[Qemu-devel] [PATCH 13/16] target-arm: A64: Add FRECPX (reciprocal exponent)

[Peter Maydell <peter.maydell@linaro.org>]

From: Alex Bennée <alex.bennee@linaro.org>

These are fairly simple exponent only estimation functions using helpers.

Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
---
 target-arm/helper-a64.c    | 59 ++++++++++++++++++++++++++++++++++++++
 target-arm/helper-a64.h    |  2 ++
 target-arm/translate-a64.c | 70 +++++++++++++++++++++++++++++++++++++++++++++-
 3 files changed, 130 insertions(+), 1 deletion(-)

diff --git a/target-arm/helper-a64.c b/target-arm/helper-a64.c
index c31c45e..cea2468 100644
--- a/target-arm/helper-a64.c
+++ b/target-arm/helper-a64.c
@@ -354,3 +354,62 @@ uint64_t HELPER(neon_addlp_u16)(uint64_t a)
     tmp += (a >> 16) & 0x0000ffff0000ffffULL;
     return tmp;
 }
+
+/* Floating-point reciprocal exponent - see FPRecpX in ARM ARM */
+float32 HELPER(frecpx_f32)(float32 a, void *fpstp)
+{
+    float_status *fpst = fpstp;
+    uint32_t val32, sbit;
+    int32_t exp;
+
+    if (float32_is_any_nan(a)) {
+        float32 nan = a;
+        if (float32_is_signaling_nan(a)) {
+            float_raise(float_flag_invalid, fpst);
+            nan = float32_maybe_silence_nan(a);
+        }
+        if (fpst->default_nan_mode) {
+            nan = float32_default_nan;
+        }
+        return nan;
+    }
+
+    val32 = float32_val(a);
+    sbit = 0x80000000ULL & val32;
+    exp = extract32(val32, 23, 8);
+
+    if (exp == 0) {
+        return make_float32(sbit | (0xfe << 23));
+    } else {
+        return make_float32(sbit | (~exp & 0xff) << 23);
+    }
+}
+
+float64 HELPER(frecpx_f64)(float64 a, void *fpstp)
+{
+    float_status *fpst = fpstp;
+    uint64_t val64, sbit;
+    int64_t exp;
+
+    if (float64_is_any_nan(a)) {
+        float64 nan = a;
+        if (float64_is_signaling_nan(a)) {
+            float_raise(float_flag_invalid, fpst);
+            nan = float64_maybe_silence_nan(a);
+        }
+        if (fpst->default_nan_mode) {
+            nan = float64_default_nan;
+        }
+        return nan;
+    }
+
+    val64 = float64_val(a);
+    sbit = 0x8000000000000000ULL & val64;
+    exp = extract64(float64_val(a), 52, 11);
+
+    if (exp == 0) {
+        return make_float64(sbit | (0x7feULL << 52));
+    } else {
+        return make_float64(sbit | (~exp & 0x7ffULL) << 52);
+    }
+}
diff --git a/target-arm/helper-a64.h b/target-arm/helper-a64.h
index 88fc9fe..8cbc349 100644
--- a/target-arm/helper-a64.h
+++ b/target-arm/helper-a64.h
@@ -43,3 +43,5 @@ DEF_HELPER_FLAGS_1(neon_addlp_s8, TCG_CALL_NO_RWG_SE, i64, i64)
 DEF_HELPER_FLAGS_1(neon_addlp_u8, TCG_CALL_NO_RWG_SE, i64, i64)
 DEF_HELPER_FLAGS_1(neon_addlp_s16, TCG_CALL_NO_RWG_SE, i64, i64)
 DEF_HELPER_FLAGS_1(neon_addlp_u16, TCG_CALL_NO_RWG_SE, i64, i64)
+DEF_HELPER_FLAGS_2(frecpx_f64, TCG_CALL_NO_RWG, f64, f64, ptr)
+DEF_HELPER_FLAGS_2(frecpx_f32, TCG_CALL_NO_RWG, f32, f32, ptr)
diff --git a/target-arm/translate-a64.c b/target-arm/translate-a64.c
index 8bf66a6..5f560eb 100644
--- a/target-arm/translate-a64.c
+++ b/target-arm/translate-a64.c
@@ -6886,6 +6886,72 @@ static void handle_2misc_fcmp_zero(DisasContext *s, int opcode,
     tcg_temp_free_ptr(fpst);
 }
 
+static void handle_2misc_reciprocal(DisasContext *s, int opcode,
+                                    bool is_scalar, bool is_u, bool is_q,
+                                    int size, int rn, int rd)
+{
+    bool is_double = (size == 3);
+    TCGv_ptr fpst = get_fpstatus_ptr();
+
+    if (is_double) {
+        TCGv_i64 tcg_op = tcg_temp_new_i64();
+        TCGv_i64 tcg_res = tcg_temp_new_i64();
+        int pass;
+
+        for (pass = 0; pass < (is_scalar ? 1 : 2); pass++) {
+            read_vec_element(s, tcg_op, rn, pass, MO_64);
+            switch (opcode) {
+            case 0x3f: /* FRECPX */
+                gen_helper_frecpx_f64(tcg_res, tcg_op, fpst);
+                break;
+            default:
+                g_assert_not_reached();
+            }
+            write_vec_element(s, tcg_res, rd, pass, MO_64);
+        }
+        if (is_scalar) {
+            clear_vec_high(s, rd);
+        }
+
+        tcg_temp_free_i64(tcg_res);
+        tcg_temp_free_i64(tcg_op);
+    } else {
+        TCGv_i32 tcg_op = tcg_temp_new_i32();
+        TCGv_i32 tcg_res = tcg_temp_new_i32();
+        int pass, maxpasses;
+
+        if (is_scalar) {
+            maxpasses = 1;
+        } else {
+            maxpasses = is_q ? 4 : 2;
+        }
+
+        for (pass = 0; pass < maxpasses; pass++) {
+            read_vec_element_i32(s, tcg_op, rn, pass, MO_32);
+
+            switch (opcode) {
+            case 0x3f: /* FRECPX */
+                gen_helper_frecpx_f32(tcg_res, tcg_op, fpst);
+                break;
+            default:
+                g_assert_not_reached();
+            }
+
+            if (is_scalar) {
+                write_fp_sreg(s, rd, tcg_res);
+            } else {
+                write_vec_element_i32(s, tcg_res, rd, pass, MO_32);
+            }
+        }
+        tcg_temp_free_i32(tcg_res);
+        tcg_temp_free_i32(tcg_op);
+        if (!is_q && !is_scalar) {
+            clear_vec_high(s, rd);
+        }
+    }
+    tcg_temp_free_ptr(fpst);
+}
+
 /* C3.6.12 AdvSIMD scalar two reg misc
  *  31 30  29 28       24 23  22 21       17 16    12 11 10 9    5 4    0
  * +-----+---+-----------+------+-----------+--------+-----+------+------+
@@ -6942,6 +7008,9 @@ static void disas_simd_scalar_two_reg_misc(DisasContext *s, uint32_t insn)
             handle_simd_intfp_conv(s, rd, rn, 1, is_signed, 0, size);
             return;
         }
+        case 0x3f: /* FRECPX */
+            handle_2misc_reciprocal(s, opcode, true, u, true, size, rn, rd);
+            return;
         case 0x1a: /* FCVTNS */
         case 0x1b: /* FCVTMS */
         case 0x3a: /* FCVTPS */
@@ -6960,7 +7029,6 @@ static void disas_simd_scalar_two_reg_misc(DisasContext *s, uint32_t insn)
             rmode = FPROUNDING_TIEAWAY;
             break;
         case 0x3d: /* FRECPE */
-        case 0x3f: /* FRECPX */
         case 0x56: /* FCVTXN, FCVTXN2 */
         case 0x7d: /* FRSQRTE */
             unsupported_encoding(s, insn);
-- 
1.9.0