kerneltest/e32test/math/t_vfp.cpp
changeset 0 a41df078684a
child 31 56f325a607ea
equal deleted inserted replaced
-1:000000000000 0:a41df078684a
       
     1 // Copyright (c) 2003-2009 Nokia Corporation and/or its subsidiary(-ies).
       
     2 // All rights reserved.
       
     3 // This component and the accompanying materials are made available
       
     4 // under the terms of the License "Eclipse Public License v1.0"
       
     5 // which accompanies this distribution, and is available
       
     6 // at the URL "http://www.eclipse.org/legal/epl-v10.html".
       
     7 //
       
     8 // Initial Contributors:
       
     9 // Nokia Corporation - initial contribution.
       
    10 //
       
    11 // Contributors:
       
    12 //
       
    13 // Description:
       
    14 // e32test\math\t_vfp.cpp
       
    15 // Overview:
       
    16 // Test the ARM Vector Floating Point operations.
       
    17 // API Information:
       
    18 // VFP
       
    19 // Details:
       
    20 // - Check that the HAL agrees with the hardware about whether
       
    21 // VFP is supported.
       
    22 // - Test setting VFP to IEEE with no exceptions mode, if IEEE mode is
       
    23 // supported, otherwise leave the mode alone.
       
    24 // - Test single and double precision vector floating point operations:
       
    25 // ABS, NEG, ADD, SUB, MUL, DIV, NMUL, SQRT, MAC, MSC, NMAC and NMSC.
       
    26 // Verify results are as expected - if IEEE mode was set, verify
       
    27 // bit-for-bit, in accordance with the IEEE specification, otherwise
       
    28 // use normal floating point equality.
       
    29 // - Test VFP context save.
       
    30 // - Test various VFP operations that cause bounces to support code if
       
    31 // IEEE mode is supported.
       
    32 // - Test setting VFP to RunFast mode if RunFast mode is supported.
       
    33 // - Test setting VFP rounding mode.
       
    34 // - Test inheriting VFP mode to created threads.
       
    35 // Platforms/Drives/Compatibility:
       
    36 // All 
       
    37 // Assumptions/Requirement/Pre-requisites:
       
    38 // Failures and causes:
       
    39 // Base Port information:
       
    40 // 
       
    41 //
       
    42 
       
    43 //! @file
       
    44 //! @SYMTestCaseID KBASE-0017-T_VFP
       
    45 //! @SYMTestCaseDesc VFPv2 general functionality and bounce handling
       
    46 //! @SYMREQ 5159
       
    47 //! @SYMTestPriority Critical
       
    48 //! @SYMTestActions Check VFP functions correctly in all modes and that mode switching works correctly.
       
    49 //! @SYMTestExpectedResults Test runs until this message is emitted: RTEST: SUCCESS : T_VFP test completed O.K.
       
    50 //! @SYMTestType UT
       
    51 
       
    52 #include "t_vfp.h"
       
    53 #define __E32TEST_EXTENSION__
       
    54 #include <e32test.h>
       
    55 #include <e32math.h>
       
    56 #include <hal.h>
       
    57 #include <e32svr.h>
       
    58 #include <u32hal.h>
       
    59 
       
    60 RTest test(_L("T_VFP"));
       
    61 TUint32 FPSID;
       
    62 TUint32 ArchVersion; 
       
    63 TBool Double;
       
    64 TBool IEEEMode;
       
    65 TInt CPUs;
       
    66 TInt CurrentCpu1;
       
    67 TInt CurrentCpu2;
       
    68 
       
    69 typedef void TSglTest(const TReal32* aArgs, TReal32* aResults);
       
    70 typedef void TDblTest(const TReal64* aArgs, TReal64* aResults);
       
    71 
       
    72 TBool DetectVFP()
       
    73 	{
       
    74 	TInt r = UserSvr::HalFunction(EHalGroupKernel, EKernelHalFloatingPointSystemId, &FPSID, NULL);
       
    75 	return (r==KErrNone);
       
    76 	}
       
    77 
       
    78 TInt TestVFPInitThreadFn(TAny* aPtr)
       
    79 	{
       
    80 	UserSvr::HalFunction(EHalGroupKernel, EKernelHalLockThreadToCpu, (TAny*)CurrentCpu1, 0);
       
    81 	TReal32* p = (TReal32*)aPtr;
       
    82 	TInt i;
       
    83 	for (i=0; i<32; ++i)
       
    84 		*p++ = Vfp::SReg(i);
       
    85 	return 0;
       
    86 	}
       
    87 
       
    88 void TestVFPInitialState()
       
    89 	{
       
    90 	for (CurrentCpu1 = 0; CurrentCpu1 < CPUs; CurrentCpu1++)
       
    91 		{
       
    92 		TReal32 f[32];
       
    93 		RThread t;
       
    94 		TInt r = t.Create(KNullDesC, &TestVFPInitThreadFn, 0x1000, NULL, f);
       
    95 		test(r==KErrNone);
       
    96 		TRequestStatus s;
       
    97 		t.Logon(s);
       
    98 		t.Resume();
       
    99 		User::WaitForRequest(s);
       
   100 		TInt xt = t.ExitType();
       
   101 		TInt xr = t.ExitReason();
       
   102 		test(xt == EExitKill && xr == KErrNone);
       
   103 		CLOSE_AND_WAIT(t);
       
   104 		UserSvr::HalFunction(EHalGroupKernel, EKernelHalLockThreadToCpu, (TAny*)CurrentCpu1, 0);
       
   105 		test.Printf(_L("FPSCR = %08x for core %d\n"), Vfp::Fpscr(), CurrentCpu1);
       
   106 		const TUint32* p = (const TUint32*)f;
       
   107 		for (TInt i=0; i<32; ++i)
       
   108 			{
       
   109 			if (f[i] != 0.0f)
       
   110 				{
       
   111 				test.Printf(_L("S%d = 0x%08x\n"), i, p[i]);
       
   112 				test(f[i] == 0.0f);
       
   113 				}
       
   114 			}
       
   115 		}
       
   116 	}
       
   117 
       
   118 void TestVFPSglRegs(TInt aIter=2)
       
   119 	{
       
   120 	TInt i;
       
   121 	TInt j;
       
   122 	TInt nSglRegs=0; 
       
   123 
       
   124 	switch(ArchVersion)	
       
   125 		{ 
       
   126 		case ARCH_VERSION_VFPV2:
       
   127 		case ARCH_VERSION_VFPV3_SUBARCH_V2:
       
   128 		case ARCH_VERSION_VFPV3_SUBARCH_NULL:
       
   129 		case ARCH_VERSION_VFPV3_SUBARCH_V3:
       
   130 			nSglRegs = 32;
       
   131 			break; 		
       
   132 		case 0:
       
   133 		default:
       
   134 			__ASSERT_ALWAYS(0, User::Panic(_L("Bad VFP version"),__LINE__)); 
       
   135 			/* NOTREACHED */
       
   136 		} 
       
   137 
       
   138 	for (i=0; i<aIter; ++i)
       
   139 		{
       
   140 		for (j=0; j<nSglRegs; ++j)
       
   141 			{
       
   142 			TInt32 f = i + j;
       
   143 			Vfp::SetSReg(f, j);
       
   144 			}
       
   145 		for (j=0; j<nSglRegs; ++j)
       
   146 			{
       
   147 			TInt32 f = i + j;
       
   148 			TInt32 g = Vfp::SRegInt(j);
       
   149 			test(f == g);
       
   150 			}
       
   151 		}
       
   152 	}
       
   153 
       
   154 TInt TestVFPSglRegsThread(TAny*)
       
   155 	{
       
   156 	UserSvr::HalFunction(EHalGroupKernel, EKernelHalLockThreadToCpu, (TAny*)CurrentCpu1, 0);
       
   157 	TestVFPSglRegs(KMaxTInt);
       
   158 	return 0;
       
   159 	}
       
   160 
       
   161 void TestVFPDblRegs(TInt aIter=2)
       
   162 	{
       
   163 	TInt i;
       
   164 	TInt j;
       
   165 	TInt nDblRegs=0; 
       
   166 
       
   167 	switch(ArchVersion)
       
   168 		{ 
       
   169 		case ARCH_VERSION_VFPV2:
       
   170 			{
       
   171 			nDblRegs = 16;
       
   172 			break;
       
   173 			}
       
   174 		case ARCH_VERSION_VFPV3_SUBARCH_V2:
       
   175 		case ARCH_VERSION_VFPV3_SUBARCH_NULL:
       
   176 		case ARCH_VERSION_VFPV3_SUBARCH_V3:
       
   177 			{
       
   178 			TInt vfpType;
       
   179 			TInt ret = HAL::Get(HALData::EHardwareFloatingPoint, vfpType);
       
   180 			if (ret == KErrNone && vfpType == EFpTypeVFPv3)
       
   181 				nDblRegs = 32;
       
   182 			else
       
   183 				nDblRegs = 16;
       
   184 			break;
       
   185 				}
       
   186 		case 0:
       
   187 		default:
       
   188 			__ASSERT_ALWAYS(0, User::Panic(_L("Bad VFP version"),__LINE__)); 
       
   189 		} 
       
   190 
       
   191 
       
   192 	for (i=0; i<aIter; ++i)
       
   193 		{
       
   194 		for (j=0; j<nDblRegs; ++j)
       
   195 			{
       
   196 			TInt64 f = i + j + KMaxTUint;
       
   197 			Vfp::SetDReg(f, j);
       
   198 			}
       
   199 		for (j=0; j<nDblRegs; ++j)
       
   200 			{
       
   201 			TInt64 f = i + j + KMaxTUint;
       
   202 			TInt64 g = Vfp::DRegInt(j);
       
   203 			test(f == g);
       
   204 			}
       
   205 		}
       
   206 	}
       
   207 
       
   208 TInt TestVFPDblRegsThread(TAny*)
       
   209 	{
       
   210 	UserSvr::HalFunction(EHalGroupKernel, EKernelHalLockThreadToCpu, (TAny*)CurrentCpu2, 0);
       
   211 	TestVFPDblRegs(KMaxTInt);
       
   212 	return 0;
       
   213 	}
       
   214 
       
   215 void TestVFPContextSave()
       
   216 	{
       
   217 	for (CurrentCpu2 = 0; CurrentCpu2 < CPUs; CurrentCpu2++)
       
   218 		{
       
   219 		for (CurrentCpu1 = 0; CurrentCpu1 < CPUs; CurrentCpu1++)
       
   220 			{
       
   221 			TThreadFunction tf1 = &TestVFPSglRegsThread;
       
   222 			TThreadFunction tf2 = Double ? &TestVFPDblRegsThread : &TestVFPSglRegsThread;
       
   223 			RThread t1, t2;
       
   224 			TInt r;
       
   225 			r = t1.Create(KNullDesC, tf1, 0x1000, 0x1000, 0x1000, NULL);
       
   226 			test(r==KErrNone);
       
   227 			t1.SetPriority(EPriorityLess);
       
   228 			r = t2.Create(KNullDesC, tf2, 0x1000, 0x1000, 0x1000, NULL);
       
   229 			test(r==KErrNone);
       
   230 			t2.SetPriority(EPriorityLess);
       
   231 			TRequestStatus s1, s2;
       
   232 			t1.Logon(s1);
       
   233 			t2.Logon(s2);
       
   234 			t1.Resume();
       
   235 			t2.Resume();
       
   236 			test.Printf(_L("Let threads run concurrently (cores %d and %d)\n"), CurrentCpu1, CurrentCpu2);
       
   237 			User::After(20*1000*1000/CPUs);
       
   238 
       
   239 			test.Printf(_L("Kill threads\n"));
       
   240 			t1.Kill(0);
       
   241 			t2.Kill(0);
       
   242 			User::WaitForRequest(s1);
       
   243 			User::WaitForRequest(s2);
       
   244 			test(t1.ExitType()==EExitKill && t1.ExitReason()==KErrNone);
       
   245 			test(t2.ExitType()==EExitKill && t2.ExitReason()==KErrNone);
       
   246 			CLOSE_AND_WAIT(t1);
       
   247 			CLOSE_AND_WAIT(t2);
       
   248 			}
       
   249 		}
       
   250 	}
       
   251 
       
   252 TInt TestBounceCtxThread1(TAny*)
       
   253 	{
       
   254 	for(TInt iter=0; iter<KMaxTInt; ++iter)
       
   255 		{
       
   256 		Vfp::SReg(0);
       
   257 		}
       
   258 	return KErrNone;
       
   259 	}
       
   260 
       
   261 TInt TestBounceCtxThread2(TAny*)
       
   262 	{
       
   263 	TInt start_rep = 0x00800000; // smallest single precision normal number, 1*2^-126
       
   264 	TReal32 start = *(TReal32*)&start_rep;
       
   265 	for(TInt iter=0; iter<KMaxTInt; ++iter)
       
   266 		{
       
   267 		Vfp::SetSReg(start, 1);
       
   268 		Vfp::SetSReg(2.0f, 2);
       
   269 		Vfp::DivS();
       
   270 		Vfp::CpyS0(1);
       
   271 		Vfp::MulS();
       
   272 		Vfp::CpyS0(1);
       
   273 		TReal32 end = Vfp::SReg(0);
       
   274 		TInt end_rep = *(TInt*)&end;
       
   275 		if (start_rep != end_rep)
       
   276 			{
       
   277 			RDebug::Printf("mismatch in iter %d, start %08x end %08x\n", iter, start_rep, end_rep);
       
   278 			test(0);
       
   279 			}
       
   280 		}
       
   281 	return KErrNone;
       
   282 	}
       
   283 
       
   284 void DoBounceContextSwitchTests()
       
   285 	{
       
   286 	RThread t1, t2;
       
   287 	TInt r;
       
   288 	r = t1.Create(KNullDesC, &TestBounceCtxThread1, 0x1000, 0x1000, 0x1000, NULL);
       
   289 	test(r==KErrNone);
       
   290 	t1.SetPriority(EPriorityLess);
       
   291 	r = t2.Create(KNullDesC, &TestBounceCtxThread2, 0x1000, 0x1000, 0x1000, NULL);
       
   292 	test(r==KErrNone);
       
   293 	t2.SetPriority(EPriorityLess);
       
   294 	TRequestStatus s1, s2;
       
   295 	t1.Logon(s1);
       
   296 	t2.Logon(s2);
       
   297 	t1.Resume();
       
   298 	t2.Resume();
       
   299 	test.Printf(_L("Let threads run concurrently ...\n"));
       
   300 	User::After(20*1000*1000);
       
   301 
       
   302 	test.Printf(_L("Kill threads\n"));
       
   303 	t1.Kill(0);
       
   304 	t2.Kill(0);
       
   305 	User::WaitForRequest(s1);
       
   306 	User::WaitForRequest(s2);
       
   307 	test(t1.ExitType()==EExitKill && t1.ExitReason()==KErrNone);
       
   308 	test(t2.ExitType()==EExitKill && t2.ExitReason()==KErrNone);
       
   309 	CLOSE_AND_WAIT(t1);
       
   310 	CLOSE_AND_WAIT(t2);
       
   311 	}
       
   312 
       
   313 void TestAbsS(const TReal32* a, TReal32* r)
       
   314 	{
       
   315 	Vfp::SetSReg(a[0], 1);
       
   316 	Vfp::AbsS();
       
   317 	r[0] = Vfp::SReg(0);
       
   318 	r[1] = Abs(a[0]);
       
   319 	}
       
   320 
       
   321 void TestAddS(const TReal32* a, TReal32* r)
       
   322 	{
       
   323 	Vfp::SetSReg(a[0], 1);
       
   324 	Vfp::SetSReg(a[1], 2);
       
   325 	Vfp::AddS();
       
   326 	r[0] = Vfp::SReg(0);
       
   327 	r[1] = a[0] + a[1];
       
   328 	}
       
   329 
       
   330 void TestDivS(const TReal32* a, TReal32* r)
       
   331 	{
       
   332 	Vfp::SetSReg(a[0], 1);
       
   333 	Vfp::SetSReg(a[1], 2);
       
   334 	Vfp::DivS();
       
   335 	r[0] = Vfp::SReg(0);
       
   336 	TRealX x(a[0]);
       
   337 	TRealX y(a[1]);
       
   338 	x.DivEq(y);
       
   339 	r[1] = (TReal32)x;
       
   340 	}
       
   341 
       
   342 void TestMacS(const TReal32* a, TReal32* r)
       
   343 	{
       
   344 	Vfp::SetSReg(a[0], 0);
       
   345 	Vfp::SetSReg(a[1], 1);
       
   346 	Vfp::SetSReg(a[2], 2);
       
   347 	Vfp::MacS();
       
   348 	r[0] = Vfp::SReg(0);
       
   349 	r[1] = a[0] + a[1] * a[2];
       
   350 	}
       
   351 
       
   352 void TestMscS(const TReal32* a, TReal32* r)
       
   353 	{
       
   354 	Vfp::SetSReg(a[0], 0);
       
   355 	Vfp::SetSReg(a[1], 1);
       
   356 	Vfp::SetSReg(a[2], 2);
       
   357 	Vfp::MscS();
       
   358 	r[0] = Vfp::SReg(0);
       
   359 	r[1] = a[1] * a[2] - a[0];
       
   360 	}
       
   361 
       
   362 void TestMulS(const TReal32* a, TReal32* r)
       
   363 	{
       
   364 	Vfp::SetSReg(a[0], 1);
       
   365 	Vfp::SetSReg(a[1], 2);
       
   366 	Vfp::MulS();
       
   367 	r[0] = Vfp::SReg(0);
       
   368 	TRealX x(a[0]);
       
   369 	TRealX y(a[1]);
       
   370 	x.MultEq(y);
       
   371 	r[1] = (TReal32)x;
       
   372 	}
       
   373 
       
   374 void TestNegS(const TReal32* a, TReal32* r)
       
   375 	{
       
   376 	Vfp::SetSReg(a[0], 1);
       
   377 	Vfp::NegS();
       
   378 	r[0] = Vfp::SReg(0);
       
   379 	r[1] = -a[0];
       
   380 	}
       
   381 
       
   382 void TestNMacS(const TReal32* a, TReal32* r)
       
   383 	{
       
   384 	Vfp::SetSReg(a[0], 0);
       
   385 	Vfp::SetSReg(a[1], 1);
       
   386 	Vfp::SetSReg(a[2], 2);
       
   387 	Vfp::NMacS();
       
   388 	r[0] = Vfp::SReg(0);
       
   389 	r[1] = a[0] - a[1] * a[2];
       
   390 	}
       
   391 
       
   392 void TestNMscS(const TReal32* a, TReal32* r)
       
   393 	{
       
   394 	Vfp::SetSReg(a[0], 0);
       
   395 	Vfp::SetSReg(a[1], 1);
       
   396 	Vfp::SetSReg(a[2], 2);
       
   397 	Vfp::NMscS();
       
   398 	r[0] = Vfp::SReg(0);
       
   399 	r[1] = -a[1] * a[2] - a[0];
       
   400 	}
       
   401 
       
   402 void TestNMulS(const TReal32* a, TReal32* r)
       
   403 	{
       
   404 	Vfp::SetSReg(a[0], 1);
       
   405 	Vfp::SetSReg(a[1], 2);
       
   406 	Vfp::NMulS();
       
   407 	r[0] = Vfp::SReg(0);
       
   408 	TRealX x(a[0]);
       
   409 	TRealX y(a[1]);
       
   410 	x.MultEq(y);
       
   411 	r[1] = -(TReal32)x;
       
   412 	}
       
   413 
       
   414 void TestSqrtS(const TReal32* a, TReal32* r)
       
   415 	{
       
   416 	Vfp::SetSReg(a[0], 1);
       
   417 	Vfp::SqrtS();
       
   418 	r[0] = Vfp::SReg(0);
       
   419 	TReal x = a[0];
       
   420 	TReal y;
       
   421 	Math::Sqrt(y, x);
       
   422 	r[1] = (TReal32)y;
       
   423 	}
       
   424 
       
   425 void TestSubS(const TReal32* a, TReal32* r)
       
   426 	{
       
   427 	Vfp::SetSReg(a[0], 1);
       
   428 	Vfp::SetSReg(a[1], 2);
       
   429 	Vfp::SubS();
       
   430 	r[0] = Vfp::SReg(0);
       
   431 	r[1] = a[0] - a[1];
       
   432 	}
       
   433 
       
   434 
       
   435 
       
   436 void TestAbsD(const TReal64* a, TReal64* r)
       
   437 	{
       
   438 	Vfp::SetDReg(a[0], 1);
       
   439 	Vfp::AbsD();
       
   440 	r[0] = Vfp::DReg(0);
       
   441 	r[1] = Abs(a[0]);
       
   442 	}
       
   443 
       
   444 void TestAddD(const TReal64* a, TReal64* r)
       
   445 	{
       
   446 	Vfp::SetDReg(a[0], 1);
       
   447 	Vfp::SetDReg(a[1], 2);
       
   448 	Vfp::AddD();
       
   449 	r[0] = Vfp::DReg(0);
       
   450 	r[1] = a[0] + a[1];
       
   451 	}
       
   452 
       
   453 void TestDivD(const TReal64* a, TReal64* r)
       
   454 	{
       
   455 	Vfp::SetDReg(a[0], 1);
       
   456 	Vfp::SetDReg(a[1], 2);
       
   457 	Vfp::DivD();
       
   458 	r[0] = Vfp::DReg(0);
       
   459 	TRealX x(a[0]);
       
   460 	TRealX y(a[1]);
       
   461 	x.DivEq(y);
       
   462 	r[1] = (TReal64)x;
       
   463 	}
       
   464 
       
   465 void TestMacD(const TReal64* a, TReal64* r)
       
   466 	{
       
   467 	Vfp::SetDReg(a[0], 0);
       
   468 	Vfp::SetDReg(a[1], 1);
       
   469 	Vfp::SetDReg(a[2], 2);
       
   470 	Vfp::MacD();
       
   471 	r[0] = Vfp::DReg(0);
       
   472 	r[1] = a[0] + a[1] * a[2];
       
   473 	}
       
   474 
       
   475 void TestMscD(const TReal64* a, TReal64* r)
       
   476 	{
       
   477 	Vfp::SetDReg(a[0], 0);
       
   478 	Vfp::SetDReg(a[1], 1);
       
   479 	Vfp::SetDReg(a[2], 2);
       
   480 	Vfp::MscD();
       
   481 	r[0] = Vfp::DReg(0);
       
   482 	r[1] = a[1] * a[2] - a[0];
       
   483 	}
       
   484 
       
   485 void TestMulD(const TReal64* a, TReal64* r)
       
   486 	{
       
   487 	Vfp::SetDReg(a[0], 1);
       
   488 	Vfp::SetDReg(a[1], 2);
       
   489 	Vfp::MulD();
       
   490 	r[0] = Vfp::DReg(0);
       
   491 	TRealX x(a[0]);
       
   492 	TRealX y(a[1]);
       
   493 	x.MultEq(y);
       
   494 	r[1] = (TReal64)x;
       
   495 	}
       
   496 
       
   497 void TestNegD(const TReal64* a, TReal64* r)
       
   498 	{
       
   499 	Vfp::SetDReg(a[0], 1);
       
   500 	Vfp::NegD();
       
   501 	r[0] = Vfp::DReg(0);
       
   502 	r[1] = -a[0];
       
   503 	}
       
   504 
       
   505 void TestNMacD(const TReal64* a, TReal64* r)
       
   506 	{
       
   507 	Vfp::SetDReg(a[0], 0);
       
   508 	Vfp::SetDReg(a[1], 1);
       
   509 	Vfp::SetDReg(a[2], 2);
       
   510 	Vfp::NMacD();
       
   511 	r[0] = Vfp::DReg(0);
       
   512 	r[1] = a[0] - a[1] * a[2];
       
   513 	}
       
   514 
       
   515 void TestNMscD(const TReal64* a, TReal64* r)
       
   516 	{
       
   517 	Vfp::SetDReg(a[0], 0);
       
   518 	Vfp::SetDReg(a[1], 1);
       
   519 	Vfp::SetDReg(a[2], 2);
       
   520 	Vfp::NMscD();
       
   521 	r[0] = Vfp::DReg(0);
       
   522 	r[1] = -a[1] * a[2] - a[0];
       
   523 	}
       
   524 
       
   525 void TestNMulD(const TReal64* a, TReal64* r)
       
   526 	{
       
   527 	Vfp::SetDReg(a[0], 1);
       
   528 	Vfp::SetDReg(a[1], 2);
       
   529 	Vfp::NMulD();
       
   530 	r[0] = Vfp::DReg(0);
       
   531 	TRealX x(a[0]);
       
   532 	TRealX y(a[1]);
       
   533 	x.MultEq(y);
       
   534 	r[1] = -(TReal64)x;
       
   535 	}
       
   536 
       
   537 void TestSqrtD(const TReal64* a, TReal64* r)
       
   538 	{
       
   539 	Vfp::SetDReg(a[0], 1);
       
   540 	Vfp::SqrtD();
       
   541 	r[0] = Vfp::DReg(0);
       
   542 	TReal x = a[0];
       
   543 	TReal y;
       
   544 	Math::Sqrt(y, x);
       
   545 	r[1] = (TReal64)y;
       
   546 	}
       
   547 
       
   548 void TestSubD(const TReal64* a, TReal64* r)
       
   549 	{
       
   550 	Vfp::SetDReg(a[0], 1);
       
   551 	Vfp::SetDReg(a[1], 2);
       
   552 	Vfp::SubD();
       
   553 	r[0] = Vfp::DReg(0);
       
   554 	r[1] = a[0] - a[1];
       
   555 	}
       
   556 
       
   557 #define DO_SGL_TEST1(name, func, a1)			DoSglTest(name, __LINE__, func, a1)
       
   558 #define DO_SGL_TEST2(name, func, a1, a2)		DoSglTest(name, __LINE__, func, a1, a2)
       
   559 #define DO_SGL_TEST3(name, func, a1, a2, a3)	DoSglTest(name, __LINE__, func, a1, a2, a3)
       
   560 void DoSglTest(const char* aName, TInt aLine, TSglTest aFunc, TReal32 a1, TReal32 a2=0.0f, TReal32 a3=0.0f)
       
   561 	{
       
   562 	TPtrC8 name8((const TText8*)aName);
       
   563 	TBuf<128> name16;
       
   564 	name16.Copy(name8);
       
   565 	test.Printf(_L("%S(%g,%g,%g)\n"), &name16, a1, a2, a3);
       
   566 	TReal32 args[3] = {a1, a2, a3};
       
   567 	TReal32 results[2];
       
   568 	(*aFunc)(args, results);
       
   569 	if (IEEEMode)
       
   570 		{
       
   571 		if (*((TUint32*)&(results[0])) == *((TUint32*)&(results[1])))
       
   572 			return;
       
   573 		}
       
   574 	else
       
   575 		{
       
   576 		if (results[0] == results[1])
       
   577 			return;
       
   578 		}
       
   579 	const TUint32* pa = (const TUint32*)args;
       
   580 	const TUint32* pr = (const TUint32*)results;
       
   581 	test.Printf(_L("a1=%08x a2=%08x a3=%08x\n"), pa[0], pa[1], pa[2]);
       
   582 	test.Printf(_L("actual result = %08x (%g)\nexpected result = %08x (%g)\n"), pr[0], results[0], pr[1], results[1]);
       
   583 	test.Printf(_L("Test at line %d failed\n"), aLine);
       
   584 	test(0);
       
   585 	}
       
   586 
       
   587 void DoSglTests()
       
   588 	{
       
   589 	// ABS
       
   590 	DO_SGL_TEST1("ABS", &TestAbsS, 1.0f);
       
   591 	DO_SGL_TEST1("ABS", &TestAbsS, -1.0f);
       
   592 	DO_SGL_TEST1("ABS", &TestAbsS, 0.0f);
       
   593 	DO_SGL_TEST1("ABS", &TestAbsS, -3.1415926536f);
       
   594 
       
   595 	// NEG
       
   596 	DO_SGL_TEST1("NEG", &TestNegS, 1.0f);
       
   597 	DO_SGL_TEST1("NEG", &TestNegS, -1.0f);
       
   598 	DO_SGL_TEST1("NEG", &TestNegS, 0.0f);
       
   599 	DO_SGL_TEST1("NEG", &TestNegS, -3.1415926536f);
       
   600 
       
   601 	// ADD
       
   602 	DO_SGL_TEST2("ADD", &TestAddS, 0.0f, 0.0f);
       
   603 	DO_SGL_TEST2("ADD", &TestAddS, 0.0f, 1.0f);
       
   604 	DO_SGL_TEST2("ADD", &TestAddS, -1.0f, 1.0f);
       
   605 	DO_SGL_TEST2("ADD", &TestAddS, 1.0f, 2.5f);
       
   606 	DO_SGL_TEST2("ADD", &TestAddS, 1.0f, 6.022045e23f);
       
   607 	DO_SGL_TEST2("ADD", &TestAddS, -7.3890561f, 1.414213562f);
       
   608 	DO_SGL_TEST2("ADD", &TestAddS, -7.3890561f, -1.414213562f);
       
   609 
       
   610 	// SUB
       
   611 	DO_SGL_TEST2("SUB", &TestSubS, 0.0f, 0.0f);
       
   612 	DO_SGL_TEST2("SUB", &TestSubS, 0.0f, 1.0f);
       
   613 	DO_SGL_TEST2("SUB", &TestSubS, 1.0f, 1.0f);
       
   614 	DO_SGL_TEST2("SUB", &TestSubS, 1.0f, 2.5f);
       
   615 	DO_SGL_TEST2("SUB", &TestSubS, 91.0f, 2.5f);
       
   616 	DO_SGL_TEST2("SUB", &TestSubS, 1.0f, 6.022045e23f);
       
   617 	DO_SGL_TEST2("SUB", &TestSubS, -7.3890561f, 1.414213562f);
       
   618 	DO_SGL_TEST2("SUB", &TestSubS, -7.3890561f, -1.414213562f);
       
   619 
       
   620 	// MUL
       
   621 	DO_SGL_TEST2("MUL", &TestMulS, 0.0f, 0.0f);
       
   622 	DO_SGL_TEST2("MUL", &TestMulS, 1.0f, 0.0f);
       
   623 	DO_SGL_TEST2("MUL", &TestMulS, 0.0f, 1.0f);
       
   624 	DO_SGL_TEST2("MUL", &TestMulS, 2.5f, 6.5f);
       
   625 	DO_SGL_TEST2("MUL", &TestMulS, -39.6f, 19.72f);
       
   626 	DO_SGL_TEST2("MUL", &TestMulS, -10.1f, -20.1f);
       
   627 	DO_SGL_TEST2("MUL", &TestMulS, 1e20f, 1e20f);
       
   628 	DO_SGL_TEST2("MUL", &TestMulS, 1e-30f, 1e-30f);
       
   629 
       
   630 	// DIV
       
   631 	DO_SGL_TEST2("DIV", &TestDivS, 0.0f, 1.0f);
       
   632 	DO_SGL_TEST2("DIV", &TestDivS, 1.0f, 5.0f);
       
   633 	DO_SGL_TEST2("DIV", &TestDivS, 1.0f, -5.0f);
       
   634 	DO_SGL_TEST2("DIV", &TestDivS, -1.0f, 5.0f);
       
   635 	DO_SGL_TEST2("DIV", &TestDivS, -1.0f, -5.0f);
       
   636 	DO_SGL_TEST2("DIV", &TestDivS, 7.3890561f, 2.7182818f);
       
   637 	DO_SGL_TEST2("DIV", &TestDivS, 1e20f, 1e-20f);
       
   638 	DO_SGL_TEST2("DIV", &TestDivS, 1e-30f, 1e30f);
       
   639 
       
   640 	// NMUL
       
   641 	DO_SGL_TEST2("NMUL", &TestNMulS, 0.0f, 0.0f);
       
   642 	DO_SGL_TEST2("NMUL", &TestNMulS, 1.0f, 0.0f);
       
   643 	DO_SGL_TEST2("NMUL", &TestNMulS, 0.0f, 1.0f);
       
   644 	DO_SGL_TEST2("NMUL", &TestNMulS, 2.5f, 6.5f);
       
   645 	DO_SGL_TEST2("NMUL", &TestNMulS, -39.6f, 19.72f);
       
   646 	DO_SGL_TEST2("NMUL", &TestNMulS, -10.1f, -20.1f);
       
   647 	DO_SGL_TEST2("NMUL", &TestNMulS, 1e20f, 1e20f);
       
   648 	DO_SGL_TEST2("NMUL", &TestNMulS, 1e-30f, 1e-30f);
       
   649 
       
   650 	// SQRT
       
   651 	DO_SGL_TEST1("SQRT", &TestSqrtS, 0.0f);
       
   652 	DO_SGL_TEST1("SQRT", &TestSqrtS, 1.0f);
       
   653 	DO_SGL_TEST1("SQRT", &TestSqrtS, 2.0f);
       
   654 	DO_SGL_TEST1("SQRT", &TestSqrtS, 3.0f);
       
   655 	DO_SGL_TEST1("SQRT", &TestSqrtS, 9096256.0f);
       
   656 	DO_SGL_TEST1("SQRT", &TestSqrtS, 1e36f);
       
   657 	DO_SGL_TEST1("SQRT", &TestSqrtS, 1e-36f);
       
   658 
       
   659 	// MAC
       
   660 	DO_SGL_TEST3("MAC", &TestMacS, 0.0f, 0.0f, 0.0f);
       
   661 	DO_SGL_TEST3("MAC", &TestMacS, 0.0f, 1.0f, 0.0f);
       
   662 	DO_SGL_TEST3("MAC", &TestMacS, 0.0f, 1.0f, 1.0f);
       
   663 	DO_SGL_TEST3("MAC", &TestMacS, -1.0f, 1.0f, 1.0f);
       
   664 	DO_SGL_TEST3("MAC", &TestMacS, 0.8f, 0.1f, 8.0f);
       
   665 	DO_SGL_TEST3("MAC", &TestMacS, 0.8f, -0.1f, 8.0f);
       
   666 	DO_SGL_TEST3("MAC", &TestMacS, -0.8f, -0.1f, -8.0f);
       
   667 	DO_SGL_TEST3("MAC", &TestMacS, 0.8f, 0.3333333333f, 3.1415926536f);
       
   668 
       
   669 	// MSC
       
   670 	DO_SGL_TEST3("MSC", &TestMscS, 0.0f, 0.0f, 0.0f);
       
   671 	DO_SGL_TEST3("MSC", &TestMscS, 0.0f, 1.0f, 0.0f);
       
   672 	DO_SGL_TEST3("MSC", &TestMscS, 0.0f, 1.0f, 1.0f);
       
   673 	DO_SGL_TEST3("MSC", &TestMscS, -1.0f, 1.0f, 1.0f);
       
   674 	DO_SGL_TEST3("MSC", &TestMscS, 0.8f, 0.1f, 8.0f);
       
   675 	DO_SGL_TEST3("MSC", &TestMscS, 0.8f, -0.1f, 8.0f);
       
   676 	DO_SGL_TEST3("MSC", &TestMscS, -0.8f, -0.1f, -8.0f);
       
   677 	DO_SGL_TEST3("MSC", &TestMscS, 0.8f, 0.3333333333f, 3.1415926536f);
       
   678 
       
   679 	// NMAC
       
   680 	DO_SGL_TEST3("NMAC", &TestNMacS, 0.0f, 0.0f, 0.0f);
       
   681 	DO_SGL_TEST3("NMAC", &TestNMacS, 0.0f, 1.0f, 0.0f);
       
   682 	DO_SGL_TEST3("NMAC", &TestNMacS, 0.0f, 1.0f, 1.0f);
       
   683 	DO_SGL_TEST3("NMAC", &TestNMacS, -1.0f, 1.0f, 1.0f);
       
   684 	DO_SGL_TEST3("NMAC", &TestNMacS, 0.8f, 0.1f, 8.0f);
       
   685 	DO_SGL_TEST3("NMAC", &TestNMacS, 0.8f, -0.1f, 8.0f);
       
   686 	DO_SGL_TEST3("NMAC", &TestNMacS, -0.8f, -0.1f, -8.0f);
       
   687 	DO_SGL_TEST3("NMAC", &TestNMacS, 0.8f, 0.3333333333f, 3.1415926536f);
       
   688 
       
   689 	// NMSC
       
   690 	DO_SGL_TEST3("NMSC", &TestNMscS, 0.0f, 0.0f, 0.0f);
       
   691 	DO_SGL_TEST3("NMSC", &TestNMscS, 0.0f, 1.0f, 0.0f);
       
   692 	DO_SGL_TEST3("NMSC", &TestNMscS, 0.0f, 1.0f, 1.0f);
       
   693 	DO_SGL_TEST3("NMSC", &TestNMscS, -1.0f, 1.0f, 1.0f);
       
   694 	DO_SGL_TEST3("NMSC", &TestNMscS, 0.8f, 0.1f, 8.0f);
       
   695 	DO_SGL_TEST3("NMSC", &TestNMscS, 0.8f, -0.1f, 8.0f);
       
   696 	DO_SGL_TEST3("NMSC", &TestNMscS, -0.8f, -0.1f, -8.0f);
       
   697 	DO_SGL_TEST3("NMSC", &TestNMscS, 0.8f, 0.3333333333f, 3.1415926536f);
       
   698 	}
       
   699 
       
   700 #define DO_DBL_TEST1(name, func, a1)			DoDblTest(name, __LINE__, func, a1)
       
   701 #define DO_DBL_TEST2(name, func, a1, a2)		DoDblTest(name, __LINE__, func, a1, a2)
       
   702 #define DO_DBL_TEST3(name, func, a1, a2, a3)	DoDblTest(name, __LINE__, func, a1, a2, a3)
       
   703 void DoDblTest(const char* aName, TInt aLine, TDblTest aFunc, TReal64 a1, TReal64 a2=0.0, TReal64 a3=0.0)
       
   704 	{
       
   705 	TPtrC8 name8((const TText8*)aName);
       
   706 	TBuf<128> name16;
       
   707 	name16.Copy(name8);
       
   708 	test.Printf(_L("%S(%g,%g,%g)\n"), &name16, a1, a2, a3);
       
   709 	TReal64 args[3] = {a1, a2, a3};
       
   710 	TReal64 results[2];
       
   711 	SDouble sargs[3];
       
   712 	sargs[0] = a1;
       
   713 	sargs[1] = a2;
       
   714 	sargs[2] = a3;
       
   715 	(*aFunc)(args, results);
       
   716 	if (IEEEMode)
       
   717 		{
       
   718 		if (*((TUint64*)&(results[0])) == *((TUint64*)&(results[1])))
       
   719 			return;
       
   720 		}
       
   721 	else
       
   722 		{
       
   723 		if (results[0] == results[1])
       
   724 			return;
       
   725 		}
       
   726 	SDouble sres[3];
       
   727 	sres[0] = results[0];
       
   728 	sres[1] = results[1];
       
   729 	test.Printf(_L("a1=%08x %08x\na2=%08x %08x\na3=%08x %08x\n"), sargs[0].iData[1], sargs[0].iData[0],
       
   730 								sargs[1].iData[1], sargs[1].iData[0], sargs[2].iData[1], sargs[2].iData[0]);
       
   731 	test.Printf(_L("actual result = %08x %08x (%g)\nexpected result = %08x %08x (%g)\n"),
       
   732 			sres[0].iData[1], sres[0].iData[0], results[0], sres[1].iData[1], sres[1].iData[0], results[1]);
       
   733 	test.Printf(_L("Test at line %d failed\n"), aLine);
       
   734 	test(0);
       
   735 	}
       
   736 
       
   737 void DoDblTests()
       
   738 	{
       
   739 	// ABS
       
   740 	DO_DBL_TEST1("ABS", &TestAbsD, 1.0);
       
   741 	DO_DBL_TEST1("ABS", &TestAbsD, -1.0);
       
   742 	DO_DBL_TEST1("ABS", &TestAbsD, 0.0);
       
   743 	DO_DBL_TEST1("ABS", &TestAbsD, -3.1415926536);
       
   744 
       
   745 	// NEG
       
   746 	DO_DBL_TEST1("NEG", &TestNegD, 1.0);
       
   747 	DO_DBL_TEST1("NEG", &TestNegD, -1.0);
       
   748 	DO_DBL_TEST1("NEG", &TestNegD, 0.0);
       
   749 	DO_DBL_TEST1("NEG", &TestNegD, -3.1415926536);
       
   750 
       
   751 	// ADD
       
   752 	DO_DBL_TEST2("ADD", &TestAddD, 0.0, 0.0);
       
   753 	DO_DBL_TEST2("ADD", &TestAddD, 0.0, 1.0);
       
   754 	DO_DBL_TEST2("ADD", &TestAddD, -1.0, 1.0);
       
   755 	DO_DBL_TEST2("ADD", &TestAddD, 1.0, 2.5);
       
   756 	DO_DBL_TEST2("ADD", &TestAddD, 1.0, 6.022045e23);
       
   757 	DO_DBL_TEST2("ADD", &TestAddD, -7.3890561, 1.414213562);
       
   758 	DO_DBL_TEST2("ADD", &TestAddD, -7.3890561, -1.414213562);
       
   759 
       
   760 	// SUB
       
   761 	DO_DBL_TEST2("SUB", &TestSubD, 0.0, 0.0);
       
   762 	DO_DBL_TEST2("SUB", &TestSubD, 0.0, 1.0);
       
   763 	DO_DBL_TEST2("SUB", &TestSubD, 1.0, 1.0);
       
   764 	DO_DBL_TEST2("SUB", &TestSubD, 1.0, 2.5);
       
   765 	DO_DBL_TEST2("SUB", &TestSubD, 91.0, 2.5);
       
   766 	DO_DBL_TEST2("SUB", &TestSubD, 1.0, 6.022045e23);
       
   767 	DO_DBL_TEST2("SUB", &TestSubD, -7.3890561, 1.414213562);
       
   768 	DO_DBL_TEST2("SUB", &TestSubD, -7.3890561, -1.414213562);
       
   769 
       
   770 	// MUL
       
   771 	DO_DBL_TEST2("MUL", &TestMulD, 0.0, 0.0);
       
   772 	DO_DBL_TEST2("MUL", &TestMulD, 1.0, 0.0);
       
   773 	DO_DBL_TEST2("MUL", &TestMulD, 0.0, 1.0);
       
   774 	DO_DBL_TEST2("MUL", &TestMulD, 2.5, 6.5);
       
   775 	DO_DBL_TEST2("MUL", &TestMulD, -39.6, 19.72);
       
   776 	DO_DBL_TEST2("MUL", &TestMulD, -10.1, -20.1);
       
   777 	DO_DBL_TEST2("MUL", &TestMulD, 1e20, 1e20);
       
   778 	DO_DBL_TEST2("MUL", &TestMulD, 1e100, 1e300);
       
   779 	DO_DBL_TEST2("MUL", &TestMulD, 1e-20, 1e-20);
       
   780 	DO_DBL_TEST2("MUL", &TestMulD, 1e-200, 1e-300);
       
   781 
       
   782 	// DIV
       
   783 	DO_DBL_TEST2("DIV", &TestDivD, 0.0, 1.0);
       
   784 	DO_DBL_TEST2("DIV", &TestDivD, 1.0, 5.0);
       
   785 	DO_DBL_TEST2("DIV", &TestDivD, 1.0, -5.0);
       
   786 	DO_DBL_TEST2("DIV", &TestDivD, -1.0, 5.0);
       
   787 	DO_DBL_TEST2("DIV", &TestDivD, -1.0, -5.0);
       
   788 	DO_DBL_TEST2("DIV", &TestDivD, 7.3890561, 2.7182818);
       
   789 	DO_DBL_TEST2("DIV", &TestDivD, 1e20, 1e-20);
       
   790 	DO_DBL_TEST2("DIV", &TestDivD, 1e-20, 1e20);
       
   791 	DO_DBL_TEST2("DIV", &TestDivD, 1e-50, 1e300);
       
   792 
       
   793 	// NMUL
       
   794 	DO_DBL_TEST2("NMUL", &TestNMulD, 0.0, 0.0);
       
   795 	DO_DBL_TEST2("NMUL", &TestNMulD, 1.0, 0.0);
       
   796 	DO_DBL_TEST2("NMUL", &TestNMulD, 0.0, 1.0);
       
   797 	DO_DBL_TEST2("NMUL", &TestNMulD, 2.5, 6.5);
       
   798 	DO_DBL_TEST2("NMUL", &TestNMulD, -39.6, 19.72);
       
   799 	DO_DBL_TEST2("NMUL", &TestNMulD, -10.1, -20.1);
       
   800 	DO_DBL_TEST2("NMUL", &TestNMulD, 1e20, 1e20);
       
   801 	DO_DBL_TEST2("NMUL", &TestNMulD, 1e100, 1e300);
       
   802 	DO_DBL_TEST2("NMUL", &TestNMulD, 1e-20, 1e-20);
       
   803 	DO_DBL_TEST2("NMUL", &TestNMulD, 1e-200, 1e-300);
       
   804 
       
   805 	// SQRT
       
   806 	DO_DBL_TEST1("SQRT", &TestSqrtD, 0.0);
       
   807 	DO_DBL_TEST1("SQRT", &TestSqrtD, 1.0);
       
   808 	DO_DBL_TEST1("SQRT", &TestSqrtD, 2.0);
       
   809 	DO_DBL_TEST1("SQRT", &TestSqrtD, 3.0);
       
   810 	DO_DBL_TEST1("SQRT", &TestSqrtD, 9096256.0);
       
   811 	DO_DBL_TEST1("SQRT", &TestSqrtD, 1e36);
       
   812 	DO_DBL_TEST1("SQRT", &TestSqrtD, 1e-36);
       
   813 
       
   814 	// MAC
       
   815 	DO_DBL_TEST3("MAC", &TestMacD, 0.0, 0.0, 0.0);
       
   816 	DO_DBL_TEST3("MAC", &TestMacD, 0.0, 1.0, 0.0);
       
   817 	DO_DBL_TEST3("MAC", &TestMacD, 0.0, 1.0, 1.0);
       
   818 	DO_DBL_TEST3("MAC", &TestMacD, -1.0, 1.0, 1.0);
       
   819 	DO_DBL_TEST3("MAC", &TestMacD, 0.8, 0.1, 8.0);
       
   820 	DO_DBL_TEST3("MAC", &TestMacD, 0.8, -0.1, 8.0);
       
   821 	DO_DBL_TEST3("MAC", &TestMacD, -0.8, -0.1, -8.0);
       
   822 	DO_DBL_TEST3("MAC", &TestMacD, 0.8, 0.3333333333, 3.1415926536);
       
   823 
       
   824 	// MSC
       
   825 	DO_DBL_TEST3("MSC", &TestMscD, 0.0, 0.0, 0.0);
       
   826 	DO_DBL_TEST3("MSC", &TestMscD, 0.0, 1.0, 0.0);
       
   827 	DO_DBL_TEST3("MSC", &TestMscD, 0.0, 1.0, 1.0);
       
   828 	DO_DBL_TEST3("MSC", &TestMscD, -1.0, 1.0, 1.0);
       
   829 	DO_DBL_TEST3("MSC", &TestMscD, 0.8, 0.1, 8.0);
       
   830 	DO_DBL_TEST3("MSC", &TestMscD, 0.8, -0.1, 8.0);
       
   831 	DO_DBL_TEST3("MSC", &TestMscD, -0.8, -0.1, -8.0);
       
   832 	DO_DBL_TEST3("MSC", &TestMscD, 0.8, 0.3333333333, 3.1415926536);
       
   833 
       
   834 	// NMAC
       
   835 	DO_DBL_TEST3("NMAC", &TestNMacD, 0.0, 0.0, 0.0);
       
   836 	DO_DBL_TEST3("NMAC", &TestNMacD, 0.0, 1.0, 0.0);
       
   837 	DO_DBL_TEST3("NMAC", &TestNMacD, 0.0, 1.0, 1.0);
       
   838 	DO_DBL_TEST3("NMAC", &TestNMacD, -1.0, 1.0, 1.0);
       
   839 	DO_DBL_TEST3("NMAC", &TestNMacD, 0.8, 0.1, 8.0);
       
   840 	DO_DBL_TEST3("NMAC", &TestNMacD, 0.8, -0.1, 8.0);
       
   841 	DO_DBL_TEST3("NMAC", &TestNMacD, -0.8, -0.1, -8.0);
       
   842 	DO_DBL_TEST3("NMAC", &TestNMacD, 0.8, 0.3333333333, 3.1415926536);
       
   843 
       
   844 	// NMSC
       
   845 	DO_DBL_TEST3("NMSC", &TestNMscD, 0.0, 0.0, 0.0);
       
   846 	DO_DBL_TEST3("NMSC", &TestNMscD, 0.0, 1.0, 0.0);
       
   847 	DO_DBL_TEST3("NMSC", &TestNMscD, 0.0, 1.0, 1.0);
       
   848 	DO_DBL_TEST3("NMSC", &TestNMscD, -1.0, 1.0, 1.0);
       
   849 	DO_DBL_TEST3("NMSC", &TestNMscD, 0.8, 0.1, 8.0);
       
   850 	DO_DBL_TEST3("NMSC", &TestNMscD, 0.8, -0.1, 8.0);
       
   851 	DO_DBL_TEST3("NMSC", &TestNMscD, -0.8, -0.1, -8.0);
       
   852 	DO_DBL_TEST3("NMSC", &TestNMscD, 0.8, 0.3333333333, 3.1415926536);
       
   853 	}
       
   854 
       
   855 void DoBounceTests()
       
   856 	{
       
   857 	test.Next(_L("Test denormal handling - single"));
       
   858 	DO_SGL_TEST2("ADD", &TestAddS, 1e-39f, 1e-39f);
       
   859 	test.Next(_L("Test potential underflow - single"));
       
   860 	DO_SGL_TEST2("MUL", &TestMulS, 3.162e-20f, 3.162e-20f);
       
   861 // fails on VFPv2 hardware. ARM's library should be fixed
       
   862 //	test.Next(_L("Test NaN input - single"));
       
   863 //	TReal32 aSingleNaN;
       
   864 //	*((TUint32*)&aSingleNaN) = 0x7F9ABCDE;
       
   865 //	Vfp::SetSReg(aSingleNaN, 1);
       
   866 //	Vfp::SetSReg(aSingleNaN, 2);
       
   867 //	Vfp::AddS();
       
   868 //	TReal32 aSingleResult = Vfp::SReg(0);
       
   869 //	test(*((TUint32*)&aSingleResult) == 0x7FDABCDE);
       
   870 
       
   871 	if (Double)
       
   872 		{
       
   873 		test.Next(_L("Test denormal handling - double"));
       
   874 		DO_DBL_TEST2("ADD", &TestAddD, 3.1234e-322, 3.1234e-322);
       
   875 		test.Next(_L("Test potential underflow - double"));
       
   876 		DO_DBL_TEST2("MUL", &TestMulD, 1.767e-161, 1.767e-161);
       
   877 // fails on VFPv2 hardware. ARM's library should be fixed
       
   878 //		test.Next(_L("Test NaN input - double"));
       
   879 //		TReal64 aDoubleNaN;
       
   880 //		*((TUint64*)&aDoubleNaN) = 0x7FF0123456789ABCll;
       
   881 //		Vfp::SetDReg(aDoubleNaN, 1);
       
   882 //		Vfp::SetDReg(aDoubleNaN, 2);
       
   883 //		Vfp::AddD();
       
   884 //		TReal64 aDoubleResult = Vfp::DReg(0);
       
   885 //		test(*((TUint64*)&aDoubleResult) == 0x7FF8123456789ABC);
       
   886 		}
       
   887 	}
       
   888 
       
   889 void DoRunFastTests()
       
   890 	{
       
   891 	test.Next(_L("Test flushing denormals to zero - single"));
       
   892 	Vfp::SetSReg(1e-39f, 1);
       
   893 	Vfp::SetSReg(1e-39f, 2);
       
   894 	Vfp::AddS();
       
   895 	test(Vfp::SReg(0)==0);
       
   896 
       
   897 	test.Next(_L("Test flushing underflow to zero - single"));
       
   898 	Vfp::SetSReg(3.162e-20f, 1);
       
   899 	Vfp::SetSReg(3.162e-20f, 2);
       
   900 	Vfp::MulS();
       
   901 	test(Vfp::SReg(0)==0);
       
   902 
       
   903 	test.Next(_L("Test default NaNs - single"));
       
   904 	TReal32 aSingleNaN;
       
   905 	*((TUint32*)&aSingleNaN) = 0x7F9ABCDE;
       
   906 	Vfp::SetSReg(aSingleNaN, 1);
       
   907 	Vfp::SetSReg(aSingleNaN, 2);
       
   908 	Vfp::AddS();
       
   909 	TReal32 aSingleResult = Vfp::SReg(0);
       
   910 	test(*((TUint32*)&aSingleResult) == 0x7FC00000);
       
   911 
       
   912 	if (Double)
       
   913 		{
       
   914 		test.Next(_L("Test flushing denormals to zero - double"));
       
   915 		Vfp::SetDReg(3.1234e-322, 1);
       
   916 		Vfp::SetDReg(3.1234e-322, 2);
       
   917 		Vfp::AddD();
       
   918 		test(Vfp::DReg(0)==0);
       
   919 	
       
   920 		test.Next(_L("Test flushing underflow to zero - double"));
       
   921 		Vfp::SetDReg(1.767e-161, 1);
       
   922 		Vfp::SetDReg(1.767e-161, 2);
       
   923 		Vfp::MulD();
       
   924 		test(Vfp::DReg(0)==0);
       
   925 
       
   926 		test.Next(_L("Test default NaNs - double"));
       
   927 		TReal64 aDoubleNaN;
       
   928 		*((TUint64*)&aDoubleNaN) = 0x7FF0123456789ABCll;
       
   929 		Vfp::SetDReg(aDoubleNaN, 1);
       
   930 		Vfp::SetDReg(aDoubleNaN, 2);
       
   931 		Vfp::AddD();
       
   932 		TReal64 aDoubleResult = Vfp::DReg(0);
       
   933 		test(*((TUint64*)&aDoubleResult) == 0x7FF8000000000000ll);
       
   934 		}
       
   935 	}
       
   936 
       
   937 void TestAddSResult(const TReal32 a, const TReal32 b, const TReal32 r)
       
   938 	{
       
   939 	Vfp::SetSReg(a, 1);
       
   940 	Vfp::SetSReg(b, 2);
       
   941 	Vfp::AddS();
       
   942 	test(Vfp::SReg(0) == r);
       
   943 	}
       
   944 
       
   945 void DoRoundingTests()
       
   946 	{
       
   947 	TFloatingPointMode fpmode = IEEEMode ? EFpModeIEEENoExceptions : EFpModeRunFast;
       
   948 	test.Next(_L("Check default rounding to nearest"));
       
   949 	test(User::SetFloatingPointMode(fpmode) == KErrNone);
       
   950 	test.Next(_L("Check nearest-downward"));
       
   951 	TestAddSResult(16777215, 0.49f, 16777215);
       
   952 	test.Next(_L("Check nearest-upward"));
       
   953 	TestAddSResult(16777215, 0.51f, 16777216);
       
   954 	test.Next(_L("Set rounding mode to toward-plus-infinity"));
       
   955 	test(User::SetFloatingPointMode(fpmode, EFpRoundToPlusInfinity) == KErrNone);
       
   956 	test.Next(_L("Check positive rounding goes upward"));
       
   957 	TestAddSResult(16777215, 0.49f, 16777216);
       
   958 	test.Next(_L("Check negative rounding goes upward"));
       
   959 	TestAddSResult(-16777215, -0.51f, -16777215);
       
   960 	test.Next(_L("Set rounding mode to toward-minus-infinity"));
       
   961 	test(User::SetFloatingPointMode(fpmode, EFpRoundToMinusInfinity) == KErrNone);
       
   962 	test.Next(_L("Check positive rounding goes downward"));
       
   963 	TestAddSResult(16777215, 0.51f, 16777215);
       
   964 	test.Next(_L("Check negative rounding goes downward"));
       
   965 	TestAddSResult(-16777215, -0.49f, -16777216);
       
   966 	test.Next(_L("Set rounding mode to toward-zero"));
       
   967 	test(User::SetFloatingPointMode(fpmode, EFpRoundToZero) == KErrNone);
       
   968 	test.Next(_L("Check positive rounding goes downward"));
       
   969 	TestAddSResult(16777215, 0.51f, 16777215);
       
   970 	test.Next(_L("Check negative rounding goes upward"));
       
   971 	TestAddSResult(-16777215, -0.51f, -16777215);
       
   972 	}
       
   973 
       
   974 TInt RunFastThread(TAny* /*unused*/)
       
   975 	{
       
   976 	Vfp::SetSReg(1e-39f, 1);
       
   977 	Vfp::SetSReg(1e-39f, 2);
       
   978 	Vfp::AddS();
       
   979 	return (Vfp::SReg(0)==0) ? KErrNone : KErrGeneral;
       
   980 	}
       
   981 
       
   982 TInt IEEECompliantThread(TAny* /*unused*/)
       
   983 	{
       
   984 	Vfp::SetSReg(1e-39f, 1);
       
   985 	Vfp::SetSReg(1e-39f, 2);
       
   986 	Vfp::AddS();
       
   987 	return (Vfp::SReg(0)==2e-39f) ? KErrNone : KErrGeneral;
       
   988 	}
       
   989 
       
   990 void TestVFPModeInheritance()
       
   991 	{
       
   992 	test.Printf(_L("Set floating point mode to RunFast\n"));
       
   993 	test(User::SetFloatingPointMode(EFpModeRunFast)==KErrNone);
       
   994 	RThread t;
       
   995 	TInt r = t.Create(KNullDesC, &RunFastThread, 0x1000, NULL, NULL);
       
   996 	test(r==KErrNone);
       
   997 	TRequestStatus s;
       
   998 	t.Logon(s);
       
   999 	test.Printf(_L("Run RunFast test in another thread...\n"));
       
  1000 	t.Resume();
       
  1001 	test.Printf(_L("Wait for other thread to terminate\n"));
       
  1002 	User::WaitForRequest(s);
       
  1003 	test(t.ExitType() == EExitKill);
       
  1004 	test(s == KErrNone);
       
  1005 	CLOSE_AND_WAIT(t);
       
  1006 	test.Printf(_L("Set floating point mode to IEEE\n"));
       
  1007 	test(User::SetFloatingPointMode(EFpModeIEEENoExceptions)==KErrNone);
       
  1008 	r = t.Create(KNullDesC, &IEEECompliantThread, 0x1000, NULL, NULL);
       
  1009 	test(r==KErrNone);
       
  1010 	t.Logon(s);
       
  1011 	test.Printf(_L("Run IEEE test in another thread...\n"));
       
  1012 	t.Resume();
       
  1013 	test.Printf(_L("Wait for other thread to terminate\n"));
       
  1014 	User::WaitForRequest(s);
       
  1015 	test(t.ExitType() == EExitKill);
       
  1016 	test(s == KErrNone);
       
  1017 	CLOSE_AND_WAIT(t);
       
  1018 	}
       
  1019 
       
  1020 
       
  1021 void TestVFPv3()
       
  1022 	{
       
  1023 	test.Next(_L("Transferring to and from fixed point"));
       
  1024 	
       
  1025 	Vfp::SetSReg(2.5f, 0);
       
  1026 	test(Vfp::SReg(0)==2.5f);
       
  1027 	Vfp::ToFixedS(3);				// Convert to fixed (3) precision
       
  1028 	test(Vfp::SRegInt(0)==0x14);	// 10.100 in binary fixed(3) format
       
  1029 	Vfp::FromFixedS(3);				//Convert from fixed (3) precision
       
  1030 	test(Vfp::SReg(0)==2.5f);
       
  1031 
       
  1032 	
       
  1033 	test.Next(_L("Setting immediate value to floating point registers"));
       
  1034 	
       
  1035 	Vfp::SetSReg(5.0f, 0);
       
  1036 	test(Vfp::SReg(0) == 5.0f);
       
  1037 	Vfp::TconstS2();
       
  1038 	test(Vfp::SReg(0) == 2.0f);
       
  1039 	Vfp::SetSReg(5.0f, 0);
       
  1040 	Vfp::TconstS2_8();
       
  1041 	test(Vfp::SReg(0) == 2.875f);
       
  1042 	
       
  1043 	Vfp::SetDReg(5.0f, 0);
       
  1044 	test(Vfp::DReg(0) == 5.0f);
       
  1045 	Vfp::TconstD2();
       
  1046 	test(Vfp::DReg(0) == 2.0f);
       
  1047 	Vfp::TconstD2_8();
       
  1048 	test(Vfp::DReg(0) == 2.875f);
       
  1049 	}
       
  1050 
       
  1051 void TestNEON()
       
  1052 	{
       
  1053 	RThread t;
       
  1054 	TRequestStatus s;
       
  1055 	test.Next(_L("Test creating a thread to execute an F2-prefix instruction"));
       
  1056 	test_KErrNone(t.Create(KNullDesC, &NeonWithF2, 0x1000, NULL, NULL));
       
  1057 	t.Logon(s);
       
  1058 	t.Resume();
       
  1059 	User::WaitForRequest(s);
       
  1060 	test(t.ExitType() == EExitKill);
       
  1061 	test(s == KErrNone);
       
  1062 	t.Close();
       
  1063 	test.Next(_L("Test creating a thread to execute an F3-prefix instruction"));
       
  1064 	test_KErrNone(t.Create(KNullDesC, &NeonWithF3, 0x1000, NULL, NULL));
       
  1065 	t.Logon(s);
       
  1066 	t.Resume();
       
  1067 	User::WaitForRequest(s);
       
  1068 	test(t.ExitType() == EExitKill);
       
  1069 	test(s == KErrNone);
       
  1070 	t.Close();
       
  1071 	test.Next(_L("Test creating a thread to execute an F4x-prefix instruction"));
       
  1072 	test_KErrNone(t.Create(KNullDesC, &NeonWithF4x, 0x1000, NULL, NULL));
       
  1073 	t.Logon(s);
       
  1074 	t.Resume();
       
  1075 	User::WaitForRequest(s);
       
  1076 	test(t.ExitType() == EExitKill);
       
  1077 	test(s == KErrNone);
       
  1078 	t.Close();
       
  1079 	}
       
  1080 
       
  1081 void TestThumb()
       
  1082 	{
       
  1083 	RThread t;
       
  1084 	TRequestStatus s;
       
  1085 	TInt testStep = 0;
       
  1086 	do {
       
  1087 		test_KErrNone(t.Create(KNullDesC, &ThumbMode, 0x1000, NULL, (TAny*)testStep++));
       
  1088 		t.Logon(s);
       
  1089 		t.Resume();
       
  1090 		User::WaitForRequest(s);
       
  1091 		test(s == KErrNone || s == 1);
       
  1092 		test(t.ExitType() == EExitKill);
       
  1093 		t.Close();
       
  1094 		}
       
  1095 	while (s == KErrNone);
       
  1096 
       
  1097 	test(s == 1);
       
  1098 	test(testStep == 7);
       
  1099 	}
       
  1100 
       
  1101 TInt E32Main()
       
  1102 	{
       
  1103 	test.Title();
       
  1104 
       
  1105 	test.Start(_L("Ask HAL if we have hardware floating point"));
       
  1106 
       
  1107 	CPUs = UserSvr::HalFunction(EHalGroupKernel, EKernelHalNumLogicalCpus, 0, 0);
       
  1108 	TInt supportedTypes;
       
  1109 	TInt HalVfp = HAL::Get(HALData::EHardwareFloatingPoint, supportedTypes);
       
  1110 	if (HalVfp == KErrNone) 
       
  1111 		{ 
       
  1112 		if (supportedTypes == EFpTypeVFPv2) 
       
  1113 			{ 
       
  1114 			test.Printf(_L("HAL reports VFPv2\n"));
       
  1115 			} 
       
  1116 		else if (supportedTypes == EFpTypeVFPv3)
       
  1117 			{ 
       
  1118 			test.Printf(_L("HAL reports VFPv3\n"));
       
  1119 			} 
       
  1120 		else if (supportedTypes == EFpTypeVFPv3D16)
       
  1121 			{ 
       
  1122 			test.Printf(_L("HAL reports VFPv3-D16\n"));
       
  1123 			} 
       
  1124 		else
       
  1125 			{
       
  1126 			test.Printf(_L("HAL reports an unknown floating point type\n"));
       
  1127 			test(0);
       
  1128 			}
       
  1129 		} 
       
  1130 	else
       
  1131 		{ 
       
  1132 		test.Printf(_L("HAL reports no VFP support\n"));
       
  1133 		} 
       
  1134 		
       
  1135 	test.Next(_L("Check VFP present"));
       
  1136 	TBool present = DetectVFP();
       
  1137 	if (!present)
       
  1138 		{
       
  1139 		test.Printf(_L("No VFP detected\n"));
       
  1140 		test(HalVfp == KErrNotSupported || 
       
  1141 						((supportedTypes != EFpTypeVFPv2) && 
       
  1142 						(supportedTypes != EFpTypeVFPv3) && 
       
  1143 						(supportedTypes != EFpTypeVFPv3D16))
       
  1144 						);
       
  1145 		test.End();
       
  1146 		return 0;
       
  1147 		}
       
  1148 	
       
  1149 	test.Printf(_L("VFP detected. FPSID = %08x\n"), FPSID);
       
  1150 	test(HalVfp == KErrNone);
       
  1151 
       
  1152 	// Verify that the HAL architecture ID matches the FPSID values
       
  1153 	// ARMv7 redefines some of these bits so the masks are different :(
       
  1154 	if (supportedTypes == EFpTypeVFPv2)
       
  1155 		{
       
  1156 		// assume armv5/6's bit definitions, where 19:16 are the arch version
       
  1157 		// and 20 is the single-precision-only bit
       
  1158 		ArchVersion = (FPSID >> 16) & 0xf;
       
  1159 		test(ArchVersion == ARCH_VERSION_VFPV2);
       
  1160 		Double = !(FPSID & VFP_FPSID_SNG);
       
  1161 		}
       
  1162 	else if (supportedTypes == EFpTypeVFPv3 || supportedTypes == EFpTypeVFPv3D16)
       
  1163 		{
       
  1164 		// assume armv7's bit definitions, where 22:16 are the arch version
       
  1165 		ArchVersion = (FPSID >> 16) & 0x7f;
       
  1166 		test(ArchVersion == ARCH_VERSION_VFPV3_SUBARCH_V2
       
  1167 		  || ArchVersion == ARCH_VERSION_VFPV3_SUBARCH_NULL
       
  1168 		  || ArchVersion == ARCH_VERSION_VFPV3_SUBARCH_V3); 
       
  1169 		// there are bits for this in MVFR0 but ARM implementations should always have it?
       
  1170 		Double = ETrue;
       
  1171 		}
       
  1172 
       
  1173 	if (Double)
       
  1174 		test.Printf(_L("Both single and double precision supported\n"), FPSID);
       
  1175 	else
       
  1176 		test.Printf(_L("Only single precision supported\n"), FPSID);
       
  1177 
       
  1178 	test.Next(_L("Test VFP Initial State"));
       
  1179 	TestVFPInitialState();
       
  1180 
       
  1181 	test.Next(_L("Test setting VFP to IEEE no exceptions mode"));
       
  1182 	IEEEMode = User::SetFloatingPointMode(EFpModeIEEENoExceptions) == KErrNone;
       
  1183 	if (!IEEEMode)
       
  1184 		test.Printf(_L("IEEE no exceptions mode not supported, continuing in RunFast\n"));
       
  1185 
       
  1186 	test.Next(_L("Test VFP calculations - single"));
       
  1187 	DoSglTests();
       
  1188 	if (Double)
       
  1189 		{
       
  1190 		test.Next(_L("Test VFP calculations - double"));
       
  1191 		DoDblTests();
       
  1192 		}
       
  1193 
       
  1194 	test.Next(_L("Test VFP Context Save"));
       
  1195 	TestVFPContextSave();
       
  1196 
       
  1197 	if (IEEEMode)
       
  1198 		{
       
  1199 		test.Next(_L("Test bounce handling"));
       
  1200 		DoBounceTests();
       
  1201 		test.Next(_L("Test bouncing while context switching"));
       
  1202 		DoBounceContextSwitchTests();
       
  1203 		test.Next(_L("Test setting VFP to RunFast mode"));
       
  1204 		test(User::SetFloatingPointMode(EFpModeRunFast) == KErrNone);
       
  1205 		DoRunFastTests();
       
  1206 		}
       
  1207 
       
  1208 	test.Next(_L("Test VFP rounding modes"));
       
  1209 	DoRoundingTests();
       
  1210 
       
  1211 	if (IEEEMode)
       
  1212 		{
       
  1213 		test.Next(_L("Test VFP mode inheritance between threads"));
       
  1214 		TestVFPModeInheritance();
       
  1215 		}
       
  1216 
       
  1217 	if (supportedTypes == EFpTypeVFPv3 || supportedTypes == EFpTypeVFPv3D16)
       
  1218 		{
       
  1219 		test.Next(_L("Test VFPv3"));
       
  1220 		TestVFPv3();
       
  1221 
       
  1222 		if (supportedTypes == EFpTypeVFPv3)
       
  1223 			{
       
  1224 			test.Next(_L("Test NEON"));
       
  1225 			TestNEON();
       
  1226 
       
  1227 #if defined(__SUPPORT_THUMB_INTERWORKING)
       
  1228 			test.Next(_L("Test Thumb Decode"));
       
  1229 			TestThumb();
       
  1230 #endif
       
  1231 			}
       
  1232 		}
       
  1233 
       
  1234 	test.End();
       
  1235 	return 0;
       
  1236 	}