Patch implements encoding for instructions used for floating point number comparison.

R=stichnot@chromium.org

Review URL: https://codereview.chromium.org/2350833002 .

Patch from Srdjan Obucina <Srdjan.Obucina@imgtec.com>.

The patch emits constant pool for float and double constants.

R=stichnot@chromium.org

Review URL: https://codereview.chromium.org/2351583002 .

Patch from Jaydeep Patil <jaydeep.patil@imgtec.com>.

Patch implements lowerUnreachable and encoding for teq
instruction. To avoid duplicated code, class describing
trap instruction is borrowed from
https://codereview.chromium.org/2339323004/

Review URL: https://codereview.chromium.org/2350903002 .

Patch from Srdjan Obucina <Srdjan.Obucina@imgtec.com>.

Use DIV/DIVU instructions provided by MIPS32 ISA instead of calling target
helper function (__divsi3 etc.). These instructions give 32-bit quotient and
remainder in 32-bit special LO/HI registers respectively. An additional
instructions to check for divide-by-zero (Trap if equal) is emitted after the
DIV/DIVU instructions.

R=stichnot@chromium.org

Review URL: https://codereview.chromium.org/2339323004 .

Patch from Jaydeep Patil <jaydeep.patil@imgtec.com>.

BUG=swiftshader:9

Change-Id: Ie58412c13991143c1ee39f3a122475bf93ead242
Reviewed-on: https://chromium-review.googlesource.com/385117Reviewed-by: Jim Stichnoth <stichnot@chromium.org>
Tested-by: Nicolas Capens <nicolascapens@google.com>

Previously all unsandboxed 64-bit code was assumed to use ILP32 and
be stored in ELF32 format using the x32 ABI.

BUG=swiftshader:9

Change-Id: I2476a09d1f0af60b1ac6f8807ee9ed37d54a99d4
Reviewed-on: https://chromium-review.googlesource.com/385277Reviewed-by: Jim Stichnoth <stichnot@chromium.org>
Tested-by: Nicolas Capens <nicolascapens@google.com>

Patch implements improvements and instruction encodings for many COP1 instructions for handling floating point values.

Patch covers load, store, basic arithmetic, data movement for FPR<->FPR, GPR<->FPR, FPR<->GPR, and format conversion instructinos.

Added instruction encodings:
Load: lb, lh, lwc1, ldc1
Store: sb, sh, swc1, sdc1
FP arith: abs_d, abs_s, add_d, add_s, div_d, div_s, mul_d, mul_s, sqrt_d, sqrt_s, sub_d, sub_s
FP movs: mfc1, mov_d, mov_s, movn_d, movn_s, movz_d, movz_s, mtc1
Conversion: cvt_d_l, cvt_d_s, cvt_d_w, cvt_s_d, cvt_s_l, cvt_s_w, trunc_l_d, trunc_l_s, trunc_w_d, trunc_w_s

R=stichnot@chromium.org

Review URL: https://codereview.chromium.org/2341713003 .

Patch from Srdjan Obucina <Srdjan.Obucina@imgtec.com>.

With fix related to stack alignment bytes increasing, it is
possible to return mips tests from nacl-other-intrinsics-mips
to its original place. However, with existing vector test, O2 test
had to be turned off. This does not affect anything imortant,
because it only tested one case (test_sqrt_ignored).

R=stichnot@chromium.org

Review URL: https://codereview.chromium.org/2342083003 .

Patch from Srdjan Obucina <Srdjan.Obucina@imgtec.com>.

This enables other implementations, such as streaming to memory
instead of a file.

BUG=swiftshader:9

Change-Id: I2a780ee67e9bccd157c120b7a0895d9764117464
Reviewed-on: https://chromium-review.googlesource.com/384911Reviewed-by: Jim Stichnoth <stichnot@chromium.org>
Tested-by: Nicolas Capens <nicolascapens@google.com>

Patch implements randomlyInsertNop for MIPS32. Separate test case
added because current test is based on function with vector
operands, which MIPS32 does not support yet.
run-pnacl-sz.py updated, to pass -z argument to objdump, so it does
not hide multiple nops (shows ... instead) between instruction.

R=stichnot@chromium.org

Review URL: https://codereview.chromium.org/2340733003 .

Patch from Srdjan Obucina <Srdjan.Obucina@imgtec.com>.

In production autoconf builds (which are the builds that matter), the SUBZERO_REVISION was being calculated based on git history of the top-level native_client repo, rather than the pnacl-subzero repo.

BUG= none
R=jpp@chromium.org, kschimpf@google.com

Review URL: https://codereview.chromium.org/2338943003 .

The test nacl-other-intrinsics.ll fails with an assertion failure: assert(SpillAreaAlignmentBytes <= MIPS32_STACK_ALIGNMENT_BYTES). This assertion triggers because MIPS32_STACK_ALIGNMENT_BYTES is 8 and SpillAreaAlignmentBytes is 16. The minimum required alignment of stack for MIPS32 is 8 bytes according to the ABI document. Since 16 is aligned to 8, we can safely keep the stack aligned to 16 bytes to fix the test.

R=stichnot@chromium.org

Review URL: https://codereview.chromium.org/2339683003 .

Patch from Sagar Thakur <sagar.thakur@imgtec.com>.

Implements lowerSelect for basic integer and float datatypes.
Support for i64 will be added as separate patch.
Crucial for running crosstests.

R=stichnot@chromium.org

Review URL: https://codereview.chromium.org/2337023003 .

Patch from Srdjan Obucina <Srdjan.Obucina@imgtec.com>.

In a similar fashion as https://codereview.chromium.org/1732233002 .

BUG= none
R=jpp@chromium.org

Review URL: https://codereview.chromium.org/2339843002 .

The patch implements mtc1/mfc1 instructions which are required for GP to/from FP registers moves.
The patch also implements fptosi and sitofp for float and i32 types to test mtc1/mfc1 instructions.

R=stichnot@chromium.org

Review URL: https://codereview.chromium.org/2316933002 .

Patch from Jaydeep Patil <jaydeep.patil@imgtec.com>.

Instructions are allocated using the ArenaAllocator which uses a
memory pool of "slabs", so we can't use the regular C++ delete to
deallocate them. Just remove them from the list. This change also
provides an override for Inst's operator delete to use the custom
allocator, which should currently not be called.

BUG=swiftshader:8

Change-Id: Ibb166910402a70e7d9276b28e19b15caf64422f2
Reviewed-on: https://chromium-review.googlesource.com/384336Tested-by: Nicolas Capens <nicolascapens@google.com>
Reviewed-by: Jim Stichnoth <stichnot@chromium.org>

The patch implements address optimization for MIPS.

R=stichnot@chromium.org

Review URL: https://codereview.chromium.org/2313293002 .

Patch from Jaydeep Patil <jaydeep.patil@imgtec.com>.

This patch implements lowerFcmp, for lowering floating point comparison.

R=stichnot@chromium.org

Review URL: https://codereview.chromium.org/2047043002 .

Patch from Srdjan Obucina <Srdjan.Obucina@imgtec.com>.

The movzx instruction in the lowering sequence should not be allowed to be elided.

BUG= https://bugs.chromium.org/p/nativeclient/issues/detail?id=4379
R=jpp@chromium.org

Review URL: https://codereview.chromium.org/2335933002 .

Patch implements truncation and extension of FP values.

R=stichnot@chromium.org

Review URL: https://codereview.chromium.org/2324903002 .

Patch from Srdjan Obucina <Srdjan.Obucina@imgtec.com>.

This patch adds ABS.fmt and SQRT.fmt instructions for intrinsic
calls. Test cases are copied from nacl-other-intrinsics.ll.

R=stichnot@chromium.org

Review URL: https://codereview.chromium.org/2325703002 .

Patch from Srdjan Obucina <Srdjan.Obucina@imgtec.com>.

genTargetHelperCallFor provides framework for calling Subzero
runtime functions. This patch implements calls and tests for
some of the available functions, but not all. Current
lowerCall implementation does not allow more (Handling of FP
return values is missing).

R=stichnot@chromium.org

Review URL: https://codereview.chromium.org/2317653004 .

Patch from Srdjan Obucina <srdjan.obucina@imgtec.com>.

Adds the right macro voodoo so that when Subzero is built for a single target, e.g.:

$ make -f Makefile.standalone SZTARGET=ARM32

we get output like this:

$ ./pnacl-sz --build-atts
target_ARM32

rather than this:

$ ./pnacl-sz --build-atts
target_SZTARGET

BUG= none
R=eholk@chromium.org

Review URL: https://codereview.chromium.org/2306273002 .

BUG= none
TBR=nicolascapens@chromium.org

Review URL: https://codereview.chromium.org/2328873003 .

MSVC's implementation of the STL allocates memory even for an empty
vector. Since we're using a custom thread-local allocator, it should
be initialized before any STL member containers get initialized.

BUG=swiftshader:7

Change-Id: I4bd977e7ee8eb87006fe08b051cbcfc9bc62342b
Reviewed-on: https://chromium-review.googlesource.com/381531Tested-by: Nicolas Capens <nicolascapens@google.com>
Reviewed-by: Jim Stichnoth <stichnot@chromium.org>

As a global, NoDefinitions could get initialized at program startup,
which happens specifically with Visual Studio. This causes the
progam to abort because its initialization depends on a TLS variable
to be (manually) initialized first. Since there's only one use of
NoDefinitions, it can be moved to that location and since it's at
function scope it only gets constructed at first use.

BUG=swiftshader:7

Change-Id: I30801ad0d0ab380ead33069f174bb78dc1b230ab
Reviewed-on: https://chromium-review.googlesource.com/379955Tested-by: Nicolas Capens <nicolascapens@google.com>
Reviewed-by: Jim Stichnoth <stichnot@chromium.org>

Visual Studio 2015 generates an error for using array_lengthof() on
a template class member array. Since the array content does not
depend on the template types, just make it a global.

BUG=swiftshader:7

Change-Id: If197308413f880484be06e6738f246be144d0ba6
Reviewed-on: https://chromium-review.googlesource.com/380501Reviewed-by: Jim Stichnoth <stichnot@chromium.org>
Tested-by: Nicolas Capens <nicolascapens@google.com>

BUG=swiftshader:7

Change-Id: If4e1f8269343161f907a62f1fccf6a547f80fc1d
Reviewed-on: https://chromium-review.googlesource.com/382371Tested-by: Nicolas Capens <nicolascapens@google.com>
Reviewed-by: Jim Stichnoth <stichnot@chromium.org>

ConstantInteger32::shouldBeRandomizedOrPooled() was defined in a
source file and not resolvable. Move its definition inline.

BUG=swiftshader:7

Change-Id: I693ef93a9986e11b4e24f0194c9260f074567b81
Reviewed-on: https://chromium-review.googlesource.com/380422Reviewed-by: Jim Stichnoth <stichnot@chromium.org>
Tested-by: Nicolas Capens <nicolascapens@google.com>

BUG=swiftshader:7

Change-Id: I0c51c67b6f19a061dfff83415df5e41ea0efd6de
Reviewed-on: https://chromium-review.googlesource.com/380596Reviewed-by: Jim Stichnoth <stichnot@chromium.org>
Tested-by: Nicolas Capens <nicolascapens@google.com>

BUG=swiftshader:7

Change-Id: I0ef3b0ab059bd48e0241eea13a5b77a193351108
Reviewed-on: https://chromium-review.googlesource.com/380275Reviewed-by: Jim Stichnoth <stichnot@chromium.org>
Tested-by: Nicolas Capens <nicolascapens@google.com>

Visual Studio doesn't recognize the use of a class member in a static
constexpr unless prefixed with the class name.

BUG=swiftshader:7

Change-Id: I4ffbaa6fc1a43a7294ecdaf426bd7fc2aab5e469
Reviewed-on: https://chromium-review.googlesource.com/380195Reviewed-by: Jim Stichnoth <stichnot@chromium.org>
Tested-by: Nicolas Capens <nicolascapens@google.com>

With Visual Studio 2015, the Traits type alias is not considered
within scope yet for the constructor's argument list. Instead Use
the template argument directly.

BUG=swiftshader:7

Change-Id: I677f5bed1a4749aba921a23ed5b1466431147069
Reviewed-on: https://chromium-review.googlesource.com/379915Reviewed-by: Jim Stichnoth <stichnot@chromium.org>
Tested-by: Nicolas Capens <nicolascapens@google.com>

BUG=swiftshader:7

Change-Id: If0cfad149ba6ba7bc17708e8a6e5183189eeef5c
Reviewed-on: https://chromium-review.googlesource.com/379876Reviewed-by: Jim Stichnoth <stichnot@chromium.org>
Tested-by: Nicolas Capens <nicolascapens@google.com>

This patch implements lowerLoad and extends existing lowerStore for byte, short and floating-point types.
The patch also modifies PostLoweringLegalizer for conversion of mov to load or store.

R=stichnot@chromium.org

Review URL: https://codereview.chromium.org/2301303003 .

Patch from Jaydeep Patil <jaydeep.patil@imgtec.com>.

Visual Studio 2015 does not support the use of typename outside of a
template definition. Instead alias the flag's storage type. This
even avoids some duplication.

BUG=swiftshader:7

Change-Id: I787701f6bfe8e36169e6ac5d63fcb6a1a114bdb0

There are two problems to address:

1. Google developers lack some tooling to test MIPS changes.

2. MIPS developers lack some tooling/packages to allow full "make presubmit" testing.

Until all this gets sorted out, we add the "make -f Makefile.standalone MIPS=1" option to control some of the testing targets:

check-lit: No changes, as all these lit tests should be runnable in both environments.

check-xtest: MIPS=1 runs just the mips32 tests, plus the x8664 native tests as a sanity-check.  Non-MIPS runs everything except the mips32 tests.

check-spec: MIPS=1 completely disables spec2k testing.

presubmit: MIPS=1 greatly reduces the number of tests (especially since so many are spec variants).

With this change, mips32 CLs should pass "make -f Makefile.standalone MIPS=1 presubmit" before upload, and other CLs should pass "make -f Makefile.standalone presubmit".

BUG= none
R=kschimpf@google.com, obucinac@gmail.com

Review URL: https://codereview.chromium.org/2271053006 .

BUG=swiftshader:7

Change-Id: I1c40c10b3c3d032d3f7e0f8634c7c9abcc21d130

Currently, TargetLowering::sortVarsByAlignment() uses the variable's required alignment as the primary key, but then std::sort() breaks ties arbitrarily.  This can give different results across different STL implementations, for example when building against LLVM trunk versus building in the PNaCl environment.

The fix is to use Variable::Number as the secondary key.

BUG= none
R=kschimpf@google.com

Review URL: https://codereview.chromium.org/2295393002 .