korg/linux
0
0
Fork 0
mirror of https://mirrors.bfsu.edu.cn/git/linux.git synced 2024-11-19 10:14:23 +08:00
Code Issues Packages Projects Releases Wiki Activity
47226ad4f4
linux/arch/arm/vfp/vfpsingle.c
Jay Foad 244b478386 ARM: 8026/1: Fix emulation of multiply accumulate instructions 
The emulation for single and double precision multiply accumulate
instructions correctly normalised any denormal values in the operand
registers, but failed to normalise the destination (accumulator)
register.

This fixes https://bugzilla.kernel.org/show_bug.cgi?id=70501

Signed-off-by: Jay Foad <jay.foad@gmail.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
2014-04-14 23:28:24 +01:00

1247 lines
29 KiB
C
Raw Blame History

/*
* linux/arch/arm/vfp/vfpsingle.c
*
* This code is derived in part from John R. Housers softfloat library, which
* carries the following notice:
*
* ===========================================================================
* This C source file is part of the SoftFloat IEC/IEEE Floating-point
* Arithmetic Package, Release 2.
*
* Written by John R. Hauser. This work was made possible in part by the
* International Computer Science Institute, located at Suite 600, 1947 Center
* Street, Berkeley, California 94704. Funding was partially provided by the
* National Science Foundation under grant MIP-9311980. The original version
* of this code was written as part of a project to build a fixed-point vector
* processor in collaboration with the University of California at Berkeley,
* overseen by Profs. Nelson Morgan and John Wawrzynek. More information
* is available through the web page `http://HTTP.CS.Berkeley.EDU/~jhauser/
* arithmetic/softfloat.html'.
*
* THIS SOFTWARE IS DISTRIBUTED AS IS, FOR FREE. Although reasonable effort
* has been made to avoid it, THIS SOFTWARE MAY CONTAIN FAULTS THAT WILL AT
* TIMES RESULT IN INCORRECT BEHAVIOR. USE OF THIS SOFTWARE IS RESTRICTED TO
* PERSONS AND ORGANIZATIONS WHO CAN AND WILL TAKE FULL RESPONSIBILITY FOR ANY
* AND ALL LOSSES, COSTS, OR OTHER PROBLEMS ARISING FROM ITS USE.
*
* Derivative works are acceptable, even for commercial purposes, so long as
* (1) they include prominent notice that the work is derivative, and (2) they
* include prominent notice akin to these three paragraphs for those parts of
* this code that are retained.
* ===========================================================================
*/
#include <linux/kernel.h>
#include <linux/bitops.h>
#include <asm/div64.h>
#include <asm/vfp.h>
#include "vfpinstr.h"
#include "vfp.h"
static struct vfp_single vfp_single_default_qnan = {
.exponent = 255,
.sign = 0,
.significand = VFP_SINGLE_SIGNIFICAND_QNAN,
};
static void vfp_single_dump(const char *str, struct vfp_single *s)
{
pr_debug("VFP: %s: sign=%d exponent=%d significand=%08x\n",
str, s->sign != 0, s->exponent, s->significand);
}
static void vfp_single_normalise_denormal(struct vfp_single *vs)
{
int bits = 31 - fls(vs->significand);
vfp_single_dump("normalise_denormal: in", vs);
if (bits) {
vs->exponent -= bits - 1;
vs->significand <<= bits;
}
vfp_single_dump("normalise_denormal: out", vs);
}
#ifndef DEBUG
#define vfp_single_normaliseround(sd,vsd,fpscr,except,func) __vfp_single_normaliseround(sd,vsd,fpscr,except)
u32 __vfp_single_normaliseround(int sd, struct vfp_single *vs, u32 fpscr, u32 exceptions)
#else
u32 vfp_single_normaliseround(int sd, struct vfp_single *vs, u32 fpscr, u32 exceptions, const char *func)
#endif
{
u32 significand, incr, rmode;
int exponent, shift, underflow;
vfp_single_dump("pack: in", vs);
/*
* Infinities and NaNs are a special case.
*/
if (vs->exponent == 255 && (vs->significand == 0 || exceptions))
goto pack;
/*
* Special-case zero.
*/
if (vs->significand == 0) {
vs->exponent = 0;
goto pack;
}
exponent = vs->exponent;
significand = vs->significand;
/*
* Normalise first. Note that we shift the significand up to
* bit 31, so we have VFP_SINGLE_LOW_BITS + 1 below the least
* significant bit.
*/
shift = 32 - fls(significand);
if (shift < 32 && shift) {
exponent -= shift;
significand <<= shift;
}
#ifdef DEBUG
vs->exponent = exponent;
vs->significand = significand;
vfp_single_dump("pack: normalised", vs);
#endif
/*
* Tiny number?
*/
underflow = exponent < 0;
if (underflow) {
significand = vfp_shiftright32jamming(significand, -exponent);
exponent = 0;
#ifdef DEBUG
vs->exponent = exponent;
vs->significand = significand;
vfp_single_dump("pack: tiny number", vs);
#endif
if (!(significand & ((1 << (VFP_SINGLE_LOW_BITS + 1)) - 1)))
underflow = 0;
}
/*
* Select rounding increment.
*/
incr = 0;
rmode = fpscr & FPSCR_RMODE_MASK;
if (rmode == FPSCR_ROUND_NEAREST) {
incr = 1 << VFP_SINGLE_LOW_BITS;
if ((significand & (1 << (VFP_SINGLE_LOW_BITS + 1))) == 0)
incr -= 1;
} else if (rmode == FPSCR_ROUND_TOZERO) {
incr = 0;
} else if ((rmode == FPSCR_ROUND_PLUSINF) ^ (vs->sign != 0))
incr = (1 << (VFP_SINGLE_LOW_BITS + 1)) - 1;
pr_debug("VFP: rounding increment = 0x%08x\n", incr);
/*
* Is our rounding going to overflow?
*/
if ((significand + incr) < significand) {
exponent += 1;
significand = (significand >> 1) | (significand & 1);
incr >>= 1;
#ifdef DEBUG
vs->exponent = exponent;
vs->significand = significand;
vfp_single_dump("pack: overflow", vs);
#endif
}
/*
* If any of the low bits (which will be shifted out of the
* number) are non-zero, the result is inexact.
*/
if (significand & ((1 << (VFP_SINGLE_LOW_BITS + 1)) - 1))
exceptions |= FPSCR_IXC;
/*
* Do our rounding.
*/
significand += incr;
/*
* Infinity?
*/
if (exponent >= 254) {
exceptions |= FPSCR_OFC | FPSCR_IXC;
if (incr == 0) {
vs->exponent = 253;
vs->significand = 0x7fffffff;
} else {
vs->exponent = 255; /* infinity */
vs->significand = 0;
}
} else {
if (significand >> (VFP_SINGLE_LOW_BITS + 1) == 0)
exponent = 0;
if (exponent || significand > 0x80000000)
underflow = 0;
if (underflow)
exceptions |= FPSCR_UFC;
vs->exponent = exponent;
vs->significand = significand >> 1;
}
pack:
vfp_single_dump("pack: final", vs);
{
s32 d = vfp_single_pack(vs);
#ifdef DEBUG
pr_debug("VFP: %s: d(s%d)=%08x exceptions=%08x\n", func,
sd, d, exceptions);
#endif
vfp_put_float(d, sd);
}
return exceptions;
}
/*
* Propagate the NaN, setting exceptions if it is signalling.
* 'n' is always a NaN. 'm' may be a number, NaN or infinity.
*/
static u32
vfp_propagate_nan(struct vfp_single *vsd, struct vfp_single *vsn,
struct vfp_single *vsm, u32 fpscr)
{
struct vfp_single *nan;
int tn, tm = 0;
tn = vfp_single_type(vsn);
if (vsm)
tm = vfp_single_type(vsm);
if (fpscr & FPSCR_DEFAULT_NAN)
/*
* Default NaN mode - always returns a quiet NaN
*/
nan = &vfp_single_default_qnan;
else {
/*
* Contemporary mode - select the first signalling
* NAN, or if neither are signalling, the first
* quiet NAN.
*/
if (tn == VFP_SNAN || (tm != VFP_SNAN && tn == VFP_QNAN))
nan = vsn;
else
nan = vsm;
/*
* Make the NaN quiet.
*/
nan->significand |= VFP_SINGLE_SIGNIFICAND_QNAN;
}
*vsd = *nan;
/*
* If one was a signalling NAN, raise invalid operation.
*/
return tn == VFP_SNAN || tm == VFP_SNAN ? FPSCR_IOC : VFP_NAN_FLAG;
}
/*
* Extended operations
*/
static u32 vfp_single_fabs(int sd, int unused, s32 m, u32 fpscr)
{
vfp_put_float(vfp_single_packed_abs(m), sd);
return 0;
}
static u32 vfp_single_fcpy(int sd, int unused, s32 m, u32 fpscr)
{
vfp_put_float(m, sd);
return 0;
}
static u32 vfp_single_fneg(int sd, int unused, s32 m, u32 fpscr)
{
vfp_put_float(vfp_single_packed_negate(m), sd);
return 0;
}
static const u16 sqrt_oddadjust[] = {
0x0004, 0x0022, 0x005d, 0x00b1, 0x011d, 0x019f, 0x0236, 0x02e0,
0x039c, 0x0468, 0x0545, 0x0631, 0x072b, 0x0832, 0x0946, 0x0a67
};
static const u16 sqrt_evenadjust[] = {
0x0a2d, 0x08af, 0x075a, 0x0629, 0x051a, 0x0429, 0x0356, 0x029e,
0x0200, 0x0179, 0x0109, 0x00af, 0x0068, 0x0034, 0x0012, 0x0002
};
u32 vfp_estimate_sqrt_significand(u32 exponent, u32 significand)
{
int index;
u32 z, a;
if ((significand & 0xc0000000) != 0x40000000) {
printk(KERN_WARNING "VFP: estimate_sqrt: invalid significand\n");
}
a = significand << 1;
index = (a >> 27) & 15;
if (exponent & 1) {
z = 0x4000 + (a >> 17) - sqrt_oddadjust[index];
z = ((a / z) << 14) + (z << 15);
a >>= 1;
} else {
z = 0x8000 + (a >> 17) - sqrt_evenadjust[index];
z = a / z + z;
z = (z >= 0x20000) ? 0xffff8000 : (z << 15);
if (z <= a)
return (s32)a >> 1;
}
{
u64 v = (u64)a << 31;
do_div(v, z);
return v + (z >> 1);
}
}
static u32 vfp_single_fsqrt(int sd, int unused, s32 m, u32 fpscr)
{
struct vfp_single vsm, vsd;
int ret, tm;
vfp_single_unpack(&vsm, m);
tm = vfp_single_type(&vsm);
if (tm & (VFP_NAN|VFP_INFINITY)) {
struct vfp_single *vsp = &vsd;
if (tm & VFP_NAN)
ret = vfp_propagate_nan(vsp, &vsm, NULL, fpscr);
else if (vsm.sign == 0) {
sqrt_copy:
vsp = &vsm;
ret = 0;
} else {
sqrt_invalid:
vsp = &vfp_single_default_qnan;
ret = FPSCR_IOC;
}
vfp_put_float(vfp_single_pack(vsp), sd);
return ret;
}
/*
* sqrt(+/- 0) == +/- 0
*/
if (tm & VFP_ZERO)
goto sqrt_copy;
/*
* Normalise a denormalised number
*/
if (tm & VFP_DENORMAL)
vfp_single_normalise_denormal(&vsm);
/*
* sqrt(<0) = invalid
*/
if (vsm.sign)
goto sqrt_invalid;
vfp_single_dump("sqrt", &vsm);
/*
* Estimate the square root.
*/
vsd.sign = 0;
vsd.exponent = ((vsm.exponent - 127) >> 1) + 127;
vsd.significand = vfp_estimate_sqrt_significand(vsm.exponent, vsm.significand) + 2;
vfp_single_dump("sqrt estimate", &vsd);
/*
* And now adjust.
*/
if ((vsd.significand & VFP_SINGLE_LOW_BITS_MASK) <= 5) {
if (vsd.significand < 2) {
vsd.significand = 0xffffffff;
} else {
u64 term;
s64 rem;
vsm.significand <<= !(vsm.exponent & 1);
term = (u64)vsd.significand * vsd.significand;
rem = ((u64)vsm.significand << 32) - term;
pr_debug("VFP: term=%016llx rem=%016llx\n", term, rem);
while (rem < 0) {
vsd.significand -= 1;
rem += ((u64)vsd.significand << 1) | 1;
}
vsd.significand |= rem != 0;
}
}
vsd.significand = vfp_shiftright32jamming(vsd.significand, 1);
return vfp_single_normaliseround(sd, &vsd, fpscr, 0, "fsqrt");
}
/*
* Equal := ZC
* Less than := N
* Greater than := C
* Unordered := CV
*/
static u32 vfp_compare(int sd, int signal_on_qnan, s32 m, u32 fpscr)
{
s32 d;
u32 ret = 0;
d = vfp_get_float(sd);
if (vfp_single_packed_exponent(m) == 255 && vfp_single_packed_mantissa(m)) {
ret |= FPSCR_C | FPSCR_V;
if (signal_on_qnan || !(vfp_single_packed_mantissa(m) & (1 << (VFP_SINGLE_MANTISSA_BITS - 1))))
/*
* Signalling NaN, or signalling on quiet NaN
*/
ret |= FPSCR_IOC;
}
if (vfp_single_packed_exponent(d) == 255 && vfp_single_packed_mantissa(d)) {
ret |= FPSCR_C | FPSCR_V;
if (signal_on_qnan || !(vfp_single_packed_mantissa(d) & (1 << (VFP_SINGLE_MANTISSA_BITS - 1))))
/*
* Signalling NaN, or signalling on quiet NaN
*/
ret |= FPSCR_IOC;
}
if (ret == 0) {
if (d == m || vfp_single_packed_abs(d | m) == 0) {
/*
* equal
*/
ret |= FPSCR_Z | FPSCR_C;
} else if (vfp_single_packed_sign(d ^ m)) {
/*
* different signs
*/
if (vfp_single_packed_sign(d))
/*
* d is negative, so d < m
*/
ret |= FPSCR_N;
else
/*
* d is positive, so d > m
*/
ret |= FPSCR_C;
} else if ((vfp_single_packed_sign(d) != 0) ^ (d < m)) {
/*
* d < m
*/
ret |= FPSCR_N;
} else if ((vfp_single_packed_sign(d) != 0) ^ (d > m)) {
/*
* d > m
*/
ret |= FPSCR_C;
}
}
return ret;
}
static u32 vfp_single_fcmp(int sd, int unused, s32 m, u32 fpscr)
{
return vfp_compare(sd, 0, m, fpscr);
}
static u32 vfp_single_fcmpe(int sd, int unused, s32 m, u32 fpscr)
{
return vfp_compare(sd, 1, m, fpscr);
}
static u32 vfp_single_fcmpz(int sd, int unused, s32 m, u32 fpscr)
{
return vfp_compare(sd, 0, 0, fpscr);
}
static u32 vfp_single_fcmpez(int sd, int unused, s32 m, u32 fpscr)
{
return vfp_compare(sd, 1, 0, fpscr);
}
static u32 vfp_single_fcvtd(int dd, int unused, s32 m, u32 fpscr)
{
struct vfp_single vsm;
struct vfp_double vdd;
int tm;
u32 exceptions = 0;
vfp_single_unpack(&vsm, m);
tm = vfp_single_type(&vsm);
/*
* If we have a signalling NaN, signal invalid operation.
*/
if (tm == VFP_SNAN)
exceptions = FPSCR_IOC;
if (tm & VFP_DENORMAL)
vfp_single_normalise_denormal(&vsm);
vdd.sign = vsm.sign;
vdd.significand = (u64)vsm.significand << 32;
/*
* If we have an infinity or NaN, the exponent must be 2047.
*/
if (tm & (VFP_INFINITY|VFP_NAN)) {
vdd.exponent = 2047;
if (tm == VFP_QNAN)
vdd.significand |= VFP_DOUBLE_SIGNIFICAND_QNAN;
goto pack_nan;
} else if (tm & VFP_ZERO)
vdd.exponent = 0;
else
vdd.exponent = vsm.exponent + (1023 - 127);
return vfp_double_normaliseround(dd, &vdd, fpscr, exceptions, "fcvtd");
pack_nan:
vfp_put_double(vfp_double_pack(&vdd), dd);
return exceptions;
}
static u32 vfp_single_fuito(int sd, int unused, s32 m, u32 fpscr)
{
struct vfp_single vs;
vs.sign = 0;
vs.exponent = 127 + 31 - 1;
vs.significand = (u32)m;
return vfp_single_normaliseround(sd, &vs, fpscr, 0, "fuito");
}
static u32 vfp_single_fsito(int sd, int unused, s32 m, u32 fpscr)
{
struct vfp_single vs;
vs.sign = (m & 0x80000000) >> 16;
vs.exponent = 127 + 31 - 1;
vs.significand = vs.sign ? -m : m;
return vfp_single_normaliseround(sd, &vs, fpscr, 0, "fsito");
}
static u32 vfp_single_ftoui(int sd, int unused, s32 m, u32 fpscr)
{
struct vfp_single vsm;
u32 d, exceptions = 0;
int rmode = fpscr & FPSCR_RMODE_MASK;
int tm;
vfp_single_unpack(&vsm, m);
vfp_single_dump("VSM", &vsm);
/*
* Do we have a denormalised number?
*/
tm = vfp_single_type(&vsm);
if (tm & VFP_DENORMAL)
exceptions |= FPSCR_IDC;
if (tm & VFP_NAN)
vsm.sign = 0;
if (vsm.exponent >= 127 + 32) {
d = vsm.sign ? 0 : 0xffffffff;
exceptions = FPSCR_IOC;
} else if (vsm.exponent >= 127 - 1) {
int shift = 127 + 31 - vsm.exponent;
u32 rem, incr = 0;
/*
* 2^0 <= m < 2^32-2^8
*/
d = (vsm.significand << 1) >> shift;
rem = vsm.significand << (33 - shift);
if (rmode == FPSCR_ROUND_NEAREST) {
incr = 0x80000000;
if ((d & 1) == 0)
incr -= 1;
} else if (rmode == FPSCR_ROUND_TOZERO) {
incr = 0;
} else if ((rmode == FPSCR_ROUND_PLUSINF) ^ (vsm.sign != 0)) {
incr = ~0;
}
if ((rem + incr) < rem) {
if (d < 0xffffffff)
d += 1;
else
exceptions |= FPSCR_IOC;
}
if (d && vsm.sign) {
d = 0;
exceptions |= FPSCR_IOC;
} else if (rem)
exceptions |= FPSCR_IXC;
} else {
d = 0;
if (vsm.exponent | vsm.significand) {
exceptions |= FPSCR_IXC;
if (rmode == FPSCR_ROUND_PLUSINF && vsm.sign == 0)
d = 1;
else if (rmode == FPSCR_ROUND_MINUSINF && vsm.sign) {
d = 0;
exceptions |= FPSCR_IOC;
}
}
}
pr_debug("VFP: ftoui: d(s%d)=%08x exceptions=%08x\n", sd, d, exceptions);
vfp_put_float(d, sd);
return exceptions;
}
static u32 vfp_single_ftouiz(int sd, int unused, s32 m, u32 fpscr)
{
return vfp_single_ftoui(sd, unused, m, FPSCR_ROUND_TOZERO);
}
static u32 vfp_single_ftosi(int sd, int unused, s32 m, u32 fpscr)
{
struct vfp_single vsm;
u32 d, exceptions = 0;
int rmode = fpscr & FPSCR_RMODE_MASK;
int tm;
vfp_single_unpack(&vsm, m);
vfp_single_dump("VSM", &vsm);
/*
* Do we have a denormalised number?
*/
tm = vfp_single_type(&vsm);
if (vfp_single_type(&vsm) & VFP_DENORMAL)
exceptions |= FPSCR_IDC;
if (tm & VFP_NAN) {
d = 0;
exceptions |= FPSCR_IOC;
} else if (vsm.exponent >= 127 + 32) {
/*
* m >= 2^31-2^7: invalid
*/
d = 0x7fffffff;
if (vsm.sign)
d = ~d;
exceptions |= FPSCR_IOC;
} else if (vsm.exponent >= 127 - 1) {
int shift = 127 + 31 - vsm.exponent;
u32 rem, incr = 0;
/* 2^0 <= m <= 2^31-2^7 */
d = (vsm.significand << 1) >> shift;
rem = vsm.significand << (33 - shift);
if (rmode == FPSCR_ROUND_NEAREST) {
incr = 0x80000000;
if ((d & 1) == 0)
incr -= 1;
} else if (rmode == FPSCR_ROUND_TOZERO) {
incr = 0;
} else if ((rmode == FPSCR_ROUND_PLUSINF) ^ (vsm.sign != 0)) {
incr = ~0;
}
if ((rem + incr) < rem && d < 0xffffffff)
d += 1;
if (d > 0x7fffffff + (vsm.sign != 0)) {
d = 0x7fffffff + (vsm.sign != 0);
exceptions |= FPSCR_IOC;
} else if (rem)
exceptions |= FPSCR_IXC;
if (vsm.sign)
d = -d;
} else {
d = 0;
if (vsm.exponent | vsm.significand) {
exceptions |= FPSCR_IXC;
if (rmode == FPSCR_ROUND_PLUSINF && vsm.sign == 0)
d = 1;
else if (rmode == FPSCR_ROUND_MINUSINF && vsm.sign)
d = -1;
}
}
pr_debug("VFP: ftosi: d(s%d)=%08x exceptions=%08x\n", sd, d, exceptions);
vfp_put_float((s32)d, sd);
return exceptions;
}
static u32 vfp_single_ftosiz(int sd, int unused, s32 m, u32 fpscr)
{
return vfp_single_ftosi(sd, unused, m, FPSCR_ROUND_TOZERO);
}
static struct op fops_ext[32] = {
[FEXT_TO_IDX(FEXT_FCPY)] = { vfp_single_fcpy, 0 },
[FEXT_TO_IDX(FEXT_FABS)] = { vfp_single_fabs, 0 },
[FEXT_TO_IDX(FEXT_FNEG)] = { vfp_single_fneg, 0 },
[FEXT_TO_IDX(FEXT_FSQRT)] = { vfp_single_fsqrt, 0 },
[FEXT_TO_IDX(FEXT_FCMP)] = { vfp_single_fcmp, OP_SCALAR },
[FEXT_TO_IDX(FEXT_FCMPE)] = { vfp_single_fcmpe, OP_SCALAR },
[FEXT_TO_IDX(FEXT_FCMPZ)] = { vfp_single_fcmpz, OP_SCALAR },
[FEXT_TO_IDX(FEXT_FCMPEZ)] = { vfp_single_fcmpez, OP_SCALAR },
[FEXT_TO_IDX(FEXT_FCVT)] = { vfp_single_fcvtd, OP_SCALAR|OP_DD },
[FEXT_TO_IDX(FEXT_FUITO)] = { vfp_single_fuito, OP_SCALAR },
[FEXT_TO_IDX(FEXT_FSITO)] = { vfp_single_fsito, OP_SCALAR },
[FEXT_TO_IDX(FEXT_FTOUI)] = { vfp_single_ftoui, OP_SCALAR },
[FEXT_TO_IDX(FEXT_FTOUIZ)] = { vfp_single_ftouiz, OP_SCALAR },
[FEXT_TO_IDX(FEXT_FTOSI)] = { vfp_single_ftosi, OP_SCALAR },
[FEXT_TO_IDX(FEXT_FTOSIZ)] = { vfp_single_ftosiz, OP_SCALAR },
};
static u32
vfp_single_fadd_nonnumber(struct vfp_single *vsd, struct vfp_single *vsn,
struct vfp_single *vsm, u32 fpscr)
{
struct vfp_single *vsp;
u32 exceptions = 0;
int tn, tm;
tn = vfp_single_type(vsn);
tm = vfp_single_type(vsm);
if (tn & tm & VFP_INFINITY) {
/*
* Two infinities. Are they different signs?
*/
if (vsn->sign ^ vsm->sign) {
/*
* different signs -> invalid
*/
exceptions = FPSCR_IOC;
vsp = &vfp_single_default_qnan;
} else {
/*
* same signs -> valid
*/
vsp = vsn;
}
} else if (tn & VFP_INFINITY && tm & VFP_NUMBER) {
/*
* One infinity and one number -> infinity
*/
vsp = vsn;
} else {
/*
* 'n' is a NaN of some type
*/
return vfp_propagate_nan(vsd, vsn, vsm, fpscr);
}
*vsd = *vsp;
return exceptions;
}
static u32
vfp_single_add(struct vfp_single *vsd, struct vfp_single *vsn,
struct vfp_single *vsm, u32 fpscr)
{
u32 exp_diff, m_sig;
if (vsn->significand & 0x80000000 ||
vsm->significand & 0x80000000) {
pr_info("VFP: bad FP values in %s\n", __func__);
vfp_single_dump("VSN", vsn);
vfp_single_dump("VSM", vsm);
}
/*
* Ensure that 'n' is the largest magnitude number. Note that
* if 'n' and 'm' have equal exponents, we do not swap them.
* This ensures that NaN propagation works correctly.
*/
if (vsn->exponent < vsm->exponent) {
struct vfp_single *t = vsn;
vsn = vsm;
vsm = t;
}
/*
* Is 'n' an infinity or a NaN? Note that 'm' may be a number,
* infinity or a NaN here.
*/
if (vsn->exponent == 255)
return vfp_single_fadd_nonnumber(vsd, vsn, vsm, fpscr);
/*
* We have two proper numbers, where 'vsn' is the larger magnitude.
*
* Copy 'n' to 'd' before doing the arithmetic.
*/
*vsd = *vsn;
/*
* Align both numbers.
*/
exp_diff = vsn->exponent - vsm->exponent;
m_sig = vfp_shiftright32jamming(vsm->significand, exp_diff);
/*
* If the signs are different, we are really subtracting.
*/
if (vsn->sign ^ vsm->sign) {
m_sig = vsn->significand - m_sig;
if ((s32)m_sig < 0) {
vsd->sign = vfp_sign_negate(vsd->sign);
m_sig = -m_sig;
} else if (m_sig == 0) {
vsd->sign = (fpscr & FPSCR_RMODE_MASK) ==
FPSCR_ROUND_MINUSINF ? 0x8000 : 0;
}
} else {
m_sig = vsn->significand + m_sig;
}
vsd->significand = m_sig;
return 0;
}
static u32
vfp_single_multiply(struct vfp_single *vsd, struct vfp_single *vsn, struct vfp_single *vsm, u32 fpscr)
{
vfp_single_dump("VSN", vsn);
vfp_single_dump("VSM", vsm);
/*
* Ensure that 'n' is the largest magnitude number. Note that
* if 'n' and 'm' have equal exponents, we do not swap them.
* This ensures that NaN propagation works correctly.
*/
if (vsn->exponent < vsm->exponent) {
struct vfp_single *t = vsn;
vsn = vsm;
vsm = t;
pr_debug("VFP: swapping M <-> N\n");
}
vsd->sign = vsn->sign ^ vsm->sign;
/*
* If 'n' is an infinity or NaN, handle it. 'm' may be anything.
*/
if (vsn->exponent == 255) {
if (vsn->significand || (vsm->exponent == 255 && vsm->significand))
return vfp_propagate_nan(vsd, vsn, vsm, fpscr);
if ((vsm->exponent | vsm->significand) == 0) {
*vsd = vfp_single_default_qnan;
return FPSCR_IOC;
}
vsd->exponent = vsn->exponent;
vsd->significand = 0;
return 0;
}
/*
* If 'm' is zero, the result is always zero. In this case,
* 'n' may be zero or a number, but it doesn't matter which.
*/
if ((vsm->exponent | vsm->significand) == 0) {
vsd->exponent = 0;
vsd->significand = 0;
return 0;
}
/*
* We add 2 to the destination exponent for the same reason as
* the addition case - though this time we have +1 from each
* input operand.
*/
vsd->exponent = vsn->exponent + vsm->exponent - 127 + 2;
vsd->significand = vfp_hi64to32jamming((u64)vsn->significand * vsm->significand);
vfp_single_dump("VSD", vsd);
return 0;
}
#define NEG_MULTIPLY (1 << 0)
#define NEG_SUBTRACT (1 << 1)
static u32
vfp_single_multiply_accumulate(int sd, int sn, s32 m, u32 fpscr, u32 negate, char *func)
{
struct vfp_single vsd, vsp, vsn, vsm;
u32 exceptions;
s32 v;
v = vfp_get_float(sn);
pr_debug("VFP: s%u = %08x\n", sn, v);
vfp_single_unpack(&vsn, v);
if (vsn.exponent == 0 && vsn.significand)
vfp_single_normalise_denormal(&vsn);
vfp_single_unpack(&vsm, m);
if (vsm.exponent == 0 && vsm.significand)
vfp_single_normalise_denormal(&vsm);
exceptions = vfp_single_multiply(&vsp, &vsn, &vsm, fpscr);
if (negate & NEG_MULTIPLY)
vsp.sign = vfp_sign_negate(vsp.sign);
v = vfp_get_float(sd);
pr_debug("VFP: s%u = %08x\n", sd, v);
vfp_single_unpack(&vsn, v);
if (vsn.exponent == 0 && vsn.significand)
vfp_single_normalise_denormal(&vsn);
if (negate & NEG_SUBTRACT)
vsn.sign = vfp_sign_negate(vsn.sign);
exceptions |= vfp_single_add(&vsd, &vsn, &vsp, fpscr);
return vfp_single_normaliseround(sd, &vsd, fpscr, exceptions, func);
}
/*
* Standard operations
*/
/*
* sd = sd + (sn * sm)
*/
static u32 vfp_single_fmac(int sd, int sn, s32 m, u32 fpscr)
{
return vfp_single_multiply_accumulate(sd, sn, m, fpscr, 0, "fmac");
}
/*
* sd = sd - (sn * sm)
*/
static u32 vfp_single_fnmac(int sd, int sn, s32 m, u32 fpscr)
{
return vfp_single_multiply_accumulate(sd, sn, m, fpscr, NEG_MULTIPLY, "fnmac");
}
/*
* sd = -sd + (sn * sm)
*/
static u32 vfp_single_fmsc(int sd, int sn, s32 m, u32 fpscr)
{
return vfp_single_multiply_accumulate(sd, sn, m, fpscr, NEG_SUBTRACT, "fmsc");
}
/*
* sd = -sd - (sn * sm)
*/
static u32 vfp_single_fnmsc(int sd, int sn, s32 m, u32 fpscr)
{
return vfp_single_multiply_accumulate(sd, sn, m, fpscr, NEG_SUBTRACT | NEG_MULTIPLY, "fnmsc");
}
/*
* sd = sn * sm
*/
static u32 vfp_single_fmul(int sd, int sn, s32 m, u32 fpscr)
{
struct vfp_single vsd, vsn, vsm;
u32 exceptions;
s32 n = vfp_get_float(sn);
pr_debug("VFP: s%u = %08x\n", sn, n);
vfp_single_unpack(&vsn, n);
if (vsn.exponent == 0 && vsn.significand)
vfp_single_normalise_denormal(&vsn);
vfp_single_unpack(&vsm, m);
if (vsm.exponent == 0 && vsm.significand)
vfp_single_normalise_denormal(&vsm);
exceptions = vfp_single_multiply(&vsd, &vsn, &vsm, fpscr);
return vfp_single_normaliseround(sd, &vsd, fpscr, exceptions, "fmul");
}
/*
* sd = -(sn * sm)
*/
static u32 vfp_single_fnmul(int sd, int sn, s32 m, u32 fpscr)
{
struct vfp_single vsd, vsn, vsm;
u32 exceptions;
s32 n = vfp_get_float(sn);
pr_debug("VFP: s%u = %08x\n", sn, n);
vfp_single_unpack(&vsn, n);
if (vsn.exponent == 0 && vsn.significand)
vfp_single_normalise_denormal(&vsn);
vfp_single_unpack(&vsm, m);
if (vsm.exponent == 0 && vsm.significand)
vfp_single_normalise_denormal(&vsm);
exceptions = vfp_single_multiply(&vsd, &vsn, &vsm, fpscr);
vsd.sign = vfp_sign_negate(vsd.sign);
return vfp_single_normaliseround(sd, &vsd, fpscr, exceptions, "fnmul");
}
/*
* sd = sn + sm
*/
static u32 vfp_single_fadd(int sd, int sn, s32 m, u32 fpscr)
{
struct vfp_single vsd, vsn, vsm;
u32 exceptions;
s32 n = vfp_get_float(sn);
pr_debug("VFP: s%u = %08x\n", sn, n);
/*
* Unpack and normalise denormals.
*/
vfp_single_unpack(&vsn, n);
if (vsn.exponent == 0 && vsn.significand)
vfp_single_normalise_denormal(&vsn);
vfp_single_unpack(&vsm, m);
if (vsm.exponent == 0 && vsm.significand)
vfp_single_normalise_denormal(&vsm);
exceptions = vfp_single_add(&vsd, &vsn, &vsm, fpscr);
return vfp_single_normaliseround(sd, &vsd, fpscr, exceptions, "fadd");
}
/*
* sd = sn - sm
*/
static u32 vfp_single_fsub(int sd, int sn, s32 m, u32 fpscr)
{
/*
* Subtraction is addition with one sign inverted.
*/
return vfp_single_fadd(sd, sn, vfp_single_packed_negate(m), fpscr);
}
/*
* sd = sn / sm
*/
static u32 vfp_single_fdiv(int sd, int sn, s32 m, u32 fpscr)
{
struct vfp_single vsd, vsn, vsm;
u32 exceptions = 0;
s32 n = vfp_get_float(sn);
int tm, tn;
pr_debug("VFP: s%u = %08x\n", sn, n);
vfp_single_unpack(&vsn, n);
vfp_single_unpack(&vsm, m);
vsd.sign = vsn.sign ^ vsm.sign;
tn = vfp_single_type(&vsn);
tm = vfp_single_type(&vsm);
/*
* Is n a NAN?
*/
if (tn & VFP_NAN)
goto vsn_nan;
/*
* Is m a NAN?
*/
if (tm & VFP_NAN)
goto vsm_nan;
/*
* If n and m are infinity, the result is invalid
* If n and m are zero, the result is invalid
*/
if (tm & tn & (VFP_INFINITY|VFP_ZERO))
goto invalid;
/*
* If n is infinity, the result is infinity
*/
if (tn & VFP_INFINITY)
goto infinity;
/*
* If m is zero, raise div0 exception
*/
if (tm & VFP_ZERO)
goto divzero;
/*
* If m is infinity, or n is zero, the result is zero
*/
if (tm & VFP_INFINITY || tn & VFP_ZERO)
goto zero;
if (tn & VFP_DENORMAL)
vfp_single_normalise_denormal(&vsn);
if (tm & VFP_DENORMAL)
vfp_single_normalise_denormal(&vsm);
/*
* Ok, we have two numbers, we can perform division.
*/
vsd.exponent = vsn.exponent - vsm.exponent + 127 - 1;
vsm.significand <<= 1;
if (vsm.significand <= (2 * vsn.significand)) {
vsn.significand >>= 1;
vsd.exponent++;
}
{
u64 significand = (u64)vsn.significand << 32;
do_div(significand, vsm.significand);
vsd.significand = significand;
}
if ((vsd.significand & 0x3f) == 0)
vsd.significand |= ((u64)vsm.significand * vsd.significand != (u64)vsn.significand << 32);
return vfp_single_normaliseround(sd, &vsd, fpscr, 0, "fdiv");
vsn_nan:
exceptions = vfp_propagate_nan(&vsd, &vsn, &vsm, fpscr);
pack:
vfp_put_float(vfp_single_pack(&vsd), sd);
return exceptions;
vsm_nan:
exceptions = vfp_propagate_nan(&vsd, &vsm, &vsn, fpscr);
goto pack;
zero:
vsd.exponent = 0;
vsd.significand = 0;
goto pack;
divzero:
exceptions = FPSCR_DZC;
infinity:
vsd.exponent = 255;
vsd.significand = 0;
goto pack;
invalid:
vfp_put_float(vfp_single_pack(&vfp_single_default_qnan), sd);
return FPSCR_IOC;
}
static struct op fops[16] = {
[FOP_TO_IDX(FOP_FMAC)] = { vfp_single_fmac, 0 },
[FOP_TO_IDX(FOP_FNMAC)] = { vfp_single_fnmac, 0 },
[FOP_TO_IDX(FOP_FMSC)] = { vfp_single_fmsc, 0 },
[FOP_TO_IDX(FOP_FNMSC)] = { vfp_single_fnmsc, 0 },
[FOP_TO_IDX(FOP_FMUL)] = { vfp_single_fmul, 0 },
[FOP_TO_IDX(FOP_FNMUL)] = { vfp_single_fnmul, 0 },
[FOP_TO_IDX(FOP_FADD)] = { vfp_single_fadd, 0 },
[FOP_TO_IDX(FOP_FSUB)] = { vfp_single_fsub, 0 },
[FOP_TO_IDX(FOP_FDIV)] = { vfp_single_fdiv, 0 },
};
#define FREG_BANK(x) ((x) & 0x18)
#define FREG_IDX(x) ((x) & 7)
u32 vfp_single_cpdo(u32 inst, u32 fpscr)
{
u32 op = inst & FOP_MASK;
u32 exceptions = 0;
unsigned int dest;
unsigned int sn = vfp_get_sn(inst);
unsigned int sm = vfp_get_sm(inst);
unsigned int vecitr, veclen, vecstride;
struct op *fop;
vecstride = 1 + ((fpscr & FPSCR_STRIDE_MASK) == FPSCR_STRIDE_MASK);
fop = (op == FOP_EXT) ? &fops_ext[FEXT_TO_IDX(inst)] : &fops[FOP_TO_IDX(op)];
/*
* fcvtsd takes a dN register number as destination, not sN.
* Technically, if bit 0 of dd is set, this is an invalid
* instruction. However, we ignore this for efficiency.
* It also only operates on scalars.
*/
if (fop->flags & OP_DD)
dest = vfp_get_dd(inst);
else
dest = vfp_get_sd(inst);
/*
* If destination bank is zero, vector length is always '1'.
* ARM DDI0100F C5.1.3, C5.3.2.
*/
if ((fop->flags & OP_SCALAR) || FREG_BANK(dest) == 0)
veclen = 0;
else
veclen = fpscr & FPSCR_LENGTH_MASK;
pr_debug("VFP: vecstride=%u veclen=%u\n", vecstride,
(veclen >> FPSCR_LENGTH_BIT) + 1);
if (!fop->fn)
goto invalid;
for (vecitr = 0; vecitr <= veclen; vecitr += 1 << FPSCR_LENGTH_BIT) {
s32 m = vfp_get_float(sm);
u32 except;
char type;
type = fop->flags & OP_DD ? 'd' : 's';
if (op == FOP_EXT)
pr_debug("VFP: itr%d (%c%u) = op[%u] (s%u=%08x)\n",
vecitr >> FPSCR_LENGTH_BIT, type, dest, sn,
sm, m);
else
pr_debug("VFP: itr%d (%c%u) = (s%u) op[%u] (s%u=%08x)\n",
vecitr >> FPSCR_LENGTH_BIT, type, dest, sn,
FOP_TO_IDX(op), sm, m);
except = fop->fn(dest, sn, m, fpscr);
pr_debug("VFP: itr%d: exceptions=%08x\n",
vecitr >> FPSCR_LENGTH_BIT, except);
exceptions |= except;
/*
* CHECK: It appears to be undefined whether we stop when
* we encounter an exception. We continue.
*/
dest = FREG_BANK(dest) + ((FREG_IDX(dest) + vecstride) & 7);
sn = FREG_BANK(sn) + ((FREG_IDX(sn) + vecstride) & 7);
if (FREG_BANK(sm) != 0)
sm = FREG_BANK(sm) + ((FREG_IDX(sm) + vecstride) & 7);
}
return exceptions;
invalid:
return (u32)-1;
}
Reference in New Issue View Git Blame Copy Permalink