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

1045 lines
28 KiB
C++
Raw Normal View History

// Copyright 2012 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 "src/init/v8.h"
#include "src/codegen/ppc/assembler-ppc-inl.h"
#include "src/diagnostics/disassembler.h"
#include "src/execution/simulator.h"
#include "src/heap/factory.h"
#include "test/cctest/cctest.h"
#include "test/common/assembler-tester.h"
namespace v8 {
namespace internal {
// TODO(ppc): Refine these signatures per test case, they can have arbitrary
// return and argument types and arbitrary number of arguments.
using F_iiiii = void*(int x, int p1, int p2, int p3, int p4);
using F_piiii = void*(void* p0, int p1, int p2, int p3, int p4);
using F_ppiii = void*(void* p0, void* p1, int p2, int p3, int p4);
using F_pppii = void*(void* p0, void* p1, void* p2, int p3, int p4);
using F_ippii = void*(int p0, void* p1, void* p2, int p3, int p4);
#define __ assm.
// Simple add parameter 1 to parameter 2 and return
TEST(0) {
CcTest::InitializeVM();
Isolate* isolate = CcTest::i_isolate();
HandleScope scope(isolate);
Assembler assm(AssemblerOptions{});
__ add(r3, r3, r4);
__ blr();
CodeDesc desc;
assm.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 DEBUG
code->Print();
#endif
auto f = GeneratedCode<F_iiiii>::FromCode(*code);
intptr_t res = reinterpret_cast<intptr_t>(f.Call(3, 4, 0, 0, 0));
::printf("f() = %" V8PRIdPTR "\n", res);
CHECK_EQ(7, static_cast<int>(res));
}
// Loop 100 times, adding loop counter to result
TEST(1) {
CcTest::InitializeVM();
Isolate* isolate = CcTest::i_isolate();
HandleScope scope(isolate);
Assembler assm(AssemblerOptions{});
Label L, C;
__ mr(r4, r3);
__ li(r3, Operand::Zero());
__ b(&C);
__ bind(&L);
__ add(r3, r3, r4);
__ subi(r4, r4, Operand(1));
__ bind(&C);
__ cmpi(r4, Operand::Zero());
__ bne(&L);
__ blr();
CodeDesc desc;
assm.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 DEBUG
code->Print();
#endif
auto f = GeneratedCode<F_iiiii>::FromCode(*code);
intptr_t res = reinterpret_cast<intptr_t>(f.Call(100, 0, 0, 0, 0));
::printf("f() = %" V8PRIdPTR "\n", res);
CHECK_EQ(5050, static_cast<int>(res));
}
TEST(2) {
CcTest::InitializeVM();
Isolate* isolate = CcTest::i_isolate();
HandleScope scope(isolate);
Assembler assm(AssemblerOptions{});
Label L, C;
__ mr(r4, r3);
__ li(r3, Operand(1));
__ b(&C);
__ bind(&L);
#if defined(V8_TARGET_ARCH_PPC64)
__ mulld(r3, r4, r3);
#else
__ mullw(r3, r4, r3);
#endif
__ subi(r4, r4, Operand(1));
__ bind(&C);
__ cmpi(r4, Operand::Zero());
__ bne(&L);
__ blr();
// some relocated stuff here, not executed
__ RecordComment("dead code, just testing relocations");
__ mov(r0, Operand(isolate->factory()->true_value()));
__ RecordComment("dead code, just testing immediate operands");
__ mov(r0, Operand(-1));
__ mov(r0, Operand(0xFF000000));
__ mov(r0, Operand(0xF0F0F0F0));
__ mov(r0, Operand(0xFFF0FFFF));
CodeDesc desc;
assm.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 DEBUG
code->Print();
#endif
auto f = GeneratedCode<F_iiiii>::FromCode(*code);
intptr_t res = reinterpret_cast<intptr_t>(f.Call(10, 0, 0, 0, 0));
::printf("f() = %" V8PRIdPTR "\n", res);
CHECK_EQ(3628800, static_cast<int>(res));
}
TEST(3) {
CcTest::InitializeVM();
Isolate* isolate = CcTest::i_isolate();
HandleScope scope(isolate);
struct T {
int i;
char c;
int16_t s;
};
T t;
Assembler assm(AssemblerOptions{});
// build a frame
#if V8_TARGET_ARCH_PPC64
__ stdu(sp, MemOperand(sp, -32));
__ std(fp, MemOperand(sp, 24));
#else
__ stwu(sp, MemOperand(sp, -16));
__ stw(fp, MemOperand(sp, 12));
#endif
__ mr(fp, sp);
// r4 points to our struct
__ mr(r4, r3);
// modify field int i of struct
__ lwz(r3, MemOperand(r4, offsetof(T, i)));
__ srwi(r5, r3, Operand(1));
__ stw(r5, MemOperand(r4, offsetof(T, i)));
// modify field char c of struct
__ lbz(r5, MemOperand(r4, offsetof(T, c)));
__ add(r3, r5, r3);
__ slwi(r5, r5, Operand(2));
__ stb(r5, MemOperand(r4, offsetof(T, c)));
// modify field int16_t s of struct
__ lhz(r5, MemOperand(r4, offsetof(T, s)));
__ add(r3, r5, r3);
__ srwi(r5, r5, Operand(3));
__ sth(r5, MemOperand(r4, offsetof(T, s)));
// restore frame
#if V8_TARGET_ARCH_PPC64
__ addi(r11, fp, Operand(32));
__ ld(fp, MemOperand(r11, -8));
#else
__ addi(r11, fp, Operand(16));
__ lwz(fp, MemOperand(r11, -4));
#endif
__ mr(sp, r11);
__ blr();
CodeDesc desc;
assm.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 DEBUG
code->Print();
#endif
auto f = GeneratedCode<F_piiii>::FromCode(*code);
t.i = 100000;
t.c = 10;
t.s = 1000;
intptr_t res = reinterpret_cast<intptr_t>(f.Call(&t, 0, 0, 0, 0));
::printf("f() = %" V8PRIdPTR "\n", res);
CHECK_EQ(101010, static_cast<int>(res));
CHECK_EQ(100000 / 2, t.i);
CHECK_EQ(10 * 4, t.c);
CHECK_EQ(1000 / 8, t.s);
}
#if 0
TEST(4) {
// Test the VFP floating point instructions.
CcTest::InitializeVM();
Isolate* isolate = CcTest::i_isolate();
HandleScope scope(isolate);
struct T {
double a;
double b;
double c;
double d;
double e;
double f;
double g;
double h;
int i;
double m;
double n;
float x;
float y;
};
T t;
// Create a function that accepts &t, and loads, manipulates, and stores
// the doubles and floats.
Assembler assm(AssemblerOptions{});
Label L, C;
if (CpuFeatures::IsSupported(VFP3)) {
CpuFeatures::Scope scope(VFP3);
__ mov(ip, Operand(sp));
__ stm(db_w, sp, r4.bit() | fp.bit() | lr.bit());
__ sub(fp, ip, Operand(4));
__ mov(r4, Operand(r0));
__ vldr(d6, r4, offsetof(T, a));
__ vldr(d7, r4, offsetof(T, b));
__ vadd(d5, d6, d7);
__ vstr(d5, r4, offsetof(T, c));
__ vmov(r2, r3, d5);
__ vmov(d4, r2, r3);
__ vstr(d4, r4, offsetof(T, b));
// Load t.x and t.y, switch values, and store back to the struct.
__ vldr(s0, r4, offsetof(T, x));
__ vldr(s31, r4, offsetof(T, y));
__ vmov(s16, s0);
__ vmov(s0, s31);
__ vmov(s31, s16);
__ vstr(s0, r4, offsetof(T, x));
__ vstr(s31, r4, offsetof(T, y));
// Move a literal into a register that can be encoded in the instruction.
__ vmov(d4, 1.0);
__ vstr(d4, r4, offsetof(T, e));
// Move a literal into a register that requires 64 bits to encode.
// 0x3FF0000010000000 = 1.000000059604644775390625
__ vmov(d4, 1.000000059604644775390625);
__ vstr(d4, r4, offsetof(T, d));
// Convert from floating point to integer.
__ vmov(d4, 2.0);
__ vcvt_s32_f64(s31, d4);
__ vstr(s31, r4, offsetof(T, i));
// Convert from integer to floating point.
__ mov(lr, Operand(42));
__ vmov(s31, lr);
__ vcvt_f64_s32(d4, s31);
__ vstr(d4, r4, offsetof(T, f));
// Test vabs.
__ vldr(d1, r4, offsetof(T, g));
__ vabs(d0, d1);
__ vstr(d0, r4, offsetof(T, g));
__ vldr(d2, r4, offsetof(T, h));
__ vabs(d0, d2);
__ vstr(d0, r4, offsetof(T, h));
// Test vneg.
__ vldr(d1, r4, offsetof(T, m));
__ vneg(d0, d1);
__ vstr(d0, r4, offsetof(T, m));
__ vldr(d1, r4, offsetof(T, n));
__ vneg(d0, d1);
__ vstr(d0, r4, offsetof(T, n));
__ ldm(ia_w, sp, r4.bit() | fp.bit() | pc.bit());
CodeDesc desc;
assm.GetCode(isolate, &desc);
Object code = isolate->heap()->CreateCode(
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
CodeKind::FOR_TESTING,
Handle<Code>())->ToObjectChecked();
CHECK(code->IsCode());
#ifdef DEBUG
Code::cast(code)->Print();
#endif
auto f = GeneratedCode<F_piiii>::FromCode(*code);
t.a = 1.5;
t.b = 2.75;
t.c = 17.17;
t.d = 0.0;
t.e = 0.0;
t.f = 0.0;
t.g = -2718.2818;
t.h = 31415926.5;
t.i = 0;
t.m = -2718.2818;
t.n = 123.456;
t.x = 4.5;
t.y = 9.0;
f.Call(&t, 0, 0, 0, 0);
CHECK_EQ(4.5, t.y);
CHECK_EQ(9.0, t.x);
CHECK_EQ(-123.456, t.n);
CHECK_EQ(2718.2818, t.m);
CHECK_EQ(2, t.i);
CHECK_EQ(2718.2818, t.g);
CHECK_EQ(31415926.5, t.h);
CHECK_EQ(42.0, t.f);
CHECK_EQ(1.0, t.e);
CHECK_EQ(1.000000059604644775390625, t.d);
CHECK_EQ(4.25, t.c);
CHECK_EQ(4.25, t.b);
CHECK_EQ(1.5, t.a);
}
}
TEST(5) {
// Test the ARMv7 bitfield instructions.
CcTest::InitializeVM();
Isolate* isolate = CcTest::i_isolate();
HandleScope scope(isolate);
Assembler assm(AssemblerOptions{});
if (CpuFeatures::IsSupported(ARMv7)) {
CpuFeatures::Scope scope(ARMv7);
// On entry, r0 = 0xAAAAAAAA = 0b10..10101010.
__ ubfx(r0, r0, 1, 12); // 0b00..010101010101 = 0x555
__ sbfx(r0, r0, 0, 5); // 0b11..111111110101 = -11
__ bfc(r0, 1, 3); // 0b11..111111110001 = -15
__ mov(r1, Operand(7));
__ bfi(r0, r1, 3, 3); // 0b11..111111111001 = -7
__ mov(pc, Operand(lr));
CodeDesc desc;
assm.GetCode(isolate, &desc);
Object code = isolate->heap()->CreateCode(
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
CodeKind::FOR_TESTING,
Handle<Code>())->ToObjectChecked();
CHECK(code->IsCode());
#ifdef DEBUG
Code::cast(code)->Print();
#endif
auto f = GeneratedCode<F_iiiii>::FromCode(*code);
int res = reinterpret_cast<int>(f.Call(0xAAAAAAAA, 0, 0, 0, 0));
::printf("f() = %d\n", res);
CHECK_EQ(-7, res);
}
}
TEST(6) {
// Test saturating instructions.
CcTest::InitializeVM();
Isolate* isolate = CcTest::i_isolate();
HandleScope scope(isolate);
Assembler assm(AssemblerOptions{});
if (CpuFeatures::IsSupported(ARMv7)) {
CpuFeatures::Scope scope(ARMv7);
__ usat(r1, 8, Operand(r0)); // Sat 0xFFFF to 0-255 = 0xFF.
__ usat(r2, 12, Operand(r0, ASR, 9)); // Sat (0xFFFF>>9) to 0-4095 = 0x7F.
__ usat(r3, 1, Operand(r0, LSL, 16)); // Sat (0xFFFF<<16) to 0-1 = 0x0.
__ addi(r0, r1, Operand(r2));
__ addi(r0, r0, Operand(r3));
__ mov(pc, Operand(lr));
CodeDesc desc;
assm.GetCode(isolate, &desc);
Object code = isolate->heap()->CreateCode(
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
CodeKind::FOR_TESTING,
Handle<Code>())->ToObjectChecked();
CHECK(code->IsCode());
#ifdef DEBUG
Code::cast(code)->Print();
#endif
auto f = GeneratedCode<F_iiiii>::FromCode(*code);
int res = reinterpret_cast<int>(f.Call(0xFFFF, 0, 0, 0, 0));
::printf("f() = %d\n", res);
CHECK_EQ(382, res);
}
}
enum VCVTTypes {
s32_f64,
u32_f64
};
static void TestRoundingMode(VCVTTypes types,
VFPRoundingMode mode,
double value,
int expected,
bool expected_exception = false) {
CcTest::InitializeVM();
Isolate* isolate = CcTest::i_isolate();
HandleScope scope(isolate);
Assembler assm(AssemblerOptions{});
if (CpuFeatures::IsSupported(VFP3)) {
CpuFeatures::Scope scope(VFP3);
Label wrong_exception;
__ vmrs(r1);
// Set custom FPSCR.
__ bic(r2, r1, Operand(kVFPRoundingModeMask | kVFPExceptionMask));
__ orr(r2, r2, Operand(mode));
__ vmsr(r2);
// Load value, convert, and move back result to r0 if everything went well.
__ vmov(d1, value);
switch (types) {
case s32_f64:
__ vcvt_s32_f64(s0, d1, kFPSCRRounding);
break;
case u32_f64:
__ vcvt_u32_f64(s0, d1, kFPSCRRounding);
break;
default:
UNREACHABLE();
break;
}
// Check for vfp exceptions
__ vmrs(r2);
__ tst(r2, Operand(kVFPExceptionMask));
// Check that we behaved as expected.
__ b(&wrong_exception,
expected_exception ? eq : ne);
// There was no exception. Retrieve the result and return.
__ vmov(r0, s0);
__ mov(pc, Operand(lr));
// The exception behaviour is not what we expected.
// Load a special value and return.
__ bind(&wrong_exception);
__ mov(r0, Operand(11223344));
__ mov(pc, Operand(lr));
CodeDesc desc;
assm.GetCode(isolate, &desc);
Object code = isolate->heap()->CreateCode(
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
CodeKind::FOR_TESTING,
Handle<Code>())->ToObjectChecked();
CHECK(code->IsCode());
#ifdef DEBUG
Code::cast(code)->Print();
#endif
auto f = GeneratedCode<F_iiiii>::FromCode(*code);
int res = reinterpret_cast<int>(f.Call(0, 0, 0, 0, 0));
::printf("res = %d\n", res);
CHECK_EQ(expected, res);
}
}
TEST(7) {
// Test vfp rounding modes.
// s32_f64 (double to integer).
TestRoundingMode(s32_f64, RN, 0, 0);
TestRoundingMode(s32_f64, RN, 0.5, 0);
TestRoundingMode(s32_f64, RN, -0.5, 0);
TestRoundingMode(s32_f64, RN, 1.5, 2);
TestRoundingMode(s32_f64, RN, -1.5, -2);
TestRoundingMode(s32_f64, RN, 123.7, 124);
TestRoundingMode(s32_f64, RN, -123.7, -124);
TestRoundingMode(s32_f64, RN, 123456.2, 123456);
TestRoundingMode(s32_f64, RN, -123456.2, -123456);
TestRoundingMode(s32_f64, RN, static_cast<double>(kMaxInt), kMaxInt);
TestRoundingMode(s32_f64, RN, (kMaxInt + 0.49), kMaxInt);
TestRoundingMode(s32_f64, RN, (kMaxInt + 1.0), kMaxInt, true);
TestRoundingMode(s32_f64, RN, (kMaxInt + 0.5), kMaxInt, true);
TestRoundingMode(s32_f64, RN, static_cast<double>(kMinInt), kMinInt);
TestRoundingMode(s32_f64, RN, (kMinInt - 0.5), kMinInt);
TestRoundingMode(s32_f64, RN, (kMinInt - 1.0), kMinInt, true);
TestRoundingMode(s32_f64, RN, (kMinInt - 0.51), kMinInt, true);
TestRoundingMode(s32_f64, RM, 0, 0);
TestRoundingMode(s32_f64, RM, 0.5, 0);
TestRoundingMode(s32_f64, RM, -0.5, -1);
TestRoundingMode(s32_f64, RM, 123.7, 123);
TestRoundingMode(s32_f64, RM, -123.7, -124);
TestRoundingMode(s32_f64, RM, 123456.2, 123456);
TestRoundingMode(s32_f64, RM, -123456.2, -123457);
TestRoundingMode(s32_f64, RM, static_cast<double>(kMaxInt), kMaxInt);
TestRoundingMode(s32_f64, RM, (kMaxInt + 0.5), kMaxInt);
TestRoundingMode(s32_f64, RM, (kMaxInt + 1.0), kMaxInt, true);
TestRoundingMode(s32_f64, RM, static_cast<double>(kMinInt), kMinInt);
TestRoundingMode(s32_f64, RM, (kMinInt - 0.5), kMinInt, true);
TestRoundingMode(s32_f64, RM, (kMinInt + 0.5), kMinInt);
TestRoundingMode(s32_f64, RZ, 0, 0);
TestRoundingMode(s32_f64, RZ, 0.5, 0);
TestRoundingMode(s32_f64, RZ, -0.5, 0);
TestRoundingMode(s32_f64, RZ, 123.7, 123);
TestRoundingMode(s32_f64, RZ, -123.7, -123);
TestRoundingMode(s32_f64, RZ, 123456.2, 123456);
TestRoundingMode(s32_f64, RZ, -123456.2, -123456);
TestRoundingMode(s32_f64, RZ, static_cast<double>(kMaxInt), kMaxInt);
TestRoundingMode(s32_f64, RZ, (kMaxInt + 0.5), kMaxInt);
TestRoundingMode(s32_f64, RZ, (kMaxInt + 1.0), kMaxInt, true);
TestRoundingMode(s32_f64, RZ, static_cast<double>(kMinInt), kMinInt);
TestRoundingMode(s32_f64, RZ, (kMinInt - 0.5), kMinInt);
TestRoundingMode(s32_f64, RZ, (kMinInt - 1.0), kMinInt, true);
// u32_f64 (double to integer).
// Negative values.
TestRoundingMode(u32_f64, RN, -0.5, 0);
TestRoundingMode(u32_f64, RN, -123456.7, 0, true);
TestRoundingMode(u32_f64, RN, static_cast<double>(kMinInt), 0, true);
TestRoundingMode(u32_f64, RN, kMinInt - 1.0, 0, true);
TestRoundingMode(u32_f64, RM, -0.5, 0, true);
TestRoundingMode(u32_f64, RM, -123456.7, 0, true);
TestRoundingMode(u32_f64, RM, static_cast<double>(kMinInt), 0, true);
TestRoundingMode(u32_f64, RM, kMinInt - 1.0, 0, true);
TestRoundingMode(u32_f64, RZ, -0.5, 0);
TestRoundingMode(u32_f64, RZ, -123456.7, 0, true);
TestRoundingMode(u32_f64, RZ, static_cast<double>(kMinInt), 0, true);
TestRoundingMode(u32_f64, RZ, kMinInt - 1.0, 0, true);
// Positive values.
// kMaxInt is the maximum *signed* integer: 0x7FFFFFFF.
static const uint32_t kMaxUInt = 0xFFFFFFFFu;
TestRoundingMode(u32_f64, RZ, 0, 0);
TestRoundingMode(u32_f64, RZ, 0.5, 0);
TestRoundingMode(u32_f64, RZ, 123.7, 123);
TestRoundingMode(u32_f64, RZ, 123456.2, 123456);
TestRoundingMode(u32_f64, RZ, static_cast<double>(kMaxInt), kMaxInt);
TestRoundingMode(u32_f64, RZ, (kMaxInt + 0.5), kMaxInt);
TestRoundingMode(u32_f64, RZ, (kMaxInt + 1.0),
static_cast<uint32_t>(kMaxInt) + 1);
TestRoundingMode(u32_f64, RZ, (kMaxUInt + 0.5), kMaxUInt);
TestRoundingMode(u32_f64, RZ, (kMaxUInt + 1.0), kMaxUInt, true);
TestRoundingMode(u32_f64, RM, 0, 0);
TestRoundingMode(u32_f64, RM, 0.5, 0);
TestRoundingMode(u32_f64, RM, 123.7, 123);
TestRoundingMode(u32_f64, RM, 123456.2, 123456);
TestRoundingMode(u32_f64, RM, static_cast<double>(kMaxInt), kMaxInt);
TestRoundingMode(u32_f64, RM, (kMaxInt + 0.5), kMaxInt);
TestRoundingMode(u32_f64, RM, (kMaxInt + 1.0),
static_cast<uint32_t>(kMaxInt) + 1);
TestRoundingMode(u32_f64, RM, (kMaxUInt + 0.5), kMaxUInt);
TestRoundingMode(u32_f64, RM, (kMaxUInt + 1.0), kMaxUInt, true);
TestRoundingMode(u32_f64, RN, 0, 0);
TestRoundingMode(u32_f64, RN, 0.5, 0);
TestRoundingMode(u32_f64, RN, 1.5, 2);
TestRoundingMode(u32_f64, RN, 123.7, 124);
TestRoundingMode(u32_f64, RN, 123456.2, 123456);
TestRoundingMode(u32_f64, RN, static_cast<double>(kMaxInt), kMaxInt);
TestRoundingMode(u32_f64, RN, (kMaxInt + 0.49), kMaxInt);
TestRoundingMode(u32_f64, RN, (kMaxInt + 0.5),
static_cast<uint32_t>(kMaxInt) + 1);
TestRoundingMode(u32_f64, RN, (kMaxUInt + 0.49), kMaxUInt);
TestRoundingMode(u32_f64, RN, (kMaxUInt + 0.5), kMaxUInt, true);
TestRoundingMode(u32_f64, RN, (kMaxUInt + 1.0), kMaxUInt, true);
}
TEST(8) {
// Test VFP multi load/store with ia_w.
CcTest::InitializeVM();
Isolate* isolate = CcTest::i_isolate();
HandleScope scope(isolate);
struct D {
double a;
double b;
double c;
double d;
double e;
double f;
double g;
double h;
};
D d;
struct F {
float a;
float b;
float c;
float d;
float e;
float f;
float g;
float h;
};
F f;
// Create a function that uses vldm/vstm to move some double and
// single precision values around in memory.
Assembler assm(AssemblerOptions{});
if (CpuFeatures::IsSupported(VFP2)) {
CpuFeatures::Scope scope(VFP2);
__ mov(ip, Operand(sp));
__ stm(db_w, sp, r4.bit() | fp.bit() | lr.bit());
__ sub(fp, ip, Operand(4));
__ addi(r4, r0, Operand(offsetof(D, a)));
__ vldm(ia_w, r4, d0, d3);
__ vldm(ia_w, r4, d4, d7);
__ addi(r4, r0, Operand(offsetof(D, a)));
__ vstm(ia_w, r4, d6, d7);
__ vstm(ia_w, r4, d0, d5);
__ addi(r4, r1, Operand(offsetof(F, a)));
__ vldm(ia_w, r4, s0, s3);
__ vldm(ia_w, r4, s4, s7);
__ addi(r4, r1, Operand(offsetof(F, a)));
__ vstm(ia_w, r4, s6, s7);
__ vstm(ia_w, r4, s0, s5);
__ ldm(ia_w, sp, r4.bit() | fp.bit() | pc.bit());
CodeDesc desc;
assm.GetCode(isolate, &desc);
Object code = isolate->heap()->CreateCode(
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
CodeKind::FOR_TESTING,
Handle<Code>())->ToObjectChecked();
CHECK(code->IsCode());
#ifdef DEBUG
Code::cast(code)->Print();
#endif
auto fn = GeneratedCode<F_ppiii>::FromCode(*code);
d.a = 1.1;
d.b = 2.2;
d.c = 3.3;
d.d = 4.4;
d.e = 5.5;
d.f = 6.6;
d.g = 7.7;
d.h = 8.8;
f.a = 1.0;
f.b = 2.0;
f.c = 3.0;
f.d = 4.0;
f.e = 5.0;
f.f = 6.0;
f.g = 7.0;
f.h = 8.0;
fn.Call(&d, &f, 0, 0, 0);
CHECK_EQ(7.7, d.a);
CHECK_EQ(8.8, d.b);
CHECK_EQ(1.1, d.c);
CHECK_EQ(2.2, d.d);
CHECK_EQ(3.3, d.e);
CHECK_EQ(4.4, d.f);
CHECK_EQ(5.5, d.g);
CHECK_EQ(6.6, d.h);
CHECK_EQ(7.0, f.a);
CHECK_EQ(8.0, f.b);
CHECK_EQ(1.0, f.c);
CHECK_EQ(2.0, f.d);
CHECK_EQ(3.0, f.e);
CHECK_EQ(4.0, f.f);
CHECK_EQ(5.0, f.g);
CHECK_EQ(6.0, f.h);
}
}
TEST(9) {
// Test VFP multi load/store with ia.
CcTest::InitializeVM();
Isolate* isolate = CcTest::i_isolate();
HandleScope scope(isolate);
struct D {
double a;
double b;
double c;
double d;
double e;
double f;
double g;
double h;
};
D d;
struct F {
float a;
float b;
float c;
float d;
float e;
float f;
float g;
float h;
};
F f;
// Create a function that uses vldm/vstm to move some double and
// single precision values around in memory.
Assembler assm(AssemblerOptions{});
if (CpuFeatures::IsSupported(VFP2)) {
CpuFeatures::Scope scope(VFP2);
__ mov(ip, Operand(sp));
__ stm(db_w, sp, r4.bit() | fp.bit() | lr.bit());
__ sub(fp, ip, Operand(4));
__ addi(r4, r0, Operand(offsetof(D, a)));
__ vldm(ia, r4, d0, d3);
__ addi(r4, r4, Operand(4 * 8));
__ vldm(ia, r4, d4, d7);
__ addi(r4, r0, Operand(offsetof(D, a)));
__ vstm(ia, r4, d6, d7);
__ addi(r4, r4, Operand(2 * 8));
__ vstm(ia, r4, d0, d5);
__ addi(r4, r1, Operand(offsetof(F, a)));
__ vldm(ia, r4, s0, s3);
__ addi(r4, r4, Operand(4 * 4));
__ vldm(ia, r4, s4, s7);
__ addi(r4, r1, Operand(offsetof(F, a)));
__ vstm(ia, r4, s6, s7);
__ addi(r4, r4, Operand(2 * 4));
__ vstm(ia, r4, s0, s5);
__ ldm(ia_w, sp, r4.bit() | fp.bit() | pc.bit());
CodeDesc desc;
assm.GetCode(isolate, &desc);
Object code = isolate->heap()->CreateCode(
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
CodeKind::FOR_TESTING,
Handle<Code>())->ToObjectChecked();
CHECK(code->IsCode());
#ifdef DEBUG
Code::cast(code)->Print();
#endif
auto fn = GeneratedCode<F_ppiii>::FromCode(*code);
d.a = 1.1;
d.b = 2.2;
d.c = 3.3;
d.d = 4.4;
d.e = 5.5;
d.f = 6.6;
d.g = 7.7;
d.h = 8.8;
f.a = 1.0;
f.b = 2.0;
f.c = 3.0;
f.d = 4.0;
f.e = 5.0;
f.f = 6.0;
f.g = 7.0;
f.h = 8.0;
fn.Call(&d, &f, 0, 0, 0);
CHECK_EQ(7.7, d.a);
CHECK_EQ(8.8, d.b);
CHECK_EQ(1.1, d.c);
CHECK_EQ(2.2, d.d);
CHECK_EQ(3.3, d.e);
CHECK_EQ(4.4, d.f);
CHECK_EQ(5.5, d.g);
CHECK_EQ(6.6, d.h);
CHECK_EQ(7.0, f.a);
CHECK_EQ(8.0, f.b);
CHECK_EQ(1.0, f.c);
CHECK_EQ(2.0, f.d);
CHECK_EQ(3.0, f.e);
CHECK_EQ(4.0, f.f);
CHECK_EQ(5.0, f.g);
CHECK_EQ(6.0, f.h);
}
}
TEST(10) {
// Test VFP multi load/store with db_w.
CcTest::InitializeVM();
Isolate* isolate = CcTest::i_isolate();
HandleScope scope(isolate);
struct D {
double a;
double b;
double c;
double d;
double e;
double f;
double g;
double h;
};
D d;
struct F {
float a;
float b;
float c;
float d;
float e;
float f;
float g;
float h;
};
F f;
// Create a function that uses vldm/vstm to move some double and
// single precision values around in memory.
Assembler assm(AssemblerOptions{});
if (CpuFeatures::IsSupported(VFP2)) {
CpuFeatures::Scope scope(VFP2);
__ mov(ip, Operand(sp));
__ stm(db_w, sp, r4.bit() | fp.bit() | lr.bit());
__ sub(fp, ip, Operand(4));
__ addi(r4, r0, Operand(offsetof(D, h) + 8));
__ vldm(db_w, r4, d4, d7);
__ vldm(db_w, r4, d0, d3);
__ addi(r4, r0, Operand(offsetof(D, h) + 8));
__ vstm(db_w, r4, d0, d5);
__ vstm(db_w, r4, d6, d7);
__ addi(r4, r1, Operand(offsetof(F, h) + 4));
__ vldm(db_w, r4, s4, s7);
__ vldm(db_w, r4, s0, s3);
__ addi(r4, r1, Operand(offsetof(F, h) + 4));
__ vstm(db_w, r4, s0, s5);
__ vstm(db_w, r4, s6, s7);
__ ldm(ia_w, sp, r4.bit() | fp.bit() | pc.bit());
CodeDesc desc;
assm.GetCode(isolate, &desc);
Object code = isolate->heap()->CreateCode(
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
CodeKind::FOR_TESTING,
Handle<Code>())->ToObjectChecked();
CHECK(code->IsCode());
#ifdef DEBUG
Code::cast(code)->Print();
#endif
auto fn = GeneratedCode<F_ppiii>::FromCode(*code);
d.a = 1.1;
d.b = 2.2;
d.c = 3.3;
d.d = 4.4;
d.e = 5.5;
d.f = 6.6;
d.g = 7.7;
d.h = 8.8;
f.a = 1.0;
f.b = 2.0;
f.c = 3.0;
f.d = 4.0;
f.e = 5.0;
f.f = 6.0;
f.g = 7.0;
f.h = 8.0;
fn.Call(&d, &f, 0, 0, 0);
CHECK_EQ(7.7, d.a);
CHECK_EQ(8.8, d.b);
CHECK_EQ(1.1, d.c);
CHECK_EQ(2.2, d.d);
CHECK_EQ(3.3, d.e);
CHECK_EQ(4.4, d.f);
CHECK_EQ(5.5, d.g);
CHECK_EQ(6.6, d.h);
CHECK_EQ(7.0, f.a);
CHECK_EQ(8.0, f.b);
CHECK_EQ(1.0, f.c);
CHECK_EQ(2.0, f.d);
CHECK_EQ(3.0, f.e);
CHECK_EQ(4.0, f.f);
CHECK_EQ(5.0, f.g);
CHECK_EQ(6.0, f.h);
}
}
TEST(11) {
// Test instructions using the carry flag.
CcTest::InitializeVM();
Isolate* isolate = CcTest::i_isolate();
HandleScope scope(isolate);
struct I {
int32_t a;
int32_t b;
int32_t c;
int32_t d;
};
I i;
i.a = 0xABCD0001;
i.b = 0xABCD0000;
Assembler assm(AssemblerOptions{});
// Test HeapObject untagging.
__ ldr(r1, MemOperand(r0, offsetof(I, a)));
__ mov(r1, Operand(r1, ASR, 1), SetCC);
__ adc(r1, r1, Operand(r1), LeaveCC, cs);
__ str(r1, MemOperand(r0, offsetof(I, a)));
__ ldr(r2, MemOperand(r0, offsetof(I, b)));
__ mov(r2, Operand(r2, ASR, 1), SetCC);
__ adc(r2, r2, Operand(r2), LeaveCC, cs);
__ str(r2, MemOperand(r0, offsetof(I, b)));
// Test corner cases.
__ mov(r1, Operand(0xFFFFFFFF));
__ mov(r2, Operand::Zero());
__ mov(r3, Operand(r1, ASR, 1), SetCC); // Set the carry.
__ adc(r3, r1, Operand(r2));
__ str(r3, MemOperand(r0, offsetof(I, c)));
__ mov(r1, Operand(0xFFFFFFFF));
__ mov(r2, Operand::Zero());
__ mov(r3, Operand(r2, ASR, 1), SetCC); // Unset the carry.
__ adc(r3, r1, Operand(r2));
__ str(r3, MemOperand(r0, offsetof(I, d)));
__ mov(pc, Operand(lr));
CodeDesc desc;
assm.GetCode(isolate, &desc);
Object code = isolate->heap()->CreateCode(
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
CodeKind::FOR_TESTING,
Handle<Code>())->ToObjectChecked();
CHECK(code->IsCode());
#ifdef DEBUG
Code::cast(code)->Print();
#endif
auto f = GeneratedCode<F_piiii>::FromCode(*code);
f.Call(&i, 0, 0, 0, 0);
CHECK_EQ(0xABCD0001, i.a);
CHECK_EQ(static_cast<int32_t>(0xABCD0000) >> 1, i.b);
CHECK_EQ(0x00000000, i.c);
CHECK_EQ(0xFFFFFFFF, i.d);
}
TEST(12) {
// Test chaining of label usages within instructions (issue 1644).
CcTest::InitializeVM();
Isolate* isolate = CcTest::i_isolate();
HandleScope scope(isolate);
Assembler assm(AssemblerOptions{});
Label target;
__ b(eq, &target);
__ b(ne, &target);
__ bind(&target);
__ nop();
}
#endif
#undef __
} // namespace internal
} // namespace v8