v8/test/cctest/test-macro-assembler-mips64.cc
ivica.bogosavljevic 2c63060f11 MIPS64: r6 compact branch optimization.
Several ports to enable r6 compact branch optimizations on MIPS64

Port 3573d3cb58

Original commit message:
MIPS: r6 compact branch optimization.

Port bddf8c9e08

Original commit message:
MIPS: Fix trampoline pool handling in MacroAssembler::BranchShort()

Port 6993cd0de5

Original commit message:
MIPS: Fix 'MIPS:r6 compact branch optimization.'

Jic and jialc compact branch ops are fixed as they does
not have 'forbidden slot' restriction. Also COP1 branches
(CTI instructions) added to IsForbiddenAfterBranchInstr().

Port bb332195d3

Original commit message:
MIPS: Fix trampoline pool handling in MacroAssembler::BranchShort()

Port c91bcf7192

Original commit message:
MIPS: Fix trampoline pool handling in MacroAssembler::BranchShort()
for r6.

BUG=

Review URL: https://codereview.chromium.org/1534183002

Cr-Commit-Position: refs/heads/master@{#33136}
2016-01-06 13:36:41 +00:00

473 lines
15 KiB
C++

// Copyright 2013 the V8 project authors. All rights reserved.
// Redistribution and use in source and binary forms, with or without
// modification, are permitted provided that the following conditions are
// met:
//
// * Redistributions of source code must retain the above copyright
// notice, this list of conditions and the following disclaimer.
// * Redistributions in binary form must reproduce the above
// copyright notice, this list of conditions and the following
// disclaimer in the documentation and/or other materials provided
// with the distribution.
// * Neither the name of Google Inc. nor the names of its
// contributors may be used to endorse or promote products derived
// from this software without specific prior written permission.
//
// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
#include <stdlib.h>
#include <iostream> // NOLINT(readability/streams)
#include "src/v8.h"
#include "test/cctest/cctest.h"
#include "src/base/utils/random-number-generator.h"
#include "src/macro-assembler.h"
#include "src/mips64/macro-assembler-mips64.h"
#include "src/mips64/simulator-mips64.h"
using namespace v8::internal;
typedef void* (*F)(int64_t x, int64_t y, int p2, int p3, int p4);
typedef Object* (*F1)(int x, int p1, int p2, int p3, int p4);
#define __ masm->
static byte to_non_zero(int n) {
return static_cast<unsigned>(n) % 255 + 1;
}
static bool all_zeroes(const byte* beg, const byte* end) {
CHECK(beg);
CHECK(beg <= end);
while (beg < end) {
if (*beg++ != 0)
return false;
}
return true;
}
TEST(CopyBytes) {
CcTest::InitializeVM();
Isolate* isolate = CcTest::i_isolate();
HandleScope handles(isolate);
const int data_size = 1 * KB;
size_t act_size;
// Allocate two blocks to copy data between.
byte* src_buffer =
static_cast<byte*>(v8::base::OS::Allocate(data_size, &act_size, 0));
CHECK(src_buffer);
CHECK(act_size >= static_cast<size_t>(data_size));
byte* dest_buffer =
static_cast<byte*>(v8::base::OS::Allocate(data_size, &act_size, 0));
CHECK(dest_buffer);
CHECK(act_size >= static_cast<size_t>(data_size));
// Storage for a0 and a1.
byte* a0_;
byte* a1_;
MacroAssembler assembler(isolate, NULL, 0,
v8::internal::CodeObjectRequired::kYes);
MacroAssembler* masm = &assembler;
// Code to be generated: The stuff in CopyBytes followed by a store of a0 and
// a1, respectively.
__ CopyBytes(a0, a1, a2, a3);
__ li(a2, Operand(reinterpret_cast<int64_t>(&a0_)));
__ li(a3, Operand(reinterpret_cast<int64_t>(&a1_)));
__ sd(a0, MemOperand(a2));
__ jr(ra);
__ sd(a1, MemOperand(a3));
CodeDesc desc;
masm->GetCode(&desc);
Handle<Code> code = isolate->factory()->NewCode(
desc, Code::ComputeFlags(Code::STUB), Handle<Code>());
::F f = FUNCTION_CAST< ::F>(code->entry());
// Initialise source data with non-zero bytes.
for (int i = 0; i < data_size; i++) {
src_buffer[i] = to_non_zero(i);
}
const int fuzz = 11;
for (int size = 0; size < 600; size++) {
for (const byte* src = src_buffer; src < src_buffer + fuzz; src++) {
for (byte* dest = dest_buffer; dest < dest_buffer + fuzz; dest++) {
memset(dest_buffer, 0, data_size);
CHECK(dest + size < dest_buffer + data_size);
(void)CALL_GENERATED_CODE(isolate, f, reinterpret_cast<int64_t>(src),
reinterpret_cast<int64_t>(dest), size, 0, 0);
// a0 and a1 should point at the first byte after the copied data.
CHECK_EQ(src + size, a0_);
CHECK_EQ(dest + size, a1_);
// Check that we haven't written outside the target area.
CHECK(all_zeroes(dest_buffer, dest));
CHECK(all_zeroes(dest + size, dest_buffer + data_size));
// Check the target area.
CHECK_EQ(0, memcmp(src, dest, size));
}
}
}
// Check that the source data hasn't been clobbered.
for (int i = 0; i < data_size; i++) {
CHECK(src_buffer[i] == to_non_zero(i));
}
}
TEST(LoadConstants) {
CcTest::InitializeVM();
Isolate* isolate = CcTest::i_isolate();
HandleScope handles(isolate);
int64_t refConstants[64];
int64_t result[64];
int64_t mask = 1;
for (int i = 0; i < 64; i++) {
refConstants[i] = ~(mask << i);
}
MacroAssembler assembler(isolate, NULL, 0,
v8::internal::CodeObjectRequired::kYes);
MacroAssembler* masm = &assembler;
__ mov(a4, a0);
for (int i = 0; i < 64; i++) {
// Load constant.
__ li(a5, Operand(refConstants[i]));
__ sd(a5, MemOperand(a4));
__ Daddu(a4, a4, Operand(kPointerSize));
}
__ jr(ra);
__ nop();
CodeDesc desc;
masm->GetCode(&desc);
Handle<Code> code = isolate->factory()->NewCode(
desc, Code::ComputeFlags(Code::STUB), Handle<Code>());
::F f = FUNCTION_CAST< ::F>(code->entry());
(void)CALL_GENERATED_CODE(isolate, f, reinterpret_cast<int64_t>(result), 0, 0,
0, 0);
// Check results.
for (int i = 0; i < 64; i++) {
CHECK(refConstants[i] == result[i]);
}
}
TEST(LoadAddress) {
CcTest::InitializeVM();
Isolate* isolate = CcTest::i_isolate();
HandleScope handles(isolate);
MacroAssembler assembler(isolate, NULL, 0,
v8::internal::CodeObjectRequired::kYes);
MacroAssembler* masm = &assembler;
Label to_jump, skip;
__ mov(a4, a0);
__ Branch(&skip);
__ bind(&to_jump);
__ nop();
__ nop();
__ jr(ra);
__ nop();
__ bind(&skip);
__ li(a4, Operand(masm->jump_address(&to_jump)), ADDRESS_LOAD);
int check_size = masm->InstructionsGeneratedSince(&skip);
CHECK_EQ(check_size, 4);
__ jr(a4);
__ nop();
__ stop("invalid");
__ stop("invalid");
__ stop("invalid");
__ stop("invalid");
__ stop("invalid");
CodeDesc desc;
masm->GetCode(&desc);
Handle<Code> code = isolate->factory()->NewCode(
desc, Code::ComputeFlags(Code::STUB), Handle<Code>());
::F f = FUNCTION_CAST< ::F>(code->entry());
(void)CALL_GENERATED_CODE(isolate, f, 0, 0, 0, 0, 0);
// Check results.
}
TEST(jump_tables4) {
// Similar to test-assembler-mips jump_tables1, with extra test for branch
// trampoline required before emission of the dd table (where trampolines are
// blocked), and proper transition to long-branch mode.
// Regression test for v8:4294.
CcTest::InitializeVM();
Isolate* isolate = CcTest::i_isolate();
HandleScope scope(isolate);
MacroAssembler assembler(isolate, NULL, 0,
v8::internal::CodeObjectRequired::kYes);
MacroAssembler* masm = &assembler;
const int kNumCases = 512;
int values[kNumCases];
isolate->random_number_generator()->NextBytes(values, sizeof(values));
Label labels[kNumCases];
Label near_start, end;
__ daddiu(sp, sp, -8);
__ sd(ra, MemOperand(sp));
__ mov(v0, zero_reg);
__ Branch(&end);
__ bind(&near_start);
// Generate slightly less than 32K instructions, which will soon require
// trampoline for branch distance fixup.
for (int i = 0; i < 32768 - 256; ++i) {
__ addiu(v0, v0, 1);
}
__ Align(8);
Label done;
{
__ BlockTrampolinePoolFor(kNumCases * 2 + 6);
PredictableCodeSizeScope predictable(
masm, (kNumCases * 2 + 6) * Assembler::kInstrSize);
Label here;
__ bal(&here);
__ dsll(at, a0, 3); // In delay slot.
__ bind(&here);
__ daddu(at, at, ra);
__ ld(at, MemOperand(at, 4 * Assembler::kInstrSize));
__ jr(at);
__ nop(); // Branch delay slot nop.
for (int i = 0; i < kNumCases; ++i) {
__ dd(&labels[i]);
}
}
for (int i = 0; i < kNumCases; ++i) {
__ bind(&labels[i]);
__ lui(v0, (values[i] >> 16) & 0xffff);
__ ori(v0, v0, values[i] & 0xffff);
__ Branch(&done);
}
__ bind(&done);
__ ld(ra, MemOperand(sp));
__ daddiu(sp, sp, 8);
__ jr(ra);
__ nop();
__ bind(&end);
__ Branch(&near_start);
CodeDesc desc;
masm->GetCode(&desc);
Handle<Code> code = isolate->factory()->NewCode(
desc, Code::ComputeFlags(Code::STUB), Handle<Code>());
#ifdef OBJECT_PRINT
code->Print(std::cout);
#endif
F1 f = FUNCTION_CAST<F1>(code->entry());
for (int i = 0; i < kNumCases; ++i) {
int64_t res = reinterpret_cast<int64_t>(
CALL_GENERATED_CODE(isolate, f, i, 0, 0, 0, 0));
::printf("f(%d) = %" PRId64 "\n", i, res);
CHECK_EQ(values[i], res);
}
}
static uint64_t run_lsa(uint32_t rt, uint32_t rs, int8_t sa) {
Isolate* isolate = CcTest::i_isolate();
HandleScope scope(isolate);
MacroAssembler assembler(isolate, nullptr, 0,
v8::internal::CodeObjectRequired::kYes);
MacroAssembler* masm = &assembler;
__ Lsa(v0, a0, a1, sa);
__ jr(ra);
__ nop();
CodeDesc desc;
assembler.GetCode(&desc);
Handle<Code> code = isolate->factory()->NewCode(
desc, Code::ComputeFlags(Code::STUB), Handle<Code>());
F1 f = FUNCTION_CAST<F1>(code->entry());
uint64_t res = reinterpret_cast<uint64_t>(
CALL_GENERATED_CODE(isolate, f, rt, rs, 0, 0, 0));
return res;
}
TEST(Lsa) {
CcTest::InitializeVM();
struct TestCaseLsa {
int32_t rt;
int32_t rs;
uint8_t sa;
uint64_t expected_res;
};
struct TestCaseLsa tc[] = {// rt, rs, sa, expected_res
{0x4, 0x1, 1, 0x6},
{0x4, 0x1, 2, 0x8},
{0x4, 0x1, 3, 0xc},
{0x4, 0x1, 4, 0x14},
{0x4, 0x1, 5, 0x24},
{0x0, 0x1, 1, 0x2},
{0x0, 0x1, 2, 0x4},
{0x0, 0x1, 3, 0x8},
{0x0, 0x1, 4, 0x10},
{0x0, 0x1, 5, 0x20},
{0x4, 0x0, 1, 0x4},
{0x4, 0x0, 2, 0x4},
{0x4, 0x0, 3, 0x4},
{0x4, 0x0, 4, 0x4},
{0x4, 0x0, 5, 0x4},
// Shift overflow.
{0x4, INT32_MAX, 1, 0x2},
{0x4, INT32_MAX >> 1, 2, 0x0},
{0x4, INT32_MAX >> 2, 3, 0xfffffffffffffffc},
{0x4, INT32_MAX >> 3, 4, 0xfffffffffffffff4},
{0x4, INT32_MAX >> 4, 5, 0xffffffffffffffe4},
// Signed addition overflow.
{INT32_MAX - 1, 0x1, 1, 0xffffffff80000000},
{INT32_MAX - 3, 0x1, 2, 0xffffffff80000000},
{INT32_MAX - 7, 0x1, 3, 0xffffffff80000000},
{INT32_MAX - 15, 0x1, 4, 0xffffffff80000000},
{INT32_MAX - 31, 0x1, 5, 0xffffffff80000000},
// Addition overflow.
{-2, 0x1, 1, 0x0},
{-4, 0x1, 2, 0x0},
{-8, 0x1, 3, 0x0},
{-16, 0x1, 4, 0x0},
{-32, 0x1, 5, 0x0}};
size_t nr_test_cases = sizeof(tc) / sizeof(TestCaseLsa);
for (size_t i = 0; i < nr_test_cases; ++i) {
uint64_t res = run_lsa(tc[i].rt, tc[i].rs, tc[i].sa);
PrintF("0x%" PRIx64 " =? 0x%" PRIx64 " == Lsa(v0, %x, %x, %hhu)\n",
tc[i].expected_res, res, tc[i].rt, tc[i].rs, tc[i].sa);
CHECK_EQ(tc[i].expected_res, res);
}
}
static uint64_t run_dlsa(uint64_t rt, uint64_t rs, int8_t sa) {
Isolate* isolate = CcTest::i_isolate();
HandleScope scope(isolate);
MacroAssembler assembler(isolate, nullptr, 0,
v8::internal::CodeObjectRequired::kYes);
MacroAssembler* masm = &assembler;
__ Dlsa(v0, a0, a1, sa);
__ jr(ra);
__ nop();
CodeDesc desc;
assembler.GetCode(&desc);
Handle<Code> code = isolate->factory()->NewCode(
desc, Code::ComputeFlags(Code::STUB), Handle<Code>());
::F f = FUNCTION_CAST<::F>(code->entry());
uint64_t res = reinterpret_cast<uint64_t>(
CALL_GENERATED_CODE(isolate, f, rt, rs, 0, 0, 0));
return res;
}
TEST(Dlsa) {
CcTest::InitializeVM();
struct TestCaseLsa {
int64_t rt;
int64_t rs;
uint8_t sa;
uint64_t expected_res;
};
struct TestCaseLsa tc[] = {// rt, rs, sa, expected_res
{0x4, 0x1, 1, 0x6},
{0x4, 0x1, 2, 0x8},
{0x4, 0x1, 3, 0xc},
{0x4, 0x1, 4, 0x14},
{0x4, 0x1, 5, 0x24},
{0x0, 0x1, 1, 0x2},
{0x0, 0x1, 2, 0x4},
{0x0, 0x1, 3, 0x8},
{0x0, 0x1, 4, 0x10},
{0x0, 0x1, 5, 0x20},
{0x4, 0x0, 1, 0x4},
{0x4, 0x0, 2, 0x4},
{0x4, 0x0, 3, 0x4},
{0x4, 0x0, 4, 0x4},
{0x4, 0x0, 5, 0x4},
// Shift overflow.
{0x4, INT64_MAX, 1, 0x2},
{0x4, INT64_MAX >> 1, 2, 0x0},
{0x4, INT64_MAX >> 2, 3, 0xfffffffffffffffc},
{0x4, INT64_MAX >> 3, 4, 0xfffffffffffffff4},
{0x4, INT64_MAX >> 4, 5, 0xffffffffffffffe4},
// Signed addition overflow.
{INT64_MAX - 1, 0x1, 1, 0x8000000000000000},
{INT64_MAX - 3, 0x1, 2, 0x8000000000000000},
{INT64_MAX - 7, 0x1, 3, 0x8000000000000000},
{INT64_MAX - 15, 0x1, 4, 0x8000000000000000},
{INT64_MAX - 31, 0x1, 5, 0x8000000000000000},
// Addition overflow.
{-2, 0x1, 1, 0x0},
{-4, 0x1, 2, 0x0},
{-8, 0x1, 3, 0x0},
{-16, 0x1, 4, 0x0},
{-32, 0x1, 5, 0x0}};
size_t nr_test_cases = sizeof(tc) / sizeof(TestCaseLsa);
for (size_t i = 0; i < nr_test_cases; ++i) {
uint64_t res = run_dlsa(tc[i].rt, tc[i].rs, tc[i].sa);
PrintF("0x%" PRIx64 " =? 0x%" PRIx64 " == Dlsa(v0, %" PRIx64 ", %" PRIx64
", %hhu)\n",
tc[i].expected_res, res, tc[i].rt, tc[i].rs, tc[i].sa);
CHECK_EQ(tc[i].expected_res, res);
}
}
#undef __