper_gvec_sssub32_mips64el
|
||||
#define helper_gvec_sssub64 helper_gvec_sssub64_mips64el
|
||||
#define helper_gvec_udot_b helper_gvec_udot_b_mips64el
|
||||
#define helper_gvec_udot_h helper_gvec_udot_h_mips64el
|
||||
#define helper_gvec_usadd8 helper_gvec_usadd8_mips64el
|
||||
#define helper_gvec_usadd16 helper_gvec_usadd16_mips64el
|
||||
#define helper_gvec_usadd32 helper_gvec_usadd32_mips64el
|
||||
|
|
|
@ -1189,6 +1189,8 @@
|
|||
#define helper_gvec_sar16i helper_gvec_sar16i_mipsel
|
||||
#define helper_gvec_sar32i helper_gvec_sar32i_mipsel
|
||||
#define helper_gvec_sar64i helper_gvec_sar64i_mipsel
|
||||
#define helper_gvec_sdot_b helper_gvec_sdot_b_mipsel
|
||||
#define helper_gvec_sdot_h helper_gvec_sdot_h_mipsel
|
||||
#define helper_gvec_shl8i helper_gvec_shl8i_mipsel
|
||||
#define helper_gvec_shl16i helper_gvec_shl16i_mipsel
|
||||
#define helper_gvec_shl32i helper_gvec_shl32i_mipsel
|
||||
|
@ -1213,6 +1215,8 @@
|
|||
#define helper_gvec_sssub16 helper_gvec_sssub16_mipsel
|
||||
#define helper_gvec_sssub32 helper_gvec_sssub32_mipsel
|
||||
#define helper_gvec_sssub64 helper_gvec_sssub64_mipsel
|
||||
#define helper_gvec_udot_b helper_gvec_udot_b_mipsel
|
||||
#define helper_gvec_udot_h helper_gvec_udot_h_mipsel
|
||||
#define helper_gvec_usadd8 helper_gvec_usadd8_mipsel
|
||||
#define helper_gvec_usadd16 helper_gvec_usadd16_mipsel
|
||||
#define helper_gvec_usadd32 helper_gvec_usadd32_mipsel
|
||||
|
|
|
@ -1189,6 +1189,8 @@
|
|||
#define helper_gvec_sar16i helper_gvec_sar16i_powerpc
|
||||
#define helper_gvec_sar32i helper_gvec_sar32i_powerpc
|
||||
#define helper_gvec_sar64i helper_gvec_sar64i_powerpc
|
||||
#define helper_gvec_sdot_b helper_gvec_sdot_b_powerpc
|
||||
#define helper_gvec_sdot_h helper_gvec_sdot_h_powerpc
|
||||
#define helper_gvec_shl8i helper_gvec_shl8i_powerpc
|
||||
#define helper_gvec_shl16i helper_gvec_shl16i_powerpc
|
||||
#define helper_gvec_shl32i helper_gvec_shl32i_powerpc
|
||||
|
@ -1213,6 +1215,8 @@
|
|||
#define helper_gvec_sssub16 helper_gvec_sssub16_powerpc
|
||||
#define helper_gvec_sssub32 helper_gvec_sssub32_powerpc
|
||||
#define helper_gvec_sssub64 helper_gvec_sssub64_powerpc
|
||||
#define helper_gvec_udot_b helper_gvec_udot_b_powerpc
|
||||
#define helper_gvec_udot_h helper_gvec_udot_h_powerpc
|
||||
#define helper_gvec_usadd8 helper_gvec_usadd8_powerpc
|
||||
#define helper_gvec_usadd16 helper_gvec_usadd16_powerpc
|
||||
#define helper_gvec_usadd32 helper_gvec_usadd32_powerpc
|
||||
|
|
|
@ -1189,6 +1189,8 @@
|
|||
#define helper_gvec_sar16i helper_gvec_sar16i_sparc
|
||||
#define helper_gvec_sar32i helper_gvec_sar32i_sparc
|
||||
#define helper_gvec_sar64i helper_gvec_sar64i_sparc
|
||||
#define helper_gvec_sdot_b helper_gvec_sdot_b_sparc
|
||||
#define helper_gvec_sdot_h helper_gvec_sdot_h_sparc
|
||||
#define helper_gvec_shl8i helper_gvec_shl8i_sparc
|
||||
#define helper_gvec_shl16i helper_gvec_shl16i_sparc
|
||||
#define helper_gvec_shl32i helper_gvec_shl32i_sparc
|
||||
|
@ -1213,6 +1215,8 @@
|
|||
#define helper_gvec_sssub16 helper_gvec_sssub16_sparc
|
||||
#define helper_gvec_sssub32 helper_gvec_sssub32_sparc
|
||||
#define helper_gvec_sssub64 helper_gvec_sssub64_sparc
|
||||
#define helper_gvec_udot_b helper_gvec_udot_b_sparc
|
||||
#define helper_gvec_udot_h helper_gvec_udot_h_sparc
|
||||
#define helper_gvec_usadd8 helper_gvec_usadd8_sparc
|
||||
#define helper_gvec_usadd16 helper_gvec_usadd16_sparc
|
||||
#define helper_gvec_usadd32 helper_gvec_usadd32_sparc
|
||||
|
|
|
@ -1189,6 +1189,8 @@
|
|||
#define helper_gvec_sar16i helper_gvec_sar16i_sparc64
|
||||
#define helper_gvec_sar32i helper_gvec_sar32i_sparc64
|
||||
#define helper_gvec_sar64i helper_gvec_sar64i_sparc64
|
||||
#define helper_gvec_sdot_b helper_gvec_sdot_b_sparc64
|
||||
#define helper_gvec_sdot_h helper_gvec_sdot_h_sparc64
|
||||
#define helper_gvec_shl8i helper_gvec_shl8i_sparc64
|
||||
#define helper_gvec_shl16i helper_gvec_shl16i_sparc64
|
||||
#define helper_gvec_shl32i helper_gvec_shl32i_sparc64
|
||||
|
@ -1213,6 +1215,8 @@
|
|||
#define helper_gvec_sssub16 helper_gvec_sssub16_sparc64
|
||||
#define helper_gvec_sssub32 helper_gvec_sssub32_sparc64
|
||||
#define helper_gvec_sssub64 helper_gvec_sssub64_sparc64
|
||||
#define helper_gvec_udot_b helper_gvec_udot_b_sparc64
|
||||
#define helper_gvec_udot_h helper_gvec_udot_h_sparc64
|
||||
#define helper_gvec_usadd8 helper_gvec_usadd8_sparc64
|
||||
#define helper_gvec_usadd16 helper_gvec_usadd16_sparc64
|
||||
#define helper_gvec_usadd32 helper_gvec_usadd32_sparc64
|
||||
|
|
|
@ -585,6 +585,11 @@ DEF_HELPER_FLAGS_5(gvec_qrdmlah_s32, TCG_CALL_NO_RWG,
|
|||
DEF_HELPER_FLAGS_5(gvec_qrdmlsh_s32, TCG_CALL_NO_RWG,
|
||||
void, ptr, ptr, ptr, ptr, i32)
|
||||
|
||||
DEF_HELPER_FLAGS_4(gvec_sdot_b, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32)
|
||||
DEF_HELPER_FLAGS_4(gvec_udot_b, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32)
|
||||
DEF_HELPER_FLAGS_4(gvec_sdot_h, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32)
|
||||
DEF_HELPER_FLAGS_4(gvec_udot_h, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32)
|
||||
|
||||
DEF_HELPER_FLAGS_5(gvec_fcaddh, TCG_CALL_NO_RWG,
|
||||
void, ptr, ptr, ptr, ptr, i32)
|
||||
DEF_HELPER_FLAGS_5(gvec_fcadds, TCG_CALL_NO_RWG,
|
||||
|
|
|
@ -735,6 +735,9 @@ UMIN_zzi 00100101 .. 101 011 110 ........ ..... @rdn_i8u
|
|||
# SVE integer multiply immediate (unpredicated)
|
||||
MUL_zzi 00100101 .. 110 000 110 ........ ..... @rdn_i8s
|
||||
|
||||
# SVE integer dot product (unpredicated)
|
||||
DOT_zzz 01000100 1 sz:1 0 rm:5 00000 u:1 rn:5 rd:5 ra=%reg_movprfx
|
||||
|
||||
# SVE floating-point complex add (predicated)
|
||||
FCADD 01100100 esz:2 00000 rot:1 100 pg:3 rm:5 rd:5 \
|
||||
rn=%reg_movprfx
|
||||
|
|
|
@ -846,6 +846,24 @@ DO_ZZW(LSL, lsl)
|
|||
|
||||
#undef DO_ZZW
|
||||
|
||||
static bool trans_DOT_zzz(DisasContext *s, arg_DOT_zzz *a, uint32_t insn)
|
||||
{
|
||||
static gen_helper_gvec_3 * const fns[2][2] = {
|
||||
{ gen_helper_gvec_sdot_b, gen_helper_gvec_sdot_h },
|
||||
{ gen_helper_gvec_udot_b, gen_helper_gvec_udot_h }
|
||||
};
|
||||
|
||||
if (sve_access_check(s)) {
|
||||
TCGContext *tcg_ctx = s->uc->tcg_ctx;
|
||||
unsigned vsz = vec_full_reg_size(s);
|
||||
tcg_gen_gvec_3_ool(tcg_ctx, vec_full_reg_offset(s, a->rd),
|
||||
vec_full_reg_offset(s, a->rn),
|
||||
vec_full_reg_offset(s, a->rm),
|
||||
vsz, vsz, 0, fns[a->u][a->sz]);
|
||||
}
|
||||
return true;
|
||||
}
|
||||
|
||||
/*
|
||||
*** SVE Integer Multiply-Add Group
|
||||
*/
|
||||
|
|
|
@ -195,6 +195,73 @@ void HELPER(gvec_qrdmlsh_s32)(void *vd, void *vn, void *vm,
|
|||
clear_tail(d, opr_sz, simd_maxsz(desc));
|
||||
}
|
||||
|
||||
/* Integer 8 and 16-bit dot-product.
|
||||
*
|
||||
* Note that for the loops herein, host endianness does not matter
|
||||
* with respect to the ordering of data within the 64-bit lanes.
|
||||
* All elements are treated equally, no matter where they are.
|
||||
*/
|
||||
|
||||
void HELPER(gvec_sdot_b)(void *vd, void *vn, void *vm, uint32_t desc)
|
||||
{
|
||||
intptr_t i, opr_sz = simd_oprsz(desc);
|
||||
uint32_t *d = vd;
|
||||
int8_t *n = vn, *m = vm;
|
||||
|
||||
for (i = 0; i < opr_sz / 4; ++i) {
|
||||
d[i] += n[i * 4 + 0] * m[i * 4 + 0]
|
||||
+ n[i * 4 + 1] * m[i * 4 + 1]
|
||||
+ n[i * 4 + 2] * m[i * 4 + 2]
|
||||
+ n[i * 4 + 3] * m[i * 4 + 3];
|
||||
}
|
||||
clear_tail(d, opr_sz, simd_maxsz(desc));
|
||||
}
|
||||
|
||||
void HELPER(gvec_udot_b)(void *vd, void *vn, void *vm, uint32_t desc)
|
||||
{
|
||||
intptr_t i, opr_sz = simd_oprsz(desc);
|
||||
uint32_t *d = vd;
|
||||
uint8_t *n = vn, *m = vm;
|
||||
|
||||
for (i = 0; i < opr_sz / 4; ++i) {
|
||||
d[i] += n[i * 4 + 0] * m[i * 4 + 0]
|
||||
+ n[i * 4 + 1] * m[i * 4 + 1]
|
||||
+ n[i * 4 + 2] * m[i * 4 + 2]
|
||||
+ n[i * 4 + 3] * m[i * 4 + 3];
|
||||
}
|
||||
clear_tail(d, opr_sz, simd_maxsz(desc));
|
||||
}
|
||||
|
||||
void HELPER(gvec_sdot_h)(void *vd, void *vn, void *vm, uint32_t desc)
|
||||
{
|
||||
intptr_t i, opr_sz = simd_oprsz(desc);
|
||||
uint64_t *d = vd;
|
||||
int16_t *n = vn, *m = vm;
|
||||
|
||||
for (i = 0; i < opr_sz / 8; ++i) {
|
||||
d[i] += (int64_t)n[i * 4 + 0] * m[i * 4 + 0]
|
||||
+ (int64_t)n[i * 4 + 1] * m[i * 4 + 1]
|
||||
+ (int64_t)n[i * 4 + 2] * m[i * 4 + 2]
|
||||
+ (int64_t)n[i * 4 + 3] * m[i * 4 + 3];
|
||||
}
|
||||
clear_tail(d, opr_sz, simd_maxsz(desc));
|
||||
}
|
||||
|
||||
void HELPER(gvec_udot_h)(void *vd, void *vn, void *vm, uint32_t desc)
|
||||
{
|
||||
intptr_t i, opr_sz = simd_oprsz(desc);
|
||||
uint64_t *d = vd;
|
||||
uint16_t *n = vn, *m = vm;
|
||||
|
||||
for (i = 0; i < opr_sz / 8; ++i) {
|
||||
d[i] += (uint64_t)n[i * 4 + 0] * m[i * 4 + 0]
|
||||
+ (uint64_t)n[i * 4 + 1] * m[i * 4 + 1]
|
||||
+ (uint64_t)n[i * 4 + 2] * m[i * 4 + 2]
|
||||
+ (uint64_t)n[i * 4 + 3] * m[i * 4 + 3];
|
||||
}
|
||||
clear_tail(d, opr_sz, simd_maxsz(desc));
|
||||
}
|
||||
|
||||
void HELPER(gvec_fcaddh)(void *vd, void *vn, void *vm,
|
||||
void *vfpst, uint32_t desc)
|
||||
{
|
||||
|
|
|
@ -1189,6 +1189,8 @@
|
|||
#define helper_gvec_sar16i helper_gvec_sar16i_x86_64
|
||||
#define helper_gvec_sar32i helper_gvec_sar32i_x86_64
|
||||
#define helper_gvec_sar64i helper_gvec_sar64i_x86_64
|
||||
#define helper_gvec_sdot_b helper_gvec_sdot_b_x86_64
|
||||
#define helper_gvec_sdot_h helper_gvec_sdot_h_x86_64
|
||||
#define helper_gvec_shl8i helper_gvec_shl8i_x86_64
|
||||
#define helper_gvec_shl16i helper_gvec_shl16i_x86_64
|
||||
#define helper_gvec_shl32i helper_gvec_shl32i_x86_64
|
||||
|
@ -1213,6 +1215,8 @@
|
|||
#define helper_gvec_sssub16 helper_gvec_sssub16_x86_64
|
||||
#define helper_gvec_sssub32 helper_gvec_sssub32_x86_64
|
||||
#define helper_gvec_sssub64 helper_gvec_sssub64_x86_64
|
||||
#define helper_gvec_udot_b helper_gvec_udot_b_x86_64
|
||||
#define helper_gvec_udot_h helper_gvec_udot_h_x86_64
|
||||
#define helper_gvec_usadd8 helper_gvec_usadd8_x86_64
|
||||
#define helper_gvec_usadd16 helper_gvec_usadd16_x86_64
|
||||
#define helper_gvec_usadd32 helper_gvec_usadd32_x86_64
|
||||
|
|
Loading…
Reference in a new issue