qemu/target-arm/helper-a64.c
Aleksandar Markovic af39bc8c49 softfloat: Implement run-time-configurable meaning of signaling NaN bit
This patch modifies SoftFloat library so that it can be configured in
run-time in relation to the meaning of signaling NaN bit, while, at the
same time, strictly preserving its behavior on all existing platforms.

Background:

In floating-point calculations, there is a need for denoting undefined or
unrepresentable values. This is achieved by defining certain floating-point
numerical values to be NaNs (which stands for "not a number"). For additional
reasons, virtually all modern floating-point unit implementations use two
kinds of NaNs: quiet and signaling. The binary representations of these two
kinds of NaNs, as a rule, differ only in one bit (that bit is, traditionally,
the first bit of mantissa).

Up to 2008, standards for floating-point did not specify all details about
binary representation of NaNs. More specifically, the meaning of the bit
that is used for distinguishing between signaling and quiet NaNs was not
strictly prescribed. (IEEE 754-2008 was the first floating-point standard
that defined that meaning clearly, see [1], p. 35) As a result, different
platforms took different approaches, and that presented considerable
challenge for multi-platform emulators like QEMU.

Mips platform represents the most complex case among QEMU-supported
platforms regarding signaling NaN bit. Up to the Release 6 of Mips
architecture, "1" in signaling NaN bit denoted signaling NaN, which is
opposite to IEEE 754-2008 standard. From Release 6 on, Mips architecture
adopted IEEE standard prescription, and "0" denotes signaling NaN. On top of
that, Mips architecture for SIMD (also known as MSA, or vector instructions)
also specifies signaling bit in accordance to IEEE standard. MSA unit can be
implemented with both pre-Release 6 and Release 6 main processor units.

QEMU uses SoftFloat library to implement various floating-point-related
instructions on all platforms. The current QEMU implementation allows for
defining meaning of signaling NaN bit during build time, and is implemented
via preprocessor macro called SNAN_BIT_IS_ONE.

On the other hand, the change in this patch enables SoftFloat library to be
configured in run-time. This configuration is meant to occur during CPU
initialization, at the moment when it is definitely known what desired
behavior for particular CPU (or any additional FPUs) is.

The change is implemented so that it is consistent with existing
implementation of similar cases. This means that structure float_status is
used for passing the information about desired signaling NaN bit on each
invocation of SoftFloat functions. The additional field in float_status is
called snan_bit_is_one, which supersedes macro SNAN_BIT_IS_ONE.

IMPORTANT:

This change is not meant to create any change in emulator behavior or
functionality on any platform. It just provides the means for SoftFloat
library to be used in a more flexible way - in other words, it will just
prepare SoftFloat library for usage related to Mips platform and its
specifics regarding signaling bit meaning, which is done in some of
subsequent patches from this series.

Further break down of changes:

  1) Added field snan_bit_is_one to the structure float_status, and
     correspondent setter function set_snan_bit_is_one().

  2) Constants <float16|float32|float64|floatx80|float128>_default_nan
     (used both internally and externally) converted to functions
     <float16|float32|float64|floatx80|float128>_default_nan(float_status*).
     This is necessary since they are dependent on signaling bit meaning.
     At the same time, for the sake of code cleanup and simplicity, constants
     <floatx80|float128>_default_nan_<low|high> (used only internally within
     SoftFloat library) are removed, as not needed.

  3) Added a float_status* argument to SoftFloat library functions
     XXX_is_quiet_nan(XXX a_), XXX_is_signaling_nan(XXX a_),
     XXX_maybe_silence_nan(XXX a_). This argument must be present in
     order to enable correct invocation of new version of functions
     XXX_default_nan(). (XXX is <float16|float32|float64|floatx80|float128>
     here)

  4) Updated code for all platforms to reflect changes in SoftFloat library.
     This change is twofolds: it includes modifications of SoftFloat library
     functions invocations, and an addition of invocation of function
     set_snan_bit_is_one() during CPU initialization, with arguments that
     are appropriate for each particular platform. It was established that
     all platforms zero their main CPU data structures, so snan_bit_is_one(0)
     in appropriate places is not added, as it is not needed.

[1] "IEEE Standard for Floating-Point Arithmetic",
    IEEE Computer Society, August 29, 2008.

Signed-off-by: Thomas Schwinge <thomas@codesourcery.com>
Signed-off-by: Maciej W. Rozycki <macro@codesourcery.com>
Signed-off-by: Aleksandar Markovic <aleksandar.markovic@imgtec.com>
Tested-by: Bastian Koppelmann <kbastian@mail.uni-paderborn.de>
Reviewed-by: Leon Alrae <leon.alrae@imgtec.com>
Tested-by: Leon Alrae <leon.alrae@imgtec.com>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
[leon.alrae@imgtec.com:
 * cherry-picked 2 chunks from patch #2 to fix compilation warnings]
Signed-off-by: Leon Alrae <leon.alrae@imgtec.com>
2016-06-24 13:40:37 +01:00

447 lines
13 KiB
C

/*
* AArch64 specific helpers
*
* Copyright (c) 2013 Alexander Graf <agraf@suse.de>
*
* 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 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/>.
*/
#include "qemu/osdep.h"
#include "cpu.h"
#include "exec/gdbstub.h"
#include "exec/helper-proto.h"
#include "qemu/host-utils.h"
#include "qemu/log.h"
#include "sysemu/sysemu.h"
#include "qemu/bitops.h"
#include "internals.h"
#include "qemu/crc32c.h"
#include <zlib.h> /* For crc32 */
/* C2.4.7 Multiply and divide */
/* special cases for 0 and LLONG_MIN are mandated by the standard */
uint64_t HELPER(udiv64)(uint64_t num, uint64_t den)
{
if (den == 0) {
return 0;
}
return num / den;
}
int64_t HELPER(sdiv64)(int64_t num, int64_t den)
{
if (den == 0) {
return 0;
}
if (num == LLONG_MIN && den == -1) {
return LLONG_MIN;
}
return num / den;
}
uint64_t HELPER(clz64)(uint64_t x)
{
return clz64(x);
}
uint64_t HELPER(cls64)(uint64_t x)
{
return clrsb64(x);
}
uint32_t HELPER(cls32)(uint32_t x)
{
return clrsb32(x);
}
uint32_t HELPER(clz32)(uint32_t x)
{
return clz32(x);
}
uint64_t HELPER(rbit64)(uint64_t x)
{
return revbit64(x);
}
/* Convert a softfloat float_relation_ (as returned by
* the float*_compare functions) to the correct ARM
* NZCV flag state.
*/
static inline uint32_t float_rel_to_flags(int res)
{
uint64_t flags;
switch (res) {
case float_relation_equal:
flags = PSTATE_Z | PSTATE_C;
break;
case float_relation_less:
flags = PSTATE_N;
break;
case float_relation_greater:
flags = PSTATE_C;
break;
case float_relation_unordered:
default:
flags = PSTATE_C | PSTATE_V;
break;
}
return flags;
}
uint64_t HELPER(vfp_cmps_a64)(float32 x, float32 y, void *fp_status)
{
return float_rel_to_flags(float32_compare_quiet(x, y, fp_status));
}
uint64_t HELPER(vfp_cmpes_a64)(float32 x, float32 y, void *fp_status)
{
return float_rel_to_flags(float32_compare(x, y, fp_status));
}
uint64_t HELPER(vfp_cmpd_a64)(float64 x, float64 y, void *fp_status)
{
return float_rel_to_flags(float64_compare_quiet(x, y, fp_status));
}
uint64_t HELPER(vfp_cmped_a64)(float64 x, float64 y, void *fp_status)
{
return float_rel_to_flags(float64_compare(x, y, fp_status));
}
float32 HELPER(vfp_mulxs)(float32 a, float32 b, void *fpstp)
{
float_status *fpst = fpstp;
a = float32_squash_input_denormal(a, fpst);
b = float32_squash_input_denormal(b, fpst);
if ((float32_is_zero(a) && float32_is_infinity(b)) ||
(float32_is_infinity(a) && float32_is_zero(b))) {
/* 2.0 with the sign bit set to sign(A) XOR sign(B) */
return make_float32((1U << 30) |
((float32_val(a) ^ float32_val(b)) & (1U << 31)));
}
return float32_mul(a, b, fpst);
}
float64 HELPER(vfp_mulxd)(float64 a, float64 b, void *fpstp)
{
float_status *fpst = fpstp;
a = float64_squash_input_denormal(a, fpst);
b = float64_squash_input_denormal(b, fpst);
if ((float64_is_zero(a) && float64_is_infinity(b)) ||
(float64_is_infinity(a) && float64_is_zero(b))) {
/* 2.0 with the sign bit set to sign(A) XOR sign(B) */
return make_float64((1ULL << 62) |
((float64_val(a) ^ float64_val(b)) & (1ULL << 63)));
}
return float64_mul(a, b, fpst);
}
uint64_t HELPER(simd_tbl)(CPUARMState *env, uint64_t result, uint64_t indices,
uint32_t rn, uint32_t numregs)
{
/* Helper function for SIMD TBL and TBX. We have to do the table
* lookup part for the 64 bits worth of indices we're passed in.
* result is the initial results vector (either zeroes for TBL
* or some guest values for TBX), rn the register number where
* the table starts, and numregs the number of registers in the table.
* We return the results of the lookups.
*/
int shift;
for (shift = 0; shift < 64; shift += 8) {
int index = extract64(indices, shift, 8);
if (index < 16 * numregs) {
/* Convert index (a byte offset into the virtual table
* which is a series of 128-bit vectors concatenated)
* into the correct vfp.regs[] element plus a bit offset
* into that element, bearing in mind that the table
* can wrap around from V31 to V0.
*/
int elt = (rn * 2 + (index >> 3)) % 64;
int bitidx = (index & 7) * 8;
uint64_t val = extract64(env->vfp.regs[elt], bitidx, 8);
result = deposit64(result, shift, 8, val);
}
}
return result;
}
/* 64bit/double versions of the neon float compare functions */
uint64_t HELPER(neon_ceq_f64)(float64 a, float64 b, void *fpstp)
{
float_status *fpst = fpstp;
return -float64_eq_quiet(a, b, fpst);
}
uint64_t HELPER(neon_cge_f64)(float64 a, float64 b, void *fpstp)
{
float_status *fpst = fpstp;
return -float64_le(b, a, fpst);
}
uint64_t HELPER(neon_cgt_f64)(float64 a, float64 b, void *fpstp)
{
float_status *fpst = fpstp;
return -float64_lt(b, a, fpst);
}
/* Reciprocal step and sqrt step. Note that unlike the A32/T32
* versions, these do a fully fused multiply-add or
* multiply-add-and-halve.
*/
#define float32_two make_float32(0x40000000)
#define float32_three make_float32(0x40400000)
#define float32_one_point_five make_float32(0x3fc00000)
#define float64_two make_float64(0x4000000000000000ULL)
#define float64_three make_float64(0x4008000000000000ULL)
#define float64_one_point_five make_float64(0x3FF8000000000000ULL)
float32 HELPER(recpsf_f32)(float32 a, float32 b, void *fpstp)
{
float_status *fpst = fpstp;
a = float32_squash_input_denormal(a, fpst);
b = float32_squash_input_denormal(b, fpst);
a = float32_chs(a);
if ((float32_is_infinity(a) && float32_is_zero(b)) ||
(float32_is_infinity(b) && float32_is_zero(a))) {
return float32_two;
}
return float32_muladd(a, b, float32_two, 0, fpst);
}
float64 HELPER(recpsf_f64)(float64 a, float64 b, void *fpstp)
{
float_status *fpst = fpstp;
a = float64_squash_input_denormal(a, fpst);
b = float64_squash_input_denormal(b, fpst);
a = float64_chs(a);
if ((float64_is_infinity(a) && float64_is_zero(b)) ||
(float64_is_infinity(b) && float64_is_zero(a))) {
return float64_two;
}
return float64_muladd(a, b, float64_two, 0, fpst);
}
float32 HELPER(rsqrtsf_f32)(float32 a, float32 b, void *fpstp)
{
float_status *fpst = fpstp;
a = float32_squash_input_denormal(a, fpst);
b = float32_squash_input_denormal(b, fpst);
a = float32_chs(a);
if ((float32_is_infinity(a) && float32_is_zero(b)) ||
(float32_is_infinity(b) && float32_is_zero(a))) {
return float32_one_point_five;
}
return float32_muladd(a, b, float32_three, float_muladd_halve_result, fpst);
}
float64 HELPER(rsqrtsf_f64)(float64 a, float64 b, void *fpstp)
{
float_status *fpst = fpstp;
a = float64_squash_input_denormal(a, fpst);
b = float64_squash_input_denormal(b, fpst);
a = float64_chs(a);
if ((float64_is_infinity(a) && float64_is_zero(b)) ||
(float64_is_infinity(b) && float64_is_zero(a))) {
return float64_one_point_five;
}
return float64_muladd(a, b, float64_three, float_muladd_halve_result, fpst);
}
/* Pairwise long add: add pairs of adjacent elements into
* double-width elements in the result (eg _s8 is an 8x8->16 op)
*/
uint64_t HELPER(neon_addlp_s8)(uint64_t a)
{
uint64_t nsignmask = 0x0080008000800080ULL;
uint64_t wsignmask = 0x8000800080008000ULL;
uint64_t elementmask = 0x00ff00ff00ff00ffULL;
uint64_t tmp1, tmp2;
uint64_t res, signres;
/* Extract odd elements, sign extend each to a 16 bit field */
tmp1 = a & elementmask;
tmp1 ^= nsignmask;
tmp1 |= wsignmask;
tmp1 = (tmp1 - nsignmask) ^ wsignmask;
/* Ditto for the even elements */
tmp2 = (a >> 8) & elementmask;
tmp2 ^= nsignmask;
tmp2 |= wsignmask;
tmp2 = (tmp2 - nsignmask) ^ wsignmask;
/* calculate the result by summing bits 0..14, 16..22, etc,
* and then adjusting the sign bits 15, 23, etc manually.
* This ensures the addition can't overflow the 16 bit field.
*/
signres = (tmp1 ^ tmp2) & wsignmask;
res = (tmp1 & ~wsignmask) + (tmp2 & ~wsignmask);
res ^= signres;
return res;
}
uint64_t HELPER(neon_addlp_u8)(uint64_t a)
{
uint64_t tmp;
tmp = a & 0x00ff00ff00ff00ffULL;
tmp += (a >> 8) & 0x00ff00ff00ff00ffULL;
return tmp;
}
uint64_t HELPER(neon_addlp_s16)(uint64_t a)
{
int32_t reslo, reshi;
reslo = (int32_t)(int16_t)a + (int32_t)(int16_t)(a >> 16);
reshi = (int32_t)(int16_t)(a >> 32) + (int32_t)(int16_t)(a >> 48);
return (uint32_t)reslo | (((uint64_t)reshi) << 32);
}
uint64_t HELPER(neon_addlp_u16)(uint64_t a)
{
uint64_t tmp;
tmp = a & 0x0000ffff0000ffffULL;
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, fpst)) {
float_raise(float_flag_invalid, fpst);
nan = float32_maybe_silence_nan(a, fpst);
}
if (fpst->default_nan_mode) {
nan = float32_default_nan(fpst);
}
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, fpst)) {
float_raise(float_flag_invalid, fpst);
nan = float64_maybe_silence_nan(a, fpst);
}
if (fpst->default_nan_mode) {
nan = float64_default_nan(fpst);
}
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);
}
}
float32 HELPER(fcvtx_f64_to_f32)(float64 a, CPUARMState *env)
{
/* Von Neumann rounding is implemented by using round-to-zero
* and then setting the LSB of the result if Inexact was raised.
*/
float32 r;
float_status *fpst = &env->vfp.fp_status;
float_status tstat = *fpst;
int exflags;
set_float_rounding_mode(float_round_to_zero, &tstat);
set_float_exception_flags(0, &tstat);
r = float64_to_float32(a, &tstat);
r = float32_maybe_silence_nan(r, &tstat);
exflags = get_float_exception_flags(&tstat);
if (exflags & float_flag_inexact) {
r = make_float32(float32_val(r) | 1);
}
exflags |= get_float_exception_flags(fpst);
set_float_exception_flags(exflags, fpst);
return r;
}
/* 64-bit versions of the CRC helpers. Note that although the operation
* (and the prototypes of crc32c() and crc32() mean that only the bottom
* 32 bits of the accumulator and result are used, we pass and return
* uint64_t for convenience of the generated code. Unlike the 32-bit
* instruction set versions, val may genuinely have 64 bits of data in it.
* The upper bytes of val (above the number specified by 'bytes') must have
* been zeroed out by the caller.
*/
uint64_t HELPER(crc32_64)(uint64_t acc, uint64_t val, uint32_t bytes)
{
uint8_t buf[8];
stq_le_p(buf, val);
/* zlib crc32 converts the accumulator and output to one's complement. */
return crc32(acc ^ 0xffffffff, buf, bytes) ^ 0xffffffff;
}
uint64_t HELPER(crc32c_64)(uint64_t acc, uint64_t val, uint32_t bytes)
{
uint8_t buf[8];
stq_le_p(buf, val);
/* Linux crc32c converts the output to one's complement. */
return crc32c(acc, buf, bytes) ^ 0xffffffff;
}