maxjhandsome
/
llvm-project
forked from OSchip/llvm-project
1
0
Fork 0
Code Issues Pull Requests Packages Releases Wiki Activity
llvm-project/llvm/lib/Target/WebAssembly/WebAssemblyInstrMemory.td

687 lines
37 KiB
TableGen
Raw Normal View History

WebAssembly: basic instructions todo, and basic register info. Summary: This code is based on AArch64 for modern backend good practice, and NVPTX for virtual ISA concerns. Reviewers: sunfish Subscribers: aemerson, llvm-commits, jfb Differential Revision: http://reviews.llvm.org/D11070 llvm-svn: 241923
2015-07-11 02:23:10 +08:00
// WebAssemblyInstrMemory.td-WebAssembly Memory codegen support -*- tablegen -*-
//
// The LLVM Compiler Infrastructure
//
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
///
/// \file
/// \brief WebAssembly Memory operand code-gen constructs.
///
//===----------------------------------------------------------------------===//
[WebAssembly] Clean up several FIXME comments. llvm-svn: 254079
2015-11-26 02:13:18 +08:00
// TODO:
WebAssembly: generate load/store Summary: This handles all load/store operations that WebAssembly defines, and handles those necessary for C++ such as i1. I left a FIXME for outstanding features which aren't required for now. Reviewers: sunfish Subscribers: jfb, llvm-commits, dschuff llvm-svn: 246500
2015-09-01 06:24:11 +08:00
// - HasAddr64
// - WebAssemblyTargetLowering having to do with atomics
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
// - Each has optional alignment.
WebAssembly: generate load/store Summary: This handles all load/store operations that WebAssembly defines, and handles those necessary for C++ such as i1. I left a FIXME for outstanding features which aren't required for now. Reviewers: sunfish Subscribers: jfb, llvm-commits, dschuff llvm-svn: 246500
2015-09-01 06:24:11 +08:00
// WebAssembly has i8/i16/i32/i64/f32/f64 memory types, but doesn't have i8/i16
// local types. These memory-only types instead zero- or sign-extend into local
// types when loading, and truncate when storing.
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
// WebAssembly constant offsets are performed as unsigned with infinite
// precision, so we need to check for NoUnsignedWrap so that we don't fold an
// offset for an add that needs wrapping.
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def regPlusImm : PatFrag<(ops node:$addr, node:$off),
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
(add node:$addr, node:$off),
[{ return N->getFlags()->hasNoUnsignedWrap(); }]>;
[WebAssembly] Teach address folding to fold bitwise-or nodes. LLVM converts adds into ors when it can prove that the operands don't share any non-zero bits. Teach address folding to recognize or instructions with constant operands with this property that can be folded into addresses as if they were adds. llvm-svn: 261562
2016-02-23 04:04:02 +08:00
// Treat an 'or' node as an 'add' if the or'ed bits are known to be zero.
def or_is_add : PatFrag<(ops node:$lhs, node:$rhs), (or node:$lhs, node:$rhs),[{
if (ConstantSDNode *CN = dyn_cast<ConstantSDNode>(N->getOperand(1)))
return CurDAG->MaskedValueIsZero(N->getOperand(0), CN->getAPIntValue());
APInt KnownZero0, KnownOne0;
CurDAG->computeKnownBits(N->getOperand(0), KnownZero0, KnownOne0, 0);
APInt KnownZero1, KnownOne1;
CurDAG->computeKnownBits(N->getOperand(1), KnownZero1, KnownOne1, 0);
return (~KnownZero0 & ~KnownZero1) == 0;
}]>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
// GlobalAddresses are conceptually unsigned values, so we can also fold them
[WebAssembly] Disable folding of GA+reg into load/store constant offsets Summary: If the register has a negative value then unsigned overflow will occur; this case is sometimes even created intentionally by LSR. For now disable GA+reg folding. Fixes PR29127 Differential Revision: https://reviews.llvm.org/D24053 llvm-svn: 280285
2016-09-01 04:27:20 +08:00
// into immediate values as long as the add is 'nuw'.
// TODO: We'd like to also match GA offsets but there are cases where the
// register can have a negative value. Find out what more we can do.
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def regPlusGA : PatFrag<(ops node:$addr, node:$off),
(add node:$addr, node:$off),
[{
[WebAssembly] Disable folding of GA+reg into load/store constant offsets Summary: If the register has a negative value then unsigned overflow will occur; this case is sometimes even created intentionally by LSR. For now disable GA+reg folding. Fixes PR29127 Differential Revision: https://reviews.llvm.org/D24053 llvm-svn: 280285
2016-09-01 04:27:20 +08:00
return N->getFlags()->hasNoUnsignedWrap();
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
}]>;
// We don't need a regPlusES because external symbols never have constant
// offsets folded into them, so we can just use add.
[WebAssembly] Use a physical register to describe ARGUMENT liveness. Instead of trying to move ARGUMENT instructions back up to the top after they've been scheduled or sunk down, use a fake physical register to create a liveness constraint that prevents ARGUMENT instructions from moving down in the first place. This is still not entirely ideal, however it is more robust than letting them move and moving them back. llvm-svn: 254084
2015-11-26 03:36:19 +08:00
let Defs = [ARGUMENTS] in {
WebAssembly: generate load/store Summary: This handles all load/store operations that WebAssembly defines, and handles those necessary for C++ such as i1. I left a FIXME for outstanding features which aren't required for now. Reviewers: sunfish Subscribers: jfb, llvm-commits, dschuff llvm-svn: 246500
2015-09-01 06:24:11 +08:00
// Basic load.
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
// FIXME: When we can break syntax compatibility, reorder the fields in the
// asmstrings to match the binary encoding.
def LOAD_I32 : I<(outs I32:$dst),
[WebAssembly] Annotate call and load/store immediates. These will be used to guide the binary encoding of these immediates. llvm-svn: 290412
2016-12-23 11:23:52 +08:00
(ins P2Align:$p2align, offset32_op:$off, I32:$addr),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
[], "i32.load\t$dst, ${off}(${addr})${p2align}", 0x28>;
def LOAD_I64 : I<(outs I64:$dst),
[WebAssembly] Annotate call and load/store immediates. These will be used to guide the binary encoding of these immediates. llvm-svn: 290412
2016-12-23 11:23:52 +08:00
(ins P2Align:$p2align, offset32_op:$off, I32:$addr),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
[], "i64.load\t$dst, ${off}(${addr})${p2align}", 0x29>;
def LOAD_F32 : I<(outs F32:$dst),
[WebAssembly] Annotate call and load/store immediates. These will be used to guide the binary encoding of these immediates. llvm-svn: 290412
2016-12-23 11:23:52 +08:00
(ins P2Align:$p2align, offset32_op:$off, I32:$addr),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
[], "f32.load\t$dst, ${off}(${addr})${p2align}", 0x2a>;
def LOAD_F64 : I<(outs F64:$dst),
[WebAssembly] Annotate call and load/store immediates. These will be used to guide the binary encoding of these immediates. llvm-svn: 290412
2016-12-23 11:23:52 +08:00
(ins P2Align:$p2align, offset32_op:$off, I32:$addr),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
[], "f64.load\t$dst, ${off}(${addr})${p2align}", 0x2b>;
WebAssembly: generate load/store Summary: This handles all load/store operations that WebAssembly defines, and handles those necessary for C++ such as i1. I left a FIXME for outstanding features which aren't required for now. Reviewers: sunfish Subscribers: jfb, llvm-commits, dschuff llvm-svn: 246500
2015-09-01 06:24:11 +08:00
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
} // Defs = [ARGUMENTS]
// Select loads with no constant offset.
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
def : Pat<(i32 (load I32:$addr)), (LOAD_I32 0, 0, $addr)>;
def : Pat<(i64 (load I32:$addr)), (LOAD_I64 0, 0, $addr)>;
def : Pat<(f32 (load I32:$addr)), (LOAD_F32 0, 0, $addr)>;
def : Pat<(f64 (load I32:$addr)), (LOAD_F64 0, 0, $addr)>;
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
// Select loads with a constant offset.
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(i32 (load (regPlusImm I32:$addr, imm:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD_I32 0, imm:$off, $addr)>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(i64 (load (regPlusImm I32:$addr, imm:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD_I64 0, imm:$off, $addr)>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(f32 (load (regPlusImm I32:$addr, imm:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD_F32 0, imm:$off, $addr)>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(f64 (load (regPlusImm I32:$addr, imm:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD_F64 0, imm:$off, $addr)>;
[WebAssembly] Teach address folding to fold bitwise-or nodes. LLVM converts adds into ors when it can prove that the operands don't share any non-zero bits. Teach address folding to recognize or instructions with constant operands with this property that can be folded into addresses as if they were adds. llvm-svn: 261562
2016-02-23 04:04:02 +08:00
def : Pat<(i32 (load (or_is_add I32:$addr, imm:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD_I32 0, imm:$off, $addr)>;
[WebAssembly] Teach address folding to fold bitwise-or nodes. LLVM converts adds into ors when it can prove that the operands don't share any non-zero bits. Teach address folding to recognize or instructions with constant operands with this property that can be folded into addresses as if they were adds. llvm-svn: 261562
2016-02-23 04:04:02 +08:00
def : Pat<(i64 (load (or_is_add I32:$addr, imm:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD_I64 0, imm:$off, $addr)>;
[WebAssembly] Teach address folding to fold bitwise-or nodes. LLVM converts adds into ors when it can prove that the operands don't share any non-zero bits. Teach address folding to recognize or instructions with constant operands with this property that can be folded into addresses as if they were adds. llvm-svn: 261562
2016-02-23 04:04:02 +08:00
def : Pat<(f32 (load (or_is_add I32:$addr, imm:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD_F32 0, imm:$off, $addr)>;
[WebAssembly] Teach address folding to fold bitwise-or nodes. LLVM converts adds into ors when it can prove that the operands don't share any non-zero bits. Teach address folding to recognize or instructions with constant operands with this property that can be folded into addresses as if they were adds. llvm-svn: 261562
2016-02-23 04:04:02 +08:00
def : Pat<(f64 (load (or_is_add I32:$addr, imm:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD_F64 0, imm:$off, $addr)>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(i32 (load (regPlusGA I32:$addr,
(WebAssemblywrapper tglobaladdr:$off)))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD_I32 0, tglobaladdr:$off, $addr)>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(i64 (load (regPlusGA I32:$addr,
(WebAssemblywrapper tglobaladdr:$off)))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD_I64 0, tglobaladdr:$off, $addr)>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(f32 (load (regPlusGA I32:$addr,
(WebAssemblywrapper tglobaladdr:$off)))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD_F32 0, tglobaladdr:$off, $addr)>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(f64 (load (regPlusGA I32:$addr,
(WebAssemblywrapper tglobaladdr:$off)))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD_F64 0, tglobaladdr:$off, $addr)>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(i32 (load (add I32:$addr, (WebAssemblywrapper texternalsym:$off)))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD_I32 0, texternalsym:$off, $addr)>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(i64 (load (add I32:$addr, (WebAssemblywrapper texternalsym:$off)))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD_I64 0, texternalsym:$off, $addr)>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(f32 (load (add I32:$addr, (WebAssemblywrapper texternalsym:$off)))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD_F32 0, texternalsym:$off, $addr)>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(f64 (load (add I32:$addr, (WebAssemblywrapper texternalsym:$off)))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD_F64 0, texternalsym:$off, $addr)>;
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
// Select loads with just a constant offset.
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
def : Pat<(i32 (load imm:$off)), (LOAD_I32 0, imm:$off, (CONST_I32 0))>;
def : Pat<(i64 (load imm:$off)), (LOAD_I64 0, imm:$off, (CONST_I32 0))>;
def : Pat<(f32 (load imm:$off)), (LOAD_F32 0, imm:$off, (CONST_I32 0))>;
def : Pat<(f64 (load imm:$off)), (LOAD_F64 0, imm:$off, (CONST_I32 0))>;
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
def : Pat<(i32 (load (WebAssemblywrapper tglobaladdr:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD_I32 0, tglobaladdr:$off, (CONST_I32 0))>;
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
def : Pat<(i64 (load (WebAssemblywrapper tglobaladdr:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD_I64 0, tglobaladdr:$off, (CONST_I32 0))>;
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
def : Pat<(f32 (load (WebAssemblywrapper tglobaladdr:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD_F32 0, tglobaladdr:$off, (CONST_I32 0))>;
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
def : Pat<(f64 (load (WebAssemblywrapper tglobaladdr:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD_F64 0, tglobaladdr:$off, (CONST_I32 0))>;
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
def : Pat<(i32 (load (WebAssemblywrapper texternalsym:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD_I32 0, texternalsym:$off, (CONST_I32 0))>;
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
def : Pat<(i64 (load (WebAssemblywrapper texternalsym:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD_I64 0, texternalsym:$off, (CONST_I32 0))>;
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
def : Pat<(f32 (load (WebAssemblywrapper texternalsym:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD_F32 0, texternalsym:$off, (CONST_I32 0))>;
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
def : Pat<(f64 (load (WebAssemblywrapper texternalsym:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD_F64 0, texternalsym:$off, (CONST_I32 0))>;
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
let Defs = [ARGUMENTS] in {
WebAssembly: generate load/store Summary: This handles all load/store operations that WebAssembly defines, and handles those necessary for C++ such as i1. I left a FIXME for outstanding features which aren't required for now. Reviewers: sunfish Subscribers: jfb, llvm-commits, dschuff llvm-svn: 246500
2015-09-01 06:24:11 +08:00
// Extending load.
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
def LOAD8_S_I32 : I<(outs I32:$dst),
[WebAssembly] Annotate call and load/store immediates. These will be used to guide the binary encoding of these immediates. llvm-svn: 290412
2016-12-23 11:23:52 +08:00
(ins P2Align:$p2align, offset32_op:$off, I32:$addr),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
[], "i32.load8_s\t$dst, ${off}(${addr})${p2align}", 0x2c>;
def LOAD8_U_I32 : I<(outs I32:$dst),
[WebAssembly] Annotate call and load/store immediates. These will be used to guide the binary encoding of these immediates. llvm-svn: 290412
2016-12-23 11:23:52 +08:00
(ins P2Align:$p2align, offset32_op:$off, I32:$addr),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
[], "i32.load8_u\t$dst, ${off}(${addr})${p2align}", 0x2d>;
def LOAD16_S_I32 : I<(outs I32:$dst),
[WebAssembly] Annotate call and load/store immediates. These will be used to guide the binary encoding of these immediates. llvm-svn: 290412
2016-12-23 11:23:52 +08:00
(ins P2Align:$p2align, offset32_op:$off, I32:$addr),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
[], "i32.load16_s\t$dst, ${off}(${addr})${p2align}", 0x2e>;
def LOAD16_U_I32 : I<(outs I32:$dst),
[WebAssembly] Annotate call and load/store immediates. These will be used to guide the binary encoding of these immediates. llvm-svn: 290412
2016-12-23 11:23:52 +08:00
(ins P2Align:$p2align, offset32_op:$off, I32:$addr),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
[], "i32.load16_u\t$dst, ${off}(${addr})${p2align}", 0x2f>;
def LOAD8_S_I64 : I<(outs I64:$dst),
[WebAssembly] Annotate call and load/store immediates. These will be used to guide the binary encoding of these immediates. llvm-svn: 290412
2016-12-23 11:23:52 +08:00
(ins P2Align:$p2align, offset32_op:$off, I32:$addr),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
[], "i64.load8_s\t$dst, ${off}(${addr})${p2align}", 0x30>;
def LOAD8_U_I64 : I<(outs I64:$dst),
[WebAssembly] Annotate call and load/store immediates. These will be used to guide the binary encoding of these immediates. llvm-svn: 290412
2016-12-23 11:23:52 +08:00
(ins P2Align:$p2align, offset32_op:$off, I32:$addr),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
[], "i64.load8_u\t$dst, ${off}(${addr})${p2align}", 0x31>;
def LOAD16_S_I64 : I<(outs I64:$dst),
[WebAssembly] Annotate call and load/store immediates. These will be used to guide the binary encoding of these immediates. llvm-svn: 290412
2016-12-23 11:23:52 +08:00
(ins P2Align:$p2align, offset32_op:$off, I32:$addr),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
[], "i64.load16_s\t$dst, ${off}(${addr})${p2align}", 0x32>;
def LOAD16_U_I64 : I<(outs I64:$dst),
[WebAssembly] Annotate call and load/store immediates. These will be used to guide the binary encoding of these immediates. llvm-svn: 290412
2016-12-23 11:23:52 +08:00
(ins P2Align:$p2align, offset32_op:$off, I32:$addr),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
[], "i64.load16_u\t$dst, ${off}(${addr})${p2align}", 0x33>;
def LOAD32_S_I64 : I<(outs I64:$dst),
[WebAssembly] Annotate call and load/store immediates. These will be used to guide the binary encoding of these immediates. llvm-svn: 290412
2016-12-23 11:23:52 +08:00
(ins P2Align:$p2align, offset32_op:$off, I32:$addr),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
[], "i64.load32_s\t$dst, ${off}(${addr})${p2align}", 0x34>;
def LOAD32_U_I64 : I<(outs I64:$dst),
[WebAssembly] Annotate call and load/store immediates. These will be used to guide the binary encoding of these immediates. llvm-svn: 290412
2016-12-23 11:23:52 +08:00
(ins P2Align:$p2align, offset32_op:$off, I32:$addr),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
[], "i64.load32_u\t$dst, ${off}(${addr})${p2align}", 0x35>;
WebAssembly: generate load/store Summary: This handles all load/store operations that WebAssembly defines, and handles those necessary for C++ such as i1. I left a FIXME for outstanding features which aren't required for now. Reviewers: sunfish Subscribers: jfb, llvm-commits, dschuff llvm-svn: 246500
2015-09-01 06:24:11 +08:00
[WebAssembly] Use a physical register to describe ARGUMENT liveness. Instead of trying to move ARGUMENT instructions back up to the top after they've been scheduled or sunk down, use a fake physical register to create a liveness constraint that prevents ARGUMENT instructions from moving down in the first place. This is still not entirely ideal, however it is more robust than letting them move and moving them back. llvm-svn: 254084
2015-11-26 03:36:19 +08:00
} // Defs = [ARGUMENTS]
[WebAssembly] Support constant offsets on loads and stores This is just prototype for load/store for i32 types. I'll add them to the rest of the types if we like this direction. Differential Revision: http://reviews.llvm.org/D15197 llvm-svn: 254807
2015-12-05 08:26:39 +08:00
// Select extending loads with no constant offset.
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
def : Pat<(i32 (sextloadi8 I32:$addr)), (LOAD8_S_I32 0, 0, $addr)>;
def : Pat<(i32 (zextloadi8 I32:$addr)), (LOAD8_U_I32 0, 0, $addr)>;
def : Pat<(i32 (sextloadi16 I32:$addr)), (LOAD16_S_I32 0, 0, $addr)>;
def : Pat<(i32 (zextloadi16 I32:$addr)), (LOAD16_U_I32 0, 0, $addr)>;
def : Pat<(i64 (sextloadi8 I32:$addr)), (LOAD8_S_I64 0, 0, $addr)>;
def : Pat<(i64 (zextloadi8 I32:$addr)), (LOAD8_U_I64 0, 0, $addr)>;
def : Pat<(i64 (sextloadi16 I32:$addr)), (LOAD16_S_I64 0, 0, $addr)>;
def : Pat<(i64 (zextloadi16 I32:$addr)), (LOAD16_U_I64 0, 0, $addr)>;
def : Pat<(i64 (sextloadi32 I32:$addr)), (LOAD32_S_I64 0, 0, $addr)>;
def : Pat<(i64 (zextloadi32 I32:$addr)), (LOAD32_U_I64 0, 0, $addr)>;
[WebAssembly] Support constant offsets on loads and stores This is just prototype for load/store for i32 types. I'll add them to the rest of the types if we like this direction. Differential Revision: http://reviews.llvm.org/D15197 llvm-svn: 254807
2015-12-05 08:26:39 +08:00
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
// Select extending loads with a constant offset.
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(i32 (sextloadi8 (regPlusImm I32:$addr, imm:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD8_S_I32 0, imm:$off, $addr)>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(i32 (zextloadi8 (regPlusImm I32:$addr, imm:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD8_U_I32 0, imm:$off, $addr)>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(i32 (sextloadi16 (regPlusImm I32:$addr, imm:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD16_S_I32 0, imm:$off, $addr)>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(i32 (zextloadi16 (regPlusImm I32:$addr, imm:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD16_U_I32 0, imm:$off, $addr)>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(i64 (sextloadi8 (regPlusImm I32:$addr, imm:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD8_S_I64 0, imm:$off, $addr)>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(i64 (zextloadi8 (regPlusImm I32:$addr, imm:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD8_U_I64 0, imm:$off, $addr)>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(i64 (sextloadi16 (regPlusImm I32:$addr, imm:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD16_S_I64 0, imm:$off, $addr)>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(i64 (zextloadi16 (regPlusImm I32:$addr, imm:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD16_U_I64 0, imm:$off, $addr)>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(i64 (sextloadi32 (regPlusImm I32:$addr, imm:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD32_S_I64 0, imm:$off, $addr)>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(i64 (zextloadi32 (regPlusImm I32:$addr, imm:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD32_U_I64 0, imm:$off, $addr)>;
[WebAssembly] Teach address folding to fold bitwise-or nodes. LLVM converts adds into ors when it can prove that the operands don't share any non-zero bits. Teach address folding to recognize or instructions with constant operands with this property that can be folded into addresses as if they were adds. llvm-svn: 261562
2016-02-23 04:04:02 +08:00
def : Pat<(i32 (sextloadi8 (or_is_add I32:$addr, imm:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD8_S_I32 0, imm:$off, $addr)>;
[WebAssembly] Teach address folding to fold bitwise-or nodes. LLVM converts adds into ors when it can prove that the operands don't share any non-zero bits. Teach address folding to recognize or instructions with constant operands with this property that can be folded into addresses as if they were adds. llvm-svn: 261562
2016-02-23 04:04:02 +08:00
def : Pat<(i32 (zextloadi8 (or_is_add I32:$addr, imm:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD8_U_I32 0, imm:$off, $addr)>;
[WebAssembly] Teach address folding to fold bitwise-or nodes. LLVM converts adds into ors when it can prove that the operands don't share any non-zero bits. Teach address folding to recognize or instructions with constant operands with this property that can be folded into addresses as if they were adds. llvm-svn: 261562
2016-02-23 04:04:02 +08:00
def : Pat<(i32 (sextloadi16 (or_is_add I32:$addr, imm:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD16_S_I32 0, imm:$off, $addr)>;
[WebAssembly] Teach address folding to fold bitwise-or nodes. LLVM converts adds into ors when it can prove that the operands don't share any non-zero bits. Teach address folding to recognize or instructions with constant operands with this property that can be folded into addresses as if they were adds. llvm-svn: 261562
2016-02-23 04:04:02 +08:00
def : Pat<(i32 (zextloadi16 (or_is_add I32:$addr, imm:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD16_U_I32 0, imm:$off, $addr)>;
[WebAssembly] Teach address folding to fold bitwise-or nodes. LLVM converts adds into ors when it can prove that the operands don't share any non-zero bits. Teach address folding to recognize or instructions with constant operands with this property that can be folded into addresses as if they were adds. llvm-svn: 261562
2016-02-23 04:04:02 +08:00
def : Pat<(i64 (sextloadi8 (or_is_add I32:$addr, imm:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD8_S_I64 0, imm:$off, $addr)>;
[WebAssembly] Teach address folding to fold bitwise-or nodes. LLVM converts adds into ors when it can prove that the operands don't share any non-zero bits. Teach address folding to recognize or instructions with constant operands with this property that can be folded into addresses as if they were adds. llvm-svn: 261562
2016-02-23 04:04:02 +08:00
def : Pat<(i64 (zextloadi8 (or_is_add I32:$addr, imm:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD8_U_I64 0, imm:$off, $addr)>;
[WebAssembly] Teach address folding to fold bitwise-or nodes. LLVM converts adds into ors when it can prove that the operands don't share any non-zero bits. Teach address folding to recognize or instructions with constant operands with this property that can be folded into addresses as if they were adds. llvm-svn: 261562
2016-02-23 04:04:02 +08:00
def : Pat<(i64 (sextloadi16 (or_is_add I32:$addr, imm:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD16_S_I64 0, imm:$off, $addr)>;
[WebAssembly] Teach address folding to fold bitwise-or nodes. LLVM converts adds into ors when it can prove that the operands don't share any non-zero bits. Teach address folding to recognize or instructions with constant operands with this property that can be folded into addresses as if they were adds. llvm-svn: 261562
2016-02-23 04:04:02 +08:00
def : Pat<(i64 (zextloadi16 (or_is_add I32:$addr, imm:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD16_U_I64 0, imm:$off, $addr)>;
[WebAssembly] Teach address folding to fold bitwise-or nodes. LLVM converts adds into ors when it can prove that the operands don't share any non-zero bits. Teach address folding to recognize or instructions with constant operands with this property that can be folded into addresses as if they were adds. llvm-svn: 261562
2016-02-23 04:04:02 +08:00
def : Pat<(i64 (sextloadi32 (or_is_add I32:$addr, imm:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD32_S_I64 0, imm:$off, $addr)>;
[WebAssembly] Teach address folding to fold bitwise-or nodes. LLVM converts adds into ors when it can prove that the operands don't share any non-zero bits. Teach address folding to recognize or instructions with constant operands with this property that can be folded into addresses as if they were adds. llvm-svn: 261562
2016-02-23 04:04:02 +08:00
def : Pat<(i64 (zextloadi32 (or_is_add I32:$addr, imm:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD32_U_I64 0, imm:$off, $addr)>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(i32 (sextloadi8 (regPlusGA I32:$addr,
(WebAssemblywrapper tglobaladdr:$off)))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD8_S_I32 0, tglobaladdr:$off, $addr)>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(i32 (zextloadi8 (regPlusGA I32:$addr,
(WebAssemblywrapper tglobaladdr:$off)))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD8_U_I32 0, tglobaladdr:$off, $addr)>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(i32 (sextloadi16 (regPlusGA I32:$addr,
(WebAssemblywrapper tglobaladdr:$off)))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD16_S_I32 0, tglobaladdr:$off, $addr)>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(i32 (zextloadi16 (regPlusGA I32:$addr,
(WebAssemblywrapper tglobaladdr:$off)))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD16_U_I32 0, tglobaladdr:$off, $addr)>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(i64 (sextloadi8 (regPlusGA I32:$addr,
(WebAssemblywrapper tglobaladdr:$off)))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD8_S_I64 0, tglobaladdr:$off, $addr)>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(i64 (zextloadi8 (regPlusGA I32:$addr,
(WebAssemblywrapper tglobaladdr:$off)))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD8_U_I64 0, tglobaladdr:$off, $addr)>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(i64 (sextloadi16 (regPlusGA I32:$addr,
(WebAssemblywrapper tglobaladdr:$off)))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD16_S_I64 0, tglobaladdr:$off, $addr)>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(i64 (zextloadi16 (regPlusGA I32:$addr,
(WebAssemblywrapper tglobaladdr:$off)))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD16_U_I64 0, tglobaladdr:$off, $addr)>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(i64 (sextloadi32 (regPlusGA I32:$addr,
(WebAssemblywrapper tglobaladdr:$off)))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD32_S_I64 0, tglobaladdr:$off, $addr)>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(i64 (zextloadi32 (regPlusGA I32:$addr,
(WebAssemblywrapper tglobaladdr:$off)))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD32_U_I64 0, tglobaladdr:$off, $addr)>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(i32 (sextloadi8 (add I32:$addr,
(WebAssemblywrapper texternalsym:$off)))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD8_S_I32 0, texternalsym:$off, $addr)>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(i32 (zextloadi8 (add I32:$addr,
(WebAssemblywrapper texternalsym:$off)))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD8_U_I32 0, texternalsym:$off, $addr)>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(i32 (sextloadi16 (add I32:$addr,
(WebAssemblywrapper texternalsym:$off)))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD16_S_I32 0, texternalsym:$off, $addr)>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(i32 (zextloadi16 (add I32:$addr,
(WebAssemblywrapper texternalsym:$off)))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD16_U_I32 0, texternalsym:$off, $addr)>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(i64 (sextloadi8 (add I32:$addr,
(WebAssemblywrapper texternalsym:$off)))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD8_S_I64 0, texternalsym:$off, $addr)>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(i64 (zextloadi8 (add I32:$addr,
(WebAssemblywrapper texternalsym:$off)))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD8_U_I64 0, texternalsym:$off, $addr)>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(i64 (sextloadi16 (add I32:$addr,
(WebAssemblywrapper texternalsym:$off)))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD16_S_I64 0, texternalsym:$off, $addr)>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(i64 (zextloadi16 (add I32:$addr,
(WebAssemblywrapper texternalsym:$off)))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD16_U_I64 0, texternalsym:$off, $addr)>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(i64 (sextloadi32 (add I32:$addr,
(WebAssemblywrapper texternalsym:$off)))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD32_S_I64 0, texternalsym:$off, $addr)>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(i64 (zextloadi32 (add I32:$addr,
(WebAssemblywrapper texternalsym:$off)))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD32_U_I64 0, texternalsym:$off, $addr)>;
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
// Select extending loads with just a constant offset.
[WebAssembly] Implement unaligned loads and stores. Differential Revision: http://reviews.llvm.org/D16534 llvm-svn: 258779
2016-01-26 11:39:31 +08:00
def : Pat<(i32 (sextloadi8 imm:$off)),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD8_S_I32 0, imm:$off, (CONST_I32 0))>;
[WebAssembly] Implement unaligned loads and stores. Differential Revision: http://reviews.llvm.org/D16534 llvm-svn: 258779
2016-01-26 11:39:31 +08:00
def : Pat<(i32 (zextloadi8 imm:$off)),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD8_U_I32 0, imm:$off, (CONST_I32 0))>;
[WebAssembly] Implement unaligned loads and stores. Differential Revision: http://reviews.llvm.org/D16534 llvm-svn: 258779
2016-01-26 11:39:31 +08:00
def : Pat<(i32 (sextloadi16 imm:$off)),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD16_S_I32 0, imm:$off, (CONST_I32 0))>;
[WebAssembly] Implement unaligned loads and stores. Differential Revision: http://reviews.llvm.org/D16534 llvm-svn: 258779
2016-01-26 11:39:31 +08:00
def : Pat<(i32 (zextloadi16 imm:$off)),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD16_U_I32 0, imm:$off, (CONST_I32 0))>;
[WebAssembly] Implement unaligned loads and stores. Differential Revision: http://reviews.llvm.org/D16534 llvm-svn: 258779
2016-01-26 11:39:31 +08:00
def : Pat<(i64 (sextloadi8 imm:$off)),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD8_S_I64 0, imm:$off, (CONST_I32 0))>;
[WebAssembly] Implement unaligned loads and stores. Differential Revision: http://reviews.llvm.org/D16534 llvm-svn: 258779
2016-01-26 11:39:31 +08:00
def : Pat<(i64 (zextloadi8 imm:$off)),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD8_U_I64 0, imm:$off, (CONST_I32 0))>;
[WebAssembly] Implement unaligned loads and stores. Differential Revision: http://reviews.llvm.org/D16534 llvm-svn: 258779
2016-01-26 11:39:31 +08:00
def : Pat<(i64 (sextloadi16 imm:$off)),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD16_S_I64 0, imm:$off, (CONST_I32 0))>;
[WebAssembly] Implement unaligned loads and stores. Differential Revision: http://reviews.llvm.org/D16534 llvm-svn: 258779
2016-01-26 11:39:31 +08:00
def : Pat<(i64 (zextloadi16 imm:$off)),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD16_U_I64 0, imm:$off, (CONST_I32 0))>;
[WebAssembly] Implement unaligned loads and stores. Differential Revision: http://reviews.llvm.org/D16534 llvm-svn: 258779
2016-01-26 11:39:31 +08:00
def : Pat<(i64 (sextloadi32 imm:$off)),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD32_S_I64 0, imm:$off, (CONST_I32 0))>;
[WebAssembly] Implement unaligned loads and stores. Differential Revision: http://reviews.llvm.org/D16534 llvm-svn: 258779
2016-01-26 11:39:31 +08:00
def : Pat<(i64 (zextloadi32 imm:$off)),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD32_U_I64 0, imm:$off, (CONST_I32 0))>;
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
def : Pat<(i32 (sextloadi8 (WebAssemblywrapper tglobaladdr:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD8_S_I32 0, tglobaladdr:$off, (CONST_I32 0))>;
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
def : Pat<(i32 (zextloadi8 (WebAssemblywrapper tglobaladdr:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD8_U_I32 0, tglobaladdr:$off, (CONST_I32 0))>;
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
def : Pat<(i32 (sextloadi16 (WebAssemblywrapper tglobaladdr:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD16_S_I32 0, tglobaladdr:$off, (CONST_I32 0))>;
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
def : Pat<(i32 (zextloadi16 (WebAssemblywrapper tglobaladdr:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD16_U_I32 0, tglobaladdr:$off, (CONST_I32 0))>;
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
def : Pat<(i64 (sextloadi8 (WebAssemblywrapper tglobaladdr:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD8_S_I64 0, tglobaladdr:$off, (CONST_I32 0))>;
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
def : Pat<(i64 (zextloadi8 (WebAssemblywrapper tglobaladdr:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD8_U_I64 0, tglobaladdr:$off, (CONST_I32 0))>;
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
def : Pat<(i64 (sextloadi16 (WebAssemblywrapper tglobaladdr:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD16_S_I64 0, tglobaladdr:$off, (CONST_I32 0))>;
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
def : Pat<(i64 (zextloadi16 (WebAssemblywrapper tglobaladdr:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD16_U_I64 0, tglobaladdr:$off, (CONST_I32 0))>;
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
def : Pat<(i64 (sextloadi32 (WebAssemblywrapper tglobaladdr:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD32_S_I64 0, tglobaladdr:$off, (CONST_I32 0))>;
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
def : Pat<(i64 (zextloadi32 (WebAssemblywrapper tglobaladdr:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD32_U_I64 0, tglobaladdr:$off, (CONST_I32 0))>;
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
def : Pat<(i32 (sextloadi8 (WebAssemblywrapper texternalsym:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD8_S_I32 0, texternalsym:$off, (CONST_I32 0))>;
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
def : Pat<(i32 (zextloadi8 (WebAssemblywrapper texternalsym:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD8_U_I32 0, texternalsym:$off, (CONST_I32 0))>;
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
def : Pat<(i32 (sextloadi16 (WebAssemblywrapper texternalsym:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD16_S_I32 0, texternalsym:$off, (CONST_I32 0))>;
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
def : Pat<(i32 (zextloadi16 (WebAssemblywrapper texternalsym:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD16_U_I32 0, texternalsym:$off, (CONST_I32 0))>;
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
def : Pat<(i64 (sextloadi8 (WebAssemblywrapper texternalsym:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD8_S_I64 0, texternalsym:$off, (CONST_I32 0))>;
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
def : Pat<(i64 (zextloadi8 (WebAssemblywrapper texternalsym:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD8_U_I64 0, texternalsym:$off, (CONST_I32 0))>;
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
def : Pat<(i64 (sextloadi16 (WebAssemblywrapper texternalsym:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD16_S_I64 0, texternalsym:$off, (CONST_I32 0))>;
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
def : Pat<(i64 (zextloadi16 (WebAssemblywrapper texternalsym:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD16_U_I64 0, texternalsym:$off, (CONST_I32 0))>;
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
def : Pat<(i64 (sextloadi32 (WebAssemblywrapper texternalsym:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD32_S_I64 0, texternalsym:$off, (CONST_I32 0))>;
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
def : Pat<(i64 (zextloadi32 (WebAssemblywrapper texternalsym:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD32_U_I64 0, texternalsym:$off, (CONST_I32 0))>;
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
// Resolve "don't care" extending loads to zero-extending loads. This is
// somewhat arbitrary, but zero-extending is conceptually simpler.
// Select "don't care" extending loads with no constant offset.
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
def : Pat<(i32 (extloadi8 I32:$addr)), (LOAD8_U_I32 0, 0, $addr)>;
def : Pat<(i32 (extloadi16 I32:$addr)), (LOAD16_U_I32 0, 0, $addr)>;
def : Pat<(i64 (extloadi8 I32:$addr)), (LOAD8_U_I64 0, 0, $addr)>;
def : Pat<(i64 (extloadi16 I32:$addr)), (LOAD16_U_I64 0, 0, $addr)>;
def : Pat<(i64 (extloadi32 I32:$addr)), (LOAD32_U_I64 0, 0, $addr)>;
WebAssembly: generate load/store Summary: This handles all load/store operations that WebAssembly defines, and handles those necessary for C++ such as i1. I left a FIXME for outstanding features which aren't required for now. Reviewers: sunfish Subscribers: jfb, llvm-commits, dschuff llvm-svn: 246500
2015-09-01 06:24:11 +08:00
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
// Select "don't care" extending loads with a constant offset.
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(i32 (extloadi8 (regPlusImm I32:$addr, imm:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD8_U_I32 0, imm:$off, $addr)>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(i32 (extloadi16 (regPlusImm I32:$addr, imm:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD16_U_I32 0, imm:$off, $addr)>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(i64 (extloadi8 (regPlusImm I32:$addr, imm:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD8_U_I64 0, imm:$off, $addr)>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(i64 (extloadi16 (regPlusImm I32:$addr, imm:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD16_U_I64 0, imm:$off, $addr)>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(i64 (extloadi32 (regPlusImm I32:$addr, imm:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD32_U_I64 0, imm:$off, $addr)>;
[WebAssembly] Teach address folding to fold bitwise-or nodes. LLVM converts adds into ors when it can prove that the operands don't share any non-zero bits. Teach address folding to recognize or instructions with constant operands with this property that can be folded into addresses as if they were adds. llvm-svn: 261562
2016-02-23 04:04:02 +08:00
def : Pat<(i32 (extloadi8 (or_is_add I32:$addr, imm:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD8_U_I32 0, imm:$off, $addr)>;
[WebAssembly] Teach address folding to fold bitwise-or nodes. LLVM converts adds into ors when it can prove that the operands don't share any non-zero bits. Teach address folding to recognize or instructions with constant operands with this property that can be folded into addresses as if they were adds. llvm-svn: 261562
2016-02-23 04:04:02 +08:00
def : Pat<(i32 (extloadi16 (or_is_add I32:$addr, imm:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD16_U_I32 0, imm:$off, $addr)>;
[WebAssembly] Teach address folding to fold bitwise-or nodes. LLVM converts adds into ors when it can prove that the operands don't share any non-zero bits. Teach address folding to recognize or instructions with constant operands with this property that can be folded into addresses as if they were adds. llvm-svn: 261562
2016-02-23 04:04:02 +08:00
def : Pat<(i64 (extloadi8 (or_is_add I32:$addr, imm:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD8_U_I64 0, imm:$off, $addr)>;
[WebAssembly] Teach address folding to fold bitwise-or nodes. LLVM converts adds into ors when it can prove that the operands don't share any non-zero bits. Teach address folding to recognize or instructions with constant operands with this property that can be folded into addresses as if they were adds. llvm-svn: 261562
2016-02-23 04:04:02 +08:00
def : Pat<(i64 (extloadi16 (or_is_add I32:$addr, imm:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD16_U_I64 0, imm:$off, $addr)>;
[WebAssembly] Teach address folding to fold bitwise-or nodes. LLVM converts adds into ors when it can prove that the operands don't share any non-zero bits. Teach address folding to recognize or instructions with constant operands with this property that can be folded into addresses as if they were adds. llvm-svn: 261562
2016-02-23 04:04:02 +08:00
def : Pat<(i64 (extloadi32 (or_is_add I32:$addr, imm:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD32_U_I64 0, imm:$off, $addr)>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(i32 (extloadi8 (regPlusGA I32:$addr,
(WebAssemblywrapper tglobaladdr:$off)))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD8_U_I32 0, tglobaladdr:$off, $addr)>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(i32 (extloadi16 (regPlusGA I32:$addr,
(WebAssemblywrapper tglobaladdr:$off)))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD16_U_I32 0, tglobaladdr:$off, $addr)>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(i64 (extloadi8 (regPlusGA I32:$addr,
(WebAssemblywrapper tglobaladdr:$off)))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD8_U_I64 0, tglobaladdr:$off, $addr)>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(i64 (extloadi16 (regPlusGA I32:$addr,
(WebAssemblywrapper tglobaladdr:$off)))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD16_U_I64 0, tglobaladdr:$off, $addr)>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(i64 (extloadi32 (regPlusGA I32:$addr,
(WebAssemblywrapper tglobaladdr:$off)))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD32_U_I64 0, tglobaladdr:$off, $addr)>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(i32 (extloadi8 (add I32:$addr,
(WebAssemblywrapper texternalsym:$off)))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD8_U_I32 0, texternalsym:$off, $addr)>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(i32 (extloadi16 (add I32:$addr,
(WebAssemblywrapper texternalsym:$off)))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD16_U_I32 0, texternalsym:$off, $addr)>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(i64 (extloadi8 (add I32:$addr,
(WebAssemblywrapper texternalsym:$off)))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD8_U_I64 0, texternalsym:$off, $addr)>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(i64 (extloadi16 (add I32:$addr,
(WebAssemblywrapper texternalsym:$off)))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD16_U_I64 0, texternalsym:$off, $addr)>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(i64 (extloadi32 (add I32:$addr,
(WebAssemblywrapper texternalsym:$off)))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD32_U_I64 0, texternalsym:$off, $addr)>;
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
// Select "don't care" extending loads with just a constant offset.
[WebAssembly] Implement unaligned loads and stores. Differential Revision: http://reviews.llvm.org/D16534 llvm-svn: 258779
2016-01-26 11:39:31 +08:00
def : Pat<(i32 (extloadi8 imm:$off)),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD8_U_I32 0, imm:$off, (CONST_I32 0))>;
[WebAssembly] Implement unaligned loads and stores. Differential Revision: http://reviews.llvm.org/D16534 llvm-svn: 258779
2016-01-26 11:39:31 +08:00
def : Pat<(i32 (extloadi16 imm:$off)),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD16_U_I32 0, imm:$off, (CONST_I32 0))>;
[WebAssembly] Implement unaligned loads and stores. Differential Revision: http://reviews.llvm.org/D16534 llvm-svn: 258779
2016-01-26 11:39:31 +08:00
def : Pat<(i64 (extloadi8 imm:$off)),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD8_U_I64 0, imm:$off, (CONST_I32 0))>;
[WebAssembly] Implement unaligned loads and stores. Differential Revision: http://reviews.llvm.org/D16534 llvm-svn: 258779
2016-01-26 11:39:31 +08:00
def : Pat<(i64 (extloadi16 imm:$off)),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD16_U_I64 0, imm:$off, (CONST_I32 0))>;
[WebAssembly] Implement unaligned loads and stores. Differential Revision: http://reviews.llvm.org/D16534 llvm-svn: 258779
2016-01-26 11:39:31 +08:00
def : Pat<(i64 (extloadi32 imm:$off)),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD32_U_I64 0, imm:$off, (CONST_I32 0))>;
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
def : Pat<(i32 (extloadi8 (WebAssemblywrapper tglobaladdr:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD8_U_I32 0, tglobaladdr:$off, (CONST_I32 0))>;
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
def : Pat<(i32 (extloadi16 (WebAssemblywrapper tglobaladdr:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD16_U_I32 0, tglobaladdr:$off, (CONST_I32 0))>;
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
def : Pat<(i64 (extloadi8 (WebAssemblywrapper tglobaladdr:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD8_U_I64 0, tglobaladdr:$off, (CONST_I32 0))>;
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
def : Pat<(i64 (extloadi16 (WebAssemblywrapper tglobaladdr:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD16_U_I64 0, tglobaladdr:$off, (CONST_I32 0))>;
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
def : Pat<(i64 (extloadi32 (WebAssemblywrapper tglobaladdr:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD32_U_I64 0, tglobaladdr:$off, (CONST_I32 0))>;
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
def : Pat<(i32 (extloadi8 (WebAssemblywrapper texternalsym:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD8_U_I32 0, texternalsym:$off, (CONST_I32 0))>;
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
def : Pat<(i32 (extloadi16 (WebAssemblywrapper texternalsym:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD16_U_I32 0, texternalsym:$off, (CONST_I32 0))>;
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
def : Pat<(i64 (extloadi8 (WebAssemblywrapper texternalsym:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD8_U_I64 0, texternalsym:$off, (CONST_I32 0))>;
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
def : Pat<(i64 (extloadi16 (WebAssemblywrapper texternalsym:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD16_U_I64 0, texternalsym:$off, (CONST_I32 0))>;
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
def : Pat<(i64 (extloadi32 (WebAssemblywrapper texternalsym:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(LOAD32_U_I64 0, tglobaladdr:$off, (CONST_I32 0))>;
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
[WebAssembly] Use a physical register to describe ARGUMENT liveness. Instead of trying to move ARGUMENT instructions back up to the top after they've been scheduled or sunk down, use a fake physical register to create a liveness constraint that prevents ARGUMENT instructions from moving down in the first place. This is still not entirely ideal, however it is more robust than letting them move and moving them back. llvm-svn: 254084
2015-11-26 03:36:19 +08:00
let Defs = [ARGUMENTS] in {
WebAssembly: generate load/store Summary: This handles all load/store operations that WebAssembly defines, and handles those necessary for C++ such as i1. I left a FIXME for outstanding features which aren't required for now. Reviewers: sunfish Subscribers: jfb, llvm-commits, dschuff llvm-svn: 246500
2015-09-01 06:24:11 +08:00
// Basic store.
// Note: WebAssembly inverts SelectionDAG's usual operand order.
[WebAssembly] Annotate call and load/store immediates. These will be used to guide the binary encoding of these immediates. llvm-svn: 290412
2016-12-23 11:23:52 +08:00
def STORE_I32 : I<(outs), (ins P2Align:$p2align, offset32_op:$off, I32:$addr,
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
I32:$val), [],
[WebAssembly] Update opcode values according to recent spec changes. This corresponds to the "0xd" opcode renumbering. llvm-svn: 285014
2016-10-25 04:21:49 +08:00
"i32.store\t${off}(${addr})${p2align}, $val", 0x36>;
[WebAssembly] Annotate call and load/store immediates. These will be used to guide the binary encoding of these immediates. llvm-svn: 290412
2016-12-23 11:23:52 +08:00
def STORE_I64 : I<(outs), (ins P2Align:$p2align, offset32_op:$off, I32:$addr,
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
I64:$val), [],
[WebAssembly] Update opcode values according to recent spec changes. This corresponds to the "0xd" opcode renumbering. llvm-svn: 285014
2016-10-25 04:21:49 +08:00
"i64.store\t${off}(${addr})${p2align}, $val", 0x37>;
[WebAssembly] Annotate call and load/store immediates. These will be used to guide the binary encoding of these immediates. llvm-svn: 290412
2016-12-23 11:23:52 +08:00
def STORE_F32 : I<(outs), (ins P2Align:$p2align, offset32_op:$off, I32:$addr,
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
F32:$val), [],
[WebAssembly] Update opcode values according to recent spec changes. This corresponds to the "0xd" opcode renumbering. llvm-svn: 285014
2016-10-25 04:21:49 +08:00
"f32.store\t${off}(${addr})${p2align}, $val", 0x38>;
[WebAssembly] Annotate call and load/store immediates. These will be used to guide the binary encoding of these immediates. llvm-svn: 290412
2016-12-23 11:23:52 +08:00
def STORE_F64 : I<(outs), (ins P2Align:$p2align, offset32_op:$off, I32:$addr,
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
F64:$val), [],
[WebAssembly] Update opcode values according to recent spec changes. This corresponds to the "0xd" opcode renumbering. llvm-svn: 285014
2016-10-25 04:21:49 +08:00
"f64.store\t${off}(${addr})${p2align}, $val", 0x39>;
[WebAssembly] Model the return value of store instructions in wasm. llvm-svn: 253916
2015-11-24 05:16:35 +08:00
[WebAssembly] Use a physical register to describe ARGUMENT liveness. Instead of trying to move ARGUMENT instructions back up to the top after they've been scheduled or sunk down, use a fake physical register to create a liveness constraint that prevents ARGUMENT instructions from moving down in the first place. This is still not entirely ideal, however it is more robust than letting them move and moving them back. llvm-svn: 254084
2015-11-26 03:36:19 +08:00
} // Defs = [ARGUMENTS]
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
// Select stores with no constant offset.
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
def : Pat<(store I32:$val, I32:$addr), (STORE_I32 0, 0, I32:$addr, I32:$val)>;
def : Pat<(store I64:$val, I32:$addr), (STORE_I64 0, 0, I32:$addr, I64:$val)>;
def : Pat<(store F32:$val, I32:$addr), (STORE_F32 0, 0, I32:$addr, F32:$val)>;
def : Pat<(store F64:$val, I32:$addr), (STORE_F64 0, 0, I32:$addr, F64:$val)>;
[WebAssembly] Support constant offsets on loads and stores This is just prototype for load/store for i32 types. I'll add them to the rest of the types if we like this direction. Differential Revision: http://reviews.llvm.org/D15197 llvm-svn: 254807
2015-12-05 08:26:39 +08:00
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
// Select stores with a constant offset.
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(store I32:$val, (regPlusImm I32:$addr, imm:$off)),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(STORE_I32 0, imm:$off, I32:$addr, I32:$val)>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(store I64:$val, (regPlusImm I32:$addr, imm:$off)),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(STORE_I64 0, imm:$off, I32:$addr, I64:$val)>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(store F32:$val, (regPlusImm I32:$addr, imm:$off)),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(STORE_F32 0, imm:$off, I32:$addr, F32:$val)>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(store F64:$val, (regPlusImm I32:$addr, imm:$off)),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(STORE_F64 0, imm:$off, I32:$addr, F64:$val)>;
[WebAssembly] Teach address folding to fold bitwise-or nodes. LLVM converts adds into ors when it can prove that the operands don't share any non-zero bits. Teach address folding to recognize or instructions with constant operands with this property that can be folded into addresses as if they were adds. llvm-svn: 261562
2016-02-23 04:04:02 +08:00
def : Pat<(store I32:$val, (or_is_add I32:$addr, imm:$off)),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(STORE_I32 0, imm:$off, I32:$addr, I32:$val)>;
[WebAssembly] Teach address folding to fold bitwise-or nodes. LLVM converts adds into ors when it can prove that the operands don't share any non-zero bits. Teach address folding to recognize or instructions with constant operands with this property that can be folded into addresses as if they were adds. llvm-svn: 261562
2016-02-23 04:04:02 +08:00
def : Pat<(store I64:$val, (or_is_add I32:$addr, imm:$off)),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(STORE_I64 0, imm:$off, I32:$addr, I64:$val)>;
[WebAssembly] Teach address folding to fold bitwise-or nodes. LLVM converts adds into ors when it can prove that the operands don't share any non-zero bits. Teach address folding to recognize or instructions with constant operands with this property that can be folded into addresses as if they were adds. llvm-svn: 261562
2016-02-23 04:04:02 +08:00
def : Pat<(store F32:$val, (or_is_add I32:$addr, imm:$off)),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(STORE_F32 0, imm:$off, I32:$addr, F32:$val)>;
[WebAssembly] Teach address folding to fold bitwise-or nodes. LLVM converts adds into ors when it can prove that the operands don't share any non-zero bits. Teach address folding to recognize or instructions with constant operands with this property that can be folded into addresses as if they were adds. llvm-svn: 261562
2016-02-23 04:04:02 +08:00
def : Pat<(store F64:$val, (or_is_add I32:$addr, imm:$off)),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(STORE_F64 0, imm:$off, I32:$addr, F64:$val)>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(store I32:$val, (regPlusGA I32:$addr,
(WebAssemblywrapper tglobaladdr:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(STORE_I32 0, tglobaladdr:$off, I32:$addr, I32:$val)>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(store I64:$val, (regPlusGA I32:$addr,
(WebAssemblywrapper tglobaladdr:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(STORE_I64 0, tglobaladdr:$off, I32:$addr, I64:$val)>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(store F32:$val, (regPlusGA I32:$addr,
(WebAssemblywrapper tglobaladdr:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(STORE_F32 0, tglobaladdr:$off, I32:$addr, F32:$val)>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(store F64:$val, (regPlusGA I32:$addr,
(WebAssemblywrapper tglobaladdr:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(STORE_F64 0, tglobaladdr:$off, I32:$addr, F64:$val)>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(store I32:$val, (add I32:$addr,
(WebAssemblywrapper texternalsym:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(STORE_I32 0, texternalsym:$off, I32:$addr, I32:$val)>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(store I64:$val, (add I32:$addr,
(WebAssemblywrapper texternalsym:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(STORE_I64 0, texternalsym:$off, I32:$addr, I64:$val)>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(store F32:$val, (add I32:$addr,
(WebAssemblywrapper texternalsym:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(STORE_F32 0, texternalsym:$off, I32:$addr, F32:$val)>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(store F64:$val, (add I32:$addr,
(WebAssemblywrapper texternalsym:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(STORE_F64 0, texternalsym:$off, I32:$addr, F64:$val)>;
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
// Select stores with just a constant offset.
def : Pat<(store I32:$val, imm:$off),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(STORE_I32 0, imm:$off, (CONST_I32 0), I32:$val)>;
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
def : Pat<(store I64:$val, imm:$off),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(STORE_I64 0, imm:$off, (CONST_I32 0), I64:$val)>;
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
def : Pat<(store F32:$val, imm:$off),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(STORE_F32 0, imm:$off, (CONST_I32 0), F32:$val)>;
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
def : Pat<(store F64:$val, imm:$off),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(STORE_F64 0, imm:$off, (CONST_I32 0), F64:$val)>;
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
def : Pat<(store I32:$val, (WebAssemblywrapper tglobaladdr:$off)),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(STORE_I32 0, tglobaladdr:$off, (CONST_I32 0), I32:$val)>;
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
def : Pat<(store I64:$val, (WebAssemblywrapper tglobaladdr:$off)),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(STORE_I64 0, tglobaladdr:$off, (CONST_I32 0), I64:$val)>;
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
def : Pat<(store F32:$val, (WebAssemblywrapper tglobaladdr:$off)),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(STORE_F32 0, tglobaladdr:$off, (CONST_I32 0), F32:$val)>;
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
def : Pat<(store F64:$val, (WebAssemblywrapper tglobaladdr:$off)),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(STORE_F64 0, tglobaladdr:$off, (CONST_I32 0), F64:$val)>;
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
def : Pat<(store I32:$val, (WebAssemblywrapper texternalsym:$off)),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(STORE_I32 0, texternalsym:$off, (CONST_I32 0), I32:$val)>;
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
def : Pat<(store I64:$val, (WebAssemblywrapper texternalsym:$off)),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(STORE_I64 0, texternalsym:$off, (CONST_I32 0), I64:$val)>;
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
def : Pat<(store F32:$val, (WebAssemblywrapper texternalsym:$off)),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(STORE_F32 0, texternalsym:$off, (CONST_I32 0), F32:$val)>;
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
def : Pat<(store F64:$val, (WebAssemblywrapper texternalsym:$off)),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(STORE_F64 0, texternalsym:$off, (CONST_I32 0), F64:$val)>;
WebAssembly: generate load/store Summary: This handles all load/store operations that WebAssembly defines, and handles those necessary for C++ such as i1. I left a FIXME for outstanding features which aren't required for now. Reviewers: sunfish Subscribers: jfb, llvm-commits, dschuff llvm-svn: 246500
2015-09-01 06:24:11 +08:00
[WebAssembly] Use a physical register to describe ARGUMENT liveness. Instead of trying to move ARGUMENT instructions back up to the top after they've been scheduled or sunk down, use a fake physical register to create a liveness constraint that prevents ARGUMENT instructions from moving down in the first place. This is still not entirely ideal, however it is more robust than letting them move and moving them back. llvm-svn: 254084
2015-11-26 03:36:19 +08:00
let Defs = [ARGUMENTS] in {
WebAssembly: generate load/store Summary: This handles all load/store operations that WebAssembly defines, and handles those necessary for C++ such as i1. I left a FIXME for outstanding features which aren't required for now. Reviewers: sunfish Subscribers: jfb, llvm-commits, dschuff llvm-svn: 246500
2015-09-01 06:24:11 +08:00
// Truncating store.
[WebAssembly] Annotate call and load/store immediates. These will be used to guide the binary encoding of these immediates. llvm-svn: 290412
2016-12-23 11:23:52 +08:00
def STORE8_I32 : I<(outs), (ins P2Align:$p2align, offset32_op:$off, I32:$addr,
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
I32:$val), [],
[WebAssembly] Update opcode values according to recent spec changes. This corresponds to the "0xd" opcode renumbering. llvm-svn: 285014
2016-10-25 04:21:49 +08:00
"i32.store8\t${off}(${addr})${p2align}, $val", 0x3a>;
[WebAssembly] Annotate call and load/store immediates. These will be used to guide the binary encoding of these immediates. llvm-svn: 290412
2016-12-23 11:23:52 +08:00
def STORE16_I32 : I<(outs), (ins P2Align:$p2align, offset32_op:$off, I32:$addr,
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
I32:$val), [],
[WebAssembly] Update opcode values according to recent spec changes. This corresponds to the "0xd" opcode renumbering. llvm-svn: 285014
2016-10-25 04:21:49 +08:00
"i32.store16\t${off}(${addr})${p2align}, $val", 0x3b>;
[WebAssembly] Annotate call and load/store immediates. These will be used to guide the binary encoding of these immediates. llvm-svn: 290412
2016-12-23 11:23:52 +08:00
def STORE8_I64 : I<(outs), (ins P2Align:$p2align, offset32_op:$off, I32:$addr,
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
I64:$val), [],
[WebAssembly] Update opcode values according to recent spec changes. This corresponds to the "0xd" opcode renumbering. llvm-svn: 285014
2016-10-25 04:21:49 +08:00
"i64.store8\t${off}(${addr})${p2align}, $val", 0x3c>;
[WebAssembly] Annotate call and load/store immediates. These will be used to guide the binary encoding of these immediates. llvm-svn: 290412
2016-12-23 11:23:52 +08:00
def STORE16_I64 : I<(outs), (ins P2Align:$p2align, offset32_op:$off, I32:$addr,
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
I64:$val), [],
[WebAssembly] Update opcode values according to recent spec changes. This corresponds to the "0xd" opcode renumbering. llvm-svn: 285014
2016-10-25 04:21:49 +08:00
"i64.store16\t${off}(${addr})${p2align}, $val", 0x3d>;
[WebAssembly] Annotate call and load/store immediates. These will be used to guide the binary encoding of these immediates. llvm-svn: 290412
2016-12-23 11:23:52 +08:00
def STORE32_I64 : I<(outs), (ins P2Align:$p2align, offset32_op:$off, I32:$addr,
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
I64:$val), [],
[WebAssembly] Update opcode values according to recent spec changes. This corresponds to the "0xd" opcode renumbering. llvm-svn: 285014
2016-10-25 04:21:49 +08:00
"i64.store32\t${off}(${addr})${p2align}, $val", 0x3e>;
[WebAssembly] Model the return value of store instructions in wasm. llvm-svn: 253916
2015-11-24 05:16:35 +08:00
[WebAssembly] Use a physical register to describe ARGUMENT liveness. Instead of trying to move ARGUMENT instructions back up to the top after they've been scheduled or sunk down, use a fake physical register to create a liveness constraint that prevents ARGUMENT instructions from moving down in the first place. This is still not entirely ideal, however it is more robust than letting them move and moving them back. llvm-svn: 254084
2015-11-26 03:36:19 +08:00
} // Defs = [ARGUMENTS]
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
// Select truncating stores with no constant offset.
[WebAssembly] Model the return value of store instructions in wasm. llvm-svn: 253916
2015-11-24 05:16:35 +08:00
def : Pat<(truncstorei8 I32:$val, I32:$addr),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(STORE8_I32 0, 0, I32:$addr, I32:$val)>;
[WebAssembly] Model the return value of store instructions in wasm. llvm-svn: 253916
2015-11-24 05:16:35 +08:00
def : Pat<(truncstorei16 I32:$val, I32:$addr),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(STORE16_I32 0, 0, I32:$addr, I32:$val)>;
[WebAssembly] Model the return value of store instructions in wasm. llvm-svn: 253916
2015-11-24 05:16:35 +08:00
def : Pat<(truncstorei8 I64:$val, I32:$addr),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(STORE8_I64 0, 0, I32:$addr, I64:$val)>;
[WebAssembly] Model the return value of store instructions in wasm. llvm-svn: 253916
2015-11-24 05:16:35 +08:00
def : Pat<(truncstorei16 I64:$val, I32:$addr),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(STORE16_I64 0, 0, I32:$addr, I64:$val)>;
[WebAssembly] Model the return value of store instructions in wasm. llvm-svn: 253916
2015-11-24 05:16:35 +08:00
def : Pat<(truncstorei32 I64:$val, I32:$addr),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(STORE32_I64 0, 0, I32:$addr, I64:$val)>;
WebAssembly: generate load/store Summary: This handles all load/store operations that WebAssembly defines, and handles those necessary for C++ such as i1. I left a FIXME for outstanding features which aren't required for now. Reviewers: sunfish Subscribers: jfb, llvm-commits, dschuff llvm-svn: 246500
2015-09-01 06:24:11 +08:00
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
// Select truncating stores with a constant offset.
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(truncstorei8 I32:$val, (regPlusImm I32:$addr, imm:$off)),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(STORE8_I32 0, imm:$off, I32:$addr, I32:$val)>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(truncstorei16 I32:$val, (regPlusImm I32:$addr, imm:$off)),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(STORE16_I32 0, imm:$off, I32:$addr, I32:$val)>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(truncstorei8 I64:$val, (regPlusImm I32:$addr, imm:$off)),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(STORE8_I64 0, imm:$off, I32:$addr, I64:$val)>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(truncstorei16 I64:$val, (regPlusImm I32:$addr, imm:$off)),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(STORE16_I64 0, imm:$off, I32:$addr, I64:$val)>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(truncstorei32 I64:$val, (regPlusImm I32:$addr, imm:$off)),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(STORE32_I64 0, imm:$off, I32:$addr, I64:$val)>;
[WebAssembly] Teach address folding to fold bitwise-or nodes. LLVM converts adds into ors when it can prove that the operands don't share any non-zero bits. Teach address folding to recognize or instructions with constant operands with this property that can be folded into addresses as if they were adds. llvm-svn: 261562
2016-02-23 04:04:02 +08:00
def : Pat<(truncstorei8 I32:$val, (or_is_add I32:$addr, imm:$off)),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(STORE8_I32 0, imm:$off, I32:$addr, I32:$val)>;
[WebAssembly] Teach address folding to fold bitwise-or nodes. LLVM converts adds into ors when it can prove that the operands don't share any non-zero bits. Teach address folding to recognize or instructions with constant operands with this property that can be folded into addresses as if they were adds. llvm-svn: 261562
2016-02-23 04:04:02 +08:00
def : Pat<(truncstorei16 I32:$val, (or_is_add I32:$addr, imm:$off)),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(STORE16_I32 0, imm:$off, I32:$addr, I32:$val)>;
[WebAssembly] Teach address folding to fold bitwise-or nodes. LLVM converts adds into ors when it can prove that the operands don't share any non-zero bits. Teach address folding to recognize or instructions with constant operands with this property that can be folded into addresses as if they were adds. llvm-svn: 261562
2016-02-23 04:04:02 +08:00
def : Pat<(truncstorei8 I64:$val, (or_is_add I32:$addr, imm:$off)),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(STORE8_I64 0, imm:$off, I32:$addr, I64:$val)>;
[WebAssembly] Teach address folding to fold bitwise-or nodes. LLVM converts adds into ors when it can prove that the operands don't share any non-zero bits. Teach address folding to recognize or instructions with constant operands with this property that can be folded into addresses as if they were adds. llvm-svn: 261562
2016-02-23 04:04:02 +08:00
def : Pat<(truncstorei16 I64:$val, (or_is_add I32:$addr, imm:$off)),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(STORE16_I64 0, imm:$off, I32:$addr, I64:$val)>;
[WebAssembly] Teach address folding to fold bitwise-or nodes. LLVM converts adds into ors when it can prove that the operands don't share any non-zero bits. Teach address folding to recognize or instructions with constant operands with this property that can be folded into addresses as if they were adds. llvm-svn: 261562
2016-02-23 04:04:02 +08:00
def : Pat<(truncstorei32 I64:$val, (or_is_add I32:$addr, imm:$off)),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(STORE32_I64 0, imm:$off, I32:$addr, I64:$val)>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(truncstorei8 I32:$val,
(regPlusGA I32:$addr,
(WebAssemblywrapper tglobaladdr:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(STORE8_I32 0, tglobaladdr:$off, I32:$addr, I32:$val)>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(truncstorei16 I32:$val,
(regPlusGA I32:$addr,
(WebAssemblywrapper tglobaladdr:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(STORE16_I32 0, tglobaladdr:$off, I32:$addr, I32:$val)>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(truncstorei8 I64:$val,
(regPlusGA I32:$addr,
(WebAssemblywrapper tglobaladdr:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(STORE8_I64 0, tglobaladdr:$off, I32:$addr, I64:$val)>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(truncstorei16 I64:$val,
(regPlusGA I32:$addr,
(WebAssemblywrapper tglobaladdr:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(STORE16_I64 0, tglobaladdr:$off, I32:$addr, I64:$val)>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(truncstorei32 I64:$val,
(regPlusGA I32:$addr,
(WebAssemblywrapper tglobaladdr:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(STORE32_I64 0, tglobaladdr:$off, I32:$addr, I64:$val)>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(truncstorei8 I32:$val, (add I32:$addr,
(WebAssemblywrapper texternalsym:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(STORE8_I32 0, texternalsym:$off, I32:$addr, I32:$val)>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(truncstorei16 I32:$val,
(add I32:$addr,
(WebAssemblywrapper texternalsym:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(STORE16_I32 0, texternalsym:$off, I32:$addr, I32:$val)>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(truncstorei8 I64:$val,
(add I32:$addr,
(WebAssemblywrapper texternalsym:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(STORE8_I64 0, texternalsym:$off, I32:$addr, I64:$val)>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(truncstorei16 I64:$val,
(add I32:$addr,
(WebAssemblywrapper texternalsym:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(STORE16_I64 0, texternalsym:$off, I32:$addr, I64:$val)>;
[WebAssembly] Reorganize address offset folding. Always expect tglobaladdr and texternalsym to be wrapped in WebAssemblywrapper nodes. Also, split out a regPlusGA from regPlusImm so that it can special-case global addresses, as they can be folded in more cases. Unfortunately this doesn't enable any new optimizations yet due to SelectionDAG limitations. I'll be submitting changes to the SelectionDAG infrastructure, along with tests, in a separate patch. llvm-svn: 257394
2016-01-12 06:05:44 +08:00
def : Pat<(truncstorei32 I64:$val,
(add I32:$addr,
(WebAssemblywrapper texternalsym:$off))),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(STORE32_I64 0, texternalsym:$off, I32:$addr, I64:$val)>;
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
// Select truncating stores with just a constant offset.
def : Pat<(truncstorei8 I32:$val, imm:$off),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(STORE8_I32 0, imm:$off, (CONST_I32 0), I32:$val)>;
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
def : Pat<(truncstorei16 I32:$val, imm:$off),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(STORE16_I32 0, imm:$off, (CONST_I32 0), I32:$val)>;
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
def : Pat<(truncstorei8 I64:$val, imm:$off),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(STORE8_I64 0, imm:$off, (CONST_I32 0), I64:$val)>;
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
def : Pat<(truncstorei16 I64:$val, imm:$off),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(STORE16_I64 0, imm:$off, (CONST_I32 0), I64:$val)>;
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
def : Pat<(truncstorei32 I64:$val, imm:$off),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(STORE32_I64 0, imm:$off, (CONST_I32 0), I64:$val)>;
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
def : Pat<(truncstorei8 I32:$val, (WebAssemblywrapper tglobaladdr:$off)),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(STORE8_I32 0, tglobaladdr:$off, (CONST_I32 0), I32:$val)>;
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
def : Pat<(truncstorei16 I32:$val, (WebAssemblywrapper tglobaladdr:$off)),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(STORE16_I32 0, tglobaladdr:$off, (CONST_I32 0), I32:$val)>;
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
def : Pat<(truncstorei8 I64:$val, (WebAssemblywrapper tglobaladdr:$off)),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(STORE8_I64 0, tglobaladdr:$off, (CONST_I32 0), I64:$val)>;
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
def : Pat<(truncstorei16 I64:$val, (WebAssemblywrapper tglobaladdr:$off)),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(STORE16_I64 0, tglobaladdr:$off, (CONST_I32 0), I64:$val)>;
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
def : Pat<(truncstorei32 I64:$val, (WebAssemblywrapper tglobaladdr:$off)),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(STORE32_I64 0, tglobaladdr:$off, (CONST_I32 0), I64:$val)>;
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
def : Pat<(truncstorei8 I32:$val, (WebAssemblywrapper texternalsym:$off)),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(STORE8_I32 0, texternalsym:$off, (CONST_I32 0), I32:$val)>;
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
def : Pat<(truncstorei16 I32:$val, (WebAssemblywrapper texternalsym:$off)),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(STORE16_I32 0, texternalsym:$off, (CONST_I32 0), I32:$val)>;
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
def : Pat<(truncstorei8 I64:$val, (WebAssemblywrapper texternalsym:$off)),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(STORE8_I64 0, texternalsym:$off, (CONST_I32 0), I64:$val)>;
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
def : Pat<(truncstorei16 I64:$val, (WebAssemblywrapper texternalsym:$off)),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(STORE16_I64 0, texternalsym:$off, (CONST_I32 0), I64:$val)>;
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
def : Pat<(truncstorei32 I64:$val, (WebAssemblywrapper texternalsym:$off)),
[WebAssembly] Reorder load/store operands to match binary encoding. The p2align operand of a load/store is encoded before the offset operand; reorder the MachineInstr operands accordingly. llvm-svn: 285044
2016-10-25 08:17:11 +08:00
(STORE32_I64 0, texternalsym:$off, (CONST_I32 0), I64:$val)>;
[WebAssembly] Implement instruction selection for constant offsets in addresses. Add instruction patterns for matching load and store instructions with constant offsets in addresses. The code is fairly redundant due to the need to replicate everything between imm, tglobaldadr, and texternalsym, but this appears to be common tablegen practice. The main alternative appears to be to introduce matching functions with C++ code, but sticking with purely generated matchers seems better for now. Also note that this doesn't yet support offsets from getelementptr, which will be the most common case; that will depend on a change in target-independent code in order to set the NoUnsignedWrap flag, which I'll submit separately. Until then, the testcase uses ptrtoint+add+inttoptr with a nuw on the add. Also implement isLegalAddressingMode with an approximation of this. Differential Revision: http://reviews.llvm.org/D15538 llvm-svn: 255681
2015-12-16 06:01:29 +08:00
[WebAssembly] Use a physical register to describe ARGUMENT liveness. Instead of trying to move ARGUMENT instructions back up to the top after they've been scheduled or sunk down, use a fake physical register to create a liveness constraint that prevents ARGUMENT instructions from moving down in the first place. This is still not entirely ideal, however it is more robust than letting them move and moving them back. llvm-svn: 254084
2015-11-26 03:36:19 +08:00
let Defs = [ARGUMENTS] in {
[WebAssembly] Rename memory_size intrinsic to current_memory This follows the recent renaming in the wasm spec. llvm-svn: 268255
2016-05-03 01:25:22 +08:00
// Current memory size.
[WebAssembly] Add immediate fields to call_indirect and memory operators. call_indirect, grow_memory, and current_memory now have immediate operands in the 0xd binary encoding. llvm-svn: 285085
2016-10-26 00:55:52 +08:00
def CURRENT_MEMORY_I32 : I<(outs I32:$dst), (ins i32imm:$flags),
[],
[WebAssembly] Update opcode values according to recent spec changes. This corresponds to the "0xd" opcode renumbering. llvm-svn: 285014
2016-10-25 04:21:49 +08:00
"current_memory\t$dst", 0x3f>,
[WebAssembly] Rename memory_size intrinsic to current_memory This follows the recent renaming in the wasm spec. llvm-svn: 268255
2016-05-03 01:25:22 +08:00
Requires<[HasAddr32]>;
[WebAssembly] Add a resize_memory intrinsic. llvm-svn: 249178
2015-10-03 04:10:26 +08:00
[WebAssembly] Update wasm builtin functions to match spec changes. The page_size operator has been removed from the spec, and the resize_memory operator has been changed to grow_memory. llvm-svn: 252202
2015-11-06 04:16:59 +08:00
// Grow memory.
[WebAssembly] Add immediate fields to call_indirect and memory operators. call_indirect, grow_memory, and current_memory now have immediate operands in the 0xd binary encoding. llvm-svn: 285085
2016-10-26 00:55:52 +08:00
def GROW_MEMORY_I32 : I<(outs), (ins i32imm:$flags, I32:$delta),
[],
[WebAssembly] Update opcode values according to recent spec changes. This corresponds to the "0xd" opcode renumbering. llvm-svn: 285014
2016-10-25 04:21:49 +08:00
"grow_memory\t$delta", 0x40>,
[WebAssembly] Update wasm builtin functions to match spec changes. The page_size operator has been removed from the spec, and the resize_memory operator has been changed to grow_memory. llvm-svn: 252202
2015-11-06 04:16:59 +08:00
Requires<[HasAddr32]>;
[WebAssembly] Use a physical register to describe ARGUMENT liveness. Instead of trying to move ARGUMENT instructions back up to the top after they've been scheduled or sunk down, use a fake physical register to create a liveness constraint that prevents ARGUMENT instructions from moving down in the first place. This is still not entirely ideal, however it is more robust than letting them move and moving them back. llvm-svn: 254084
2015-11-26 03:36:19 +08:00
} // Defs = [ARGUMENTS]
[WebAssembly] Add immediate fields to call_indirect and memory operators. call_indirect, grow_memory, and current_memory now have immediate operands in the 0xd binary encoding. llvm-svn: 285085
2016-10-26 00:55:52 +08:00
def : Pat<(int_wasm_current_memory),
(CURRENT_MEMORY_I32 0)>;
def : Pat<(int_wasm_grow_memory I32:$delta),
(GROW_MEMORY_I32 0, $delta)>;