qemu-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Qemu-devel] [PATCH v2 3/4] target-tilegx: Add double floating point imp


From: Chen Gang
Subject: [Qemu-devel] [PATCH v2 3/4] target-tilegx: Add double floating point implementation
Date: Tue, 17 Nov 2015 03:41:58 +0800

>From db171c94e8c9df446c091d9f42004c80ed8c6ccc Mon Sep 17 00:00:00 2001
From: Chen Gang <address@hidden>
Date: Tue, 17 Nov 2015 03:08:38 +0800
Subject: [PATCH v2 3/4] target-tilegx: Add double floating point implementation

Signed-off-by: Chen Gang <address@hidden>
---
 target-tilegx/helper-fdouble.c | 400 +++++++++++++++++++++++++++++++++++++++++
 1 file changed, 400 insertions(+)
 create mode 100644 target-tilegx/helper-fdouble.c

diff --git a/target-tilegx/helper-fdouble.c b/target-tilegx/helper-fdouble.c
new file mode 100644
index 0000000..f005c80
--- /dev/null
+++ b/target-tilegx/helper-fdouble.c
@@ -0,0 +1,400 @@
+/*
+ * QEMU TILE-Gx helpers
+ *
+ *  Copyright (c) 2015 Chen Gang
+ *
+ * This library is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * This library is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with this library; if not, see
+ * <http://www.gnu.org/licenses/lgpl-2.1.html>
+ */
+
+#include "cpu.h"
+#include "qemu-common.h"
+#include "exec/helper-proto.h"
+#include "fpu/softfloat.h"
+
+#include "helper-fshared.c"
+
+/*
+ * FDouble instructions implemenation:
+ *
+ * fdouble_unpack_min   ; srca and srcb are float_64 value.
+ *                      ; get the min absolute value's mantissa.
+ *                      ; move "mantissa>> (exp_max - exp_min)" to dest.
+ *
+ * fdouble_unpack_max   ; srca and srcb are float_64 value.
+ *                      ; get the max absolute value's mantissa.
+ *                      ; move mantissa to dest.
+ *
+ * fdouble_add_flags    ; srca and srcb are float_64 value.
+ *                      ; calc exp (exp_max), sign, and comp bits for flags.
+ *                      ; set addsub bit to flags and move flags to dest.
+ *
+ * fdouble_sub_flags    ; srca and srcb are float_64 value.
+ *                      ; calc exp (exp_max), sign, and comp bits for flags.
+ *                      ; set addsub bit to flags and move flags to dest.
+ *
+ * fdouble_addsub:      ; dest, srca (max, min mantissa), and srcb (flags).
+ *                      ; "dest +/- srca" depend on the add/sub bit of flags.
+ *                      ; move result mantissa to dest.
+ *
+ * fdouble_mul_flags:   ; srca and srcb are float_64 value.
+ *                      ; calc sign (xor), exp (min + max), and comp bits.
+ *                      ; mix sign, exp, and comp bits as flags to dest.
+ *
+ * fdouble_pack1        ; move srcb (flags) to dest.
+ *
+ * fdouble_pack2        ; srca, srcb (high, low mantissa), and dest (flags)
+ *                      ; normalize and pack result from srca, srcb, and dest.
+ *                      ; move result to dest.
+ */
+
+#define TILEGX_F_EXP_DZERO  0x3ff /* Zero exp for double 11-bits */
+#define TILEGX_F_EXP_DMAX   0x7fe /* max exp for double 11-bits */
+#define TILEGX_F_EXP_DUF    0x1000/* underflow exp bit for double */
+
+#define TILEGX_F_MAN_HBIT   (1ULL << 59)
+
+#define TILEGX_F_CALC_ADD   1     /* Perform absolute add operation */
+#define TILEGX_F_CALC_SUB   2     /* Perform absolute sub operation */
+#define TILEGX_F_CALC_MUL   3     /* Perform absolute mul operation */
+
+static uint32_t get_f64_exp(float64 d)
+{
+    return extract64(float64_val(d), 52, 11);
+}
+
+static void set_f64_exp(float64 *d, uint32_t exp)
+{
+    *d = make_float64(deposit64(float64_val(*d), 52, 11, exp));
+}
+
+static uint64_t get_f64_man(float64 d)
+{
+    return extract64(float64_val(d), 0, 52);
+}
+
+static uint64_t fr_to_man(float64 d)
+{
+    uint64_t val = get_f64_man(d) << 7;
+
+    if (get_f64_exp(d)) {
+        val |= TILEGX_F_MAN_HBIT;
+    }
+ 
+    return val;
+}
+
+static uint64_t get_fdouble_man(uint64_t n)
+{
+    return extract64(n, 0, 60);
+}
+
+static void set_fdouble_man(uint64_t *n, uint64_t man)
+{
+    *n = deposit64(*n, 0, 60, man);
+}
+
+static uint64_t get_fdouble_man_of(uint64_t n)
+{
+    return test_bit(60, &n);
+}
+
+static void clear_fdouble_man_of(uint64_t *n)
+{
+    return clear_bit(60, n);
+}
+
+static uint32_t get_fdouble_nan(uint64_t n)
+{
+    return test_bit(24, &n);
+}
+
+static void set_fdouble_nan(uint64_t *n)
+{
+    set_bit(24, n);
+}
+
+static uint32_t get_fdouble_inf(uint64_t n)
+{
+    return test_bit(23, &n);
+}
+
+static void set_fdouble_inf(uint64_t *n)
+{
+    set_bit(23, n);
+}
+
+static uint32_t get_fdouble_calc(uint64_t n)
+{
+    return extract32(n, 21, 2);
+}
+
+static void set_fdouble_calc(uint64_t *n, uint32_t calc)
+{
+    *n = deposit64(*n, 21, 2, calc);
+}
+
+static uint32_t get_fdouble_sign(uint64_t n)
+{
+    return test_bit(20, &n);
+}
+
+static void set_fdouble_sign(uint64_t *n)
+{
+    set_bit(20, n);
+}
+
+static uint32_t get_fdouble_vexp(uint64_t n)
+{
+    return extract32(n, 7, 13);
+}
+
+static void set_fdouble_vexp(uint64_t *n, uint32_t vexp)
+{
+    *n = deposit64(*n, 7, 13, vexp);
+}
+
+uint64_t helper_fdouble_unpack_min(CPUTLGState *env,
+                                   uint64_t srca, uint64_t srcb)
+{
+    uint64_t v = 0;
+    uint32_t expa = get_f64_exp(srca);
+    uint32_t expb = get_f64_exp(srcb);
+
+    if (float64_is_any_nan(srca) || float64_is_any_nan(srcb)
+        || float64_is_infinity(srca) || float64_is_infinity(srcb)) {
+        return 0;
+    } else if (expa> expb) {
+        if (expa - expb < 64) {
+            set_fdouble_man(&v, fr_to_man(srcb)>> (expa - expb));
+        } else {
+            return 0;
+        }
+    } else if (expa < expb) {
+        if (expb - expa < 64) {
+            set_fdouble_man(&v, fr_to_man(srca)>> (expb - expa));
+        } else {
+            return 0;
+        }
+    } else if (get_f64_man(srca)> get_f64_man(srcb)) {
+        set_fdouble_man(&v, fr_to_man(srcb));
+    } else {
+        set_fdouble_man(&v, fr_to_man(srca));
+    }
+
+    return v;
+}
+
+uint64_t helper_fdouble_unpack_max(CPUTLGState *env,
+                                   uint64_t srca, uint64_t srcb)
+{
+    uint64_t v = 0;
+    uint32_t expa = get_f64_exp(srca);
+    uint32_t expb = get_f64_exp(srcb);
+
+    if (float64_is_any_nan(srca) || float64_is_any_nan(srcb)
+        || float64_is_infinity(srca) || float64_is_infinity(srcb)) {
+        return 0;
+    } else if (expa> expb) {
+        set_fdouble_man(&v, fr_to_man(srca));
+    } else if (expa < expb) {
+        set_fdouble_man(&v, fr_to_man(srcb));
+    } else if (get_f64_man(srca)> get_f64_man(srcb)) {
+        set_fdouble_man(&v, fr_to_man(srca));
+    } else {
+        set_fdouble_man(&v, fr_to_man(srcb));
+    }
+
+    return v;
+}
+
+uint64_t helper_fdouble_addsub(CPUTLGState *env,
+                               uint64_t dest, uint64_t srca, uint64_t srcb)
+{
+    if (get_fdouble_calc(srcb) == TILEGX_F_CALC_ADD) {
+        return dest + srca; /* maybe set addsub overflow bit */
+    } else {
+        return dest - srca;
+    }
+}
+
+/* absolute-add/mul may cause add/mul carry or overflow */
+static bool proc_oflow(uint64_t *flags, uint64_t *v, uint64_t *srcb)
+{
+    if (get_fdouble_man_of(*v)) {
+        set_fdouble_vexp(flags, get_fdouble_vexp(*flags) + 1);
+        *srcb>>= 1;
+        *srcb |= *v << 63;
+        *v>>= 1;
+        clear_fdouble_man_of(v);
+    }
+    return get_fdouble_vexp(*flags)> TILEGX_F_EXP_DMAX;
+}
+
+uint64_t helper_fdouble_pack2(CPUTLGState *env, uint64_t flags /* dest */,
+                              uint64_t srca, uint64_t srcb)
+{
+    uint64_t v = srca;
+    float64 d = float64_set_sign(float64_zero, get_fdouble_sign(flags));
+
+    /*
+     * fdouble_add_flags, fdouble_sub_flags, or fdouble_mul_flags have
+     * processed exceptions. So need not process fp_status, again.
+     */
+
+    if (get_fdouble_nan(flags)) {
+        return float64_val(float64_default_nan);
+    } else if (get_fdouble_inf(flags)) {
+        return float64_val(d |= float64_infinity);
+    }
+
+    /* absolute-mul needs left shift 4 + 1 bytes to match the real mantissa */
+    if (get_fdouble_calc(flags) == TILEGX_F_CALC_MUL) {
+        v <<= 5;
+        v |= srcb>> 59;
+        srcb <<= 5;
+    }
+
+    /* must check underflow, firstly */
+    if (get_fdouble_vexp(flags) & TILEGX_F_EXP_DUF) {
+        return float64_val(d);
+    }
+
+    if (proc_oflow(&flags, &v, &srcb)) {
+        return float64_val(d |= float64_infinity);
+    }
+
+    while (!(get_fdouble_man(v) & TILEGX_F_MAN_HBIT)
+           && (get_fdouble_man(v)| srcb)) {
+        set_fdouble_vexp(&flags, get_fdouble_vexp(flags) - 1);
+        set_fdouble_man(&v, get_fdouble_man(v) << 1);
+        set_fdouble_man(&v, get_fdouble_man(v) | (srcb>> 63));
+        srcb <<= 1;
+    }
+
+    /* check underflow, again, after format */
+    if ((get_fdouble_vexp(flags) & TILEGX_F_EXP_DUF) || !get_fdouble_man(v)) {
+        return float64_val(d);
+    }
+
+    if (get_fdouble_sign(flags)) {
+        d = int64_to_float64(0 - get_fdouble_man(v), &env->fp_status);
+    } else {
+        d = uint64_to_float64(get_fdouble_man(v), &env->fp_status);
+    }
+
+    if (get_f64_exp(d) == 59 + TILEGX_F_EXP_DZERO) {
+        set_f64_exp(&d, get_fdouble_vexp(flags));
+    } else {                            /* for carry and overflow again */
+        set_f64_exp(&d, get_fdouble_vexp(flags) + 1);
+        if (get_f64_exp(d) == TILEGX_F_EXP_DMAX) {
+            d = float64_infinity;
+        }
+    }
+
+    d = float64_set_sign(d, get_fdouble_sign(flags));
+
+    return float64_val(d);
+}
+
+static void ana_bits(float_status *fp_status,
+                     float64 fsrca, float64 fsrcb, uint64_t *dfmt)
+{
+    if (float64_eq(fsrca, fsrcb, fp_status)) {
+        *dfmt |= create_fsfd_flag_eq();
+    } else {
+        *dfmt |= create_fsfd_flag_ne();
+    }
+
+    if (float64_lt(fsrca, fsrcb, fp_status)) {
+        *dfmt |= create_fsfd_flag_lt();
+    }
+    if (float64_le(fsrca, fsrcb, fp_status)) {
+        *dfmt |= create_fsfd_flag_le();
+    }
+
+    if (float64_lt(fsrcb, fsrca, fp_status)) {
+        *dfmt |= create_fsfd_flag_gt();
+    }
+    if (float64_le(fsrcb, fsrca, fp_status)) {
+        *dfmt |= create_fsfd_flag_ge();
+    }
+
+    if (float64_unordered(fsrca, fsrcb, fp_status)) {
+        *dfmt |= create_fsfd_flag_un();
+    }
+}
+
+static uint64_t main_calc(float_status *fp_status,
+                          float64 fsrca, float64 fsrcb,
+                          float64 (*calc)(float64, float64, float_status *))
+{
+    float64 d;
+    uint64_t flags = 0;
+    uint32_t expa = get_f64_exp(fsrca);
+    uint32_t expb = get_f64_exp(fsrcb);
+
+    ana_bits(fp_status, fsrca, fsrcb, &flags);
+
+    d = calc(fsrca, fsrcb, fp_status); /* also check exceptions */
+    if (float64_is_neg(d)) {
+        set_fdouble_sign(&flags);
+    }
+
+    if (float64_is_any_nan(d)) {
+        set_fdouble_nan(&flags);
+    } else if (float64_is_infinity(d)) {
+        set_fdouble_inf(&flags);
+    } else if (calc == float64_add) {
+        set_fdouble_vexp(&flags, (expa> expb) ? expa : expb);
+        set_fdouble_calc(&flags,
+                         (float64_is_neg(fsrca) == float64_is_neg(fsrcb))
+                             ? TILEGX_F_CALC_ADD : TILEGX_F_CALC_SUB);
+
+    } else if (calc == float64_sub) {
+        set_fdouble_vexp(&flags, (expa> expb) ? expa : expb);
+        set_fdouble_calc(&flags,
+                         (float64_is_neg(fsrca) != float64_is_neg(fsrcb))
+                             ? TILEGX_F_CALC_ADD : TILEGX_F_CALC_SUB);
+
+    } else {
+        set_fdouble_vexp(&flags, (int64_t)(expa - TILEGX_F_EXP_DZERO)
+                                 + (int64_t)(expb - TILEGX_F_EXP_DZERO)
+                                 + TILEGX_F_EXP_DZERO);
+        set_fdouble_calc(&flags, TILEGX_F_CALC_MUL);
+    }
+
+    return flags;
+}
+
+uint64_t helper_fdouble_add_flags(CPUTLGState *env,
+                                  uint64_t srca, uint64_t srcb)
+{
+    return main_calc(&env->fp_status,
+                     make_float64(srca), make_float64(srcb), float64_add);
+}
+
+uint64_t helper_fdouble_sub_flags(CPUTLGState *env,
+                                  uint64_t srca, uint64_t srcb)
+{
+    return main_calc(&env->fp_status,
+                     make_float64(srca), make_float64(srcb), float64_sub);
+}
+
+uint64_t helper_fdouble_mul_flags(CPUTLGState *env,
+                                  uint64_t srca, uint64_t srcb)
+{
+    return main_calc(&env->fp_status,
+                     make_float64(srca), make_float64(srcb), float64_mul);
+}
-- 
1.9.3

                                          

Attachment: v2-0003-target-tilegx-Add-double-floating-point-implement.patch
Description: Binary data


reply via email to

[Prev in Thread] Current Thread [Next in Thread]