From 0fa1ab0302badabc3581aefcbb2f189ef52c4985 Mon Sep 17 00:00:00 2001 From: Peter Maydell Date: Tue, 16 Jun 2020 10:32:25 +0100 Subject: [PATCH] target/arm: Convert Neon 3-reg-diff narrowing ops to decodetree Convert the narrow-to-high-half insns VADDHN, VSUBHN, VRADDHN, VRSUBHN in the Neon 3-registers-different-lengths group to decodetree. Signed-off-by: Peter Maydell Reviewed-by: Richard Henderson --- target/arm/neon-dp.decode | 6 +++ target/arm/translate-neon.inc.c | 87 +++++++++++++++++++++++++++++++++++++++ target/arm/translate.c | 91 +++++------------------------------------ 3 files changed, 104 insertions(+), 80 deletions(-) diff --git a/target/arm/neon-dp.decode b/target/arm/neon-dp.decode index 144a527ee6..a2234dfa4f 100644 --- a/target/arm/neon-dp.decode +++ b/target/arm/neon-dp.decode @@ -438,5 +438,11 @@ Vimm_1r 1111 001 . 1 . 000 ... .... cmode:4 0 . op:1 1 .... @1reg_imm VSUBW_S_3d 1111 001 0 1 . .. .... .... 0011 . 0 . 0 .... @3diff VSUBW_U_3d 1111 001 1 1 . .. .... .... 0011 . 0 . 0 .... @3diff + + VADDHN_3d 1111 001 0 1 . .. .... .... 0100 . 0 . 0 .... @3diff + VRADDHN_3d 1111 001 1 1 . .. .... .... 0100 . 0 . 0 .... @3diff + + VSUBHN_3d 1111 001 0 1 . .. .... .... 0110 . 0 . 0 .... @3diff + VRSUBHN_3d 1111 001 1 1 . .. .... .... 0110 . 0 . 0 .... @3diff ] } diff --git a/target/arm/translate-neon.inc.c b/target/arm/translate-neon.inc.c index 9b9d411107..0c3965802a 100644 --- a/target/arm/translate-neon.inc.c +++ b/target/arm/translate-neon.inc.c @@ -1932,3 +1932,90 @@ DO_PREWIDEN(VADDW_S, s, ext, add, true) DO_PREWIDEN(VADDW_U, u, extu, add, true) DO_PREWIDEN(VSUBW_S, s, ext, sub, true) DO_PREWIDEN(VSUBW_U, u, extu, sub, true) + +static bool do_narrow_3d(DisasContext *s, arg_3diff *a, + NeonGenTwo64OpFn *opfn, NeonGenNarrowFn *narrowfn) +{ + /* 3-regs different lengths, narrowing (VADDHN/VSUBHN/VRADDHN/VRSUBHN) */ + TCGv_i64 rn_64, rm_64; + TCGv_i32 rd0, rd1; + + if (!arm_dc_feature(s, ARM_FEATURE_NEON)) { + return false; + } + + /* UNDEF accesses to D16-D31 if they don't exist. */ + if (!dc_isar_feature(aa32_simd_r32, s) && + ((a->vd | a->vn | a->vm) & 0x10)) { + return false; + } + + if (!opfn || !narrowfn) { + /* size == 3 case, which is an entirely different insn group */ + return false; + } + + if ((a->vn | a->vm) & 1) { + return false; + } + + if (!vfp_access_check(s)) { + return true; + } + + rn_64 = tcg_temp_new_i64(); + rm_64 = tcg_temp_new_i64(); + rd0 = tcg_temp_new_i32(); + rd1 = tcg_temp_new_i32(); + + neon_load_reg64(rn_64, a->vn); + neon_load_reg64(rm_64, a->vm); + + opfn(rn_64, rn_64, rm_64); + + narrowfn(rd0, rn_64); + + neon_load_reg64(rn_64, a->vn + 1); + neon_load_reg64(rm_64, a->vm + 1); + + opfn(rn_64, rn_64, rm_64); + + narrowfn(rd1, rn_64); + + neon_store_reg(a->vd, 0, rd0); + neon_store_reg(a->vd, 1, rd1); + + tcg_temp_free_i64(rn_64); + tcg_temp_free_i64(rm_64); + + return true; +} + +#define DO_NARROW_3D(INSN, OP, NARROWTYPE, EXTOP) \ + static bool trans_##INSN##_3d(DisasContext *s, arg_3diff *a) \ + { \ + static NeonGenTwo64OpFn * const addfn[] = { \ + gen_helper_neon_##OP##l_u16, \ + gen_helper_neon_##OP##l_u32, \ + tcg_gen_##OP##_i64, \ + NULL, \ + }; \ + static NeonGenNarrowFn * const narrowfn[] = { \ + gen_helper_neon_##NARROWTYPE##_high_u8, \ + gen_helper_neon_##NARROWTYPE##_high_u16, \ + EXTOP, \ + NULL, \ + }; \ + return do_narrow_3d(s, a, addfn[a->size], narrowfn[a->size]); \ + } + +static void gen_narrow_round_high_u32(TCGv_i32 rd, TCGv_i64 rn) +{ + tcg_gen_addi_i64(rn, rn, 1u << 31); + tcg_gen_extrh_i64_i32(rd, rn); +} + +DO_NARROW_3D(VADDHN, add, narrow, tcg_gen_extrh_i64_i32) +DO_NARROW_3D(VSUBHN, sub, narrow, tcg_gen_extrh_i64_i32) +DO_NARROW_3D(VRADDHN, add, narrow_round, gen_narrow_round_high_u32) +DO_NARROW_3D(VRSUBHN, sub, narrow_round, gen_narrow_round_high_u32) diff --git a/target/arm/translate.c b/target/arm/translate.c index 9376534441..3fe39cd4f4 100644 --- a/target/arm/translate.c +++ b/target/arm/translate.c @@ -3231,16 +3231,6 @@ static inline void gen_neon_addl(int size) } } -static inline void gen_neon_subl(int size) -{ - switch (size) { - case 0: gen_helper_neon_subl_u16(CPU_V001); break; - case 1: gen_helper_neon_subl_u32(CPU_V001); break; - case 2: tcg_gen_sub_i64(CPU_V001); break; - default: abort(); - } -} - static inline void gen_neon_negl(TCGv_i64 var, int size) { switch (size) { @@ -5239,8 +5229,6 @@ static int disas_neon_data_insn(DisasContext *s, uint32_t insn) op = (insn >> 8) & 0xf; if ((insn & (1 << 6)) == 0) { /* Three registers of different lengths. */ - int src1_wide; - int src2_wide; /* undefreq: bit 0 : UNDEF if size == 0 * bit 1 : UNDEF if size == 1 * bit 2 : UNDEF if size == 2 @@ -5254,9 +5242,9 @@ static int disas_neon_data_insn(DisasContext *s, uint32_t insn) {0, 0, 0, 7}, /* VADDW: handled by decodetree */ {0, 0, 0, 7}, /* VSUBL: handled by decodetree */ {0, 0, 0, 7}, /* VSUBW: handled by decodetree */ - {0, 1, 1, 0}, /* VADDHN */ + {0, 0, 0, 7}, /* VADDHN: handled by decodetree */ {0, 0, 0, 0}, /* VABAL */ - {0, 1, 1, 0}, /* VSUBHN */ + {0, 0, 0, 7}, /* VSUBHN: handled by decodetree */ {0, 0, 0, 0}, /* VABDL */ {0, 0, 0, 0}, /* VMLAL */ {0, 0, 0, 9}, /* VQDMLAL */ @@ -5268,17 +5256,13 @@ static int disas_neon_data_insn(DisasContext *s, uint32_t insn) {0, 0, 0, 7}, /* Reserved: always UNDEF */ }; - src1_wide = neon_3reg_wide[op][1]; - src2_wide = neon_3reg_wide[op][2]; undefreq = neon_3reg_wide[op][3]; if ((undefreq & (1 << size)) || ((undefreq & 8) && u)) { return 1; } - if ((src1_wide && (rn & 1)) || - (src2_wide && (rm & 1)) || - (!src2_wide && (rd & 1))) { + if (rd & 1) { return 1; } @@ -5302,42 +5286,26 @@ static int disas_neon_data_insn(DisasContext *s, uint32_t insn) /* Avoid overlapping operands. Wide source operands are always aligned so will never overlap with wide destinations in problematic ways. */ - if (rd == rm && !src2_wide) { + if (rd == rm) { tmp = neon_load_reg(rm, 1); neon_store_scratch(2, tmp); - } else if (rd == rn && !src1_wide) { + } else if (rd == rn) { tmp = neon_load_reg(rn, 1); neon_store_scratch(2, tmp); } tmp3 = NULL; for (pass = 0; pass < 2; pass++) { - if (src1_wide) { - neon_load_reg64(cpu_V0, rn + pass); - tmp = NULL; + if (pass == 1 && rd == rn) { + tmp = neon_load_scratch(2); } else { - if (pass == 1 && rd == rn) { - tmp = neon_load_scratch(2); - } else { - tmp = neon_load_reg(rn, pass); - } + tmp = neon_load_reg(rn, pass); } - if (src2_wide) { - neon_load_reg64(cpu_V1, rm + pass); - tmp2 = NULL; + if (pass == 1 && rd == rm) { + tmp2 = neon_load_scratch(2); } else { - if (pass == 1 && rd == rm) { - tmp2 = neon_load_scratch(2); - } else { - tmp2 = neon_load_reg(rm, pass); - } + tmp2 = neon_load_reg(rm, pass); } switch (op) { - case 0: case 1: case 4: /* VADDL, VADDW, VADDHN, VRADDHN */ - gen_neon_addl(size); - break; - case 2: case 3: case 6: /* VSUBL, VSUBW, VSUBHN, VRSUBHN */ - gen_neon_subl(size); - break; case 5: case 7: /* VABAL, VABDL */ switch ((size << 1) | u) { case 0: @@ -5395,43 +5363,6 @@ static int disas_neon_data_insn(DisasContext *s, uint32_t insn) abort(); } neon_store_reg64(cpu_V0, rd + pass); - } else if (op == 4 || op == 6) { - /* Narrowing operation. */ - tmp = tcg_temp_new_i32(); - if (!u) { - switch (size) { - case 0: - gen_helper_neon_narrow_high_u8(tmp, cpu_V0); - break; - case 1: - gen_helper_neon_narrow_high_u16(tmp, cpu_V0); - break; - case 2: - tcg_gen_extrh_i64_i32(tmp, cpu_V0); - break; - default: abort(); - } - } else { - switch (size) { - case 0: - gen_helper_neon_narrow_round_high_u8(tmp, cpu_V0); - break; - case 1: - gen_helper_neon_narrow_round_high_u16(tmp, cpu_V0); - break; - case 2: - tcg_gen_addi_i64(cpu_V0, cpu_V0, 1u << 31); - tcg_gen_extrh_i64_i32(tmp, cpu_V0); - break; - default: abort(); - } - } - if (pass == 0) { - tmp3 = tmp; - } else { - neon_store_reg(rd, 0, tmp3); - neon_store_reg(rd, 1, tmp); - } } else { /* Write back the result. */ neon_store_reg64(cpu_V0, rd + pass); -- 2.11.4.GIT