v8/test/cctest/test-assembler-x64.cc

Ignoring revisions in .git-blame-ignore-revs. Click here to bypass and see the normal blame view.

2826 lines
81 KiB
C++
Raw Normal View History

// Copyright 2009 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 <cstdlib>
#include <cstring>
#include <iostream>
Reland "[include] Split out v8.h" This is a reland of d1b27019d3bf86360ea838c317f8505fac6d3a7e Fixes include: Adding missing file to bazel build Forward-declaring classing before friend-classing them to fix win/gcc Add missing v8-isolate.h include for vtune builds Original change's description: > [include] Split out v8.h > > This moves every single class/function out of include/v8.h into a > separate header in include/, which v8.h then includes so that > externally nothing appears to have changed. > > Every include of v8.h from inside v8 has been changed to a more > fine-grained include. > > Previously inline functions defined at the bottom of v8.h would call > private non-inline functions in the V8 class. Since that class is now > in v8-initialization.h and is rarely included (as that would create > dependency cycles), this is not possible and so those methods have been > moved out of the V8 class into the namespace v8::api_internal. > > None of the previous files in include/ now #include v8.h, which means > if embedders were relying on this transitive dependency then it will > give compile failures. > > v8-inspector.h does depend on v8-scripts.h for the time being to ensure > that Chrome continue to compile but that change will be reverted once > those transitive #includes in chrome are changed to include it directly. > > Full design: > https://docs.google.com/document/d/1rTD--I8hCAr-Rho1WTumZzFKaDpEp0IJ8ejZtk4nJdA/edit?usp=sharing > > Bug: v8:11965 > Change-Id: I53b84b29581632710edc80eb11f819c2097a2877 > Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3097448 > Reviewed-by: Yang Guo <yangguo@chromium.org> > Reviewed-by: Camillo Bruni <cbruni@chromium.org> > Reviewed-by: Jakob Kummerow <jkummerow@chromium.org> > Reviewed-by: Leszek Swirski <leszeks@chromium.org> > Reviewed-by: Michael Lippautz <mlippautz@chromium.org> > Commit-Queue: Dan Elphick <delphick@chromium.org> > Cr-Commit-Position: refs/heads/main@{#76424} Cq-Include-Trybots: luci.v8.try:v8_linux_vtunejit Bug: v8:11965 Change-Id: I99f5d3a73bf8fe25b650adfaf9567dc4e44a09e6 Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3113629 Reviewed-by: Leszek Swirski <leszeks@chromium.org> Reviewed-by: Camillo Bruni <cbruni@chromium.org> Reviewed-by: Michael Lippautz <mlippautz@chromium.org> Reviewed-by: Jakob Kummerow <jkummerow@chromium.org> Reviewed-by: Simon Zünd <szuend@chromium.org> Commit-Queue: Dan Elphick <delphick@chromium.org> Cr-Commit-Position: refs/heads/main@{#76460}
2021-08-23 13:01:06 +00:00
#include "include/v8-function.h"
#include "src/base/numbers/double.h"
#include "src/base/platform/platform.h"
#include "src/base/utils/random-number-generator.h"
#include "src/codegen/macro-assembler.h"
#include "src/execution/simulator.h"
#include "src/heap/factory.h"
#include "src/init/v8.h"
#include "src/objects/objects-inl.h"
#include "src/utils/ostreams.h"
#include "test/cctest/cctest.h"
Reland "[turboassembler] Introduce hard-abort mode" This is a reland of a462a7854a081f4f34bb4c112ee33f3d69efa309 Original change's description: > [turboassembler] Introduce hard-abort mode > > For checks and assertions (mostly for debug code, like stack alignment > or zero extension), we had two modes: Emit a call to the {Abort} > runtime function (the default), and emit a debug break (used for > testing, enabled via --trap-on-abort). > In wasm, where we cannot just call a runtime function because code must > be isolate independent, we always used the trap-on-abort behaviour. > This causes problems for our fuzzers, which do not catch SIGTRAP, and > hence do not detect debug code failures. > > This CL introduces a third mode ("hard abort"), which calls a C > function via {ExternalReference}. The C function still outputs the > abort reason, but does not print the stack trace. It then aborts via > "OS::Abort", just like the runtime function. > This will allow fuzzers to detect the crash and even find a nice error > message. > > Even though this looks like a lot of code churn, it is actually not. > Most added lines are new tests, and other changes are minimal. > > R=mstarzinger@chromium.org > > Bug: chromium:863799 > Change-Id: I77c58ff72db552d49014614436259ccfb49ba87b > Reviewed-on: https://chromium-review.googlesource.com/1142163 > Commit-Queue: Clemens Hammacher <clemensh@chromium.org> > Reviewed-by: Michael Starzinger <mstarzinger@chromium.org> > Cr-Commit-Position: refs/heads/master@{#54592} Bug: chromium:863799 Change-Id: I7729a47b4823a982a8e201df36520aa2b6ef5326 Reviewed-on: https://chromium-review.googlesource.com/1146100 Reviewed-by: Michael Starzinger <mstarzinger@chromium.org> Commit-Queue: Clemens Hammacher <clemensh@chromium.org> Cr-Commit-Position: refs/heads/master@{#54656}
2018-07-24 14:12:47 +00:00
#include "test/common/assembler-tester.h"
namespace v8 {
namespace internal {
// Test the x64 assembler by compiling some simple functions into
// a buffer and executing them. These tests do not initialize the
// V8 library, create a context, or use any V8 objects.
// The AMD64 calling convention is used, with the first six arguments
// in RDI, RSI, RDX, RCX, R8, and R9, and floating point arguments in
// the XMM registers. The return value is in RAX.
// This calling convention is used on Linux, with GCC, and on Mac OS,
// with GCC. A different convention is used on 64-bit windows,
// where the first four integer arguments are passed in RCX, RDX, R8 and R9.
using F0 = int();
using F1 = int(int64_t x);
using F2 = int(int64_t x, int64_t y);
using F3 = unsigned(double x);
using F4 = uint64_t(uint64_t* x, uint64_t* y);
using F5 = uint64_t(uint64_t x);
#ifdef _WIN64
static const Register arg1 = rcx;
static const Register arg2 = rdx;
#else
static const Register arg1 = rdi;
static const Register arg2 = rsi;
#endif
Reland "[Memory] Use OS::Allocate for all OS memory allocations." This is a reland of 4899bcb66db1c8c8272b02fb1cedf9886e010f82 This is a reland of b73ee3344ac96f92b4418c3c07779eb95a8541cc Original change's description: > [Memory] Use OS::Allocate for all OS memory allocations. > > - Eliminates OS::ReserveRegion and OS::ReserveAlignedRegion. > - Changes OS::Allocate to take alignment parameter, reorders parameters > to match page_allocator. > - Since the size of memory allocation can be deduced, don't return the > amount of memory allocated. > - Changes reservation of aligned address space. Before we would reserve > (size + alignment) rounded up to page size. This is too much, because > maximum misalignment is (alignment - page_size). > - On Windows and Cygwin, we release an oversize allocation and > immediately retry at the aligned address in the allocation. If we > lose the address due to a race, we just retry. > - Clean up all the calls to OS::Allocate in codegen and tests by adding > helper AllocateSystemPage function (allocation.h) and > AllocateAssemblerBuffer (cctest.h). > - Changes 'assm' to 'masm' in some targets for consistency when using > a macro-assembler. > > - Eliminates OS::ReleaseRegion, replacing with calls to OS::Free. > - Adds bool return value to OS::Free. > - Cleans up types of flags, protection on Windows and Cygwin. > Bug: chromium:756050 > Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng > Change-Id: I306dbe042cc867670fdc935abca29db074b0da71 Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng Change-Id: Iad3c025334e8f8d7d647be99a36a11ee449c9087 Reviewed-on: https://chromium-review.googlesource.com/767014 Commit-Queue: Bill Budge <bbudge@chromium.org> Reviewed-by: Michael Lippautz <mlippautz@chromium.org> Cr-Commit-Position: refs/heads/master@{#49363}
2017-11-14 15:55:09 +00:00
#define __ masm.
TEST(AssemblerX64ReturnOperation) {
CcTest::InitializeVM();
auto buffer = AllocateAssemblerBuffer();
Assembler masm(AssemblerOptions{}, buffer->CreateView());
// Assemble a simple function that copies argument 2 and returns it.
__ movq(rax, arg2);
__ nop();
__ ret(0);
CodeDesc desc;
Reland "[Memory] Use OS::Allocate for all OS memory allocations." This is a reland of 4899bcb66db1c8c8272b02fb1cedf9886e010f82 This is a reland of b73ee3344ac96f92b4418c3c07779eb95a8541cc Original change's description: > [Memory] Use OS::Allocate for all OS memory allocations. > > - Eliminates OS::ReserveRegion and OS::ReserveAlignedRegion. > - Changes OS::Allocate to take alignment parameter, reorders parameters > to match page_allocator. > - Since the size of memory allocation can be deduced, don't return the > amount of memory allocated. > - Changes reservation of aligned address space. Before we would reserve > (size + alignment) rounded up to page size. This is too much, because > maximum misalignment is (alignment - page_size). > - On Windows and Cygwin, we release an oversize allocation and > immediately retry at the aligned address in the allocation. If we > lose the address due to a race, we just retry. > - Clean up all the calls to OS::Allocate in codegen and tests by adding > helper AllocateSystemPage function (allocation.h) and > AllocateAssemblerBuffer (cctest.h). > - Changes 'assm' to 'masm' in some targets for consistency when using > a macro-assembler. > > - Eliminates OS::ReleaseRegion, replacing with calls to OS::Free. > - Adds bool return value to OS::Free. > - Cleans up types of flags, protection on Windows and Cygwin. > Bug: chromium:756050 > Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng > Change-Id: I306dbe042cc867670fdc935abca29db074b0da71 Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng Change-Id: Iad3c025334e8f8d7d647be99a36a11ee449c9087 Reviewed-on: https://chromium-review.googlesource.com/767014 Commit-Queue: Bill Budge <bbudge@chromium.org> Reviewed-by: Michael Lippautz <mlippautz@chromium.org> Cr-Commit-Position: refs/heads/master@{#49363}
2017-11-14 15:55:09 +00:00
masm.GetCode(CcTest::i_isolate(), &desc);
buffer->MakeExecutable();
// Call the function from C++.
auto f = GeneratedCode<F2>::FromBuffer(CcTest::i_isolate(), buffer->start());
int result = f.Call(3, 2);
CHECK_EQ(2, result);
}
TEST(AssemblerX64StackOperations) {
CcTest::InitializeVM();
auto buffer = AllocateAssemblerBuffer();
Assembler masm(AssemblerOptions{}, buffer->CreateView());
// Assemble a simple function that copies argument 2 and returns it.
// We compile without stack frame pointers, so the gdb debugger shows
// incorrect stack frames when debugging this function (which has them).
__ pushq(rbp);
__ movq(rbp, rsp);
__ pushq(arg2); // Value at (rbp - 8)
__ pushq(arg2); // Value at (rbp - 16)
__ pushq(arg1); // Value at (rbp - 24)
__ popq(rax);
__ popq(rax);
__ popq(rax);
__ popq(rbp);
__ nop();
__ ret(0);
CodeDesc desc;
Reland "[Memory] Use OS::Allocate for all OS memory allocations." This is a reland of 4899bcb66db1c8c8272b02fb1cedf9886e010f82 This is a reland of b73ee3344ac96f92b4418c3c07779eb95a8541cc Original change's description: > [Memory] Use OS::Allocate for all OS memory allocations. > > - Eliminates OS::ReserveRegion and OS::ReserveAlignedRegion. > - Changes OS::Allocate to take alignment parameter, reorders parameters > to match page_allocator. > - Since the size of memory allocation can be deduced, don't return the > amount of memory allocated. > - Changes reservation of aligned address space. Before we would reserve > (size + alignment) rounded up to page size. This is too much, because > maximum misalignment is (alignment - page_size). > - On Windows and Cygwin, we release an oversize allocation and > immediately retry at the aligned address in the allocation. If we > lose the address due to a race, we just retry. > - Clean up all the calls to OS::Allocate in codegen and tests by adding > helper AllocateSystemPage function (allocation.h) and > AllocateAssemblerBuffer (cctest.h). > - Changes 'assm' to 'masm' in some targets for consistency when using > a macro-assembler. > > - Eliminates OS::ReleaseRegion, replacing with calls to OS::Free. > - Adds bool return value to OS::Free. > - Cleans up types of flags, protection on Windows and Cygwin. > Bug: chromium:756050 > Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng > Change-Id: I306dbe042cc867670fdc935abca29db074b0da71 Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng Change-Id: Iad3c025334e8f8d7d647be99a36a11ee449c9087 Reviewed-on: https://chromium-review.googlesource.com/767014 Commit-Queue: Bill Budge <bbudge@chromium.org> Reviewed-by: Michael Lippautz <mlippautz@chromium.org> Cr-Commit-Position: refs/heads/master@{#49363}
2017-11-14 15:55:09 +00:00
masm.GetCode(CcTest::i_isolate(), &desc);
buffer->MakeExecutable();
// Call the function from C++.
auto f = GeneratedCode<F2>::FromBuffer(CcTest::i_isolate(), buffer->start());
int result = f.Call(3, 2);
CHECK_EQ(2, result);
}
TEST(AssemblerX64ArithmeticOperations) {
CcTest::InitializeVM();
auto buffer = AllocateAssemblerBuffer();
Assembler masm(AssemblerOptions{}, buffer->CreateView());
// Assemble a simple function that adds arguments returning the sum.
__ movq(rax, arg2);
__ addq(rax, arg1);
__ ret(0);
CodeDesc desc;
Reland "[Memory] Use OS::Allocate for all OS memory allocations." This is a reland of 4899bcb66db1c8c8272b02fb1cedf9886e010f82 This is a reland of b73ee3344ac96f92b4418c3c07779eb95a8541cc Original change's description: > [Memory] Use OS::Allocate for all OS memory allocations. > > - Eliminates OS::ReserveRegion and OS::ReserveAlignedRegion. > - Changes OS::Allocate to take alignment parameter, reorders parameters > to match page_allocator. > - Since the size of memory allocation can be deduced, don't return the > amount of memory allocated. > - Changes reservation of aligned address space. Before we would reserve > (size + alignment) rounded up to page size. This is too much, because > maximum misalignment is (alignment - page_size). > - On Windows and Cygwin, we release an oversize allocation and > immediately retry at the aligned address in the allocation. If we > lose the address due to a race, we just retry. > - Clean up all the calls to OS::Allocate in codegen and tests by adding > helper AllocateSystemPage function (allocation.h) and > AllocateAssemblerBuffer (cctest.h). > - Changes 'assm' to 'masm' in some targets for consistency when using > a macro-assembler. > > - Eliminates OS::ReleaseRegion, replacing with calls to OS::Free. > - Adds bool return value to OS::Free. > - Cleans up types of flags, protection on Windows and Cygwin. > Bug: chromium:756050 > Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng > Change-Id: I306dbe042cc867670fdc935abca29db074b0da71 Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng Change-Id: Iad3c025334e8f8d7d647be99a36a11ee449c9087 Reviewed-on: https://chromium-review.googlesource.com/767014 Commit-Queue: Bill Budge <bbudge@chromium.org> Reviewed-by: Michael Lippautz <mlippautz@chromium.org> Cr-Commit-Position: refs/heads/master@{#49363}
2017-11-14 15:55:09 +00:00
masm.GetCode(CcTest::i_isolate(), &desc);
buffer->MakeExecutable();
// Call the function from C++.
auto f = GeneratedCode<F2>::FromBuffer(CcTest::i_isolate(), buffer->start());
int result = f.Call(3, 2);
CHECK_EQ(5, result);
}
TEST(AssemblerX64CmpbOperation) {
CcTest::InitializeVM();
auto buffer = AllocateAssemblerBuffer();
Assembler masm(AssemblerOptions{}, buffer->CreateView());
// Assemble a function that compare argument byte returing 1 if equal else 0.
// On Windows, it compares rcx with rdx which does not require REX prefix;
// on Linux, it compares rdi with rsi which requires REX prefix.
Label done;
__ movq(rax, Immediate(1));
__ cmpb(arg1, arg2);
__ j(equal, &done);
__ movq(rax, Immediate(0));
__ bind(&done);
__ ret(0);
CodeDesc desc;
Reland "[Memory] Use OS::Allocate for all OS memory allocations." This is a reland of 4899bcb66db1c8c8272b02fb1cedf9886e010f82 This is a reland of b73ee3344ac96f92b4418c3c07779eb95a8541cc Original change's description: > [Memory] Use OS::Allocate for all OS memory allocations. > > - Eliminates OS::ReserveRegion and OS::ReserveAlignedRegion. > - Changes OS::Allocate to take alignment parameter, reorders parameters > to match page_allocator. > - Since the size of memory allocation can be deduced, don't return the > amount of memory allocated. > - Changes reservation of aligned address space. Before we would reserve > (size + alignment) rounded up to page size. This is too much, because > maximum misalignment is (alignment - page_size). > - On Windows and Cygwin, we release an oversize allocation and > immediately retry at the aligned address in the allocation. If we > lose the address due to a race, we just retry. > - Clean up all the calls to OS::Allocate in codegen and tests by adding > helper AllocateSystemPage function (allocation.h) and > AllocateAssemblerBuffer (cctest.h). > - Changes 'assm' to 'masm' in some targets for consistency when using > a macro-assembler. > > - Eliminates OS::ReleaseRegion, replacing with calls to OS::Free. > - Adds bool return value to OS::Free. > - Cleans up types of flags, protection on Windows and Cygwin. > Bug: chromium:756050 > Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng > Change-Id: I306dbe042cc867670fdc935abca29db074b0da71 Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng Change-Id: Iad3c025334e8f8d7d647be99a36a11ee449c9087 Reviewed-on: https://chromium-review.googlesource.com/767014 Commit-Queue: Bill Budge <bbudge@chromium.org> Reviewed-by: Michael Lippautz <mlippautz@chromium.org> Cr-Commit-Position: refs/heads/master@{#49363}
2017-11-14 15:55:09 +00:00
masm.GetCode(CcTest::i_isolate(), &desc);
buffer->MakeExecutable();
// Call the function from C++.
auto f = GeneratedCode<F2>::FromBuffer(CcTest::i_isolate(), buffer->start());
int result = f.Call(0x1002, 0x2002);
CHECK_EQ(1, result);
result = f.Call(0x1002, 0x2003);
CHECK_EQ(0, result);
}
TEST(AssemblerX64ImulOperation) {
CcTest::InitializeVM();
auto buffer = AllocateAssemblerBuffer();
Assembler masm(AssemblerOptions{}, buffer->CreateView());
// Assemble a simple function that multiplies arguments returning the high
// word.
__ movq(rax, arg2);
__ imulq(arg1);
__ movq(rax, rdx);
__ ret(0);
CodeDesc desc;
Reland "[Memory] Use OS::Allocate for all OS memory allocations." This is a reland of 4899bcb66db1c8c8272b02fb1cedf9886e010f82 This is a reland of b73ee3344ac96f92b4418c3c07779eb95a8541cc Original change's description: > [Memory] Use OS::Allocate for all OS memory allocations. > > - Eliminates OS::ReserveRegion and OS::ReserveAlignedRegion. > - Changes OS::Allocate to take alignment parameter, reorders parameters > to match page_allocator. > - Since the size of memory allocation can be deduced, don't return the > amount of memory allocated. > - Changes reservation of aligned address space. Before we would reserve > (size + alignment) rounded up to page size. This is too much, because > maximum misalignment is (alignment - page_size). > - On Windows and Cygwin, we release an oversize allocation and > immediately retry at the aligned address in the allocation. If we > lose the address due to a race, we just retry. > - Clean up all the calls to OS::Allocate in codegen and tests by adding > helper AllocateSystemPage function (allocation.h) and > AllocateAssemblerBuffer (cctest.h). > - Changes 'assm' to 'masm' in some targets for consistency when using > a macro-assembler. > > - Eliminates OS::ReleaseRegion, replacing with calls to OS::Free. > - Adds bool return value to OS::Free. > - Cleans up types of flags, protection on Windows and Cygwin. > Bug: chromium:756050 > Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng > Change-Id: I306dbe042cc867670fdc935abca29db074b0da71 Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng Change-Id: Iad3c025334e8f8d7d647be99a36a11ee449c9087 Reviewed-on: https://chromium-review.googlesource.com/767014 Commit-Queue: Bill Budge <bbudge@chromium.org> Reviewed-by: Michael Lippautz <mlippautz@chromium.org> Cr-Commit-Position: refs/heads/master@{#49363}
2017-11-14 15:55:09 +00:00
masm.GetCode(CcTest::i_isolate(), &desc);
buffer->MakeExecutable();
// Call the function from C++.
auto f = GeneratedCode<F2>::FromBuffer(CcTest::i_isolate(), buffer->start());
int result = f.Call(3, 2);
CHECK_EQ(0, result);
result = f.Call(0x100000000l, 0x100000000l);
CHECK_EQ(1, result);
result = f.Call(-0x100000000l, 0x100000000l);
CHECK_EQ(-1, result);
}
TEST(AssemblerX64testbwqOperation) {
CcTest::InitializeVM();
v8::HandleScope scope(CcTest::isolate());
auto buffer = AllocateAssemblerBuffer();
Assembler masm(AssemblerOptions{}, buffer->CreateView());
__ pushq(rbx);
__ pushq(rdi);
__ pushq(rsi);
__ pushq(r12);
__ pushq(r13);
__ pushq(r14);
__ pushq(r15);
// Assemble a simple function that tests testb and testw
Label bad;
Label done;
// Test immediate testb and testw
__ movq(rax, Immediate(2));
__ movq(rbx, Immediate(4));
__ movq(rcx, Immediate(8));
__ movq(rdx, Immediate(16));
__ movq(rsi, Immediate(32));
__ movq(rdi, Immediate(64));
__ movq(r10, Immediate(128));
__ movq(r11, Immediate(0));
__ movq(r12, Immediate(0));
__ movq(r13, Immediate(0));
__ testb(rax, Immediate(2));
__ j(zero, &bad);
__ testb(rbx, Immediate(4));
__ j(zero, &bad);
__ testb(rcx, Immediate(8));
__ j(zero, &bad);
__ testb(rdx, Immediate(16));
__ j(zero, &bad);
__ testb(rsi, Immediate(32));
__ j(zero, &bad);
__ testb(rdi, Immediate(64));
__ j(zero, &bad);
__ testb(r10, Immediate(128));
__ j(zero, &bad);
__ testw(rax, Immediate(2));
__ j(zero, &bad);
__ testw(rbx, Immediate(4));
__ j(zero, &bad);
__ testw(rcx, Immediate(8));
__ j(zero, &bad);
__ testw(rdx, Immediate(16));
__ j(zero, &bad);
__ testw(rsi, Immediate(32));
__ j(zero, &bad);
__ testw(rdi, Immediate(64));
__ j(zero, &bad);
__ testw(r10, Immediate(128));
__ j(zero, &bad);
// Test reg, reg testb and testw
__ movq(rax, Immediate(2));
__ movq(rbx, Immediate(2));
__ testb(rax, rbx);
__ j(zero, &bad);
__ movq(rbx, Immediate(4));
__ movq(rax, Immediate(4));
__ testb(rbx, rax);
__ j(zero, &bad);
__ movq(rax, Immediate(8));
__ testb(rcx, rax);
__ j(zero, &bad);
__ movq(rax, Immediate(16));
__ testb(rdx, rax);
__ j(zero, &bad);
__ movq(rax, Immediate(32));
__ testb(rsi, rax);
__ j(zero, &bad);
__ movq(rax, Immediate(64));
__ testb(rdi, rax);
__ j(zero, &bad);
__ movq(rax, Immediate(128));
__ testb(r10, rax);
__ j(zero, &bad);
__ movq(rax, Immediate(2));
__ movq(rbx, Immediate(2));
__ testw(rax, rbx);
__ j(zero, &bad);
__ movq(rbx, Immediate(4));
__ movq(rax, Immediate(4));
__ testw(rbx, rax);
__ j(zero, &bad);
__ movq(rax, Immediate(8));
__ testw(rcx, rax);
__ j(zero, &bad);
__ movq(rax, Immediate(16));
__ testw(rdx, rax);
__ j(zero, &bad);
__ movq(rax, Immediate(32));
__ testw(rsi, rax);
__ j(zero, &bad);
__ movq(rax, Immediate(64));
__ testw(rdi, rax);
__ j(zero, &bad);
__ movq(rax, Immediate(128));
__ testw(r10, rax);
__ j(zero, &bad);
// Test diffrrent extended register coding combinations.
__ movq(rax, Immediate(5));
__ movq(r11, Immediate(5));
__ testb(r11, rax);
__ j(zero, &bad);
__ testb(rax, r11);
__ j(zero, &bad);
__ testw(r11, rax);
__ j(zero, &bad);
__ testw(rax, r11);
__ j(zero, &bad);
__ movq(r11, Immediate(3));
__ movq(r12, Immediate(3));
__ movq(rdi, Immediate(3));
__ testb(r12, rdi);
__ j(zero, &bad);
__ testb(rdi, r12);
__ j(zero, &bad);
__ testb(r12, r11);
__ j(zero, &bad);
__ testb(r11, r12);
__ j(zero, &bad);
__ testw(r12, r11);
__ j(zero, &bad);
__ testw(r11, r12);
__ j(zero, &bad);
// Test sign-extended imediate tests
__ movq(r11, Immediate(2));
__ shlq(r11, Immediate(32));
__ testq(r11, Immediate(-1));
__ j(zero, &bad);
// All tests passed
__ movq(rax, Immediate(1));
__ jmp(&done);
__ bind(&bad);
__ movq(rax, Immediate(0));
__ bind(&done);
__ popq(r15);
__ popq(r14);
__ popq(r13);
__ popq(r12);
__ popq(rsi);
__ popq(rdi);
__ popq(rbx);
__ ret(0);
CodeDesc desc;
Reland "[Memory] Use OS::Allocate for all OS memory allocations." This is a reland of 4899bcb66db1c8c8272b02fb1cedf9886e010f82 This is a reland of b73ee3344ac96f92b4418c3c07779eb95a8541cc Original change's description: > [Memory] Use OS::Allocate for all OS memory allocations. > > - Eliminates OS::ReserveRegion and OS::ReserveAlignedRegion. > - Changes OS::Allocate to take alignment parameter, reorders parameters > to match page_allocator. > - Since the size of memory allocation can be deduced, don't return the > amount of memory allocated. > - Changes reservation of aligned address space. Before we would reserve > (size + alignment) rounded up to page size. This is too much, because > maximum misalignment is (alignment - page_size). > - On Windows and Cygwin, we release an oversize allocation and > immediately retry at the aligned address in the allocation. If we > lose the address due to a race, we just retry. > - Clean up all the calls to OS::Allocate in codegen and tests by adding > helper AllocateSystemPage function (allocation.h) and > AllocateAssemblerBuffer (cctest.h). > - Changes 'assm' to 'masm' in some targets for consistency when using > a macro-assembler. > > - Eliminates OS::ReleaseRegion, replacing with calls to OS::Free. > - Adds bool return value to OS::Free. > - Cleans up types of flags, protection on Windows and Cygwin. > Bug: chromium:756050 > Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng > Change-Id: I306dbe042cc867670fdc935abca29db074b0da71 Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng Change-Id: Iad3c025334e8f8d7d647be99a36a11ee449c9087 Reviewed-on: https://chromium-review.googlesource.com/767014 Commit-Queue: Bill Budge <bbudge@chromium.org> Reviewed-by: Michael Lippautz <mlippautz@chromium.org> Cr-Commit-Position: refs/heads/master@{#49363}
2017-11-14 15:55:09 +00:00
masm.GetCode(CcTest::i_isolate(), &desc);
buffer->MakeExecutable();
// Call the function from C++.
auto f = GeneratedCode<F2>::FromBuffer(CcTest::i_isolate(), buffer->start());
int result = f.Call(0, 0);
CHECK_EQ(1, result);
}
TEST(AssemblerX64XchglOperations) {
CcTest::InitializeVM();
auto buffer = AllocateAssemblerBuffer();
Assembler masm(AssemblerOptions{}, buffer->CreateView());
__ movq(rax, Operand(arg1, 0));
__ movq(r11, Operand(arg2, 0));
__ xchgl(rax, r11);
__ movq(Operand(arg1, 0), rax);
__ movq(Operand(arg2, 0), r11);
__ ret(0);
CodeDesc desc;
Reland "[Memory] Use OS::Allocate for all OS memory allocations." This is a reland of 4899bcb66db1c8c8272b02fb1cedf9886e010f82 This is a reland of b73ee3344ac96f92b4418c3c07779eb95a8541cc Original change's description: > [Memory] Use OS::Allocate for all OS memory allocations. > > - Eliminates OS::ReserveRegion and OS::ReserveAlignedRegion. > - Changes OS::Allocate to take alignment parameter, reorders parameters > to match page_allocator. > - Since the size of memory allocation can be deduced, don't return the > amount of memory allocated. > - Changes reservation of aligned address space. Before we would reserve > (size + alignment) rounded up to page size. This is too much, because > maximum misalignment is (alignment - page_size). > - On Windows and Cygwin, we release an oversize allocation and > immediately retry at the aligned address in the allocation. If we > lose the address due to a race, we just retry. > - Clean up all the calls to OS::Allocate in codegen and tests by adding > helper AllocateSystemPage function (allocation.h) and > AllocateAssemblerBuffer (cctest.h). > - Changes 'assm' to 'masm' in some targets for consistency when using > a macro-assembler. > > - Eliminates OS::ReleaseRegion, replacing with calls to OS::Free. > - Adds bool return value to OS::Free. > - Cleans up types of flags, protection on Windows and Cygwin. > Bug: chromium:756050 > Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng > Change-Id: I306dbe042cc867670fdc935abca29db074b0da71 Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng Change-Id: Iad3c025334e8f8d7d647be99a36a11ee449c9087 Reviewed-on: https://chromium-review.googlesource.com/767014 Commit-Queue: Bill Budge <bbudge@chromium.org> Reviewed-by: Michael Lippautz <mlippautz@chromium.org> Cr-Commit-Position: refs/heads/master@{#49363}
2017-11-14 15:55:09 +00:00
masm.GetCode(CcTest::i_isolate(), &desc);
buffer->MakeExecutable();
// Call the function from C++.
uint64_t left = 0x1000'0000'2000'0000;
uint64_t right = 0x3000'0000'4000'0000;
auto f = GeneratedCode<F4>::FromBuffer(CcTest::i_isolate(), buffer->start());
uint64_t result = f.Call(&left, &right);
CHECK_EQ(0x0000'0000'4000'0000, left);
CHECK_EQ(0x0000'0000'2000'0000, right);
USE(result);
}
TEST(AssemblerX64OrlOperations) {
CcTest::InitializeVM();
auto buffer = AllocateAssemblerBuffer();
Assembler masm(AssemblerOptions{}, buffer->CreateView());
__ movq(rax, Operand(arg2, 0));
__ orl(Operand(arg1, 0), rax);
__ ret(0);
CodeDesc desc;
Reland "[Memory] Use OS::Allocate for all OS memory allocations." This is a reland of 4899bcb66db1c8c8272b02fb1cedf9886e010f82 This is a reland of b73ee3344ac96f92b4418c3c07779eb95a8541cc Original change's description: > [Memory] Use OS::Allocate for all OS memory allocations. > > - Eliminates OS::ReserveRegion and OS::ReserveAlignedRegion. > - Changes OS::Allocate to take alignment parameter, reorders parameters > to match page_allocator. > - Since the size of memory allocation can be deduced, don't return the > amount of memory allocated. > - Changes reservation of aligned address space. Before we would reserve > (size + alignment) rounded up to page size. This is too much, because > maximum misalignment is (alignment - page_size). > - On Windows and Cygwin, we release an oversize allocation and > immediately retry at the aligned address in the allocation. If we > lose the address due to a race, we just retry. > - Clean up all the calls to OS::Allocate in codegen and tests by adding > helper AllocateSystemPage function (allocation.h) and > AllocateAssemblerBuffer (cctest.h). > - Changes 'assm' to 'masm' in some targets for consistency when using > a macro-assembler. > > - Eliminates OS::ReleaseRegion, replacing with calls to OS::Free. > - Adds bool return value to OS::Free. > - Cleans up types of flags, protection on Windows and Cygwin. > Bug: chromium:756050 > Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng > Change-Id: I306dbe042cc867670fdc935abca29db074b0da71 Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng Change-Id: Iad3c025334e8f8d7d647be99a36a11ee449c9087 Reviewed-on: https://chromium-review.googlesource.com/767014 Commit-Queue: Bill Budge <bbudge@chromium.org> Reviewed-by: Michael Lippautz <mlippautz@chromium.org> Cr-Commit-Position: refs/heads/master@{#49363}
2017-11-14 15:55:09 +00:00
masm.GetCode(CcTest::i_isolate(), &desc);
buffer->MakeExecutable();
// Call the function from C++.
uint64_t left = 0x1000'0000'2000'0000;
uint64_t right = 0x3000'0000'4000'0000;
auto f = GeneratedCode<F4>::FromBuffer(CcTest::i_isolate(), buffer->start());
uint64_t result = f.Call(&left, &right);
CHECK_EQ(0x1000'0000'6000'0000, left);
USE(result);
}
TEST(AssemblerX64RollOperations) {
CcTest::InitializeVM();
auto buffer = AllocateAssemblerBuffer();
Assembler masm(AssemblerOptions{}, buffer->CreateView());
__ movq(rax, arg1);
__ roll(rax, Immediate(1));
__ ret(0);
CodeDesc desc;
Reland "[Memory] Use OS::Allocate for all OS memory allocations." This is a reland of 4899bcb66db1c8c8272b02fb1cedf9886e010f82 This is a reland of b73ee3344ac96f92b4418c3c07779eb95a8541cc Original change's description: > [Memory] Use OS::Allocate for all OS memory allocations. > > - Eliminates OS::ReserveRegion and OS::ReserveAlignedRegion. > - Changes OS::Allocate to take alignment parameter, reorders parameters > to match page_allocator. > - Since the size of memory allocation can be deduced, don't return the > amount of memory allocated. > - Changes reservation of aligned address space. Before we would reserve > (size + alignment) rounded up to page size. This is too much, because > maximum misalignment is (alignment - page_size). > - On Windows and Cygwin, we release an oversize allocation and > immediately retry at the aligned address in the allocation. If we > lose the address due to a race, we just retry. > - Clean up all the calls to OS::Allocate in codegen and tests by adding > helper AllocateSystemPage function (allocation.h) and > AllocateAssemblerBuffer (cctest.h). > - Changes 'assm' to 'masm' in some targets for consistency when using > a macro-assembler. > > - Eliminates OS::ReleaseRegion, replacing with calls to OS::Free. > - Adds bool return value to OS::Free. > - Cleans up types of flags, protection on Windows and Cygwin. > Bug: chromium:756050 > Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng > Change-Id: I306dbe042cc867670fdc935abca29db074b0da71 Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng Change-Id: Iad3c025334e8f8d7d647be99a36a11ee449c9087 Reviewed-on: https://chromium-review.googlesource.com/767014 Commit-Queue: Bill Budge <bbudge@chromium.org> Reviewed-by: Michael Lippautz <mlippautz@chromium.org> Cr-Commit-Position: refs/heads/master@{#49363}
2017-11-14 15:55:09 +00:00
masm.GetCode(CcTest::i_isolate(), &desc);
buffer->MakeExecutable();
// Call the function from C++.
uint64_t src = 0x1000'0000'C000'0000;
auto f = GeneratedCode<F5>::FromBuffer(CcTest::i_isolate(), buffer->start());
uint64_t result = f.Call(src);
CHECK_EQ(0x0000'0000'8000'0001, result);
}
TEST(AssemblerX64SublOperations) {
CcTest::InitializeVM();
auto buffer = AllocateAssemblerBuffer();
Assembler masm(AssemblerOptions{}, buffer->CreateView());
__ movq(rax, Operand(arg2, 0));
__ subl(Operand(arg1, 0), rax);
__ ret(0);
CodeDesc desc;
Reland "[Memory] Use OS::Allocate for all OS memory allocations." This is a reland of 4899bcb66db1c8c8272b02fb1cedf9886e010f82 This is a reland of b73ee3344ac96f92b4418c3c07779eb95a8541cc Original change's description: > [Memory] Use OS::Allocate for all OS memory allocations. > > - Eliminates OS::ReserveRegion and OS::ReserveAlignedRegion. > - Changes OS::Allocate to take alignment parameter, reorders parameters > to match page_allocator. > - Since the size of memory allocation can be deduced, don't return the > amount of memory allocated. > - Changes reservation of aligned address space. Before we would reserve > (size + alignment) rounded up to page size. This is too much, because > maximum misalignment is (alignment - page_size). > - On Windows and Cygwin, we release an oversize allocation and > immediately retry at the aligned address in the allocation. If we > lose the address due to a race, we just retry. > - Clean up all the calls to OS::Allocate in codegen and tests by adding > helper AllocateSystemPage function (allocation.h) and > AllocateAssemblerBuffer (cctest.h). > - Changes 'assm' to 'masm' in some targets for consistency when using > a macro-assembler. > > - Eliminates OS::ReleaseRegion, replacing with calls to OS::Free. > - Adds bool return value to OS::Free. > - Cleans up types of flags, protection on Windows and Cygwin. > Bug: chromium:756050 > Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng > Change-Id: I306dbe042cc867670fdc935abca29db074b0da71 Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng Change-Id: Iad3c025334e8f8d7d647be99a36a11ee449c9087 Reviewed-on: https://chromium-review.googlesource.com/767014 Commit-Queue: Bill Budge <bbudge@chromium.org> Reviewed-by: Michael Lippautz <mlippautz@chromium.org> Cr-Commit-Position: refs/heads/master@{#49363}
2017-11-14 15:55:09 +00:00
masm.GetCode(CcTest::i_isolate(), &desc);
buffer->MakeExecutable();
// Call the function from C++.
uint64_t left = 0x1000'0000'2000'0000;
uint64_t right = 0x3000'0000'4000'0000;
auto f = GeneratedCode<F4>::FromBuffer(CcTest::i_isolate(), buffer->start());
uint64_t result = f.Call(&left, &right);
CHECK_EQ(0x1000'0000'E000'0000, left);
USE(result);
}
TEST(AssemblerX64TestlOperations) {
CcTest::InitializeVM();
auto buffer = AllocateAssemblerBuffer();
Assembler masm(AssemblerOptions{}, buffer->CreateView());
// Set rax with the ZF flag of the testl instruction.
Label done;
__ movq(rax, Immediate(1));
__ movq(r11, Operand(arg2, 0));
__ testl(Operand(arg1, 0), r11);
__ j(zero, &done, Label::kNear);
__ movq(rax, Immediate(0));
__ bind(&done);
__ ret(0);
CodeDesc desc;
Reland "[Memory] Use OS::Allocate for all OS memory allocations." This is a reland of 4899bcb66db1c8c8272b02fb1cedf9886e010f82 This is a reland of b73ee3344ac96f92b4418c3c07779eb95a8541cc Original change's description: > [Memory] Use OS::Allocate for all OS memory allocations. > > - Eliminates OS::ReserveRegion and OS::ReserveAlignedRegion. > - Changes OS::Allocate to take alignment parameter, reorders parameters > to match page_allocator. > - Since the size of memory allocation can be deduced, don't return the > amount of memory allocated. > - Changes reservation of aligned address space. Before we would reserve > (size + alignment) rounded up to page size. This is too much, because > maximum misalignment is (alignment - page_size). > - On Windows and Cygwin, we release an oversize allocation and > immediately retry at the aligned address in the allocation. If we > lose the address due to a race, we just retry. > - Clean up all the calls to OS::Allocate in codegen and tests by adding > helper AllocateSystemPage function (allocation.h) and > AllocateAssemblerBuffer (cctest.h). > - Changes 'assm' to 'masm' in some targets for consistency when using > a macro-assembler. > > - Eliminates OS::ReleaseRegion, replacing with calls to OS::Free. > - Adds bool return value to OS::Free. > - Cleans up types of flags, protection on Windows and Cygwin. > Bug: chromium:756050 > Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng > Change-Id: I306dbe042cc867670fdc935abca29db074b0da71 Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng Change-Id: Iad3c025334e8f8d7d647be99a36a11ee449c9087 Reviewed-on: https://chromium-review.googlesource.com/767014 Commit-Queue: Bill Budge <bbudge@chromium.org> Reviewed-by: Michael Lippautz <mlippautz@chromium.org> Cr-Commit-Position: refs/heads/master@{#49363}
2017-11-14 15:55:09 +00:00
masm.GetCode(CcTest::i_isolate(), &desc);
buffer->MakeExecutable();
// Call the function from C++.
uint64_t left = 0x1000'0000'2000'0000;
uint64_t right = 0x3000'0000'0000'0000;
auto f = GeneratedCode<F4>::FromBuffer(CcTest::i_isolate(), buffer->start());
uint64_t result = f.Call(&left, &right);
CHECK_EQ(1u, result);
}
TEST(AssemblerX64TestwOperations) {
using F = uint16_t(uint16_t * x);
CcTest::InitializeVM();
auto buffer = AllocateAssemblerBuffer();
Assembler masm(AssemblerOptions{}, buffer->CreateView());
// Set rax with the ZF flag of the testl instruction.
Label done;
__ movq(rax, Immediate(1));
__ testw(Operand(arg1, 0), Immediate(0xF0F0));
__ j(not_zero, &done, Label::kNear);
__ movq(rax, Immediate(0));
__ bind(&done);
__ ret(0);
CodeDesc desc;
Reland "[Memory] Use OS::Allocate for all OS memory allocations." This is a reland of 4899bcb66db1c8c8272b02fb1cedf9886e010f82 This is a reland of b73ee3344ac96f92b4418c3c07779eb95a8541cc Original change's description: > [Memory] Use OS::Allocate for all OS memory allocations. > > - Eliminates OS::ReserveRegion and OS::ReserveAlignedRegion. > - Changes OS::Allocate to take alignment parameter, reorders parameters > to match page_allocator. > - Since the size of memory allocation can be deduced, don't return the > amount of memory allocated. > - Changes reservation of aligned address space. Before we would reserve > (size + alignment) rounded up to page size. This is too much, because > maximum misalignment is (alignment - page_size). > - On Windows and Cygwin, we release an oversize allocation and > immediately retry at the aligned address in the allocation. If we > lose the address due to a race, we just retry. > - Clean up all the calls to OS::Allocate in codegen and tests by adding > helper AllocateSystemPage function (allocation.h) and > AllocateAssemblerBuffer (cctest.h). > - Changes 'assm' to 'masm' in some targets for consistency when using > a macro-assembler. > > - Eliminates OS::ReleaseRegion, replacing with calls to OS::Free. > - Adds bool return value to OS::Free. > - Cleans up types of flags, protection on Windows and Cygwin. > Bug: chromium:756050 > Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng > Change-Id: I306dbe042cc867670fdc935abca29db074b0da71 Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng Change-Id: Iad3c025334e8f8d7d647be99a36a11ee449c9087 Reviewed-on: https://chromium-review.googlesource.com/767014 Commit-Queue: Bill Budge <bbudge@chromium.org> Reviewed-by: Michael Lippautz <mlippautz@chromium.org> Cr-Commit-Position: refs/heads/master@{#49363}
2017-11-14 15:55:09 +00:00
masm.GetCode(CcTest::i_isolate(), &desc);
buffer->MakeExecutable();
// Call the function from C++.
uint16_t operand = 0x8000;
auto f = GeneratedCode<F>::FromBuffer(CcTest::i_isolate(), buffer->start());
uint16_t result = f.Call(&operand);
CHECK_EQ(1u, result);
}
TEST(AssemblerX64XorlOperations) {
CcTest::InitializeVM();
auto buffer = AllocateAssemblerBuffer();
Assembler masm(AssemblerOptions{}, buffer->CreateView());
__ movq(rax, Operand(arg2, 0));
__ xorl(Operand(arg1, 0), rax);
__ ret(0);
CodeDesc desc;
Reland "[Memory] Use OS::Allocate for all OS memory allocations." This is a reland of 4899bcb66db1c8c8272b02fb1cedf9886e010f82 This is a reland of b73ee3344ac96f92b4418c3c07779eb95a8541cc Original change's description: > [Memory] Use OS::Allocate for all OS memory allocations. > > - Eliminates OS::ReserveRegion and OS::ReserveAlignedRegion. > - Changes OS::Allocate to take alignment parameter, reorders parameters > to match page_allocator. > - Since the size of memory allocation can be deduced, don't return the > amount of memory allocated. > - Changes reservation of aligned address space. Before we would reserve > (size + alignment) rounded up to page size. This is too much, because > maximum misalignment is (alignment - page_size). > - On Windows and Cygwin, we release an oversize allocation and > immediately retry at the aligned address in the allocation. If we > lose the address due to a race, we just retry. > - Clean up all the calls to OS::Allocate in codegen and tests by adding > helper AllocateSystemPage function (allocation.h) and > AllocateAssemblerBuffer (cctest.h). > - Changes 'assm' to 'masm' in some targets for consistency when using > a macro-assembler. > > - Eliminates OS::ReleaseRegion, replacing with calls to OS::Free. > - Adds bool return value to OS::Free. > - Cleans up types of flags, protection on Windows and Cygwin. > Bug: chromium:756050 > Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng > Change-Id: I306dbe042cc867670fdc935abca29db074b0da71 Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng Change-Id: Iad3c025334e8f8d7d647be99a36a11ee449c9087 Reviewed-on: https://chromium-review.googlesource.com/767014 Commit-Queue: Bill Budge <bbudge@chromium.org> Reviewed-by: Michael Lippautz <mlippautz@chromium.org> Cr-Commit-Position: refs/heads/master@{#49363}
2017-11-14 15:55:09 +00:00
masm.GetCode(CcTest::i_isolate(), &desc);
buffer->MakeExecutable();
// Call the function from C++.
uint64_t left = 0x1000'0000'2000'0000;
uint64_t right = 0x3000'0000'6000'0000;
auto f = GeneratedCode<F4>::FromBuffer(CcTest::i_isolate(), buffer->start());
uint64_t result = f.Call(&left, &right);
CHECK_EQ(0x1000'0000'4000'0000, left);
USE(result);
}
TEST(AssemblerX64MemoryOperands) {
CcTest::InitializeVM();
auto buffer = AllocateAssemblerBuffer();
Assembler masm(AssemblerOptions{}, buffer->CreateView());
// Assemble a simple function that copies argument 2 and returns it.
__ pushq(rbp);
__ movq(rbp, rsp);
__ pushq(arg2); // Value at (rbp - 8)
__ pushq(arg2); // Value at (rbp - 16)
__ pushq(arg1); // Value at (rbp - 24)
const int kStackElementSize = 8;
__ movq(rax, Operand(rbp, -3 * kStackElementSize));
__ popq(arg2);
__ popq(arg2);
__ popq(arg2);
__ popq(rbp);
__ nop();
__ ret(0);
CodeDesc desc;
Reland "[Memory] Use OS::Allocate for all OS memory allocations." This is a reland of 4899bcb66db1c8c8272b02fb1cedf9886e010f82 This is a reland of b73ee3344ac96f92b4418c3c07779eb95a8541cc Original change's description: > [Memory] Use OS::Allocate for all OS memory allocations. > > - Eliminates OS::ReserveRegion and OS::ReserveAlignedRegion. > - Changes OS::Allocate to take alignment parameter, reorders parameters > to match page_allocator. > - Since the size of memory allocation can be deduced, don't return the > amount of memory allocated. > - Changes reservation of aligned address space. Before we would reserve > (size + alignment) rounded up to page size. This is too much, because > maximum misalignment is (alignment - page_size). > - On Windows and Cygwin, we release an oversize allocation and > immediately retry at the aligned address in the allocation. If we > lose the address due to a race, we just retry. > - Clean up all the calls to OS::Allocate in codegen and tests by adding > helper AllocateSystemPage function (allocation.h) and > AllocateAssemblerBuffer (cctest.h). > - Changes 'assm' to 'masm' in some targets for consistency when using > a macro-assembler. > > - Eliminates OS::ReleaseRegion, replacing with calls to OS::Free. > - Adds bool return value to OS::Free. > - Cleans up types of flags, protection on Windows and Cygwin. > Bug: chromium:756050 > Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng > Change-Id: I306dbe042cc867670fdc935abca29db074b0da71 Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng Change-Id: Iad3c025334e8f8d7d647be99a36a11ee449c9087 Reviewed-on: https://chromium-review.googlesource.com/767014 Commit-Queue: Bill Budge <bbudge@chromium.org> Reviewed-by: Michael Lippautz <mlippautz@chromium.org> Cr-Commit-Position: refs/heads/master@{#49363}
2017-11-14 15:55:09 +00:00
masm.GetCode(CcTest::i_isolate(), &desc);
buffer->MakeExecutable();
// Call the function from C++.
auto f = GeneratedCode<F2>::FromBuffer(CcTest::i_isolate(), buffer->start());
int result = f.Call(3, 2);
CHECK_EQ(3, result);
}
TEST(AssemblerX64ControlFlow) {
CcTest::InitializeVM();
auto buffer = AllocateAssemblerBuffer();
Assembler masm(AssemblerOptions{}, buffer->CreateView());
// Assemble a simple function that copies argument 1 and returns it.
__ pushq(rbp);
__ movq(rbp, rsp);
__ movq(rax, arg1);
Label target;
__ jmp(&target);
__ movq(rax, arg2);
__ bind(&target);
__ popq(rbp);
__ ret(0);
CodeDesc desc;
Reland "[Memory] Use OS::Allocate for all OS memory allocations." This is a reland of 4899bcb66db1c8c8272b02fb1cedf9886e010f82 This is a reland of b73ee3344ac96f92b4418c3c07779eb95a8541cc Original change's description: > [Memory] Use OS::Allocate for all OS memory allocations. > > - Eliminates OS::ReserveRegion and OS::ReserveAlignedRegion. > - Changes OS::Allocate to take alignment parameter, reorders parameters > to match page_allocator. > - Since the size of memory allocation can be deduced, don't return the > amount of memory allocated. > - Changes reservation of aligned address space. Before we would reserve > (size + alignment) rounded up to page size. This is too much, because > maximum misalignment is (alignment - page_size). > - On Windows and Cygwin, we release an oversize allocation and > immediately retry at the aligned address in the allocation. If we > lose the address due to a race, we just retry. > - Clean up all the calls to OS::Allocate in codegen and tests by adding > helper AllocateSystemPage function (allocation.h) and > AllocateAssemblerBuffer (cctest.h). > - Changes 'assm' to 'masm' in some targets for consistency when using > a macro-assembler. > > - Eliminates OS::ReleaseRegion, replacing with calls to OS::Free. > - Adds bool return value to OS::Free. > - Cleans up types of flags, protection on Windows and Cygwin. > Bug: chromium:756050 > Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng > Change-Id: I306dbe042cc867670fdc935abca29db074b0da71 Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng Change-Id: Iad3c025334e8f8d7d647be99a36a11ee449c9087 Reviewed-on: https://chromium-review.googlesource.com/767014 Commit-Queue: Bill Budge <bbudge@chromium.org> Reviewed-by: Michael Lippautz <mlippautz@chromium.org> Cr-Commit-Position: refs/heads/master@{#49363}
2017-11-14 15:55:09 +00:00
masm.GetCode(CcTest::i_isolate(), &desc);
buffer->MakeExecutable();
// Call the function from C++.
auto f = GeneratedCode<F2>::FromBuffer(CcTest::i_isolate(), buffer->start());
int result = f.Call(3, 2);
CHECK_EQ(3, result);
}
TEST(AssemblerX64LoopImmediates) {
CcTest::InitializeVM();
auto buffer = AllocateAssemblerBuffer();
Assembler masm(AssemblerOptions{}, buffer->CreateView());
// Assemble two loops using rax as counter, and verify the ending counts.
Label Fail;
__ movq(rax, Immediate(-3));
Label Loop1_test;
Label Loop1_body;
__ jmp(&Loop1_test);
__ bind(&Loop1_body);
__ addq(rax, Immediate(7));
__ bind(&Loop1_test);
__ cmpq(rax, Immediate(20));
__ j(less_equal, &Loop1_body);
// Did the loop terminate with the expected value?
__ cmpq(rax, Immediate(25));
__ j(not_equal, &Fail);
Label Loop2_test;
Label Loop2_body;
__ movq(rax, Immediate(0x11FEED00));
__ jmp(&Loop2_test);
__ bind(&Loop2_body);
__ addq(rax, Immediate(-0x1100));
__ bind(&Loop2_test);
__ cmpq(rax, Immediate(0x11FE8000));
__ j(greater, &Loop2_body);
// Did the loop terminate with the expected value?
__ cmpq(rax, Immediate(0x11FE7600));
__ j(not_equal, &Fail);
__ movq(rax, Immediate(1));
__ ret(0);
__ bind(&Fail);
__ movq(rax, Immediate(0));
__ ret(0);
CodeDesc desc;
Reland "[Memory] Use OS::Allocate for all OS memory allocations." This is a reland of 4899bcb66db1c8c8272b02fb1cedf9886e010f82 This is a reland of b73ee3344ac96f92b4418c3c07779eb95a8541cc Original change's description: > [Memory] Use OS::Allocate for all OS memory allocations. > > - Eliminates OS::ReserveRegion and OS::ReserveAlignedRegion. > - Changes OS::Allocate to take alignment parameter, reorders parameters > to match page_allocator. > - Since the size of memory allocation can be deduced, don't return the > amount of memory allocated. > - Changes reservation of aligned address space. Before we would reserve > (size + alignment) rounded up to page size. This is too much, because > maximum misalignment is (alignment - page_size). > - On Windows and Cygwin, we release an oversize allocation and > immediately retry at the aligned address in the allocation. If we > lose the address due to a race, we just retry. > - Clean up all the calls to OS::Allocate in codegen and tests by adding > helper AllocateSystemPage function (allocation.h) and > AllocateAssemblerBuffer (cctest.h). > - Changes 'assm' to 'masm' in some targets for consistency when using > a macro-assembler. > > - Eliminates OS::ReleaseRegion, replacing with calls to OS::Free. > - Adds bool return value to OS::Free. > - Cleans up types of flags, protection on Windows and Cygwin. > Bug: chromium:756050 > Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng > Change-Id: I306dbe042cc867670fdc935abca29db074b0da71 Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng Change-Id: Iad3c025334e8f8d7d647be99a36a11ee449c9087 Reviewed-on: https://chromium-review.googlesource.com/767014 Commit-Queue: Bill Budge <bbudge@chromium.org> Reviewed-by: Michael Lippautz <mlippautz@chromium.org> Cr-Commit-Position: refs/heads/master@{#49363}
2017-11-14 15:55:09 +00:00
masm.GetCode(CcTest::i_isolate(), &desc);
buffer->MakeExecutable();
// Call the function from C++.
auto f = GeneratedCode<F0>::FromBuffer(CcTest::i_isolate(), buffer->start());
int result = f.Call();
CHECK_EQ(1, result);
}
TEST(OperandRegisterDependency) {
int offsets[4] = {0, 1, 0xFED, 0xBEEFCAD};
for (int i = 0; i < 4; i++) {
int offset = offsets[i];
CHECK(Operand(rax, offset).AddressUsesRegister(rax));
CHECK(!Operand(rax, offset).AddressUsesRegister(r8));
CHECK(!Operand(rax, offset).AddressUsesRegister(rcx));
CHECK(Operand(rax, rax, times_1, offset).AddressUsesRegister(rax));
CHECK(!Operand(rax, rax, times_1, offset).AddressUsesRegister(r8));
CHECK(!Operand(rax, rax, times_1, offset).AddressUsesRegister(rcx));
CHECK(Operand(rax, rcx, times_1, offset).AddressUsesRegister(rax));
CHECK(Operand(rax, rcx, times_1, offset).AddressUsesRegister(rcx));
CHECK(!Operand(rax, rcx, times_1, offset).AddressUsesRegister(r8));
CHECK(!Operand(rax, rcx, times_1, offset).AddressUsesRegister(r9));
CHECK(!Operand(rax, rcx, times_1, offset).AddressUsesRegister(rdx));
CHECK(!Operand(rax, rcx, times_1, offset).AddressUsesRegister(rsp));
CHECK(Operand(rsp, offset).AddressUsesRegister(rsp));
CHECK(!Operand(rsp, offset).AddressUsesRegister(rax));
CHECK(!Operand(rsp, offset).AddressUsesRegister(r15));
CHECK(Operand(rbp, offset).AddressUsesRegister(rbp));
CHECK(!Operand(rbp, offset).AddressUsesRegister(rax));
CHECK(!Operand(rbp, offset).AddressUsesRegister(r13));
CHECK(Operand(rbp, rax, times_1, offset).AddressUsesRegister(rbp));
CHECK(Operand(rbp, rax, times_1, offset).AddressUsesRegister(rax));
CHECK(!Operand(rbp, rax, times_1, offset).AddressUsesRegister(rcx));
CHECK(!Operand(rbp, rax, times_1, offset).AddressUsesRegister(r13));
CHECK(!Operand(rbp, rax, times_1, offset).AddressUsesRegister(r8));
CHECK(!Operand(rbp, rax, times_1, offset).AddressUsesRegister(rsp));
CHECK(Operand(rsp, rbp, times_1, offset).AddressUsesRegister(rsp));
CHECK(Operand(rsp, rbp, times_1, offset).AddressUsesRegister(rbp));
CHECK(!Operand(rsp, rbp, times_1, offset).AddressUsesRegister(rax));
CHECK(!Operand(rsp, rbp, times_1, offset).AddressUsesRegister(r15));
CHECK(!Operand(rsp, rbp, times_1, offset).AddressUsesRegister(r13));
}
}
TEST(AssemblerX64LabelChaining) {
// Test chaining of label usages within instructions (issue 1644).
CcTest::InitializeVM();
v8::HandleScope scope(CcTest::isolate());
Assembler masm(AssemblerOptions{});
Label target;
__ j(equal, &target);
__ j(not_equal, &target);
__ bind(&target);
__ nop();
}
TEST(AssemblerMultiByteNop) {
CcTest::InitializeVM();
v8::HandleScope scope(CcTest::isolate());
byte buffer[1024];
Isolate* isolate = CcTest::i_isolate();
Assembler masm(AssemblerOptions{},
ExternalAssemblerBuffer(buffer, sizeof(buffer)));
__ pushq(rbx);
__ pushq(rcx);
__ pushq(rdx);
__ pushq(rdi);
__ pushq(rsi);
__ movq(rax, Immediate(1));
__ movq(rbx, Immediate(2));
__ movq(rcx, Immediate(3));
__ movq(rdx, Immediate(4));
__ movq(rdi, Immediate(5));
__ movq(rsi, Immediate(6));
for (int i = 0; i < 16; i++) {
Reland "[Memory] Use OS::Allocate for all OS memory allocations." This is a reland of 4899bcb66db1c8c8272b02fb1cedf9886e010f82 This is a reland of b73ee3344ac96f92b4418c3c07779eb95a8541cc Original change's description: > [Memory] Use OS::Allocate for all OS memory allocations. > > - Eliminates OS::ReserveRegion and OS::ReserveAlignedRegion. > - Changes OS::Allocate to take alignment parameter, reorders parameters > to match page_allocator. > - Since the size of memory allocation can be deduced, don't return the > amount of memory allocated. > - Changes reservation of aligned address space. Before we would reserve > (size + alignment) rounded up to page size. This is too much, because > maximum misalignment is (alignment - page_size). > - On Windows and Cygwin, we release an oversize allocation and > immediately retry at the aligned address in the allocation. If we > lose the address due to a race, we just retry. > - Clean up all the calls to OS::Allocate in codegen and tests by adding > helper AllocateSystemPage function (allocation.h) and > AllocateAssemblerBuffer (cctest.h). > - Changes 'assm' to 'masm' in some targets for consistency when using > a macro-assembler. > > - Eliminates OS::ReleaseRegion, replacing with calls to OS::Free. > - Adds bool return value to OS::Free. > - Cleans up types of flags, protection on Windows and Cygwin. > Bug: chromium:756050 > Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng > Change-Id: I306dbe042cc867670fdc935abca29db074b0da71 Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng Change-Id: Iad3c025334e8f8d7d647be99a36a11ee449c9087 Reviewed-on: https://chromium-review.googlesource.com/767014 Commit-Queue: Bill Budge <bbudge@chromium.org> Reviewed-by: Michael Lippautz <mlippautz@chromium.org> Cr-Commit-Position: refs/heads/master@{#49363}
2017-11-14 15:55:09 +00:00
int before = masm.pc_offset();
__ Nop(i);
Reland "[Memory] Use OS::Allocate for all OS memory allocations." This is a reland of 4899bcb66db1c8c8272b02fb1cedf9886e010f82 This is a reland of b73ee3344ac96f92b4418c3c07779eb95a8541cc Original change's description: > [Memory] Use OS::Allocate for all OS memory allocations. > > - Eliminates OS::ReserveRegion and OS::ReserveAlignedRegion. > - Changes OS::Allocate to take alignment parameter, reorders parameters > to match page_allocator. > - Since the size of memory allocation can be deduced, don't return the > amount of memory allocated. > - Changes reservation of aligned address space. Before we would reserve > (size + alignment) rounded up to page size. This is too much, because > maximum misalignment is (alignment - page_size). > - On Windows and Cygwin, we release an oversize allocation and > immediately retry at the aligned address in the allocation. If we > lose the address due to a race, we just retry. > - Clean up all the calls to OS::Allocate in codegen and tests by adding > helper AllocateSystemPage function (allocation.h) and > AllocateAssemblerBuffer (cctest.h). > - Changes 'assm' to 'masm' in some targets for consistency when using > a macro-assembler. > > - Eliminates OS::ReleaseRegion, replacing with calls to OS::Free. > - Adds bool return value to OS::Free. > - Cleans up types of flags, protection on Windows and Cygwin. > Bug: chromium:756050 > Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng > Change-Id: I306dbe042cc867670fdc935abca29db074b0da71 Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng Change-Id: Iad3c025334e8f8d7d647be99a36a11ee449c9087 Reviewed-on: https://chromium-review.googlesource.com/767014 Commit-Queue: Bill Budge <bbudge@chromium.org> Reviewed-by: Michael Lippautz <mlippautz@chromium.org> Cr-Commit-Position: refs/heads/master@{#49363}
2017-11-14 15:55:09 +00:00
CHECK_EQ(masm.pc_offset() - before, i);
}
Label fail;
__ cmpq(rax, Immediate(1));
__ j(not_equal, &fail);
__ cmpq(rbx, Immediate(2));
__ j(not_equal, &fail);
__ cmpq(rcx, Immediate(3));
__ j(not_equal, &fail);
__ cmpq(rdx, Immediate(4));
__ j(not_equal, &fail);
__ cmpq(rdi, Immediate(5));
__ j(not_equal, &fail);
__ cmpq(rsi, Immediate(6));
__ j(not_equal, &fail);
__ movq(rax, Immediate(42));
__ popq(rsi);
__ popq(rdi);
__ popq(rdx);
__ popq(rcx);
__ popq(rbx);
__ ret(0);
__ bind(&fail);
__ movq(rax, Immediate(13));
__ popq(rsi);
__ popq(rdi);
__ popq(rdx);
__ popq(rcx);
__ popq(rbx);
__ ret(0);
CodeDesc desc;
Reland "[Memory] Use OS::Allocate for all OS memory allocations." This is a reland of 4899bcb66db1c8c8272b02fb1cedf9886e010f82 This is a reland of b73ee3344ac96f92b4418c3c07779eb95a8541cc Original change's description: > [Memory] Use OS::Allocate for all OS memory allocations. > > - Eliminates OS::ReserveRegion and OS::ReserveAlignedRegion. > - Changes OS::Allocate to take alignment parameter, reorders parameters > to match page_allocator. > - Since the size of memory allocation can be deduced, don't return the > amount of memory allocated. > - Changes reservation of aligned address space. Before we would reserve > (size + alignment) rounded up to page size. This is too much, because > maximum misalignment is (alignment - page_size). > - On Windows and Cygwin, we release an oversize allocation and > immediately retry at the aligned address in the allocation. If we > lose the address due to a race, we just retry. > - Clean up all the calls to OS::Allocate in codegen and tests by adding > helper AllocateSystemPage function (allocation.h) and > AllocateAssemblerBuffer (cctest.h). > - Changes 'assm' to 'masm' in some targets for consistency when using > a macro-assembler. > > - Eliminates OS::ReleaseRegion, replacing with calls to OS::Free. > - Adds bool return value to OS::Free. > - Cleans up types of flags, protection on Windows and Cygwin. > Bug: chromium:756050 > Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng > Change-Id: I306dbe042cc867670fdc935abca29db074b0da71 Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng Change-Id: Iad3c025334e8f8d7d647be99a36a11ee449c9087 Reviewed-on: https://chromium-review.googlesource.com/767014 Commit-Queue: Bill Budge <bbudge@chromium.org> Reviewed-by: Michael Lippautz <mlippautz@chromium.org> Cr-Commit-Position: refs/heads/master@{#49363}
2017-11-14 15:55:09 +00:00
masm.GetCode(isolate, &desc);
Reland "Reland "[deoptimizer] Change deopt entries into builtins"" This is a reland of fbfa9bf4ec72b1b73a96b70ccb68cd98c321511b The arm64 was missing proper codegen for CFI, thus sizes were off. Original change's description: > Reland "[deoptimizer] Change deopt entries into builtins" > > This is a reland of 7f58ced72eb65b6b5530ccabaf2eaebe45bf9d33 > > It fixes the different exit size emitted on x64/Atom CPUs due to > performance tuning in TurboAssembler::Call. Additionally, add > cctests to verify the fixed size exits. > > Original change's description: > > [deoptimizer] Change deopt entries into builtins > > > > While the overall goal of this commit is to change deoptimization > > entries into builtins, there are multiple related things happening: > > > > - Deoptimization entries, formerly stubs (i.e. Code objects generated > > at runtime, guaranteed to be immovable), have been converted into > > builtins. The major restriction is that we now need to preserve the > > kRootRegister, which was formerly used on most architectures to pass > > the deoptimization id. The solution differs based on platform. > > - Renamed DEOPT_ENTRIES_OR_FOR_TESTING code kind to FOR_TESTING. > > - Removed heap/ support for immovable Code generation. > > - Removed the DeserializerData class (no longer needed). > > - arm64: to preserve 4-byte deopt exits, introduced a new optimization > > in which the final jump to the deoptimization entry is generated > > once per Code object, and deopt exits can continue to emit a > > near-call. > > - arm,ia32,x64: change to fixed-size deopt exits. This reduces exit > > sizes by 4/8, 5, and 5 bytes, respectively. > > > > On arm the deopt exit size is reduced from 12 (or 16) bytes to 8 bytes > > by using the same strategy as on arm64 (recalc deopt id from return > > address). Before: > > > > e300a002 movw r10, <id> > > e59fc024 ldr ip, [pc, <entry offset>] > > e12fff3c blx ip > > > > After: > > > > e59acb35 ldr ip, [r10, <entry offset>] > > e12fff3c blx ip > > > > On arm64 the deopt exit size remains 4 bytes (or 8 bytes in same cases > > with CFI). Additionally, up to 4 builtin jumps are emitted per Code > > object (max 32 bytes added overhead per Code object). Before: > > > > 9401cdae bl <entry offset> > > > > After: > > > > # eager deoptimization entry jump. > > f95b1f50 ldr x16, [x26, <eager entry offset>] > > d61f0200 br x16 > > # lazy deoptimization entry jump. > > f95b2b50 ldr x16, [x26, <lazy entry offset>] > > d61f0200 br x16 > > # the deopt exit. > > 97fffffc bl <eager deoptimization entry jump offset> > > > > On ia32 the deopt exit size is reduced from 10 to 5 bytes. Before: > > > > bb00000000 mov ebx,<id> > > e825f5372b call <entry> > > > > After: > > > > e8ea2256ba call <entry> > > > > On x64 the deopt exit size is reduced from 12 to 7 bytes. Before: > > > > 49c7c511000000 REX.W movq r13,<id> > > e8ea2f0700 call <entry> > > > > After: > > > > 41ff9560360000 call [r13+<entry offset>] > > > > Bug: v8:8661,v8:8768 > > Change-Id: I13e30aedc360474dc818fecc528ce87c3bfeed42 > > Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2465834 > > Commit-Queue: Jakob Gruber <jgruber@chromium.org> > > Reviewed-by: Ross McIlroy <rmcilroy@chromium.org> > > Reviewed-by: Tobias Tebbi <tebbi@chromium.org> > > Reviewed-by: Ulan Degenbaev <ulan@chromium.org> > > Cr-Commit-Position: refs/heads/master@{#70597} > > Tbr: ulan@chromium.org, tebbi@chromium.org, rmcilroy@chromium.org > Bug: v8:8661,v8:8768,chromium:1140165 > Change-Id: Ibcd5c39c58a70bf2b2ac221aa375fc68d495e144 > Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2485506 > Reviewed-by: Jakob Gruber <jgruber@chromium.org> > Reviewed-by: Tobias Tebbi <tebbi@chromium.org> > Commit-Queue: Jakob Gruber <jgruber@chromium.org> > Cr-Commit-Position: refs/heads/master@{#70655} Tbr: ulan@chromium.org, tebbi@chromium.org, rmcilroy@chromium.org Bug: v8:8661 Bug: v8:8768 Bug: chromium:1140165 Change-Id: I471cc94fc085e527dc9bfb5a84b96bd907c2333f Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2488682 Reviewed-by: Jakob Gruber <jgruber@chromium.org> Commit-Queue: Jakob Gruber <jgruber@chromium.org> Cr-Commit-Position: refs/heads/master@{#70672}
2020-10-21 05:12:25 +00:00
Handle<Code> code =
Factory::CodeBuilder(isolate, desc, CodeKind::FOR_TESTING).Build();
auto f = GeneratedCode<F0>::FromCode(*code);
int res = f.Call();
CHECK_EQ(42, res);
}
#ifdef __GNUC__
#define ELEMENT_COUNT 4u
void DoSSE2(const v8::FunctionCallbackInfo<v8::Value>& args) {
v8::HandleScope scope(CcTest::isolate());
v8::Local<v8::Context> context = CcTest::isolate()->GetCurrentContext();
byte buffer[1024];
CHECK(args[0]->IsArray());
v8::Local<v8::Array> vec = v8::Local<v8::Array>::Cast(args[0]);
CHECK_EQ(ELEMENT_COUNT, vec->Length());
Isolate* isolate = CcTest::i_isolate();
Assembler masm(AssemblerOptions{},
ExternalAssemblerBuffer(buffer, sizeof(buffer)));
// Remove return address from the stack for fix stack frame alignment.
__ popq(rcx);
// Store input vector on the stack.
for (unsigned i = 0; i < ELEMENT_COUNT; i++) {
__ movl(rax, Immediate(vec->Get(context, i)
.ToLocalChecked()
->Int32Value(context)
.FromJust()));
__ shlq(rax, Immediate(0x20));
__ orq(rax, Immediate(vec->Get(context, ++i)
.ToLocalChecked()
->Int32Value(context)
.FromJust()));
__ pushq(rax);
}
// Read vector into a xmm register.
__ xorps(xmm0, xmm0);
__ movdqa(xmm0, Operand(rsp, 0));
// Create mask and store it in the return register.
__ movmskps(rax, xmm0);
// Remove unused data from the stack.
__ addq(rsp, Immediate(ELEMENT_COUNT * sizeof(int32_t)));
// Restore return address.
__ pushq(rcx);
__ ret(0);
CodeDesc desc;
Reland "[Memory] Use OS::Allocate for all OS memory allocations." This is a reland of 4899bcb66db1c8c8272b02fb1cedf9886e010f82 This is a reland of b73ee3344ac96f92b4418c3c07779eb95a8541cc Original change's description: > [Memory] Use OS::Allocate for all OS memory allocations. > > - Eliminates OS::ReserveRegion and OS::ReserveAlignedRegion. > - Changes OS::Allocate to take alignment parameter, reorders parameters > to match page_allocator. > - Since the size of memory allocation can be deduced, don't return the > amount of memory allocated. > - Changes reservation of aligned address space. Before we would reserve > (size + alignment) rounded up to page size. This is too much, because > maximum misalignment is (alignment - page_size). > - On Windows and Cygwin, we release an oversize allocation and > immediately retry at the aligned address in the allocation. If we > lose the address due to a race, we just retry. > - Clean up all the calls to OS::Allocate in codegen and tests by adding > helper AllocateSystemPage function (allocation.h) and > AllocateAssemblerBuffer (cctest.h). > - Changes 'assm' to 'masm' in some targets for consistency when using > a macro-assembler. > > - Eliminates OS::ReleaseRegion, replacing with calls to OS::Free. > - Adds bool return value to OS::Free. > - Cleans up types of flags, protection on Windows and Cygwin. > Bug: chromium:756050 > Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng > Change-Id: I306dbe042cc867670fdc935abca29db074b0da71 Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng Change-Id: Iad3c025334e8f8d7d647be99a36a11ee449c9087 Reviewed-on: https://chromium-review.googlesource.com/767014 Commit-Queue: Bill Budge <bbudge@chromium.org> Reviewed-by: Michael Lippautz <mlippautz@chromium.org> Cr-Commit-Position: refs/heads/master@{#49363}
2017-11-14 15:55:09 +00:00
masm.GetCode(isolate, &desc);
Reland "Reland "[deoptimizer] Change deopt entries into builtins"" This is a reland of fbfa9bf4ec72b1b73a96b70ccb68cd98c321511b The arm64 was missing proper codegen for CFI, thus sizes were off. Original change's description: > Reland "[deoptimizer] Change deopt entries into builtins" > > This is a reland of 7f58ced72eb65b6b5530ccabaf2eaebe45bf9d33 > > It fixes the different exit size emitted on x64/Atom CPUs due to > performance tuning in TurboAssembler::Call. Additionally, add > cctests to verify the fixed size exits. > > Original change's description: > > [deoptimizer] Change deopt entries into builtins > > > > While the overall goal of this commit is to change deoptimization > > entries into builtins, there are multiple related things happening: > > > > - Deoptimization entries, formerly stubs (i.e. Code objects generated > > at runtime, guaranteed to be immovable), have been converted into > > builtins. The major restriction is that we now need to preserve the > > kRootRegister, which was formerly used on most architectures to pass > > the deoptimization id. The solution differs based on platform. > > - Renamed DEOPT_ENTRIES_OR_FOR_TESTING code kind to FOR_TESTING. > > - Removed heap/ support for immovable Code generation. > > - Removed the DeserializerData class (no longer needed). > > - arm64: to preserve 4-byte deopt exits, introduced a new optimization > > in which the final jump to the deoptimization entry is generated > > once per Code object, and deopt exits can continue to emit a > > near-call. > > - arm,ia32,x64: change to fixed-size deopt exits. This reduces exit > > sizes by 4/8, 5, and 5 bytes, respectively. > > > > On arm the deopt exit size is reduced from 12 (or 16) bytes to 8 bytes > > by using the same strategy as on arm64 (recalc deopt id from return > > address). Before: > > > > e300a002 movw r10, <id> > > e59fc024 ldr ip, [pc, <entry offset>] > > e12fff3c blx ip > > > > After: > > > > e59acb35 ldr ip, [r10, <entry offset>] > > e12fff3c blx ip > > > > On arm64 the deopt exit size remains 4 bytes (or 8 bytes in same cases > > with CFI). Additionally, up to 4 builtin jumps are emitted per Code > > object (max 32 bytes added overhead per Code object). Before: > > > > 9401cdae bl <entry offset> > > > > After: > > > > # eager deoptimization entry jump. > > f95b1f50 ldr x16, [x26, <eager entry offset>] > > d61f0200 br x16 > > # lazy deoptimization entry jump. > > f95b2b50 ldr x16, [x26, <lazy entry offset>] > > d61f0200 br x16 > > # the deopt exit. > > 97fffffc bl <eager deoptimization entry jump offset> > > > > On ia32 the deopt exit size is reduced from 10 to 5 bytes. Before: > > > > bb00000000 mov ebx,<id> > > e825f5372b call <entry> > > > > After: > > > > e8ea2256ba call <entry> > > > > On x64 the deopt exit size is reduced from 12 to 7 bytes. Before: > > > > 49c7c511000000 REX.W movq r13,<id> > > e8ea2f0700 call <entry> > > > > After: > > > > 41ff9560360000 call [r13+<entry offset>] > > > > Bug: v8:8661,v8:8768 > > Change-Id: I13e30aedc360474dc818fecc528ce87c3bfeed42 > > Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2465834 > > Commit-Queue: Jakob Gruber <jgruber@chromium.org> > > Reviewed-by: Ross McIlroy <rmcilroy@chromium.org> > > Reviewed-by: Tobias Tebbi <tebbi@chromium.org> > > Reviewed-by: Ulan Degenbaev <ulan@chromium.org> > > Cr-Commit-Position: refs/heads/master@{#70597} > > Tbr: ulan@chromium.org, tebbi@chromium.org, rmcilroy@chromium.org > Bug: v8:8661,v8:8768,chromium:1140165 > Change-Id: Ibcd5c39c58a70bf2b2ac221aa375fc68d495e144 > Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2485506 > Reviewed-by: Jakob Gruber <jgruber@chromium.org> > Reviewed-by: Tobias Tebbi <tebbi@chromium.org> > Commit-Queue: Jakob Gruber <jgruber@chromium.org> > Cr-Commit-Position: refs/heads/master@{#70655} Tbr: ulan@chromium.org, tebbi@chromium.org, rmcilroy@chromium.org Bug: v8:8661 Bug: v8:8768 Bug: chromium:1140165 Change-Id: I471cc94fc085e527dc9bfb5a84b96bd907c2333f Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2488682 Reviewed-by: Jakob Gruber <jgruber@chromium.org> Commit-Queue: Jakob Gruber <jgruber@chromium.org> Cr-Commit-Position: refs/heads/master@{#70672}
2020-10-21 05:12:25 +00:00
Handle<Code> code =
Factory::CodeBuilder(isolate, desc, CodeKind::FOR_TESTING).Build();
auto f = GeneratedCode<F0>::FromCode(*code);
int res = f.Call();
args.GetReturnValue().Set(v8::Integer::New(CcTest::isolate(), res));
}
TEST(StackAlignmentForSSE2) {
CcTest::InitializeVM();
CHECK_EQ(0, v8::base::OS::ActivationFrameAlignment() % 16);
v8::Isolate* isolate = CcTest::isolate();
v8::HandleScope handle_scope(isolate);
v8::Local<v8::ObjectTemplate> global_template =
v8::ObjectTemplate::New(isolate);
global_template->Set(isolate, "do_sse2",
v8::FunctionTemplate::New(isolate, DoSSE2));
LocalContext env(nullptr, global_template);
CompileRun(
"function foo(vec) {"
" return do_sse2(vec);"
"}");
v8::Local<v8::Object> global_object = env->Global();
v8::Local<v8::Function> foo = v8::Local<v8::Function>::Cast(
global_object->Get(env.local(), v8_str("foo")).ToLocalChecked());
int32_t vec[ELEMENT_COUNT] = { -1, 1, 1, 1 };
v8::Local<v8::Array> v8_vec = v8::Array::New(isolate, ELEMENT_COUNT);
for (unsigned i = 0; i < ELEMENT_COUNT; i++) {
v8_vec->Set(env.local(), i, v8_num(vec[i])).FromJust();
}
v8::Local<v8::Value> args[] = { v8_vec };
v8::Local<v8::Value> result =
foo->Call(env.local(), global_object, 1, args).ToLocalChecked();
// The mask should be 0b1000.
CHECK_EQ(8, result->Int32Value(env.local()).FromJust());
}
#undef ELEMENT_COUNT
#endif // __GNUC__
TEST(AssemblerX64Extractps) {
CcTest::InitializeVM();
if (!CpuFeatures::IsSupported(SSE4_1)) return;
v8::HandleScope scope(CcTest::isolate());
byte buffer[256];
Isolate* isolate = CcTest::i_isolate();
Assembler masm(AssemblerOptions{},
ExternalAssemblerBuffer(buffer, sizeof(buffer)));
Reland "[Memory] Use OS::Allocate for all OS memory allocations." This is a reland of 4899bcb66db1c8c8272b02fb1cedf9886e010f82 This is a reland of b73ee3344ac96f92b4418c3c07779eb95a8541cc Original change's description: > [Memory] Use OS::Allocate for all OS memory allocations. > > - Eliminates OS::ReserveRegion and OS::ReserveAlignedRegion. > - Changes OS::Allocate to take alignment parameter, reorders parameters > to match page_allocator. > - Since the size of memory allocation can be deduced, don't return the > amount of memory allocated. > - Changes reservation of aligned address space. Before we would reserve > (size + alignment) rounded up to page size. This is too much, because > maximum misalignment is (alignment - page_size). > - On Windows and Cygwin, we release an oversize allocation and > immediately retry at the aligned address in the allocation. If we > lose the address due to a race, we just retry. > - Clean up all the calls to OS::Allocate in codegen and tests by adding > helper AllocateSystemPage function (allocation.h) and > AllocateAssemblerBuffer (cctest.h). > - Changes 'assm' to 'masm' in some targets for consistency when using > a macro-assembler. > > - Eliminates OS::ReleaseRegion, replacing with calls to OS::Free. > - Adds bool return value to OS::Free. > - Cleans up types of flags, protection on Windows and Cygwin. > Bug: chromium:756050 > Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng > Change-Id: I306dbe042cc867670fdc935abca29db074b0da71 Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng Change-Id: Iad3c025334e8f8d7d647be99a36a11ee449c9087 Reviewed-on: https://chromium-review.googlesource.com/767014 Commit-Queue: Bill Budge <bbudge@chromium.org> Reviewed-by: Michael Lippautz <mlippautz@chromium.org> Cr-Commit-Position: refs/heads/master@{#49363}
2017-11-14 15:55:09 +00:00
{
CpuFeatureScope fscope2(&masm, SSE4_1);
__ extractps(rax, xmm0, 0x1);
__ ret(0);
}
CodeDesc desc;
Reland "[Memory] Use OS::Allocate for all OS memory allocations." This is a reland of 4899bcb66db1c8c8272b02fb1cedf9886e010f82 This is a reland of b73ee3344ac96f92b4418c3c07779eb95a8541cc Original change's description: > [Memory] Use OS::Allocate for all OS memory allocations. > > - Eliminates OS::ReserveRegion and OS::ReserveAlignedRegion. > - Changes OS::Allocate to take alignment parameter, reorders parameters > to match page_allocator. > - Since the size of memory allocation can be deduced, don't return the > amount of memory allocated. > - Changes reservation of aligned address space. Before we would reserve > (size + alignment) rounded up to page size. This is too much, because > maximum misalignment is (alignment - page_size). > - On Windows and Cygwin, we release an oversize allocation and > immediately retry at the aligned address in the allocation. If we > lose the address due to a race, we just retry. > - Clean up all the calls to OS::Allocate in codegen and tests by adding > helper AllocateSystemPage function (allocation.h) and > AllocateAssemblerBuffer (cctest.h). > - Changes 'assm' to 'masm' in some targets for consistency when using > a macro-assembler. > > - Eliminates OS::ReleaseRegion, replacing with calls to OS::Free. > - Adds bool return value to OS::Free. > - Cleans up types of flags, protection on Windows and Cygwin. > Bug: chromium:756050 > Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng > Change-Id: I306dbe042cc867670fdc935abca29db074b0da71 Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng Change-Id: Iad3c025334e8f8d7d647be99a36a11ee449c9087 Reviewed-on: https://chromium-review.googlesource.com/767014 Commit-Queue: Bill Budge <bbudge@chromium.org> Reviewed-by: Michael Lippautz <mlippautz@chromium.org> Cr-Commit-Position: refs/heads/master@{#49363}
2017-11-14 15:55:09 +00:00
masm.GetCode(isolate, &desc);
Reland "Reland "[deoptimizer] Change deopt entries into builtins"" This is a reland of fbfa9bf4ec72b1b73a96b70ccb68cd98c321511b The arm64 was missing proper codegen for CFI, thus sizes were off. Original change's description: > Reland "[deoptimizer] Change deopt entries into builtins" > > This is a reland of 7f58ced72eb65b6b5530ccabaf2eaebe45bf9d33 > > It fixes the different exit size emitted on x64/Atom CPUs due to > performance tuning in TurboAssembler::Call. Additionally, add > cctests to verify the fixed size exits. > > Original change's description: > > [deoptimizer] Change deopt entries into builtins > > > > While the overall goal of this commit is to change deoptimization > > entries into builtins, there are multiple related things happening: > > > > - Deoptimization entries, formerly stubs (i.e. Code objects generated > > at runtime, guaranteed to be immovable), have been converted into > > builtins. The major restriction is that we now need to preserve the > > kRootRegister, which was formerly used on most architectures to pass > > the deoptimization id. The solution differs based on platform. > > - Renamed DEOPT_ENTRIES_OR_FOR_TESTING code kind to FOR_TESTING. > > - Removed heap/ support for immovable Code generation. > > - Removed the DeserializerData class (no longer needed). > > - arm64: to preserve 4-byte deopt exits, introduced a new optimization > > in which the final jump to the deoptimization entry is generated > > once per Code object, and deopt exits can continue to emit a > > near-call. > > - arm,ia32,x64: change to fixed-size deopt exits. This reduces exit > > sizes by 4/8, 5, and 5 bytes, respectively. > > > > On arm the deopt exit size is reduced from 12 (or 16) bytes to 8 bytes > > by using the same strategy as on arm64 (recalc deopt id from return > > address). Before: > > > > e300a002 movw r10, <id> > > e59fc024 ldr ip, [pc, <entry offset>] > > e12fff3c blx ip > > > > After: > > > > e59acb35 ldr ip, [r10, <entry offset>] > > e12fff3c blx ip > > > > On arm64 the deopt exit size remains 4 bytes (or 8 bytes in same cases > > with CFI). Additionally, up to 4 builtin jumps are emitted per Code > > object (max 32 bytes added overhead per Code object). Before: > > > > 9401cdae bl <entry offset> > > > > After: > > > > # eager deoptimization entry jump. > > f95b1f50 ldr x16, [x26, <eager entry offset>] > > d61f0200 br x16 > > # lazy deoptimization entry jump. > > f95b2b50 ldr x16, [x26, <lazy entry offset>] > > d61f0200 br x16 > > # the deopt exit. > > 97fffffc bl <eager deoptimization entry jump offset> > > > > On ia32 the deopt exit size is reduced from 10 to 5 bytes. Before: > > > > bb00000000 mov ebx,<id> > > e825f5372b call <entry> > > > > After: > > > > e8ea2256ba call <entry> > > > > On x64 the deopt exit size is reduced from 12 to 7 bytes. Before: > > > > 49c7c511000000 REX.W movq r13,<id> > > e8ea2f0700 call <entry> > > > > After: > > > > 41ff9560360000 call [r13+<entry offset>] > > > > Bug: v8:8661,v8:8768 > > Change-Id: I13e30aedc360474dc818fecc528ce87c3bfeed42 > > Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2465834 > > Commit-Queue: Jakob Gruber <jgruber@chromium.org> > > Reviewed-by: Ross McIlroy <rmcilroy@chromium.org> > > Reviewed-by: Tobias Tebbi <tebbi@chromium.org> > > Reviewed-by: Ulan Degenbaev <ulan@chromium.org> > > Cr-Commit-Position: refs/heads/master@{#70597} > > Tbr: ulan@chromium.org, tebbi@chromium.org, rmcilroy@chromium.org > Bug: v8:8661,v8:8768,chromium:1140165 > Change-Id: Ibcd5c39c58a70bf2b2ac221aa375fc68d495e144 > Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2485506 > Reviewed-by: Jakob Gruber <jgruber@chromium.org> > Reviewed-by: Tobias Tebbi <tebbi@chromium.org> > Commit-Queue: Jakob Gruber <jgruber@chromium.org> > Cr-Commit-Position: refs/heads/master@{#70655} Tbr: ulan@chromium.org, tebbi@chromium.org, rmcilroy@chromium.org Bug: v8:8661 Bug: v8:8768 Bug: chromium:1140165 Change-Id: I471cc94fc085e527dc9bfb5a84b96bd907c2333f Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2488682 Reviewed-by: Jakob Gruber <jgruber@chromium.org> Commit-Queue: Jakob Gruber <jgruber@chromium.org> Cr-Commit-Position: refs/heads/master@{#70672}
2020-10-21 05:12:25 +00:00
Handle<Code> code =
Factory::CodeBuilder(isolate, desc, CodeKind::FOR_TESTING).Build();
#ifdef OBJECT_PRINT
StdoutStream os;
code->Print(os);
#endif
auto f = GeneratedCode<F3>::FromCode(*code);
uint64_t value1 = 0x1234'5678'8765'4321;
CHECK_EQ(0x12345678u, f.Call(base::uint64_to_double(value1)));
uint64_t value2 = 0x8765'4321'1234'5678;
CHECK_EQ(0x87654321u, f.Call(base::uint64_to_double(value2)));
}
using F6 = int(float x, float y);
TEST(AssemblerX64SSE) {
CcTest::InitializeVM();
Isolate* isolate = reinterpret_cast<Isolate*>(CcTest::isolate());
HandleScope scope(isolate);
v8::internal::byte buffer[256];
MacroAssembler masm(isolate, v8::internal::CodeObjectRequired::kYes,
ExternalAssemblerBuffer(buffer, sizeof(buffer)));
{
__ shufps(xmm0, xmm0, 0x0); // brocast first argument
__ shufps(xmm1, xmm1, 0x0); // brocast second argument
__ movaps(xmm2, xmm1);
__ addps(xmm2, xmm0);
__ mulps(xmm2, xmm1);
__ subps(xmm2, xmm0);
__ divps(xmm2, xmm1);
__ cvttss2si(rax, xmm2);
__ ret(0);
}
CodeDesc desc;
Reland "[Memory] Use OS::Allocate for all OS memory allocations." This is a reland of 4899bcb66db1c8c8272b02fb1cedf9886e010f82 This is a reland of b73ee3344ac96f92b4418c3c07779eb95a8541cc Original change's description: > [Memory] Use OS::Allocate for all OS memory allocations. > > - Eliminates OS::ReserveRegion and OS::ReserveAlignedRegion. > - Changes OS::Allocate to take alignment parameter, reorders parameters > to match page_allocator. > - Since the size of memory allocation can be deduced, don't return the > amount of memory allocated. > - Changes reservation of aligned address space. Before we would reserve > (size + alignment) rounded up to page size. This is too much, because > maximum misalignment is (alignment - page_size). > - On Windows and Cygwin, we release an oversize allocation and > immediately retry at the aligned address in the allocation. If we > lose the address due to a race, we just retry. > - Clean up all the calls to OS::Allocate in codegen and tests by adding > helper AllocateSystemPage function (allocation.h) and > AllocateAssemblerBuffer (cctest.h). > - Changes 'assm' to 'masm' in some targets for consistency when using > a macro-assembler. > > - Eliminates OS::ReleaseRegion, replacing with calls to OS::Free. > - Adds bool return value to OS::Free. > - Cleans up types of flags, protection on Windows and Cygwin. > Bug: chromium:756050 > Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng > Change-Id: I306dbe042cc867670fdc935abca29db074b0da71 Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng Change-Id: Iad3c025334e8f8d7d647be99a36a11ee449c9087 Reviewed-on: https://chromium-review.googlesource.com/767014 Commit-Queue: Bill Budge <bbudge@chromium.org> Reviewed-by: Michael Lippautz <mlippautz@chromium.org> Cr-Commit-Position: refs/heads/master@{#49363}
2017-11-14 15:55:09 +00:00
masm.GetCode(isolate, &desc);
Reland "Reland "[deoptimizer] Change deopt entries into builtins"" This is a reland of fbfa9bf4ec72b1b73a96b70ccb68cd98c321511b The arm64 was missing proper codegen for CFI, thus sizes were off. Original change's description: > Reland "[deoptimizer] Change deopt entries into builtins" > > This is a reland of 7f58ced72eb65b6b5530ccabaf2eaebe45bf9d33 > > It fixes the different exit size emitted on x64/Atom CPUs due to > performance tuning in TurboAssembler::Call. Additionally, add > cctests to verify the fixed size exits. > > Original change's description: > > [deoptimizer] Change deopt entries into builtins > > > > While the overall goal of this commit is to change deoptimization > > entries into builtins, there are multiple related things happening: > > > > - Deoptimization entries, formerly stubs (i.e. Code objects generated > > at runtime, guaranteed to be immovable), have been converted into > > builtins. The major restriction is that we now need to preserve the > > kRootRegister, which was formerly used on most architectures to pass > > the deoptimization id. The solution differs based on platform. > > - Renamed DEOPT_ENTRIES_OR_FOR_TESTING code kind to FOR_TESTING. > > - Removed heap/ support for immovable Code generation. > > - Removed the DeserializerData class (no longer needed). > > - arm64: to preserve 4-byte deopt exits, introduced a new optimization > > in which the final jump to the deoptimization entry is generated > > once per Code object, and deopt exits can continue to emit a > > near-call. > > - arm,ia32,x64: change to fixed-size deopt exits. This reduces exit > > sizes by 4/8, 5, and 5 bytes, respectively. > > > > On arm the deopt exit size is reduced from 12 (or 16) bytes to 8 bytes > > by using the same strategy as on arm64 (recalc deopt id from return > > address). Before: > > > > e300a002 movw r10, <id> > > e59fc024 ldr ip, [pc, <entry offset>] > > e12fff3c blx ip > > > > After: > > > > e59acb35 ldr ip, [r10, <entry offset>] > > e12fff3c blx ip > > > > On arm64 the deopt exit size remains 4 bytes (or 8 bytes in same cases > > with CFI). Additionally, up to 4 builtin jumps are emitted per Code > > object (max 32 bytes added overhead per Code object). Before: > > > > 9401cdae bl <entry offset> > > > > After: > > > > # eager deoptimization entry jump. > > f95b1f50 ldr x16, [x26, <eager entry offset>] > > d61f0200 br x16 > > # lazy deoptimization entry jump. > > f95b2b50 ldr x16, [x26, <lazy entry offset>] > > d61f0200 br x16 > > # the deopt exit. > > 97fffffc bl <eager deoptimization entry jump offset> > > > > On ia32 the deopt exit size is reduced from 10 to 5 bytes. Before: > > > > bb00000000 mov ebx,<id> > > e825f5372b call <entry> > > > > After: > > > > e8ea2256ba call <entry> > > > > On x64 the deopt exit size is reduced from 12 to 7 bytes. Before: > > > > 49c7c511000000 REX.W movq r13,<id> > > e8ea2f0700 call <entry> > > > > After: > > > > 41ff9560360000 call [r13+<entry offset>] > > > > Bug: v8:8661,v8:8768 > > Change-Id: I13e30aedc360474dc818fecc528ce87c3bfeed42 > > Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2465834 > > Commit-Queue: Jakob Gruber <jgruber@chromium.org> > > Reviewed-by: Ross McIlroy <rmcilroy@chromium.org> > > Reviewed-by: Tobias Tebbi <tebbi@chromium.org> > > Reviewed-by: Ulan Degenbaev <ulan@chromium.org> > > Cr-Commit-Position: refs/heads/master@{#70597} > > Tbr: ulan@chromium.org, tebbi@chromium.org, rmcilroy@chromium.org > Bug: v8:8661,v8:8768,chromium:1140165 > Change-Id: Ibcd5c39c58a70bf2b2ac221aa375fc68d495e144 > Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2485506 > Reviewed-by: Jakob Gruber <jgruber@chromium.org> > Reviewed-by: Tobias Tebbi <tebbi@chromium.org> > Commit-Queue: Jakob Gruber <jgruber@chromium.org> > Cr-Commit-Position: refs/heads/master@{#70655} Tbr: ulan@chromium.org, tebbi@chromium.org, rmcilroy@chromium.org Bug: v8:8661 Bug: v8:8768 Bug: chromium:1140165 Change-Id: I471cc94fc085e527dc9bfb5a84b96bd907c2333f Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2488682 Reviewed-by: Jakob Gruber <jgruber@chromium.org> Commit-Queue: Jakob Gruber <jgruber@chromium.org> Cr-Commit-Position: refs/heads/master@{#70672}
2020-10-21 05:12:25 +00:00
Handle<Code> code =
Factory::CodeBuilder(isolate, desc, CodeKind::FOR_TESTING).Build();
#ifdef OBJECT_PRINT
StdoutStream os;
code->Print(os);
#endif
auto f = GeneratedCode<F6>::FromCode(*code);
CHECK_EQ(2, f.Call(1.0, 2.0));
}
TEST(AssemblerX64SSE3) {
CcTest::InitializeVM();
if (!CpuFeatures::IsSupported(SSE3)) return;
Isolate* isolate = reinterpret_cast<Isolate*>(CcTest::isolate());
HandleScope scope(isolate);
v8::internal::byte buffer[256];
MacroAssembler masm(isolate, v8::internal::CodeObjectRequired::kYes,
ExternalAssemblerBuffer(buffer, sizeof(buffer)));
{
CpuFeatureScope fscope(&masm, SSE3);
__ shufps(xmm0, xmm0, 0x0); // brocast first argument
__ shufps(xmm1, xmm1, 0x0); // brocast second argument
__ haddps(xmm1, xmm0);
__ cvttss2si(rax, xmm1);
__ ret(0);
}
CodeDesc desc;
masm.GetCode(isolate, &desc);
Reland "Reland "[deoptimizer] Change deopt entries into builtins"" This is a reland of fbfa9bf4ec72b1b73a96b70ccb68cd98c321511b The arm64 was missing proper codegen for CFI, thus sizes were off. Original change's description: > Reland "[deoptimizer] Change deopt entries into builtins" > > This is a reland of 7f58ced72eb65b6b5530ccabaf2eaebe45bf9d33 > > It fixes the different exit size emitted on x64/Atom CPUs due to > performance tuning in TurboAssembler::Call. Additionally, add > cctests to verify the fixed size exits. > > Original change's description: > > [deoptimizer] Change deopt entries into builtins > > > > While the overall goal of this commit is to change deoptimization > > entries into builtins, there are multiple related things happening: > > > > - Deoptimization entries, formerly stubs (i.e. Code objects generated > > at runtime, guaranteed to be immovable), have been converted into > > builtins. The major restriction is that we now need to preserve the > > kRootRegister, which was formerly used on most architectures to pass > > the deoptimization id. The solution differs based on platform. > > - Renamed DEOPT_ENTRIES_OR_FOR_TESTING code kind to FOR_TESTING. > > - Removed heap/ support for immovable Code generation. > > - Removed the DeserializerData class (no longer needed). > > - arm64: to preserve 4-byte deopt exits, introduced a new optimization > > in which the final jump to the deoptimization entry is generated > > once per Code object, and deopt exits can continue to emit a > > near-call. > > - arm,ia32,x64: change to fixed-size deopt exits. This reduces exit > > sizes by 4/8, 5, and 5 bytes, respectively. > > > > On arm the deopt exit size is reduced from 12 (or 16) bytes to 8 bytes > > by using the same strategy as on arm64 (recalc deopt id from return > > address). Before: > > > > e300a002 movw r10, <id> > > e59fc024 ldr ip, [pc, <entry offset>] > > e12fff3c blx ip > > > > After: > > > > e59acb35 ldr ip, [r10, <entry offset>] > > e12fff3c blx ip > > > > On arm64 the deopt exit size remains 4 bytes (or 8 bytes in same cases > > with CFI). Additionally, up to 4 builtin jumps are emitted per Code > > object (max 32 bytes added overhead per Code object). Before: > > > > 9401cdae bl <entry offset> > > > > After: > > > > # eager deoptimization entry jump. > > f95b1f50 ldr x16, [x26, <eager entry offset>] > > d61f0200 br x16 > > # lazy deoptimization entry jump. > > f95b2b50 ldr x16, [x26, <lazy entry offset>] > > d61f0200 br x16 > > # the deopt exit. > > 97fffffc bl <eager deoptimization entry jump offset> > > > > On ia32 the deopt exit size is reduced from 10 to 5 bytes. Before: > > > > bb00000000 mov ebx,<id> > > e825f5372b call <entry> > > > > After: > > > > e8ea2256ba call <entry> > > > > On x64 the deopt exit size is reduced from 12 to 7 bytes. Before: > > > > 49c7c511000000 REX.W movq r13,<id> > > e8ea2f0700 call <entry> > > > > After: > > > > 41ff9560360000 call [r13+<entry offset>] > > > > Bug: v8:8661,v8:8768 > > Change-Id: I13e30aedc360474dc818fecc528ce87c3bfeed42 > > Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2465834 > > Commit-Queue: Jakob Gruber <jgruber@chromium.org> > > Reviewed-by: Ross McIlroy <rmcilroy@chromium.org> > > Reviewed-by: Tobias Tebbi <tebbi@chromium.org> > > Reviewed-by: Ulan Degenbaev <ulan@chromium.org> > > Cr-Commit-Position: refs/heads/master@{#70597} > > Tbr: ulan@chromium.org, tebbi@chromium.org, rmcilroy@chromium.org > Bug: v8:8661,v8:8768,chromium:1140165 > Change-Id: Ibcd5c39c58a70bf2b2ac221aa375fc68d495e144 > Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2485506 > Reviewed-by: Jakob Gruber <jgruber@chromium.org> > Reviewed-by: Tobias Tebbi <tebbi@chromium.org> > Commit-Queue: Jakob Gruber <jgruber@chromium.org> > Cr-Commit-Position: refs/heads/master@{#70655} Tbr: ulan@chromium.org, tebbi@chromium.org, rmcilroy@chromium.org Bug: v8:8661 Bug: v8:8768 Bug: chromium:1140165 Change-Id: I471cc94fc085e527dc9bfb5a84b96bd907c2333f Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2488682 Reviewed-by: Jakob Gruber <jgruber@chromium.org> Commit-Queue: Jakob Gruber <jgruber@chromium.org> Cr-Commit-Position: refs/heads/master@{#70672}
2020-10-21 05:12:25 +00:00
Handle<Code> code =
Factory::CodeBuilder(isolate, desc, CodeKind::FOR_TESTING).Build();
#ifdef OBJECT_PRINT
StdoutStream os;
code->Print(os);
#endif
auto f = GeneratedCode<F6>::FromCode(*code);
CHECK_EQ(4, f.Call(1.0, 2.0));
}
using F7 = int(double x, double y, double z);
TEST(AssemblerX64FMA_sd) {
CcTest::InitializeVM();
if (!CpuFeatures::IsSupported(FMA3)) return;
Isolate* isolate = reinterpret_cast<Isolate*>(CcTest::isolate());
HandleScope scope(isolate);
v8::internal::byte buffer[1024];
MacroAssembler masm(isolate, v8::internal::CodeObjectRequired::kYes,
ExternalAssemblerBuffer(buffer, sizeof(buffer)));
{
Reland "[Memory] Use OS::Allocate for all OS memory allocations." This is a reland of 4899bcb66db1c8c8272b02fb1cedf9886e010f82 This is a reland of b73ee3344ac96f92b4418c3c07779eb95a8541cc Original change's description: > [Memory] Use OS::Allocate for all OS memory allocations. > > - Eliminates OS::ReserveRegion and OS::ReserveAlignedRegion. > - Changes OS::Allocate to take alignment parameter, reorders parameters > to match page_allocator. > - Since the size of memory allocation can be deduced, don't return the > amount of memory allocated. > - Changes reservation of aligned address space. Before we would reserve > (size + alignment) rounded up to page size. This is too much, because > maximum misalignment is (alignment - page_size). > - On Windows and Cygwin, we release an oversize allocation and > immediately retry at the aligned address in the allocation. If we > lose the address due to a race, we just retry. > - Clean up all the calls to OS::Allocate in codegen and tests by adding > helper AllocateSystemPage function (allocation.h) and > AllocateAssemblerBuffer (cctest.h). > - Changes 'assm' to 'masm' in some targets for consistency when using > a macro-assembler. > > - Eliminates OS::ReleaseRegion, replacing with calls to OS::Free. > - Adds bool return value to OS::Free. > - Cleans up types of flags, protection on Windows and Cygwin. > Bug: chromium:756050 > Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng > Change-Id: I306dbe042cc867670fdc935abca29db074b0da71 Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng Change-Id: Iad3c025334e8f8d7d647be99a36a11ee449c9087 Reviewed-on: https://chromium-review.googlesource.com/767014 Commit-Queue: Bill Budge <bbudge@chromium.org> Reviewed-by: Michael Lippautz <mlippautz@chromium.org> Cr-Commit-Position: refs/heads/master@{#49363}
2017-11-14 15:55:09 +00:00
CpuFeatureScope fscope(&masm, FMA3);
Label exit;
// argument in xmm0, xmm1 and xmm2
// xmm0 * xmm1 + xmm2
__ movaps(xmm3, xmm0);
__ mulsd(xmm3, xmm1);
__ addsd(xmm3, xmm2); // Expected result in xmm3
__ AllocateStackSpace(kDoubleSize); // For memory operand
// vfmadd132sd
__ movl(rax, Immediate(1)); // Test number
__ movaps(xmm8, xmm0);
__ vfmadd132sd(xmm8, xmm2, xmm1);
__ ucomisd(xmm8, xmm3);
__ j(not_equal, &exit);
// vfmadd213sd
__ incq(rax);
__ movaps(xmm8, xmm1);
__ vfmadd213sd(xmm8, xmm0, xmm2);
__ ucomisd(xmm8, xmm3);
__ j(not_equal, &exit);
// vfmadd231sd
__ incq(rax);
__ movaps(xmm8, xmm2);
__ vfmadd231sd(xmm8, xmm0, xmm1);
__ ucomisd(xmm8, xmm3);
__ j(not_equal, &exit);
// vfmadd132sd
__ incq(rax);
__ movaps(xmm8, xmm0);
__ movsd(Operand(rsp, 0), xmm1);
__ vfmadd132sd(xmm8, xmm2, Operand(rsp, 0));
__ ucomisd(xmm8, xmm3);
__ j(not_equal, &exit);
// vfmadd213sd
__ incq(rax);
__ movaps(xmm8, xmm1);
__ movsd(Operand(rsp, 0), xmm2);
__ vfmadd213sd(xmm8, xmm0, Operand(rsp, 0));
__ ucomisd(xmm8, xmm3);
__ j(not_equal, &exit);
// vfmadd231sd
__ incq(rax);
__ movaps(xmm8, xmm2);
__ movsd(Operand(rsp, 0), xmm1);
__ vfmadd231sd(xmm8, xmm0, Operand(rsp, 0));
__ ucomisd(xmm8, xmm3);
__ j(not_equal, &exit);
// xmm0 * xmm1 - xmm2
__ movaps(xmm3, xmm0);
__ mulsd(xmm3, xmm1);
__ subsd(xmm3, xmm2); // Expected result in xmm3
// vfmsub132sd
__ incq(rax);
__ movaps(xmm8, xmm0);
__ vfmsub132sd(xmm8, xmm2, xmm1);
__ ucomisd(xmm8, xmm3);
__ j(not_equal, &exit);
// vfmadd213sd
__ incq(rax);
__ movaps(xmm8, xmm1);
__ vfmsub213sd(xmm8, xmm0, xmm2);
__ ucomisd(xmm8, xmm3);
__ j(not_equal, &exit);
// vfmsub231sd
__ incq(rax);
__ movaps(xmm8, xmm2);
__ vfmsub231sd(xmm8, xmm0, xmm1);
__ ucomisd(xmm8, xmm3);
__ j(not_equal, &exit);
// vfmsub132sd
__ incq(rax);
__ movaps(xmm8, xmm0);
__ movsd(Operand(rsp, 0), xmm1);
__ vfmsub132sd(xmm8, xmm2, Operand(rsp, 0));
__ ucomisd(xmm8, xmm3);
__ j(not_equal, &exit);
// vfmsub213sd
__ incq(rax);
__ movaps(xmm8, xmm1);
__ movsd(Operand(rsp, 0), xmm2);
__ vfmsub213sd(xmm8, xmm0, Operand(rsp, 0));
__ ucomisd(xmm8, xmm3);
__ j(not_equal, &exit);
// vfmsub231sd
__ incq(rax);
__ movaps(xmm8, xmm2);
__ movsd(Operand(rsp, 0), xmm1);
__ vfmsub231sd(xmm8, xmm0, Operand(rsp, 0));
__ ucomisd(xmm8, xmm3);
__ j(not_equal, &exit);
// - xmm0 * xmm1 + xmm2
__ movaps(xmm3, xmm0);
__ mulsd(xmm3, xmm1);
__ Move(xmm4, static_cast<uint64_t>(1) << 63);
__ xorpd(xmm3, xmm4);
__ addsd(xmm3, xmm2); // Expected result in xmm3
// vfnmadd132sd
__ incq(rax);
__ movaps(xmm8, xmm0);
__ vfnmadd132sd(xmm8, xmm2, xmm1);
__ ucomisd(xmm8, xmm3);
__ j(not_equal, &exit);
// vfmadd213sd
__ incq(rax);
__ movaps(xmm8, xmm1);
__ vfnmadd213sd(xmm8, xmm0, xmm2);
__ ucomisd(xmm8, xmm3);
__ j(not_equal, &exit);
// vfnmadd231sd
__ incq(rax);
__ movaps(xmm8, xmm2);
__ vfnmadd231sd(xmm8, xmm0, xmm1);
__ ucomisd(xmm8, xmm3);
__ j(not_equal, &exit);
// vfnmadd132sd
__ incq(rax);
__ movaps(xmm8, xmm0);
__ movsd(Operand(rsp, 0), xmm1);
__ vfnmadd132sd(xmm8, xmm2, Operand(rsp, 0));
__ ucomisd(xmm8, xmm3);
__ j(not_equal, &exit);
// vfnmadd213sd
__ incq(rax);
__ movaps(xmm8, xmm1);
__ movsd(Operand(rsp, 0), xmm2);
__ vfnmadd213sd(xmm8, xmm0, Operand(rsp, 0));
__ ucomisd(xmm8, xmm3);
__ j(not_equal, &exit);
// vfnmadd231sd
__ incq(rax);
__ movaps(xmm8, xmm2);
__ movsd(Operand(rsp, 0), xmm1);
__ vfnmadd231sd(xmm8, xmm0, Operand(rsp, 0));
__ ucomisd(xmm8, xmm3);
__ j(not_equal, &exit);
// - xmm0 * xmm1 - xmm2
__ movaps(xmm3, xmm0);
__ mulsd(xmm3, xmm1);
__ Move(xmm4, static_cast<uint64_t>(1) << 63);
__ xorpd(xmm3, xmm4);
__ subsd(xmm3, xmm2); // Expected result in xmm3
// vfnmsub132sd
__ incq(rax);
__ movaps(xmm8, xmm0);
__ vfnmsub132sd(xmm8, xmm2, xmm1);
__ ucomisd(xmm8, xmm3);
__ j(not_equal, &exit);
// vfmsub213sd
__ incq(rax);
__ movaps(xmm8, xmm1);
__ vfnmsub213sd(xmm8, xmm0, xmm2);
__ ucomisd(xmm8, xmm3);
__ j(not_equal, &exit);
// vfnmsub231sd
__ incq(rax);
__ movaps(xmm8, xmm2);
__ vfnmsub231sd(xmm8, xmm0, xmm1);
__ ucomisd(xmm8, xmm3);
__ j(not_equal, &exit);
// vfnmsub132sd
__ incq(rax);
__ movaps(xmm8, xmm0);
__ movsd(Operand(rsp, 0), xmm1);
__ vfnmsub132sd(xmm8, xmm2, Operand(rsp, 0));
__ ucomisd(xmm8, xmm3);
__ j(not_equal, &exit);
// vfnmsub213sd
__ incq(rax);
__ movaps(xmm8, xmm1);
__ movsd(Operand(rsp, 0), xmm2);
__ vfnmsub213sd(xmm8, xmm0, Operand(rsp, 0));
__ ucomisd(xmm8, xmm3);
__ j(not_equal, &exit);
// vfnmsub231sd
__ incq(rax);
__ movaps(xmm8, xmm2);
__ movsd(Operand(rsp, 0), xmm1);
__ vfnmsub231sd(xmm8, xmm0, Operand(rsp, 0));
__ ucomisd(xmm8, xmm3);
__ j(not_equal, &exit);
__ xorl(rax, rax);
__ bind(&exit);
__ addq(rsp, Immediate(kDoubleSize));
__ ret(0);
}
CodeDesc desc;
Reland "[Memory] Use OS::Allocate for all OS memory allocations." This is a reland of 4899bcb66db1c8c8272b02fb1cedf9886e010f82 This is a reland of b73ee3344ac96f92b4418c3c07779eb95a8541cc Original change's description: > [Memory] Use OS::Allocate for all OS memory allocations. > > - Eliminates OS::ReserveRegion and OS::ReserveAlignedRegion. > - Changes OS::Allocate to take alignment parameter, reorders parameters > to match page_allocator. > - Since the size of memory allocation can be deduced, don't return the > amount of memory allocated. > - Changes reservation of aligned address space. Before we would reserve > (size + alignment) rounded up to page size. This is too much, because > maximum misalignment is (alignment - page_size). > - On Windows and Cygwin, we release an oversize allocation and > immediately retry at the aligned address in the allocation. If we > lose the address due to a race, we just retry. > - Clean up all the calls to OS::Allocate in codegen and tests by adding > helper AllocateSystemPage function (allocation.h) and > AllocateAssemblerBuffer (cctest.h). > - Changes 'assm' to 'masm' in some targets for consistency when using > a macro-assembler. > > - Eliminates OS::ReleaseRegion, replacing with calls to OS::Free. > - Adds bool return value to OS::Free. > - Cleans up types of flags, protection on Windows and Cygwin. > Bug: chromium:756050 > Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng > Change-Id: I306dbe042cc867670fdc935abca29db074b0da71 Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng Change-Id: Iad3c025334e8f8d7d647be99a36a11ee449c9087 Reviewed-on: https://chromium-review.googlesource.com/767014 Commit-Queue: Bill Budge <bbudge@chromium.org> Reviewed-by: Michael Lippautz <mlippautz@chromium.org> Cr-Commit-Position: refs/heads/master@{#49363}
2017-11-14 15:55:09 +00:00
masm.GetCode(isolate, &desc);
Reland "Reland "[deoptimizer] Change deopt entries into builtins"" This is a reland of fbfa9bf4ec72b1b73a96b70ccb68cd98c321511b The arm64 was missing proper codegen for CFI, thus sizes were off. Original change's description: > Reland "[deoptimizer] Change deopt entries into builtins" > > This is a reland of 7f58ced72eb65b6b5530ccabaf2eaebe45bf9d33 > > It fixes the different exit size emitted on x64/Atom CPUs due to > performance tuning in TurboAssembler::Call. Additionally, add > cctests to verify the fixed size exits. > > Original change's description: > > [deoptimizer] Change deopt entries into builtins > > > > While the overall goal of this commit is to change deoptimization > > entries into builtins, there are multiple related things happening: > > > > - Deoptimization entries, formerly stubs (i.e. Code objects generated > > at runtime, guaranteed to be immovable), have been converted into > > builtins. The major restriction is that we now need to preserve the > > kRootRegister, which was formerly used on most architectures to pass > > the deoptimization id. The solution differs based on platform. > > - Renamed DEOPT_ENTRIES_OR_FOR_TESTING code kind to FOR_TESTING. > > - Removed heap/ support for immovable Code generation. > > - Removed the DeserializerData class (no longer needed). > > - arm64: to preserve 4-byte deopt exits, introduced a new optimization > > in which the final jump to the deoptimization entry is generated > > once per Code object, and deopt exits can continue to emit a > > near-call. > > - arm,ia32,x64: change to fixed-size deopt exits. This reduces exit > > sizes by 4/8, 5, and 5 bytes, respectively. > > > > On arm the deopt exit size is reduced from 12 (or 16) bytes to 8 bytes > > by using the same strategy as on arm64 (recalc deopt id from return > > address). Before: > > > > e300a002 movw r10, <id> > > e59fc024 ldr ip, [pc, <entry offset>] > > e12fff3c blx ip > > > > After: > > > > e59acb35 ldr ip, [r10, <entry offset>] > > e12fff3c blx ip > > > > On arm64 the deopt exit size remains 4 bytes (or 8 bytes in same cases > > with CFI). Additionally, up to 4 builtin jumps are emitted per Code > > object (max 32 bytes added overhead per Code object). Before: > > > > 9401cdae bl <entry offset> > > > > After: > > > > # eager deoptimization entry jump. > > f95b1f50 ldr x16, [x26, <eager entry offset>] > > d61f0200 br x16 > > # lazy deoptimization entry jump. > > f95b2b50 ldr x16, [x26, <lazy entry offset>] > > d61f0200 br x16 > > # the deopt exit. > > 97fffffc bl <eager deoptimization entry jump offset> > > > > On ia32 the deopt exit size is reduced from 10 to 5 bytes. Before: > > > > bb00000000 mov ebx,<id> > > e825f5372b call <entry> > > > > After: > > > > e8ea2256ba call <entry> > > > > On x64 the deopt exit size is reduced from 12 to 7 bytes. Before: > > > > 49c7c511000000 REX.W movq r13,<id> > > e8ea2f0700 call <entry> > > > > After: > > > > 41ff9560360000 call [r13+<entry offset>] > > > > Bug: v8:8661,v8:8768 > > Change-Id: I13e30aedc360474dc818fecc528ce87c3bfeed42 > > Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2465834 > > Commit-Queue: Jakob Gruber <jgruber@chromium.org> > > Reviewed-by: Ross McIlroy <rmcilroy@chromium.org> > > Reviewed-by: Tobias Tebbi <tebbi@chromium.org> > > Reviewed-by: Ulan Degenbaev <ulan@chromium.org> > > Cr-Commit-Position: refs/heads/master@{#70597} > > Tbr: ulan@chromium.org, tebbi@chromium.org, rmcilroy@chromium.org > Bug: v8:8661,v8:8768,chromium:1140165 > Change-Id: Ibcd5c39c58a70bf2b2ac221aa375fc68d495e144 > Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2485506 > Reviewed-by: Jakob Gruber <jgruber@chromium.org> > Reviewed-by: Tobias Tebbi <tebbi@chromium.org> > Commit-Queue: Jakob Gruber <jgruber@chromium.org> > Cr-Commit-Position: refs/heads/master@{#70655} Tbr: ulan@chromium.org, tebbi@chromium.org, rmcilroy@chromium.org Bug: v8:8661 Bug: v8:8768 Bug: chromium:1140165 Change-Id: I471cc94fc085e527dc9bfb5a84b96bd907c2333f Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2488682 Reviewed-by: Jakob Gruber <jgruber@chromium.org> Commit-Queue: Jakob Gruber <jgruber@chromium.org> Cr-Commit-Position: refs/heads/master@{#70672}
2020-10-21 05:12:25 +00:00
Handle<Code> code =
Factory::CodeBuilder(isolate, desc, CodeKind::FOR_TESTING).Build();
#ifdef OBJECT_PRINT
StdoutStream os;
code->Print(os);
#endif
auto f = GeneratedCode<F7>::FromCode(*code);
CHECK_EQ(
0, f.Call(0.000092662107262076, -2.460774966188315, -1.0958787393627414));
}
using F8 = int(float x, float y, float z);
TEST(AssemblerX64FMA_ss) {
CcTest::InitializeVM();
if (!CpuFeatures::IsSupported(FMA3)) return;
Isolate* isolate = reinterpret_cast<Isolate*>(CcTest::isolate());
HandleScope scope(isolate);
v8::internal::byte buffer[1024];
MacroAssembler masm(isolate, v8::internal::CodeObjectRequired::kYes,
ExternalAssemblerBuffer(buffer, sizeof(buffer)));
{
Reland "[Memory] Use OS::Allocate for all OS memory allocations." This is a reland of 4899bcb66db1c8c8272b02fb1cedf9886e010f82 This is a reland of b73ee3344ac96f92b4418c3c07779eb95a8541cc Original change's description: > [Memory] Use OS::Allocate for all OS memory allocations. > > - Eliminates OS::ReserveRegion and OS::ReserveAlignedRegion. > - Changes OS::Allocate to take alignment parameter, reorders parameters > to match page_allocator. > - Since the size of memory allocation can be deduced, don't return the > amount of memory allocated. > - Changes reservation of aligned address space. Before we would reserve > (size + alignment) rounded up to page size. This is too much, because > maximum misalignment is (alignment - page_size). > - On Windows and Cygwin, we release an oversize allocation and > immediately retry at the aligned address in the allocation. If we > lose the address due to a race, we just retry. > - Clean up all the calls to OS::Allocate in codegen and tests by adding > helper AllocateSystemPage function (allocation.h) and > AllocateAssemblerBuffer (cctest.h). > - Changes 'assm' to 'masm' in some targets for consistency when using > a macro-assembler. > > - Eliminates OS::ReleaseRegion, replacing with calls to OS::Free. > - Adds bool return value to OS::Free. > - Cleans up types of flags, protection on Windows and Cygwin. > Bug: chromium:756050 > Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng > Change-Id: I306dbe042cc867670fdc935abca29db074b0da71 Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng Change-Id: Iad3c025334e8f8d7d647be99a36a11ee449c9087 Reviewed-on: https://chromium-review.googlesource.com/767014 Commit-Queue: Bill Budge <bbudge@chromium.org> Reviewed-by: Michael Lippautz <mlippautz@chromium.org> Cr-Commit-Position: refs/heads/master@{#49363}
2017-11-14 15:55:09 +00:00
CpuFeatureScope fscope(&masm, FMA3);
Label exit;
// arguments in xmm0, xmm1 and xmm2
// xmm0 * xmm1 + xmm2
__ movaps(xmm3, xmm0);
__ mulss(xmm3, xmm1);
__ addss(xmm3, xmm2); // Expected result in xmm3
__ AllocateStackSpace(kDoubleSize); // For memory operand
// vfmadd132ss
__ movl(rax, Immediate(1)); // Test number
__ movaps(xmm8, xmm0);
__ vfmadd132ss(xmm8, xmm2, xmm1);
__ ucomiss(xmm8, xmm3);
__ j(not_equal, &exit);
// vfmadd213ss
__ incq(rax);
__ movaps(xmm8, xmm1);
__ vfmadd213ss(xmm8, xmm0, xmm2);
__ ucomiss(xmm8, xmm3);
__ j(not_equal, &exit);
// vfmadd231ss
__ incq(rax);
__ movaps(xmm8, xmm2);
__ vfmadd231ss(xmm8, xmm0, xmm1);
__ ucomiss(xmm8, xmm3);
__ j(not_equal, &exit);
// vfmadd132ss
__ incq(rax);
__ movaps(xmm8, xmm0);
__ movss(Operand(rsp, 0), xmm1);
__ vfmadd132ss(xmm8, xmm2, Operand(rsp, 0));
__ ucomiss(xmm8, xmm3);
__ j(not_equal, &exit);
// vfmadd213ss
__ incq(rax);
__ movaps(xmm8, xmm1);
__ movss(Operand(rsp, 0), xmm2);
__ vfmadd213ss(xmm8, xmm0, Operand(rsp, 0));
__ ucomiss(xmm8, xmm3);
__ j(not_equal, &exit);
// vfmadd231ss
__ incq(rax);
__ movaps(xmm8, xmm2);
__ movss(Operand(rsp, 0), xmm1);
__ vfmadd231ss(xmm8, xmm0, Operand(rsp, 0));
__ ucomiss(xmm8, xmm3);
__ j(not_equal, &exit);
// xmm0 * xmm1 - xmm2
__ movaps(xmm3, xmm0);
__ mulss(xmm3, xmm1);
__ subss(xmm3, xmm2); // Expected result in xmm3
// vfmsub132ss
__ incq(rax);
__ movaps(xmm8, xmm0);
__ vfmsub132ss(xmm8, xmm2, xmm1);
__ ucomiss(xmm8, xmm3);
__ j(not_equal, &exit);
// vfmadd213ss
__ incq(rax);
__ movaps(xmm8, xmm1);
__ vfmsub213ss(xmm8, xmm0, xmm2);
__ ucomiss(xmm8, xmm3);
__ j(not_equal, &exit);
// vfmsub231ss
__ incq(rax);
__ movaps(xmm8, xmm2);
__ vfmsub231ss(xmm8, xmm0, xmm1);
__ ucomiss(xmm8, xmm3);
__ j(not_equal, &exit);
// vfmsub132ss
__ incq(rax);
__ movaps(xmm8, xmm0);
__ movss(Operand(rsp, 0), xmm1);
__ vfmsub132ss(xmm8, xmm2, Operand(rsp, 0));
__ ucomiss(xmm8, xmm3);
__ j(not_equal, &exit);
// vfmsub213ss
__ incq(rax);
__ movaps(xmm8, xmm1);
__ movss(Operand(rsp, 0), xmm2);
__ vfmsub213ss(xmm8, xmm0, Operand(rsp, 0));
__ ucomiss(xmm8, xmm3);
__ j(not_equal, &exit);
// vfmsub231ss
__ incq(rax);
__ movaps(xmm8, xmm2);
__ movss(Operand(rsp, 0), xmm1);
__ vfmsub231ss(xmm8, xmm0, Operand(rsp, 0));
__ ucomiss(xmm8, xmm3);
__ j(not_equal, &exit);
// - xmm0 * xmm1 + xmm2
__ movaps(xmm3, xmm0);
__ mulss(xmm3, xmm1);
__ Move(xmm4, static_cast<uint32_t>(1) << 31);
__ xorps(xmm3, xmm4);
__ addss(xmm3, xmm2); // Expected result in xmm3
// vfnmadd132ss
__ incq(rax);
__ movaps(xmm8, xmm0);
__ vfnmadd132ss(xmm8, xmm2, xmm1);
__ ucomiss(xmm8, xmm3);
__ j(not_equal, &exit);
// vfmadd213ss
__ incq(rax);
__ movaps(xmm8, xmm1);
__ vfnmadd213ss(xmm8, xmm0, xmm2);
__ ucomiss(xmm8, xmm3);
__ j(not_equal, &exit);
// vfnmadd231ss
__ incq(rax);
__ movaps(xmm8, xmm2);
__ vfnmadd231ss(xmm8, xmm0, xmm1);
__ ucomiss(xmm8, xmm3);
__ j(not_equal, &exit);
// vfnmadd132ss
__ incq(rax);
__ movaps(xmm8, xmm0);
__ movss(Operand(rsp, 0), xmm1);
__ vfnmadd132ss(xmm8, xmm2, Operand(rsp, 0));
__ ucomiss(xmm8, xmm3);
__ j(not_equal, &exit);
// vfnmadd213ss
__ incq(rax);
__ movaps(xmm8, xmm1);
__ movss(Operand(rsp, 0), xmm2);
__ vfnmadd213ss(xmm8, xmm0, Operand(rsp, 0));
__ ucomiss(xmm8, xmm3);
__ j(not_equal, &exit);
// vfnmadd231ss
__ incq(rax);
__ movaps(xmm8, xmm2);
__ movss(Operand(rsp, 0), xmm1);
__ vfnmadd231ss(xmm8, xmm0, Operand(rsp, 0));
__ ucomiss(xmm8, xmm3);
__ j(not_equal, &exit);
// - xmm0 * xmm1 - xmm2
__ movaps(xmm3, xmm0);
__ mulss(xmm3, xmm1);
__ Move(xmm4, static_cast<uint32_t>(1) << 31);
__ xorps(xmm3, xmm4);
__ subss(xmm3, xmm2); // Expected result in xmm3
// vfnmsub132ss
__ incq(rax);
__ movaps(xmm8, xmm0);
__ vfnmsub132ss(xmm8, xmm2, xmm1);
__ ucomiss(xmm8, xmm3);
__ j(not_equal, &exit);
// vfmsub213ss
__ incq(rax);
__ movaps(xmm8, xmm1);
__ vfnmsub213ss(xmm8, xmm0, xmm2);
__ ucomiss(xmm8, xmm3);
__ j(not_equal, &exit);
// vfnmsub231ss
__ incq(rax);
__ movaps(xmm8, xmm2);
__ vfnmsub231ss(xmm8, xmm0, xmm1);
__ ucomiss(xmm8, xmm3);
__ j(not_equal, &exit);
// vfnmsub132ss
__ incq(rax);
__ movaps(xmm8, xmm0);
__ movss(Operand(rsp, 0), xmm1);
__ vfnmsub132ss(xmm8, xmm2, Operand(rsp, 0));
__ ucomiss(xmm8, xmm3);
__ j(not_equal, &exit);
// vfnmsub213ss
__ incq(rax);
__ movaps(xmm8, xmm1);
__ movss(Operand(rsp, 0), xmm2);
__ vfnmsub213ss(xmm8, xmm0, Operand(rsp, 0));
__ ucomiss(xmm8, xmm3);
__ j(not_equal, &exit);
// vfnmsub231ss
__ incq(rax);
__ movaps(xmm8, xmm2);
__ movss(Operand(rsp, 0), xmm1);
__ vfnmsub231ss(xmm8, xmm0, Operand(rsp, 0));
__ ucomiss(xmm8, xmm3);
__ j(not_equal, &exit);
__ xorl(rax, rax);
__ bind(&exit);
__ addq(rsp, Immediate(kDoubleSize));
__ ret(0);
}
CodeDesc desc;
Reland "[Memory] Use OS::Allocate for all OS memory allocations." This is a reland of 4899bcb66db1c8c8272b02fb1cedf9886e010f82 This is a reland of b73ee3344ac96f92b4418c3c07779eb95a8541cc Original change's description: > [Memory] Use OS::Allocate for all OS memory allocations. > > - Eliminates OS::ReserveRegion and OS::ReserveAlignedRegion. > - Changes OS::Allocate to take alignment parameter, reorders parameters > to match page_allocator. > - Since the size of memory allocation can be deduced, don't return the > amount of memory allocated. > - Changes reservation of aligned address space. Before we would reserve > (size + alignment) rounded up to page size. This is too much, because > maximum misalignment is (alignment - page_size). > - On Windows and Cygwin, we release an oversize allocation and > immediately retry at the aligned address in the allocation. If we > lose the address due to a race, we just retry. > - Clean up all the calls to OS::Allocate in codegen and tests by adding > helper AllocateSystemPage function (allocation.h) and > AllocateAssemblerBuffer (cctest.h). > - Changes 'assm' to 'masm' in some targets for consistency when using > a macro-assembler. > > - Eliminates OS::ReleaseRegion, replacing with calls to OS::Free. > - Adds bool return value to OS::Free. > - Cleans up types of flags, protection on Windows and Cygwin. > Bug: chromium:756050 > Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng > Change-Id: I306dbe042cc867670fdc935abca29db074b0da71 Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng Change-Id: Iad3c025334e8f8d7d647be99a36a11ee449c9087 Reviewed-on: https://chromium-review.googlesource.com/767014 Commit-Queue: Bill Budge <bbudge@chromium.org> Reviewed-by: Michael Lippautz <mlippautz@chromium.org> Cr-Commit-Position: refs/heads/master@{#49363}
2017-11-14 15:55:09 +00:00
masm.GetCode(isolate, &desc);
Reland "Reland "[deoptimizer] Change deopt entries into builtins"" This is a reland of fbfa9bf4ec72b1b73a96b70ccb68cd98c321511b The arm64 was missing proper codegen for CFI, thus sizes were off. Original change's description: > Reland "[deoptimizer] Change deopt entries into builtins" > > This is a reland of 7f58ced72eb65b6b5530ccabaf2eaebe45bf9d33 > > It fixes the different exit size emitted on x64/Atom CPUs due to > performance tuning in TurboAssembler::Call. Additionally, add > cctests to verify the fixed size exits. > > Original change's description: > > [deoptimizer] Change deopt entries into builtins > > > > While the overall goal of this commit is to change deoptimization > > entries into builtins, there are multiple related things happening: > > > > - Deoptimization entries, formerly stubs (i.e. Code objects generated > > at runtime, guaranteed to be immovable), have been converted into > > builtins. The major restriction is that we now need to preserve the > > kRootRegister, which was formerly used on most architectures to pass > > the deoptimization id. The solution differs based on platform. > > - Renamed DEOPT_ENTRIES_OR_FOR_TESTING code kind to FOR_TESTING. > > - Removed heap/ support for immovable Code generation. > > - Removed the DeserializerData class (no longer needed). > > - arm64: to preserve 4-byte deopt exits, introduced a new optimization > > in which the final jump to the deoptimization entry is generated > > once per Code object, and deopt exits can continue to emit a > > near-call. > > - arm,ia32,x64: change to fixed-size deopt exits. This reduces exit > > sizes by 4/8, 5, and 5 bytes, respectively. > > > > On arm the deopt exit size is reduced from 12 (or 16) bytes to 8 bytes > > by using the same strategy as on arm64 (recalc deopt id from return > > address). Before: > > > > e300a002 movw r10, <id> > > e59fc024 ldr ip, [pc, <entry offset>] > > e12fff3c blx ip > > > > After: > > > > e59acb35 ldr ip, [r10, <entry offset>] > > e12fff3c blx ip > > > > On arm64 the deopt exit size remains 4 bytes (or 8 bytes in same cases > > with CFI). Additionally, up to 4 builtin jumps are emitted per Code > > object (max 32 bytes added overhead per Code object). Before: > > > > 9401cdae bl <entry offset> > > > > After: > > > > # eager deoptimization entry jump. > > f95b1f50 ldr x16, [x26, <eager entry offset>] > > d61f0200 br x16 > > # lazy deoptimization entry jump. > > f95b2b50 ldr x16, [x26, <lazy entry offset>] > > d61f0200 br x16 > > # the deopt exit. > > 97fffffc bl <eager deoptimization entry jump offset> > > > > On ia32 the deopt exit size is reduced from 10 to 5 bytes. Before: > > > > bb00000000 mov ebx,<id> > > e825f5372b call <entry> > > > > After: > > > > e8ea2256ba call <entry> > > > > On x64 the deopt exit size is reduced from 12 to 7 bytes. Before: > > > > 49c7c511000000 REX.W movq r13,<id> > > e8ea2f0700 call <entry> > > > > After: > > > > 41ff9560360000 call [r13+<entry offset>] > > > > Bug: v8:8661,v8:8768 > > Change-Id: I13e30aedc360474dc818fecc528ce87c3bfeed42 > > Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2465834 > > Commit-Queue: Jakob Gruber <jgruber@chromium.org> > > Reviewed-by: Ross McIlroy <rmcilroy@chromium.org> > > Reviewed-by: Tobias Tebbi <tebbi@chromium.org> > > Reviewed-by: Ulan Degenbaev <ulan@chromium.org> > > Cr-Commit-Position: refs/heads/master@{#70597} > > Tbr: ulan@chromium.org, tebbi@chromium.org, rmcilroy@chromium.org > Bug: v8:8661,v8:8768,chromium:1140165 > Change-Id: Ibcd5c39c58a70bf2b2ac221aa375fc68d495e144 > Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2485506 > Reviewed-by: Jakob Gruber <jgruber@chromium.org> > Reviewed-by: Tobias Tebbi <tebbi@chromium.org> > Commit-Queue: Jakob Gruber <jgruber@chromium.org> > Cr-Commit-Position: refs/heads/master@{#70655} Tbr: ulan@chromium.org, tebbi@chromium.org, rmcilroy@chromium.org Bug: v8:8661 Bug: v8:8768 Bug: chromium:1140165 Change-Id: I471cc94fc085e527dc9bfb5a84b96bd907c2333f Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2488682 Reviewed-by: Jakob Gruber <jgruber@chromium.org> Commit-Queue: Jakob Gruber <jgruber@chromium.org> Cr-Commit-Position: refs/heads/master@{#70672}
2020-10-21 05:12:25 +00:00
Handle<Code> code =
Factory::CodeBuilder(isolate, desc, CodeKind::FOR_TESTING).Build();
#ifdef OBJECT_PRINT
StdoutStream os;
code->Print(os);
#endif
auto f = GeneratedCode<F8>::FromCode(*code);
CHECK_EQ(0, f.Call(9.26621069e-05f, -2.4607749f, -1.09587872f));
}
TEST(AssemblerX64SSE_ss) {
CcTest::InitializeVM();
Isolate* isolate = reinterpret_cast<Isolate*>(CcTest::isolate());
HandleScope scope(isolate);
v8::internal::byte buffer[1024];
Assembler masm(AssemblerOptions{},
ExternalAssemblerBuffer(buffer, sizeof(buffer)));
{
Label exit;
// arguments in xmm0, xmm1 and xmm2
__ movl(rax, Immediate(0));
__ movaps(xmm3, xmm0);
__ maxss(xmm3, xmm1);
__ ucomiss(xmm3, xmm1);
__ j(parity_even, &exit);
__ j(not_equal, &exit);
__ movl(rax, Immediate(1));
__ movaps(xmm3, xmm1);
__ minss(xmm3, xmm2);
__ ucomiss(xmm3, xmm1);
__ j(parity_even, &exit);
__ j(not_equal, &exit);
__ movl(rax, Immediate(2));
__ movaps(xmm3, xmm2);
__ subss(xmm3, xmm1);
__ ucomiss(xmm3, xmm0);
__ j(parity_even, &exit);
__ j(not_equal, &exit);
__ movl(rax, Immediate(3));
__ movaps(xmm3, xmm0);
__ addss(xmm3, xmm1);
__ ucomiss(xmm3, xmm2);
__ j(parity_even, &exit);
__ j(not_equal, &exit);
__ movl(rax, Immediate(4));
__ movaps(xmm3, xmm0);
__ mulss(xmm3, xmm1);
__ ucomiss(xmm3, xmm1);
__ j(parity_even, &exit);
__ j(not_equal, &exit);
__ movl(rax, Immediate(5));
__ movaps(xmm3, xmm0);
__ divss(xmm3, xmm1);
__ mulss(xmm3, xmm2);
__ mulss(xmm3, xmm1);
__ ucomiss(xmm3, xmm2);
__ j(parity_even, &exit);
__ j(not_equal, &exit);
__ movl(rax, Immediate(6));
// result in eax
__ bind(&exit);
__ ret(0);
}
CodeDesc desc;
Reland "[Memory] Use OS::Allocate for all OS memory allocations." This is a reland of 4899bcb66db1c8c8272b02fb1cedf9886e010f82 This is a reland of b73ee3344ac96f92b4418c3c07779eb95a8541cc Original change's description: > [Memory] Use OS::Allocate for all OS memory allocations. > > - Eliminates OS::ReserveRegion and OS::ReserveAlignedRegion. > - Changes OS::Allocate to take alignment parameter, reorders parameters > to match page_allocator. > - Since the size of memory allocation can be deduced, don't return the > amount of memory allocated. > - Changes reservation of aligned address space. Before we would reserve > (size + alignment) rounded up to page size. This is too much, because > maximum misalignment is (alignment - page_size). > - On Windows and Cygwin, we release an oversize allocation and > immediately retry at the aligned address in the allocation. If we > lose the address due to a race, we just retry. > - Clean up all the calls to OS::Allocate in codegen and tests by adding > helper AllocateSystemPage function (allocation.h) and > AllocateAssemblerBuffer (cctest.h). > - Changes 'assm' to 'masm' in some targets for consistency when using > a macro-assembler. > > - Eliminates OS::ReleaseRegion, replacing with calls to OS::Free. > - Adds bool return value to OS::Free. > - Cleans up types of flags, protection on Windows and Cygwin. > Bug: chromium:756050 > Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng > Change-Id: I306dbe042cc867670fdc935abca29db074b0da71 Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng Change-Id: Iad3c025334e8f8d7d647be99a36a11ee449c9087 Reviewed-on: https://chromium-review.googlesource.com/767014 Commit-Queue: Bill Budge <bbudge@chromium.org> Reviewed-by: Michael Lippautz <mlippautz@chromium.org> Cr-Commit-Position: refs/heads/master@{#49363}
2017-11-14 15:55:09 +00:00
masm.GetCode(isolate, &desc);
Reland "Reland "[deoptimizer] Change deopt entries into builtins"" This is a reland of fbfa9bf4ec72b1b73a96b70ccb68cd98c321511b The arm64 was missing proper codegen for CFI, thus sizes were off. Original change's description: > Reland "[deoptimizer] Change deopt entries into builtins" > > This is a reland of 7f58ced72eb65b6b5530ccabaf2eaebe45bf9d33 > > It fixes the different exit size emitted on x64/Atom CPUs due to > performance tuning in TurboAssembler::Call. Additionally, add > cctests to verify the fixed size exits. > > Original change's description: > > [deoptimizer] Change deopt entries into builtins > > > > While the overall goal of this commit is to change deoptimization > > entries into builtins, there are multiple related things happening: > > > > - Deoptimization entries, formerly stubs (i.e. Code objects generated > > at runtime, guaranteed to be immovable), have been converted into > > builtins. The major restriction is that we now need to preserve the > > kRootRegister, which was formerly used on most architectures to pass > > the deoptimization id. The solution differs based on platform. > > - Renamed DEOPT_ENTRIES_OR_FOR_TESTING code kind to FOR_TESTING. > > - Removed heap/ support for immovable Code generation. > > - Removed the DeserializerData class (no longer needed). > > - arm64: to preserve 4-byte deopt exits, introduced a new optimization > > in which the final jump to the deoptimization entry is generated > > once per Code object, and deopt exits can continue to emit a > > near-call. > > - arm,ia32,x64: change to fixed-size deopt exits. This reduces exit > > sizes by 4/8, 5, and 5 bytes, respectively. > > > > On arm the deopt exit size is reduced from 12 (or 16) bytes to 8 bytes > > by using the same strategy as on arm64 (recalc deopt id from return > > address). Before: > > > > e300a002 movw r10, <id> > > e59fc024 ldr ip, [pc, <entry offset>] > > e12fff3c blx ip > > > > After: > > > > e59acb35 ldr ip, [r10, <entry offset>] > > e12fff3c blx ip > > > > On arm64 the deopt exit size remains 4 bytes (or 8 bytes in same cases > > with CFI). Additionally, up to 4 builtin jumps are emitted per Code > > object (max 32 bytes added overhead per Code object). Before: > > > > 9401cdae bl <entry offset> > > > > After: > > > > # eager deoptimization entry jump. > > f95b1f50 ldr x16, [x26, <eager entry offset>] > > d61f0200 br x16 > > # lazy deoptimization entry jump. > > f95b2b50 ldr x16, [x26, <lazy entry offset>] > > d61f0200 br x16 > > # the deopt exit. > > 97fffffc bl <eager deoptimization entry jump offset> > > > > On ia32 the deopt exit size is reduced from 10 to 5 bytes. Before: > > > > bb00000000 mov ebx,<id> > > e825f5372b call <entry> > > > > After: > > > > e8ea2256ba call <entry> > > > > On x64 the deopt exit size is reduced from 12 to 7 bytes. Before: > > > > 49c7c511000000 REX.W movq r13,<id> > > e8ea2f0700 call <entry> > > > > After: > > > > 41ff9560360000 call [r13+<entry offset>] > > > > Bug: v8:8661,v8:8768 > > Change-Id: I13e30aedc360474dc818fecc528ce87c3bfeed42 > > Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2465834 > > Commit-Queue: Jakob Gruber <jgruber@chromium.org> > > Reviewed-by: Ross McIlroy <rmcilroy@chromium.org> > > Reviewed-by: Tobias Tebbi <tebbi@chromium.org> > > Reviewed-by: Ulan Degenbaev <ulan@chromium.org> > > Cr-Commit-Position: refs/heads/master@{#70597} > > Tbr: ulan@chromium.org, tebbi@chromium.org, rmcilroy@chromium.org > Bug: v8:8661,v8:8768,chromium:1140165 > Change-Id: Ibcd5c39c58a70bf2b2ac221aa375fc68d495e144 > Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2485506 > Reviewed-by: Jakob Gruber <jgruber@chromium.org> > Reviewed-by: Tobias Tebbi <tebbi@chromium.org> > Commit-Queue: Jakob Gruber <jgruber@chromium.org> > Cr-Commit-Position: refs/heads/master@{#70655} Tbr: ulan@chromium.org, tebbi@chromium.org, rmcilroy@chromium.org Bug: v8:8661 Bug: v8:8768 Bug: chromium:1140165 Change-Id: I471cc94fc085e527dc9bfb5a84b96bd907c2333f Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2488682 Reviewed-by: Jakob Gruber <jgruber@chromium.org> Commit-Queue: Jakob Gruber <jgruber@chromium.org> Cr-Commit-Position: refs/heads/master@{#70672}
2020-10-21 05:12:25 +00:00
Handle<Code> code =
Factory::CodeBuilder(isolate, desc, CodeKind::FOR_TESTING).Build();
#ifdef OBJECT_PRINT
StdoutStream os;
code->Print(os);
#endif
auto f = GeneratedCode<F8>::FromCode(*code);
int res = f.Call(1.0f, 2.0f, 3.0f);
PrintF("f(1,2,3) = %d\n", res);
CHECK_EQ(6, res);
}
TEST(AssemblerX64AVX_ss) {
CcTest::InitializeVM();
if (!CpuFeatures::IsSupported(AVX)) return;
Isolate* isolate = reinterpret_cast<Isolate*>(CcTest::isolate());
HandleScope scope(isolate);
v8::internal::byte buffer[1024];
Assembler masm(AssemblerOptions{},
ExternalAssemblerBuffer(buffer, sizeof(buffer)));
{
Reland "[Memory] Use OS::Allocate for all OS memory allocations." This is a reland of 4899bcb66db1c8c8272b02fb1cedf9886e010f82 This is a reland of b73ee3344ac96f92b4418c3c07779eb95a8541cc Original change's description: > [Memory] Use OS::Allocate for all OS memory allocations. > > - Eliminates OS::ReserveRegion and OS::ReserveAlignedRegion. > - Changes OS::Allocate to take alignment parameter, reorders parameters > to match page_allocator. > - Since the size of memory allocation can be deduced, don't return the > amount of memory allocated. > - Changes reservation of aligned address space. Before we would reserve > (size + alignment) rounded up to page size. This is too much, because > maximum misalignment is (alignment - page_size). > - On Windows and Cygwin, we release an oversize allocation and > immediately retry at the aligned address in the allocation. If we > lose the address due to a race, we just retry. > - Clean up all the calls to OS::Allocate in codegen and tests by adding > helper AllocateSystemPage function (allocation.h) and > AllocateAssemblerBuffer (cctest.h). > - Changes 'assm' to 'masm' in some targets for consistency when using > a macro-assembler. > > - Eliminates OS::ReleaseRegion, replacing with calls to OS::Free. > - Adds bool return value to OS::Free. > - Cleans up types of flags, protection on Windows and Cygwin. > Bug: chromium:756050 > Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng > Change-Id: I306dbe042cc867670fdc935abca29db074b0da71 Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng Change-Id: Iad3c025334e8f8d7d647be99a36a11ee449c9087 Reviewed-on: https://chromium-review.googlesource.com/767014 Commit-Queue: Bill Budge <bbudge@chromium.org> Reviewed-by: Michael Lippautz <mlippautz@chromium.org> Cr-Commit-Position: refs/heads/master@{#49363}
2017-11-14 15:55:09 +00:00
CpuFeatureScope avx_scope(&masm, AVX);
Label exit;
// arguments in xmm0, xmm1 and xmm2
__ subq(rsp, Immediate(kDoubleSize * 2)); // For memory operand
__ movl(rdx, Immediate(0xC2F64000)); // -123.125
__ vmovd(xmm4, rdx);
__ vmovss(Operand(rsp, 0), xmm4);
__ vmovss(xmm5, Operand(rsp, 0));
__ vmovaps(xmm6, xmm5);
__ vmovd(rcx, xmm6);
__ cmpl(rcx, rdx);
__ movl(rax, Immediate(9));
__ j(not_equal, &exit);
__ movl(rax, Immediate(0));
__ vmaxss(xmm3, xmm0, xmm1);
__ vucomiss(xmm3, xmm1);
__ j(parity_even, &exit);
__ j(not_equal, &exit);
__ movl(rax, Immediate(1));
__ vminss(xmm3, xmm1, xmm2);
__ vucomiss(xmm3, xmm1);
__ j(parity_even, &exit);
__ j(not_equal, &exit);
__ movl(rax, Immediate(2));
__ vsubss(xmm3, xmm2, xmm1);
__ vucomiss(xmm3, xmm0);
__ j(parity_even, &exit);
__ j(not_equal, &exit);
__ movl(rax, Immediate(3));
__ vaddss(xmm3, xmm0, xmm1);
__ vucomiss(xmm3, xmm2);
__ j(parity_even, &exit);
__ j(not_equal, &exit);
__ movl(rax, Immediate(4));
__ vmulss(xmm3, xmm0, xmm1);
__ vucomiss(xmm3, xmm1);
__ j(parity_even, &exit);
__ j(not_equal, &exit);
__ movl(rax, Immediate(5));
__ vdivss(xmm3, xmm0, xmm1);
__ vmulss(xmm3, xmm3, xmm2);
__ vmulss(xmm3, xmm3, xmm1);
__ vucomiss(xmm3, xmm2);
__ j(parity_even, &exit);
__ j(not_equal, &exit);
__ movl(rax, Immediate(6));
// result in eax
__ bind(&exit);
__ addq(rsp, Immediate(kDoubleSize * 2));
__ ret(0);
}
CodeDesc desc;
Reland "[Memory] Use OS::Allocate for all OS memory allocations." This is a reland of 4899bcb66db1c8c8272b02fb1cedf9886e010f82 This is a reland of b73ee3344ac96f92b4418c3c07779eb95a8541cc Original change's description: > [Memory] Use OS::Allocate for all OS memory allocations. > > - Eliminates OS::ReserveRegion and OS::ReserveAlignedRegion. > - Changes OS::Allocate to take alignment parameter, reorders parameters > to match page_allocator. > - Since the size of memory allocation can be deduced, don't return the > amount of memory allocated. > - Changes reservation of aligned address space. Before we would reserve > (size + alignment) rounded up to page size. This is too much, because > maximum misalignment is (alignment - page_size). > - On Windows and Cygwin, we release an oversize allocation and > immediately retry at the aligned address in the allocation. If we > lose the address due to a race, we just retry. > - Clean up all the calls to OS::Allocate in codegen and tests by adding > helper AllocateSystemPage function (allocation.h) and > AllocateAssemblerBuffer (cctest.h). > - Changes 'assm' to 'masm' in some targets for consistency when using > a macro-assembler. > > - Eliminates OS::ReleaseRegion, replacing with calls to OS::Free. > - Adds bool return value to OS::Free. > - Cleans up types of flags, protection on Windows and Cygwin. > Bug: chromium:756050 > Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng > Change-Id: I306dbe042cc867670fdc935abca29db074b0da71 Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng Change-Id: Iad3c025334e8f8d7d647be99a36a11ee449c9087 Reviewed-on: https://chromium-review.googlesource.com/767014 Commit-Queue: Bill Budge <bbudge@chromium.org> Reviewed-by: Michael Lippautz <mlippautz@chromium.org> Cr-Commit-Position: refs/heads/master@{#49363}
2017-11-14 15:55:09 +00:00
masm.GetCode(isolate, &desc);
Reland "Reland "[deoptimizer] Change deopt entries into builtins"" This is a reland of fbfa9bf4ec72b1b73a96b70ccb68cd98c321511b The arm64 was missing proper codegen for CFI, thus sizes were off. Original change's description: > Reland "[deoptimizer] Change deopt entries into builtins" > > This is a reland of 7f58ced72eb65b6b5530ccabaf2eaebe45bf9d33 > > It fixes the different exit size emitted on x64/Atom CPUs due to > performance tuning in TurboAssembler::Call. Additionally, add > cctests to verify the fixed size exits. > > Original change's description: > > [deoptimizer] Change deopt entries into builtins > > > > While the overall goal of this commit is to change deoptimization > > entries into builtins, there are multiple related things happening: > > > > - Deoptimization entries, formerly stubs (i.e. Code objects generated > > at runtime, guaranteed to be immovable), have been converted into > > builtins. The major restriction is that we now need to preserve the > > kRootRegister, which was formerly used on most architectures to pass > > the deoptimization id. The solution differs based on platform. > > - Renamed DEOPT_ENTRIES_OR_FOR_TESTING code kind to FOR_TESTING. > > - Removed heap/ support for immovable Code generation. > > - Removed the DeserializerData class (no longer needed). > > - arm64: to preserve 4-byte deopt exits, introduced a new optimization > > in which the final jump to the deoptimization entry is generated > > once per Code object, and deopt exits can continue to emit a > > near-call. > > - arm,ia32,x64: change to fixed-size deopt exits. This reduces exit > > sizes by 4/8, 5, and 5 bytes, respectively. > > > > On arm the deopt exit size is reduced from 12 (or 16) bytes to 8 bytes > > by using the same strategy as on arm64 (recalc deopt id from return > > address). Before: > > > > e300a002 movw r10, <id> > > e59fc024 ldr ip, [pc, <entry offset>] > > e12fff3c blx ip > > > > After: > > > > e59acb35 ldr ip, [r10, <entry offset>] > > e12fff3c blx ip > > > > On arm64 the deopt exit size remains 4 bytes (or 8 bytes in same cases > > with CFI). Additionally, up to 4 builtin jumps are emitted per Code > > object (max 32 bytes added overhead per Code object). Before: > > > > 9401cdae bl <entry offset> > > > > After: > > > > # eager deoptimization entry jump. > > f95b1f50 ldr x16, [x26, <eager entry offset>] > > d61f0200 br x16 > > # lazy deoptimization entry jump. > > f95b2b50 ldr x16, [x26, <lazy entry offset>] > > d61f0200 br x16 > > # the deopt exit. > > 97fffffc bl <eager deoptimization entry jump offset> > > > > On ia32 the deopt exit size is reduced from 10 to 5 bytes. Before: > > > > bb00000000 mov ebx,<id> > > e825f5372b call <entry> > > > > After: > > > > e8ea2256ba call <entry> > > > > On x64 the deopt exit size is reduced from 12 to 7 bytes. Before: > > > > 49c7c511000000 REX.W movq r13,<id> > > e8ea2f0700 call <entry> > > > > After: > > > > 41ff9560360000 call [r13+<entry offset>] > > > > Bug: v8:8661,v8:8768 > > Change-Id: I13e30aedc360474dc818fecc528ce87c3bfeed42 > > Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2465834 > > Commit-Queue: Jakob Gruber <jgruber@chromium.org> > > Reviewed-by: Ross McIlroy <rmcilroy@chromium.org> > > Reviewed-by: Tobias Tebbi <tebbi@chromium.org> > > Reviewed-by: Ulan Degenbaev <ulan@chromium.org> > > Cr-Commit-Position: refs/heads/master@{#70597} > > Tbr: ulan@chromium.org, tebbi@chromium.org, rmcilroy@chromium.org > Bug: v8:8661,v8:8768,chromium:1140165 > Change-Id: Ibcd5c39c58a70bf2b2ac221aa375fc68d495e144 > Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2485506 > Reviewed-by: Jakob Gruber <jgruber@chromium.org> > Reviewed-by: Tobias Tebbi <tebbi@chromium.org> > Commit-Queue: Jakob Gruber <jgruber@chromium.org> > Cr-Commit-Position: refs/heads/master@{#70655} Tbr: ulan@chromium.org, tebbi@chromium.org, rmcilroy@chromium.org Bug: v8:8661 Bug: v8:8768 Bug: chromium:1140165 Change-Id: I471cc94fc085e527dc9bfb5a84b96bd907c2333f Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2488682 Reviewed-by: Jakob Gruber <jgruber@chromium.org> Commit-Queue: Jakob Gruber <jgruber@chromium.org> Cr-Commit-Position: refs/heads/master@{#70672}
2020-10-21 05:12:25 +00:00
Handle<Code> code =
Factory::CodeBuilder(isolate, desc, CodeKind::FOR_TESTING).Build();
#ifdef OBJECT_PRINT
StdoutStream os;
code->Print(os);
#endif
auto f = GeneratedCode<F8>::FromCode(*code);
int res = f.Call(1.0f, 2.0f, 3.0f);
PrintF("f(1,2,3) = %d\n", res);
CHECK_EQ(6, res);
}
TEST(AssemblerX64AVX_sd) {
CcTest::InitializeVM();
if (!CpuFeatures::IsSupported(AVX)) return;
Isolate* isolate = reinterpret_cast<Isolate*>(CcTest::isolate());
HandleScope scope(isolate);
v8::internal::byte buffer[1024];
Assembler masm(AssemblerOptions{},
ExternalAssemblerBuffer(buffer, sizeof(buffer)));
{
Reland "[Memory] Use OS::Allocate for all OS memory allocations." This is a reland of 4899bcb66db1c8c8272b02fb1cedf9886e010f82 This is a reland of b73ee3344ac96f92b4418c3c07779eb95a8541cc Original change's description: > [Memory] Use OS::Allocate for all OS memory allocations. > > - Eliminates OS::ReserveRegion and OS::ReserveAlignedRegion. > - Changes OS::Allocate to take alignment parameter, reorders parameters > to match page_allocator. > - Since the size of memory allocation can be deduced, don't return the > amount of memory allocated. > - Changes reservation of aligned address space. Before we would reserve > (size + alignment) rounded up to page size. This is too much, because > maximum misalignment is (alignment - page_size). > - On Windows and Cygwin, we release an oversize allocation and > immediately retry at the aligned address in the allocation. If we > lose the address due to a race, we just retry. > - Clean up all the calls to OS::Allocate in codegen and tests by adding > helper AllocateSystemPage function (allocation.h) and > AllocateAssemblerBuffer (cctest.h). > - Changes 'assm' to 'masm' in some targets for consistency when using > a macro-assembler. > > - Eliminates OS::ReleaseRegion, replacing with calls to OS::Free. > - Adds bool return value to OS::Free. > - Cleans up types of flags, protection on Windows and Cygwin. > Bug: chromium:756050 > Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng > Change-Id: I306dbe042cc867670fdc935abca29db074b0da71 Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng Change-Id: Iad3c025334e8f8d7d647be99a36a11ee449c9087 Reviewed-on: https://chromium-review.googlesource.com/767014 Commit-Queue: Bill Budge <bbudge@chromium.org> Reviewed-by: Michael Lippautz <mlippautz@chromium.org> Cr-Commit-Position: refs/heads/master@{#49363}
2017-11-14 15:55:09 +00:00
CpuFeatureScope avx_scope(&masm, AVX);
Label exit;
// arguments in xmm0, xmm1 and xmm2
__ subq(rsp, Immediate(kDoubleSize * 2)); // For memory operand
__ movl(rax, Immediate(0));
__ vmaxsd(xmm4, xmm0, xmm1);
__ vmovsd(Operand(rsp, kDoubleSize), xmm4);
__ vmovsd(xmm5, Operand(rsp, kDoubleSize));
__ vmovsd(xmm6, xmm6, xmm5);
__ vmovapd(xmm3, xmm6);
// Test vcvtss2sd & vcvtsd2ss
__ movl(rax, Immediate(9));
__ movq(rdx, uint64_t{0x426D1A0000000000});
__ movq(Operand(rsp, 0), rdx);
__ vcvtsd2ss(xmm6, xmm6, Operand(rsp, 0));
__ vcvtss2sd(xmm7, xmm6, xmm6);
__ vcvtsd2ss(xmm8, xmm7, xmm7);
__ vmovss(Operand(rsp, 0), xmm8);
__ vcvtss2sd(xmm9, xmm8, Operand(rsp, 0));
__ vmovq(rcx, xmm9);
__ cmpq(rcx, rdx);
__ j(not_equal, &exit);
// Test vcvttsd2si
__ movl(rax, Immediate(10));
__ movl(rdx, Immediate(123));
__ vcvtlsi2sd(xmm6, xmm6, rdx);
__ vcvttsd2si(rcx, xmm6);
__ cmpl(rcx, rdx);
__ j(not_equal, &exit);
__ xorl(rcx, rcx);
__ vmovsd(Operand(rsp, 0), xmm6);
__ vcvttsd2si(rcx, Operand(rsp, 0));
__ cmpl(rcx, rdx);
__ j(not_equal, &exit);
// Test vcvttsd2siq
__ movl(rax, Immediate(11));
__ movq(rdx, uint64_t{0x426D1A94A2000000}); // 1.0e12
__ vmovq(xmm6, rdx);
__ vcvttsd2siq(rcx, xmm6);
__ movq(rdx, uint64_t{1000000000000});
__ cmpq(rcx, rdx);
__ j(not_equal, &exit);
__ xorq(rcx, rcx);
__ vmovsd(Operand(rsp, 0), xmm6);
__ vcvttsd2siq(rcx, Operand(rsp, 0));
__ cmpq(rcx, rdx);
__ j(not_equal, &exit);
// Test vmovmskpd
__ movl(rax, Immediate(12));
__ movq(rdx, uint64_t{0x426D1A94A2000000}); // 1.0e12
__ vmovq(xmm6, rdx);
__ movq(rdx, uint64_t{0xC26D1A94A2000000}); // -1.0e12
__ vmovq(xmm7, rdx);
__ shufps(xmm6, xmm7, 0x44);
__ vmovmskpd(rdx, xmm6);
__ cmpl(rdx, Immediate(2));
__ j(not_equal, &exit);
// Test vpcmpeqd
__ movq(rdx, uint64_t{0x0123456789ABCDEF});
__ movq(rcx, uint64_t{0x0123456788888888});
__ vmovq(xmm6, rdx);
__ vmovq(xmm7, rcx);
__ vpcmpeqd(xmm8, xmm6, xmm7);
__ vmovq(rdx, xmm8);
__ movq(rcx, uint64_t{0xFFFFFFFF00000000});
__ cmpq(rcx, rdx);
__ movl(rax, Immediate(13));
__ j(not_equal, &exit);
// Test vpsllq, vpsrlq
__ movl(rax, Immediate(13));
__ movq(rdx, uint64_t{0x0123456789ABCDEF});
__ vmovq(xmm6, rdx);
__ vpsrlq(xmm7, xmm6, 4);
__ vmovq(rdx, xmm7);
__ movq(rcx, uint64_t{0x00123456789ABCDE});
__ cmpq(rdx, rcx);
__ j(not_equal, &exit);
__ vpsllq(xmm7, xmm6, 12);
__ vmovq(rdx, xmm7);
__ movq(rcx, uint64_t{0x3456789ABCDEF000});
__ cmpq(rdx, rcx);
__ j(not_equal, &exit);
// Test vandpd, vorpd, vxorpd
__ movl(rax, Immediate(14));
__ movl(rdx, Immediate(0x00FF00FF));
__ movl(rcx, Immediate(0x0F0F0F0F));
__ vmovd(xmm4, rdx);
__ vmovd(xmm5, rcx);
__ vandpd(xmm6, xmm4, xmm5);
__ vmovd(rdx, xmm6);
__ cmpl(rdx, Immediate(0x000F000F));
__ j(not_equal, &exit);
__ vorpd(xmm6, xmm4, xmm5);
__ vmovd(rdx, xmm6);
__ cmpl(rdx, Immediate(0x0FFF0FFF));
__ j(not_equal, &exit);
__ vxorpd(xmm6, xmm4, xmm5);
__ vmovd(rdx, xmm6);
__ cmpl(rdx, Immediate(0x0FF00FF0));
__ j(not_equal, &exit);
// Test vsqrtsd
__ movl(rax, Immediate(15));
__ movq(rdx, uint64_t{0x4004000000000000}); // 2.5
__ vmovq(xmm4, rdx);
__ vmulsd(xmm5, xmm4, xmm4);
__ vmovsd(Operand(rsp, 0), xmm5);
__ vsqrtsd(xmm6, xmm5, xmm5);
__ vmovq(rcx, xmm6);
__ cmpq(rcx, rdx);
__ j(not_equal, &exit);
__ vsqrtsd(xmm7, xmm7, Operand(rsp, 0));
__ vmovq(rcx, xmm7);
__ cmpq(rcx, rdx);
__ j(not_equal, &exit);
// Test vroundsd
__ movl(rax, Immediate(16));
__ movq(rdx, uint64_t{0x4002000000000000}); // 2.25
__ vmovq(xmm4, rdx);
__ vroundsd(xmm5, xmm4, xmm4, kRoundUp);
__ movq(rcx, uint64_t{0x4008000000000000}); // 3.0
__ vmovq(xmm6, rcx);
__ vucomisd(xmm5, xmm6);
__ j(not_equal, &exit);
// Test vcvtlsi2sd
__ movl(rax, Immediate(17));
__ movl(rdx, Immediate(6));
__ movq(rcx, uint64_t{0x4018000000000000}); // 6.0
__ vmovq(xmm5, rcx);
__ vcvtlsi2sd(xmm6, xmm6, rdx);
__ vucomisd(xmm5, xmm6);
__ j(not_equal, &exit);
__ movl(Operand(rsp, 0), rdx);
__ vcvtlsi2sd(xmm7, xmm7, Operand(rsp, 0));
__ vucomisd(xmm5, xmm6);
__ j(not_equal, &exit);
// Test vcvtqsi2sd
__ movl(rax, Immediate(18));
__ movq(rdx, uint64_t{0x2000000000000000}); // 2 << 0x3C
__ movq(rcx, uint64_t{0x43C0000000000000});
__ vmovq(xmm5, rcx);
__ vcvtqsi2sd(xmm6, xmm6, rdx);
__ vucomisd(xmm5, xmm6);
__ j(not_equal, &exit);
// Test vcvtsd2si
__ movl(rax, Immediate(19));
__ movq(rdx, uint64_t{0x4018000000000000}); // 6.0
__ vmovq(xmm5, rdx);
__ vcvtsd2si(rcx, xmm5);
__ cmpl(rcx, Immediate(6));
__ j(not_equal, &exit);
__ movq(rdx, uint64_t{0x3FF0000000000000}); // 1.0
__ vmovq(xmm7, rdx);
__ vmulsd(xmm1, xmm1, xmm7);
__ movq(Operand(rsp, 0), rdx);
__ vmovq(xmm6, Operand(rsp, 0));
__ vmulsd(xmm1, xmm1, xmm6);
__ vucomisd(xmm3, xmm1);
__ j(parity_even, &exit);
__ j(not_equal, &exit);
__ movl(rax, Immediate(1));
__ vminsd(xmm3, xmm1, xmm2);
__ vucomisd(xmm3, xmm1);
__ j(parity_even, &exit);
__ j(not_equal, &exit);
__ movl(rax, Immediate(2));
__ vsubsd(xmm3, xmm2, xmm1);
__ vucomisd(xmm3, xmm0);
__ j(parity_even, &exit);
__ j(not_equal, &exit);
__ movl(rax, Immediate(3));
__ vaddsd(xmm3, xmm0, xmm1);
__ vucomisd(xmm3, xmm2);
__ j(parity_even, &exit);
__ j(not_equal, &exit);
__ movl(rax, Immediate(4));
__ vmulsd(xmm3, xmm0, xmm1);
__ vucomisd(xmm3, xmm1);
__ j(parity_even, &exit);
__ j(not_equal, &exit);
__ movl(rax, Immediate(5));
__ vdivsd(xmm3, xmm0, xmm1);
__ vmulsd(xmm3, xmm3, xmm2);
__ vmulsd(xmm3, xmm3, xmm1);
__ vucomisd(xmm3, xmm2);
__ j(parity_even, &exit);
__ j(not_equal, &exit);
__ movl(rax, Immediate(6));
// result in eax
__ bind(&exit);
__ addq(rsp, Immediate(kDoubleSize * 2));
__ ret(0);
}
CodeDesc desc;
Reland "[Memory] Use OS::Allocate for all OS memory allocations." This is a reland of 4899bcb66db1c8c8272b02fb1cedf9886e010f82 This is a reland of b73ee3344ac96f92b4418c3c07779eb95a8541cc Original change's description: > [Memory] Use OS::Allocate for all OS memory allocations. > > - Eliminates OS::ReserveRegion and OS::ReserveAlignedRegion. > - Changes OS::Allocate to take alignment parameter, reorders parameters > to match page_allocator. > - Since the size of memory allocation can be deduced, don't return the > amount of memory allocated. > - Changes reservation of aligned address space. Before we would reserve > (size + alignment) rounded up to page size. This is too much, because > maximum misalignment is (alignment - page_size). > - On Windows and Cygwin, we release an oversize allocation and > immediately retry at the aligned address in the allocation. If we > lose the address due to a race, we just retry. > - Clean up all the calls to OS::Allocate in codegen and tests by adding > helper AllocateSystemPage function (allocation.h) and > AllocateAssemblerBuffer (cctest.h). > - Changes 'assm' to 'masm' in some targets for consistency when using > a macro-assembler. > > - Eliminates OS::ReleaseRegion, replacing with calls to OS::Free. > - Adds bool return value to OS::Free. > - Cleans up types of flags, protection on Windows and Cygwin. > Bug: chromium:756050 > Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng > Change-Id: I306dbe042cc867670fdc935abca29db074b0da71 Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng Change-Id: Iad3c025334e8f8d7d647be99a36a11ee449c9087 Reviewed-on: https://chromium-review.googlesource.com/767014 Commit-Queue: Bill Budge <bbudge@chromium.org> Reviewed-by: Michael Lippautz <mlippautz@chromium.org> Cr-Commit-Position: refs/heads/master@{#49363}
2017-11-14 15:55:09 +00:00
masm.GetCode(isolate, &desc);
Reland "Reland "[deoptimizer] Change deopt entries into builtins"" This is a reland of fbfa9bf4ec72b1b73a96b70ccb68cd98c321511b The arm64 was missing proper codegen for CFI, thus sizes were off. Original change's description: > Reland "[deoptimizer] Change deopt entries into builtins" > > This is a reland of 7f58ced72eb65b6b5530ccabaf2eaebe45bf9d33 > > It fixes the different exit size emitted on x64/Atom CPUs due to > performance tuning in TurboAssembler::Call. Additionally, add > cctests to verify the fixed size exits. > > Original change's description: > > [deoptimizer] Change deopt entries into builtins > > > > While the overall goal of this commit is to change deoptimization > > entries into builtins, there are multiple related things happening: > > > > - Deoptimization entries, formerly stubs (i.e. Code objects generated > > at runtime, guaranteed to be immovable), have been converted into > > builtins. The major restriction is that we now need to preserve the > > kRootRegister, which was formerly used on most architectures to pass > > the deoptimization id. The solution differs based on platform. > > - Renamed DEOPT_ENTRIES_OR_FOR_TESTING code kind to FOR_TESTING. > > - Removed heap/ support for immovable Code generation. > > - Removed the DeserializerData class (no longer needed). > > - arm64: to preserve 4-byte deopt exits, introduced a new optimization > > in which the final jump to the deoptimization entry is generated > > once per Code object, and deopt exits can continue to emit a > > near-call. > > - arm,ia32,x64: change to fixed-size deopt exits. This reduces exit > > sizes by 4/8, 5, and 5 bytes, respectively. > > > > On arm the deopt exit size is reduced from 12 (or 16) bytes to 8 bytes > > by using the same strategy as on arm64 (recalc deopt id from return > > address). Before: > > > > e300a002 movw r10, <id> > > e59fc024 ldr ip, [pc, <entry offset>] > > e12fff3c blx ip > > > > After: > > > > e59acb35 ldr ip, [r10, <entry offset>] > > e12fff3c blx ip > > > > On arm64 the deopt exit size remains 4 bytes (or 8 bytes in same cases > > with CFI). Additionally, up to 4 builtin jumps are emitted per Code > > object (max 32 bytes added overhead per Code object). Before: > > > > 9401cdae bl <entry offset> > > > > After: > > > > # eager deoptimization entry jump. > > f95b1f50 ldr x16, [x26, <eager entry offset>] > > d61f0200 br x16 > > # lazy deoptimization entry jump. > > f95b2b50 ldr x16, [x26, <lazy entry offset>] > > d61f0200 br x16 > > # the deopt exit. > > 97fffffc bl <eager deoptimization entry jump offset> > > > > On ia32 the deopt exit size is reduced from 10 to 5 bytes. Before: > > > > bb00000000 mov ebx,<id> > > e825f5372b call <entry> > > > > After: > > > > e8ea2256ba call <entry> > > > > On x64 the deopt exit size is reduced from 12 to 7 bytes. Before: > > > > 49c7c511000000 REX.W movq r13,<id> > > e8ea2f0700 call <entry> > > > > After: > > > > 41ff9560360000 call [r13+<entry offset>] > > > > Bug: v8:8661,v8:8768 > > Change-Id: I13e30aedc360474dc818fecc528ce87c3bfeed42 > > Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2465834 > > Commit-Queue: Jakob Gruber <jgruber@chromium.org> > > Reviewed-by: Ross McIlroy <rmcilroy@chromium.org> > > Reviewed-by: Tobias Tebbi <tebbi@chromium.org> > > Reviewed-by: Ulan Degenbaev <ulan@chromium.org> > > Cr-Commit-Position: refs/heads/master@{#70597} > > Tbr: ulan@chromium.org, tebbi@chromium.org, rmcilroy@chromium.org > Bug: v8:8661,v8:8768,chromium:1140165 > Change-Id: Ibcd5c39c58a70bf2b2ac221aa375fc68d495e144 > Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2485506 > Reviewed-by: Jakob Gruber <jgruber@chromium.org> > Reviewed-by: Tobias Tebbi <tebbi@chromium.org> > Commit-Queue: Jakob Gruber <jgruber@chromium.org> > Cr-Commit-Position: refs/heads/master@{#70655} Tbr: ulan@chromium.org, tebbi@chromium.org, rmcilroy@chromium.org Bug: v8:8661 Bug: v8:8768 Bug: chromium:1140165 Change-Id: I471cc94fc085e527dc9bfb5a84b96bd907c2333f Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2488682 Reviewed-by: Jakob Gruber <jgruber@chromium.org> Commit-Queue: Jakob Gruber <jgruber@chromium.org> Cr-Commit-Position: refs/heads/master@{#70672}
2020-10-21 05:12:25 +00:00
Handle<Code> code =
Factory::CodeBuilder(isolate, desc, CodeKind::FOR_TESTING).Build();
#ifdef OBJECT_PRINT
StdoutStream os;
code->Print(os);
#endif
auto f = GeneratedCode<F7>::FromCode(*code);
int res = f.Call(1.0, 2.0, 3.0);
PrintF("f(1,2,3) = %d\n", res);
CHECK_EQ(6, res);
}
TEST(AssemblerX64BMI1) {
CcTest::InitializeVM();
if (!CpuFeatures::IsSupported(BMI1)) return;
Isolate* isolate = reinterpret_cast<Isolate*>(CcTest::isolate());
HandleScope scope(isolate);
v8::internal::byte buffer[1024];
MacroAssembler masm(isolate, v8::internal::CodeObjectRequired::kYes,
ExternalAssemblerBuffer(buffer, sizeof(buffer)));
{
Reland "[Memory] Use OS::Allocate for all OS memory allocations." This is a reland of 4899bcb66db1c8c8272b02fb1cedf9886e010f82 This is a reland of b73ee3344ac96f92b4418c3c07779eb95a8541cc Original change's description: > [Memory] Use OS::Allocate for all OS memory allocations. > > - Eliminates OS::ReserveRegion and OS::ReserveAlignedRegion. > - Changes OS::Allocate to take alignment parameter, reorders parameters > to match page_allocator. > - Since the size of memory allocation can be deduced, don't return the > amount of memory allocated. > - Changes reservation of aligned address space. Before we would reserve > (size + alignment) rounded up to page size. This is too much, because > maximum misalignment is (alignment - page_size). > - On Windows and Cygwin, we release an oversize allocation and > immediately retry at the aligned address in the allocation. If we > lose the address due to a race, we just retry. > - Clean up all the calls to OS::Allocate in codegen and tests by adding > helper AllocateSystemPage function (allocation.h) and > AllocateAssemblerBuffer (cctest.h). > - Changes 'assm' to 'masm' in some targets for consistency when using > a macro-assembler. > > - Eliminates OS::ReleaseRegion, replacing with calls to OS::Free. > - Adds bool return value to OS::Free. > - Cleans up types of flags, protection on Windows and Cygwin. > Bug: chromium:756050 > Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng > Change-Id: I306dbe042cc867670fdc935abca29db074b0da71 Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng Change-Id: Iad3c025334e8f8d7d647be99a36a11ee449c9087 Reviewed-on: https://chromium-review.googlesource.com/767014 Commit-Queue: Bill Budge <bbudge@chromium.org> Reviewed-by: Michael Lippautz <mlippautz@chromium.org> Cr-Commit-Position: refs/heads/master@{#49363}
2017-11-14 15:55:09 +00:00
CpuFeatureScope fscope(&masm, BMI1);
Label exit;
__ movq(rcx, uint64_t{0x1122334455667788}); // source operand
__ pushq(rcx); // For memory operand
// andn
__ movq(rdx, uint64_t{0x1000000020000000});
__ movl(rax, Immediate(1)); // Test number
__ andnq(r8, rdx, rcx);
__ movq(r9, uint64_t{0x0122334455667788}); // expected result
__ cmpq(r8, r9);
__ j(not_equal, &exit);
__ incq(rax);
__ andnq(r8, rdx, Operand(rsp, 0));
__ movq(r9, uint64_t{0x0122334455667788}); // expected result
__ cmpq(r8, r9);
__ j(not_equal, &exit);
__ incq(rax);
__ andnl(r8, rdx, rcx);
__ movq(r9, uint64_t{0x0000000055667788}); // expected result
__ cmpq(r8, r9);
__ j(not_equal, &exit);
__ incq(rax);
__ andnl(r8, rdx, Operand(rsp, 0));
__ movq(r9, uint64_t{0x0000000055667788}); // expected result
__ cmpq(r8, r9);
__ j(not_equal, &exit);
// bextr
__ movq(rdx, uint64_t{0x0000000000002808});
__ incq(rax);
__ bextrq(r8, rcx, rdx);
__ movq(r9, uint64_t{0x0000003344556677}); // expected result
__ cmpq(r8, r9);
__ j(not_equal, &exit);
__ incq(rax);
__ bextrq(r8, Operand(rsp, 0), rdx);
__ movq(r9, uint64_t{0x0000003344556677}); // expected result
__ cmpq(r8, r9);
__ j(not_equal, &exit);
__ incq(rax);
__ bextrl(r8, rcx, rdx);
__ movq(r9, uint64_t{0x0000000000556677}); // expected result
__ cmpq(r8, r9);
__ j(not_equal, &exit);
__ incq(rax);
__ bextrl(r8, Operand(rsp, 0), rdx);
__ movq(r9, uint64_t{0x0000000000556677}); // expected result
__ cmpq(r8, r9);
__ j(not_equal, &exit);
// blsi
__ incq(rax);
__ blsiq(r8, rcx);
__ movq(r9, uint64_t{0x0000000000000008}); // expected result
__ cmpq(r8, r9);
__ j(not_equal, &exit);
__ incq(rax);
__ blsiq(r8, Operand(rsp, 0));
__ movq(r9, uint64_t{0x0000000000000008}); // expected result
__ cmpq(r8, r9);
__ j(not_equal, &exit);
__ incq(rax);
__ blsil(r8, rcx);
__ movq(r9, uint64_t{0x0000000000000008}); // expected result
__ cmpq(r8, r9);
__ j(not_equal, &exit);
__ incq(rax);
__ blsil(r8, Operand(rsp, 0));
__ movq(r9, uint64_t{0x0000000000000008}); // expected result
__ cmpq(r8, r9);
__ j(not_equal, &exit);
// blsmsk
__ incq(rax);
__ blsmskq(r8, rcx);
__ movq(r9, uint64_t{0x000000000000000F}); // expected result
__ cmpq(r8, r9);
__ j(not_equal, &exit);
__ incq(rax);
__ blsmskq(r8, Operand(rsp, 0));
__ movq(r9, uint64_t{0x000000000000000F}); // expected result
__ cmpq(r8, r9);
__ j(not_equal, &exit);
__ incq(rax);
__ blsmskl(r8, rcx);
__ movq(r9, uint64_t{0x000000000000000F}); // expected result
__ cmpq(r8, r9);
__ j(not_equal, &exit);
__ incq(rax);
__ blsmskl(r8, Operand(rsp, 0));
__ movq(r9, uint64_t{0x000000000000000F}); // expected result
__ cmpq(r8, r9);
__ j(not_equal, &exit);
// blsr
__ incq(rax);
__ blsrq(r8, rcx);
__ movq(r9, uint64_t{0x1122334455667780}); // expected result
__ cmpq(r8, r9);
__ j(not_equal, &exit);
__ incq(rax);
__ blsrq(r8, Operand(rsp, 0));
__ movq(r9, uint64_t{0x1122334455667780}); // expected result
__ cmpq(r8, r9);
__ j(not_equal, &exit);
__ incq(rax);
__ blsrl(r8, rcx);
__ movq(r9, uint64_t{0x0000000055667780}); // expected result
__ cmpq(r8, r9);
__ j(not_equal, &exit);
__ incq(rax);
__ blsrl(r8, Operand(rsp, 0));
__ movq(r9, uint64_t{0x0000000055667780}); // expected result
__ cmpq(r8, r9);
__ j(not_equal, &exit);
// tzcnt
__ incq(rax);
__ tzcntq(r8, rcx);
__ movq(r9, uint64_t{0x0000000000000003}); // expected result
__ cmpq(r8, r9);
__ j(not_equal, &exit);
__ incq(rax);
__ tzcntq(r8, Operand(rsp, 0));
__ movq(r9, uint64_t{0x0000000000000003}); // expected result
__ cmpq(r8, r9);
__ j(not_equal, &exit);
__ incq(rax);
__ tzcntl(r8, rcx);
__ movq(r9, uint64_t{0x0000000000000003}); // expected result
__ cmpq(r8, r9);
__ j(not_equal, &exit);
__ incq(rax);
__ tzcntl(r8, Operand(rsp, 0));
__ movq(r9, uint64_t{0x0000000000000003}); // expected result
__ cmpq(r8, r9);
__ j(not_equal, &exit);
__ xorl(rax, rax);
__ bind(&exit);
__ popq(rcx);
__ ret(0);
}
CodeDesc desc;
Reland "[Memory] Use OS::Allocate for all OS memory allocations." This is a reland of 4899bcb66db1c8c8272b02fb1cedf9886e010f82 This is a reland of b73ee3344ac96f92b4418c3c07779eb95a8541cc Original change's description: > [Memory] Use OS::Allocate for all OS memory allocations. > > - Eliminates OS::ReserveRegion and OS::ReserveAlignedRegion. > - Changes OS::Allocate to take alignment parameter, reorders parameters > to match page_allocator. > - Since the size of memory allocation can be deduced, don't return the > amount of memory allocated. > - Changes reservation of aligned address space. Before we would reserve > (size + alignment) rounded up to page size. This is too much, because > maximum misalignment is (alignment - page_size). > - On Windows and Cygwin, we release an oversize allocation and > immediately retry at the aligned address in the allocation. If we > lose the address due to a race, we just retry. > - Clean up all the calls to OS::Allocate in codegen and tests by adding > helper AllocateSystemPage function (allocation.h) and > AllocateAssemblerBuffer (cctest.h). > - Changes 'assm' to 'masm' in some targets for consistency when using > a macro-assembler. > > - Eliminates OS::ReleaseRegion, replacing with calls to OS::Free. > - Adds bool return value to OS::Free. > - Cleans up types of flags, protection on Windows and Cygwin. > Bug: chromium:756050 > Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng > Change-Id: I306dbe042cc867670fdc935abca29db074b0da71 Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng Change-Id: Iad3c025334e8f8d7d647be99a36a11ee449c9087 Reviewed-on: https://chromium-review.googlesource.com/767014 Commit-Queue: Bill Budge <bbudge@chromium.org> Reviewed-by: Michael Lippautz <mlippautz@chromium.org> Cr-Commit-Position: refs/heads/master@{#49363}
2017-11-14 15:55:09 +00:00
masm.GetCode(isolate, &desc);
Reland "Reland "[deoptimizer] Change deopt entries into builtins"" This is a reland of fbfa9bf4ec72b1b73a96b70ccb68cd98c321511b The arm64 was missing proper codegen for CFI, thus sizes were off. Original change's description: > Reland "[deoptimizer] Change deopt entries into builtins" > > This is a reland of 7f58ced72eb65b6b5530ccabaf2eaebe45bf9d33 > > It fixes the different exit size emitted on x64/Atom CPUs due to > performance tuning in TurboAssembler::Call. Additionally, add > cctests to verify the fixed size exits. > > Original change's description: > > [deoptimizer] Change deopt entries into builtins > > > > While the overall goal of this commit is to change deoptimization > > entries into builtins, there are multiple related things happening: > > > > - Deoptimization entries, formerly stubs (i.e. Code objects generated > > at runtime, guaranteed to be immovable), have been converted into > > builtins. The major restriction is that we now need to preserve the > > kRootRegister, which was formerly used on most architectures to pass > > the deoptimization id. The solution differs based on platform. > > - Renamed DEOPT_ENTRIES_OR_FOR_TESTING code kind to FOR_TESTING. > > - Removed heap/ support for immovable Code generation. > > - Removed the DeserializerData class (no longer needed). > > - arm64: to preserve 4-byte deopt exits, introduced a new optimization > > in which the final jump to the deoptimization entry is generated > > once per Code object, and deopt exits can continue to emit a > > near-call. > > - arm,ia32,x64: change to fixed-size deopt exits. This reduces exit > > sizes by 4/8, 5, and 5 bytes, respectively. > > > > On arm the deopt exit size is reduced from 12 (or 16) bytes to 8 bytes > > by using the same strategy as on arm64 (recalc deopt id from return > > address). Before: > > > > e300a002 movw r10, <id> > > e59fc024 ldr ip, [pc, <entry offset>] > > e12fff3c blx ip > > > > After: > > > > e59acb35 ldr ip, [r10, <entry offset>] > > e12fff3c blx ip > > > > On arm64 the deopt exit size remains 4 bytes (or 8 bytes in same cases > > with CFI). Additionally, up to 4 builtin jumps are emitted per Code > > object (max 32 bytes added overhead per Code object). Before: > > > > 9401cdae bl <entry offset> > > > > After: > > > > # eager deoptimization entry jump. > > f95b1f50 ldr x16, [x26, <eager entry offset>] > > d61f0200 br x16 > > # lazy deoptimization entry jump. > > f95b2b50 ldr x16, [x26, <lazy entry offset>] > > d61f0200 br x16 > > # the deopt exit. > > 97fffffc bl <eager deoptimization entry jump offset> > > > > On ia32 the deopt exit size is reduced from 10 to 5 bytes. Before: > > > > bb00000000 mov ebx,<id> > > e825f5372b call <entry> > > > > After: > > > > e8ea2256ba call <entry> > > > > On x64 the deopt exit size is reduced from 12 to 7 bytes. Before: > > > > 49c7c511000000 REX.W movq r13,<id> > > e8ea2f0700 call <entry> > > > > After: > > > > 41ff9560360000 call [r13+<entry offset>] > > > > Bug: v8:8661,v8:8768 > > Change-Id: I13e30aedc360474dc818fecc528ce87c3bfeed42 > > Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2465834 > > Commit-Queue: Jakob Gruber <jgruber@chromium.org> > > Reviewed-by: Ross McIlroy <rmcilroy@chromium.org> > > Reviewed-by: Tobias Tebbi <tebbi@chromium.org> > > Reviewed-by: Ulan Degenbaev <ulan@chromium.org> > > Cr-Commit-Position: refs/heads/master@{#70597} > > Tbr: ulan@chromium.org, tebbi@chromium.org, rmcilroy@chromium.org > Bug: v8:8661,v8:8768,chromium:1140165 > Change-Id: Ibcd5c39c58a70bf2b2ac221aa375fc68d495e144 > Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2485506 > Reviewed-by: Jakob Gruber <jgruber@chromium.org> > Reviewed-by: Tobias Tebbi <tebbi@chromium.org> > Commit-Queue: Jakob Gruber <jgruber@chromium.org> > Cr-Commit-Position: refs/heads/master@{#70655} Tbr: ulan@chromium.org, tebbi@chromium.org, rmcilroy@chromium.org Bug: v8:8661 Bug: v8:8768 Bug: chromium:1140165 Change-Id: I471cc94fc085e527dc9bfb5a84b96bd907c2333f Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2488682 Reviewed-by: Jakob Gruber <jgruber@chromium.org> Commit-Queue: Jakob Gruber <jgruber@chromium.org> Cr-Commit-Position: refs/heads/master@{#70672}
2020-10-21 05:12:25 +00:00
Handle<Code> code =
Factory::CodeBuilder(isolate, desc, CodeKind::FOR_TESTING).Build();
#ifdef OBJECT_PRINT
StdoutStream os;
code->Print(os);
#endif
auto f = GeneratedCode<F0>::FromCode(*code);
CHECK_EQ(0, f.Call());
}
TEST(AssemblerX64LZCNT) {
CcTest::InitializeVM();
if (!CpuFeatures::IsSupported(LZCNT)) return;
Isolate* isolate = reinterpret_cast<Isolate*>(CcTest::isolate());
HandleScope scope(isolate);
v8::internal::byte buffer[256];
MacroAssembler masm(isolate, v8::internal::CodeObjectRequired::kYes,
ExternalAssemblerBuffer(buffer, sizeof(buffer)));
{
Reland "[Memory] Use OS::Allocate for all OS memory allocations." This is a reland of 4899bcb66db1c8c8272b02fb1cedf9886e010f82 This is a reland of b73ee3344ac96f92b4418c3c07779eb95a8541cc Original change's description: > [Memory] Use OS::Allocate for all OS memory allocations. > > - Eliminates OS::ReserveRegion and OS::ReserveAlignedRegion. > - Changes OS::Allocate to take alignment parameter, reorders parameters > to match page_allocator. > - Since the size of memory allocation can be deduced, don't return the > amount of memory allocated. > - Changes reservation of aligned address space. Before we would reserve > (size + alignment) rounded up to page size. This is too much, because > maximum misalignment is (alignment - page_size). > - On Windows and Cygwin, we release an oversize allocation and > immediately retry at the aligned address in the allocation. If we > lose the address due to a race, we just retry. > - Clean up all the calls to OS::Allocate in codegen and tests by adding > helper AllocateSystemPage function (allocation.h) and > AllocateAssemblerBuffer (cctest.h). > - Changes 'assm' to 'masm' in some targets for consistency when using > a macro-assembler. > > - Eliminates OS::ReleaseRegion, replacing with calls to OS::Free. > - Adds bool return value to OS::Free. > - Cleans up types of flags, protection on Windows and Cygwin. > Bug: chromium:756050 > Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng > Change-Id: I306dbe042cc867670fdc935abca29db074b0da71 Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng Change-Id: Iad3c025334e8f8d7d647be99a36a11ee449c9087 Reviewed-on: https://chromium-review.googlesource.com/767014 Commit-Queue: Bill Budge <bbudge@chromium.org> Reviewed-by: Michael Lippautz <mlippautz@chromium.org> Cr-Commit-Position: refs/heads/master@{#49363}
2017-11-14 15:55:09 +00:00
CpuFeatureScope fscope(&masm, LZCNT);
Label exit;
__ movq(rcx, uint64_t{0x1122334455667788}); // source operand
__ pushq(rcx); // For memory operand
__ movl(rax, Immediate(1)); // Test number
__ lzcntq(r8, rcx);
__ movq(r9, uint64_t{0x0000000000000003}); // expected result
__ cmpq(r8, r9);
__ j(not_equal, &exit);
__ incq(rax);
__ lzcntq(r8, Operand(rsp, 0));
__ movq(r9, uint64_t{0x0000000000000003}); // expected result
__ cmpq(r8, r9);
__ j(not_equal, &exit);
__ incq(rax);
__ lzcntl(r8, rcx);
__ movq(r9, uint64_t{0x0000000000000001}); // expected result
__ cmpq(r8, r9);
__ j(not_equal, &exit);
__ incq(rax);
__ lzcntl(r8, Operand(rsp, 0));
__ movq(r9, uint64_t{0x0000000000000001}); // expected result
__ cmpq(r8, r9);
__ j(not_equal, &exit);
__ xorl(rax, rax);
__ bind(&exit);
__ popq(rcx);
__ ret(0);
}
CodeDesc desc;
Reland "[Memory] Use OS::Allocate for all OS memory allocations." This is a reland of 4899bcb66db1c8c8272b02fb1cedf9886e010f82 This is a reland of b73ee3344ac96f92b4418c3c07779eb95a8541cc Original change's description: > [Memory] Use OS::Allocate for all OS memory allocations. > > - Eliminates OS::ReserveRegion and OS::ReserveAlignedRegion. > - Changes OS::Allocate to take alignment parameter, reorders parameters > to match page_allocator. > - Since the size of memory allocation can be deduced, don't return the > amount of memory allocated. > - Changes reservation of aligned address space. Before we would reserve > (size + alignment) rounded up to page size. This is too much, because > maximum misalignment is (alignment - page_size). > - On Windows and Cygwin, we release an oversize allocation and > immediately retry at the aligned address in the allocation. If we > lose the address due to a race, we just retry. > - Clean up all the calls to OS::Allocate in codegen and tests by adding > helper AllocateSystemPage function (allocation.h) and > AllocateAssemblerBuffer (cctest.h). > - Changes 'assm' to 'masm' in some targets for consistency when using > a macro-assembler. > > - Eliminates OS::ReleaseRegion, replacing with calls to OS::Free. > - Adds bool return value to OS::Free. > - Cleans up types of flags, protection on Windows and Cygwin. > Bug: chromium:756050 > Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng > Change-Id: I306dbe042cc867670fdc935abca29db074b0da71 Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng Change-Id: Iad3c025334e8f8d7d647be99a36a11ee449c9087 Reviewed-on: https://chromium-review.googlesource.com/767014 Commit-Queue: Bill Budge <bbudge@chromium.org> Reviewed-by: Michael Lippautz <mlippautz@chromium.org> Cr-Commit-Position: refs/heads/master@{#49363}
2017-11-14 15:55:09 +00:00
masm.GetCode(isolate, &desc);
Reland "Reland "[deoptimizer] Change deopt entries into builtins"" This is a reland of fbfa9bf4ec72b1b73a96b70ccb68cd98c321511b The arm64 was missing proper codegen for CFI, thus sizes were off. Original change's description: > Reland "[deoptimizer] Change deopt entries into builtins" > > This is a reland of 7f58ced72eb65b6b5530ccabaf2eaebe45bf9d33 > > It fixes the different exit size emitted on x64/Atom CPUs due to > performance tuning in TurboAssembler::Call. Additionally, add > cctests to verify the fixed size exits. > > Original change's description: > > [deoptimizer] Change deopt entries into builtins > > > > While the overall goal of this commit is to change deoptimization > > entries into builtins, there are multiple related things happening: > > > > - Deoptimization entries, formerly stubs (i.e. Code objects generated > > at runtime, guaranteed to be immovable), have been converted into > > builtins. The major restriction is that we now need to preserve the > > kRootRegister, which was formerly used on most architectures to pass > > the deoptimization id. The solution differs based on platform. > > - Renamed DEOPT_ENTRIES_OR_FOR_TESTING code kind to FOR_TESTING. > > - Removed heap/ support for immovable Code generation. > > - Removed the DeserializerData class (no longer needed). > > - arm64: to preserve 4-byte deopt exits, introduced a new optimization > > in which the final jump to the deoptimization entry is generated > > once per Code object, and deopt exits can continue to emit a > > near-call. > > - arm,ia32,x64: change to fixed-size deopt exits. This reduces exit > > sizes by 4/8, 5, and 5 bytes, respectively. > > > > On arm the deopt exit size is reduced from 12 (or 16) bytes to 8 bytes > > by using the same strategy as on arm64 (recalc deopt id from return > > address). Before: > > > > e300a002 movw r10, <id> > > e59fc024 ldr ip, [pc, <entry offset>] > > e12fff3c blx ip > > > > After: > > > > e59acb35 ldr ip, [r10, <entry offset>] > > e12fff3c blx ip > > > > On arm64 the deopt exit size remains 4 bytes (or 8 bytes in same cases > > with CFI). Additionally, up to 4 builtin jumps are emitted per Code > > object (max 32 bytes added overhead per Code object). Before: > > > > 9401cdae bl <entry offset> > > > > After: > > > > # eager deoptimization entry jump. > > f95b1f50 ldr x16, [x26, <eager entry offset>] > > d61f0200 br x16 > > # lazy deoptimization entry jump. > > f95b2b50 ldr x16, [x26, <lazy entry offset>] > > d61f0200 br x16 > > # the deopt exit. > > 97fffffc bl <eager deoptimization entry jump offset> > > > > On ia32 the deopt exit size is reduced from 10 to 5 bytes. Before: > > > > bb00000000 mov ebx,<id> > > e825f5372b call <entry> > > > > After: > > > > e8ea2256ba call <entry> > > > > On x64 the deopt exit size is reduced from 12 to 7 bytes. Before: > > > > 49c7c511000000 REX.W movq r13,<id> > > e8ea2f0700 call <entry> > > > > After: > > > > 41ff9560360000 call [r13+<entry offset>] > > > > Bug: v8:8661,v8:8768 > > Change-Id: I13e30aedc360474dc818fecc528ce87c3bfeed42 > > Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2465834 > > Commit-Queue: Jakob Gruber <jgruber@chromium.org> > > Reviewed-by: Ross McIlroy <rmcilroy@chromium.org> > > Reviewed-by: Tobias Tebbi <tebbi@chromium.org> > > Reviewed-by: Ulan Degenbaev <ulan@chromium.org> > > Cr-Commit-Position: refs/heads/master@{#70597} > > Tbr: ulan@chromium.org, tebbi@chromium.org, rmcilroy@chromium.org > Bug: v8:8661,v8:8768,chromium:1140165 > Change-Id: Ibcd5c39c58a70bf2b2ac221aa375fc68d495e144 > Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2485506 > Reviewed-by: Jakob Gruber <jgruber@chromium.org> > Reviewed-by: Tobias Tebbi <tebbi@chromium.org> > Commit-Queue: Jakob Gruber <jgruber@chromium.org> > Cr-Commit-Position: refs/heads/master@{#70655} Tbr: ulan@chromium.org, tebbi@chromium.org, rmcilroy@chromium.org Bug: v8:8661 Bug: v8:8768 Bug: chromium:1140165 Change-Id: I471cc94fc085e527dc9bfb5a84b96bd907c2333f Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2488682 Reviewed-by: Jakob Gruber <jgruber@chromium.org> Commit-Queue: Jakob Gruber <jgruber@chromium.org> Cr-Commit-Position: refs/heads/master@{#70672}
2020-10-21 05:12:25 +00:00
Handle<Code> code =
Factory::CodeBuilder(isolate, desc, CodeKind::FOR_TESTING).Build();
#ifdef OBJECT_PRINT
StdoutStream os;
code->Print(os);
#endif
auto f = GeneratedCode<F0>::FromCode(*code);
CHECK_EQ(0, f.Call());
}
TEST(AssemblerX64POPCNT) {
CcTest::InitializeVM();
if (!CpuFeatures::IsSupported(POPCNT)) return;
Isolate* isolate = reinterpret_cast<Isolate*>(CcTest::isolate());
HandleScope scope(isolate);
v8::internal::byte buffer[256];
MacroAssembler masm(isolate, v8::internal::CodeObjectRequired::kYes,
ExternalAssemblerBuffer(buffer, sizeof(buffer)));
{
Reland "[Memory] Use OS::Allocate for all OS memory allocations." This is a reland of 4899bcb66db1c8c8272b02fb1cedf9886e010f82 This is a reland of b73ee3344ac96f92b4418c3c07779eb95a8541cc Original change's description: > [Memory] Use OS::Allocate for all OS memory allocations. > > - Eliminates OS::ReserveRegion and OS::ReserveAlignedRegion. > - Changes OS::Allocate to take alignment parameter, reorders parameters > to match page_allocator. > - Since the size of memory allocation can be deduced, don't return the > amount of memory allocated. > - Changes reservation of aligned address space. Before we would reserve > (size + alignment) rounded up to page size. This is too much, because > maximum misalignment is (alignment - page_size). > - On Windows and Cygwin, we release an oversize allocation and > immediately retry at the aligned address in the allocation. If we > lose the address due to a race, we just retry. > - Clean up all the calls to OS::Allocate in codegen and tests by adding > helper AllocateSystemPage function (allocation.h) and > AllocateAssemblerBuffer (cctest.h). > - Changes 'assm' to 'masm' in some targets for consistency when using > a macro-assembler. > > - Eliminates OS::ReleaseRegion, replacing with calls to OS::Free. > - Adds bool return value to OS::Free. > - Cleans up types of flags, protection on Windows and Cygwin. > Bug: chromium:756050 > Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng > Change-Id: I306dbe042cc867670fdc935abca29db074b0da71 Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng Change-Id: Iad3c025334e8f8d7d647be99a36a11ee449c9087 Reviewed-on: https://chromium-review.googlesource.com/767014 Commit-Queue: Bill Budge <bbudge@chromium.org> Reviewed-by: Michael Lippautz <mlippautz@chromium.org> Cr-Commit-Position: refs/heads/master@{#49363}
2017-11-14 15:55:09 +00:00
CpuFeatureScope fscope(&masm, POPCNT);
Label exit;
__ movq(rcx, uint64_t{0x1111111111111100}); // source operand
__ pushq(rcx); // For memory operand
__ movl(rax, Immediate(1)); // Test number
__ popcntq(r8, rcx);
__ movq(r9, uint64_t{0x000000000000000E}); // expected result
__ cmpq(r8, r9);
__ j(not_equal, &exit);
__ incq(rax);
__ popcntq(r8, Operand(rsp, 0));
__ movq(r9, uint64_t{0x000000000000000E}); // expected result
__ cmpq(r8, r9);
__ j(not_equal, &exit);
__ incq(rax);
__ popcntl(r8, rcx);
__ movq(r9, uint64_t{0x0000000000000006}); // expected result
__ cmpq(r8, r9);
__ j(not_equal, &exit);
__ incq(rax);
__ popcntl(r8, Operand(rsp, 0));
__ movq(r9, uint64_t{0x0000000000000006}); // expected result
__ cmpq(r8, r9);
__ j(not_equal, &exit);
__ xorl(rax, rax);
__ bind(&exit);
__ popq(rcx);
__ ret(0);
}
CodeDesc desc;
Reland "[Memory] Use OS::Allocate for all OS memory allocations." This is a reland of 4899bcb66db1c8c8272b02fb1cedf9886e010f82 This is a reland of b73ee3344ac96f92b4418c3c07779eb95a8541cc Original change's description: > [Memory] Use OS::Allocate for all OS memory allocations. > > - Eliminates OS::ReserveRegion and OS::ReserveAlignedRegion. > - Changes OS::Allocate to take alignment parameter, reorders parameters > to match page_allocator. > - Since the size of memory allocation can be deduced, don't return the > amount of memory allocated. > - Changes reservation of aligned address space. Before we would reserve > (size + alignment) rounded up to page size. This is too much, because > maximum misalignment is (alignment - page_size). > - On Windows and Cygwin, we release an oversize allocation and > immediately retry at the aligned address in the allocation. If we > lose the address due to a race, we just retry. > - Clean up all the calls to OS::Allocate in codegen and tests by adding > helper AllocateSystemPage function (allocation.h) and > AllocateAssemblerBuffer (cctest.h). > - Changes 'assm' to 'masm' in some targets for consistency when using > a macro-assembler. > > - Eliminates OS::ReleaseRegion, replacing with calls to OS::Free. > - Adds bool return value to OS::Free. > - Cleans up types of flags, protection on Windows and Cygwin. > Bug: chromium:756050 > Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng > Change-Id: I306dbe042cc867670fdc935abca29db074b0da71 Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng Change-Id: Iad3c025334e8f8d7d647be99a36a11ee449c9087 Reviewed-on: https://chromium-review.googlesource.com/767014 Commit-Queue: Bill Budge <bbudge@chromium.org> Reviewed-by: Michael Lippautz <mlippautz@chromium.org> Cr-Commit-Position: refs/heads/master@{#49363}
2017-11-14 15:55:09 +00:00
masm.GetCode(isolate, &desc);
Reland "Reland "[deoptimizer] Change deopt entries into builtins"" This is a reland of fbfa9bf4ec72b1b73a96b70ccb68cd98c321511b The arm64 was missing proper codegen for CFI, thus sizes were off. Original change's description: > Reland "[deoptimizer] Change deopt entries into builtins" > > This is a reland of 7f58ced72eb65b6b5530ccabaf2eaebe45bf9d33 > > It fixes the different exit size emitted on x64/Atom CPUs due to > performance tuning in TurboAssembler::Call. Additionally, add > cctests to verify the fixed size exits. > > Original change's description: > > [deoptimizer] Change deopt entries into builtins > > > > While the overall goal of this commit is to change deoptimization > > entries into builtins, there are multiple related things happening: > > > > - Deoptimization entries, formerly stubs (i.e. Code objects generated > > at runtime, guaranteed to be immovable), have been converted into > > builtins. The major restriction is that we now need to preserve the > > kRootRegister, which was formerly used on most architectures to pass > > the deoptimization id. The solution differs based on platform. > > - Renamed DEOPT_ENTRIES_OR_FOR_TESTING code kind to FOR_TESTING. > > - Removed heap/ support for immovable Code generation. > > - Removed the DeserializerData class (no longer needed). > > - arm64: to preserve 4-byte deopt exits, introduced a new optimization > > in which the final jump to the deoptimization entry is generated > > once per Code object, and deopt exits can continue to emit a > > near-call. > > - arm,ia32,x64: change to fixed-size deopt exits. This reduces exit > > sizes by 4/8, 5, and 5 bytes, respectively. > > > > On arm the deopt exit size is reduced from 12 (or 16) bytes to 8 bytes > > by using the same strategy as on arm64 (recalc deopt id from return > > address). Before: > > > > e300a002 movw r10, <id> > > e59fc024 ldr ip, [pc, <entry offset>] > > e12fff3c blx ip > > > > After: > > > > e59acb35 ldr ip, [r10, <entry offset>] > > e12fff3c blx ip > > > > On arm64 the deopt exit size remains 4 bytes (or 8 bytes in same cases > > with CFI). Additionally, up to 4 builtin jumps are emitted per Code > > object (max 32 bytes added overhead per Code object). Before: > > > > 9401cdae bl <entry offset> > > > > After: > > > > # eager deoptimization entry jump. > > f95b1f50 ldr x16, [x26, <eager entry offset>] > > d61f0200 br x16 > > # lazy deoptimization entry jump. > > f95b2b50 ldr x16, [x26, <lazy entry offset>] > > d61f0200 br x16 > > # the deopt exit. > > 97fffffc bl <eager deoptimization entry jump offset> > > > > On ia32 the deopt exit size is reduced from 10 to 5 bytes. Before: > > > > bb00000000 mov ebx,<id> > > e825f5372b call <entry> > > > > After: > > > > e8ea2256ba call <entry> > > > > On x64 the deopt exit size is reduced from 12 to 7 bytes. Before: > > > > 49c7c511000000 REX.W movq r13,<id> > > e8ea2f0700 call <entry> > > > > After: > > > > 41ff9560360000 call [r13+<entry offset>] > > > > Bug: v8:8661,v8:8768 > > Change-Id: I13e30aedc360474dc818fecc528ce87c3bfeed42 > > Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2465834 > > Commit-Queue: Jakob Gruber <jgruber@chromium.org> > > Reviewed-by: Ross McIlroy <rmcilroy@chromium.org> > > Reviewed-by: Tobias Tebbi <tebbi@chromium.org> > > Reviewed-by: Ulan Degenbaev <ulan@chromium.org> > > Cr-Commit-Position: refs/heads/master@{#70597} > > Tbr: ulan@chromium.org, tebbi@chromium.org, rmcilroy@chromium.org > Bug: v8:8661,v8:8768,chromium:1140165 > Change-Id: Ibcd5c39c58a70bf2b2ac221aa375fc68d495e144 > Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2485506 > Reviewed-by: Jakob Gruber <jgruber@chromium.org> > Reviewed-by: Tobias Tebbi <tebbi@chromium.org> > Commit-Queue: Jakob Gruber <jgruber@chromium.org> > Cr-Commit-Position: refs/heads/master@{#70655} Tbr: ulan@chromium.org, tebbi@chromium.org, rmcilroy@chromium.org Bug: v8:8661 Bug: v8:8768 Bug: chromium:1140165 Change-Id: I471cc94fc085e527dc9bfb5a84b96bd907c2333f Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2488682 Reviewed-by: Jakob Gruber <jgruber@chromium.org> Commit-Queue: Jakob Gruber <jgruber@chromium.org> Cr-Commit-Position: refs/heads/master@{#70672}
2020-10-21 05:12:25 +00:00
Handle<Code> code =
Factory::CodeBuilder(isolate, desc, CodeKind::FOR_TESTING).Build();
#ifdef OBJECT_PRINT
StdoutStream os;
code->Print(os);
#endif
auto f = GeneratedCode<F0>::FromCode(*code);
CHECK_EQ(0, f.Call());
}
TEST(AssemblerX64BMI2) {
CcTest::InitializeVM();
if (!CpuFeatures::IsSupported(BMI2)) return;
Isolate* isolate = reinterpret_cast<Isolate*>(CcTest::isolate());
HandleScope scope(isolate);
v8::internal::byte buffer[2048];
MacroAssembler masm(isolate, v8::internal::CodeObjectRequired::kYes,
ExternalAssemblerBuffer(buffer, sizeof(buffer)));
{
Reland "[Memory] Use OS::Allocate for all OS memory allocations." This is a reland of 4899bcb66db1c8c8272b02fb1cedf9886e010f82 This is a reland of b73ee3344ac96f92b4418c3c07779eb95a8541cc Original change's description: > [Memory] Use OS::Allocate for all OS memory allocations. > > - Eliminates OS::ReserveRegion and OS::ReserveAlignedRegion. > - Changes OS::Allocate to take alignment parameter, reorders parameters > to match page_allocator. > - Since the size of memory allocation can be deduced, don't return the > amount of memory allocated. > - Changes reservation of aligned address space. Before we would reserve > (size + alignment) rounded up to page size. This is too much, because > maximum misalignment is (alignment - page_size). > - On Windows and Cygwin, we release an oversize allocation and > immediately retry at the aligned address in the allocation. If we > lose the address due to a race, we just retry. > - Clean up all the calls to OS::Allocate in codegen and tests by adding > helper AllocateSystemPage function (allocation.h) and > AllocateAssemblerBuffer (cctest.h). > - Changes 'assm' to 'masm' in some targets for consistency when using > a macro-assembler. > > - Eliminates OS::ReleaseRegion, replacing with calls to OS::Free. > - Adds bool return value to OS::Free. > - Cleans up types of flags, protection on Windows and Cygwin. > Bug: chromium:756050 > Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng > Change-Id: I306dbe042cc867670fdc935abca29db074b0da71 Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng Change-Id: Iad3c025334e8f8d7d647be99a36a11ee449c9087 Reviewed-on: https://chromium-review.googlesource.com/767014 Commit-Queue: Bill Budge <bbudge@chromium.org> Reviewed-by: Michael Lippautz <mlippautz@chromium.org> Cr-Commit-Position: refs/heads/master@{#49363}
2017-11-14 15:55:09 +00:00
CpuFeatureScope fscope(&masm, BMI2);
Label exit;
__ pushq(rbx); // save rbx
__ movq(rcx, uint64_t{0x1122334455667788}); // source operand
__ pushq(rcx); // For memory operand
// bzhi
__ movq(rdx, uint64_t{0x0000000000000009});
__ movl(rax, Immediate(1)); // Test number
__ bzhiq(r8, rcx, rdx);
__ movq(r9, uint64_t{0x0000000000000188}); // expected result
__ cmpq(r8, r9);
__ j(not_equal, &exit);
__ incq(rax);
__ bzhiq(r8, Operand(rsp, 0), rdx);
__ movq(r9, uint64_t{0x0000000000000188}); // expected result
__ cmpq(r8, r9);
__ j(not_equal, &exit);
__ incq(rax);
__ bzhil(r8, rcx, rdx);
__ movq(r9, uint64_t{0x0000000000000188}); // expected result
__ cmpq(r8, r9);
__ j(not_equal, &exit);
__ incq(rax);
__ bzhil(r8, Operand(rsp, 0), rdx);
__ movq(r9, uint64_t{0x0000000000000188}); // expected result
__ cmpq(r8, r9);
__ j(not_equal, &exit);
// mulx
__ movq(rdx, uint64_t{0x0000000000001000});
__ incq(rax);
__ mulxq(r8, r9, rcx);
__ movq(rbx, uint64_t{0x0000000000000112}); // expected result
__ cmpq(r8, rbx);
__ j(not_equal, &exit);
__ movq(rbx, uint64_t{0x2334455667788000}); // expected result
__ cmpq(r9, rbx);
__ j(not_equal, &exit);
__ incq(rax);
__ mulxq(r8, r9, Operand(rsp, 0));
__ movq(rbx, uint64_t{0x0000000000000112}); // expected result
__ cmpq(r8, rbx);
__ j(not_equal, &exit);
__ movq(rbx, uint64_t{0x2334455667788000}); // expected result
__ cmpq(r9, rbx);
__ j(not_equal, &exit);
__ incq(rax);
__ mulxl(r8, r9, rcx);
__ movq(rbx, uint64_t{0x0000000000000556}); // expected result
__ cmpq(r8, rbx);
__ j(not_equal, &exit);
__ movq(rbx, uint64_t{0x0000000067788000}); // expected result
__ cmpq(r9, rbx);
__ j(not_equal, &exit);
__ incq(rax);
__ mulxl(r8, r9, Operand(rsp, 0));
__ movq(rbx, uint64_t{0x0000000000000556}); // expected result
__ cmpq(r8, rbx);
__ j(not_equal, &exit);
__ movq(rbx, uint64_t{0x0000000067788000}); // expected result
__ cmpq(r9, rbx);
__ j(not_equal, &exit);
// pdep
__ movq(rdx, uint64_t{0xFFFFFFFFFFFFFFF0});
__ incq(rax);
__ pdepq(r8, rdx, rcx);
__ movq(r9, uint64_t{0x1122334455667400}); // expected result
__ cmpq(r8, r9);
__ j(not_equal, &exit);
__ incq(rax);
__ pdepq(r8, rdx, Operand(rsp, 0));
__ movq(r9, uint64_t{0x1122334455667400}); // expected result
__ cmpq(r8, r9);
__ j(not_equal, &exit);
__ incq(rax);
__ pdepl(r8, rdx, rcx);
__ movq(r9, uint64_t{0x0000000055667400}); // expected result
__ cmpq(r8, r9);
__ j(not_equal, &exit);
__ incq(rax);
__ pdepl(r8, rdx, Operand(rsp, 0));
__ movq(r9, uint64_t{0x0000000055667400}); // expected result
__ cmpq(r8, r9);
__ j(not_equal, &exit);
// pext
__ movq(rdx, uint64_t{0xFFFFFFFFFFFFFFF0});
__ incq(rax);
__ pextq(r8, rdx, rcx);
__ movq(r9, uint64_t{0x0000000003FFFFFE}); // expected result
__ cmpq(r8, r9);
__ j(not_equal, &exit);
__ incq(rax);
__ pextq(r8, rdx, Operand(rsp, 0));
__ movq(r9, uint64_t{0x0000000003FFFFFE}); // expected result
__ cmpq(r8, r9);
__ j(not_equal, &exit);
__ incq(rax);
__ pextl(r8, rdx, rcx);
__ movq(r9, uint64_t{0x000000000000FFFE}); // expected result
__ cmpq(r8, r9);
__ j(not_equal, &exit);
__ incq(rax);
__ pextl(r8, rdx, Operand(rsp, 0));
__ movq(r9, uint64_t{0x000000000000FFFE}); // expected result
__ cmpq(r8, r9);
__ j(not_equal, &exit);
// sarx
__ movq(rdx, uint64_t{0x0000000000000004});
__ incq(rax);
__ sarxq(r8, rcx, rdx);
__ movq(r9, uint64_t{0x0112233445566778}); // expected result
__ cmpq(r8, r9);
__ j(not_equal, &exit);
__ incq(rax);
__ sarxq(r8, Operand(rsp, 0), rdx);
__ movq(r9, uint64_t{0x0112233445566778}); // expected result
__ cmpq(r8, r9);
__ j(not_equal, &exit);
__ incq(rax);
__ sarxl(r8, rcx, rdx);
__ movq(r9, uint64_t{0x0000000005566778}); // expected result
__ cmpq(r8, r9);
__ j(not_equal, &exit);
__ incq(rax);
__ sarxl(r8, Operand(rsp, 0), rdx);
__ movq(r9, uint64_t{0x0000000005566778}); // expected result
__ cmpq(r8, r9);
__ j(not_equal, &exit);
// shlx
__ movq(rdx, uint64_t{0x0000000000000004});
__ incq(rax);
__ shlxq(r8, rcx, rdx);
__ movq(r9, uint64_t{0x1223344556677880}); // expected result
__ cmpq(r8, r9);
__ j(not_equal, &exit);
__ incq(rax);
__ shlxq(r8, Operand(rsp, 0), rdx);
__ movq(r9, uint64_t{0x1223344556677880}); // expected result
__ cmpq(r8, r9);
__ j(not_equal, &exit);
__ incq(rax);
__ shlxl(r8, rcx, rdx);
__ movq(r9, uint64_t{0x0000000056677880}); // expected result
__ cmpq(r8, r9);
__ j(not_equal, &exit);
__ incq(rax);
__ shlxl(r8, Operand(rsp, 0), rdx);
__ movq(r9, uint64_t{0x0000000056677880}); // expected result
__ cmpq(r8, r9);
__ j(not_equal, &exit);
// shrx
__ movq(rdx, uint64_t{0x0000000000000004});
__ incq(rax);
__ shrxq(r8, rcx, rdx);
__ movq(r9, uint64_t{0x0112233445566778}); // expected result
__ cmpq(r8, r9);
__ j(not_equal, &exit);
__ incq(rax);
__ shrxq(r8, Operand(rsp, 0), rdx);
__ movq(r9, uint64_t{0x0112233445566778}); // expected result
__ cmpq(r8, r9);
__ j(not_equal, &exit);
__ incq(rax);
__ shrxl(r8, rcx, rdx);
__ movq(r9, uint64_t{0x0000000005566778}); // expected result
__ cmpq(r8, r9);
__ j(not_equal, &exit);
__ incq(rax);
__ shrxl(r8, Operand(rsp, 0), rdx);
__ movq(r9, uint64_t{0x0000000005566778}); // expected result
__ cmpq(r8, r9);
__ j(not_equal, &exit);
// rorx
__ incq(rax);
__ rorxq(r8, rcx, 0x4);
__ movq(r9, uint64_t{0x8112233445566778}); // expected result
__ cmpq(r8, r9);
__ j(not_equal, &exit);
__ incq(rax);
__ rorxq(r8, Operand(rsp, 0), 0x4);
__ movq(r9, uint64_t{0x8112233445566778}); // expected result
__ cmpq(r8, r9);
__ j(not_equal, &exit);
__ incq(rax);
__ rorxl(r8, rcx, 0x4);
__ movq(r9, uint64_t{0x0000000085566778}); // expected result
__ cmpq(r8, r9);
__ j(not_equal, &exit);
__ incq(rax);
__ rorxl(r8, Operand(rsp, 0), 0x4);
__ movq(r9, uint64_t{0x0000000085566778}); // expected result
__ cmpq(r8, r9);
__ j(not_equal, &exit);
__ xorl(rax, rax);
__ bind(&exit);
__ popq(rcx);
__ popq(rbx);
__ ret(0);
}
CodeDesc desc;
Reland "[Memory] Use OS::Allocate for all OS memory allocations." This is a reland of 4899bcb66db1c8c8272b02fb1cedf9886e010f82 This is a reland of b73ee3344ac96f92b4418c3c07779eb95a8541cc Original change's description: > [Memory] Use OS::Allocate for all OS memory allocations. > > - Eliminates OS::ReserveRegion and OS::ReserveAlignedRegion. > - Changes OS::Allocate to take alignment parameter, reorders parameters > to match page_allocator. > - Since the size of memory allocation can be deduced, don't return the > amount of memory allocated. > - Changes reservation of aligned address space. Before we would reserve > (size + alignment) rounded up to page size. This is too much, because > maximum misalignment is (alignment - page_size). > - On Windows and Cygwin, we release an oversize allocation and > immediately retry at the aligned address in the allocation. If we > lose the address due to a race, we just retry. > - Clean up all the calls to OS::Allocate in codegen and tests by adding > helper AllocateSystemPage function (allocation.h) and > AllocateAssemblerBuffer (cctest.h). > - Changes 'assm' to 'masm' in some targets for consistency when using > a macro-assembler. > > - Eliminates OS::ReleaseRegion, replacing with calls to OS::Free. > - Adds bool return value to OS::Free. > - Cleans up types of flags, protection on Windows and Cygwin. > Bug: chromium:756050 > Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng > Change-Id: I306dbe042cc867670fdc935abca29db074b0da71 Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng Change-Id: Iad3c025334e8f8d7d647be99a36a11ee449c9087 Reviewed-on: https://chromium-review.googlesource.com/767014 Commit-Queue: Bill Budge <bbudge@chromium.org> Reviewed-by: Michael Lippautz <mlippautz@chromium.org> Cr-Commit-Position: refs/heads/master@{#49363}
2017-11-14 15:55:09 +00:00
masm.GetCode(isolate, &desc);
Reland "Reland "[deoptimizer] Change deopt entries into builtins"" This is a reland of fbfa9bf4ec72b1b73a96b70ccb68cd98c321511b The arm64 was missing proper codegen for CFI, thus sizes were off. Original change's description: > Reland "[deoptimizer] Change deopt entries into builtins" > > This is a reland of 7f58ced72eb65b6b5530ccabaf2eaebe45bf9d33 > > It fixes the different exit size emitted on x64/Atom CPUs due to > performance tuning in TurboAssembler::Call. Additionally, add > cctests to verify the fixed size exits. > > Original change's description: > > [deoptimizer] Change deopt entries into builtins > > > > While the overall goal of this commit is to change deoptimization > > entries into builtins, there are multiple related things happening: > > > > - Deoptimization entries, formerly stubs (i.e. Code objects generated > > at runtime, guaranteed to be immovable), have been converted into > > builtins. The major restriction is that we now need to preserve the > > kRootRegister, which was formerly used on most architectures to pass > > the deoptimization id. The solution differs based on platform. > > - Renamed DEOPT_ENTRIES_OR_FOR_TESTING code kind to FOR_TESTING. > > - Removed heap/ support for immovable Code generation. > > - Removed the DeserializerData class (no longer needed). > > - arm64: to preserve 4-byte deopt exits, introduced a new optimization > > in which the final jump to the deoptimization entry is generated > > once per Code object, and deopt exits can continue to emit a > > near-call. > > - arm,ia32,x64: change to fixed-size deopt exits. This reduces exit > > sizes by 4/8, 5, and 5 bytes, respectively. > > > > On arm the deopt exit size is reduced from 12 (or 16) bytes to 8 bytes > > by using the same strategy as on arm64 (recalc deopt id from return > > address). Before: > > > > e300a002 movw r10, <id> > > e59fc024 ldr ip, [pc, <entry offset>] > > e12fff3c blx ip > > > > After: > > > > e59acb35 ldr ip, [r10, <entry offset>] > > e12fff3c blx ip > > > > On arm64 the deopt exit size remains 4 bytes (or 8 bytes in same cases > > with CFI). Additionally, up to 4 builtin jumps are emitted per Code > > object (max 32 bytes added overhead per Code object). Before: > > > > 9401cdae bl <entry offset> > > > > After: > > > > # eager deoptimization entry jump. > > f95b1f50 ldr x16, [x26, <eager entry offset>] > > d61f0200 br x16 > > # lazy deoptimization entry jump. > > f95b2b50 ldr x16, [x26, <lazy entry offset>] > > d61f0200 br x16 > > # the deopt exit. > > 97fffffc bl <eager deoptimization entry jump offset> > > > > On ia32 the deopt exit size is reduced from 10 to 5 bytes. Before: > > > > bb00000000 mov ebx,<id> > > e825f5372b call <entry> > > > > After: > > > > e8ea2256ba call <entry> > > > > On x64 the deopt exit size is reduced from 12 to 7 bytes. Before: > > > > 49c7c511000000 REX.W movq r13,<id> > > e8ea2f0700 call <entry> > > > > After: > > > > 41ff9560360000 call [r13+<entry offset>] > > > > Bug: v8:8661,v8:8768 > > Change-Id: I13e30aedc360474dc818fecc528ce87c3bfeed42 > > Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2465834 > > Commit-Queue: Jakob Gruber <jgruber@chromium.org> > > Reviewed-by: Ross McIlroy <rmcilroy@chromium.org> > > Reviewed-by: Tobias Tebbi <tebbi@chromium.org> > > Reviewed-by: Ulan Degenbaev <ulan@chromium.org> > > Cr-Commit-Position: refs/heads/master@{#70597} > > Tbr: ulan@chromium.org, tebbi@chromium.org, rmcilroy@chromium.org > Bug: v8:8661,v8:8768,chromium:1140165 > Change-Id: Ibcd5c39c58a70bf2b2ac221aa375fc68d495e144 > Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2485506 > Reviewed-by: Jakob Gruber <jgruber@chromium.org> > Reviewed-by: Tobias Tebbi <tebbi@chromium.org> > Commit-Queue: Jakob Gruber <jgruber@chromium.org> > Cr-Commit-Position: refs/heads/master@{#70655} Tbr: ulan@chromium.org, tebbi@chromium.org, rmcilroy@chromium.org Bug: v8:8661 Bug: v8:8768 Bug: chromium:1140165 Change-Id: I471cc94fc085e527dc9bfb5a84b96bd907c2333f Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2488682 Reviewed-by: Jakob Gruber <jgruber@chromium.org> Commit-Queue: Jakob Gruber <jgruber@chromium.org> Cr-Commit-Position: refs/heads/master@{#70672}
2020-10-21 05:12:25 +00:00
Handle<Code> code =
Factory::CodeBuilder(isolate, desc, CodeKind::FOR_TESTING).Build();
#ifdef OBJECT_PRINT
StdoutStream os;
code->Print(os);
#endif
auto f = GeneratedCode<F0>::FromCode(*code);
CHECK_EQ(0, f.Call());
}
TEST(AssemblerX64JumpTables1) {
// Test jump tables with forward jumps.
CcTest::InitializeVM();
Isolate* isolate = reinterpret_cast<Isolate*>(CcTest::isolate());
HandleScope scope(isolate);
MacroAssembler masm(isolate, v8::internal::CodeObjectRequired::kYes);
const int kNumCases = 512;
int values[kNumCases];
isolate->random_number_generator()->NextBytes(values, sizeof(values));
Label labels[kNumCases];
Label done, table;
__ leaq(arg2, Operand(&table));
__ jmp(Operand(arg2, arg1, times_8, 0));
__ ud2();
__ bind(&table);
for (int i = 0; i < kNumCases; ++i) {
__ dq(&labels[i]);
}
for (int i = 0; i < kNumCases; ++i) {
__ bind(&labels[i]);
__ movq(rax, Immediate(values[i]));
__ jmp(&done);
}
__ bind(&done);
__ ret(0);
CodeDesc desc;
Reland "[Memory] Use OS::Allocate for all OS memory allocations." This is a reland of 4899bcb66db1c8c8272b02fb1cedf9886e010f82 This is a reland of b73ee3344ac96f92b4418c3c07779eb95a8541cc Original change's description: > [Memory] Use OS::Allocate for all OS memory allocations. > > - Eliminates OS::ReserveRegion and OS::ReserveAlignedRegion. > - Changes OS::Allocate to take alignment parameter, reorders parameters > to match page_allocator. > - Since the size of memory allocation can be deduced, don't return the > amount of memory allocated. > - Changes reservation of aligned address space. Before we would reserve > (size + alignment) rounded up to page size. This is too much, because > maximum misalignment is (alignment - page_size). > - On Windows and Cygwin, we release an oversize allocation and > immediately retry at the aligned address in the allocation. If we > lose the address due to a race, we just retry. > - Clean up all the calls to OS::Allocate in codegen and tests by adding > helper AllocateSystemPage function (allocation.h) and > AllocateAssemblerBuffer (cctest.h). > - Changes 'assm' to 'masm' in some targets for consistency when using > a macro-assembler. > > - Eliminates OS::ReleaseRegion, replacing with calls to OS::Free. > - Adds bool return value to OS::Free. > - Cleans up types of flags, protection on Windows and Cygwin. > Bug: chromium:756050 > Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng > Change-Id: I306dbe042cc867670fdc935abca29db074b0da71 Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng Change-Id: Iad3c025334e8f8d7d647be99a36a11ee449c9087 Reviewed-on: https://chromium-review.googlesource.com/767014 Commit-Queue: Bill Budge <bbudge@chromium.org> Reviewed-by: Michael Lippautz <mlippautz@chromium.org> Cr-Commit-Position: refs/heads/master@{#49363}
2017-11-14 15:55:09 +00:00
masm.GetCode(isolate, &desc);
Reland "Reland "[deoptimizer] Change deopt entries into builtins"" This is a reland of fbfa9bf4ec72b1b73a96b70ccb68cd98c321511b The arm64 was missing proper codegen for CFI, thus sizes were off. Original change's description: > Reland "[deoptimizer] Change deopt entries into builtins" > > This is a reland of 7f58ced72eb65b6b5530ccabaf2eaebe45bf9d33 > > It fixes the different exit size emitted on x64/Atom CPUs due to > performance tuning in TurboAssembler::Call. Additionally, add > cctests to verify the fixed size exits. > > Original change's description: > > [deoptimizer] Change deopt entries into builtins > > > > While the overall goal of this commit is to change deoptimization > > entries into builtins, there are multiple related things happening: > > > > - Deoptimization entries, formerly stubs (i.e. Code objects generated > > at runtime, guaranteed to be immovable), have been converted into > > builtins. The major restriction is that we now need to preserve the > > kRootRegister, which was formerly used on most architectures to pass > > the deoptimization id. The solution differs based on platform. > > - Renamed DEOPT_ENTRIES_OR_FOR_TESTING code kind to FOR_TESTING. > > - Removed heap/ support for immovable Code generation. > > - Removed the DeserializerData class (no longer needed). > > - arm64: to preserve 4-byte deopt exits, introduced a new optimization > > in which the final jump to the deoptimization entry is generated > > once per Code object, and deopt exits can continue to emit a > > near-call. > > - arm,ia32,x64: change to fixed-size deopt exits. This reduces exit > > sizes by 4/8, 5, and 5 bytes, respectively. > > > > On arm the deopt exit size is reduced from 12 (or 16) bytes to 8 bytes > > by using the same strategy as on arm64 (recalc deopt id from return > > address). Before: > > > > e300a002 movw r10, <id> > > e59fc024 ldr ip, [pc, <entry offset>] > > e12fff3c blx ip > > > > After: > > > > e59acb35 ldr ip, [r10, <entry offset>] > > e12fff3c blx ip > > > > On arm64 the deopt exit size remains 4 bytes (or 8 bytes in same cases > > with CFI). Additionally, up to 4 builtin jumps are emitted per Code > > object (max 32 bytes added overhead per Code object). Before: > > > > 9401cdae bl <entry offset> > > > > After: > > > > # eager deoptimization entry jump. > > f95b1f50 ldr x16, [x26, <eager entry offset>] > > d61f0200 br x16 > > # lazy deoptimization entry jump. > > f95b2b50 ldr x16, [x26, <lazy entry offset>] > > d61f0200 br x16 > > # the deopt exit. > > 97fffffc bl <eager deoptimization entry jump offset> > > > > On ia32 the deopt exit size is reduced from 10 to 5 bytes. Before: > > > > bb00000000 mov ebx,<id> > > e825f5372b call <entry> > > > > After: > > > > e8ea2256ba call <entry> > > > > On x64 the deopt exit size is reduced from 12 to 7 bytes. Before: > > > > 49c7c511000000 REX.W movq r13,<id> > > e8ea2f0700 call <entry> > > > > After: > > > > 41ff9560360000 call [r13+<entry offset>] > > > > Bug: v8:8661,v8:8768 > > Change-Id: I13e30aedc360474dc818fecc528ce87c3bfeed42 > > Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2465834 > > Commit-Queue: Jakob Gruber <jgruber@chromium.org> > > Reviewed-by: Ross McIlroy <rmcilroy@chromium.org> > > Reviewed-by: Tobias Tebbi <tebbi@chromium.org> > > Reviewed-by: Ulan Degenbaev <ulan@chromium.org> > > Cr-Commit-Position: refs/heads/master@{#70597} > > Tbr: ulan@chromium.org, tebbi@chromium.org, rmcilroy@chromium.org > Bug: v8:8661,v8:8768,chromium:1140165 > Change-Id: Ibcd5c39c58a70bf2b2ac221aa375fc68d495e144 > Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2485506 > Reviewed-by: Jakob Gruber <jgruber@chromium.org> > Reviewed-by: Tobias Tebbi <tebbi@chromium.org> > Commit-Queue: Jakob Gruber <jgruber@chromium.org> > Cr-Commit-Position: refs/heads/master@{#70655} Tbr: ulan@chromium.org, tebbi@chromium.org, rmcilroy@chromium.org Bug: v8:8661 Bug: v8:8768 Bug: chromium:1140165 Change-Id: I471cc94fc085e527dc9bfb5a84b96bd907c2333f Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2488682 Reviewed-by: Jakob Gruber <jgruber@chromium.org> Commit-Queue: Jakob Gruber <jgruber@chromium.org> Cr-Commit-Position: refs/heads/master@{#70672}
2020-10-21 05:12:25 +00:00
Handle<Code> code =
Factory::CodeBuilder(isolate, desc, CodeKind::FOR_TESTING).Build();
#ifdef OBJECT_PRINT
code->Print(std::cout);
#endif
auto f = GeneratedCode<F1>::FromCode(*code);
for (int i = 0; i < kNumCases; ++i) {
int res = f.Call(i);
PrintF("f(%d) = %d\n", i, res);
CHECK_EQ(values[i], res);
}
}
TEST(AssemblerX64JumpTables2) {
// Test jump tables with backwards jumps.
CcTest::InitializeVM();
Isolate* isolate = reinterpret_cast<Isolate*>(CcTest::isolate());
HandleScope scope(isolate);
MacroAssembler masm(isolate, v8::internal::CodeObjectRequired::kYes);
const int kNumCases = 512;
int values[kNumCases];
isolate->random_number_generator()->NextBytes(values, sizeof(values));
Label labels[kNumCases];
Label done, table;
__ leaq(arg2, Operand(&table));
__ jmp(Operand(arg2, arg1, times_8, 0));
__ ud2();
for (int i = 0; i < kNumCases; ++i) {
__ bind(&labels[i]);
__ movq(rax, Immediate(values[i]));
__ jmp(&done);
}
__ bind(&done);
__ ret(0);
__ bind(&table);
for (int i = 0; i < kNumCases; ++i) {
__ dq(&labels[i]);
}
CodeDesc desc;
Reland "[Memory] Use OS::Allocate for all OS memory allocations." This is a reland of 4899bcb66db1c8c8272b02fb1cedf9886e010f82 This is a reland of b73ee3344ac96f92b4418c3c07779eb95a8541cc Original change's description: > [Memory] Use OS::Allocate for all OS memory allocations. > > - Eliminates OS::ReserveRegion and OS::ReserveAlignedRegion. > - Changes OS::Allocate to take alignment parameter, reorders parameters > to match page_allocator. > - Since the size of memory allocation can be deduced, don't return the > amount of memory allocated. > - Changes reservation of aligned address space. Before we would reserve > (size + alignment) rounded up to page size. This is too much, because > maximum misalignment is (alignment - page_size). > - On Windows and Cygwin, we release an oversize allocation and > immediately retry at the aligned address in the allocation. If we > lose the address due to a race, we just retry. > - Clean up all the calls to OS::Allocate in codegen and tests by adding > helper AllocateSystemPage function (allocation.h) and > AllocateAssemblerBuffer (cctest.h). > - Changes 'assm' to 'masm' in some targets for consistency when using > a macro-assembler. > > - Eliminates OS::ReleaseRegion, replacing with calls to OS::Free. > - Adds bool return value to OS::Free. > - Cleans up types of flags, protection on Windows and Cygwin. > Bug: chromium:756050 > Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng > Change-Id: I306dbe042cc867670fdc935abca29db074b0da71 Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng Change-Id: Iad3c025334e8f8d7d647be99a36a11ee449c9087 Reviewed-on: https://chromium-review.googlesource.com/767014 Commit-Queue: Bill Budge <bbudge@chromium.org> Reviewed-by: Michael Lippautz <mlippautz@chromium.org> Cr-Commit-Position: refs/heads/master@{#49363}
2017-11-14 15:55:09 +00:00
masm.GetCode(isolate, &desc);
Reland "Reland "[deoptimizer] Change deopt entries into builtins"" This is a reland of fbfa9bf4ec72b1b73a96b70ccb68cd98c321511b The arm64 was missing proper codegen for CFI, thus sizes were off. Original change's description: > Reland "[deoptimizer] Change deopt entries into builtins" > > This is a reland of 7f58ced72eb65b6b5530ccabaf2eaebe45bf9d33 > > It fixes the different exit size emitted on x64/Atom CPUs due to > performance tuning in TurboAssembler::Call. Additionally, add > cctests to verify the fixed size exits. > > Original change's description: > > [deoptimizer] Change deopt entries into builtins > > > > While the overall goal of this commit is to change deoptimization > > entries into builtins, there are multiple related things happening: > > > > - Deoptimization entries, formerly stubs (i.e. Code objects generated > > at runtime, guaranteed to be immovable), have been converted into > > builtins. The major restriction is that we now need to preserve the > > kRootRegister, which was formerly used on most architectures to pass > > the deoptimization id. The solution differs based on platform. > > - Renamed DEOPT_ENTRIES_OR_FOR_TESTING code kind to FOR_TESTING. > > - Removed heap/ support for immovable Code generation. > > - Removed the DeserializerData class (no longer needed). > > - arm64: to preserve 4-byte deopt exits, introduced a new optimization > > in which the final jump to the deoptimization entry is generated > > once per Code object, and deopt exits can continue to emit a > > near-call. > > - arm,ia32,x64: change to fixed-size deopt exits. This reduces exit > > sizes by 4/8, 5, and 5 bytes, respectively. > > > > On arm the deopt exit size is reduced from 12 (or 16) bytes to 8 bytes > > by using the same strategy as on arm64 (recalc deopt id from return > > address). Before: > > > > e300a002 movw r10, <id> > > e59fc024 ldr ip, [pc, <entry offset>] > > e12fff3c blx ip > > > > After: > > > > e59acb35 ldr ip, [r10, <entry offset>] > > e12fff3c blx ip > > > > On arm64 the deopt exit size remains 4 bytes (or 8 bytes in same cases > > with CFI). Additionally, up to 4 builtin jumps are emitted per Code > > object (max 32 bytes added overhead per Code object). Before: > > > > 9401cdae bl <entry offset> > > > > After: > > > > # eager deoptimization entry jump. > > f95b1f50 ldr x16, [x26, <eager entry offset>] > > d61f0200 br x16 > > # lazy deoptimization entry jump. > > f95b2b50 ldr x16, [x26, <lazy entry offset>] > > d61f0200 br x16 > > # the deopt exit. > > 97fffffc bl <eager deoptimization entry jump offset> > > > > On ia32 the deopt exit size is reduced from 10 to 5 bytes. Before: > > > > bb00000000 mov ebx,<id> > > e825f5372b call <entry> > > > > After: > > > > e8ea2256ba call <entry> > > > > On x64 the deopt exit size is reduced from 12 to 7 bytes. Before: > > > > 49c7c511000000 REX.W movq r13,<id> > > e8ea2f0700 call <entry> > > > > After: > > > > 41ff9560360000 call [r13+<entry offset>] > > > > Bug: v8:8661,v8:8768 > > Change-Id: I13e30aedc360474dc818fecc528ce87c3bfeed42 > > Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2465834 > > Commit-Queue: Jakob Gruber <jgruber@chromium.org> > > Reviewed-by: Ross McIlroy <rmcilroy@chromium.org> > > Reviewed-by: Tobias Tebbi <tebbi@chromium.org> > > Reviewed-by: Ulan Degenbaev <ulan@chromium.org> > > Cr-Commit-Position: refs/heads/master@{#70597} > > Tbr: ulan@chromium.org, tebbi@chromium.org, rmcilroy@chromium.org > Bug: v8:8661,v8:8768,chromium:1140165 > Change-Id: Ibcd5c39c58a70bf2b2ac221aa375fc68d495e144 > Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2485506 > Reviewed-by: Jakob Gruber <jgruber@chromium.org> > Reviewed-by: Tobias Tebbi <tebbi@chromium.org> > Commit-Queue: Jakob Gruber <jgruber@chromium.org> > Cr-Commit-Position: refs/heads/master@{#70655} Tbr: ulan@chromium.org, tebbi@chromium.org, rmcilroy@chromium.org Bug: v8:8661 Bug: v8:8768 Bug: chromium:1140165 Change-Id: I471cc94fc085e527dc9bfb5a84b96bd907c2333f Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2488682 Reviewed-by: Jakob Gruber <jgruber@chromium.org> Commit-Queue: Jakob Gruber <jgruber@chromium.org> Cr-Commit-Position: refs/heads/master@{#70672}
2020-10-21 05:12:25 +00:00
Handle<Code> code =
Factory::CodeBuilder(isolate, desc, CodeKind::FOR_TESTING).Build();
#ifdef OBJECT_PRINT
code->Print(std::cout);
#endif
auto f = GeneratedCode<F1>::FromCode(*code);
for (int i = 0; i < kNumCases; ++i) {
int res = f.Call(i);
PrintF("f(%d) = %d\n", i, res);
CHECK_EQ(values[i], res);
}
}
TEST(AssemblerX64PslldWithXmm15) {
CcTest::InitializeVM();
auto buffer = AllocateAssemblerBuffer();
Assembler masm(AssemblerOptions{}, buffer->CreateView());
__ movq(xmm15, arg1);
__ pslld(xmm15, 1);
__ movq(rax, xmm15);
__ ret(0);
CodeDesc desc;
Reland "[Memory] Use OS::Allocate for all OS memory allocations." This is a reland of 4899bcb66db1c8c8272b02fb1cedf9886e010f82 This is a reland of b73ee3344ac96f92b4418c3c07779eb95a8541cc Original change's description: > [Memory] Use OS::Allocate for all OS memory allocations. > > - Eliminates OS::ReserveRegion and OS::ReserveAlignedRegion. > - Changes OS::Allocate to take alignment parameter, reorders parameters > to match page_allocator. > - Since the size of memory allocation can be deduced, don't return the > amount of memory allocated. > - Changes reservation of aligned address space. Before we would reserve > (size + alignment) rounded up to page size. This is too much, because > maximum misalignment is (alignment - page_size). > - On Windows and Cygwin, we release an oversize allocation and > immediately retry at the aligned address in the allocation. If we > lose the address due to a race, we just retry. > - Clean up all the calls to OS::Allocate in codegen and tests by adding > helper AllocateSystemPage function (allocation.h) and > AllocateAssemblerBuffer (cctest.h). > - Changes 'assm' to 'masm' in some targets for consistency when using > a macro-assembler. > > - Eliminates OS::ReleaseRegion, replacing with calls to OS::Free. > - Adds bool return value to OS::Free. > - Cleans up types of flags, protection on Windows and Cygwin. > Bug: chromium:756050 > Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng > Change-Id: I306dbe042cc867670fdc935abca29db074b0da71 Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng Change-Id: Iad3c025334e8f8d7d647be99a36a11ee449c9087 Reviewed-on: https://chromium-review.googlesource.com/767014 Commit-Queue: Bill Budge <bbudge@chromium.org> Reviewed-by: Michael Lippautz <mlippautz@chromium.org> Cr-Commit-Position: refs/heads/master@{#49363}
2017-11-14 15:55:09 +00:00
masm.GetCode(CcTest::i_isolate(), &desc);
buffer->MakeExecutable();
auto f = GeneratedCode<F5>::FromBuffer(CcTest::i_isolate(), buffer->start());
uint64_t result = f.Call(uint64_t{0x1122334455667788});
CHECK_EQ(uint64_t{0x22446688AACCEF10}, result);
}
using F9 = float(float x, float y);
TEST(AssemblerX64vmovups) {
CcTest::InitializeVM();
if (!CpuFeatures::IsSupported(AVX)) return;
Isolate* isolate = reinterpret_cast<Isolate*>(CcTest::isolate());
HandleScope scope(isolate);
v8::internal::byte buffer[256];
MacroAssembler masm(isolate, v8::internal::CodeObjectRequired::kYes,
ExternalAssemblerBuffer(buffer, sizeof(buffer)));
{
Reland "[Memory] Use OS::Allocate for all OS memory allocations." This is a reland of 4899bcb66db1c8c8272b02fb1cedf9886e010f82 This is a reland of b73ee3344ac96f92b4418c3c07779eb95a8541cc Original change's description: > [Memory] Use OS::Allocate for all OS memory allocations. > > - Eliminates OS::ReserveRegion and OS::ReserveAlignedRegion. > - Changes OS::Allocate to take alignment parameter, reorders parameters > to match page_allocator. > - Since the size of memory allocation can be deduced, don't return the > amount of memory allocated. > - Changes reservation of aligned address space. Before we would reserve > (size + alignment) rounded up to page size. This is too much, because > maximum misalignment is (alignment - page_size). > - On Windows and Cygwin, we release an oversize allocation and > immediately retry at the aligned address in the allocation. If we > lose the address due to a race, we just retry. > - Clean up all the calls to OS::Allocate in codegen and tests by adding > helper AllocateSystemPage function (allocation.h) and > AllocateAssemblerBuffer (cctest.h). > - Changes 'assm' to 'masm' in some targets for consistency when using > a macro-assembler. > > - Eliminates OS::ReleaseRegion, replacing with calls to OS::Free. > - Adds bool return value to OS::Free. > - Cleans up types of flags, protection on Windows and Cygwin. > Bug: chromium:756050 > Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng > Change-Id: I306dbe042cc867670fdc935abca29db074b0da71 Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng Change-Id: Iad3c025334e8f8d7d647be99a36a11ee449c9087 Reviewed-on: https://chromium-review.googlesource.com/767014 Commit-Queue: Bill Budge <bbudge@chromium.org> Reviewed-by: Michael Lippautz <mlippautz@chromium.org> Cr-Commit-Position: refs/heads/master@{#49363}
2017-11-14 15:55:09 +00:00
CpuFeatureScope avx_scope(&masm, AVX);
__ shufps(xmm0, xmm0, 0x0); // brocast first argument
__ shufps(xmm1, xmm1, 0x0); // brocast second argument
// copy xmm1 to xmm0 through the stack to test the "vmovups reg, mem".
__ AllocateStackSpace(kSimd128Size);
__ vmovups(Operand(rsp, 0), xmm1);
__ vmovups(xmm0, Operand(rsp, 0));
__ addq(rsp, Immediate(kSimd128Size));
__ ret(0);
}
CodeDesc desc;
Reland "[Memory] Use OS::Allocate for all OS memory allocations." This is a reland of 4899bcb66db1c8c8272b02fb1cedf9886e010f82 This is a reland of b73ee3344ac96f92b4418c3c07779eb95a8541cc Original change's description: > [Memory] Use OS::Allocate for all OS memory allocations. > > - Eliminates OS::ReserveRegion and OS::ReserveAlignedRegion. > - Changes OS::Allocate to take alignment parameter, reorders parameters > to match page_allocator. > - Since the size of memory allocation can be deduced, don't return the > amount of memory allocated. > - Changes reservation of aligned address space. Before we would reserve > (size + alignment) rounded up to page size. This is too much, because > maximum misalignment is (alignment - page_size). > - On Windows and Cygwin, we release an oversize allocation and > immediately retry at the aligned address in the allocation. If we > lose the address due to a race, we just retry. > - Clean up all the calls to OS::Allocate in codegen and tests by adding > helper AllocateSystemPage function (allocation.h) and > AllocateAssemblerBuffer (cctest.h). > - Changes 'assm' to 'masm' in some targets for consistency when using > a macro-assembler. > > - Eliminates OS::ReleaseRegion, replacing with calls to OS::Free. > - Adds bool return value to OS::Free. > - Cleans up types of flags, protection on Windows and Cygwin. > Bug: chromium:756050 > Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng > Change-Id: I306dbe042cc867670fdc935abca29db074b0da71 Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng Change-Id: Iad3c025334e8f8d7d647be99a36a11ee449c9087 Reviewed-on: https://chromium-review.googlesource.com/767014 Commit-Queue: Bill Budge <bbudge@chromium.org> Reviewed-by: Michael Lippautz <mlippautz@chromium.org> Cr-Commit-Position: refs/heads/master@{#49363}
2017-11-14 15:55:09 +00:00
masm.GetCode(isolate, &desc);
Reland "Reland "[deoptimizer] Change deopt entries into builtins"" This is a reland of fbfa9bf4ec72b1b73a96b70ccb68cd98c321511b The arm64 was missing proper codegen for CFI, thus sizes were off. Original change's description: > Reland "[deoptimizer] Change deopt entries into builtins" > > This is a reland of 7f58ced72eb65b6b5530ccabaf2eaebe45bf9d33 > > It fixes the different exit size emitted on x64/Atom CPUs due to > performance tuning in TurboAssembler::Call. Additionally, add > cctests to verify the fixed size exits. > > Original change's description: > > [deoptimizer] Change deopt entries into builtins > > > > While the overall goal of this commit is to change deoptimization > > entries into builtins, there are multiple related things happening: > > > > - Deoptimization entries, formerly stubs (i.e. Code objects generated > > at runtime, guaranteed to be immovable), have been converted into > > builtins. The major restriction is that we now need to preserve the > > kRootRegister, which was formerly used on most architectures to pass > > the deoptimization id. The solution differs based on platform. > > - Renamed DEOPT_ENTRIES_OR_FOR_TESTING code kind to FOR_TESTING. > > - Removed heap/ support for immovable Code generation. > > - Removed the DeserializerData class (no longer needed). > > - arm64: to preserve 4-byte deopt exits, introduced a new optimization > > in which the final jump to the deoptimization entry is generated > > once per Code object, and deopt exits can continue to emit a > > near-call. > > - arm,ia32,x64: change to fixed-size deopt exits. This reduces exit > > sizes by 4/8, 5, and 5 bytes, respectively. > > > > On arm the deopt exit size is reduced from 12 (or 16) bytes to 8 bytes > > by using the same strategy as on arm64 (recalc deopt id from return > > address). Before: > > > > e300a002 movw r10, <id> > > e59fc024 ldr ip, [pc, <entry offset>] > > e12fff3c blx ip > > > > After: > > > > e59acb35 ldr ip, [r10, <entry offset>] > > e12fff3c blx ip > > > > On arm64 the deopt exit size remains 4 bytes (or 8 bytes in same cases > > with CFI). Additionally, up to 4 builtin jumps are emitted per Code > > object (max 32 bytes added overhead per Code object). Before: > > > > 9401cdae bl <entry offset> > > > > After: > > > > # eager deoptimization entry jump. > > f95b1f50 ldr x16, [x26, <eager entry offset>] > > d61f0200 br x16 > > # lazy deoptimization entry jump. > > f95b2b50 ldr x16, [x26, <lazy entry offset>] > > d61f0200 br x16 > > # the deopt exit. > > 97fffffc bl <eager deoptimization entry jump offset> > > > > On ia32 the deopt exit size is reduced from 10 to 5 bytes. Before: > > > > bb00000000 mov ebx,<id> > > e825f5372b call <entry> > > > > After: > > > > e8ea2256ba call <entry> > > > > On x64 the deopt exit size is reduced from 12 to 7 bytes. Before: > > > > 49c7c511000000 REX.W movq r13,<id> > > e8ea2f0700 call <entry> > > > > After: > > > > 41ff9560360000 call [r13+<entry offset>] > > > > Bug: v8:8661,v8:8768 > > Change-Id: I13e30aedc360474dc818fecc528ce87c3bfeed42 > > Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2465834 > > Commit-Queue: Jakob Gruber <jgruber@chromium.org> > > Reviewed-by: Ross McIlroy <rmcilroy@chromium.org> > > Reviewed-by: Tobias Tebbi <tebbi@chromium.org> > > Reviewed-by: Ulan Degenbaev <ulan@chromium.org> > > Cr-Commit-Position: refs/heads/master@{#70597} > > Tbr: ulan@chromium.org, tebbi@chromium.org, rmcilroy@chromium.org > Bug: v8:8661,v8:8768,chromium:1140165 > Change-Id: Ibcd5c39c58a70bf2b2ac221aa375fc68d495e144 > Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2485506 > Reviewed-by: Jakob Gruber <jgruber@chromium.org> > Reviewed-by: Tobias Tebbi <tebbi@chromium.org> > Commit-Queue: Jakob Gruber <jgruber@chromium.org> > Cr-Commit-Position: refs/heads/master@{#70655} Tbr: ulan@chromium.org, tebbi@chromium.org, rmcilroy@chromium.org Bug: v8:8661 Bug: v8:8768 Bug: chromium:1140165 Change-Id: I471cc94fc085e527dc9bfb5a84b96bd907c2333f Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2488682 Reviewed-by: Jakob Gruber <jgruber@chromium.org> Commit-Queue: Jakob Gruber <jgruber@chromium.org> Cr-Commit-Position: refs/heads/master@{#70672}
2020-10-21 05:12:25 +00:00
Handle<Code> code =
Factory::CodeBuilder(isolate, desc, CodeKind::FOR_TESTING).Build();
#ifdef OBJECT_PRINT
StdoutStream os;
code->Print(os);
#endif
auto f = GeneratedCode<F9>::FromCode(*code);
CHECK_EQ(-1.5, f.Call(1.5, -1.5));
}
TEST(AssemblerX64Regmove256bit) {
if (!CpuFeatures::IsSupported(AVX)) return;
CcTest::InitializeVM();
v8::HandleScope scope(CcTest::isolate());
auto buffer = AllocateAssemblerBuffer();
Isolate* isolate = CcTest::i_isolate();
Assembler masm(AssemblerOptions{}, buffer->CreateView());
CpuFeatureScope fscope(&masm, AVX);
__ vmovdqa(ymm0, ymm1);
__ vmovdqa(ymm4, Operand(rbx, rcx, times_4, 10000));
__ vmovdqu(ymm10, ymm11);
__ vmovdqu(ymm9, Operand(rbx, rcx, times_4, 10000));
__ vmovdqu(Operand(rbx, rcx, times_4, 10000), ymm0);
__ vbroadcastss(ymm7, Operand(rbx, rcx, times_4, 10000));
__ vmovddup(ymm3, ymm2);
__ vmovddup(ymm4, Operand(rbx, rcx, times_4, 10000));
__ vmovshdup(ymm1, ymm2);
CodeDesc desc;
masm.GetCode(isolate, &desc);
#ifdef OBJECT_PRINT
Handle<Code> code =
Factory::CodeBuilder(isolate, desc, CodeKind::FOR_TESTING).Build();
StdoutStream os;
code->Print(os);
#endif
byte expected[] = {// VMOVDQA
// vmovdqa ymm0,ymm1
0xC5, 0xFD, 0x6F, 0xC1,
// vmovdqa ymm4,YMMWORD PTR [rbx+rcx*4+0x2710]
0xC5, 0xFD, 0x6F, 0xA4, 0x8B, 0x10, 0x27, 0x00, 0x00,
// VMOVDQU
// vmovdqu ymm10,ymm11
0xC4, 0x41, 0x7E, 0x7F, 0xDA,
// vmovdqu ymm9,YMMWORD PTR [rbx+rcx*4+0x2710]
0xC5, 0x7E, 0x6F, 0x8C, 0x8B, 0x10, 0x27, 0x00, 0x00,
// vmovdqu YMMWORD PTR [rbx+rcx*4+0x2710],ymm0
0xC5, 0xFE, 0x7F, 0x84, 0x8B, 0x10, 0x27, 0x00, 0x00,
// vbroadcastss ymm7, DWORD PTR [rbx+rcx*4+0x2710]
0xc4, 0xe2, 0x7d, 0x18, 0xbc, 0x8b, 0x10, 0x27, 0x00, 0x00,
// vmovddup ymm3, ymm2
0xc5, 0xff, 0x12, 0xda,
// vmovddup ymm4, YMMWORD PTR [rbx+rcx*4+0x2710]
0xc5, 0xff, 0x12, 0xa4, 0x8b, 0x10, 0x27, 0x00, 0x00,
// vmovshdup ymm1, ymm2
0xc5, 0xfe, 0x16, 0xca};
CHECK_EQ(0, memcmp(expected, desc.buffer, sizeof(expected)));
}
TEST(AssemblerX64AVX2Op256bit) {
if (!CpuFeatures::IsSupported(AVX2)) return;
CcTest::InitializeVM();
v8::HandleScope scope(CcTest::isolate());
auto buffer = AllocateAssemblerBuffer();
Isolate* isolate = CcTest::i_isolate();
Assembler masm(AssemblerOptions{}, buffer->CreateView());
CpuFeatureScope fscope(&masm, AVX2);
__ vpshufd(ymm1, ymm2, 85);
__ vpshufd(ymm1, Operand(rbx, rcx, times_4, 10000), 85);
__ vpshuflw(ymm9, ymm10, 85);
__ vpshuflw(ymm9, Operand(rbx, rcx, times_4, 10000), 85);
__ vpshufhw(ymm1, ymm2, 85);
__ vpshufhw(ymm1, Operand(rbx, rcx, times_4, 10000), 85);
__ vpblendw(ymm2, ymm3, ymm4, 23);
__ vpblendw(ymm2, ymm3, Operand(rbx, rcx, times_4, 10000), 23);
__ vpalignr(ymm10, ymm11, ymm12, 4);
__ vpalignr(ymm10, ymm11, Operand(rbx, rcx, times_4, 10000), 4);
__ vbroadcastss(ymm7, xmm0);
__ vpbroadcastb(ymm2, xmm1);
__ vpbroadcastb(ymm3, Operand(rbx, rcx, times_4, 10000));
__ vpbroadcastw(ymm15, xmm4);
__ vpbroadcastw(ymm5, Operand(rbx, rcx, times_4, 10000));
CodeDesc desc;
masm.GetCode(isolate, &desc);
#ifdef OBJECT_PRINT
Handle<Code> code =
Factory::CodeBuilder(isolate, desc, CodeKind::FOR_TESTING).Build();
StdoutStream os;
code->Print(os);
#endif
byte expected[] = {
// vpshufd ymm1, ymm2, 85
0xC5, 0xFD, 0x70, 0xCA, 0x55,
// vpshufd ymm1,YMMWORD PTR [rbx+rcx*4+0x2710], 85
0xC5, 0xFD, 0x70, 0x8C, 0x8B, 0x10, 0x27, 0x00, 0x00, 0x55,
// vpshuflw ymm9, ymm10, 85,
0xC4, 0x41, 0x7F, 0x70, 0xCA, 0x55,
// vpshuflw ymm9,YMMWORD PTR [rbx+rcx*4+0x2710], 85
0xC5, 0x7F, 0x70, 0x8C, 0x8B, 0x10, 0x27, 0x00, 0x00, 0x55,
// vpshufhw ymm1, ymm2, 85
0xC5, 0xFE, 0x70, 0xCA, 0x55,
// vpshufhw ymm1,YMMWORD PTR [rbx+rcx*4+0x2710], 85
0xC5, 0xFE, 0x70, 0x8C, 0x8B, 0x10, 0x27, 0x00, 0x00, 0x55,
// vpblendw ymm2, ymm3, ymm4, 23
0xC4, 0xE3, 0x65, 0x0E, 0xD4, 0x17,
// vpblendw ymm2, ymm3, YMMWORD PTR [rbx+rcx*4+0x2710], 23
0xC4, 0xE3, 0x65, 0x0E, 0x94, 0x8B, 0x10, 0x27, 0x00, 0x00, 0x17,
// vpalignr ymm10, ymm11, ymm12, 4
0xC4, 0x43, 0x25, 0x0F, 0xD4, 0x04,
// vpalignr ymm10, ymm11, YMMWORD PTR [rbx+rcx*4+0x2710], 4
0xC4, 0x63, 0x25, 0x0F, 0x94, 0x8B, 0x10, 0x27, 0x00, 0x00, 0x04,
// vbroadcastss ymm7, xmm0
0xc4, 0xe2, 0x7d, 0x18, 0xf8,
// vpbroadcastb ymm2, xmm1
0xc4, 0xe2, 0x7d, 0x78, 0xd1,
// vpbroadcastb ymm3, BYTE PTR [rbx+rcx*4+0x2710]
0xc4, 0xe2, 0x7d, 0x78, 0x9c, 0x8b, 0x10, 0x27, 0x00, 0x00,
// vpbroadcastw ymm15, xmm4
0xc4, 0x62, 0x7d, 0x79, 0xfc,
// vpbroadcastw ymm5, WORD PTR [rbx+rcx*4+0x2710]
0xc4, 0xe2, 0x7d, 0x79, 0xac, 0x8b, 0x10, 0x27, 0x00, 0x00};
CHECK_EQ(0, memcmp(expected, desc.buffer, sizeof(expected)));
}
TEST(AssemblerX64FloatingPoint256bit) {
if (!CpuFeatures::IsSupported(AVX)) return;
CcTest::InitializeVM();
v8::HandleScope scope(CcTest::isolate());
auto buffer = AllocateAssemblerBuffer();
Isolate* isolate = CcTest::i_isolate();
Assembler masm(AssemblerOptions{}, buffer->CreateView());
CpuFeatureScope fscope(&masm, AVX);
__ vandpd(ymm1, ymm3, ymm5);
__ vminpd(ymm2, ymm3, Operand(r8, r9, times_4, 10000));
__ vsqrtps(ymm0, ymm1);
__ vunpcklps(ymm2, ymm3, ymm14);
__ vsubps(ymm10, ymm11, ymm12);
__ vroundps(ymm9, ymm2, kRoundUp);
__ vroundpd(ymm9, ymm2, kRoundToNearest);
__ vhaddps(ymm1, ymm2, ymm3);
__ vhaddps(ymm0, ymm1, Operand(rbx, rcx, times_4, 10000));
CodeDesc desc;
masm.GetCode(isolate, &desc);
#ifdef OBJECT_PRINT
Handle<Code> code =
Factory::CodeBuilder(isolate, desc, CodeKind::FOR_TESTING).Build();
StdoutStream os;
code->Print(os);
#endif
byte expected[] = {// vandpd ymm1, ymm3, ymm5
0xC5, 0xE5, 0x54, 0xCD,
// vminpd ymm2, ymm3, YMMWORD PTR [r8+r9*4+0x2710]
0xC4, 0x81, 0x65, 0x5D, 0x94, 0x88, 0x10, 0x27, 0x00, 0x00,
// VSQRTPS
0xC5, 0xFC, 0x51, 0xC1,
// VUNPCKLPS
0xC4, 0xC1, 0x64, 0x14, 0xD6,
// VSUBPS
0xC4, 0x41, 0x24, 0x5C, 0xD4,
// vroundps ymm9, ymm2, 0xA
0xC4, 0x63, 0x7D, 0x08, 0xCA, 0x0A,
// vroundpd ymm9, ymm2, 0x8
0xC4, 0x63, 0x7D, 0x09, 0xCA, 0x08,
// VHADDPS ymm1, ymm2, ymm3
0xC5, 0xEF, 0x7C, 0xCB,
// VHADDPS ymm0, ymm1, YMMWORD PTR [rbx+rcx*4+0x2710]
0xc5, 0xf7, 0x7c, 0x84, 0x8b, 0x10, 0x27, 0x00, 0x00};
CHECK_EQ(0, memcmp(expected, desc.buffer, sizeof(expected)));
}
TEST(AssemblerX64Integer256bit) {
if (!CpuFeatures::IsSupported(AVX2)) return;
CcTest::InitializeVM();
v8::HandleScope scope(CcTest::isolate());
auto buffer = AllocateAssemblerBuffer();
Isolate* isolate = CcTest::i_isolate();
Assembler masm(AssemblerOptions{}, buffer->CreateView());
CpuFeatureScope fscope(&masm, AVX2);
// SSE2_AVX_INSTRUCTION
__ vpunpcklbw(ymm9, ymm2, ymm0);
__ vpacksswb(ymm8, ymm3, ymm1);
__ vpcmpgtw(ymm2, ymm7, ymm9);
__ vpand(ymm2, ymm3, ymm4);
__ vpmaxsw(ymm10, ymm11, Operand(rbx, rcx, times_4, 10000));
__ vpaddb(ymm1, ymm2, ymm3);
__ vpsraw(ymm7, ymm1, xmm4);
__ vpsllq(ymm3, ymm2, xmm1);
// SSSE3_AVX_INSTRUCTION
__ vpshufb(ymm1, ymm2, ymm3);
__ vphaddw(ymm8, ymm9, Operand(rbx, rcx, times_4, 10000));
__ vpmaddubsw(ymm5, ymm7, ymm9);
__ vpsignd(ymm7, ymm0, ymm1);
__ vpmulhrsw(ymm4, ymm3, ymm1);
// SSE4_AVX_INSTRUCTION
__ vpmuldq(ymm1, ymm5, ymm6);
__ vpcmpeqq(ymm0, ymm2, ymm3);
__ vpackusdw(ymm4, ymm2, ymm0);
__ vpminud(ymm8, ymm9, Operand(rbx, rcx, times_4, 10000));
__ vpmaxsb(ymm3, ymm4, ymm7);
__ vpmulld(ymm6, ymm5, ymm3);
// SSE4_2_AVX_INSTRUCTION
__ vpcmpgtq(ymm3, ymm2, ymm0);
CodeDesc desc;
masm.GetCode(isolate, &desc);
#ifdef OBJECT_PRINT
Handle<Code> code =
Factory::CodeBuilder(isolate, desc, CodeKind::FOR_TESTING).Build();
StdoutStream os;
code->Print(os);
#endif
byte expected[] = {// SSE2_AVX_INSTRUCTION
// vpunpcklbw ymm9, ymm2, ymm0
0xC5, 0x6D, 0x60, 0xC8,
// vpacksswb ymm8, ymm3, ymm1
0xC5, 0x65, 0x63, 0xC1,
// vpcmpgtw ymm2, ymm7, ymm9
0xC4, 0xC1, 0x45, 0x65, 0xD1,
// vpand ymm2, ymm3, ymm4
0xC5, 0xE5, 0xDB, 0xD4,
// vpmaxsw ymm10, ymm11, YMMWORD PTR [rbx+rcx*4+0x2710]
0xC5, 0x25, 0xEE, 0x94, 0x8B, 0x10, 0x27, 0x00, 0x00,
// vpaddb ymm1, ymm2, ymm3
0xC5, 0xED, 0xFC, 0xCB,
// vpsraw ymm7, ymm1, xmm4
0xC5, 0xF5, 0xE1, 0xFC,
// vpsllq ymm3, ymm2, xmm1
0xC5, 0xED, 0xF3, 0xD9,
// SSSE3_AVX_INSTRUCTION
// vpshufb ymm1, ymm2, ymm3
0xC4, 0xE2, 0x6D, 0x00, 0xCB,
// vphaddw ymm8, ymm9, YMMWORD PTR [rbx+rcx*4+0x2710]
0xC4, 0x62, 0x35, 0x01, 0x84, 0x8B, 0x10, 0x27, 0x00, 0x00,
// vpmaddubsw ymm5, ymm7, ymm9
0xC4, 0xC2, 0x45, 0x04, 0xE9,
// vpsignd ymm7, ymm0, ymm1
0xC4, 0xE2, 0x7D, 0x0A, 0xF9,
// vpmulhrsw ymm4, ymm3, ymm1
0xC4, 0xE2, 0x65, 0x0B, 0xE1,
// SSE4_AVX_INSTRUCTION
// vpmuldq ymm1, ymm5, ymm6
0xC4, 0xE2, 0x55, 0x28, 0xCE,
// vpcmpeqq ymm0, ymm2, ymm3
0xC4, 0xE2, 0x6D, 0x29, 0xC3,
// vpackusdw ymm4, ymm2, ymm0
0xC4, 0xE2, 0x6D, 0x2B, 0xE0,
// vpminud ymm8, ymm9, YMMWORD PTR [rbx+rcx*4+0x2710]
0xC4, 0x62, 0x35, 0x3B, 0x84, 0x8B, 0x10, 0x27, 0x0, 0x0,
// vpmaxsb ymm3, ymm4, ymm7
0xC4, 0xE2, 0x5D, 0x3C, 0xDF,
// vpmulld ymm6, ymm5, ymm3
0xC4, 0xE2, 0x55, 0x40, 0xF3,
// SSE4_2_AVX_INSTRUCTION
// vpcmpgtq ymm3, ymm2, ymm0
0xC4, 0xE2, 0x6D, 0x37, 0xD8};
CHECK_EQ(0, memcmp(expected, desc.buffer, sizeof(expected)));
}
TEST(CpuFeatures_ProbeImpl) {
// Support for a newer extension implies support for the older extensions.
CHECK_IMPLIES(CpuFeatures::IsSupported(FMA3), CpuFeatures::IsSupported(AVX));
CHECK_IMPLIES(CpuFeatures::IsSupported(AVX2), CpuFeatures::IsSupported(AVX));
CHECK_IMPLIES(CpuFeatures::IsSupported(AVX),
CpuFeatures::IsSupported(SSE4_2));
CHECK_IMPLIES(CpuFeatures::IsSupported(SSE4_2),
CpuFeatures::IsSupported(SSE4_1));
CHECK_IMPLIES(CpuFeatures::IsSupported(SSE4_1),
CpuFeatures::IsSupported(SSSE3));
CHECK_IMPLIES(CpuFeatures::IsSupported(SSSE3),
CpuFeatures::IsSupported(SSE3));
// Check the reverse, if an older extension is not supported, a newer
// extension cannot be supported.
CHECK_IMPLIES(!CpuFeatures::IsSupported(SSE3),
!CpuFeatures::IsSupported(SSSE3));
CHECK_IMPLIES(!CpuFeatures::IsSupported(SSSE3),
!CpuFeatures::IsSupported(SSE4_1));
CHECK_IMPLIES(!CpuFeatures::IsSupported(SSE4_1),
!CpuFeatures::IsSupported(SSE4_2));
CHECK_IMPLIES(!CpuFeatures::IsSupported(SSE4_2),
!CpuFeatures::IsSupported(AVX));
CHECK_IMPLIES(!CpuFeatures::IsSupported(AVX),
!CpuFeatures::IsSupported(AVX2));
CHECK_IMPLIES(!CpuFeatures::IsSupported(AVX),
!CpuFeatures::IsSupported(FMA3));
}
#undef __
} // namespace internal
} // namespace v8