diff mbox series

[52/67] target/arm: Convert FABS, FNEG (vector) to decodetree

Message ID 20241201150607.12812-53-richard.henderson@linaro.org
State Superseded
Headers show
Series target/arm: AArch64 decodetree conversion, final part | expand

Commit Message

Richard Henderson Dec. 1, 2024, 3:05 p.m. UTC 
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
 target/arm/tcg/translate-a64.c | 61 ++++++++++++++++++----------------
 target/arm/tcg/a64.decode      |  7 ++++
 2 files changed, 39 insertions(+), 29 deletions(-)

Comments

Peter Maydell Dec. 6, 2024, 4:03 p.m. UTC | #1 
On Sun, 1 Dec 2024 at 15:19, Richard Henderson
<richard.henderson@linaro.org> wrote:
>
> Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
> ---
>  target/arm/tcg/translate-a64.c | 61 ++++++++++++++++++----------------
>  target/arm/tcg/a64.decode      |  7 ++++
>  2 files changed, 39 insertions(+), 29 deletions(-)
>
> diff --git a/target/arm/tcg/translate-a64.c b/target/arm/tcg/translate-a64.c
> index 613dcdb9a2..31272c1878 100644
> --- a/target/arm/tcg/translate-a64.c
> +++ b/target/arm/tcg/translate-a64.c
> @@ -9153,6 +9153,28 @@ static bool trans_SHLL_v(DisasContext *s, arg_qrr_e *a)
>      return true;
>  }
>
> +static bool do_fabs_fneg_v(DisasContext *s, arg_qrr_e *a, bool neg)
> +{
> +    int check = fp_access_check_vector_hsd(s, a->q, a->esz);
> +    uint64_t sign;
> +
> +    if (check <= 0) {
> +        return check == 0;
> +    }
> +
> +    sign = 1ull << ((8 << a->esz) - 1);
> +    if (neg) {
> +        gen_gvec_fn2i(s, a->q, a->rd, a->rn, sign,
> +                      tcg_gen_gvec_xori, a->esz);
> +    } else {
> +        gen_gvec_fn2i(s, a->q, a->rd, a->rn, sign - 1,
> +                      tcg_gen_gvec_andi, a->esz);
> +    }
> +    return true;
> +}

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>

Annoying FEAT_AFP wrinkle: for FPCR.AH=1 we will need to make
fabs and fneg not flip the sign bit for NaNs. I guess that means
AH will need to be a tbflags bit so we can generate the nice
vector code for AH=0 and fall back to something else for AH=1.

thanks
-- PMM
diff mbox series

Patch

diff --git a/target/arm/tcg/translate-a64.c b/target/arm/tcg/translate-a64.c
index 613dcdb9a2..31272c1878 100644
--- a/target/arm/tcg/translate-a64.c
+++ b/target/arm/tcg/translate-a64.c
@@ -9153,6 +9153,28 @@  static bool trans_SHLL_v(DisasContext *s, arg_qrr_e *a)
     return true;
 }
 
+static bool do_fabs_fneg_v(DisasContext *s, arg_qrr_e *a, bool neg)
+{
+    int check = fp_access_check_vector_hsd(s, a->q, a->esz);
+    uint64_t sign;
+
+    if (check <= 0) {
+        return check == 0;
+    }
+
+    sign = 1ull << ((8 << a->esz) - 1);
+    if (neg) {
+        gen_gvec_fn2i(s, a->q, a->rd, a->rn, sign,
+                      tcg_gen_gvec_xori, a->esz);
+    } else {
+        gen_gvec_fn2i(s, a->q, a->rd, a->rn, sign - 1,
+                      tcg_gen_gvec_andi, a->esz);
+    }
+    return true;
+}
+
+TRANS(FABS_v, do_fabs_fneg_v, a, false)
+TRANS(FNEG_v, do_fabs_fneg_v, a, true)
 
 /* Common vector code for handling integer to FP conversion */
 static void handle_simd_intfp_conv(DisasContext *s, int rd, int rn,
@@ -9451,12 +9473,6 @@  static void handle_2misc_64(DisasContext *s, int opcode, bool u,
      * requires them.
      */
     switch (opcode) {
-    case 0x2f: /* FABS */
-        gen_vfp_absd(tcg_rd, tcg_rn);
-        break;
-    case 0x6f: /* FNEG */
-        gen_vfp_negd(tcg_rd, tcg_rn);
-        break;
     case 0x7f: /* FSQRT */
         gen_helper_vfp_sqrtd(tcg_rd, tcg_rn, tcg_fpstatus);
         break;
@@ -9501,6 +9517,8 @@  static void handle_2misc_64(DisasContext *s, int opcode, bool u,
     case 0x9: /* CMEQ, CMLE */
     case 0xa: /* CMLT */
     case 0xb: /* ABS, NEG */
+    case 0x2f: /* FABS */
+    case 0x6f: /* FNEG */
         g_assert_not_reached();
     }
 }
@@ -9972,13 +9990,6 @@  static void disas_simd_two_reg_misc(DisasContext *s, uint32_t insn)
         opcode |= (extract32(size, 1, 1) << 5) | (u << 6);
         size = is_double ? 3 : 2;
         switch (opcode) {
-        case 0x2f: /* FABS */
-        case 0x6f: /* FNEG */
-            if (size == 3 && !is_q) {
-                unallocated_encoding(s);
-                return;
-            }
-            break;
         case 0x1d: /* SCVTF */
         case 0x5d: /* UCVTF */
         {
@@ -10103,6 +10114,8 @@  static void disas_simd_two_reg_misc(DisasContext *s, uint32_t insn)
         case 0x16: /* FCVTN, FCVTN2 */
         case 0x36: /* BFCVTN, BFCVTN2 */
         case 0x56: /* FCVTXN, FCVTXN2 */
+        case 0x2f: /* FABS */
+        case 0x6f: /* FNEG */
             unallocated_encoding(s);
             return;
         }
@@ -10175,12 +10188,6 @@  static void disas_simd_two_reg_misc(DisasContext *s, uint32_t insn)
             {
                 /* Special cases for 32 bit elements */
                 switch (opcode) {
-                case 0x2f: /* FABS */
-                    gen_vfp_abss(tcg_res, tcg_op);
-                    break;
-                case 0x6f: /* FNEG */
-                    gen_vfp_negs(tcg_res, tcg_op);
-                    break;
                 case 0x7f: /* FSQRT */
                     gen_helper_vfp_sqrts(tcg_res, tcg_op, tcg_fpstatus);
                     break;
@@ -10224,6 +10231,8 @@  static void disas_simd_two_reg_misc(DisasContext *s, uint32_t insn)
                     break;
                 default:
                 case 0x7: /* SQABS, SQNEG */
+                case 0x2f: /* FABS */
+                case 0x6f: /* FNEG */
                     g_assert_not_reached();
                 }
             }
@@ -10366,14 +10375,12 @@  static void disas_simd_two_reg_misc_fp16(DisasContext *s, uint32_t insn)
     case 0x7b: /* FCVTZU */
         rmode = FPROUNDING_ZERO;
         break;
-    case 0x2f: /* FABS */
-    case 0x6f: /* FNEG */
-        need_fpst = false;
-        break;
     case 0x7d: /* FRSQRTE */
     case 0x7f: /* FSQRT (vector) */
         break;
     default:
+    case 0x2f: /* FABS */
+    case 0x6f: /* FNEG */
         unallocated_encoding(s);
         return;
     }
@@ -10476,12 +10483,6 @@  static void disas_simd_two_reg_misc_fp16(DisasContext *s, uint32_t insn)
             case 0x59: /* FRINTX */
                 gen_helper_advsimd_rinth_exact(tcg_res, tcg_op, tcg_fpstatus);
                 break;
-            case 0x2f: /* FABS */
-                tcg_gen_andi_i32(tcg_res, tcg_op, 0x7fff);
-                break;
-            case 0x6f: /* FNEG */
-                tcg_gen_xori_i32(tcg_res, tcg_op, 0x8000);
-                break;
             case 0x7d: /* FRSQRTE */
                 gen_helper_rsqrte_f16(tcg_res, tcg_op, tcg_fpstatus);
                 break;
@@ -10489,6 +10490,8 @@  static void disas_simd_two_reg_misc_fp16(DisasContext *s, uint32_t insn)
                 gen_helper_vfp_sqrth(tcg_res, tcg_op, tcg_fpstatus);
                 break;
             default:
+            case 0x2f: /* FABS */
+            case 0x6f: /* FNEG */
                 g_assert_not_reached();
             }
 
diff --git a/target/arm/tcg/a64.decode b/target/arm/tcg/a64.decode
index 8f91b9af1a..fec6aa7a23 100644
--- a/target/arm/tcg/a64.decode
+++ b/target/arm/tcg/a64.decode
@@ -77,6 +77,7 @@ 
 @qrr_s          . q:1 ...... .. ...... ...... rn:5 rd:5  &qrr_e esz=2
 @qrr_bh         . q:1 ...... . esz:1 ...... ...... rn:5 rd:5  &qrr_e
 @qrr_hs         . q:1 ...... .. ...... ...... rn:5 rd:5  &qrr_e esz=%esz_hs
+@qrr_sd         . q:1 ...... .. ...... ...... rn:5 rd:5  &qrr_e esz=%esz_sd
 @qrr_e          . q:1 ...... esz:2 ...... ...... rn:5 rd:5  &qrr_e
 
 @qrrr_b         . q:1 ...... ... rm:5 ...... rn:5 rd:5  &qrrr_e esz=0
@@ -1687,3 +1688,9 @@  FCVTXN_v        0.10 1110 011 00001 01101 0 ..... .....     @qrr_s
 BFCVTN_v        0.00 1110 101 00001 01101 0 ..... .....     @qrr_h
 
 SHLL_v          0.10 1110 ..1 00001 00111 0 ..... .....     @qrr_e
+
+FABS_v          0.00 1110 111 11000 11111 0 ..... .....     @qrr_h
+FABS_v          0.00 1110 1.1 00000 11111 0 ..... .....     @qrr_sd
+
+FNEG_v          0.10 1110 111 11000 11111 0 ..... .....     @qrr_h
+FNEG_v          0.10 1110 1.1 00000 11111 0 ..... .....     @qrr_sd