maxjhandsome
/
llvm-project
forked from OSchip/llvm-project
1
0
Fork 0
Code Issues Pull Requests Packages Releases Wiki Activity
llvm-project/llvm/lib/Target/SystemZ/SystemZAsmPrinter.cpp

528 lines
17 KiB
C++
Raw Normal View History

[SystemZ] Add back end This adds the actual lib/Target/SystemZ target files necessary to implement the SystemZ target. Note that at this point, the target cannot yet be built since the configure bits are missing. Those will be provided shortly by a follow-on patch. This version of the patch incorporates feedback from reviews by Chris Lattner and Anton Korobeynikov. Thanks to all reviewers! Patch by Richard Sandiford. llvm-svn: 181203
2013-05-07 00:15:19 +08:00
//===-- SystemZAsmPrinter.cpp - SystemZ LLVM assembly printer -------------===//
//
// The LLVM Compiler Infrastructure
//
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
//
// Streams SystemZ assembly language and associated data, in the form of
// MCInsts and MCExprs respectively.
//
//===----------------------------------------------------------------------===//
#include "SystemZAsmPrinter.h"
#include "InstPrinter/SystemZInstPrinter.h"
#include "SystemZConstantPoolValue.h"
#include "SystemZMCInstLower.h"
#include "llvm/CodeGen/MachineModuleInfoImpls.h"
#include "llvm/CodeGen/TargetLoweringObjectFileImpl.h"
Move the llvm mangler to lib/IR. This makes it available to tools that don't link with target (like llvm-ar). llvm-svn: 198708
2014-01-08 05:19:40 +08:00
#include "llvm/IR/Mangler.h"
[SystemZ] Add back end This adds the actual lib/Target/SystemZ target files necessary to implement the SystemZ target. Note that at this point, the target cannot yet be built since the configure bits are missing. Those will be provided shortly by a follow-on patch. This version of the patch incorporates feedback from reviews by Chris Lattner and Anton Korobeynikov. Thanks to all reviewers! Patch by Richard Sandiford. llvm-svn: 181203
2013-05-07 00:15:19 +08:00
#include "llvm/MC/MCExpr.h"
[SystemZ] Define the return instruction as a pseudo alias of BR This is the first of a few patches to reduce the dupliation of encoding information. The return instruction is a normal BR in which one of the registers is fixed. llvm-svn: 191364
2013-09-25 18:20:08 +08:00
#include "llvm/MC/MCInstBuilder.h"
[SystemZ] Add back end This adds the actual lib/Target/SystemZ target files necessary to implement the SystemZ target. Note that at this point, the target cannot yet be built since the configure bits are missing. Those will be provided shortly by a follow-on patch. This version of the patch incorporates feedback from reviews by Chris Lattner and Anton Korobeynikov. Thanks to all reviewers! Patch by Richard Sandiford. llvm-svn: 181203
2013-05-07 00:15:19 +08:00
#include "llvm/MC/MCStreamer.h"
#include "llvm/Support/TargetRegistry.h"
using namespace llvm;
[SystemZ] Define the GR64 low-word logic instructions as pseudo aliases. Another patch to avoid duplication of encoding information. Things like NILF, NILL and NILH are used as both 32-bit and 64-bit instructions. Here the 64-bit versions are defined as aliases of the 32-bit ones. llvm-svn: 191369
2013-09-25 19:11:53 +08:00
// Return an RI instruction like MI with opcode Opcode, but with the
// GR64 register operands turned into GR32s.
static MCInst lowerRILow(const MachineInstr *MI, unsigned Opcode) {
[SystemZ] Fix TMHH and TMHL usage for z10 with -O0 I've no idea why I decided to handle TMxx differently from all the other high/low logic operations, but it was a stupid thing to do. The high registers aren't available as separate 32-bit registers on z10, so subreg_h32 can't be used on a GR64 there. I've normally been testing with z196 and with -O3 and so hadn't noticed this until now. llvm-svn: 195473
2013-11-23 01:28:28 +08:00
if (MI->isCompare())
return MCInstBuilder(Opcode)
.addReg(SystemZMC::getRegAsGR32(MI->getOperand(0).getReg()))
.addImm(MI->getOperand(1).getImm());
else
return MCInstBuilder(Opcode)
.addReg(SystemZMC::getRegAsGR32(MI->getOperand(0).getReg()))
.addReg(SystemZMC::getRegAsGR32(MI->getOperand(1).getReg()))
.addImm(MI->getOperand(2).getImm());
[SystemZ] Define the GR64 low-word logic instructions as pseudo aliases. Another patch to avoid duplication of encoding information. Things like NILF, NILL and NILH are used as both 32-bit and 64-bit instructions. Here the 64-bit versions are defined as aliases of the 32-bit ones. llvm-svn: 191369
2013-09-25 19:11:53 +08:00
}
[SystemZ] Allow integer insertions with a high-word destination llvm-svn: 191753
2013-10-01 21:18:56 +08:00
// Return an RI instruction like MI with opcode Opcode, but with the
// GR64 register operands turned into GRH32s.
static MCInst lowerRIHigh(const MachineInstr *MI, unsigned Opcode) {
[SystemZ] Fix TMHH and TMHL usage for z10 with -O0 I've no idea why I decided to handle TMxx differently from all the other high/low logic operations, but it was a stupid thing to do. The high registers aren't available as separate 32-bit registers on z10, so subreg_h32 can't be used on a GR64 there. I've normally been testing with z196 and with -O3 and so hadn't noticed this until now. llvm-svn: 195473
2013-11-23 01:28:28 +08:00
if (MI->isCompare())
return MCInstBuilder(Opcode)
.addReg(SystemZMC::getRegAsGRH32(MI->getOperand(0).getReg()))
.addImm(MI->getOperand(1).getImm());
else
return MCInstBuilder(Opcode)
.addReg(SystemZMC::getRegAsGRH32(MI->getOperand(0).getReg()))
.addReg(SystemZMC::getRegAsGRH32(MI->getOperand(1).getReg()))
.addImm(MI->getOperand(2).getImm());
[SystemZ] Allow integer insertions with a high-word destination llvm-svn: 191753
2013-10-01 21:18:56 +08:00
}
[SystemZ] Use upper words of GR64s for codegen This just adds the basics necessary for allocating the upper words to virtual registers (move, load and store). The move support is parameterised in a way that makes it easy to handle zero extensions, but the associated zero-extend patterns are added by a later patch. The easiest way of testing this seemed to be add a new "h" register constraint for high words. I don't expect the constraint to be useful in real inline asms, but it should work, so I didn't try to hide it behind an option. llvm-svn: 191739
2013-10-01 19:26:28 +08:00
// Return an RI instruction like MI with opcode Opcode, but with the
// R2 register turned into a GR64.
static MCInst lowerRIEfLow(const MachineInstr *MI, unsigned Opcode) {
return MCInstBuilder(Opcode)
.addReg(MI->getOperand(0).getReg())
.addReg(MI->getOperand(1).getReg())
.addReg(SystemZMC::getRegAsGR64(MI->getOperand(2).getReg()))
.addImm(MI->getOperand(3).getImm())
.addImm(MI->getOperand(4).getImm())
.addImm(MI->getOperand(5).getImm());
}
[SystemZ] Support all TLS access models - CodeGen part The current SystemZ back-end only supports the local-exec TLS access model. This patch adds all required CodeGen support for the other TLS models, which means in particular: - Expand initial-exec TLS accesses by loading TLS offsets from the GOT using @indntpoff relocations. - Expand general-dynamic and local-dynamic accesses by generating the appropriate calls to __tls_get_offset. Note that this routine has a non-standard ABI and requires loading the GOT pointer into %r12, so the patch also adds support for the GLOBAL_OFFSET_TABLE ISD node. - Add a new platform-specific optimization pass to remove redundant __tls_get_offset calls in the local-dynamic model (modeled after the corresponding X86 pass). - Add test cases verifying all access models and optimizations. llvm-svn: 229654
2015-02-18 17:13:27 +08:00
static const MCSymbolRefExpr *getTLSGetOffset(MCContext &Context) {
StringRef Name = "__tls_get_offset";
MC: Clean up MCExpr naming. NFC. llvm-svn: 238634
2015-05-30 09:25:56 +08:00
return MCSymbolRefExpr::create(Context.getOrCreateSymbol(Name),
[SystemZ] Support all TLS access models - CodeGen part The current SystemZ back-end only supports the local-exec TLS access model. This patch adds all required CodeGen support for the other TLS models, which means in particular: - Expand initial-exec TLS accesses by loading TLS offsets from the GOT using @indntpoff relocations. - Expand general-dynamic and local-dynamic accesses by generating the appropriate calls to __tls_get_offset. Note that this routine has a non-standard ABI and requires loading the GOT pointer into %r12, so the patch also adds support for the GLOBAL_OFFSET_TABLE ISD node. - Add a new platform-specific optimization pass to remove redundant __tls_get_offset calls in the local-dynamic model (modeled after the corresponding X86 pass). - Add test cases verifying all access models and optimizations. llvm-svn: 229654
2015-02-18 17:13:27 +08:00
MCSymbolRefExpr::VK_PLT,
Context);
}
static const MCSymbolRefExpr *getGlobalOffsetTable(MCContext &Context) {
StringRef Name = "_GLOBAL_OFFSET_TABLE_";
MC: Clean up MCExpr naming. NFC. llvm-svn: 238634
2015-05-30 09:25:56 +08:00
return MCSymbolRefExpr::create(Context.getOrCreateSymbol(Name),
[SystemZ] Support all TLS access models - CodeGen part The current SystemZ back-end only supports the local-exec TLS access model. This patch adds all required CodeGen support for the other TLS models, which means in particular: - Expand initial-exec TLS accesses by loading TLS offsets from the GOT using @indntpoff relocations. - Expand general-dynamic and local-dynamic accesses by generating the appropriate calls to __tls_get_offset. Note that this routine has a non-standard ABI and requires loading the GOT pointer into %r12, so the patch also adds support for the GLOBAL_OFFSET_TABLE ISD node. - Add a new platform-specific optimization pass to remove redundant __tls_get_offset calls in the local-dynamic model (modeled after the corresponding X86 pass). - Add test cases verifying all access models and optimizations. llvm-svn: 229654
2015-02-18 17:13:27 +08:00
MCSymbolRefExpr::VK_None,
Context);
}
[SystemZ] Add CodeGen support for scalar f64 ops in vector registers The z13 vector facility includes some instructions that operate only on the high f64 in a v2f64, effectively extending the FP register set from 16 to 32 registers. It's still better to use the old instructions if the operands happen to fit though, since the older instructions have a shorter encoding. Based on a patch by Richard Sandiford. llvm-svn: 236524
2015-05-06 03:28:34 +08:00
// MI loads the high part of a vector from memory. Return an instruction
// that uses replicating vector load Opcode to do the same thing.
static MCInst lowerSubvectorLoad(const MachineInstr *MI, unsigned Opcode) {
return MCInstBuilder(Opcode)
.addReg(SystemZMC::getRegAsVR128(MI->getOperand(0).getReg()))
.addReg(MI->getOperand(1).getReg())
.addImm(MI->getOperand(2).getImm())
.addReg(MI->getOperand(3).getReg());
}
// MI stores the high part of a vector to memory. Return an instruction
// that uses elemental vector store Opcode to do the same thing.
static MCInst lowerSubvectorStore(const MachineInstr *MI, unsigned Opcode) {
return MCInstBuilder(Opcode)
.addReg(SystemZMC::getRegAsVR128(MI->getOperand(0).getReg()))
.addReg(MI->getOperand(1).getReg())
.addImm(MI->getOperand(2).getImm())
.addReg(MI->getOperand(3).getReg())
.addImm(0);
}
[SystemZ] Add back end This adds the actual lib/Target/SystemZ target files necessary to implement the SystemZ target. Note that at this point, the target cannot yet be built since the configure bits are missing. Those will be provided shortly by a follow-on patch. This version of the patch incorporates feedback from reviews by Chris Lattner and Anton Korobeynikov. Thanks to all reviewers! Patch by Richard Sandiford. llvm-svn: 181203
2013-05-07 00:15:19 +08:00
void SystemZAsmPrinter::EmitInstruction(const MachineInstr *MI) {
The AsmPrinter has a Mangler. Use it. llvm-svn: 193617
2013-10-30 00:18:15 +08:00
SystemZMCInstLower Lower(MF->getContext(), *this);
[SystemZ] Add back end This adds the actual lib/Target/SystemZ target files necessary to implement the SystemZ target. Note that at this point, the target cannot yet be built since the configure bits are missing. Those will be provided shortly by a follow-on patch. This version of the patch incorporates feedback from reviews by Chris Lattner and Anton Korobeynikov. Thanks to all reviewers! Patch by Richard Sandiford. llvm-svn: 181203
2013-05-07 00:15:19 +08:00
MCInst LoweredMI;
[SystemZ] Define the return instruction as a pseudo alias of BR This is the first of a few patches to reduce the dupliation of encoding information. The return instruction is a normal BR in which one of the registers is fixed. llvm-svn: 191364
2013-09-25 18:20:08 +08:00
switch (MI->getOpcode()) {
case SystemZ::Return:
LoweredMI = MCInstBuilder(SystemZ::BR).addReg(SystemZ::R14D);
break;
[SystemZ] Implement conditional returns Return is now considered a predicable instruction, and is converted to a newly-added CondReturn (which maps to BCR to %r14) instruction by the if conversion pass. Also, fused compare-and-branch transform knows about conditional returns, emitting the proper fused instructions for them. This transform triggers on a *lot* of tests, hence the huge diffstat. The changes are mostly jX to br %r14 -> bXr %r14. Author: koriakin Differential Revision: http://reviews.llvm.org/D17339 llvm-svn: 265689
2016-04-08 00:11:44 +08:00
case SystemZ::CondReturn:
LoweredMI = MCInstBuilder(SystemZ::BCR)
.addImm(MI->getOperand(0).getImm())
.addImm(MI->getOperand(1).getImm())
.addReg(SystemZ::R14D);
break;
case SystemZ::CRBReturn:
LoweredMI = MCInstBuilder(SystemZ::CRB)
.addReg(MI->getOperand(0).getReg())
.addReg(MI->getOperand(1).getReg())
.addImm(MI->getOperand(2).getImm())
.addReg(SystemZ::R14D)
.addImm(0);
break;
case SystemZ::CGRBReturn:
LoweredMI = MCInstBuilder(SystemZ::CGRB)
.addReg(MI->getOperand(0).getReg())
.addReg(MI->getOperand(1).getReg())
.addImm(MI->getOperand(2).getImm())
.addReg(SystemZ::R14D)
.addImm(0);
break;
case SystemZ::CIBReturn:
LoweredMI = MCInstBuilder(SystemZ::CIB)
.addReg(MI->getOperand(0).getReg())
.addImm(MI->getOperand(1).getImm())
.addImm(MI->getOperand(2).getImm())
.addReg(SystemZ::R14D)
.addImm(0);
break;
case SystemZ::CGIBReturn:
LoweredMI = MCInstBuilder(SystemZ::CGIB)
.addReg(MI->getOperand(0).getReg())
.addImm(MI->getOperand(1).getImm())
.addImm(MI->getOperand(2).getImm())
.addReg(SystemZ::R14D)
.addImm(0);
break;
case SystemZ::CLRBReturn:
LoweredMI = MCInstBuilder(SystemZ::CLRB)
.addReg(MI->getOperand(0).getReg())
.addReg(MI->getOperand(1).getReg())
.addImm(MI->getOperand(2).getImm())
.addReg(SystemZ::R14D)
.addImm(0);
break;
case SystemZ::CLGRBReturn:
LoweredMI = MCInstBuilder(SystemZ::CLGRB)
.addReg(MI->getOperand(0).getReg())
.addReg(MI->getOperand(1).getReg())
.addImm(MI->getOperand(2).getImm())
.addReg(SystemZ::R14D)
.addImm(0);
break;
case SystemZ::CLIBReturn:
LoweredMI = MCInstBuilder(SystemZ::CLIB)
.addReg(MI->getOperand(0).getReg())
.addImm(MI->getOperand(1).getImm())
.addImm(MI->getOperand(2).getImm())
.addReg(SystemZ::R14D)
.addImm(0);
break;
case SystemZ::CLGIBReturn:
LoweredMI = MCInstBuilder(SystemZ::CLGIB)
.addReg(MI->getOperand(0).getReg())
.addImm(MI->getOperand(1).getImm())
.addImm(MI->getOperand(2).getImm())
.addReg(SystemZ::R14D)
.addImm(0);
break;
[SystemZ] Define the call instructions as pseudo aliases. Similar to r191364, but for calls. This patch also removes the shortening of BRASL to BRAS within a TU. Doing that was a bit controversial internally, since there's a strong expectation with the z assembler that WYWIWYG. llvm-svn: 191366
2013-09-25 18:37:17 +08:00
case SystemZ::CallBRASL:
LoweredMI = MCInstBuilder(SystemZ::BRASL)
.addReg(SystemZ::R14D)
.addExpr(Lower.getExpr(MI->getOperand(0), MCSymbolRefExpr::VK_PLT));
break;
case SystemZ::CallBASR:
LoweredMI = MCInstBuilder(SystemZ::BASR)
.addReg(SystemZ::R14D)
.addReg(MI->getOperand(0).getReg());
break;
case SystemZ::CallJG:
LoweredMI = MCInstBuilder(SystemZ::JG)
.addExpr(Lower.getExpr(MI->getOperand(0), MCSymbolRefExpr::VK_PLT));
break;
[SystemZ] Support conditional sibling calls via BRCL This adds a conditional variant of CallJG instruction, CallBRCL. It can be used for conditional sibling calls. Unfortunately, due to IfCvt limitations, it only really works well for functions without arguments. Author: koriakin Differential Revision: http://reviews.llvm.org/D18864 llvm-svn: 265814
2016-04-09 01:22:19 +08:00
case SystemZ::CallBRCL:
LoweredMI = MCInstBuilder(SystemZ::BRCL)
.addImm(MI->getOperand(0).getImm())
.addImm(MI->getOperand(1).getImm())
.addExpr(Lower.getExpr(MI->getOperand(2), MCSymbolRefExpr::VK_PLT));
break;
[SystemZ] Define the call instructions as pseudo aliases. Similar to r191364, but for calls. This patch also removes the shortening of BRASL to BRAS within a TU. Doing that was a bit controversial internally, since there's a strong expectation with the z assembler that WYWIWYG. llvm-svn: 191366
2013-09-25 18:37:17 +08:00
case SystemZ::CallBR:
LoweredMI = MCInstBuilder(SystemZ::BR).addReg(SystemZ::R1D);
break;
[SystemZ] Support conditional indirect sibling calls via BCR This adds a conditional variant of CallBR instruction, CallBCR. Also, it can be fused with integer comparisons, resulting in one of the new C*BCall instructions. In addition to CallBRCL limitations, this has another one: it won't trigger if the function to call isn't already in %r1 - see f22 in the test for an example (it's also why the loads in tests are volatile). Author: koriakin Differential Revision: http://reviews.llvm.org/D18928 llvm-svn: 265933
2016-04-11 20:12:32 +08:00
case SystemZ::CallBCR:
LoweredMI = MCInstBuilder(SystemZ::BCR)
.addImm(MI->getOperand(0).getImm())
.addImm(MI->getOperand(1).getImm())
.addReg(SystemZ::R1D);
break;
case SystemZ::CRBCall:
LoweredMI = MCInstBuilder(SystemZ::CRB)
.addReg(MI->getOperand(0).getReg())
.addReg(MI->getOperand(1).getReg())
.addImm(MI->getOperand(2).getImm())
.addReg(SystemZ::R1D)
.addImm(0);
break;
case SystemZ::CGRBCall:
LoweredMI = MCInstBuilder(SystemZ::CGRB)
.addReg(MI->getOperand(0).getReg())
.addReg(MI->getOperand(1).getReg())
.addImm(MI->getOperand(2).getImm())
.addReg(SystemZ::R1D)
.addImm(0);
break;
case SystemZ::CIBCall:
LoweredMI = MCInstBuilder(SystemZ::CIB)
.addReg(MI->getOperand(0).getReg())
.addImm(MI->getOperand(1).getImm())
.addImm(MI->getOperand(2).getImm())
.addReg(SystemZ::R1D)
.addImm(0);
break;
case SystemZ::CGIBCall:
LoweredMI = MCInstBuilder(SystemZ::CGIB)
.addReg(MI->getOperand(0).getReg())
.addImm(MI->getOperand(1).getImm())
.addImm(MI->getOperand(2).getImm())
.addReg(SystemZ::R1D)
.addImm(0);
break;
case SystemZ::CLRBCall:
LoweredMI = MCInstBuilder(SystemZ::CLRB)
.addReg(MI->getOperand(0).getReg())
.addReg(MI->getOperand(1).getReg())
.addImm(MI->getOperand(2).getImm())
.addReg(SystemZ::R1D)
.addImm(0);
break;
case SystemZ::CLGRBCall:
LoweredMI = MCInstBuilder(SystemZ::CLGRB)
.addReg(MI->getOperand(0).getReg())
.addReg(MI->getOperand(1).getReg())
.addImm(MI->getOperand(2).getImm())
.addReg(SystemZ::R1D)
.addImm(0);
break;
case SystemZ::CLIBCall:
LoweredMI = MCInstBuilder(SystemZ::CLIB)
.addReg(MI->getOperand(0).getReg())
.addImm(MI->getOperand(1).getImm())
.addImm(MI->getOperand(2).getImm())
.addReg(SystemZ::R1D)
.addImm(0);
break;
case SystemZ::CLGIBCall:
LoweredMI = MCInstBuilder(SystemZ::CLGIB)
.addReg(MI->getOperand(0).getReg())
.addImm(MI->getOperand(1).getImm())
.addImm(MI->getOperand(2).getImm())
.addReg(SystemZ::R1D)
.addImm(0);
break;
[SystemZ] Support all TLS access models - CodeGen part The current SystemZ back-end only supports the local-exec TLS access model. This patch adds all required CodeGen support for the other TLS models, which means in particular: - Expand initial-exec TLS accesses by loading TLS offsets from the GOT using @indntpoff relocations. - Expand general-dynamic and local-dynamic accesses by generating the appropriate calls to __tls_get_offset. Note that this routine has a non-standard ABI and requires loading the GOT pointer into %r12, so the patch also adds support for the GLOBAL_OFFSET_TABLE ISD node. - Add a new platform-specific optimization pass to remove redundant __tls_get_offset calls in the local-dynamic model (modeled after the corresponding X86 pass). - Add test cases verifying all access models and optimizations. llvm-svn: 229654
2015-02-18 17:13:27 +08:00
case SystemZ::TLS_GDCALL:
LoweredMI = MCInstBuilder(SystemZ::BRASL)
.addReg(SystemZ::R14D)
.addExpr(getTLSGetOffset(MF->getContext()))
.addExpr(Lower.getExpr(MI->getOperand(0), MCSymbolRefExpr::VK_TLSGD));
break;
case SystemZ::TLS_LDCALL:
LoweredMI = MCInstBuilder(SystemZ::BRASL)
.addReg(SystemZ::R14D)
.addExpr(getTLSGetOffset(MF->getContext()))
.addExpr(Lower.getExpr(MI->getOperand(0), MCSymbolRefExpr::VK_TLSLDM));
break;
case SystemZ::GOT:
LoweredMI = MCInstBuilder(SystemZ::LARL)
.addReg(MI->getOperand(0).getReg())
.addExpr(getGlobalOffsetTable(MF->getContext()));
break;
[SystemZ] Define the GR64 low-word logic instructions as pseudo aliases. Another patch to avoid duplication of encoding information. Things like NILF, NILL and NILH are used as both 32-bit and 64-bit instructions. Here the 64-bit versions are defined as aliases of the 32-bit ones. llvm-svn: 191369
2013-09-25 19:11:53 +08:00
case SystemZ::IILF64:
LoweredMI = MCInstBuilder(SystemZ::IILF)
.addReg(SystemZMC::getRegAsGR32(MI->getOperand(0).getReg()))
.addImm(MI->getOperand(2).getImm());
break;
[SystemZ] Add patterns to load a constant into a high word (IIHF) Similar to low words, we can use the shorter LLIHL and LLIHH if it turns out that the other half of the GR64 isn't live. llvm-svn: 191750
2013-10-01 21:02:28 +08:00
case SystemZ::IIHF64:
LoweredMI = MCInstBuilder(SystemZ::IIHF)
.addReg(SystemZMC::getRegAsGRH32(MI->getOperand(0).getReg()))
.addImm(MI->getOperand(2).getImm());
break;
[SystemZ] Use upper words of GR64s for codegen This just adds the basics necessary for allocating the upper words to virtual registers (move, load and store). The move support is parameterised in a way that makes it easy to handle zero extensions, but the associated zero-extend patterns are added by a later patch. The easiest way of testing this seemed to be add a new "h" register constraint for high words. I don't expect the constraint to be useful in real inline asms, but it should work, so I didn't try to hide it behind an option. llvm-svn: 191739
2013-10-01 19:26:28 +08:00
case SystemZ::RISBHH:
case SystemZ::RISBHL:
LoweredMI = lowerRIEfLow(MI, SystemZ::RISBHG);
break;
case SystemZ::RISBLH:
case SystemZ::RISBLL:
LoweredMI = lowerRIEfLow(MI, SystemZ::RISBLG);
break;
[SystemZ] Add CodeGen support for integer vector types This the first of a series of patches to add CodeGen support exploiting the instructions of the z13 vector facility. This patch adds support for the native integer vector types (v16i8, v8i16, v4i32, v2i64). When the vector facility is present, we default to the new vector ABI. This is characterized by two major differences: - Vector types are passed/returned in vector registers (except for unnamed arguments of a variable-argument list function). - Vector types are at most 8-byte aligned. The reason for the choice of 8-byte vector alignment is that the hardware is able to efficiently load vectors at 8-byte alignment, and the ABI only guarantees 8-byte alignment of the stack pointer, so requiring any higher alignment for vectors would require dynamic stack re-alignment code. However, for compatibility with old code that may use vector types, when *not* using the vector facility, the old alignment rules (vector types are naturally aligned) remain in use. These alignment rules are not only implemented at the C language level (implemented in clang), but also at the LLVM IR level. This is done by selecting a different DataLayout string depending on whether the vector ABI is in effect or not. Based on a patch by Richard Sandiford. llvm-svn: 236521
2015-05-06 03:25:42 +08:00
case SystemZ::VLVGP32:
LoweredMI = MCInstBuilder(SystemZ::VLVGP)
.addReg(MI->getOperand(0).getReg())
.addReg(SystemZMC::getRegAsGR64(MI->getOperand(1).getReg()))
.addReg(SystemZMC::getRegAsGR64(MI->getOperand(2).getReg()));
break;
[SystemZ] Add CodeGen support for scalar f64 ops in vector registers The z13 vector facility includes some instructions that operate only on the high f64 in a v2f64, effectively extending the FP register set from 16 to 32 registers. It's still better to use the old instructions if the operands happen to fit though, since the older instructions have a shorter encoding. Based on a patch by Richard Sandiford. llvm-svn: 236524
2015-05-06 03:28:34 +08:00
case SystemZ::VLR32:
case SystemZ::VLR64:
LoweredMI = MCInstBuilder(SystemZ::VLR)
.addReg(SystemZMC::getRegAsVR128(MI->getOperand(0).getReg()))
.addReg(SystemZMC::getRegAsVR128(MI->getOperand(1).getReg()));
break;
case SystemZ::VL32:
LoweredMI = lowerSubvectorLoad(MI, SystemZ::VLREPF);
break;
case SystemZ::VL64:
LoweredMI = lowerSubvectorLoad(MI, SystemZ::VLREPG);
break;
case SystemZ::VST32:
LoweredMI = lowerSubvectorStore(MI, SystemZ::VSTEF);
break;
case SystemZ::VST64:
LoweredMI = lowerSubvectorStore(MI, SystemZ::VSTEG);
break;
[SystemZ] Add CodeGen support for v4f32 The architecture doesn't really have any native v4f32 operations except v4f32->v2f64 and v2f64->v4f32 conversions, with only half of the v4f32 elements being used. Even so, using vector registers for <4 x float> and scalarising individual operations is much better than generating completely scalar code, since there's much less register pressure. It's also more efficient to do v4f32 comparisons by extending to 2 v2f64s, comparing those, then packing the result. This particularly helps with llvmpipe. Based on a patch by Richard Sandiford. llvm-svn: 236523
2015-05-06 03:27:45 +08:00
case SystemZ::LFER:
LoweredMI = MCInstBuilder(SystemZ::VLGVF)
.addReg(SystemZMC::getRegAsGR64(MI->getOperand(0).getReg()))
.addReg(SystemZMC::getRegAsVR128(MI->getOperand(1).getReg()))
.addReg(0).addImm(0);
break;
case SystemZ::LEFR:
LoweredMI = MCInstBuilder(SystemZ::VLVGF)
.addReg(SystemZMC::getRegAsVR128(MI->getOperand(0).getReg()))
.addReg(SystemZMC::getRegAsVR128(MI->getOperand(0).getReg()))
.addReg(MI->getOperand(1).getReg())
.addReg(0).addImm(0);
break;
[SystemZ] Define the GR64 low-word logic instructions as pseudo aliases. Another patch to avoid duplication of encoding information. Things like NILF, NILL and NILH are used as both 32-bit and 64-bit instructions. Here the 64-bit versions are defined as aliases of the 32-bit ones. llvm-svn: 191369
2013-09-25 19:11:53 +08:00
#define LOWER_LOW(NAME) \
case SystemZ::NAME##64: LoweredMI = lowerRILow(MI, SystemZ::NAME); break
LOWER_LOW(IILL);
LOWER_LOW(IILH);
[SystemZ] Fix TMHH and TMHL usage for z10 with -O0 I've no idea why I decided to handle TMxx differently from all the other high/low logic operations, but it was a stupid thing to do. The high registers aren't available as separate 32-bit registers on z10, so subreg_h32 can't be used on a GR64 there. I've normally been testing with z196 and with -O3 and so hadn't noticed this until now. llvm-svn: 195473
2013-11-23 01:28:28 +08:00
LOWER_LOW(TMLL);
LOWER_LOW(TMLH);
[SystemZ] Define the GR64 low-word logic instructions as pseudo aliases. Another patch to avoid duplication of encoding information. Things like NILF, NILL and NILH are used as both 32-bit and 64-bit instructions. Here the 64-bit versions are defined as aliases of the 32-bit ones. llvm-svn: 191369
2013-09-25 19:11:53 +08:00
LOWER_LOW(NILL);
LOWER_LOW(NILH);
LOWER_LOW(NILF);
LOWER_LOW(OILL);
LOWER_LOW(OILH);
LOWER_LOW(OILF);
LOWER_LOW(XILF);
#undef LOWER_LOW
[SystemZ] Allow integer insertions with a high-word destination llvm-svn: 191753
2013-10-01 21:18:56 +08:00
#define LOWER_HIGH(NAME) \
case SystemZ::NAME##64: LoweredMI = lowerRIHigh(MI, SystemZ::NAME); break
LOWER_HIGH(IIHL);
LOWER_HIGH(IIHH);
[SystemZ] Fix TMHH and TMHL usage for z10 with -O0 I've no idea why I decided to handle TMxx differently from all the other high/low logic operations, but it was a stupid thing to do. The high registers aren't available as separate 32-bit registers on z10, so subreg_h32 can't be used on a GR64 there. I've normally been testing with z196 and with -O3 and so hadn't noticed this until now. llvm-svn: 195473
2013-11-23 01:28:28 +08:00
LOWER_HIGH(TMHL);
LOWER_HIGH(TMHH);
[SystemZ] Allow integer AND involving high words llvm-svn: 191762
2013-10-01 22:20:41 +08:00
LOWER_HIGH(NIHL);
LOWER_HIGH(NIHH);
LOWER_HIGH(NIHF);
[SystemZ] Allow integer OR involving high words llvm-svn: 191755
2013-10-01 21:22:41 +08:00
LOWER_HIGH(OIHL);
LOWER_HIGH(OIHH);
LOWER_HIGH(OIHF);
[SystemZ] Allow integer XOR involving high words llvm-svn: 191759
2013-10-01 22:08:44 +08:00
LOWER_HIGH(XIHF);
[SystemZ] Allow integer insertions with a high-word destination llvm-svn: 191753
2013-10-01 21:18:56 +08:00
#undef LOWER_HIGH
Add TargetLowering::prepareVolatileOrAtomicLoad One unusual feature of the z architecture is that the result of a previous load can be reused indefinitely for subsequent loads, even if a cache-coherent store to that location is performed by another CPU. A special serializing instruction must be used if you want to force a load to be reattempted. Since volatile loads are not supposed to be omitted in this way, we should insert a serializing instruction before each such load. The same goes for atomic loads. The patch implements this at the IR->DAG boundary, in a similar way to atomic fences. It is a no-op for targets other than SystemZ. llvm-svn: 196905
2013-12-10 18:36:34 +08:00
case SystemZ::Serialize:
Remove the local subtarget variable from the SystemZ asm printer and update the two calls accordingly. llvm-svn: 229805
2015-02-19 09:26:28 +08:00
if (MF->getSubtarget<SystemZSubtarget>().hasFastSerialization())
[SystemZ] Refactor branch and conditional instruction patterns Rework patterns for branches, call & return instructions, compare-and-branch, compare-and-trap, and conditional move instructions. In particular, simplify creation of patterns for the extended opcodes of instructions that take a CC mask. Also, use semantical instruction classes for all the instructions instead of open-coding them in SystemZInstrInfo.td. Adds a couple of the basic branch instructions (that are unused for codegen) for the assembler/disassembler. llvm-svn: 286263
2016-11-09 02:30:50 +08:00
LoweredMI = MCInstBuilder(SystemZ::BCRAsm)
Add TargetLowering::prepareVolatileOrAtomicLoad One unusual feature of the z architecture is that the result of a previous load can be reused indefinitely for subsequent loads, even if a cache-coherent store to that location is performed by another CPU. A special serializing instruction must be used if you want to force a load to be reattempted. Since volatile loads are not supposed to be omitted in this way, we should insert a serializing instruction before each such load. The same goes for atomic loads. The patch implements this at the IR->DAG boundary, in a similar way to atomic fences. It is a no-op for targets other than SystemZ. llvm-svn: 196905
2013-12-10 18:36:34 +08:00
.addImm(14).addReg(SystemZ::R0D);
else
[SystemZ] Refactor branch and conditional instruction patterns Rework patterns for branches, call & return instructions, compare-and-branch, compare-and-trap, and conditional move instructions. In particular, simplify creation of patterns for the extended opcodes of instructions that take a CC mask. Also, use semantical instruction classes for all the instructions instead of open-coding them in SystemZInstrInfo.td. Adds a couple of the basic branch instructions (that are unused for codegen) for the assembler/disassembler. llvm-svn: 286263
2016-11-09 02:30:50 +08:00
LoweredMI = MCInstBuilder(SystemZ::BCRAsm)
Add TargetLowering::prepareVolatileOrAtomicLoad One unusual feature of the z architecture is that the result of a previous load can be reused indefinitely for subsequent loads, even if a cache-coherent store to that location is performed by another CPU. A special serializing instruction must be used if you want to force a load to be reattempted. Since volatile loads are not supposed to be omitted in this way, we should insert a serializing instruction before each such load. The same goes for atomic loads. The patch implements this at the IR->DAG boundary, in a similar way to atomic fences. It is a no-op for targets other than SystemZ. llvm-svn: 196905
2013-12-10 18:36:34 +08:00
.addImm(15).addReg(SystemZ::R0D);
break;
[SystemZ] Support ATOMIC_FENCE A cross-thread sequentially consistent fence should be lowered into z/Architecture's BCR serialization instruction, instead of causing a fatal error in the back-end. Author: bryanpkc Differential Revision: http://reviews.llvm.org/D18644 llvm-svn: 265292
2016-04-04 20:45:44 +08:00
// Emit nothing here but a comment if we can.
case SystemZ::MemBarrier:
OutStreamer->emitRawComment("MEMBARRIER");
return;
[SystemZ] Support Compare and Traps Support and generate Compare and Traps like CRT, CIT, etc. Support Trap as legal DAG opcodes and generate "j .+2" for them by default. Add support for Conditional Traps and use the If Converter to convert them into the corresponding compare and trap opcodes. Differential Revision: http://reviews.llvm.org/D21155 llvm-svn: 272419
2016-06-11 03:58:10 +08:00
// We want to emit "j .+2" for traps, jumping to the relative immediate field
// of the jump instruction, which is an illegal instruction. We cannot emit a
// "." symbol, so create and emit a temp label before the instruction and use
// that instead.
case SystemZ::Trap: {
MCSymbol *DotSym = OutContext.createTempSymbol();
OutStreamer->EmitLabel(DotSym);
const MCSymbolRefExpr *Expr = MCSymbolRefExpr::create(DotSym, OutContext);
const MCConstantExpr *ConstExpr = MCConstantExpr::create(2, OutContext);
LoweredMI = MCInstBuilder(SystemZ::J)
.addExpr(MCBinaryExpr::createAdd(Expr, ConstExpr, OutContext));
}
break;
// Conditional traps will create a branch on condition instruction that jumps
// to the relative immediate field of the jump instruction. (eg. "jo .+2")
case SystemZ::CondTrap: {
MCSymbol *DotSym = OutContext.createTempSymbol();
OutStreamer->EmitLabel(DotSym);
const MCSymbolRefExpr *Expr = MCSymbolRefExpr::create(DotSym, OutContext);
const MCConstantExpr *ConstExpr = MCConstantExpr::create(2, OutContext);
LoweredMI = MCInstBuilder(SystemZ::BRC)
.addImm(MI->getOperand(0).getImm())
.addImm(MI->getOperand(1).getImm())
.addExpr(MCBinaryExpr::createAdd(Expr, ConstExpr, OutContext));
}
break;
[SystemZ] Define the return instruction as a pseudo alias of BR This is the first of a few patches to reduce the dupliation of encoding information. The return instruction is a normal BR in which one of the registers is fixed. llvm-svn: 191364
2013-09-25 18:20:08 +08:00
default:
[SystemZ] Define the call instructions as pseudo aliases. Similar to r191364, but for calls. This patch also removes the shortening of BRASL to BRAS within a TU. Doing that was a bit controversial internally, since there's a strong expectation with the z assembler that WYWIWYG. llvm-svn: 191366
2013-09-25 18:37:17 +08:00
Lower.lower(MI, LoweredMI);
[SystemZ] Define the return instruction as a pseudo alias of BR This is the first of a few patches to reduce the dupliation of encoding information. The return instruction is a normal BR in which one of the registers is fixed. llvm-svn: 191364
2013-09-25 18:20:08 +08:00
break;
}
[AsmPrinter] Make AsmPrinter's OutStreamer member a unique_ptr. AsmPrinter owns the OutStreamer, so an owning pointer makes sense here. Using a reference for this is crufty. llvm-svn: 235752
2015-04-25 03:11:51 +08:00
EmitToStreamer(*OutStreamer, LoweredMI);
[SystemZ] Add back end This adds the actual lib/Target/SystemZ target files necessary to implement the SystemZ target. Note that at this point, the target cannot yet be built since the configure bits are missing. Those will be provided shortly by a follow-on patch. This version of the patch incorporates feedback from reviews by Chris Lattner and Anton Korobeynikov. Thanks to all reviewers! Patch by Richard Sandiford. llvm-svn: 181203
2013-05-07 00:15:19 +08:00
}
// Convert a SystemZ-specific constant pool modifier into the associated
// MCSymbolRefExpr variant kind.
static MCSymbolRefExpr::VariantKind
getModifierVariantKind(SystemZCP::SystemZCPModifier Modifier) {
switch (Modifier) {
[SystemZ] Support all TLS access models - CodeGen part The current SystemZ back-end only supports the local-exec TLS access model. This patch adds all required CodeGen support for the other TLS models, which means in particular: - Expand initial-exec TLS accesses by loading TLS offsets from the GOT using @indntpoff relocations. - Expand general-dynamic and local-dynamic accesses by generating the appropriate calls to __tls_get_offset. Note that this routine has a non-standard ABI and requires loading the GOT pointer into %r12, so the patch also adds support for the GLOBAL_OFFSET_TABLE ISD node. - Add a new platform-specific optimization pass to remove redundant __tls_get_offset calls in the local-dynamic model (modeled after the corresponding X86 pass). - Add test cases verifying all access models and optimizations. llvm-svn: 229654
2015-02-18 17:13:27 +08:00
case SystemZCP::TLSGD: return MCSymbolRefExpr::VK_TLSGD;
case SystemZCP::TLSLDM: return MCSymbolRefExpr::VK_TLSLDM;
case SystemZCP::DTPOFF: return MCSymbolRefExpr::VK_DTPOFF;
[SystemZ] Add back end This adds the actual lib/Target/SystemZ target files necessary to implement the SystemZ target. Note that at this point, the target cannot yet be built since the configure bits are missing. Those will be provided shortly by a follow-on patch. This version of the patch incorporates feedback from reviews by Chris Lattner and Anton Korobeynikov. Thanks to all reviewers! Patch by Richard Sandiford. llvm-svn: 181203
2013-05-07 00:15:19 +08:00
case SystemZCP::NTPOFF: return MCSymbolRefExpr::VK_NTPOFF;
}
llvm_unreachable("Invalid SystemCPModifier!");
}
void SystemZAsmPrinter::
EmitMachineConstantPoolValue(MachineConstantPoolValue *MCPV) {
[SystemZ] Use "auto" for cast results No functional change intended. llvm-svn: 203106
2014-03-06 19:22:58 +08:00
auto *ZCPV = static_cast<SystemZConstantPoolValue*>(MCPV);
[SystemZ] Add back end This adds the actual lib/Target/SystemZ target files necessary to implement the SystemZ target. Note that at this point, the target cannot yet be built since the configure bits are missing. Those will be provided shortly by a follow-on patch. This version of the patch incorporates feedback from reviews by Chris Lattner and Anton Korobeynikov. Thanks to all reviewers! Patch by Richard Sandiford. llvm-svn: 181203
2013-05-07 00:15:19 +08:00
const MCExpr *Expr =
MC: Clean up MCExpr naming. NFC. llvm-svn: 238634
2015-05-30 09:25:56 +08:00
MCSymbolRefExpr::create(getSymbol(ZCPV->getGlobalValue()),
[SystemZ] Add back end This adds the actual lib/Target/SystemZ target files necessary to implement the SystemZ target. Note that at this point, the target cannot yet be built since the configure bits are missing. Those will be provided shortly by a follow-on patch. This version of the patch incorporates feedback from reviews by Chris Lattner and Anton Korobeynikov. Thanks to all reviewers! Patch by Richard Sandiford. llvm-svn: 181203
2013-05-07 00:15:19 +08:00
getModifierVariantKind(ZCPV->getModifier()),
OutContext);
Move most user of TargetMachine::getDataLayout to the Module one Summary: This change is part of a series of commits dedicated to have a single DataLayout during compilation by using always the one owned by the module. This patch is quite boring overall, except for some uglyness in ASMPrinter which has a getDataLayout function but has some clients that use it without a Module (llmv-dsymutil, llvm-dwarfdump), so some methods are taking a DataLayout as parameter. Reviewers: echristo Subscribers: yaron.keren, rafael, llvm-commits, jholewinski Differential Revision: http://reviews.llvm.org/D11090 From: Mehdi Amini <mehdi.amini@apple.com> llvm-svn: 242386
2015-07-16 14:11:10 +08:00
uint64_t Size = getDataLayout().getTypeAllocSize(ZCPV->getType());
[SystemZ] Add back end This adds the actual lib/Target/SystemZ target files necessary to implement the SystemZ target. Note that at this point, the target cannot yet be built since the configure bits are missing. Those will be provided shortly by a follow-on patch. This version of the patch incorporates feedback from reviews by Chris Lattner and Anton Korobeynikov. Thanks to all reviewers! Patch by Richard Sandiford. llvm-svn: 181203
2013-05-07 00:15:19 +08:00
[AsmPrinter] Make AsmPrinter's OutStreamer member a unique_ptr. AsmPrinter owns the OutStreamer, so an owning pointer makes sense here. Using a reference for this is crufty. llvm-svn: 235752
2015-04-25 03:11:51 +08:00
OutStreamer->EmitValue(Expr, Size);
[SystemZ] Add back end This adds the actual lib/Target/SystemZ target files necessary to implement the SystemZ target. Note that at this point, the target cannot yet be built since the configure bits are missing. Those will be provided shortly by a follow-on patch. This version of the patch incorporates feedback from reviews by Chris Lattner and Anton Korobeynikov. Thanks to all reviewers! Patch by Richard Sandiford. llvm-svn: 181203
2013-05-07 00:15:19 +08:00
}
bool SystemZAsmPrinter::PrintAsmOperand(const MachineInstr *MI,
unsigned OpNo,
unsigned AsmVariant,
const char *ExtraCode,
raw_ostream &OS) {
if (ExtraCode && *ExtraCode == 'n') {
if (!MI->getOperand(OpNo).isImm())
return true;
OS << -int64_t(MI->getOperand(OpNo).getImm());
} else {
The AsmPrinter has a Mangler. Use it. llvm-svn: 193617
2013-10-30 00:18:15 +08:00
SystemZMCInstLower Lower(MF->getContext(), *this);
[SystemZ] Add back end This adds the actual lib/Target/SystemZ target files necessary to implement the SystemZ target. Note that at this point, the target cannot yet be built since the configure bits are missing. Those will be provided shortly by a follow-on patch. This version of the patch incorporates feedback from reviews by Chris Lattner and Anton Korobeynikov. Thanks to all reviewers! Patch by Richard Sandiford. llvm-svn: 181203
2013-05-07 00:15:19 +08:00
MCOperand MO(Lower.lowerOperand(MI->getOperand(OpNo)));
MC: Add target hook to control symbol quoting llvm-svn: 239370
2015-06-09 08:31:39 +08:00
SystemZInstPrinter::printOperand(MO, MAI, OS);
[SystemZ] Add back end This adds the actual lib/Target/SystemZ target files necessary to implement the SystemZ target. Note that at this point, the target cannot yet be built since the configure bits are missing. Those will be provided shortly by a follow-on patch. This version of the patch incorporates feedback from reviews by Chris Lattner and Anton Korobeynikov. Thanks to all reviewers! Patch by Richard Sandiford. llvm-svn: 181203
2013-05-07 00:15:19 +08:00
}
return false;
}
bool SystemZAsmPrinter::PrintAsmMemoryOperand(const MachineInstr *MI,
unsigned OpNo,
unsigned AsmVariant,
const char *ExtraCode,
raw_ostream &OS) {
SystemZInstPrinter::printAddress(MI->getOperand(OpNo).getReg(),
MI->getOperand(OpNo + 1).getImm(),
MI->getOperand(OpNo + 2).getReg(), OS);
return false;
}
// Force static initialization.
extern "C" void LLVMInitializeSystemZAsmPrinter() {
Move the global variables representing each Target behind accessor function This avoids "static initialization order fiasco" Differential Revision: https://reviews.llvm.org/D25412 llvm-svn: 283702
2016-10-10 07:00:34 +08:00
RegisterAsmPrinter<SystemZAsmPrinter> X(getTheSystemZTarget());
[SystemZ] Add back end This adds the actual lib/Target/SystemZ target files necessary to implement the SystemZ target. Note that at this point, the target cannot yet be built since the configure bits are missing. Those will be provided shortly by a follow-on patch. This version of the patch incorporates feedback from reviews by Chris Lattner and Anton Korobeynikov. Thanks to all reviewers! Patch by Richard Sandiford. llvm-svn: 181203
2013-05-07 00:15:19 +08:00
}