aboutsummaryrefslogtreecommitdiffstatshomepage
path: root/arch/x86/math-emu/poly_atan.c
diff options
context:
space:
mode:
authorIngo Molnar <mingo@elte.hu>2008-01-30 13:30:11 +0100
committerIngo Molnar <mingo@elte.hu>2008-01-30 13:30:11 +0100
commit3d0d14f983b55a570b976976284df4c434af3223 (patch)
tree864f11c0ce5ee1e15acdd196018b79d0d0e2685d /arch/x86/math-emu/poly_atan.c
parentx86: mach-voyager, lindent (diff)
downloadwireguard-linux-3d0d14f983b55a570b976976284df4c434af3223.tar.xz
wireguard-linux-3d0d14f983b55a570b976976284df4c434af3223.zip
x86: lindent arch/i386/math-emu
lindent these files: errors lines of code errors/KLOC arch/x86/math-emu/ 2236 9424 237.2 arch/x86/math-emu/ 128 8706 14.7 no other changes. No code changed: text data bss dec hex filename 5589802 612739 3833856 10036397 9924ad vmlinux.before 5589802 612739 3833856 10036397 9924ad vmlinux.after the intent of this patch is to ease the automated tracking of kernel code quality - it's just much easier for us to maintain it if every file in arch/x86 is supposed to be clean. NOTE: it is a known problem of lindent that it causes some style damage of its own, but it's a safe tool (well, except for the gcc array range initializers extension), so we did the bulk of the changes via lindent, and did the manual fixups in a followup patch. the resulting math-emu code has been tested by Thomas Gleixner on a real 386 DX CPU as well, and it works fine. Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Diffstat (limited to 'arch/x86/math-emu/poly_atan.c')
-rw-r--r--arch/x86/math-emu/poly_atan.c353
1 files changed, 166 insertions, 187 deletions
diff --git a/arch/x86/math-emu/poly_atan.c b/arch/x86/math-emu/poly_atan.c
index 82f702952f69..2f4ac8143fc3 100644
--- a/arch/x86/math-emu/poly_atan.c
+++ b/arch/x86/math-emu/poly_atan.c
@@ -18,28 +18,25 @@
#include "control_w.h"
#include "poly.h"
-
#define HIPOWERon 6 /* odd poly, negative terms */
-static const unsigned long long oddnegterms[HIPOWERon] =
-{
- 0x0000000000000000LL, /* Dummy (not for - 1.0) */
- 0x015328437f756467LL,
- 0x0005dda27b73dec6LL,
- 0x0000226bf2bfb91aLL,
- 0x000000ccc439c5f7LL,
- 0x0000000355438407LL
-} ;
+static const unsigned long long oddnegterms[HIPOWERon] = {
+ 0x0000000000000000LL, /* Dummy (not for - 1.0) */
+ 0x015328437f756467LL,
+ 0x0005dda27b73dec6LL,
+ 0x0000226bf2bfb91aLL,
+ 0x000000ccc439c5f7LL,
+ 0x0000000355438407LL
+};
#define HIPOWERop 6 /* odd poly, positive terms */
-static const unsigned long long oddplterms[HIPOWERop] =
-{
+static const unsigned long long oddplterms[HIPOWERop] = {
/* 0xaaaaaaaaaaaaaaabLL, transferred to fixedpterm[] */
- 0x0db55a71875c9ac2LL,
- 0x0029fce2d67880b0LL,
- 0x0000dfd3908b4596LL,
- 0x00000550fd61dab4LL,
- 0x0000001c9422b3f9LL,
- 0x000000003e3301e1LL
+ 0x0db55a71875c9ac2LL,
+ 0x0029fce2d67880b0LL,
+ 0x0000dfd3908b4596LL,
+ 0x00000550fd61dab4LL,
+ 0x0000001c9422b3f9LL,
+ 0x000000003e3301e1LL
};
static const unsigned long long denomterm = 0xebd9b842c5c53a0eLL;
@@ -48,182 +45,164 @@ static const Xsig fixedpterm = MK_XSIG(0xaaaaaaaa, 0xaaaaaaaa, 0xaaaaaaaa);
static const Xsig pi_signif = MK_XSIG(0xc90fdaa2, 0x2168c234, 0xc4c6628b);
-
/*--- poly_atan() -----------------------------------------------------------+
| |
+---------------------------------------------------------------------------*/
-void poly_atan(FPU_REG *st0_ptr, u_char st0_tag,
- FPU_REG *st1_ptr, u_char st1_tag)
+void poly_atan(FPU_REG * st0_ptr, u_char st0_tag,
+ FPU_REG * st1_ptr, u_char st1_tag)
{
- u_char transformed, inverted,
- sign1, sign2;
- int exponent;
- long int dummy_exp;
- Xsig accumulator, Numer, Denom, accumulatore, argSignif,
- argSq, argSqSq;
- u_char tag;
-
- sign1 = getsign(st0_ptr);
- sign2 = getsign(st1_ptr);
- if ( st0_tag == TAG_Valid )
- {
- exponent = exponent(st0_ptr);
- }
- else
- {
- /* This gives non-compatible stack contents... */
- FPU_to_exp16(st0_ptr, st0_ptr);
- exponent = exponent16(st0_ptr);
- }
- if ( st1_tag == TAG_Valid )
- {
- exponent -= exponent(st1_ptr);
- }
- else
- {
- /* This gives non-compatible stack contents... */
- FPU_to_exp16(st1_ptr, st1_ptr);
- exponent -= exponent16(st1_ptr);
- }
-
- if ( (exponent < 0) || ((exponent == 0) &&
- ((st0_ptr->sigh < st1_ptr->sigh) ||
- ((st0_ptr->sigh == st1_ptr->sigh) &&
- (st0_ptr->sigl < st1_ptr->sigl))) ) )
- {
- inverted = 1;
- Numer.lsw = Denom.lsw = 0;
- XSIG_LL(Numer) = significand(st0_ptr);
- XSIG_LL(Denom) = significand(st1_ptr);
- }
- else
- {
- inverted = 0;
- exponent = -exponent;
- Numer.lsw = Denom.lsw = 0;
- XSIG_LL(Numer) = significand(st1_ptr);
- XSIG_LL(Denom) = significand(st0_ptr);
- }
- div_Xsig(&Numer, &Denom, &argSignif);
- exponent += norm_Xsig(&argSignif);
-
- if ( (exponent >= -1)
- || ((exponent == -2) && (argSignif.msw > 0xd413ccd0)) )
- {
- /* The argument is greater than sqrt(2)-1 (=0.414213562...) */
- /* Convert the argument by an identity for atan */
- transformed = 1;
-
- if ( exponent >= 0 )
- {
+ u_char transformed, inverted, sign1, sign2;
+ int exponent;
+ long int dummy_exp;
+ Xsig accumulator, Numer, Denom, accumulatore, argSignif, argSq, argSqSq;
+ u_char tag;
+
+ sign1 = getsign(st0_ptr);
+ sign2 = getsign(st1_ptr);
+ if (st0_tag == TAG_Valid) {
+ exponent = exponent(st0_ptr);
+ } else {
+ /* This gives non-compatible stack contents... */
+ FPU_to_exp16(st0_ptr, st0_ptr);
+ exponent = exponent16(st0_ptr);
+ }
+ if (st1_tag == TAG_Valid) {
+ exponent -= exponent(st1_ptr);
+ } else {
+ /* This gives non-compatible stack contents... */
+ FPU_to_exp16(st1_ptr, st1_ptr);
+ exponent -= exponent16(st1_ptr);
+ }
+
+ if ((exponent < 0) || ((exponent == 0) &&
+ ((st0_ptr->sigh < st1_ptr->sigh) ||
+ ((st0_ptr->sigh == st1_ptr->sigh) &&
+ (st0_ptr->sigl < st1_ptr->sigl))))) {
+ inverted = 1;
+ Numer.lsw = Denom.lsw = 0;
+ XSIG_LL(Numer) = significand(st0_ptr);
+ XSIG_LL(Denom) = significand(st1_ptr);
+ } else {
+ inverted = 0;
+ exponent = -exponent;
+ Numer.lsw = Denom.lsw = 0;
+ XSIG_LL(Numer) = significand(st1_ptr);
+ XSIG_LL(Denom) = significand(st0_ptr);
+ }
+ div_Xsig(&Numer, &Denom, &argSignif);
+ exponent += norm_Xsig(&argSignif);
+
+ if ((exponent >= -1)
+ || ((exponent == -2) && (argSignif.msw > 0xd413ccd0))) {
+ /* The argument is greater than sqrt(2)-1 (=0.414213562...) */
+ /* Convert the argument by an identity for atan */
+ transformed = 1;
+
+ if (exponent >= 0) {
#ifdef PARANOID
- if ( !( (exponent == 0) &&
- (argSignif.lsw == 0) && (argSignif.midw == 0) &&
- (argSignif.msw == 0x80000000) ) )
- {
- EXCEPTION(EX_INTERNAL|0x104); /* There must be a logic error */
- return;
- }
+ if (!((exponent == 0) &&
+ (argSignif.lsw == 0) && (argSignif.midw == 0) &&
+ (argSignif.msw == 0x80000000))) {
+ EXCEPTION(EX_INTERNAL | 0x104); /* There must be a logic error */
+ return;
+ }
#endif /* PARANOID */
- argSignif.msw = 0; /* Make the transformed arg -> 0.0 */
+ argSignif.msw = 0; /* Make the transformed arg -> 0.0 */
+ } else {
+ Numer.lsw = Denom.lsw = argSignif.lsw;
+ XSIG_LL(Numer) = XSIG_LL(Denom) = XSIG_LL(argSignif);
+
+ if (exponent < -1)
+ shr_Xsig(&Numer, -1 - exponent);
+ negate_Xsig(&Numer);
+
+ shr_Xsig(&Denom, -exponent);
+ Denom.msw |= 0x80000000;
+
+ div_Xsig(&Numer, &Denom, &argSignif);
+
+ exponent = -1 + norm_Xsig(&argSignif);
+ }
+ } else {
+ transformed = 0;
+ }
+
+ argSq.lsw = argSignif.lsw;
+ argSq.midw = argSignif.midw;
+ argSq.msw = argSignif.msw;
+ mul_Xsig_Xsig(&argSq, &argSq);
+
+ argSqSq.lsw = argSq.lsw;
+ argSqSq.midw = argSq.midw;
+ argSqSq.msw = argSq.msw;
+ mul_Xsig_Xsig(&argSqSq, &argSqSq);
+
+ accumulatore.lsw = argSq.lsw;
+ XSIG_LL(accumulatore) = XSIG_LL(argSq);
+
+ shr_Xsig(&argSq, 2 * (-1 - exponent - 1));
+ shr_Xsig(&argSqSq, 4 * (-1 - exponent - 1));
+
+ /* Now have argSq etc with binary point at the left
+ .1xxxxxxxx */
+
+ /* Do the basic fixed point polynomial evaluation */
+ accumulator.msw = accumulator.midw = accumulator.lsw = 0;
+ polynomial_Xsig(&accumulator, &XSIG_LL(argSqSq),
+ oddplterms, HIPOWERop - 1);
+ mul64_Xsig(&accumulator, &XSIG_LL(argSq));
+ negate_Xsig(&accumulator);
+ polynomial_Xsig(&accumulator, &XSIG_LL(argSqSq), oddnegterms,
+ HIPOWERon - 1);
+ negate_Xsig(&accumulator);
+ add_two_Xsig(&accumulator, &fixedpterm, &dummy_exp);
+
+ mul64_Xsig(&accumulatore, &denomterm);
+ shr_Xsig(&accumulatore, 1 + 2 * (-1 - exponent));
+ accumulatore.msw |= 0x80000000;
+
+ div_Xsig(&accumulator, &accumulatore, &accumulator);
+
+ mul_Xsig_Xsig(&accumulator, &argSignif);
+ mul_Xsig_Xsig(&accumulator, &argSq);
+
+ shr_Xsig(&accumulator, 3);
+ negate_Xsig(&accumulator);
+ add_Xsig_Xsig(&accumulator, &argSignif);
+
+ if (transformed) {
+ /* compute pi/4 - accumulator */
+ shr_Xsig(&accumulator, -1 - exponent);
+ negate_Xsig(&accumulator);
+ add_Xsig_Xsig(&accumulator, &pi_signif);
+ exponent = -1;
+ }
+
+ if (inverted) {
+ /* compute pi/2 - accumulator */
+ shr_Xsig(&accumulator, -exponent);
+ negate_Xsig(&accumulator);
+ add_Xsig_Xsig(&accumulator, &pi_signif);
+ exponent = 0;
}
- else
- {
- Numer.lsw = Denom.lsw = argSignif.lsw;
- XSIG_LL(Numer) = XSIG_LL(Denom) = XSIG_LL(argSignif);
-
- if ( exponent < -1 )
- shr_Xsig(&Numer, -1-exponent);
- negate_Xsig(&Numer);
-
- shr_Xsig(&Denom, -exponent);
- Denom.msw |= 0x80000000;
-
- div_Xsig(&Numer, &Denom, &argSignif);
-
- exponent = -1 + norm_Xsig(&argSignif);
+
+ if (sign1) {
+ /* compute pi - accumulator */
+ shr_Xsig(&accumulator, 1 - exponent);
+ negate_Xsig(&accumulator);
+ add_Xsig_Xsig(&accumulator, &pi_signif);
+ exponent = 1;
}
- }
- else
- {
- transformed = 0;
- }
-
- argSq.lsw = argSignif.lsw; argSq.midw = argSignif.midw;
- argSq.msw = argSignif.msw;
- mul_Xsig_Xsig(&argSq, &argSq);
-
- argSqSq.lsw = argSq.lsw; argSqSq.midw = argSq.midw; argSqSq.msw = argSq.msw;
- mul_Xsig_Xsig(&argSqSq, &argSqSq);
-
- accumulatore.lsw = argSq.lsw;
- XSIG_LL(accumulatore) = XSIG_LL(argSq);
-
- shr_Xsig(&argSq, 2*(-1-exponent-1));
- shr_Xsig(&argSqSq, 4*(-1-exponent-1));
-
- /* Now have argSq etc with binary point at the left
- .1xxxxxxxx */
-
- /* Do the basic fixed point polynomial evaluation */
- accumulator.msw = accumulator.midw = accumulator.lsw = 0;
- polynomial_Xsig(&accumulator, &XSIG_LL(argSqSq),
- oddplterms, HIPOWERop-1);
- mul64_Xsig(&accumulator, &XSIG_LL(argSq));
- negate_Xsig(&accumulator);
- polynomial_Xsig(&accumulator, &XSIG_LL(argSqSq), oddnegterms, HIPOWERon-1);
- negate_Xsig(&accumulator);
- add_two_Xsig(&accumulator, &fixedpterm, &dummy_exp);
-
- mul64_Xsig(&accumulatore, &denomterm);
- shr_Xsig(&accumulatore, 1 + 2*(-1-exponent));
- accumulatore.msw |= 0x80000000;
-
- div_Xsig(&accumulator, &accumulatore, &accumulator);
-
- mul_Xsig_Xsig(&accumulator, &argSignif);
- mul_Xsig_Xsig(&accumulator, &argSq);
-
- shr_Xsig(&accumulator, 3);
- negate_Xsig(&accumulator);
- add_Xsig_Xsig(&accumulator, &argSignif);
-
- if ( transformed )
- {
- /* compute pi/4 - accumulator */
- shr_Xsig(&accumulator, -1-exponent);
- negate_Xsig(&accumulator);
- add_Xsig_Xsig(&accumulator, &pi_signif);
- exponent = -1;
- }
-
- if ( inverted )
- {
- /* compute pi/2 - accumulator */
- shr_Xsig(&accumulator, -exponent);
- negate_Xsig(&accumulator);
- add_Xsig_Xsig(&accumulator, &pi_signif);
- exponent = 0;
- }
-
- if ( sign1 )
- {
- /* compute pi - accumulator */
- shr_Xsig(&accumulator, 1 - exponent);
- negate_Xsig(&accumulator);
- add_Xsig_Xsig(&accumulator, &pi_signif);
- exponent = 1;
- }
-
- exponent += round_Xsig(&accumulator);
-
- significand(st1_ptr) = XSIG_LL(accumulator);
- setexponent16(st1_ptr, exponent);
-
- tag = FPU_round(st1_ptr, 1, 0, FULL_PRECISION, sign2);
- FPU_settagi(1, tag);
-
- set_precision_flag_up(); /* We do not really know if up or down,
- use this as the default. */
+
+ exponent += round_Xsig(&accumulator);
+
+ significand(st1_ptr) = XSIG_LL(accumulator);
+ setexponent16(st1_ptr, exponent);
+
+ tag = FPU_round(st1_ptr, 1, 0, FULL_PRECISION, sign2);
+ FPU_settagi(1, tag);
+
+ set_precision_flag_up(); /* We do not really know if up or down,
+ use this as the default. */
}