Files
kobalicek b56f4176cb Codebase update and improvements, instruction DB update
* Denested src folder to root, renamed testing to asmjit-testing

  * Refactored how headers are included into <asmjit/...> form. This
    is necessary as compilers would never simplify a path once a ..
    appears in include directory - then paths such as ../core/../core
    appeared in asserts, which was ugly

  * Moved support utilities into asmjit/support/... (still included
    by asmjit/core.h for convenience and compatibility)

  * Added CMakePresets.json for making it easy to develop AsmJit

  * Reworked CMakeLists to be shorter and use CMake option(),
    etc... This simplifies it and makes it using more standard
    features

  * ASMJIT_EMBED now creates asmjit_embed INTERFACE library,
    which is accessible via asmjit::asmjit target - this simplifies
    embedding and makes it the same as library targets from a CMake
    perspective

  * Removed ASMJIT_DEPS - this is now provided by cmake target
    aliases - 'asmjit::asmjit' so users should not need this variable

  * Changed meaning of ASMJIT_LIBS - this now contains only AsmJit
    dependencies without asmjit::asmjit target alias. Don't rely on
    ASMJIT_LIBS anymore as it's only used internally

  * Removed ASMJIT_NO_DEPRECATED option - AsmJit is not going
    to provide controllable deprecations in the future

  * Removed ASMJIT_NO_VALIDATION in favor of ASMJIT_NO_INTROSPECTION,
    which now controls query, features, and validation API presence

  * Removed ASMJIT_DIR option - it was never really needed

  * Removed AMX_TRANSPOSE feature from instruction database (X86).
    Intel has removed it as well, so it's a feature that won't
    be siliconized
2025-11-02 22:31:46 +01:00

1939 lines
73 KiB
C++

// This file is part of AsmJit project <https://asmjit.com>
//
// See <asmjit/core.h> or LICENSE.md for license and copyright information
// SPDX-License-Identifier: Zlib
#include <asmjit/core/api-build_p.h>
#if !defined(ASMJIT_NO_X86)
#include <asmjit/core/cpuinfo.h>
#include <asmjit/core/instdb_p.h>
#include <asmjit/core/misc_p.h>
#include <asmjit/x86/x86instapi_p.h>
#include <asmjit/x86/x86instdb_p.h>
#include <asmjit/x86/x86opcode_p.h>
#include <asmjit/x86/x86operand.h>
ASMJIT_BEGIN_SUB_NAMESPACE(x86)
namespace InstInternal {
// x86::InstInternal - Text
// ========================
#ifndef ASMJIT_NO_TEXT
Error inst_id_to_string(InstId inst_id, InstStringifyOptions options, String& output) noexcept {
if (ASMJIT_UNLIKELY(!Inst::is_defined_id(inst_id)))
return make_error(Error::kInvalidInstruction);
return InstNameUtils::decode(InstDB::_inst_name_index_table[inst_id], options, InstDB::_inst_name_string_table, output);
}
InstId string_to_inst_id(const char* s, size_t len) noexcept {
if (ASMJIT_UNLIKELY(!s)) {
return BaseInst::kIdNone;
}
if (len == SIZE_MAX) {
len = strlen(s);
}
if (len == 0u || len > InstDB::_inst_name_index.max_name_length) {
return BaseInst::kIdNone;
}
InstId inst_id = InstNameUtils::find_instruction(s, len, InstDB::_inst_name_index_table, InstDB::_inst_name_string_table, InstDB::_inst_name_index);
if (inst_id != BaseInst::kIdNone) {
return inst_id;
}
uint32_t alias_index = InstNameUtils::find_alias(s, len, InstDB::alias_name_index_table, InstDB::alias_name_string_table, InstDB::kAliasTableSize);
if (alias_index != Globals::kInvalidId) {
return InstDB::alias_index_to_inst_id_table[alias_index];
}
return BaseInst::kIdNone;
}
#endif // !ASMJIT_NO_TEXT
// x86::InstInternal - Validate
// ============================
#ifndef ASMJIT_NO_INTROSPECTION
struct X86ValidationData {
//! Allowed registers by \ref RegType.
RegMask allowed_reg_mask[uint32_t(RegType::kMaxValue) + 1];
uint32_t allowed_mem_base_regs;
uint32_t allowed_mem_index_regs;
};
#define VALUE(x) \
(x == uint32_t(RegType::kPC )) ? InstDB::OpFlags::kNone : \
(x == uint32_t(RegType::kGp8Lo )) ? InstDB::OpFlags::kRegGpbLo : \
(x == uint32_t(RegType::kGp8Hi )) ? InstDB::OpFlags::kRegGpbHi : \
(x == uint32_t(RegType::kGp16 )) ? InstDB::OpFlags::kRegGpw : \
(x == uint32_t(RegType::kGp32 )) ? InstDB::OpFlags::kRegGpd : \
(x == uint32_t(RegType::kGp64 )) ? InstDB::OpFlags::kRegGpq : \
(x == uint32_t(RegType::kVec128 )) ? InstDB::OpFlags::kRegXmm : \
(x == uint32_t(RegType::kVec256 )) ? InstDB::OpFlags::kRegYmm : \
(x == uint32_t(RegType::kVec512 )) ? InstDB::OpFlags::kRegZmm : \
(x == uint32_t(RegType::kMask )) ? InstDB::OpFlags::kRegKReg : \
(x == uint32_t(RegType::kX86_Mm )) ? InstDB::OpFlags::kRegMm : \
(x == uint32_t(RegType::kSegment )) ? InstDB::OpFlags::kRegSReg : \
(x == uint32_t(RegType::kControl )) ? InstDB::OpFlags::kRegCReg : \
(x == uint32_t(RegType::kDebug )) ? InstDB::OpFlags::kRegDReg : \
(x == uint32_t(RegType::kX86_St )) ? InstDB::OpFlags::kRegSt : \
(x == uint32_t(RegType::kX86_Bnd )) ? InstDB::OpFlags::kRegBnd : \
(x == uint32_t(RegType::kTile )) ? InstDB::OpFlags::kRegTmm : InstDB::OpFlags::kNone
static const InstDB::OpFlags op_flag_from_reg_type_table[uint32_t(RegType::kMaxValue) + 1] = { ASMJIT_LOOKUP_TABLE_32(VALUE, 0) };
#undef VALUE
#define REG_MASK_FROM_REG_TYPE_X86(x) \
(x == uint32_t(RegType::kPC )) ? 0x00000001u : \
(x == uint32_t(RegType::kGp8Lo )) ? 0x0000000Fu : \
(x == uint32_t(RegType::kGp8Hi )) ? 0x0000000Fu : \
(x == uint32_t(RegType::kGp16 )) ? 0x000000FFu : \
(x == uint32_t(RegType::kGp32 )) ? 0x000000FFu : \
(x == uint32_t(RegType::kGp64 )) ? 0x000000FFu : \
(x == uint32_t(RegType::kVec128 )) ? 0x000000FFu : \
(x == uint32_t(RegType::kVec256 )) ? 0x000000FFu : \
(x == uint32_t(RegType::kVec512 )) ? 0x000000FFu : \
(x == uint32_t(RegType::kMask )) ? 0x000000FFu : \
(x == uint32_t(RegType::kX86_Mm )) ? 0x000000FFu : \
(x == uint32_t(RegType::kSegment )) ? 0x0000007Eu : \
(x == uint32_t(RegType::kControl )) ? 0x0000FFFFu : \
(x == uint32_t(RegType::kDebug )) ? 0x000000FFu : \
(x == uint32_t(RegType::kX86_St )) ? 0x000000FFu : \
(x == uint32_t(RegType::kX86_Bnd )) ? 0x0000000Fu : \
(x == uint32_t(RegType::kTile )) ? 0x000000FFu : 0u
#define REG_MASK_FROM_REG_TYPE_X64(x) \
(x == uint32_t(RegType::kPC )) ? 0x00000001u : \
(x == uint32_t(RegType::kGp8Lo )) ? 0x0000FFFFu : \
(x == uint32_t(RegType::kGp8Hi )) ? 0x0000000Fu : \
(x == uint32_t(RegType::kGp16 )) ? 0x0000FFFFu : \
(x == uint32_t(RegType::kGp32 )) ? 0x0000FFFFu : \
(x == uint32_t(RegType::kGp64 )) ? 0x0000FFFFu : \
(x == uint32_t(RegType::kVec128 )) ? 0xFFFFFFFFu : \
(x == uint32_t(RegType::kVec256 )) ? 0xFFFFFFFFu : \
(x == uint32_t(RegType::kVec512 )) ? 0xFFFFFFFFu : \
(x == uint32_t(RegType::kMask )) ? 0x000000FFu : \
(x == uint32_t(RegType::kX86_Mm )) ? 0x000000FFu : \
(x == uint32_t(RegType::kSegment )) ? 0x0000007Eu : \
(x == uint32_t(RegType::kControl )) ? 0x0000FFFFu : \
(x == uint32_t(RegType::kDebug )) ? 0x0000FFFFu : \
(x == uint32_t(RegType::kX86_St )) ? 0x000000FFu : \
(x == uint32_t(RegType::kX86_Bnd )) ? 0x0000000Fu : \
(x == uint32_t(RegType::kTile )) ? 0x000000FFu : 0u
#define B(RegType) (uint32_t(1) << uint32_t(RegType))
static const X86ValidationData x86_validation_data = {
{ ASMJIT_LOOKUP_TABLE_32(REG_MASK_FROM_REG_TYPE_X86, 0) },
B(RegType::kGp16) | B(RegType::kGp32) | B(RegType::kPC) | B(RegType::kLabelTag),
B(RegType::kGp16) | B(RegType::kGp32) | B(RegType::kVec128) | B(RegType::kVec256) | B(RegType::kVec512)
};
static const X86ValidationData x64_validation_data = {
{ ASMJIT_LOOKUP_TABLE_32(REG_MASK_FROM_REG_TYPE_X64, 0) },
B(RegType::kGp32) | B(RegType::kGp64) | B(RegType::kPC) | B(RegType::kLabelTag),
B(RegType::kGp32) | B(RegType::kGp64) | B(RegType::kVec128) | B(RegType::kVec256) | B(RegType::kVec512)
};
#undef B
#undef REG_MASK_FROM_REG_TYPE_X64
#undef REG_MASK_FROM_REG_TYPE_X86
static ASMJIT_INLINE bool is_zmm_or_m512(const Operand_& op) noexcept {
return op.is_vec512() || (op.is_mem() && op.as<Mem>().size() == 64);
}
static ASMJIT_INLINE bool check_op_sig(const InstDB::OpSignature& op, const InstDB::OpSignature& ref, bool& imm_out_of_range) noexcept {
// Fail if operand types are incompatible.
InstDB::OpFlags common_flags = op.flags() & ref.flags();
if (!Support::test(common_flags, InstDB::OpFlags::kOpMask)) {
// Mark temporarily `imm_out_of_range` so we can return a more descriptive error later.
if (op.has_imm() && ref.has_imm()) {
imm_out_of_range = true;
return true;
}
return false;
}
// Fail if some memory specific flags do not match.
if (Support::test(common_flags, InstDB::OpFlags::kMemMask)) {
if (ref.has_flag(InstDB::OpFlags::kFlagMemBase) && !op.has_flag(InstDB::OpFlags::kFlagMemBase)) {
return false;
}
}
// Fail if register indexes do not match.
if (Support::test(common_flags, InstDB::OpFlags::kRegMask)) {
if (ref.reg_mask() && !Support::test(op.reg_mask(), ref.reg_mask())) {
return false;
}
}
return true;
}
static ASMJIT_FAVOR_SIZE Error validate(InstDB::Mode mode, const BaseInst& inst, const Operand_* operands, size_t op_count, ValidationFlags validation_flags) noexcept {
uint32_t i;
// Get the instruction data.
const X86ValidationData* vd = (mode == InstDB::Mode::kX86) ? &x86_validation_data : &x64_validation_data;
InstId inst_id = inst.inst_id();
InstOptions options = inst.options();
if (ASMJIT_UNLIKELY(!Inst::is_defined_id(inst_id))) {
return make_error(Error::kInvalidInstruction);
}
const InstDB::InstInfo& inst_info = InstDB::inst_info_by_id(inst_id);
const InstDB::CommonInfo& common_info = inst_info.common_info();
InstDB::InstFlags inst_flags = inst_info.flags();
constexpr InstOptions kRepAny = InstOptions::kX86_Rep | InstOptions::kX86_Repne;
constexpr InstOptions kXAcqXRel = InstOptions::kX86_XAcquire | InstOptions::kX86_XRelease;
constexpr InstOptions kAvx512Options = InstOptions::kX86_ZMask | InstOptions::kX86_ER | InstOptions::kX86_SAE;
// Validate LOCK|XACQUIRE|XRELEASE Prefixes
// ----------------------------------------
if (Support::test(options, InstOptions::kX86_Lock | kXAcqXRel)) {
if (Support::test(options, InstOptions::kX86_Lock)) {
if (ASMJIT_UNLIKELY(!Support::test(inst_flags, InstDB::InstFlags::kLock) && !Support::test(options, kXAcqXRel))) {
return make_error(Error::kInvalidLockPrefix);
}
if (ASMJIT_UNLIKELY(op_count < 1 || !operands[0].is_mem())) {
return make_error(Error::kInvalidLockPrefix);
}
}
if (Support::test(options, kXAcqXRel)) {
if (ASMJIT_UNLIKELY(!Support::test(options, InstOptions::kX86_Lock) || (options & kXAcqXRel) == kXAcqXRel)) {
return make_error(Error::kInvalidPrefixCombination);
}
if (ASMJIT_UNLIKELY(Support::test(options, InstOptions::kX86_XAcquire) && !Support::test(inst_flags, InstDB::InstFlags::kXAcquire))) {
return make_error(Error::kInvalidXAcquirePrefix);
}
if (ASMJIT_UNLIKELY(Support::test(options, InstOptions::kX86_XRelease) && !Support::test(inst_flags, InstDB::InstFlags::kXRelease))) {
return make_error(Error::kInvalidXReleasePrefix);
}
}
}
// Validate REP and REPNE Prefixes
// -------------------------------
if (Support::test(options, kRepAny)) {
if (ASMJIT_UNLIKELY((options & kRepAny) == kRepAny)) {
return make_error(Error::kInvalidPrefixCombination);
}
if (ASMJIT_UNLIKELY(!Support::test(inst_flags, InstDB::InstFlags::kRep))) {
return make_error(Error::kInvalidRepPrefix);
}
}
// Translate Each Operand to the Corresponding OpSignature
// -------------------------------------------------------
InstDB::OpSignature op_sig_translated[Globals::kMaxOpCount];
InstDB::OpFlags combined_op_flags = InstDB::OpFlags::kNone;
RegMask combined_reg_mask = 0;
const Mem* mem_op = nullptr;
for (i = 0; i < op_count; i++) {
const Operand_& op = operands[i];
if (op.op_type() == OperandType::kNone) {
break;
}
InstDB::OpFlags op_flags = InstDB::OpFlags::kNone;
RegMask reg_mask = 0;
switch (op.op_type()) {
case OperandType::kReg: {
RegType reg_type = op.as<Reg>().reg_type();
op_flags = op_flag_from_reg_type_table[size_t(reg_type)];
if (ASMJIT_UNLIKELY(op_flags == InstDB::OpFlags::kNone)) {
return make_error(Error::kInvalidRegType);
}
// If `reg_id` is equal or greater than Operand::kVirtIdMin it means that the register is virtual and its
// index will be assigned later by the register allocator. We must pass unless asked to disallow virtual
// registers.
uint32_t reg_id = op.id();
if (reg_id < Operand::kVirtIdMin) {
if (ASMJIT_UNLIKELY(reg_id >= 32)) {
return make_error(Error::kInvalidPhysId);
}
if (ASMJIT_UNLIKELY(Support::bit_test(vd->allowed_reg_mask[size_t(reg_type)], reg_id) == 0)) {
return make_error(Error::kInvalidPhysId);
}
reg_mask = Support::bit_mask<RegMask>(reg_id);
combined_reg_mask |= reg_mask;
}
else {
if (uint32_t(validation_flags & ValidationFlags::kEnableVirtRegs) == 0) {
return make_error(Error::kIllegalVirtReg);
}
reg_mask = 0xFFFFFFFFu;
}
break;
}
// TODO: Validate base and index and combine these with `combined_reg_mask`.
case OperandType::kMem: {
const Mem& m = op.as<Mem>();
mem_op = &m;
uint32_t mem_size = m.size();
RegType base_type = m.base_type();
RegType index_type = m.index_type();
if (m.segment_id() > 6) {
return make_error(Error::kInvalidSegment);
}
// Validate AVX-512 broadcast {1tox}.
if (m.has_broadcast()) {
if (mem_size != 0) {
// If the size is specified it has to match the broadcast size.
if (ASMJIT_UNLIKELY(common_info.has_avx512_bcst32() && mem_size != 4)) {
return make_error(Error::kInvalidBroadcast);
}
if (ASMJIT_UNLIKELY(common_info.has_avx512_bcst64() && mem_size != 8)) {
return make_error(Error::kInvalidBroadcast);
}
}
else {
// If there is no size we implicitly calculate it so we can validate N in {1toN} properly.
mem_size = common_info.has_avx512_bcst64() ? 8 :
common_info.has_avx512_bcst32() ? 4 : 2;
}
mem_size <<= uint32_t(m.get_broadcast());
}
if (base_type != RegType::kNone && base_type > RegType::kLabelTag) {
uint32_t base_id = m.base_id();
if (m.is_reg_home()) {
// Home address of a virtual register. In such case we don't want to validate the type of the
// base register as it will always be patched to ESP|RSP.
}
else if (ASMJIT_UNLIKELY(!Support::bit_test(vd->allowed_mem_base_regs, base_type))) {
return make_error(Error::kInvalidAddress);
}
// Create information that will be validated only if this is an implicit memory operand. Basically
// only usable for string instructions and other instructions where memory operand is implicit and
// has 'seg:[reg]' form.
if (base_id < Operand::kVirtIdMin) {
if (ASMJIT_UNLIKELY(base_id >= 32)) {
return make_error(Error::kInvalidPhysId);
}
// Physical base id.
reg_mask = Support::bit_mask<RegMask>(base_id);
combined_reg_mask |= reg_mask;
}
else {
// Virtual base id - fill the whole mask for implicit mem validation. The register is not assigned
// yet, so we cannot predict the phys id.
if (uint32_t(validation_flags & ValidationFlags::kEnableVirtRegs) == 0) {
return make_error(Error::kIllegalVirtReg);
}
reg_mask = 0xFFFFFFFFu;
}
if (index_type == RegType::kNone && !m.offset_lo32()) {
op_flags |= InstDB::OpFlags::kFlagMemBase;
}
}
else if (base_type == RegType::kLabelTag) {
// [Label] - there is no need to validate the base as it's label.
}
else {
// Base is a 64-bit address.
int64_t offset = m.offset();
if (!Support::is_int_n<32>(offset)) {
if (mode == InstDB::Mode::kX86) {
// 32-bit mode: Make sure that the address is either `int32_t` or `uint32_t`.
if (!Support::is_uint_n<32>(offset)) {
return make_error(Error::kInvalidAddress64Bit);
}
}
else {
// 64-bit mode: Zero extension is allowed if the address has 32-bit index register or the address
// has no index register (it's still encodable).
if (index_type != RegType::kNone) {
if (!Support::is_uint_n<32>(offset)) {
return make_error(Error::kInvalidAddress64Bit);
}
if (index_type != RegType::kGp32) {
return make_error(Error::kInvalidAddress64BitZeroExtension);
}
}
else {
// We don't validate absolute 64-bit addresses without an index register as this also depends
// on the target's base address. We don't have the information to do it at this moment.
}
}
}
}
if (index_type != RegType::kNone) {
if (ASMJIT_UNLIKELY(!Support::bit_test(vd->allowed_mem_index_regs, index_type))) {
return make_error(Error::kInvalidAddress);
}
if (index_type == RegType::kVec128) {
op_flags |= InstDB::OpFlags::kVm32x | InstDB::OpFlags::kVm64x;
}
else if (index_type == RegType::kVec256) {
op_flags |= InstDB::OpFlags::kVm32y | InstDB::OpFlags::kVm64y;
}
else if (index_type == RegType::kVec512) {
op_flags |= InstDB::OpFlags::kVm32z | InstDB::OpFlags::kVm64z;
}
else {
if (base_type != RegType::kNone)
op_flags |= InstDB::OpFlags::kFlagMib;
}
// [RIP + {XMM|YMM|ZMM}] is not allowed.
if (base_type == RegType::kPC && Support::test(op_flags, InstDB::OpFlags::kVmMask)) {
return make_error(Error::kInvalidAddress);
}
uint32_t index_id = m.index_id();
if (index_id < Operand::kVirtIdMin) {
if (ASMJIT_UNLIKELY(index_id >= 32)) {
return make_error(Error::kInvalidPhysId);
}
combined_reg_mask |= Support::bit_mask<RegMask>(index_id);
}
else if (uint32_t(validation_flags & ValidationFlags::kEnableVirtRegs) == 0) {
return make_error(Error::kIllegalVirtReg);
}
// Only used for implicit memory operands having 'seg:[reg]' form, so clear it.
reg_mask = 0;
}
switch (mem_size) {
case 0: op_flags |= InstDB::OpFlags::kMemUnspecified; break;
case 1: op_flags |= InstDB::OpFlags::kMem8; break;
case 2: op_flags |= InstDB::OpFlags::kMem16; break;
case 4: op_flags |= InstDB::OpFlags::kMem32; break;
case 6: op_flags |= InstDB::OpFlags::kMem48; break;
case 8: op_flags |= InstDB::OpFlags::kMem64; break;
case 10: op_flags |= InstDB::OpFlags::kMem80; break;
case 16: op_flags |= InstDB::OpFlags::kMem128; break;
case 32: op_flags |= InstDB::OpFlags::kMem256; break;
case 64: op_flags |= InstDB::OpFlags::kMem512; break;
default:
return make_error(Error::kInvalidOperandSize);
}
break;
}
case OperandType::kImm: {
uint64_t imm_value = op.as<Imm>().value_as<uint64_t>();
if (int64_t(imm_value) >= 0) {
if (imm_value <= 0x7u) {
op_flags = InstDB::OpFlags::kImmI64 | InstDB::OpFlags::kImmU64 | InstDB::OpFlags::kImmI32 | InstDB::OpFlags::kImmU32 |
InstDB::OpFlags::kImmI16 | InstDB::OpFlags::kImmU16 | InstDB::OpFlags::kImmI8 | InstDB::OpFlags::kImmU8 |
InstDB::OpFlags::kImmI4 | InstDB::OpFlags::kImmU4 ;
}
else if (imm_value <= 0xFu) {
op_flags = InstDB::OpFlags::kImmI64 | InstDB::OpFlags::kImmU64 | InstDB::OpFlags::kImmI32 | InstDB::OpFlags::kImmU32 |
InstDB::OpFlags::kImmI16 | InstDB::OpFlags::kImmU16 | InstDB::OpFlags::kImmI8 | InstDB::OpFlags::kImmU8 |
InstDB::OpFlags::kImmU4 ;
}
else if (imm_value <= 0x7Fu) {
op_flags = InstDB::OpFlags::kImmI64 | InstDB::OpFlags::kImmU64 | InstDB::OpFlags::kImmI32 | InstDB::OpFlags::kImmU32 |
InstDB::OpFlags::kImmI16 | InstDB::OpFlags::kImmU16 | InstDB::OpFlags::kImmI8 | InstDB::OpFlags::kImmU8 ;
}
else if (imm_value <= 0xFFu) {
op_flags = InstDB::OpFlags::kImmI64 | InstDB::OpFlags::kImmU64 | InstDB::OpFlags::kImmI32 | InstDB::OpFlags::kImmU32 |
InstDB::OpFlags::kImmI16 | InstDB::OpFlags::kImmU16 | InstDB::OpFlags::kImmU8 ;
}
else if (imm_value <= 0x7FFFu) {
op_flags = InstDB::OpFlags::kImmI64 | InstDB::OpFlags::kImmU64 | InstDB::OpFlags::kImmI32 | InstDB::OpFlags::kImmU32 |
InstDB::OpFlags::kImmI16 | InstDB::OpFlags::kImmU16 ;
}
else if (imm_value <= 0xFFFFu) {
op_flags = InstDB::OpFlags::kImmI64 | InstDB::OpFlags::kImmU64 | InstDB::OpFlags::kImmI32 | InstDB::OpFlags::kImmU32 |
InstDB::OpFlags::kImmU16 ;
}
else if (imm_value <= 0x7FFFFFFFu) {
op_flags = InstDB::OpFlags::kImmI64 | InstDB::OpFlags::kImmU64 | InstDB::OpFlags::kImmI32 | InstDB::OpFlags::kImmU32;
}
else if (imm_value <= 0xFFFFFFFFu) {
op_flags = InstDB::OpFlags::kImmI64 | InstDB::OpFlags::kImmU64 | InstDB::OpFlags::kImmU32;
}
else if (imm_value <= 0x7FFFFFFFFFFFFFFFu) {
op_flags = InstDB::OpFlags::kImmI64 | InstDB::OpFlags::kImmU64;
}
else {
op_flags = InstDB::OpFlags::kImmU64;
}
}
else {
imm_value = Support::neg(imm_value);
if (imm_value <= 0x8u) {
op_flags = InstDB::OpFlags::kImmI64 | InstDB::OpFlags::kImmI32 | InstDB::OpFlags::kImmI16 | InstDB::OpFlags::kImmI8 | InstDB::OpFlags::kImmI4;
}
else if (imm_value <= 0x80u) {
op_flags = InstDB::OpFlags::kImmI64 | InstDB::OpFlags::kImmI32 | InstDB::OpFlags::kImmI16 | InstDB::OpFlags::kImmI8;
}
else if (imm_value <= 0x8000u) {
op_flags = InstDB::OpFlags::kImmI64 | InstDB::OpFlags::kImmI32 | InstDB::OpFlags::kImmI16;
}
else if (imm_value <= 0x80000000u) {
op_flags = InstDB::OpFlags::kImmI64 | InstDB::OpFlags::kImmI32;
}
else {
op_flags = InstDB::OpFlags::kImmI64;
}
}
break;
}
case OperandType::kLabel: {
op_flags |= InstDB::OpFlags::kRel8 | InstDB::OpFlags::kRel32;
break;
}
default:
return make_error(Error::kInvalidState);
}
InstDB::OpSignature& op_sig_dst = op_sig_translated[i];
op_sig_dst._flags = uint64_t(op_flags) & 0x00FFFFFFFFFFFFFFu;
op_sig_dst._reg_mask = uint8_t(reg_mask & 0xFFu);
combined_op_flags |= op_flags;
}
// Decrease the number of operands of those that are none. This is important as Assembler and Compiler may just pass
// more operands padded with none (which means that no operand is given at that index). However, validate that there
// are no gaps (like [reg, none, reg] or [none, reg]).
if (i < op_count) {
while (--op_count > i) {
if (ASMJIT_UNLIKELY(!operands[op_count].is_none())) {
return make_error(Error::kInvalidInstruction);
}
}
}
// Validate X86 and X64 specific cases.
if (mode == InstDB::Mode::kX86) {
// Illegal use of 64-bit register in 32-bit mode.
if (ASMJIT_UNLIKELY(Support::test(combined_op_flags, InstDB::OpFlags::kRegGpq))) {
return make_error(Error::kInvalidUseOfGpq);
}
}
else {
// Illegal use of a high 8-bit register with REX prefix.
bool has_rex = inst.has_option(InstOptions::kX86_Rex) || (combined_reg_mask & 0xFFFFFF00u) != 0;
if (ASMJIT_UNLIKELY(has_rex && Support::test(combined_op_flags, InstDB::OpFlags::kRegGpbHi))) {
return make_error(Error::kInvalidUseOfGpbHi);
}
}
// Validate Instruction Signature by Comparing Against All Signature Records
// -------------------------------------------------------------------------
bool inst_signature_matched = false;
Span<const InstDB::InstSignature> inst_signatures = common_info.inst_signatures();
if (!inst_signatures.is_empty()) {
const InstDB::OpSignature* op_signature_table = InstDB::_op_signature_table;
// If set it means that we matched a signature where only immediate value
// was out of bounds. We can return a more descriptive error if we know this.
bool global_imm_out_of_range = false;
for (const InstDB::InstSignature& inst_signature : inst_signatures) {
// Only match signatures that are compatible with the requested mode.
if (!inst_signature.supports_mode(mode)) {
continue;
}
// Compare the operands table with reference operands.
uint32_t j = 0;
uint32_t inst_op_count = inst_signature.op_count();
bool local_imm_out_of_range = false;
if (inst_op_count == op_count) {
for (j = 0; j < op_count; j++) {
if (!check_op_sig(op_sig_translated[j], inst_signature.op_signature(j), local_imm_out_of_range)) {
break;
}
}
}
else if (inst_op_count - inst_signature.implicit_op_count() == op_count) {
uint32_t r = 0;
for (j = 0; j < op_count && r < inst_op_count; j++, r++) {
const InstDB::OpSignature* op_chk = op_sig_translated + j;
const InstDB::OpSignature* op_ref;
Next:
op_ref = op_signature_table + inst_signature.op_signature_index(r);
// Skip implicit operands.
if (op_ref->is_implicit()) {
if (++r >= inst_op_count) {
break;
}
else {
goto Next;
}
}
if (!check_op_sig(*op_chk, *op_ref, local_imm_out_of_range)) {
break;
}
}
}
if (j == op_count) {
if (!local_imm_out_of_range) {
// The match must clear `global_imm_out_of_range` as we have matched all operands.
global_imm_out_of_range = false;
inst_signature_matched = true;
break;
}
global_imm_out_of_range = local_imm_out_of_range;
}
}
if (!inst_signature_matched) {
return make_error(global_imm_out_of_range ? Error::kInvalidImmediate : Error::kInvalidInstruction);
}
}
// Validate AVX512 Options
// -----------------------
const RegOnly& extra_reg = inst.extra_reg();
if (Support::test(options, kAvx512Options)) {
if (common_info.has_flag(InstDB::InstFlags::kEvex)) {
// Validate AVX-512 {z}.
if (Support::test(options, InstOptions::kX86_ZMask)) {
if (ASMJIT_UNLIKELY(Support::test(options, InstOptions::kX86_ZMask) && !common_info.has_avx512_z())) {
return make_error(Error::kInvalidKZeroUse);
}
}
// Validate AVX-512 {sae} and {er}.
if (Support::test(options, InstOptions::kX86_SAE | InstOptions::kX86_ER)) {
// Rounding control is impossible if the instruction is not reg-to-reg.
if (ASMJIT_UNLIKELY(mem_op)) {
return make_error(Error::kInvalidEROrSAE);
}
// Check if {sae} or {er} is supported by the instruction.
if (Support::test(options, InstOptions::kX86_ER)) {
// NOTE: if both {sae} and {er} are set, we don't care, as {sae} is implied.
if (ASMJIT_UNLIKELY(!common_info.has_avx512_er())) {
return make_error(Error::kInvalidEROrSAE);
}
}
else {
if (ASMJIT_UNLIKELY(!common_info.has_avx512_sae())) {
return make_error(Error::kInvalidEROrSAE);
}
}
// {sae} and {er} are defined for either scalar ops or vector ops that require LL to be 10 (512-bit vector
// operations). We don't need any more bits in the instruction database to be able to validate this, as
// each AVX512 instruction that has broadcast is vector instruction (in this case we require zmm registers),
// otherwise it's a scalar instruction, which is valid.
if (common_info.has_avx512_bcst()) {
// Supports broadcast, thus we require LL to be '10', which means there have to be ZMM registers used. We
// don't calculate LL here, but we know that it would be '10' if there is at least one ZMM register used.
// There is no {er}/{sae}-enabled instruction with less than two operands.
ASMJIT_ASSERT(op_count >= 2);
if (ASMJIT_UNLIKELY(!is_zmm_or_m512(operands[0]) && !is_zmm_or_m512(operands[1]))) {
return make_error(Error::kInvalidEROrSAE);
}
}
}
}
else {
// Not an AVX512 instruction - maybe OpExtra is xCX register used by REP/REPNE prefix.
if (Support::test(options, kAvx512Options) || !Support::test(options, kRepAny)) {
return make_error(Error::kInvalidInstruction);
}
}
}
// Validate {Extra} Register
// -------------------------
if (extra_reg.is_reg()) {
if (Support::test(options, kRepAny)) {
// Validate REP|REPNE {cx|ecx|rcx}.
if (ASMJIT_UNLIKELY(Support::test(inst_flags, InstDB::InstFlags::kRepIgnored))) {
return make_error(Error::kInvalidExtraReg);
}
if (extra_reg.is_phys_reg()) {
if (ASMJIT_UNLIKELY(extra_reg.id() != Gp::kIdCx)) {
return make_error(Error::kInvalidExtraReg);
}
}
// The type of the {...} register must match the type of the base register
// of memory operand. So if the memory operand uses 32-bit register the
// count register must also be 32-bit, etc...
if (ASMJIT_UNLIKELY(!mem_op || extra_reg.type() != mem_op->base_type())) {
return make_error(Error::kInvalidExtraReg);
}
}
else if (common_info.has_flag(InstDB::InstFlags::kEvex)) {
// Validate AVX-512 {k}.
if (ASMJIT_UNLIKELY(extra_reg.type() != RegType::kMask)) {
return make_error(Error::kInvalidExtraReg);
}
if (ASMJIT_UNLIKELY(extra_reg.id() == 0 || !common_info.has_avx512_k())) {
return make_error(Error::kInvalidKMaskUse);
}
}
else {
return make_error(Error::kInvalidExtraReg);
}
}
return Error::kOk;
}
Error validate_x86(const BaseInst& inst, const Operand_* operands, size_t op_count, ValidationFlags validation_flags) noexcept {
return validate(InstDB::Mode::kX86, inst, operands, op_count, validation_flags);
}
Error validate_x64(const BaseInst& inst, const Operand_* operands, size_t op_count, ValidationFlags validation_flags) noexcept {
return validate(InstDB::Mode::kX64, inst, operands, op_count, validation_flags);
}
#endif // !ASMJIT_NO_INTROSPECTION
// x86::InstInternal - QueryRWInfo
// ===============================
#ifndef ASMJIT_NO_INTROSPECTION
static const Support::Array<uint64_t, uint32_t(RegGroup::kMaxValue) + 1> rw_reg_group_byte_mask_table = {{
0x00000000000000FFu, // GP.
0xFFFFFFFFFFFFFFFFu, // XMM|YMM|ZMM.
0x00000000000000FFu, // MM.
0x00000000000000FFu, // KReg.
0x0000000000000003u, // SReg.
0x00000000000000FFu, // CReg.
0x00000000000000FFu, // DReg.
0x00000000000003FFu, // St().
0x000000000000FFFFu, // BND.
0x00000000000000FFu // RIP.
}};
static ASMJIT_INLINE void rw_zero_extend_gp(OpRWInfo& op_rw_info, const Gp& reg, uint32_t native_gp_size) noexcept {
if (reg.size() + 4 == native_gp_size) {
op_rw_info.add_op_flags(OpRWFlags::kZExt);
op_rw_info.set_extend_byte_mask(~op_rw_info.write_byte_mask() & 0xFFu);
}
}
static ASMJIT_INLINE void rw_zero_extend_avx_vec(OpRWInfo& op_rw_info, const Vec& reg) noexcept {
Support::maybe_unused(reg);
uint64_t msk = ~Support::fill_trailing_bits(op_rw_info.write_byte_mask());
if (msk) {
op_rw_info.add_op_flags(OpRWFlags::kZExt);
op_rw_info.set_extend_byte_mask(msk);
}
}
static ASMJIT_INLINE void rw_zero_extend_non_vec(OpRWInfo& op_rw_info, const Reg& reg) noexcept {
uint64_t msk = ~Support::fill_trailing_bits(op_rw_info.write_byte_mask()) & rw_reg_group_byte_mask_table[reg.reg_group()];
if (msk) {
op_rw_info.add_op_flags(OpRWFlags::kZExt);
op_rw_info.set_extend_byte_mask(msk);
}
}
static ASMJIT_INLINE Error rw_handle_avx512(const BaseInst& inst, const InstDB::CommonInfo& common_info, InstRWInfo* out) noexcept {
if (inst.has_extra_reg() && inst.extra_reg().type() == RegType::kMask && out->op_count() > 0) {
// AVX-512 instruction that uses a destination with {k} register (zeroing vs masking).
out->_extra_reg.add_op_flags(OpRWFlags::kRead);
out->_extra_reg.set_read_byte_mask(0xFF);
if (!inst.has_option(InstOptions::kX86_ZMask) && !common_info.has_avx512_flag(InstDB::Avx512Flags::kImplicitZ)) {
out->_operands[0].add_op_flags(OpRWFlags::kRead);
out->_operands[0]._read_byte_mask |= out->_operands[0]._write_byte_mask;
}
}
return Error::kOk;
}
static ASMJIT_INLINE bool has_same_reg_type(const Reg* regs, size_t op_count) noexcept {
ASMJIT_ASSERT(op_count > 0);
RegType reg_type = regs[0].reg_type();
for (size_t i = 1; i < op_count; i++) {
if (regs[i].reg_type() != reg_type) {
return false;
}
}
return true;
}
Error query_rw_info(Arch arch, const BaseInst& inst, const Operand_* operands, size_t op_count, InstRWInfo* out) noexcept {
// Only called when `arch` matches X86 family.
ASMJIT_ASSERT(Environment::is_family_x86(arch));
// Get the instruction data.
InstId inst_id = inst.inst_id();
if (ASMJIT_UNLIKELY(!Inst::is_defined_id(inst_id))) {
return make_error(Error::kInvalidInstruction);
}
// Read/Write flags.
const InstDB::InstInfo& inst_info = InstDB::_inst_info_table[inst_id];
const InstDB::CommonInfo& common_info = InstDB::_inst_common_info_table[inst_info._common_info_index];
const InstDB::AdditionalInfo& additional_info = InstDB::additional_info_table[inst_info._additional_info_index];
const InstDB::RWFlagsInfoTable& rw_flags = InstDB::rw_flags_info_table[additional_info._rw_flags_index];
// There are two data tables, one for `op_count == 2` and the second for
// `op_count != 2`. There are two reasons for that:
// - There are instructions that share the same name that have both 2 or 3 operands, which have different
// RW information / semantics.
// - There must be 2 tables otherwise the lookup index won't fit into 8 bits (there is more than 256 records
// of combined rw_info A and B).
const InstDB::RWInfo& inst_rw_info = op_count == 2 ? InstDB::rw_info_a_table[InstDB::rw_info_index_a_table[inst_id]]
: InstDB::rw_info_b_table[InstDB::rw_info_index_b_table[inst_id]];
const InstDB::RWInfoRm& inst_rm_info = InstDB::rw_info_rm_table[inst_rw_info.rm_info];
out->_inst_flags = InstDB::inst_flags_table[additional_info._inst_flags_index];
out->_op_count = uint8_t(op_count);
out->_rm_feature = inst_rm_info.rm_feature;
out->_extra_reg.reset();
out->_read_flags = CpuRWFlags(rw_flags.read_flags);
out->_write_flags = CpuRWFlags(rw_flags.write_flags);
uint32_t op_type_mask = 0u;
uint32_t native_gp_size = Environment::reg_size_of_arch(arch);
constexpr OpRWFlags R = OpRWFlags::kRead;
constexpr OpRWFlags W = OpRWFlags::kWrite;
constexpr OpRWFlags X = OpRWFlags::kRW;
constexpr OpRWFlags RegM = OpRWFlags::kRegMem;
constexpr OpRWFlags RegPhys = OpRWFlags::kRegPhysId;
constexpr OpRWFlags MibRead = OpRWFlags::kMemBaseRead | OpRWFlags::kMemIndexRead;
if (inst_rw_info.category <= uint32_t(InstDB::RWInfo::kCategoryGenericEx)) {
uint32_t i;
uint32_t rm_ops_mask = 0;
uint32_t rm_max_size = 0;
for (i = 0; i < op_count; i++) {
OpRWInfo& op = out->_operands[i];
const Operand_& src_op = operands[i];
const InstDB::RWInfoOp& rw_op_data = InstDB::rw_info_op_table[inst_rw_info.op_info_index[i]];
op_type_mask |= Support::bit_mask<uint32_t>(src_op.op_type());
if (!src_op.is_reg_or_mem()) {
op.reset();
continue;
}
op._op_flags = rw_op_data.flags & ~OpRWFlags::kZExt;
op._phys_id = rw_op_data.phys_id;
op._rm_size = 0;
op._reset_reserved();
uint64_t r_byte_mask = rw_op_data.r_byte_mask;
uint64_t w_byte_mask = rw_op_data.w_byte_mask;
if (op.is_read() && !r_byte_mask) {
r_byte_mask = Support::lsb_mask<uint64_t>(src_op.x86_rm_size());
}
if (op.is_write() && !w_byte_mask) {
w_byte_mask = Support::lsb_mask<uint64_t>(src_op.x86_rm_size());
}
op._read_byte_mask = r_byte_mask;
op._write_byte_mask = w_byte_mask;
op._extend_byte_mask = 0;
op._consecutive_lead_count = rw_op_data.consecutive_lead_count;
if (src_op.is_reg()) {
// Zero extension.
if (op.is_write()) {
if (src_op.as<Reg>().is_gp()) {
// GP registers on X64 are special:
// - 8-bit and 16-bit writes aren't zero extended.
// - 32-bit writes ARE zero extended.
rw_zero_extend_gp(op, src_op.as<Gp>(), native_gp_size);
}
else if (Support::test(rw_op_data.flags, OpRWFlags::kZExt)) {
// Otherwise follow ZExt.
rw_zero_extend_non_vec(op, src_op.as<Gp>());
}
}
// Aggregate values required to calculate valid Reg/M info.
rm_max_size = Support::max(rm_max_size, src_op.x86_rm_size());
rm_ops_mask |= Support::bit_mask<uint32_t>(i);
}
else {
const x86::Mem& mem_op = src_op.as<x86::Mem>();
// The RW flags of BASE+INDEX are either provided by the data, which means
// that the instruction is border-case, or they are deduced from the operand.
if (mem_op.has_base_reg() && !op.has_op_flag(OpRWFlags::kMemBaseRW)) {
op.add_op_flags(OpRWFlags::kMemBaseRead);
}
if (mem_op.has_index_reg() && !op.has_op_flag(OpRWFlags::kMemIndexRW)) {
op.add_op_flags(OpRWFlags::kMemIndexRead);
}
}
}
// Only keep kMovOp if the instruction is actually register to register move of the same kind.
if (out->has_inst_flag(InstRWFlags::kMovOp)) {
if (!(op_count >= 2 && op_type_mask == Support::bit_mask<uint32_t>(OperandType::kReg) && has_same_reg_type(reinterpret_cast<const Reg*>(operands), op_count))) {
out->_inst_flags &= ~InstRWFlags::kMovOp;
}
}
// Special cases require more logic.
if (inst_rm_info.flags & (InstDB::RWInfoRm::kFlagMovssMovsd | InstDB::RWInfoRm::kFlagPextrw | InstDB::RWInfoRm::kFlagFeatureIfRMI)) {
if (inst_rm_info.flags & InstDB::RWInfoRm::kFlagMovssMovsd) {
if (op_count == 2) {
if (operands[0].is_reg() && operands[1].is_reg()) {
// Doesn't zero extend the destination.
out->_operands[0]._extend_byte_mask = 0;
}
}
}
else if (inst_rm_info.flags & InstDB::RWInfoRm::kFlagPextrw) {
if (op_count == 3 && operands[1].is_mm_reg()) {
out->_rm_feature = 0;
rm_ops_mask = 0;
}
}
else if (inst_rm_info.flags & InstDB::RWInfoRm::kFlagFeatureIfRMI) {
if (op_count != 3 || !operands[2].is_imm()) {
out->_rm_feature = 0;
}
}
}
rm_ops_mask &= uint32_t(inst_rm_info.rm_ops_mask);
if (rm_ops_mask && !inst.has_option(InstOptions::kX86_ER)) {
Support::BitWordIterator<uint32_t> it(rm_ops_mask);
do {
i = it.next();
OpRWInfo& op = out->_operands[i];
op.add_op_flags(RegM);
switch (inst_rm_info.category) {
case InstDB::RWInfoRm::kCategoryFixed:
op.set_rm_size(inst_rm_info.fixed_size);
break;
case InstDB::RWInfoRm::kCategoryConsistent:
op.set_rm_size(operands[i].x86_rm_size());
break;
case InstDB::RWInfoRm::kCategoryHalf:
op.set_rm_size(rm_max_size / 2u);
break;
case InstDB::RWInfoRm::kCategoryQuarter:
op.set_rm_size(rm_max_size / 4u);
break;
case InstDB::RWInfoRm::kCategoryEighth:
op.set_rm_size(rm_max_size / 8u);
break;
}
} while (it.has_next());
}
// Special cases per instruction.
if (inst_rw_info.category == InstDB::RWInfo::kCategoryGenericEx) {
switch (inst.inst_id()) {
case Inst::kIdVpternlogd:
case Inst::kIdVpternlogq: {
if (op_count == 4 && operands[3].is_imm()) {
uint32_t predicate = operands[3].as<Imm>().value_as<uint8_t>();
if ((predicate >> 4) == (predicate & 0xF)) {
out->_operands[0].clear_op_flags(OpRWFlags::kRead);
out->_operands[0].set_read_byte_mask(0);
}
}
break;
}
default:
break;
}
}
return rw_handle_avx512(inst, common_info, out);
}
switch (inst_rw_info.category) {
case InstDB::RWInfo::kCategoryMov: {
// Special case for 'mov' instruction. Here there are some variants that we have to handle as 'mov' can be
// used to move between GP, segment, control and debug registers. Moving between GP registers also allow to
// use memory operand.
// We will again set the flag if it's actually a move from GP to GP register, otherwise this flag cannot be set.
out->_inst_flags &= ~InstRWFlags::kMovOp;
if (op_count == 2) {
if (operands[0].is_reg() && operands[1].is_reg()) {
const Reg& o0 = operands[0].as<Reg>();
const Reg& o1 = operands[1].as<Reg>();
if (o0.is_gp() && o1.is_gp()) {
out->_operands[0].reset(W | RegM, operands[0].x86_rm_size());
out->_operands[1].reset(R | RegM, operands[1].x86_rm_size());
rw_zero_extend_gp(out->_operands[0], operands[0].as<Gp>(), native_gp_size);
out->_inst_flags |= InstRWFlags::kMovOp;
return Error::kOk;
}
if (o0.is_gp() && o1.is_segment_reg()) {
out->_operands[0].reset(W | RegM, native_gp_size);
out->_operands[0].set_rm_size(2);
out->_operands[1].reset(R, 2);
return Error::kOk;
}
if (o0.is_segment_reg() && o1.is_gp()) {
out->_operands[0].reset(W, 2);
out->_operands[1].reset(R | RegM, 2);
out->_operands[1].set_rm_size(2);
return Error::kOk;
}
if (o0.is_gp() && (o1.is_control_reg() || o1.is_debug_reg())) {
out->_operands[0].reset(W, native_gp_size);
out->_operands[1].reset(R, native_gp_size);
out->_write_flags = CpuRWFlags::kX86_OF |
CpuRWFlags::kX86_SF |
CpuRWFlags::kX86_ZF |
CpuRWFlags::kX86_AF |
CpuRWFlags::kX86_PF |
CpuRWFlags::kX86_CF;
return Error::kOk;
}
if ((o0.is_control_reg() || o0.is_debug_reg()) && o1.is_gp()) {
out->_operands[0].reset(W, native_gp_size);
out->_operands[1].reset(R, native_gp_size);
out->_write_flags = CpuRWFlags::kX86_OF |
CpuRWFlags::kX86_SF |
CpuRWFlags::kX86_ZF |
CpuRWFlags::kX86_AF |
CpuRWFlags::kX86_PF |
CpuRWFlags::kX86_CF;
return Error::kOk;
}
}
if (operands[0].is_reg() && operands[1].is_mem()) {
const Reg& o0 = operands[0].as<Reg>();
const Mem& o1 = operands[1].as<Mem>();
if (o0.is_gp()) {
if (!o1.is_offset_64bit()) {
out->_operands[0].reset(W, o0.size());
}
else {
out->_operands[0].reset(W | RegPhys, o0.size(), Gp::kIdAx);
}
out->_operands[1].reset(R | MibRead, o0.size());
rw_zero_extend_gp(out->_operands[0], operands[0].as<Gp>(), native_gp_size);
return Error::kOk;
}
if (o0.is_segment_reg()) {
out->_operands[0].reset(W, 2);
out->_operands[1].reset(R, 2);
return Error::kOk;
}
}
if (operands[0].is_mem() && operands[1].is_reg()) {
const Mem& o0 = operands[0].as<Mem>();
const Reg& o1 = operands[1].as<Reg>();
if (o1.is_gp()) {
out->_operands[0].reset(W | MibRead, o1.size());
if (!o0.is_offset_64bit()) {
out->_operands[1].reset(R, o1.size());
}
else {
out->_operands[1].reset(R | RegPhys, o1.size(), Gp::kIdAx);
}
return Error::kOk;
}
if (o1.is_segment_reg()) {
out->_operands[0].reset(W | MibRead, 2);
out->_operands[1].reset(R, 2);
return Error::kOk;
}
}
if (operands[0].is_gp() && operands[1].is_imm()) {
const Reg& o0 = operands[0].as<Reg>();
out->_operands[0].reset(W | RegM, o0.size());
out->_operands[1].reset();
rw_zero_extend_gp(out->_operands[0], operands[0].as<Gp>(), native_gp_size);
return Error::kOk;
}
if (operands[0].is_mem() && operands[1].is_imm()) {
const Reg& o0 = operands[0].as<Reg>();
out->_operands[0].reset(W | MibRead, o0.size());
out->_operands[1].reset();
return Error::kOk;
}
}
break;
}
case InstDB::RWInfo::kCategoryMovabs: {
if (op_count == 2) {
if (operands[0].is_gp() && operands[1].is_mem()) {
const Reg& o0 = operands[0].as<Reg>();
out->_operands[0].reset(W | RegPhys, o0.size(), Gp::kIdAx);
out->_operands[1].reset(R | MibRead, o0.size());
rw_zero_extend_gp(out->_operands[0], operands[0].as<Gp>(), native_gp_size);
return Error::kOk;
}
if (operands[0].is_mem() && operands[1].is_gp()) {
const Reg& o1 = operands[1].as<Reg>();
out->_operands[0].reset(W | MibRead, o1.size());
out->_operands[1].reset(R | RegPhys, o1.size(), Gp::kIdAx);
return Error::kOk;
}
if (operands[0].is_gp() && operands[1].is_imm()) {
const Reg& o0 = operands[0].as<Reg>();
out->_operands[0].reset(W, o0.size());
out->_operands[1].reset();
rw_zero_extend_gp(out->_operands[0], operands[0].as<Gp>(), native_gp_size);
return Error::kOk;
}
}
break;
}
case InstDB::RWInfo::kCategoryImul: {
// Special case for 'imul' instruction.
//
// There are 3 variants in general:
//
// 1. Standard multiplication: 'A = A * B'.
// 2. Multiplication with imm: 'A = B * C'.
// 3. Extended multiplication: 'A:B = B * C'.
if (op_count == 2) {
if (operands[0].is_reg() && operands[1].is_imm()) {
out->_operands[0].reset(X, operands[0].as<Reg>().size());
out->_operands[1].reset();
rw_zero_extend_gp(out->_operands[0], operands[0].as<Gp>(), native_gp_size);
return Error::kOk;
}
if (operands[0].is_gp16() && operands[1].x86_rm_size() == 1) {
// imul ax, r8/m8 <- AX = AL * r8/m8
out->_operands[0].reset(X | RegPhys, 2, Gp::kIdAx);
out->_operands[0].set_read_byte_mask(Support::lsb_mask<uint64_t>(1));
out->_operands[1].reset(R | RegM, 1);
}
else {
// imul r?, r?/m?
out->_operands[0].reset(X, operands[0].as<Gp>().size());
out->_operands[1].reset(R | RegM, operands[0].as<Gp>().size());
rw_zero_extend_gp(out->_operands[0], operands[0].as<Gp>(), native_gp_size);
}
if (operands[1].is_mem()) {
out->_operands[1].add_op_flags(MibRead);
}
return Error::kOk;
}
if (op_count == 3) {
if (operands[2].is_imm()) {
out->_operands[0].reset(W, operands[0].x86_rm_size());
out->_operands[1].reset(R | RegM, operands[1].x86_rm_size());
out->_operands[2].reset();
rw_zero_extend_gp(out->_operands[0], operands[0].as<Gp>(), native_gp_size);
if (operands[1].is_mem()) {
out->_operands[1].add_op_flags(MibRead);
}
return Error::kOk;
}
else {
out->_operands[0].reset(W | RegPhys, operands[0].x86_rm_size(), Gp::kIdDx);
out->_operands[1].reset(X | RegPhys, operands[1].x86_rm_size(), Gp::kIdAx);
out->_operands[2].reset(R | RegM, operands[2].x86_rm_size());
rw_zero_extend_gp(out->_operands[0], operands[0].as<Gp>(), native_gp_size);
rw_zero_extend_gp(out->_operands[1], operands[1].as<Gp>(), native_gp_size);
if (operands[2].is_mem()) {
out->_operands[2].add_op_flags(MibRead);
}
return Error::kOk;
}
}
break;
}
case InstDB::RWInfo::kCategoryMovh64: {
// Special case for 'movhpd|movhps' instructions. Note that this is only required for legacy (non-AVX)
// variants as AVX instructions use either 2 or 3 operands that are in `kCategoryGeneric` category.
if (op_count == 2) {
if (operands[0].is_vec() && operands[1].is_mem()) {
out->_operands[0].reset(W, 8);
out->_operands[0].set_write_byte_mask(Support::lsb_mask<uint64_t>(8) << 8);
out->_operands[1].reset(R | MibRead, 8);
return Error::kOk;
}
if (operands[0].is_mem() && operands[1].is_vec()) {
out->_operands[0].reset(W | MibRead, 8);
out->_operands[1].reset(R, 8);
out->_operands[1].set_read_byte_mask(Support::lsb_mask<uint64_t>(8) << 8);
return Error::kOk;
}
}
break;
}
case InstDB::RWInfo::kCategoryPunpcklxx: {
// Special case for 'punpcklbw|punpckldq|punpcklwd' instructions.
if (op_count == 2) {
if (operands[0].is_vec128()) {
out->_operands[0].reset(X, 16);
out->_operands[0].set_read_byte_mask(0x0F0Fu);
out->_operands[0].set_write_byte_mask(0xFFFFu);
out->_operands[1].reset(R, 16);
out->_operands[1].set_write_byte_mask(0x0F0Fu);
if (operands[1].is_vec128()) {
return Error::kOk;
}
if (operands[1].is_mem()) {
out->_operands[1].add_op_flags(MibRead);
return Error::kOk;
}
}
if (operands[0].is_mm_reg()) {
out->_operands[0].reset(X, 8);
out->_operands[0].set_read_byte_mask(0x0Fu);
out->_operands[0].set_write_byte_mask(0xFFu);
out->_operands[1].reset(R, 4);
out->_operands[1].set_read_byte_mask(0x0Fu);
if (operands[1].is_mm_reg()) {
return Error::kOk;
}
if (operands[1].is_mem()) {
out->_operands[1].add_op_flags(MibRead);
return Error::kOk;
}
}
}
break;
}
case InstDB::RWInfo::kCategoryVmaskmov: {
// Special case for 'vmaskmovpd|vmaskmovps|vpmaskmovd|vpmaskmovq' instructions.
if (op_count == 3) {
if (operands[0].is_vec() && operands[1].is_vec() && operands[2].is_mem()) {
out->_operands[0].reset(W, operands[0].x86_rm_size());
out->_operands[1].reset(R, operands[1].x86_rm_size());
out->_operands[2].reset(R | MibRead, operands[1].x86_rm_size());
rw_zero_extend_avx_vec(out->_operands[0], operands[0].as<Vec>());
return Error::kOk;
}
if (operands[0].is_mem() && operands[1].is_vec() && operands[2].is_vec()) {
out->_operands[0].reset(X | MibRead, operands[1].x86_rm_size());
out->_operands[1].reset(R, operands[1].x86_rm_size());
out->_operands[2].reset(R, operands[2].x86_rm_size());
return Error::kOk;
}
}
break;
}
case InstDB::RWInfo::kCategoryVmovddup: {
// Special case for 'vmovddup' instruction. This instruction has an interesting semantic as 128-bit XMM
// version only uses 64-bit memory operand (m64), however, 256/512-bit versions use 256/512-bit memory
// operand, respectively.
if (op_count == 2) {
if (operands[0].is_vec() && operands[1].is_vec()) {
uint32_t o0_size = operands[0].x86_rm_size();
uint32_t o1_size = o0_size == 16 ? 8 : o0_size;
out->_operands[0].reset(W, o0_size);
out->_operands[1].reset(R | RegM, o1_size);
out->_operands[1]._read_byte_mask &= 0x00FF00FF00FF00FFu;
rw_zero_extend_avx_vec(out->_operands[0], operands[0].as<Vec>());
return rw_handle_avx512(inst, common_info, out);
}
if (operands[0].is_vec() && operands[1].is_mem()) {
uint32_t o0_size = operands[0].x86_rm_size();
uint32_t o1_size = o0_size == 16 ? 8 : o0_size;
out->_operands[0].reset(W, o0_size);
out->_operands[1].reset(R | MibRead, o1_size);
rw_zero_extend_avx_vec(out->_operands[0], operands[0].as<Vec>());
return rw_handle_avx512(inst, common_info, out);
}
}
break;
}
case InstDB::RWInfo::kCategoryVmovmskpd:
case InstDB::RWInfo::kCategoryVmovmskps: {
// Special case for 'vmovmskpd|vmovmskps' instructions.
if (op_count == 2) {
if (operands[0].is_gp() && operands[1].is_vec()) {
out->_operands[0].reset(W, 1);
out->_operands[0].set_extend_byte_mask(Support::lsb_mask<uint32_t>(native_gp_size - 1) << 1);
out->_operands[1].reset(R, operands[1].x86_rm_size());
return Error::kOk;
}
}
break;
}
case InstDB::RWInfo::kCategoryVmov1_2:
case InstDB::RWInfo::kCategoryVmov1_4:
case InstDB::RWInfo::kCategoryVmov1_8: {
// Special case for instructions where the destination is 1:N (narrowing).
//
// Vmov1_2:
// vcvtpd2dq|vcvttpd2dq
// vcvtpd2udq|vcvttpd2udq
// vcvtpd2ps|vcvtps2ph
// vcvtqq2ps|vcvtuqq2ps
// vpmovwb|vpmovswb|vpmovuswb
// vpmovdw|vpmovsdw|vpmovusdw
// vpmovqd|vpmovsqd|vpmovusqd
//
// Vmov1_4:
// vpmovdb|vpmovsdb|vpmovusdb
// vpmovqw|vpmovsqw|vpmovusqw
//
// Vmov1_8:
// pmovmskb|vpmovmskb
// vpmovqb|vpmovsqb|vpmovusqb
uint32_t shift = inst_rw_info.category - InstDB::RWInfo::kCategoryVmov1_2 + 1;
if (op_count >= 2) {
if (op_count >= 3) {
if (op_count > 3) {
return make_error(Error::kInvalidInstruction);
}
out->_operands[2].reset();
}
if (operands[0].is_reg() && operands[1].is_reg()) {
uint32_t size1 = operands[1].x86_rm_size();
uint32_t size0 = size1 >> shift;
out->_operands[0].reset(W, size0);
out->_operands[1].reset(R, size1);
if (inst_rm_info.rm_ops_mask & 0x1) {
out->_operands[0].add_op_flags(RegM);
out->_operands[0].set_rm_size(size0);
}
if (inst_rm_info.rm_ops_mask & 0x2) {
out->_operands[1].add_op_flags(RegM);
out->_operands[1].set_rm_size(size1);
}
// Handle 'pmovmskb|vpmovmskb'.
if (operands[0].is_gp()) {
rw_zero_extend_gp(out->_operands[0], operands[0].as<Gp>(), native_gp_size);
}
if (operands[0].is_vec()) {
rw_zero_extend_avx_vec(out->_operands[0], operands[0].as<Vec>());
}
return rw_handle_avx512(inst, common_info, out);
}
if (operands[0].is_reg() && operands[1].is_mem()) {
uint32_t size1 = operands[1].x86_rm_size() ? operands[1].x86_rm_size() : uint32_t(16);
uint32_t size0 = size1 >> shift;
out->_operands[0].reset(W, size0);
out->_operands[1].reset(R | MibRead, size1);
if (operands[0].is_vec()) {
rw_zero_extend_avx_vec(out->_operands[0], operands[0].as<Vec>());
}
return Error::kOk;
}
if (operands[0].is_mem() && operands[1].is_reg()) {
uint32_t size1 = operands[1].x86_rm_size();
uint32_t size0 = size1 >> shift;
out->_operands[0].reset(W | MibRead, size0);
out->_operands[1].reset(R, size1);
return rw_handle_avx512(inst, common_info, out);
}
}
break;
}
case InstDB::RWInfo::kCategoryVmov2_1:
case InstDB::RWInfo::kCategoryVmov4_1:
case InstDB::RWInfo::kCategoryVmov8_1: {
// Special case for instructions where the destination is N:1 (widening).
//
// Vmov2_1:
// vcvtdq2pd|vcvtudq2pd
// vcvtps2pd|vcvtph2ps
// vcvtps2qq|vcvtps2uqq
// vcvttps2qq|vcvttps2uqq
// vpmovsxbw|vpmovzxbw
// vpmovsxwd|vpmovzxwd
// vpmovsxdq|vpmovzxdq
//
// Vmov4_1:
// vpmovsxbd|vpmovzxbd
// vpmovsxwq|vpmovzxwq
//
// Vmov8_1:
// vpmovsxbq|vpmovzxbq
uint32_t shift = inst_rw_info.category - InstDB::RWInfo::kCategoryVmov2_1 + 1;
if (op_count >= 2) {
if (op_count >= 3) {
if (op_count > 3) {
return make_error(Error::kInvalidInstruction);
}
out->_operands[2].reset();
}
uint32_t size0 = operands[0].x86_rm_size();
uint32_t size1 = size0 >> shift;
out->_operands[0].reset(W, size0);
out->_operands[1].reset(R, size1);
if (operands[0].is_vec()) {
rw_zero_extend_avx_vec(out->_operands[0], operands[0].as<Vec>());
}
if (operands[0].is_reg() && operands[1].is_reg()) {
if (inst_rm_info.rm_ops_mask & 0x1) {
out->_operands[0].add_op_flags(RegM);
out->_operands[0].set_rm_size(size0);
}
if (inst_rm_info.rm_ops_mask & 0x2) {
out->_operands[1].add_op_flags(RegM);
out->_operands[1].set_rm_size(size1);
}
return rw_handle_avx512(inst, common_info, out);
}
if (operands[0].is_reg() && operands[1].is_mem()) {
out->_operands[1].add_op_flags(MibRead);
return rw_handle_avx512(inst, common_info, out);
}
}
break;
}
}
return make_error(Error::kInvalidInstruction);
}
#endif // !ASMJIT_NO_INTROSPECTION
// x86::InstInternal - QueryFeatures
// =================================
#ifndef ASMJIT_NO_INTROSPECTION
struct RegAnalysis {
uint32_t reg_type_mask;
uint32_t high_vec_used;
inline bool has_reg_type(RegType reg_type) const noexcept {
return Support::bit_test(reg_type_mask, reg_type);
}
};
static RegAnalysis InstInternal_reg_analysis(const Operand_* operands, size_t op_count) noexcept {
uint32_t mask = 0;
uint32_t high_vec_used = 0;
for (uint32_t i = 0; i < op_count; i++) {
const Operand_& op = operands[i];
if (op.is_reg()) {
const Reg& reg = op.as<Reg>();
mask |= Support::bit_mask<uint32_t>(reg.reg_type());
if (reg.is_vec()) {
high_vec_used |= uint32_t(reg.id() >= 16 && reg.id() < 32);
}
}
else if (op.is_mem()) {
const BaseMem& mem = op.as<BaseMem>();
if (mem.has_base_reg()) {
mask |= Support::bit_mask<uint32_t>(mem.base_type());
}
if (mem.has_index_reg()) {
mask |= Support::bit_mask<uint32_t>(mem.index_type());
high_vec_used |= uint32_t(mem.index_id() >= 16 && mem.index_id() < 32);
}
}
}
return RegAnalysis { mask, high_vec_used };
}
static inline uint32_t InstInternal_usesAvx512(InstOptions inst_options, const RegOnly& extra_reg, const RegAnalysis& reg_analysis) noexcept {
uint32_t has_evex = uint32_t(inst_options & (InstOptions::kX86_Evex | InstOptions::kX86_AVX512Mask));
uint32_t has_kmask = extra_reg.type() == RegType::kMask;
uint32_t has_k_or_zmm = reg_analysis.reg_type_mask & Support::bit_mask<uint32_t>(RegType::kVec512, RegType::kMask);
return has_evex | has_kmask | has_k_or_zmm;
}
Error query_features(Arch arch, const BaseInst& inst, const Operand_* operands, size_t op_count, CpuFeatures* out) noexcept {
using Ext = CpuFeatures::X86;
// Only called when `arch` matches X86 family.
Support::maybe_unused(arch);
ASMJIT_ASSERT(Environment::is_family_x86(arch));
// Get the instruction data.
InstId inst_id = inst.inst_id();
InstOptions options = inst.options();
if (ASMJIT_UNLIKELY(!Inst::is_defined_id(inst_id))) {
return make_error(Error::kInvalidInstruction);
}
const InstDB::InstInfo& inst_info = InstDB::inst_info_by_id(inst_id);
const InstDB::AdditionalInfo& additional_info = InstDB::additional_info_table[inst_info._additional_info_index];
const uint8_t* feature_data = additional_info.features_begin();
const uint8_t* feature_data_end = additional_info.features_end();
// Copy all features to `out`.
out->reset();
do {
uint32_t feature = feature_data[0];
if (!feature) {
break;
}
out->add(feature);
} while (++feature_data != feature_data_end);
// Since AsmJit aggregates instructions that share the same name we have to
// deal with some special cases and also with MMX/SSE and AVX/AVX2 overlaps.
if (feature_data != additional_info.features_begin()) {
RegAnalysis reg_analysis = InstInternal_reg_analysis(operands, op_count);
// Handle MMX vs SSE overlap.
if (out->has(Ext::kMMX) || out->has(Ext::kMMX2)) {
// Only instructions defined by SSE and SSE2 overlap. Instructions introduced by newer instruction sets like
// SSE3+ don't state MMX as they require SSE3+.
if (out->has(Ext::kSSE) || out->has(Ext::kSSE2)) {
if (!reg_analysis.has_reg_type(RegType::kVec128)) {
// The instruction doesn't use XMM register(s), thus it's MMX/MMX2 only.
out->remove(Ext::kSSE);
out->remove(Ext::kSSE2);
out->remove(Ext::kSSE4_1);
}
else {
out->remove(Ext::kMMX);
out->remove(Ext::kMMX2);
}
// Special case: PEXTRW instruction is MMX/SSE2 instruction. However, MMX/SSE version cannot access memory
// (only register to register extract) so when SSE4.1 introduced the whole family of PEXTR/PINSR instructions
// they also introduced PEXTRW with a new opcode 0x15 that can extract directly to memory. This instruction
// is, of course, not compatible with MMX/SSE2 and would #UD if SSE4.1 is not supported.
if (inst_id == Inst::kIdPextrw) {
if (op_count >= 1 && operands[0].is_mem())
out->remove(Ext::kSSE2);
else
out->remove(Ext::kSSE4_1);
}
}
}
// Handle PCLMULQDQ vs VPCLMULQDQ.
if (out->has(Ext::kVPCLMULQDQ)) {
if (reg_analysis.has_reg_type(RegType::kVec512) || Support::test(options, InstOptions::kX86_Evex)) {
// AVX512_F & VPCLMULQDQ.
out->remove(Ext::kAVX, Ext::kPCLMULQDQ);
}
else if (reg_analysis.has_reg_type(RegType::kVec256)) {
out->remove(Ext::kAVX512_F, Ext::kAVX512_VL);
}
else {
// AVX & PCLMULQDQ.
out->remove(Ext::kAVX512_F, Ext::kAVX512_VL, Ext::kVPCLMULQDQ);
}
}
// Handle AVX vs AVX2 overlap.
if (out->has(Ext::kAVX) && out->has(Ext::kAVX2)) {
bool is_avx2 = true;
// Special case: VBROADCASTSS and VBROADCASTSD were introduced in AVX, but only version that uses memory as a
// source operand. AVX2 then added support for register source operand.
if (inst_id == Inst::kIdVbroadcastss || inst_id == Inst::kIdVbroadcastsd) {
if (op_count > 1 && operands[1].is_mem()) {
is_avx2 = false;
}
}
else {
// AVX instruction set doesn't support integer operations on YMM registers as these were later introcuced by
// AVX2. In our case we have to check if YMM register(s) are in use and if that is the case this is an AVX2
// instruction.
if (!(reg_analysis.reg_type_mask & Support::bit_mask<uint32_t>(RegType::kVec256, RegType::kVec512))) {
is_avx2 = false;
}
}
out->remove(is_avx2 ? Ext::kAVX : Ext::kAVX2);
}
// Handle AVX vs AVX512 overlap.
//
// In general, non-AVX encoding is preferred, however, AVX encoded instructions that were initially provided
// as AVX-512 instructions must naturally prefer AVX-512 encoding, as that was the first one provided.
if (out->has_any(Ext::kAVX,
Ext::kAVX_IFMA,
Ext::kAVX_NE_CONVERT,
Ext::kAVX_VNNI,
Ext::kAVX2,
Ext::kF16C,
Ext::kFMA) &&
out->has_any(Ext::kAVX512_BF16,
Ext::kAVX512_BW,
Ext::kAVX512_DQ,
Ext::kAVX512_F,
Ext::kAVX512_IFMA,
Ext::kAVX512_VNNI)) {
uint32_t use_evex = InstInternal_usesAvx512(options, inst.extra_reg(), reg_analysis) | reg_analysis.high_vec_used;
switch (inst_id) {
// Special case: VPBROADCAST[B|D|Q|W] only supports r32/r64 with EVEX prefix.
case Inst::kIdVpbroadcastb:
case Inst::kIdVpbroadcastd:
case Inst::kIdVpbroadcastq:
case Inst::kIdVpbroadcastw:
use_evex |= uint32_t(op_count >= 2 && operands[1].is_gp());
break;
case Inst::kIdVcvtpd2dq:
case Inst::kIdVcvtpd2ps:
case Inst::kIdVcvttpd2dq:
use_evex |= uint32_t(op_count >= 2 && operands[0].is_vec256());
break;
case Inst::kIdVgatherdpd:
case Inst::kIdVgatherdps:
case Inst::kIdVgatherqpd:
case Inst::kIdVgatherqps:
case Inst::kIdVpgatherdd:
case Inst::kIdVpgatherdq:
case Inst::kIdVpgatherqd:
case Inst::kIdVpgatherqq:
use_evex |= uint32_t(op_count == 2);
break;
// Special case: These instructions only allow `reg, reg. imm` combination in AVX|AVX2 mode, then
// AVX-512 introduced `reg, reg/mem, imm` combination that uses EVEX prefix. This means that if
// the second operand is memory then this is AVX-512_BW instruction and not AVX/AVX2 instruction.
case Inst::kIdVpslldq:
case Inst::kIdVpslld:
case Inst::kIdVpsllq:
case Inst::kIdVpsllw:
case Inst::kIdVpsrad:
case Inst::kIdVpsraq:
case Inst::kIdVpsraw:
case Inst::kIdVpsrld:
case Inst::kIdVpsrldq:
case Inst::kIdVpsrlq:
case Inst::kIdVpsrlw:
use_evex |= uint32_t(op_count >= 2 && operands[1].is_mem());
break;
// Special case: VPERMPD - AVX2 vs AVX512-F case.
case Inst::kIdVpermpd:
use_evex |= uint32_t(op_count >= 3 && !operands[2].is_imm());
break;
// Special case: VPERMQ - AVX2 vs AVX512-F case.
case Inst::kIdVpermq:
use_evex |= uint32_t(op_count >= 3 && (operands[1].is_mem() || !operands[2].is_imm()));
break;
}
if (inst_info.common_info().prefer_evex() && !Support::test(options, InstOptions::kX86_Vex | InstOptions::kX86_Vex3))
use_evex = 1;
if (use_evex) {
out->remove(Ext::kAVX,
Ext::kAVX_IFMA,
Ext::kAVX_NE_CONVERT,
Ext::kAVX_VNNI,
Ext::kAVX2,
Ext::kF16C,
Ext::kFMA);
}
else {
out->remove(Ext::kAVX512_BF16,
Ext::kAVX512_BW,
Ext::kAVX512_DQ,
Ext::kAVX512_F,
Ext::kAVX512_IFMA,
Ext::kAVX512_VL,
Ext::kAVX512_VNNI);
}
}
// Clear AVX512_VL if ZMM register is used.
if (reg_analysis.has_reg_type(RegType::kVec512)) {
out->remove(Ext::kAVX512_VL);
}
}
return Error::kOk;
}
#endif // !ASMJIT_NO_INTROSPECTION
} // {InstInternal}
// x86::InstInternal - Tests
// =========================
#if defined(ASMJIT_TEST)
#ifndef ASMJIT_NO_TEXT
UNIT(x86_inst_api_text) {
// All known instructions should be matched.
INFO("Matching all X86 instructions");
for (uint32_t a = 1; a < Inst::_kIdCount; a++) {
StringTmp<128> a_name;
EXPECT_EQ(InstInternal::inst_id_to_string(a, InstStringifyOptions::kNone, a_name), Error::kOk)
.message("Failed to get the name of instruction #%u", a);
uint32_t b = InstInternal::string_to_inst_id(a_name.data(), a_name.size());
StringTmp<128> b_name;
InstInternal::inst_id_to_string(b, InstStringifyOptions::kNone, b_name);
EXPECT_EQ(a, b)
.message("Instructions do not match \"%s\" (#%u) != \"%s\" (#%u)", a_name.data(), a, b_name.data(), b);
}
}
#endif // !ASMJIT_NO_TEXT
#ifndef ASMJIT_NO_INTROSPECTION
template<typename... Args>
static Error query_features_inline(CpuFeatures* out, Arch arch, BaseInst inst, Args&&... args) {
Operand_ op_array[] = { std::forward<Args>(args)... };
return InstInternal::query_features(arch, inst, op_array, sizeof...(args), out);
}
UNIT(x86_inst_api_cpu_features) {
INFO("Verifying whether SSE2+ features are reported correctly for legacy instructions");
{
CpuFeatures f;
query_features_inline(&f, Arch::kX64, BaseInst(Inst::kIdPaddd), xmm1, xmm2);
EXPECT_TRUE(f.x86().has_sse2());
query_features_inline(&f, Arch::kX64, BaseInst(Inst::kIdAddsubpd), xmm1, xmm2);
EXPECT_TRUE(f.x86().has_sse3());
query_features_inline(&f, Arch::kX64, BaseInst(Inst::kIdPshufb), xmm1, xmm2);
EXPECT_TRUE(f.x86().has_ssse3());
query_features_inline(&f, Arch::kX64, BaseInst(Inst::kIdBlendpd), xmm1, xmm2, Imm(1));
EXPECT_TRUE(f.x86().has_sse4_1());
query_features_inline(&f, Arch::kX64, BaseInst(Inst::kIdCrc32), eax, al);
EXPECT_TRUE(f.x86().has_sse4_2());
}
INFO("Verifying whether AVX+ features are reported correctly for AVX instructions");
{
CpuFeatures f;
query_features_inline(&f, Arch::kX64, BaseInst(Inst::kIdVpaddd), xmm1, xmm2, xmm3);
EXPECT_TRUE(f.x86().has_avx());
query_features_inline(&f, Arch::kX64, BaseInst(Inst::kIdVpaddd), ymm1, ymm2, ymm3);
EXPECT_TRUE(f.x86().has_avx2());
query_features_inline(&f, Arch::kX64, BaseInst(Inst::kIdVaddsubpd), xmm1, xmm2, xmm3);
EXPECT_TRUE(f.x86().has_avx());
query_features_inline(&f, Arch::kX64, BaseInst(Inst::kIdVaddsubpd), ymm1, ymm2, ymm3);
EXPECT_TRUE(f.x86().has_avx());
query_features_inline(&f, Arch::kX64, BaseInst(Inst::kIdVpshufb), xmm1, xmm2, xmm3);
EXPECT_TRUE(f.x86().has_avx());
query_features_inline(&f, Arch::kX64, BaseInst(Inst::kIdVpshufb), ymm1, ymm2, ymm3);
EXPECT_TRUE(f.x86().has_avx2());
query_features_inline(&f, Arch::kX64, BaseInst(Inst::kIdVblendpd), xmm1, xmm2, xmm3, Imm(1));
EXPECT_TRUE(f.x86().has_avx());
query_features_inline(&f, Arch::kX64, BaseInst(Inst::kIdVblendpd), ymm1, ymm2, ymm3, Imm(1));
EXPECT_TRUE(f.x86().has_avx());
query_features_inline(&f, Arch::kX64, BaseInst(Inst::kIdVpunpcklbw), xmm1, xmm2, xmm3);
EXPECT_TRUE(f.x86().has_avx());
query_features_inline(&f, Arch::kX64, BaseInst(Inst::kIdVpunpcklbw), ymm1, ymm2, ymm3);
EXPECT_TRUE(f.x86().has_avx2());
}
INFO("Verifying whether AVX2 / AVX512 features are reported correctly for vpgatherxx instructions");
{
CpuFeatures f;
query_features_inline(&f, Arch::kX64, BaseInst(Inst::kIdVpgatherdd), xmm1, ptr(rax, xmm2), xmm3);
EXPECT_TRUE(f.x86().has_avx2());
EXPECT_FALSE(f.x86().has_avx512_f());
// NOTE: This instruction is unencodable, but sometimes this signature is used to check the support (without the {k}).
query_features_inline(&f, Arch::kX64, BaseInst(Inst::kIdVpgatherdd), xmm1, ptr(rax, xmm2));
EXPECT_FALSE(f.x86().has_avx2());
EXPECT_TRUE(f.x86().has_avx512_f());
query_features_inline(&f, Arch::kX64, BaseInst(Inst::kIdVpgatherdd, InstOptions::kNone, k1), xmm1, ptr(rax, xmm2));
EXPECT_FALSE(f.x86().has_avx2());
EXPECT_TRUE(f.x86().has_avx512_f());
}
}
#endif // !ASMJIT_NO_INTROSPECTION
#ifndef ASMJIT_NO_INTROSPECTION
template<typename... Args>
static Error query_rw_info_inline(InstRWInfo* out, Arch arch, BaseInst inst, Args&&... args) {
Operand_ op_array[] = { std::forward<Args>(args)... };
return InstInternal::query_rw_info(arch, inst, op_array, sizeof...(args), out);
}
UNIT(x86_inst_api_rm_features) {
INFO("Verifying whether RM/feature is reported correctly for PEXTRW instruction");
{
InstRWInfo rwi;
query_rw_info_inline(&rwi, Arch::kX64, BaseInst(Inst::kIdPextrw), eax, mm1, imm(1));
EXPECT_EQ(rwi.rm_feature(), 0u);
query_rw_info_inline(&rwi, Arch::kX64, BaseInst(Inst::kIdPextrw), eax, xmm1, imm(1));
EXPECT_EQ(rwi.rm_feature(), CpuFeatures::X86::kSSE4_1);
}
INFO("Verifying whether RM/feature is reported correctly for AVX512 shift instructions");
{
InstRWInfo rwi;
query_rw_info_inline(&rwi, Arch::kX64, BaseInst(Inst::kIdVpslld), xmm1, xmm2, imm(8));
EXPECT_EQ(rwi.rm_feature(), CpuFeatures::X86::kAVX512_F);
query_rw_info_inline(&rwi, Arch::kX64, BaseInst(Inst::kIdVpsllq), ymm1, ymm2, imm(8));
EXPECT_EQ(rwi.rm_feature(), CpuFeatures::X86::kAVX512_F);
query_rw_info_inline(&rwi, Arch::kX64, BaseInst(Inst::kIdVpsrad), xmm1, xmm2, imm(8));
EXPECT_EQ(rwi.rm_feature(), CpuFeatures::X86::kAVX512_F);
query_rw_info_inline(&rwi, Arch::kX64, BaseInst(Inst::kIdVpsrld), ymm1, ymm2, imm(8));
EXPECT_EQ(rwi.rm_feature(), CpuFeatures::X86::kAVX512_F);
query_rw_info_inline(&rwi, Arch::kX64, BaseInst(Inst::kIdVpsrlq), xmm1, xmm2, imm(8));
EXPECT_EQ(rwi.rm_feature(), CpuFeatures::X86::kAVX512_F);
query_rw_info_inline(&rwi, Arch::kX64, BaseInst(Inst::kIdVpslldq), xmm1, xmm2, imm(8));
EXPECT_EQ(rwi.rm_feature(), CpuFeatures::X86::kAVX512_BW);
query_rw_info_inline(&rwi, Arch::kX64, BaseInst(Inst::kIdVpsllw), ymm1, ymm2, imm(8));
EXPECT_EQ(rwi.rm_feature(), CpuFeatures::X86::kAVX512_BW);
query_rw_info_inline(&rwi, Arch::kX64, BaseInst(Inst::kIdVpsraw), xmm1, xmm2, imm(8));
EXPECT_EQ(rwi.rm_feature(), CpuFeatures::X86::kAVX512_BW);
query_rw_info_inline(&rwi, Arch::kX64, BaseInst(Inst::kIdVpsrldq), ymm1, ymm2, imm(8));
EXPECT_EQ(rwi.rm_feature(), CpuFeatures::X86::kAVX512_BW);
query_rw_info_inline(&rwi, Arch::kX64, BaseInst(Inst::kIdVpsrlw), xmm1, xmm2, imm(8));
EXPECT_EQ(rwi.rm_feature(), CpuFeatures::X86::kAVX512_BW);
query_rw_info_inline(&rwi, Arch::kX64, BaseInst(Inst::kIdVpslld), xmm1, xmm2, xmm3);
EXPECT_EQ(rwi.rm_feature(), 0u);
query_rw_info_inline(&rwi, Arch::kX64, BaseInst(Inst::kIdVpsllw), xmm1, xmm2, xmm3);
EXPECT_EQ(rwi.rm_feature(), 0u);
}
}
#endif // !ASMJIT_NO_INTROSPECTION
#endif // ASMJIT_TEST
ASMJIT_END_SUB_NAMESPACE
#endif // !ASMJIT_NO_X86