RPCS3/llvm-mirror
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-10-22 20:43:44 +02:00
Code Issues Projects Releases Wiki Activity
ce9450a337
llvm-mirror/lib/Target/ARM/ARMLegalizerInfo.cpp

403 lines
16 KiB
C++
Raw Normal View History

[ARM] Add plumbing for GlobalISel Add GlobalISel skeleton, up to the point where we can select a ret void. llvm-svn: 286573
2016-11-11 09:27:37 +01:00
//===- ARMLegalizerInfo.cpp --------------------------------------*- C++ -*-==//
//
// The LLVM Compiler Infrastructure
//
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
/// \file
/// This file implements the targeting of the Machinelegalizer class for ARM.
/// \todo This should be generated by TableGen.
//===----------------------------------------------------------------------===//
#include "ARMLegalizerInfo.h"
[ARM] GlobalISel: Add support for i32 modulo Add support for modulo for targets that have hardware division and for those that don't. When hardware division is not available, we have to choose the correct libcall to use. This is generally straightforward, except for AEABI. The AEABI variant is trickier than the other libcalls because it returns { quotient, remainder }, instead of just one value like the other libcalls that we've seen so far. Therefore, we need to use custom lowering for it. However, we don't want to have too much special code, so we refactor the target-independent code in the legalizer by adding a helper for replacing an instruction with a libcall. This helper is used by the legalizer itself when dealing with simple calls, and also by the custom ARM legalization for the more complicated AEABI divmod calls. llvm-svn: 305459
2017-06-15 12:53:31 +02:00
#include "ARMCallLowering.h"
[ARM] GlobalISel: Use Subtarget in Legalizer Start using the Subtarget to make decisions about what's legal. In particular, we only mark floating point operations as legal if we have VFP2, which is something we should've done from the very start. llvm-svn: 295439
2017-02-17 12:25:17 +01:00
#include "ARMSubtarget.h"
[ARM] GlobalISel: Legalize s8 and s16 G_(S|U)DIV We have to widen the operands to 32 bits and then we can either use hardware division if it is available or lower to a libcall otherwise. At the moment it is not enough to set the Legalizer action to WidenScalar, since for libcalls it won't know what to do (it won't be able to find what size to widen to, because it will find Libcall and not Legal for 32 bits). To hack around this limitation, we request Custom lowering, and as part of that we widen first and then we run another legalizeInstrStep on the widened DIV. llvm-svn: 301166
2017-04-24 11:12:19 +02:00
#include "llvm/CodeGen/GlobalISel/LegalizerHelper.h"
[ARM] GlobalISel: Add support for i32 modulo Add support for modulo for targets that have hardware division and for those that don't. When hardware division is not available, we have to choose the correct libcall to use. This is generally straightforward, except for AEABI. The AEABI variant is trickier than the other libcalls because it returns { quotient, remainder }, instead of just one value like the other libcalls that we've seen so far. Therefore, we need to use custom lowering for it. However, we don't want to have too much special code, so we refactor the target-independent code in the legalizer by adding a helper for replacing an instruction with a libcall. This helper is used by the legalizer itself when dealing with simple calls, and also by the custom ARM legalization for the more complicated AEABI divmod calls. llvm-svn: 305459
2017-06-15 12:53:31 +02:00
#include "llvm/CodeGen/LowLevelType.h"
[ARM] GlobalISel: Legalize s8 and s16 G_(S|U)DIV We have to widen the operands to 32 bits and then we can either use hardware division if it is available or lower to a libcall otherwise. At the moment it is not enough to set the Legalizer action to WidenScalar, since for libcalls it won't know what to do (it won't be able to find what size to widen to, because it will find Libcall and not Legal for 32 bits). To hack around this limitation, we request Custom lowering, and as part of that we widen first and then we run another legalizeInstrStep on the widened DIV. llvm-svn: 301166
2017-04-24 11:12:19 +02:00
#include "llvm/CodeGen/MachineRegisterInfo.h"
Fix a bunch more layering of CodeGen headers that are in Target All these headers already depend on CodeGen headers so moving them into CodeGen fixes the layering (since CodeGen depends on Target, not the other way around). llvm-svn: 318490
2017-11-17 02:07:10 +01:00
#include "llvm/CodeGen/TargetOpcodes.h"
[ARM] Add plumbing for GlobalISel Add GlobalISel skeleton, up to the point where we can select a ret void. llvm-svn: 286573
2016-11-11 09:27:37 +01:00
#include "llvm/CodeGen/ValueTypes.h"
#include "llvm/IR/DerivedTypes.h"
#include "llvm/IR/Type.h"
using namespace llvm;
[GlobalISel] Enable legalizing non-power-of-2 sized types. This changes the interface of how targets describe how to legalize, see the below description. 1. Interface for targets to describe how to legalize. In GlobalISel, the API in the LegalizerInfo class is the main interface for targets to specify which types are legal for which operations, and what to do to turn illegal type/operation combinations into legal ones. For each operation the type sizes that can be legalized without having to change the size of the type are specified with a call to setAction. This isn't different to how GlobalISel worked before. For example, for a target that supports 32 and 64 bit adds natively: for (auto Ty : {s32, s64}) setAction({G_ADD, 0, s32}, Legal); or for a target that needs a library call for a 32 bit division: setAction({G_SDIV, s32}, Libcall); The main conceptual change to the LegalizerInfo API, is in specifying how to legalize the type sizes for which a change of size is needed. For example, in the above example, how to specify how all types from i1 to i8388607 (apart from s32 and s64 which are legal) need to be legalized and expressed in terms of operations on the available legal sizes (again, i32 and i64 in this case). Before, the implementation only allowed specifying power-of-2-sized types (e.g. setAction({G_ADD, 0, s128}, NarrowScalar). A worse limitation was that if you'd wanted to specify how to legalize all the sized types as allowed by the LLVM-IR LangRef, i1 to i8388607, you'd have to call setAction 8388607-3 times and probably would need a lot of memory to store all of these specifications. Instead, the legalization actions that need to change the size of the type are specified now using a "SizeChangeStrategy". For example: setLegalizeScalarToDifferentSizeStrategy( G_ADD, 0, widenToLargerAndNarrowToLargest); This example indicates that for type sizes for which there is a larger size that can be legalized towards, do it by Widening the size. For example, G_ADD on s17 will be legalized by first doing WidenScalar to make it s32, after which it's legal. The "NarrowToLargest" indicates what to do if there is no larger size that can be legalized towards. E.g. G_ADD on s92 will be legalized by doing NarrowScalar to s64. Another example, taken from the ARM backend is: for (unsigned Op : {G_SDIV, G_UDIV}) { setLegalizeScalarToDifferentSizeStrategy(Op, 0, widenToLargerTypesUnsupportedOtherwise); if (ST.hasDivideInARMMode()) setAction({Op, s32}, Legal); else setAction({Op, s32}, Libcall); } For this example, G_SDIV on s8, on a target without a divide instruction, would be legalized by first doing action (WidenScalar, s32), followed by (Libcall, s32). The same principle is also followed for when the number of vector lanes on vector data types need to be changed, e.g.: setAction({G_ADD, LLT::vector(8, 8)}, LegalizerInfo::Legal); setAction({G_ADD, LLT::vector(16, 8)}, LegalizerInfo::Legal); setAction({G_ADD, LLT::vector(4, 16)}, LegalizerInfo::Legal); setAction({G_ADD, LLT::vector(8, 16)}, LegalizerInfo::Legal); setAction({G_ADD, LLT::vector(2, 32)}, LegalizerInfo::Legal); setAction({G_ADD, LLT::vector(4, 32)}, LegalizerInfo::Legal); setLegalizeVectorElementToDifferentSizeStrategy( G_ADD, 0, widenToLargerTypesUnsupportedOtherwise); As currently implemented here, vector types are legalized by first making the vector element size legal, followed by then making the number of lanes legal. The strategy to follow in the first step is set by a call to setLegalizeVectorElementToDifferentSizeStrategy, see example above. The strategy followed in the second step "moreToWiderTypesAndLessToWidest" (see code for its definition), indicating that vectors are widened to more elements so they map to natively supported vector widths, or when there isn't a legal wider vector, split the vector to map it to the widest vector supported. Therefore, for the above specification, some example legalizations are: * getAction({G_ADD, LLT::vector(3, 3)}) returns {WidenScalar, LLT::vector(3, 8)} * getAction({G_ADD, LLT::vector(3, 8)}) then returns {MoreElements, LLT::vector(8, 8)} * getAction({G_ADD, LLT::vector(20, 8)}) returns {FewerElements, LLT::vector(16, 8)} 2. Key implementation aspects. How to legalize a specific (operation, type index, size) tuple is represented by mapping intervals of integers representing a range of size types to an action to take, e.g.: setScalarAction({G_ADD, LLT:scalar(1)}, {{1, WidenScalar}, // bit sizes [ 1, 31[ {32, Legal}, // bit sizes [32, 33[ {33, WidenScalar}, // bit sizes [33, 64[ {64, Legal}, // bit sizes [64, 65[ {65, NarrowScalar} // bit sizes [65, +inf[ }); Please note that most of the code to do the actual lowering of non-power-of-2 sized types is currently missing, this is just trying to make it possible for targets to specify what is legal, and how non-legal types should be legalized. Probably quite a bit of further work is needed in the actual legalizing and the other passes in GlobalISel to support non-power-of-2 sized types. I hope the documentation in LegalizerInfo.h and the examples provided in the various {Target}LegalizerInfo.cpp and LegalizerInfoTest.cpp explains well enough how this is meant to be used. This drops the need for LLT::{half,double}...Size(). Differential Revision: https://reviews.llvm.org/D30529 llvm-svn: 317560
2017-11-07 11:34:34 +01:00
/// FIXME: The following static functions are SizeChangeStrategy functions
/// that are meant to temporarily mimic the behaviour of the old legalization
/// based on doubling/halving non-legal types as closely as possible. This is
/// not entirly possible as only legalizing the types that are exactly a power
/// of 2 times the size of the legal types would require specifying all those
/// sizes explicitly.
/// In practice, not specifying those isn't a problem, and the below functions
/// should disappear quickly as we add support for legalizing non-power-of-2
/// sized types further.
static void
addAndInterleaveWithUnsupported(LegalizerInfo::SizeAndActionsVec &result,
const LegalizerInfo::SizeAndActionsVec &v) {
for (unsigned i = 0; i < v.size(); ++i) {
result.push_back(v[i]);
if (i + 1 < v[i].first && i + 1 < v.size() &&
v[i + 1].first != v[i].first + 1)
result.push_back({v[i].first + 1, LegalizerInfo::Unsupported});
}
}
static LegalizerInfo::SizeAndActionsVec
widen_8_16(const LegalizerInfo::SizeAndActionsVec &v) {
assert(v.size() >= 1);
assert(v[0].first > 17);
LegalizerInfo::SizeAndActionsVec result = {
{1, LegalizerInfo::Unsupported},
{8, LegalizerInfo::WidenScalar}, {9, LegalizerInfo::Unsupported},
{16, LegalizerInfo::WidenScalar}, {17, LegalizerInfo::Unsupported}};
addAndInterleaveWithUnsupported(result, v);
auto Largest = result.back().first;
result.push_back({Largest + 1, LegalizerInfo::Unsupported});
return result;
}
static LegalizerInfo::SizeAndActionsVec
widen_1_8_16(const LegalizerInfo::SizeAndActionsVec &v) {
assert(v.size() >= 1);
assert(v[0].first > 17);
LegalizerInfo::SizeAndActionsVec result = {
{1, LegalizerInfo::WidenScalar}, {2, LegalizerInfo::Unsupported},
{8, LegalizerInfo::WidenScalar}, {9, LegalizerInfo::Unsupported},
{16, LegalizerInfo::WidenScalar}, {17, LegalizerInfo::Unsupported}};
addAndInterleaveWithUnsupported(result, v);
auto Largest = result.back().first;
result.push_back({Largest + 1, LegalizerInfo::Unsupported});
return result;
}
[ARM] GlobalISel: Extract tiny helper. NFC Extract functionality for determining if the target uses AEABI. llvm-svn: 307145
2017-07-05 13:53:51 +02:00
static bool AEABI(const ARMSubtarget &ST) {
return ST.isTargetAEABI() || ST.isTargetGNUAEABI() || ST.isTargetMuslAEABI();
}
[ARM] GlobalISel: Use Subtarget in Legalizer Start using the Subtarget to make decisions about what's legal. In particular, we only mark floating point operations as legal if we have VFP2, which is something we should've done from the very start. llvm-svn: 295439
2017-02-17 12:25:17 +01:00
ARMLegalizerInfo::ARMLegalizerInfo(const ARMSubtarget &ST) {
[ARM] GlobalISel: Select add i32, i32 Add the minimal support necessary to select a function that returns the sum of two i32 values. This includes some support for argument/return lowering of i32 values through registers, as well as the handling of copy and add instructions throughout the GlobalISel pipeline. Differential Revision: https://reviews.llvm.org/D26677 llvm-svn: 289940
2016-12-16 13:54:46 +01:00
using namespace TargetOpcode;
[ARM] GlobalISel: Allow i8 and i16 adds Teach the instruction selector and legalizer that it's ok to have adds with 8 or 16-bit integers. This is the second part of https://reviews.llvm.org/D27704 llvm-svn: 290105
2016-12-19 15:07:56 +01:00
[ARM] GlobalISel: Support loading from the stack Add support for selecting simple G_LOAD and G_FRAME_INDEX instructions (32-bit scalars only). This will be useful for functions that need to pass arguments on the stack. First part of https://reviews.llvm.org/D27195. llvm-svn: 290096
2016-12-19 12:26:31 +01:00
const LLT p0 = LLT::pointer(0, 32);
[ARM] GlobalISel: Allow i8 and i16 adds Teach the instruction selector and legalizer that it's ok to have adds with 8 or 16-bit integers. This is the second part of https://reviews.llvm.org/D27704 llvm-svn: 290105
2016-12-19 15:07:56 +01:00
[ARM] GlobalISel: Support i1 add and ABI extensions Add support for: * i1 add * i1 function arguments, if passed through registers * i1 returns, with ABI signext/zeroext Differential Revision: https://reviews.llvm.org/D27706 llvm-svn: 293035
2017-01-25 09:47:40 +01:00
const LLT s1 = LLT::scalar(1);
[ARM] GlobalISel: Allow i8 and i16 adds Teach the instruction selector and legalizer that it's ok to have adds with 8 or 16-bit integers. This is the second part of https://reviews.llvm.org/D27704 llvm-svn: 290105
2016-12-19 15:07:56 +01:00
const LLT s8 = LLT::scalar(8);
const LLT s16 = LLT::scalar(16);
[ARM] GlobalISel: Select add i32, i32 Add the minimal support necessary to select a function that returns the sum of two i32 values. This includes some support for argument/return lowering of i32 values through registers, as well as the handling of copy and add instructions throughout the GlobalISel pipeline. Differential Revision: https://reviews.llvm.org/D26677 llvm-svn: 289940
2016-12-16 13:54:46 +01:00
const LLT s32 = LLT::scalar(32);
[ARM] GlobalISel: Legalize 64-bit G_FADD and G_LOAD For now we just mark them as legal all the time and let the other passes bail out if they can't handle it. In the future, we'll want to move more of the brains into the legalizer. llvm-svn: 295300
2017-02-16 10:09:49 +01:00
const LLT s64 = LLT::scalar(64);
[ARM] GlobalISel: Select add i32, i32 Add the minimal support necessary to select a function that returns the sum of two i32 values. This includes some support for argument/return lowering of i32 values through registers, as well as the handling of copy and add instructions throughout the GlobalISel pipeline. Differential Revision: https://reviews.llvm.org/D26677 llvm-svn: 289940
2016-12-16 13:54:46 +01:00
[ARM] GlobalISel: Mark G_GLOBAL_VALUE as legal llvm-svn: 309090
2017-07-26 11:25:15 +02:00
setAction({G_GLOBAL_VALUE, p0}, Legal);
[ARM] GlobalISel: Support loading from the stack Add support for selecting simple G_LOAD and G_FRAME_INDEX instructions (32-bit scalars only). This will be useful for functions that need to pass arguments on the stack. First part of https://reviews.llvm.org/D27195. llvm-svn: 290096
2016-12-19 12:26:31 +01:00
setAction({G_FRAME_INDEX, p0}, Legal);
[ARM] GlobalISel: Legalize stores Allow the same types that we allow for loads. llvm-svn: 296108
2017-02-24 12:28:24 +01:00
for (unsigned Op : {G_LOAD, G_STORE}) {
for (auto Ty : {s1, s8, s16, s32, p0})
setAction({Op, Ty}, Legal);
setAction({Op, 1, p0}, Legal);
}
[ARM] GlobalISel: Support loading from the stack Add support for selecting simple G_LOAD and G_FRAME_INDEX instructions (32-bit scalars only). This will be useful for functions that need to pass arguments on the stack. First part of https://reviews.llvm.org/D27195. llvm-svn: 290096
2016-12-19 12:26:31 +01:00
[ARM] GlobalISel: Support G_XOR Same as the other binary operators: - legalize to 32 bits - map to GPRs - select to EORrr via TableGen'erated code llvm-svn: 304898
2017-06-07 13:57:30 +02:00
for (unsigned Op : {G_ADD, G_SUB, G_MUL, G_AND, G_OR, G_XOR}) {
[GlobalISel] Enable legalizing non-power-of-2 sized types. This changes the interface of how targets describe how to legalize, see the below description. 1. Interface for targets to describe how to legalize. In GlobalISel, the API in the LegalizerInfo class is the main interface for targets to specify which types are legal for which operations, and what to do to turn illegal type/operation combinations into legal ones. For each operation the type sizes that can be legalized without having to change the size of the type are specified with a call to setAction. This isn't different to how GlobalISel worked before. For example, for a target that supports 32 and 64 bit adds natively: for (auto Ty : {s32, s64}) setAction({G_ADD, 0, s32}, Legal); or for a target that needs a library call for a 32 bit division: setAction({G_SDIV, s32}, Libcall); The main conceptual change to the LegalizerInfo API, is in specifying how to legalize the type sizes for which a change of size is needed. For example, in the above example, how to specify how all types from i1 to i8388607 (apart from s32 and s64 which are legal) need to be legalized and expressed in terms of operations on the available legal sizes (again, i32 and i64 in this case). Before, the implementation only allowed specifying power-of-2-sized types (e.g. setAction({G_ADD, 0, s128}, NarrowScalar). A worse limitation was that if you'd wanted to specify how to legalize all the sized types as allowed by the LLVM-IR LangRef, i1 to i8388607, you'd have to call setAction 8388607-3 times and probably would need a lot of memory to store all of these specifications. Instead, the legalization actions that need to change the size of the type are specified now using a "SizeChangeStrategy". For example: setLegalizeScalarToDifferentSizeStrategy( G_ADD, 0, widenToLargerAndNarrowToLargest); This example indicates that for type sizes for which there is a larger size that can be legalized towards, do it by Widening the size. For example, G_ADD on s17 will be legalized by first doing WidenScalar to make it s32, after which it's legal. The "NarrowToLargest" indicates what to do if there is no larger size that can be legalized towards. E.g. G_ADD on s92 will be legalized by doing NarrowScalar to s64. Another example, taken from the ARM backend is: for (unsigned Op : {G_SDIV, G_UDIV}) { setLegalizeScalarToDifferentSizeStrategy(Op, 0, widenToLargerTypesUnsupportedOtherwise); if (ST.hasDivideInARMMode()) setAction({Op, s32}, Legal); else setAction({Op, s32}, Libcall); } For this example, G_SDIV on s8, on a target without a divide instruction, would be legalized by first doing action (WidenScalar, s32), followed by (Libcall, s32). The same principle is also followed for when the number of vector lanes on vector data types need to be changed, e.g.: setAction({G_ADD, LLT::vector(8, 8)}, LegalizerInfo::Legal); setAction({G_ADD, LLT::vector(16, 8)}, LegalizerInfo::Legal); setAction({G_ADD, LLT::vector(4, 16)}, LegalizerInfo::Legal); setAction({G_ADD, LLT::vector(8, 16)}, LegalizerInfo::Legal); setAction({G_ADD, LLT::vector(2, 32)}, LegalizerInfo::Legal); setAction({G_ADD, LLT::vector(4, 32)}, LegalizerInfo::Legal); setLegalizeVectorElementToDifferentSizeStrategy( G_ADD, 0, widenToLargerTypesUnsupportedOtherwise); As currently implemented here, vector types are legalized by first making the vector element size legal, followed by then making the number of lanes legal. The strategy to follow in the first step is set by a call to setLegalizeVectorElementToDifferentSizeStrategy, see example above. The strategy followed in the second step "moreToWiderTypesAndLessToWidest" (see code for its definition), indicating that vectors are widened to more elements so they map to natively supported vector widths, or when there isn't a legal wider vector, split the vector to map it to the widest vector supported. Therefore, for the above specification, some example legalizations are: * getAction({G_ADD, LLT::vector(3, 3)}) returns {WidenScalar, LLT::vector(3, 8)} * getAction({G_ADD, LLT::vector(3, 8)}) then returns {MoreElements, LLT::vector(8, 8)} * getAction({G_ADD, LLT::vector(20, 8)}) returns {FewerElements, LLT::vector(16, 8)} 2. Key implementation aspects. How to legalize a specific (operation, type index, size) tuple is represented by mapping intervals of integers representing a range of size types to an action to take, e.g.: setScalarAction({G_ADD, LLT:scalar(1)}, {{1, WidenScalar}, // bit sizes [ 1, 31[ {32, Legal}, // bit sizes [32, 33[ {33, WidenScalar}, // bit sizes [33, 64[ {64, Legal}, // bit sizes [64, 65[ {65, NarrowScalar} // bit sizes [65, +inf[ }); Please note that most of the code to do the actual lowering of non-power-of-2 sized types is currently missing, this is just trying to make it possible for targets to specify what is legal, and how non-legal types should be legalized. Probably quite a bit of further work is needed in the actual legalizing and the other passes in GlobalISel to support non-power-of-2 sized types. I hope the documentation in LegalizerInfo.h and the examples provided in the various {Target}LegalizerInfo.cpp and LegalizerInfoTest.cpp explains well enough how this is meant to be used. This drops the need for LLT::{half,double}...Size(). Differential Revision: https://reviews.llvm.org/D30529 llvm-svn: 317560
2017-11-07 11:34:34 +01:00
if (Op != G_ADD)
setLegalizeScalarToDifferentSizeStrategy(
Op, 0, widenToLargerTypesUnsupportedOtherwise);
[ARM][GlobalISel] Legalize narrow scalar ops by widening This is the same as r292827 for AArch64: we widen 8- and 16-bit ADD, SUB and MUL to 32 bits since we only have TableGen patterns for 32 bits. See the commit message for r292827 for more details. At this point we could just remove some of the tests for regbankselect and instruction-select, since we're not going to see any narrow operations at those levels anymore. Instead I decided to update them with G_ANYEXT/G_TRUNC operations, so we can validate the full sequences generated by the legalizer. llvm-svn: 302782
2017-05-11 11:45:57 +02:00
setAction({Op, s32}, Legal);
}
[ARM] GlobalISel: Select add i32, i32 Add the minimal support necessary to select a function that returns the sum of two i32 values. This includes some support for argument/return lowering of i32 values through registers, as well as the handling of copy and add instructions throughout the GlobalISel pipeline. Differential Revision: https://reviews.llvm.org/D26677 llvm-svn: 289940
2016-12-16 13:54:46 +01:00
[ARM] GlobalISel: Support G_(S|U)DIV for s32 Add support for both targets with hardware division and without. For hardware division we have to add support throughout the pipeline (legalizer, reg bank select, instruction select). For targets without hardware division, we only need to mark it as a libcall. llvm-svn: 301164
2017-04-24 10:20:05 +02:00
for (unsigned Op : {G_SDIV, G_UDIV}) {
[GlobalISel] Enable legalizing non-power-of-2 sized types. This changes the interface of how targets describe how to legalize, see the below description. 1. Interface for targets to describe how to legalize. In GlobalISel, the API in the LegalizerInfo class is the main interface for targets to specify which types are legal for which operations, and what to do to turn illegal type/operation combinations into legal ones. For each operation the type sizes that can be legalized without having to change the size of the type are specified with a call to setAction. This isn't different to how GlobalISel worked before. For example, for a target that supports 32 and 64 bit adds natively: for (auto Ty : {s32, s64}) setAction({G_ADD, 0, s32}, Legal); or for a target that needs a library call for a 32 bit division: setAction({G_SDIV, s32}, Libcall); The main conceptual change to the LegalizerInfo API, is in specifying how to legalize the type sizes for which a change of size is needed. For example, in the above example, how to specify how all types from i1 to i8388607 (apart from s32 and s64 which are legal) need to be legalized and expressed in terms of operations on the available legal sizes (again, i32 and i64 in this case). Before, the implementation only allowed specifying power-of-2-sized types (e.g. setAction({G_ADD, 0, s128}, NarrowScalar). A worse limitation was that if you'd wanted to specify how to legalize all the sized types as allowed by the LLVM-IR LangRef, i1 to i8388607, you'd have to call setAction 8388607-3 times and probably would need a lot of memory to store all of these specifications. Instead, the legalization actions that need to change the size of the type are specified now using a "SizeChangeStrategy". For example: setLegalizeScalarToDifferentSizeStrategy( G_ADD, 0, widenToLargerAndNarrowToLargest); This example indicates that for type sizes for which there is a larger size that can be legalized towards, do it by Widening the size. For example, G_ADD on s17 will be legalized by first doing WidenScalar to make it s32, after which it's legal. The "NarrowToLargest" indicates what to do if there is no larger size that can be legalized towards. E.g. G_ADD on s92 will be legalized by doing NarrowScalar to s64. Another example, taken from the ARM backend is: for (unsigned Op : {G_SDIV, G_UDIV}) { setLegalizeScalarToDifferentSizeStrategy(Op, 0, widenToLargerTypesUnsupportedOtherwise); if (ST.hasDivideInARMMode()) setAction({Op, s32}, Legal); else setAction({Op, s32}, Libcall); } For this example, G_SDIV on s8, on a target without a divide instruction, would be legalized by first doing action (WidenScalar, s32), followed by (Libcall, s32). The same principle is also followed for when the number of vector lanes on vector data types need to be changed, e.g.: setAction({G_ADD, LLT::vector(8, 8)}, LegalizerInfo::Legal); setAction({G_ADD, LLT::vector(16, 8)}, LegalizerInfo::Legal); setAction({G_ADD, LLT::vector(4, 16)}, LegalizerInfo::Legal); setAction({G_ADD, LLT::vector(8, 16)}, LegalizerInfo::Legal); setAction({G_ADD, LLT::vector(2, 32)}, LegalizerInfo::Legal); setAction({G_ADD, LLT::vector(4, 32)}, LegalizerInfo::Legal); setLegalizeVectorElementToDifferentSizeStrategy( G_ADD, 0, widenToLargerTypesUnsupportedOtherwise); As currently implemented here, vector types are legalized by first making the vector element size legal, followed by then making the number of lanes legal. The strategy to follow in the first step is set by a call to setLegalizeVectorElementToDifferentSizeStrategy, see example above. The strategy followed in the second step "moreToWiderTypesAndLessToWidest" (see code for its definition), indicating that vectors are widened to more elements so they map to natively supported vector widths, or when there isn't a legal wider vector, split the vector to map it to the widest vector supported. Therefore, for the above specification, some example legalizations are: * getAction({G_ADD, LLT::vector(3, 3)}) returns {WidenScalar, LLT::vector(3, 8)} * getAction({G_ADD, LLT::vector(3, 8)}) then returns {MoreElements, LLT::vector(8, 8)} * getAction({G_ADD, LLT::vector(20, 8)}) returns {FewerElements, LLT::vector(16, 8)} 2. Key implementation aspects. How to legalize a specific (operation, type index, size) tuple is represented by mapping intervals of integers representing a range of size types to an action to take, e.g.: setScalarAction({G_ADD, LLT:scalar(1)}, {{1, WidenScalar}, // bit sizes [ 1, 31[ {32, Legal}, // bit sizes [32, 33[ {33, WidenScalar}, // bit sizes [33, 64[ {64, Legal}, // bit sizes [64, 65[ {65, NarrowScalar} // bit sizes [65, +inf[ }); Please note that most of the code to do the actual lowering of non-power-of-2 sized types is currently missing, this is just trying to make it possible for targets to specify what is legal, and how non-legal types should be legalized. Probably quite a bit of further work is needed in the actual legalizing and the other passes in GlobalISel to support non-power-of-2 sized types. I hope the documentation in LegalizerInfo.h and the examples provided in the various {Target}LegalizerInfo.cpp and LegalizerInfoTest.cpp explains well enough how this is meant to be used. This drops the need for LLT::{half,double}...Size(). Differential Revision: https://reviews.llvm.org/D30529 llvm-svn: 317560
2017-11-07 11:34:34 +01:00
setLegalizeScalarToDifferentSizeStrategy(Op, 0,
widenToLargerTypesUnsupportedOtherwise);
[ARM] GlobalISel: Support G_(S|U)DIV for s32 Add support for both targets with hardware division and without. For hardware division we have to add support throughout the pipeline (legalizer, reg bank select, instruction select). For targets without hardware division, we only need to mark it as a libcall. llvm-svn: 301164
2017-04-24 10:20:05 +02:00
if (ST.hasDivideInARMMode())
setAction({Op, s32}, Legal);
else
setAction({Op, s32}, Libcall);
}
[ARM] GlobalISel: Support G_(S|U)REM for s8 and s16 Widen to s32, and then do whatever Lowering/Custom/Libcall action the subtarget wants. llvm-svn: 308285
2017-07-18 12:07:01 +02:00
for (unsigned Op : {G_SREM, G_UREM}) {
[GlobalISel] Enable legalizing non-power-of-2 sized types. This changes the interface of how targets describe how to legalize, see the below description. 1. Interface for targets to describe how to legalize. In GlobalISel, the API in the LegalizerInfo class is the main interface for targets to specify which types are legal for which operations, and what to do to turn illegal type/operation combinations into legal ones. For each operation the type sizes that can be legalized without having to change the size of the type are specified with a call to setAction. This isn't different to how GlobalISel worked before. For example, for a target that supports 32 and 64 bit adds natively: for (auto Ty : {s32, s64}) setAction({G_ADD, 0, s32}, Legal); or for a target that needs a library call for a 32 bit division: setAction({G_SDIV, s32}, Libcall); The main conceptual change to the LegalizerInfo API, is in specifying how to legalize the type sizes for which a change of size is needed. For example, in the above example, how to specify how all types from i1 to i8388607 (apart from s32 and s64 which are legal) need to be legalized and expressed in terms of operations on the available legal sizes (again, i32 and i64 in this case). Before, the implementation only allowed specifying power-of-2-sized types (e.g. setAction({G_ADD, 0, s128}, NarrowScalar). A worse limitation was that if you'd wanted to specify how to legalize all the sized types as allowed by the LLVM-IR LangRef, i1 to i8388607, you'd have to call setAction 8388607-3 times and probably would need a lot of memory to store all of these specifications. Instead, the legalization actions that need to change the size of the type are specified now using a "SizeChangeStrategy". For example: setLegalizeScalarToDifferentSizeStrategy( G_ADD, 0, widenToLargerAndNarrowToLargest); This example indicates that for type sizes for which there is a larger size that can be legalized towards, do it by Widening the size. For example, G_ADD on s17 will be legalized by first doing WidenScalar to make it s32, after which it's legal. The "NarrowToLargest" indicates what to do if there is no larger size that can be legalized towards. E.g. G_ADD on s92 will be legalized by doing NarrowScalar to s64. Another example, taken from the ARM backend is: for (unsigned Op : {G_SDIV, G_UDIV}) { setLegalizeScalarToDifferentSizeStrategy(Op, 0, widenToLargerTypesUnsupportedOtherwise); if (ST.hasDivideInARMMode()) setAction({Op, s32}, Legal); else setAction({Op, s32}, Libcall); } For this example, G_SDIV on s8, on a target without a divide instruction, would be legalized by first doing action (WidenScalar, s32), followed by (Libcall, s32). The same principle is also followed for when the number of vector lanes on vector data types need to be changed, e.g.: setAction({G_ADD, LLT::vector(8, 8)}, LegalizerInfo::Legal); setAction({G_ADD, LLT::vector(16, 8)}, LegalizerInfo::Legal); setAction({G_ADD, LLT::vector(4, 16)}, LegalizerInfo::Legal); setAction({G_ADD, LLT::vector(8, 16)}, LegalizerInfo::Legal); setAction({G_ADD, LLT::vector(2, 32)}, LegalizerInfo::Legal); setAction({G_ADD, LLT::vector(4, 32)}, LegalizerInfo::Legal); setLegalizeVectorElementToDifferentSizeStrategy( G_ADD, 0, widenToLargerTypesUnsupportedOtherwise); As currently implemented here, vector types are legalized by first making the vector element size legal, followed by then making the number of lanes legal. The strategy to follow in the first step is set by a call to setLegalizeVectorElementToDifferentSizeStrategy, see example above. The strategy followed in the second step "moreToWiderTypesAndLessToWidest" (see code for its definition), indicating that vectors are widened to more elements so they map to natively supported vector widths, or when there isn't a legal wider vector, split the vector to map it to the widest vector supported. Therefore, for the above specification, some example legalizations are: * getAction({G_ADD, LLT::vector(3, 3)}) returns {WidenScalar, LLT::vector(3, 8)} * getAction({G_ADD, LLT::vector(3, 8)}) then returns {MoreElements, LLT::vector(8, 8)} * getAction({G_ADD, LLT::vector(20, 8)}) returns {FewerElements, LLT::vector(16, 8)} 2. Key implementation aspects. How to legalize a specific (operation, type index, size) tuple is represented by mapping intervals of integers representing a range of size types to an action to take, e.g.: setScalarAction({G_ADD, LLT:scalar(1)}, {{1, WidenScalar}, // bit sizes [ 1, 31[ {32, Legal}, // bit sizes [32, 33[ {33, WidenScalar}, // bit sizes [33, 64[ {64, Legal}, // bit sizes [64, 65[ {65, NarrowScalar} // bit sizes [65, +inf[ }); Please note that most of the code to do the actual lowering of non-power-of-2 sized types is currently missing, this is just trying to make it possible for targets to specify what is legal, and how non-legal types should be legalized. Probably quite a bit of further work is needed in the actual legalizing and the other passes in GlobalISel to support non-power-of-2 sized types. I hope the documentation in LegalizerInfo.h and the examples provided in the various {Target}LegalizerInfo.cpp and LegalizerInfoTest.cpp explains well enough how this is meant to be used. This drops the need for LLT::{half,double}...Size(). Differential Revision: https://reviews.llvm.org/D30529 llvm-svn: 317560
2017-11-07 11:34:34 +01:00
setLegalizeScalarToDifferentSizeStrategy(Op, 0, widen_8_16);
[ARM] GlobalISel: Add support for i32 modulo Add support for modulo for targets that have hardware division and for those that don't. When hardware division is not available, we have to choose the correct libcall to use. This is generally straightforward, except for AEABI. The AEABI variant is trickier than the other libcalls because it returns { quotient, remainder }, instead of just one value like the other libcalls that we've seen so far. Therefore, we need to use custom lowering for it. However, we don't want to have too much special code, so we refactor the target-independent code in the legalizer by adding a helper for replacing an instruction with a libcall. This helper is used by the legalizer itself when dealing with simple calls, and also by the custom ARM legalization for the more complicated AEABI divmod calls. llvm-svn: 305459
2017-06-15 12:53:31 +02:00
if (ST.hasDivideInARMMode())
setAction({Op, s32}, Lower);
[ARM] GlobalISel: Extract tiny helper. NFC Extract functionality for determining if the target uses AEABI. llvm-svn: 307145
2017-07-05 13:53:51 +02:00
else if (AEABI(ST))
[ARM] GlobalISel: Add support for i32 modulo Add support for modulo for targets that have hardware division and for those that don't. When hardware division is not available, we have to choose the correct libcall to use. This is generally straightforward, except for AEABI. The AEABI variant is trickier than the other libcalls because it returns { quotient, remainder }, instead of just one value like the other libcalls that we've seen so far. Therefore, we need to use custom lowering for it. However, we don't want to have too much special code, so we refactor the target-independent code in the legalizer by adding a helper for replacing an instruction with a libcall. This helper is used by the legalizer itself when dealing with simple calls, and also by the custom ARM legalization for the more complicated AEABI divmod calls. llvm-svn: 305459
2017-06-15 12:53:31 +02:00
setAction({Op, s32}, Custom);
else
setAction({Op, s32}, Libcall);
[ARM] GlobalISel: Support G_(S|U)REM for s8 and s16 Widen to s32, and then do whatever Lowering/Custom/Libcall action the subtarget wants. llvm-svn: 308285
2017-07-18 12:07:01 +02:00
}
[ARM] GlobalISel: Add support for i32 modulo Add support for modulo for targets that have hardware division and for those that don't. When hardware division is not available, we have to choose the correct libcall to use. This is generally straightforward, except for AEABI. The AEABI variant is trickier than the other libcalls because it returns { quotient, remainder }, instead of just one value like the other libcalls that we've seen so far. Therefore, we need to use custom lowering for it. However, we don't want to have too much special code, so we refactor the target-independent code in the legalizer by adding a helper for replacing an instruction with a libcall. This helper is used by the legalizer itself when dealing with simple calls, and also by the custom ARM legalization for the more complicated AEABI divmod calls. llvm-svn: 305459
2017-06-15 12:53:31 +02:00
[GlobalISel] Enable legalizing non-power-of-2 sized types. This changes the interface of how targets describe how to legalize, see the below description. 1. Interface for targets to describe how to legalize. In GlobalISel, the API in the LegalizerInfo class is the main interface for targets to specify which types are legal for which operations, and what to do to turn illegal type/operation combinations into legal ones. For each operation the type sizes that can be legalized without having to change the size of the type are specified with a call to setAction. This isn't different to how GlobalISel worked before. For example, for a target that supports 32 and 64 bit adds natively: for (auto Ty : {s32, s64}) setAction({G_ADD, 0, s32}, Legal); or for a target that needs a library call for a 32 bit division: setAction({G_SDIV, s32}, Libcall); The main conceptual change to the LegalizerInfo API, is in specifying how to legalize the type sizes for which a change of size is needed. For example, in the above example, how to specify how all types from i1 to i8388607 (apart from s32 and s64 which are legal) need to be legalized and expressed in terms of operations on the available legal sizes (again, i32 and i64 in this case). Before, the implementation only allowed specifying power-of-2-sized types (e.g. setAction({G_ADD, 0, s128}, NarrowScalar). A worse limitation was that if you'd wanted to specify how to legalize all the sized types as allowed by the LLVM-IR LangRef, i1 to i8388607, you'd have to call setAction 8388607-3 times and probably would need a lot of memory to store all of these specifications. Instead, the legalization actions that need to change the size of the type are specified now using a "SizeChangeStrategy". For example: setLegalizeScalarToDifferentSizeStrategy( G_ADD, 0, widenToLargerAndNarrowToLargest); This example indicates that for type sizes for which there is a larger size that can be legalized towards, do it by Widening the size. For example, G_ADD on s17 will be legalized by first doing WidenScalar to make it s32, after which it's legal. The "NarrowToLargest" indicates what to do if there is no larger size that can be legalized towards. E.g. G_ADD on s92 will be legalized by doing NarrowScalar to s64. Another example, taken from the ARM backend is: for (unsigned Op : {G_SDIV, G_UDIV}) { setLegalizeScalarToDifferentSizeStrategy(Op, 0, widenToLargerTypesUnsupportedOtherwise); if (ST.hasDivideInARMMode()) setAction({Op, s32}, Legal); else setAction({Op, s32}, Libcall); } For this example, G_SDIV on s8, on a target without a divide instruction, would be legalized by first doing action (WidenScalar, s32), followed by (Libcall, s32). The same principle is also followed for when the number of vector lanes on vector data types need to be changed, e.g.: setAction({G_ADD, LLT::vector(8, 8)}, LegalizerInfo::Legal); setAction({G_ADD, LLT::vector(16, 8)}, LegalizerInfo::Legal); setAction({G_ADD, LLT::vector(4, 16)}, LegalizerInfo::Legal); setAction({G_ADD, LLT::vector(8, 16)}, LegalizerInfo::Legal); setAction({G_ADD, LLT::vector(2, 32)}, LegalizerInfo::Legal); setAction({G_ADD, LLT::vector(4, 32)}, LegalizerInfo::Legal); setLegalizeVectorElementToDifferentSizeStrategy( G_ADD, 0, widenToLargerTypesUnsupportedOtherwise); As currently implemented here, vector types are legalized by first making the vector element size legal, followed by then making the number of lanes legal. The strategy to follow in the first step is set by a call to setLegalizeVectorElementToDifferentSizeStrategy, see example above. The strategy followed in the second step "moreToWiderTypesAndLessToWidest" (see code for its definition), indicating that vectors are widened to more elements so they map to natively supported vector widths, or when there isn't a legal wider vector, split the vector to map it to the widest vector supported. Therefore, for the above specification, some example legalizations are: * getAction({G_ADD, LLT::vector(3, 3)}) returns {WidenScalar, LLT::vector(3, 8)} * getAction({G_ADD, LLT::vector(3, 8)}) then returns {MoreElements, LLT::vector(8, 8)} * getAction({G_ADD, LLT::vector(20, 8)}) returns {FewerElements, LLT::vector(16, 8)} 2. Key implementation aspects. How to legalize a specific (operation, type index, size) tuple is represented by mapping intervals of integers representing a range of size types to an action to take, e.g.: setScalarAction({G_ADD, LLT:scalar(1)}, {{1, WidenScalar}, // bit sizes [ 1, 31[ {32, Legal}, // bit sizes [32, 33[ {33, WidenScalar}, // bit sizes [33, 64[ {64, Legal}, // bit sizes [64, 65[ {65, NarrowScalar} // bit sizes [65, +inf[ }); Please note that most of the code to do the actual lowering of non-power-of-2 sized types is currently missing, this is just trying to make it possible for targets to specify what is legal, and how non-legal types should be legalized. Probably quite a bit of further work is needed in the actual legalizing and the other passes in GlobalISel to support non-power-of-2 sized types. I hope the documentation in LegalizerInfo.h and the examples provided in the various {Target}LegalizerInfo.cpp and LegalizerInfoTest.cpp explains well enough how this is meant to be used. This drops the need for LLT::{half,double}...Size(). Differential Revision: https://reviews.llvm.org/D30529 llvm-svn: 317560
2017-11-07 11:34:34 +01:00
for (unsigned Op : {G_SEXT, G_ZEXT, G_ANYEXT}) {
[ARM] GlobalISel: Support i8/i16 ABI extensions At the moment, this means supporting the signext/zeroext attribute on the return type of the function. For function arguments, signext/zeroext should be handled by the caller, so there's nothing for us to do until we start lowering calls. Note that this does not include support for other extensions (i8 to i16), those will be added later. Differential Revision: https://reviews.llvm.org/D27705 llvm-svn: 293034
2017-01-25 09:10:40 +01:00
setAction({Op, s32}, Legal);
}
[ARM] GlobalISel: Mark shifts as legal for s32 The new legalize combiner introduces shifts all over the place, so we should support them sooner rather than later. llvm-svn: 315064
2017-10-06 16:30:05 +02:00
for (unsigned Op : {G_ASHR, G_LSHR, G_SHL})
setAction({Op, s32}, Legal);
[ARM] GlobalISel: Legalize G_GEP with 32-bit offsets At the moment we're only interested in GEPs for putting call parameters on the stack, so we'll stick to 32-bit offsets. llvm-svn: 296452
2017-02-28 10:02:42 +01:00
setAction({G_GEP, p0}, Legal);
setAction({G_GEP, 1, s32}, Legal);
[ARM] GlobalISel: Support G_SELECT for i32 * Mark as legal for (s32, i1, s32, s32) * Map everything into GPRs * Select to two instructions: a CMP of the condition against 0, to set the flags, and a MOVCCr to select between the two inputs based on the flags that we've just set llvm-svn: 306382
2017-06-27 11:19:51 +02:00
setAction({G_SELECT, s32}, Legal);
[ARM] GlobalISel: Support G_SELECT for pointers All we need to do is mark it as legal, otherwise it's just like s32. llvm-svn: 306390
2017-06-27 12:29:50 +02:00
setAction({G_SELECT, p0}, Legal);
[ARM] GlobalISel: Support G_SELECT for i32 * Mark as legal for (s32, i1, s32, s32) * Map everything into GPRs * Select to two instructions: a CMP of the condition against 0, to set the flags, and a MOVCCr to select between the two inputs based on the flags that we've just set llvm-svn: 306382
2017-06-27 11:19:51 +02:00
setAction({G_SELECT, 1, s1}, Legal);
[ARM] GlobalISel: Support G_BRCOND Insert a TSTri to set the flags and a Bcc to branch based on their values. This is a bit inefficient in the (common) cases where the condition for the branch comes from a compare right before the branch, since we set the flags both as part of the compare lowering and as part of the branch lowering. We're going to live with that until we settle on a principled way to handle this kind of situation, which occurs with other patterns as well (combines might be the way forward here). llvm-svn: 308009
2017-07-14 11:46:06 +02:00
setAction({G_BRCOND, s1}, Legal);
[ARM] GlobalISel: Legalize 32-bit constants llvm-svn: 296468
2017-02-28 12:33:46 +01:00
setAction({G_CONSTANT, s32}, Legal);
[GlobalISel] Enable legalizing non-power-of-2 sized types. This changes the interface of how targets describe how to legalize, see the below description. 1. Interface for targets to describe how to legalize. In GlobalISel, the API in the LegalizerInfo class is the main interface for targets to specify which types are legal for which operations, and what to do to turn illegal type/operation combinations into legal ones. For each operation the type sizes that can be legalized without having to change the size of the type are specified with a call to setAction. This isn't different to how GlobalISel worked before. For example, for a target that supports 32 and 64 bit adds natively: for (auto Ty : {s32, s64}) setAction({G_ADD, 0, s32}, Legal); or for a target that needs a library call for a 32 bit division: setAction({G_SDIV, s32}, Libcall); The main conceptual change to the LegalizerInfo API, is in specifying how to legalize the type sizes for which a change of size is needed. For example, in the above example, how to specify how all types from i1 to i8388607 (apart from s32 and s64 which are legal) need to be legalized and expressed in terms of operations on the available legal sizes (again, i32 and i64 in this case). Before, the implementation only allowed specifying power-of-2-sized types (e.g. setAction({G_ADD, 0, s128}, NarrowScalar). A worse limitation was that if you'd wanted to specify how to legalize all the sized types as allowed by the LLVM-IR LangRef, i1 to i8388607, you'd have to call setAction 8388607-3 times and probably would need a lot of memory to store all of these specifications. Instead, the legalization actions that need to change the size of the type are specified now using a "SizeChangeStrategy". For example: setLegalizeScalarToDifferentSizeStrategy( G_ADD, 0, widenToLargerAndNarrowToLargest); This example indicates that for type sizes for which there is a larger size that can be legalized towards, do it by Widening the size. For example, G_ADD on s17 will be legalized by first doing WidenScalar to make it s32, after which it's legal. The "NarrowToLargest" indicates what to do if there is no larger size that can be legalized towards. E.g. G_ADD on s92 will be legalized by doing NarrowScalar to s64. Another example, taken from the ARM backend is: for (unsigned Op : {G_SDIV, G_UDIV}) { setLegalizeScalarToDifferentSizeStrategy(Op, 0, widenToLargerTypesUnsupportedOtherwise); if (ST.hasDivideInARMMode()) setAction({Op, s32}, Legal); else setAction({Op, s32}, Libcall); } For this example, G_SDIV on s8, on a target without a divide instruction, would be legalized by first doing action (WidenScalar, s32), followed by (Libcall, s32). The same principle is also followed for when the number of vector lanes on vector data types need to be changed, e.g.: setAction({G_ADD, LLT::vector(8, 8)}, LegalizerInfo::Legal); setAction({G_ADD, LLT::vector(16, 8)}, LegalizerInfo::Legal); setAction({G_ADD, LLT::vector(4, 16)}, LegalizerInfo::Legal); setAction({G_ADD, LLT::vector(8, 16)}, LegalizerInfo::Legal); setAction({G_ADD, LLT::vector(2, 32)}, LegalizerInfo::Legal); setAction({G_ADD, LLT::vector(4, 32)}, LegalizerInfo::Legal); setLegalizeVectorElementToDifferentSizeStrategy( G_ADD, 0, widenToLargerTypesUnsupportedOtherwise); As currently implemented here, vector types are legalized by first making the vector element size legal, followed by then making the number of lanes legal. The strategy to follow in the first step is set by a call to setLegalizeVectorElementToDifferentSizeStrategy, see example above. The strategy followed in the second step "moreToWiderTypesAndLessToWidest" (see code for its definition), indicating that vectors are widened to more elements so they map to natively supported vector widths, or when there isn't a legal wider vector, split the vector to map it to the widest vector supported. Therefore, for the above specification, some example legalizations are: * getAction({G_ADD, LLT::vector(3, 3)}) returns {WidenScalar, LLT::vector(3, 8)} * getAction({G_ADD, LLT::vector(3, 8)}) then returns {MoreElements, LLT::vector(8, 8)} * getAction({G_ADD, LLT::vector(20, 8)}) returns {FewerElements, LLT::vector(16, 8)} 2. Key implementation aspects. How to legalize a specific (operation, type index, size) tuple is represented by mapping intervals of integers representing a range of size types to an action to take, e.g.: setScalarAction({G_ADD, LLT:scalar(1)}, {{1, WidenScalar}, // bit sizes [ 1, 31[ {32, Legal}, // bit sizes [32, 33[ {33, WidenScalar}, // bit sizes [33, 64[ {64, Legal}, // bit sizes [64, 65[ {65, NarrowScalar} // bit sizes [65, +inf[ }); Please note that most of the code to do the actual lowering of non-power-of-2 sized types is currently missing, this is just trying to make it possible for targets to specify what is legal, and how non-legal types should be legalized. Probably quite a bit of further work is needed in the actual legalizing and the other passes in GlobalISel to support non-power-of-2 sized types. I hope the documentation in LegalizerInfo.h and the examples provided in the various {Target}LegalizerInfo.cpp and LegalizerInfoTest.cpp explains well enough how this is meant to be used. This drops the need for LLT::{half,double}...Size(). Differential Revision: https://reviews.llvm.org/D30529 llvm-svn: 317560
2017-11-07 11:34:34 +01:00
setLegalizeScalarToDifferentSizeStrategy(G_CONSTANT, 0, widen_1_8_16);
[ARM] GlobalISel: Legalize 32-bit constants llvm-svn: 296468
2017-02-28 12:33:46 +01:00
[ARM] GlobalISel: Support G_ICMP for i32 and pointers Add support throughout the pipeline: - mark as legal for s32 and pointers - map to GPRs - lower to a sequence of instructions, which moves 0 or 1 into the result register based on the flags set by a CMPrr We have copied from FastISel a helper function which maps CmpInst predicates into ARMCC codes. Ideally, we should be able to move it somewhere that both FastISel and GlobalISel can use. llvm-svn: 305672
2017-06-19 11:40:51 +02:00
setAction({G_ICMP, s1}, Legal);
[GlobalISel] Enable legalizing non-power-of-2 sized types. This changes the interface of how targets describe how to legalize, see the below description. 1. Interface for targets to describe how to legalize. In GlobalISel, the API in the LegalizerInfo class is the main interface for targets to specify which types are legal for which operations, and what to do to turn illegal type/operation combinations into legal ones. For each operation the type sizes that can be legalized without having to change the size of the type are specified with a call to setAction. This isn't different to how GlobalISel worked before. For example, for a target that supports 32 and 64 bit adds natively: for (auto Ty : {s32, s64}) setAction({G_ADD, 0, s32}, Legal); or for a target that needs a library call for a 32 bit division: setAction({G_SDIV, s32}, Libcall); The main conceptual change to the LegalizerInfo API, is in specifying how to legalize the type sizes for which a change of size is needed. For example, in the above example, how to specify how all types from i1 to i8388607 (apart from s32 and s64 which are legal) need to be legalized and expressed in terms of operations on the available legal sizes (again, i32 and i64 in this case). Before, the implementation only allowed specifying power-of-2-sized types (e.g. setAction({G_ADD, 0, s128}, NarrowScalar). A worse limitation was that if you'd wanted to specify how to legalize all the sized types as allowed by the LLVM-IR LangRef, i1 to i8388607, you'd have to call setAction 8388607-3 times and probably would need a lot of memory to store all of these specifications. Instead, the legalization actions that need to change the size of the type are specified now using a "SizeChangeStrategy". For example: setLegalizeScalarToDifferentSizeStrategy( G_ADD, 0, widenToLargerAndNarrowToLargest); This example indicates that for type sizes for which there is a larger size that can be legalized towards, do it by Widening the size. For example, G_ADD on s17 will be legalized by first doing WidenScalar to make it s32, after which it's legal. The "NarrowToLargest" indicates what to do if there is no larger size that can be legalized towards. E.g. G_ADD on s92 will be legalized by doing NarrowScalar to s64. Another example, taken from the ARM backend is: for (unsigned Op : {G_SDIV, G_UDIV}) { setLegalizeScalarToDifferentSizeStrategy(Op, 0, widenToLargerTypesUnsupportedOtherwise); if (ST.hasDivideInARMMode()) setAction({Op, s32}, Legal); else setAction({Op, s32}, Libcall); } For this example, G_SDIV on s8, on a target without a divide instruction, would be legalized by first doing action (WidenScalar, s32), followed by (Libcall, s32). The same principle is also followed for when the number of vector lanes on vector data types need to be changed, e.g.: setAction({G_ADD, LLT::vector(8, 8)}, LegalizerInfo::Legal); setAction({G_ADD, LLT::vector(16, 8)}, LegalizerInfo::Legal); setAction({G_ADD, LLT::vector(4, 16)}, LegalizerInfo::Legal); setAction({G_ADD, LLT::vector(8, 16)}, LegalizerInfo::Legal); setAction({G_ADD, LLT::vector(2, 32)}, LegalizerInfo::Legal); setAction({G_ADD, LLT::vector(4, 32)}, LegalizerInfo::Legal); setLegalizeVectorElementToDifferentSizeStrategy( G_ADD, 0, widenToLargerTypesUnsupportedOtherwise); As currently implemented here, vector types are legalized by first making the vector element size legal, followed by then making the number of lanes legal. The strategy to follow in the first step is set by a call to setLegalizeVectorElementToDifferentSizeStrategy, see example above. The strategy followed in the second step "moreToWiderTypesAndLessToWidest" (see code for its definition), indicating that vectors are widened to more elements so they map to natively supported vector widths, or when there isn't a legal wider vector, split the vector to map it to the widest vector supported. Therefore, for the above specification, some example legalizations are: * getAction({G_ADD, LLT::vector(3, 3)}) returns {WidenScalar, LLT::vector(3, 8)} * getAction({G_ADD, LLT::vector(3, 8)}) then returns {MoreElements, LLT::vector(8, 8)} * getAction({G_ADD, LLT::vector(20, 8)}) returns {FewerElements, LLT::vector(16, 8)} 2. Key implementation aspects. How to legalize a specific (operation, type index, size) tuple is represented by mapping intervals of integers representing a range of size types to an action to take, e.g.: setScalarAction({G_ADD, LLT:scalar(1)}, {{1, WidenScalar}, // bit sizes [ 1, 31[ {32, Legal}, // bit sizes [32, 33[ {33, WidenScalar}, // bit sizes [33, 64[ {64, Legal}, // bit sizes [64, 65[ {65, NarrowScalar} // bit sizes [65, +inf[ }); Please note that most of the code to do the actual lowering of non-power-of-2 sized types is currently missing, this is just trying to make it possible for targets to specify what is legal, and how non-legal types should be legalized. Probably quite a bit of further work is needed in the actual legalizing and the other passes in GlobalISel to support non-power-of-2 sized types. I hope the documentation in LegalizerInfo.h and the examples provided in the various {Target}LegalizerInfo.cpp and LegalizerInfoTest.cpp explains well enough how this is meant to be used. This drops the need for LLT::{half,double}...Size(). Differential Revision: https://reviews.llvm.org/D30529 llvm-svn: 317560
2017-11-07 11:34:34 +01:00
setLegalizeScalarToDifferentSizeStrategy(G_ICMP, 1,
widenToLargerTypesUnsupportedOtherwise);
[ARM] GlobalISel: Support G_ICMP for i32 and pointers Add support throughout the pipeline: - mark as legal for s32 and pointers - map to GPRs - lower to a sequence of instructions, which moves 0 or 1 into the result register based on the flags set by a CMPrr We have copied from FastISel a helper function which maps CmpInst predicates into ARMCC codes. Ideally, we should be able to move it somewhere that both FastISel and GlobalISel can use. llvm-svn: 305672
2017-06-19 11:40:51 +02:00
for (auto Ty : {s32, p0})
setAction({G_ICMP, 1, Ty}, Legal);
[ARM] GlobalISel: Support frem for 32-bit values Legalize to a libcall. On this occasion, also start allowing soft float subtargets. For the moment G_FREM is the only legal floating point operation for them. llvm-svn: 299753
2017-04-07 11:41:39 +02:00
if (!ST.useSoftFloat() && ST.hasVFP2()) {
[GlobalISel|ARM] : Allow legalizing G_FSUB Adding support for VSUB. Reviewed by: @rovka Differential Revision: https://reviews.llvm.org/D39261 llvm-svn: 316902
2017-10-30 14:51:56 +01:00
for (unsigned BinOp : {G_FADD, G_FSUB})
for (auto Ty : {s32, s64})
setAction({BinOp, Ty}, Legal);
[ARM] GlobalISel: Use Subtarget in Legalizer Start using the Subtarget to make decisions about what's legal. In particular, we only mark floating point operations as legal if we have VFP2, which is something we should've done from the very start. llvm-svn: 295439
2017-02-17 12:25:17 +01:00
setAction({G_LOAD, s64}, Legal);
[ARM] GlobalISel: Legalize stores Allow the same types that we allow for loads. llvm-svn: 296108
2017-02-24 12:28:24 +01:00
setAction({G_STORE, s64}, Legal);
[ARM] GlobalISel: Legalize G_FCMP for s32 This covers both hard and soft float. Hard float is easy, since it's just Legal. Soft float is more involved, because there are several different ways to handle it based on the predicate: one and ueq need not only one, but two libcalls to get a result. Furthermore, we have large differences between the values returned by the AEABI and GNU functions. AEABI functions return a nice 1 or 0 representing true and respectively false. GNU functions generally return a value that needs to be compared against 0 (e.g. for ogt, the value returned by the libcall is > 0 for true). We could introduce redundant comparisons for AEABI as well, but they don't seem easy to remove afterwards, so we do different processing based on whether or not the result really needs to be compared against something (and just truncate if it doesn't). llvm-svn: 307243
2017-07-06 11:09:33 +02:00
setAction({G_FCMP, s1}, Legal);
setAction({G_FCMP, 1, s32}, Legal);
[ARM] GlobalISel: Legalize s64 G_FCMP Same as the s32 version, for both hard and soft float. llvm-svn: 307633
2017-07-11 10:50:01 +02:00
setAction({G_FCMP, 1, s64}, Legal);
GlobalISel: Allow legalizing G_FADD to a libcall Use the same handling in the generic legalizer code as for the other libcalls (G_FREM, G_FPOW). Enable it on ARM for float and double so we can test it. llvm-svn: 299931
2017-04-11 12:52:34 +02:00
} else {
[GlobalISel|ARM] : Allow legalizing G_FSUB Adding support for VSUB. Reviewed by: @rovka Differential Revision: https://reviews.llvm.org/D39261 llvm-svn: 316902
2017-10-30 14:51:56 +01:00
for (unsigned BinOp : {G_FADD, G_FSUB})
for (auto Ty : {s32, s64})
setAction({BinOp, Ty}, Libcall);
[ARM] GlobalISel: Legalize G_FCMP for s32 This covers both hard and soft float. Hard float is easy, since it's just Legal. Soft float is more involved, because there are several different ways to handle it based on the predicate: one and ueq need not only one, but two libcalls to get a result. Furthermore, we have large differences between the values returned by the AEABI and GNU functions. AEABI functions return a nice 1 or 0 representing true and respectively false. GNU functions generally return a value that needs to be compared against 0 (e.g. for ogt, the value returned by the libcall is > 0 for true). We could introduce redundant comparisons for AEABI as well, but they don't seem easy to remove afterwards, so we do different processing based on whether or not the result really needs to be compared against something (and just truncate if it doesn't). llvm-svn: 307243
2017-07-06 11:09:33 +02:00
setAction({G_FCMP, s1}, Legal);
setAction({G_FCMP, 1, s32}, Custom);
[ARM] GlobalISel: Legalize s64 G_FCMP Same as the s32 version, for both hard and soft float. llvm-svn: 307633
2017-07-11 10:50:01 +02:00
setAction({G_FCMP, 1, s64}, Custom);
[ARM] GlobalISel: Legalize G_FCMP for s32 This covers both hard and soft float. Hard float is easy, since it's just Legal. Soft float is more involved, because there are several different ways to handle it based on the predicate: one and ueq need not only one, but two libcalls to get a result. Furthermore, we have large differences between the values returned by the AEABI and GNU functions. AEABI functions return a nice 1 or 0 representing true and respectively false. GNU functions generally return a value that needs to be compared against 0 (e.g. for ogt, the value returned by the libcall is > 0 for true). We could introduce redundant comparisons for AEABI as well, but they don't seem easy to remove afterwards, so we do different processing based on whether or not the result really needs to be compared against something (and just truncate if it doesn't). llvm-svn: 307243
2017-07-06 11:09:33 +02:00
if (AEABI(ST))
setFCmpLibcallsAEABI();
else
setFCmpLibcallsGNU();
[ARM] GlobalISel: Use Subtarget in Legalizer Start using the Subtarget to make decisions about what's legal. In particular, we only mark floating point operations as legal if we have VFP2, which is something we should've done from the very start. llvm-svn: 295439
2017-02-17 12:25:17 +01:00
}
[ARM] GlobalISel: Add FPR reg bank Add a register bank for floating point values and select simple instructions using them (add, copies from GPR). This assumes that the hardware can cope with a single precision add (VADDS) instruction, so the legalizer will treat G_FADD as legal and the instruction selector will refuse to select if the hardware doesn't support it. In the future we'll want to be more careful about this, and legalize to libcalls if we have to use soft float. llvm-svn: 294442
2017-02-08 14:23:04 +01:00
[ARM] GlobalISel: Support G_FPOW for float and double Legalize to a libcall. llvm-svn: 299841
2017-04-10 11:27:39 +02:00
for (unsigned Op : {G_FREM, G_FPOW})
for (auto Ty : {s32, s64})
setAction({Op, Ty}, Libcall);
[ARM] GlobalISel: Support frem for 32-bit values Legalize to a libcall. On this occasion, also start allowing soft float subtargets. For the moment G_FREM is the only legal floating point operation for them. llvm-svn: 299753
2017-04-07 11:41:39 +02:00
[ARM] Add plumbing for GlobalISel Add GlobalISel skeleton, up to the point where we can select a ret void. llvm-svn: 286573
2016-11-11 09:27:37 +01:00
computeTables();
}
[ARM] GlobalISel: Legalize s8 and s16 G_(S|U)DIV We have to widen the operands to 32 bits and then we can either use hardware division if it is available or lower to a libcall otherwise. At the moment it is not enough to set the Legalizer action to WidenScalar, since for libcalls it won't know what to do (it won't be able to find what size to widen to, because it will find Libcall and not Legal for 32 bits). To hack around this limitation, we request Custom lowering, and as part of that we widen first and then we run another legalizeInstrStep on the widened DIV. llvm-svn: 301166
2017-04-24 11:12:19 +02:00
[ARM] GlobalISel: Legalize G_FCMP for s32 This covers both hard and soft float. Hard float is easy, since it's just Legal. Soft float is more involved, because there are several different ways to handle it based on the predicate: one and ueq need not only one, but two libcalls to get a result. Furthermore, we have large differences between the values returned by the AEABI and GNU functions. AEABI functions return a nice 1 or 0 representing true and respectively false. GNU functions generally return a value that needs to be compared against 0 (e.g. for ogt, the value returned by the libcall is > 0 for true). We could introduce redundant comparisons for AEABI as well, but they don't seem easy to remove afterwards, so we do different processing based on whether or not the result really needs to be compared against something (and just truncate if it doesn't). llvm-svn: 307243
2017-07-06 11:09:33 +02:00
void ARMLegalizerInfo::setFCmpLibcallsAEABI() {
// FCMP_TRUE and FCMP_FALSE don't need libcalls, they should be
// default-initialized.
FCmp32Libcalls.resize(CmpInst::LAST_FCMP_PREDICATE + 1);
FCmp32Libcalls[CmpInst::FCMP_OEQ] = {
{RTLIB::OEQ_F32, CmpInst::BAD_ICMP_PREDICATE}};
FCmp32Libcalls[CmpInst::FCMP_OGE] = {
{RTLIB::OGE_F32, CmpInst::BAD_ICMP_PREDICATE}};
FCmp32Libcalls[CmpInst::FCMP_OGT] = {
{RTLIB::OGT_F32, CmpInst::BAD_ICMP_PREDICATE}};
FCmp32Libcalls[CmpInst::FCMP_OLE] = {
{RTLIB::OLE_F32, CmpInst::BAD_ICMP_PREDICATE}};
FCmp32Libcalls[CmpInst::FCMP_OLT] = {
{RTLIB::OLT_F32, CmpInst::BAD_ICMP_PREDICATE}};
FCmp32Libcalls[CmpInst::FCMP_ORD] = {{RTLIB::O_F32, CmpInst::ICMP_EQ}};
FCmp32Libcalls[CmpInst::FCMP_UGE] = {{RTLIB::OLT_F32, CmpInst::ICMP_EQ}};
FCmp32Libcalls[CmpInst::FCMP_UGT] = {{RTLIB::OLE_F32, CmpInst::ICMP_EQ}};
FCmp32Libcalls[CmpInst::FCMP_ULE] = {{RTLIB::OGT_F32, CmpInst::ICMP_EQ}};
FCmp32Libcalls[CmpInst::FCMP_ULT] = {{RTLIB::OGE_F32, CmpInst::ICMP_EQ}};
FCmp32Libcalls[CmpInst::FCMP_UNE] = {{RTLIB::UNE_F32, CmpInst::ICMP_EQ}};
FCmp32Libcalls[CmpInst::FCMP_UNO] = {
{RTLIB::UO_F32, CmpInst::BAD_ICMP_PREDICATE}};
FCmp32Libcalls[CmpInst::FCMP_ONE] = {
{RTLIB::OGT_F32, CmpInst::BAD_ICMP_PREDICATE},
{RTLIB::OLT_F32, CmpInst::BAD_ICMP_PREDICATE}};
FCmp32Libcalls[CmpInst::FCMP_UEQ] = {
{RTLIB::OEQ_F32, CmpInst::BAD_ICMP_PREDICATE},
{RTLIB::UO_F32, CmpInst::BAD_ICMP_PREDICATE}};
[ARM] GlobalISel: Legalize s64 G_FCMP Same as the s32 version, for both hard and soft float. llvm-svn: 307633
2017-07-11 10:50:01 +02:00
FCmp64Libcalls.resize(CmpInst::LAST_FCMP_PREDICATE + 1);
FCmp64Libcalls[CmpInst::FCMP_OEQ] = {
{RTLIB::OEQ_F64, CmpInst::BAD_ICMP_PREDICATE}};
FCmp64Libcalls[CmpInst::FCMP_OGE] = {
{RTLIB::OGE_F64, CmpInst::BAD_ICMP_PREDICATE}};
FCmp64Libcalls[CmpInst::FCMP_OGT] = {
{RTLIB::OGT_F64, CmpInst::BAD_ICMP_PREDICATE}};
FCmp64Libcalls[CmpInst::FCMP_OLE] = {
{RTLIB::OLE_F64, CmpInst::BAD_ICMP_PREDICATE}};
FCmp64Libcalls[CmpInst::FCMP_OLT] = {
{RTLIB::OLT_F64, CmpInst::BAD_ICMP_PREDICATE}};
FCmp64Libcalls[CmpInst::FCMP_ORD] = {{RTLIB::O_F64, CmpInst::ICMP_EQ}};
FCmp64Libcalls[CmpInst::FCMP_UGE] = {{RTLIB::OLT_F64, CmpInst::ICMP_EQ}};
FCmp64Libcalls[CmpInst::FCMP_UGT] = {{RTLIB::OLE_F64, CmpInst::ICMP_EQ}};
FCmp64Libcalls[CmpInst::FCMP_ULE] = {{RTLIB::OGT_F64, CmpInst::ICMP_EQ}};
FCmp64Libcalls[CmpInst::FCMP_ULT] = {{RTLIB::OGE_F64, CmpInst::ICMP_EQ}};
FCmp64Libcalls[CmpInst::FCMP_UNE] = {{RTLIB::UNE_F64, CmpInst::ICMP_EQ}};
FCmp64Libcalls[CmpInst::FCMP_UNO] = {
{RTLIB::UO_F64, CmpInst::BAD_ICMP_PREDICATE}};
FCmp64Libcalls[CmpInst::FCMP_ONE] = {
{RTLIB::OGT_F64, CmpInst::BAD_ICMP_PREDICATE},
{RTLIB::OLT_F64, CmpInst::BAD_ICMP_PREDICATE}};
FCmp64Libcalls[CmpInst::FCMP_UEQ] = {
{RTLIB::OEQ_F64, CmpInst::BAD_ICMP_PREDICATE},
{RTLIB::UO_F64, CmpInst::BAD_ICMP_PREDICATE}};
[ARM] GlobalISel: Legalize G_FCMP for s32 This covers both hard and soft float. Hard float is easy, since it's just Legal. Soft float is more involved, because there are several different ways to handle it based on the predicate: one and ueq need not only one, but two libcalls to get a result. Furthermore, we have large differences between the values returned by the AEABI and GNU functions. AEABI functions return a nice 1 or 0 representing true and respectively false. GNU functions generally return a value that needs to be compared against 0 (e.g. for ogt, the value returned by the libcall is > 0 for true). We could introduce redundant comparisons for AEABI as well, but they don't seem easy to remove afterwards, so we do different processing based on whether or not the result really needs to be compared against something (and just truncate if it doesn't). llvm-svn: 307243
2017-07-06 11:09:33 +02:00
}
void ARMLegalizerInfo::setFCmpLibcallsGNU() {
// FCMP_TRUE and FCMP_FALSE don't need libcalls, they should be
// default-initialized.
FCmp32Libcalls.resize(CmpInst::LAST_FCMP_PREDICATE + 1);
FCmp32Libcalls[CmpInst::FCMP_OEQ] = {{RTLIB::OEQ_F32, CmpInst::ICMP_EQ}};
FCmp32Libcalls[CmpInst::FCMP_OGE] = {{RTLIB::OGE_F32, CmpInst::ICMP_SGE}};
FCmp32Libcalls[CmpInst::FCMP_OGT] = {{RTLIB::OGT_F32, CmpInst::ICMP_SGT}};
FCmp32Libcalls[CmpInst::FCMP_OLE] = {{RTLIB::OLE_F32, CmpInst::ICMP_SLE}};
FCmp32Libcalls[CmpInst::FCMP_OLT] = {{RTLIB::OLT_F32, CmpInst::ICMP_SLT}};
FCmp32Libcalls[CmpInst::FCMP_ORD] = {{RTLIB::O_F32, CmpInst::ICMP_EQ}};
FCmp32Libcalls[CmpInst::FCMP_UGE] = {{RTLIB::OLT_F32, CmpInst::ICMP_SGE}};
FCmp32Libcalls[CmpInst::FCMP_UGT] = {{RTLIB::OLE_F32, CmpInst::ICMP_SGT}};
FCmp32Libcalls[CmpInst::FCMP_ULE] = {{RTLIB::OGT_F32, CmpInst::ICMP_SLE}};
FCmp32Libcalls[CmpInst::FCMP_ULT] = {{RTLIB::OGE_F32, CmpInst::ICMP_SLT}};
FCmp32Libcalls[CmpInst::FCMP_UNE] = {{RTLIB::UNE_F32, CmpInst::ICMP_NE}};
FCmp32Libcalls[CmpInst::FCMP_UNO] = {{RTLIB::UO_F32, CmpInst::ICMP_NE}};
FCmp32Libcalls[CmpInst::FCMP_ONE] = {{RTLIB::OGT_F32, CmpInst::ICMP_SGT},
{RTLIB::OLT_F32, CmpInst::ICMP_SLT}};
FCmp32Libcalls[CmpInst::FCMP_UEQ] = {{RTLIB::OEQ_F32, CmpInst::ICMP_EQ},
{RTLIB::UO_F32, CmpInst::ICMP_NE}};
[ARM] GlobalISel: Legalize s64 G_FCMP Same as the s32 version, for both hard and soft float. llvm-svn: 307633
2017-07-11 10:50:01 +02:00
FCmp64Libcalls.resize(CmpInst::LAST_FCMP_PREDICATE + 1);
FCmp64Libcalls[CmpInst::FCMP_OEQ] = {{RTLIB::OEQ_F64, CmpInst::ICMP_EQ}};
FCmp64Libcalls[CmpInst::FCMP_OGE] = {{RTLIB::OGE_F64, CmpInst::ICMP_SGE}};
FCmp64Libcalls[CmpInst::FCMP_OGT] = {{RTLIB::OGT_F64, CmpInst::ICMP_SGT}};
FCmp64Libcalls[CmpInst::FCMP_OLE] = {{RTLIB::OLE_F64, CmpInst::ICMP_SLE}};
FCmp64Libcalls[CmpInst::FCMP_OLT] = {{RTLIB::OLT_F64, CmpInst::ICMP_SLT}};
FCmp64Libcalls[CmpInst::FCMP_ORD] = {{RTLIB::O_F64, CmpInst::ICMP_EQ}};
FCmp64Libcalls[CmpInst::FCMP_UGE] = {{RTLIB::OLT_F64, CmpInst::ICMP_SGE}};
FCmp64Libcalls[CmpInst::FCMP_UGT] = {{RTLIB::OLE_F64, CmpInst::ICMP_SGT}};
FCmp64Libcalls[CmpInst::FCMP_ULE] = {{RTLIB::OGT_F64, CmpInst::ICMP_SLE}};
FCmp64Libcalls[CmpInst::FCMP_ULT] = {{RTLIB::OGE_F64, CmpInst::ICMP_SLT}};
FCmp64Libcalls[CmpInst::FCMP_UNE] = {{RTLIB::UNE_F64, CmpInst::ICMP_NE}};
FCmp64Libcalls[CmpInst::FCMP_UNO] = {{RTLIB::UO_F64, CmpInst::ICMP_NE}};
FCmp64Libcalls[CmpInst::FCMP_ONE] = {{RTLIB::OGT_F64, CmpInst::ICMP_SGT},
{RTLIB::OLT_F64, CmpInst::ICMP_SLT}};
FCmp64Libcalls[CmpInst::FCMP_UEQ] = {{RTLIB::OEQ_F64, CmpInst::ICMP_EQ},
{RTLIB::UO_F64, CmpInst::ICMP_NE}};
[ARM] GlobalISel: Legalize G_FCMP for s32 This covers both hard and soft float. Hard float is easy, since it's just Legal. Soft float is more involved, because there are several different ways to handle it based on the predicate: one and ueq need not only one, but two libcalls to get a result. Furthermore, we have large differences between the values returned by the AEABI and GNU functions. AEABI functions return a nice 1 or 0 representing true and respectively false. GNU functions generally return a value that needs to be compared against 0 (e.g. for ogt, the value returned by the libcall is > 0 for true). We could introduce redundant comparisons for AEABI as well, but they don't seem easy to remove afterwards, so we do different processing based on whether or not the result really needs to be compared against something (and just truncate if it doesn't). llvm-svn: 307243
2017-07-06 11:09:33 +02:00
}
ARMLegalizerInfo::FCmpLibcallsList
[ARM] GlobalISel: Legalize s64 G_FCMP Same as the s32 version, for both hard and soft float. llvm-svn: 307633
2017-07-11 10:50:01 +02:00
ARMLegalizerInfo::getFCmpLibcalls(CmpInst::Predicate Predicate,
unsigned Size) const {
[ARM] GlobalISel: Legalize G_FCMP for s32 This covers both hard and soft float. Hard float is easy, since it's just Legal. Soft float is more involved, because there are several different ways to handle it based on the predicate: one and ueq need not only one, but two libcalls to get a result. Furthermore, we have large differences between the values returned by the AEABI and GNU functions. AEABI functions return a nice 1 or 0 representing true and respectively false. GNU functions generally return a value that needs to be compared against 0 (e.g. for ogt, the value returned by the libcall is > 0 for true). We could introduce redundant comparisons for AEABI as well, but they don't seem easy to remove afterwards, so we do different processing based on whether or not the result really needs to be compared against something (and just truncate if it doesn't). llvm-svn: 307243
2017-07-06 11:09:33 +02:00
assert(CmpInst::isFPPredicate(Predicate) && "Unsupported FCmp predicate");
[ARM] GlobalISel: Legalize s64 G_FCMP Same as the s32 version, for both hard and soft float. llvm-svn: 307633
2017-07-11 10:50:01 +02:00
if (Size == 32)
return FCmp32Libcalls[Predicate];
if (Size == 64)
return FCmp64Libcalls[Predicate];
llvm_unreachable("Unsupported size for FCmp predicate");
[ARM] GlobalISel: Legalize G_FCMP for s32 This covers both hard and soft float. Hard float is easy, since it's just Legal. Soft float is more involved, because there are several different ways to handle it based on the predicate: one and ueq need not only one, but two libcalls to get a result. Furthermore, we have large differences between the values returned by the AEABI and GNU functions. AEABI functions return a nice 1 or 0 representing true and respectively false. GNU functions generally return a value that needs to be compared against 0 (e.g. for ogt, the value returned by the libcall is > 0 for true). We could introduce redundant comparisons for AEABI as well, but they don't seem easy to remove afterwards, so we do different processing based on whether or not the result really needs to be compared against something (and just truncate if it doesn't). llvm-svn: 307243
2017-07-06 11:09:33 +02:00
}
[ARM] GlobalISel: Legalize s8 and s16 G_(S|U)DIV We have to widen the operands to 32 bits and then we can either use hardware division if it is available or lower to a libcall otherwise. At the moment it is not enough to set the Legalizer action to WidenScalar, since for libcalls it won't know what to do (it won't be able to find what size to widen to, because it will find Libcall and not Legal for 32 bits). To hack around this limitation, we request Custom lowering, and as part of that we widen first and then we run another legalizeInstrStep on the widened DIV. llvm-svn: 301166
2017-04-24 11:12:19 +02:00
bool ARMLegalizerInfo::legalizeCustom(MachineInstr &MI,
MachineRegisterInfo &MRI,
MachineIRBuilder &MIRBuilder) const {
using namespace TargetOpcode;
[GlobalISel] Refactor Legalizer helpers for libcalls We used to have a helper that replaced an instruction with a libcall. That turns out to be too aggressive, since sometimes we need to replace the instruction with at least two libcalls. Therefore, change our existing helper to only create the libcall and leave the instruction removal as a separate step. Also rename the helper accordingly. llvm-svn: 307149
2017-07-05 14:57:24 +02:00
MIRBuilder.setInstr(MI);
[ARM] GlobalISel: Legalize s8 and s16 G_(S|U)DIV We have to widen the operands to 32 bits and then we can either use hardware division if it is available or lower to a libcall otherwise. At the moment it is not enough to set the Legalizer action to WidenScalar, since for libcalls it won't know what to do (it won't be able to find what size to widen to, because it will find Libcall and not Legal for 32 bits). To hack around this limitation, we request Custom lowering, and as part of that we widen first and then we run another legalizeInstrStep on the widened DIV. llvm-svn: 301166
2017-04-24 11:12:19 +02:00
switch (MI.getOpcode()) {
default:
return false;
[ARM] GlobalISel: Add support for i32 modulo Add support for modulo for targets that have hardware division and for those that don't. When hardware division is not available, we have to choose the correct libcall to use. This is generally straightforward, except for AEABI. The AEABI variant is trickier than the other libcalls because it returns { quotient, remainder }, instead of just one value like the other libcalls that we've seen so far. Therefore, we need to use custom lowering for it. However, we don't want to have too much special code, so we refactor the target-independent code in the legalizer by adding a helper for replacing an instruction with a libcall. This helper is used by the legalizer itself when dealing with simple calls, and also by the custom ARM legalization for the more complicated AEABI divmod calls. llvm-svn: 305459
2017-06-15 12:53:31 +02:00
case G_SREM:
case G_UREM: {
unsigned OriginalResult = MI.getOperand(0).getReg();
auto Size = MRI.getType(OriginalResult).getSizeInBits();
if (Size != 32)
return false;
auto Libcall =
MI.getOpcode() == G_SREM ? RTLIB::SDIVREM_I32 : RTLIB::UDIVREM_I32;
// Our divmod libcalls return a struct containing the quotient and the
// remainder. We need to create a virtual register for it.
auto &Ctx = MIRBuilder.getMF().getFunction()->getContext();
Type *ArgTy = Type::getInt32Ty(Ctx);
StructType *RetTy = StructType::get(Ctx, {ArgTy, ArgTy}, /* Packed */ true);
auto RetVal = MRI.createGenericVirtualRegister(
getLLTForType(*RetTy, MIRBuilder.getMF().getDataLayout()));
[GlobalISel] Refactor Legalizer helpers for libcalls We used to have a helper that replaced an instruction with a libcall. That turns out to be too aggressive, since sometimes we need to replace the instruction with at least two libcalls. Therefore, change our existing helper to only create the libcall and leave the instruction removal as a separate step. Also rename the helper accordingly. llvm-svn: 307149
2017-07-05 14:57:24 +02:00
auto Status = createLibcall(MIRBuilder, Libcall, {RetVal, RetTy},
{{MI.getOperand(1).getReg(), ArgTy},
{MI.getOperand(2).getReg(), ArgTy}});
[ARM] GlobalISel: Add support for i32 modulo Add support for modulo for targets that have hardware division and for those that don't. When hardware division is not available, we have to choose the correct libcall to use. This is generally straightforward, except for AEABI. The AEABI variant is trickier than the other libcalls because it returns { quotient, remainder }, instead of just one value like the other libcalls that we've seen so far. Therefore, we need to use custom lowering for it. However, we don't want to have too much special code, so we refactor the target-independent code in the legalizer by adding a helper for replacing an instruction with a libcall. This helper is used by the legalizer itself when dealing with simple calls, and also by the custom ARM legalization for the more complicated AEABI divmod calls. llvm-svn: 305459
2017-06-15 12:53:31 +02:00
if (Status != LegalizerHelper::Legalized)
return false;
// The remainder is the second result of divmod. Split the return value into
// a new, unused register for the quotient and the destination of the
// original instruction for the remainder.
MIRBuilder.buildUnmerge(
{MRI.createGenericVirtualRegister(LLT::scalar(32)), OriginalResult},
RetVal);
[ARM] GlobalISel: Legalize G_FCMP for s32 This covers both hard and soft float. Hard float is easy, since it's just Legal. Soft float is more involved, because there are several different ways to handle it based on the predicate: one and ueq need not only one, but two libcalls to get a result. Furthermore, we have large differences between the values returned by the AEABI and GNU functions. AEABI functions return a nice 1 or 0 representing true and respectively false. GNU functions generally return a value that needs to be compared against 0 (e.g. for ogt, the value returned by the libcall is > 0 for true). We could introduce redundant comparisons for AEABI as well, but they don't seem easy to remove afterwards, so we do different processing based on whether or not the result really needs to be compared against something (and just truncate if it doesn't). llvm-svn: 307243
2017-07-06 11:09:33 +02:00
break;
}
case G_FCMP: {
[ARM] GlobalISel: Legalize s64 G_FCMP Same as the s32 version, for both hard and soft float. llvm-svn: 307633
2017-07-11 10:50:01 +02:00
assert(MRI.getType(MI.getOperand(2).getReg()) ==
MRI.getType(MI.getOperand(3).getReg()) &&
"Mismatched operands for G_FCMP");
auto OpSize = MRI.getType(MI.getOperand(2).getReg()).getSizeInBits();
[ARM] GlobalISel: Legalize G_FCMP for s32 This covers both hard and soft float. Hard float is easy, since it's just Legal. Soft float is more involved, because there are several different ways to handle it based on the predicate: one and ueq need not only one, but two libcalls to get a result. Furthermore, we have large differences between the values returned by the AEABI and GNU functions. AEABI functions return a nice 1 or 0 representing true and respectively false. GNU functions generally return a value that needs to be compared against 0 (e.g. for ogt, the value returned by the libcall is > 0 for true). We could introduce redundant comparisons for AEABI as well, but they don't seem easy to remove afterwards, so we do different processing based on whether or not the result really needs to be compared against something (and just truncate if it doesn't). llvm-svn: 307243
2017-07-06 11:09:33 +02:00
auto OriginalResult = MI.getOperand(0).getReg();
auto Predicate =
static_cast<CmpInst::Predicate>(MI.getOperand(1).getPredicate());
[ARM] GlobalISel: Legalize s64 G_FCMP Same as the s32 version, for both hard and soft float. llvm-svn: 307633
2017-07-11 10:50:01 +02:00
auto Libcalls = getFCmpLibcalls(Predicate, OpSize);
[ARM] GlobalISel: Legalize G_FCMP for s32 This covers both hard and soft float. Hard float is easy, since it's just Legal. Soft float is more involved, because there are several different ways to handle it based on the predicate: one and ueq need not only one, but two libcalls to get a result. Furthermore, we have large differences between the values returned by the AEABI and GNU functions. AEABI functions return a nice 1 or 0 representing true and respectively false. GNU functions generally return a value that needs to be compared against 0 (e.g. for ogt, the value returned by the libcall is > 0 for true). We could introduce redundant comparisons for AEABI as well, but they don't seem easy to remove afterwards, so we do different processing based on whether or not the result really needs to be compared against something (and just truncate if it doesn't). llvm-svn: 307243
2017-07-06 11:09:33 +02:00
if (Libcalls.empty()) {
assert((Predicate == CmpInst::FCMP_TRUE ||
Predicate == CmpInst::FCMP_FALSE) &&
"Predicate needs libcalls, but none specified");
MIRBuilder.buildConstant(OriginalResult,
Predicate == CmpInst::FCMP_TRUE ? 1 : 0);
[ARM] GlobalISel: Fix oversight in G_FCMP legalization We used to forget to erase the original instruction when replacing a G_FCMP true/false. Fix this bug and make sure the tests check for it. llvm-svn: 307639
2017-07-11 11:43:51 +02:00
MI.eraseFromParent();
[ARM] GlobalISel: Legalize G_FCMP for s32 This covers both hard and soft float. Hard float is easy, since it's just Legal. Soft float is more involved, because there are several different ways to handle it based on the predicate: one and ueq need not only one, but two libcalls to get a result. Furthermore, we have large differences between the values returned by the AEABI and GNU functions. AEABI functions return a nice 1 or 0 representing true and respectively false. GNU functions generally return a value that needs to be compared against 0 (e.g. for ogt, the value returned by the libcall is > 0 for true). We could introduce redundant comparisons for AEABI as well, but they don't seem easy to remove afterwards, so we do different processing based on whether or not the result really needs to be compared against something (and just truncate if it doesn't). llvm-svn: 307243
2017-07-06 11:09:33 +02:00
return true;
}
auto &Ctx = MIRBuilder.getMF().getFunction()->getContext();
[ARM] GlobalISel: Legalize s64 G_FCMP Same as the s32 version, for both hard and soft float. llvm-svn: 307633
2017-07-11 10:50:01 +02:00
assert((OpSize == 32 || OpSize == 64) && "Unsupported operand size");
auto *ArgTy = OpSize == 32 ? Type::getFloatTy(Ctx) : Type::getDoubleTy(Ctx);
[ARM] GlobalISel: Legalize G_FCMP for s32 This covers both hard and soft float. Hard float is easy, since it's just Legal. Soft float is more involved, because there are several different ways to handle it based on the predicate: one and ueq need not only one, but two libcalls to get a result. Furthermore, we have large differences between the values returned by the AEABI and GNU functions. AEABI functions return a nice 1 or 0 representing true and respectively false. GNU functions generally return a value that needs to be compared against 0 (e.g. for ogt, the value returned by the libcall is > 0 for true). We could introduce redundant comparisons for AEABI as well, but they don't seem easy to remove afterwards, so we do different processing based on whether or not the result really needs to be compared against something (and just truncate if it doesn't). llvm-svn: 307243
2017-07-06 11:09:33 +02:00
auto *RetTy = Type::getInt32Ty(Ctx);
SmallVector<unsigned, 2> Results;
for (auto Libcall : Libcalls) {
auto LibcallResult = MRI.createGenericVirtualRegister(LLT::scalar(32));
auto Status =
createLibcall(MIRBuilder, Libcall.LibcallID, {LibcallResult, RetTy},
{{MI.getOperand(2).getReg(), ArgTy},
{MI.getOperand(3).getReg(), ArgTy}});
if (Status != LegalizerHelper::Legalized)
return false;
auto ProcessedResult =
Libcalls.size() == 1
? OriginalResult
: MRI.createGenericVirtualRegister(MRI.getType(OriginalResult));
// We have a result, but we need to transform it into a proper 1-bit 0 or
// 1, taking into account the different peculiarities of the values
// returned by the comparison functions.
CmpInst::Predicate ResultPred = Libcall.Predicate;
if (ResultPred == CmpInst::BAD_ICMP_PREDICATE) {
// We have a nice 0 or 1, and we just need to truncate it back to 1 bit
// to keep the types consistent.
MIRBuilder.buildTrunc(ProcessedResult, LibcallResult);
} else {
// We need to compare against 0.
assert(CmpInst::isIntPredicate(ResultPred) && "Unsupported predicate");
auto Zero = MRI.createGenericVirtualRegister(LLT::scalar(32));
MIRBuilder.buildConstant(Zero, 0);
MIRBuilder.buildICmp(ResultPred, ProcessedResult, LibcallResult, Zero);
}
Results.push_back(ProcessedResult);
}
[ARM] GlobalISel: Add support for i32 modulo Add support for modulo for targets that have hardware division and for those that don't. When hardware division is not available, we have to choose the correct libcall to use. This is generally straightforward, except for AEABI. The AEABI variant is trickier than the other libcalls because it returns { quotient, remainder }, instead of just one value like the other libcalls that we've seen so far. Therefore, we need to use custom lowering for it. However, we don't want to have too much special code, so we refactor the target-independent code in the legalizer by adding a helper for replacing an instruction with a libcall. This helper is used by the legalizer itself when dealing with simple calls, and also by the custom ARM legalization for the more complicated AEABI divmod calls. llvm-svn: 305459
2017-06-15 12:53:31 +02:00
[ARM] GlobalISel: Legalize G_FCMP for s32 This covers both hard and soft float. Hard float is easy, since it's just Legal. Soft float is more involved, because there are several different ways to handle it based on the predicate: one and ueq need not only one, but two libcalls to get a result. Furthermore, we have large differences between the values returned by the AEABI and GNU functions. AEABI functions return a nice 1 or 0 representing true and respectively false. GNU functions generally return a value that needs to be compared against 0 (e.g. for ogt, the value returned by the libcall is > 0 for true). We could introduce redundant comparisons for AEABI as well, but they don't seem easy to remove afterwards, so we do different processing based on whether or not the result really needs to be compared against something (and just truncate if it doesn't). llvm-svn: 307243
2017-07-06 11:09:33 +02:00
if (Results.size() != 1) {
assert(Results.size() == 2 && "Unexpected number of results");
MIRBuilder.buildOr(OriginalResult, Results[0], Results[1]);
}
[GlobalISel] Refactor Legalizer helpers for libcalls We used to have a helper that replaced an instruction with a libcall. That turns out to be too aggressive, since sometimes we need to replace the instruction with at least two libcalls. Therefore, change our existing helper to only create the libcall and leave the instruction removal as a separate step. Also rename the helper accordingly. llvm-svn: 307149
2017-07-05 14:57:24 +02:00
break;
[ARM] GlobalISel: Add support for i32 modulo Add support for modulo for targets that have hardware division and for those that don't. When hardware division is not available, we have to choose the correct libcall to use. This is generally straightforward, except for AEABI. The AEABI variant is trickier than the other libcalls because it returns { quotient, remainder }, instead of just one value like the other libcalls that we've seen so far. Therefore, we need to use custom lowering for it. However, we don't want to have too much special code, so we refactor the target-independent code in the legalizer by adding a helper for replacing an instruction with a libcall. This helper is used by the legalizer itself when dealing with simple calls, and also by the custom ARM legalization for the more complicated AEABI divmod calls. llvm-svn: 305459
2017-06-15 12:53:31 +02:00
}
[ARM] GlobalISel: Legalize s8 and s16 G_(S|U)DIV We have to widen the operands to 32 bits and then we can either use hardware division if it is available or lower to a libcall otherwise. At the moment it is not enough to set the Legalizer action to WidenScalar, since for libcalls it won't know what to do (it won't be able to find what size to widen to, because it will find Libcall and not Legal for 32 bits). To hack around this limitation, we request Custom lowering, and as part of that we widen first and then we run another legalizeInstrStep on the widened DIV. llvm-svn: 301166
2017-04-24 11:12:19 +02:00
}
[GlobalISel] Refactor Legalizer helpers for libcalls We used to have a helper that replaced an instruction with a libcall. That turns out to be too aggressive, since sometimes we need to replace the instruction with at least two libcalls. Therefore, change our existing helper to only create the libcall and leave the instruction removal as a separate step. Also rename the helper accordingly. llvm-svn: 307149
2017-07-05 14:57:24 +02:00
MI.eraseFromParent();
return true;
[ARM] GlobalISel: Legalize s8 and s16 G_(S|U)DIV We have to widen the operands to 32 bits and then we can either use hardware division if it is available or lower to a libcall otherwise. At the moment it is not enough to set the Legalizer action to WidenScalar, since for libcalls it won't know what to do (it won't be able to find what size to widen to, because it will find Libcall and not Legal for 32 bits). To hack around this limitation, we request Custom lowering, and as part of that we widen first and then we run another legalizeInstrStep on the widened DIV. llvm-svn: 301166
2017-04-24 11:12:19 +02:00
}
Reference in New Issue Copy Permalink