6502/llvm-6502
1
0
Fork 0
mirror of https://github.com/c64scene-ar/llvm-6502.git synced 2024-12-30 17:33:24 +00:00
Code Issues Projects Releases Wiki Activity
2712c87cfe
llvm-6502/lib/Target/SystemZ/SystemZOperators.td

386 lines
20 KiB
TableGen
Raw Normal View History

[SystemZ] Add back end This adds the actual lib/Target/SystemZ target files necessary to implement the SystemZ target. Note that at this point, the target cannot yet be built since the configure bits are missing. Those will be provided shortly by a follow-on patch. This version of the patch incorporates feedback from reviews by Chris Lattner and Anton Korobeynikov. Thanks to all reviewers! Patch by Richard Sandiford. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@181203 91177308-0d34-0410-b5e6-96231b3b80d8
2013-05-06 16:15:19 +00:00
//===-- SystemZOperators.td - SystemZ-specific operators ------*- tblgen-*-===//
//
// The LLVM Compiler Infrastructure
//
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
//===----------------------------------------------------------------------===//
// Type profiles
//===----------------------------------------------------------------------===//
def SDT_CallSeqStart : SDCallSeqStart<[SDTCisVT<0, i64>]>;
def SDT_CallSeqEnd : SDCallSeqEnd<[SDTCisVT<0, i64>,
SDTCisVT<1, i64>]>;
def SDT_ZCall : SDTypeProfile<0, -1, [SDTCisPtrTy<0>]>;
def SDT_ZCmp : SDTypeProfile<0, 2, [SDTCisSameAs<0, 1>]>;
[SystemZ] Tweak integer comparison code The architecture has many comparison instructions, including some that extend one of the operands. The signed comparison instructions use sign extensions and the unsigned comparison instructions use zero extensions. In cases where we had a free choice between signed or unsigned comparisons, we were trying to decide at lowering time which would best fit the available instructions, taking things like extension type into account. The code to do that was getting increasingly hairy and was also making some bad decisions. E.g. when comparing the result of two LLCs, it is better to use CR rather than CLR, since CR can be fused with a branch while CLR can't. This patch removes the lowering code and instead adds an operand to integer comparisons to say whether signed comparison is required, whether unsigned comparison is required, or whether either is OK. We can then leave the choice of instruction up to the normal isel code. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@190138 91177308-0d34-0410-b5e6-96231b3b80d8
2013-09-06 11:51:39 +00:00
def SDT_ZICmp : SDTypeProfile<0, 3,
[SDTCisSameAs<0, 1>,
SDTCisVT<2, i32>]>;
[SystemZ] Be more careful about inverting CC masks System z branches have a mask to select which of the 4 CC values should cause the branch to be taken. We can invert a branch by inverting the mask. However, not all instructions can produce all 4 CC values, so inverting the branch like this can lead to some oddities. For example, integer comparisons only produce a CC of 0 (equal), 1 (less) or 2 (greater). If an integer EQ is reversed to NE before instruction selection, the branch will test for 1 or 2. If instead the branch is reversed after instruction selection (by inverting the mask), it will test for 1, 2 or 3. Both are correct, but the second isn't really canonical. This patch therefore keeps track of which CC values are possible and uses this when inverting a mask. Although this is mostly cosmestic, it fixes undefined behavior for the CIJNLH in branch-08.ll. Another fix would have been to mask out bit 0 when generating the fused compare and branch, but the point of this patch is that we shouldn't need to do that in the first place. The patch also makes it easier to reuse CC results from other instructions. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@187495 91177308-0d34-0410-b5e6-96231b3b80d8
2013-07-31 12:30:20 +00:00
def SDT_ZBRCCMask : SDTypeProfile<0, 3,
[SystemZ] Add back end This adds the actual lib/Target/SystemZ target files necessary to implement the SystemZ target. Note that at this point, the target cannot yet be built since the configure bits are missing. Those will be provided shortly by a follow-on patch. This version of the patch incorporates feedback from reviews by Chris Lattner and Anton Korobeynikov. Thanks to all reviewers! Patch by Richard Sandiford. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@181203 91177308-0d34-0410-b5e6-96231b3b80d8
2013-05-06 16:15:19 +00:00
[SDTCisVT<0, i8>,
[SystemZ] Be more careful about inverting CC masks System z branches have a mask to select which of the 4 CC values should cause the branch to be taken. We can invert a branch by inverting the mask. However, not all instructions can produce all 4 CC values, so inverting the branch like this can lead to some oddities. For example, integer comparisons only produce a CC of 0 (equal), 1 (less) or 2 (greater). If an integer EQ is reversed to NE before instruction selection, the branch will test for 1 or 2. If instead the branch is reversed after instruction selection (by inverting the mask), it will test for 1, 2 or 3. Both are correct, but the second isn't really canonical. This patch therefore keeps track of which CC values are possible and uses this when inverting a mask. Although this is mostly cosmestic, it fixes undefined behavior for the CIJNLH in branch-08.ll. Another fix would have been to mask out bit 0 when generating the fused compare and branch, but the point of this patch is that we shouldn't need to do that in the first place. The patch also makes it easier to reuse CC results from other instructions. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@187495 91177308-0d34-0410-b5e6-96231b3b80d8
2013-07-31 12:30:20 +00:00
SDTCisVT<1, i8>,
SDTCisVT<2, OtherVT>]>;
def SDT_ZSelectCCMask : SDTypeProfile<1, 4,
[SystemZ] Add back end This adds the actual lib/Target/SystemZ target files necessary to implement the SystemZ target. Note that at this point, the target cannot yet be built since the configure bits are missing. Those will be provided shortly by a follow-on patch. This version of the patch incorporates feedback from reviews by Chris Lattner and Anton Korobeynikov. Thanks to all reviewers! Patch by Richard Sandiford. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@181203 91177308-0d34-0410-b5e6-96231b3b80d8
2013-05-06 16:15:19 +00:00
[SDTCisSameAs<0, 1>,
SDTCisSameAs<1, 2>,
[SystemZ] Be more careful about inverting CC masks System z branches have a mask to select which of the 4 CC values should cause the branch to be taken. We can invert a branch by inverting the mask. However, not all instructions can produce all 4 CC values, so inverting the branch like this can lead to some oddities. For example, integer comparisons only produce a CC of 0 (equal), 1 (less) or 2 (greater). If an integer EQ is reversed to NE before instruction selection, the branch will test for 1 or 2. If instead the branch is reversed after instruction selection (by inverting the mask), it will test for 1, 2 or 3. Both are correct, but the second isn't really canonical. This patch therefore keeps track of which CC values are possible and uses this when inverting a mask. Although this is mostly cosmestic, it fixes undefined behavior for the CIJNLH in branch-08.ll. Another fix would have been to mask out bit 0 when generating the fused compare and branch, but the point of this patch is that we shouldn't need to do that in the first place. The patch also makes it easier to reuse CC results from other instructions. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@187495 91177308-0d34-0410-b5e6-96231b3b80d8
2013-07-31 12:30:20 +00:00
SDTCisVT<3, i8>,
SDTCisVT<4, i8>]>;
[SystemZ] Add back end This adds the actual lib/Target/SystemZ target files necessary to implement the SystemZ target. Note that at this point, the target cannot yet be built since the configure bits are missing. Those will be provided shortly by a follow-on patch. This version of the patch incorporates feedback from reviews by Chris Lattner and Anton Korobeynikov. Thanks to all reviewers! Patch by Richard Sandiford. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@181203 91177308-0d34-0410-b5e6-96231b3b80d8
2013-05-06 16:15:19 +00:00
def SDT_ZWrapPtr : SDTypeProfile<1, 1,
[SDTCisSameAs<0, 1>,
SDTCisPtrTy<0>]>;
[SystemZ] Improve handling of PC-relative addresses The backend previously folded offsets into PC-relative addresses whereever possible. That's the right thing to do when the address can be used directly in a PC-relative memory reference (using things like LRL). But if we have a register-based memory reference and need to load the PC-relative address separately, it's better to use an anchor point that could be shared with other accesses to the same area of the variable. Fixes a FIXME. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@191524 91177308-0d34-0410-b5e6-96231b3b80d8
2013-09-27 15:14:04 +00:00
def SDT_ZWrapOffset : SDTypeProfile<1, 2,
[SDTCisSameAs<0, 1>,
SDTCisSameAs<0, 2>,
SDTCisPtrTy<0>]>;
[SystemZ] Add back end This adds the actual lib/Target/SystemZ target files necessary to implement the SystemZ target. Note that at this point, the target cannot yet be built since the configure bits are missing. Those will be provided shortly by a follow-on patch. This version of the patch incorporates feedback from reviews by Chris Lattner and Anton Korobeynikov. Thanks to all reviewers! Patch by Richard Sandiford. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@181203 91177308-0d34-0410-b5e6-96231b3b80d8
2013-05-06 16:15:19 +00:00
def SDT_ZAdjDynAlloc : SDTypeProfile<1, 0, [SDTCisVT<0, i64>]>;
def SDT_ZExtractAccess : SDTypeProfile<1, 1,
[SDTCisVT<0, i32>,
SDTCisVT<1, i8>]>;
def SDT_ZGR128Binary32 : SDTypeProfile<1, 2,
[SDTCisVT<0, untyped>,
SDTCisVT<1, untyped>,
SDTCisVT<2, i32>]>;
def SDT_ZGR128Binary64 : SDTypeProfile<1, 2,
[SDTCisVT<0, untyped>,
SDTCisVT<1, untyped>,
SDTCisVT<2, i64>]>;
def SDT_ZAtomicLoadBinaryW : SDTypeProfile<1, 5,
[SDTCisVT<0, i32>,
SDTCisPtrTy<1>,
SDTCisVT<2, i32>,
SDTCisVT<3, i32>,
SDTCisVT<4, i32>,
SDTCisVT<5, i32>]>;
def SDT_ZAtomicCmpSwapW : SDTypeProfile<1, 6,
[SDTCisVT<0, i32>,
SDTCisPtrTy<1>,
SDTCisVT<2, i32>,
SDTCisVT<3, i32>,
SDTCisVT<4, i32>,
SDTCisVT<5, i32>,
SDTCisVT<6, i32>]>;
[SystemZ] Add a definition of the CLC instruction git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@188162 91177308-0d34-0410-b5e6-96231b3b80d8
2013-08-12 10:17:33 +00:00
def SDT_ZMemMemLength : SDTypeProfile<0, 3,
[SystemZ] Use MVC for memcpy Use MVC for memcpy in cases where a single MVC is enough. Using MVC is a win for longer copies too, but I'll leave that for later. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@185802 91177308-0d34-0410-b5e6-96231b3b80d8
2013-07-08 09:35:23 +00:00
[SDTCisPtrTy<0>,
SDTCisPtrTy<1>,
[SystemZ] Extend memcpy and memset support to all constant lengths Lengths up to a certain threshold (currently 6 * 256) use a series of MVCs. Lengths above that threshold use a loop to handle X*256 bytes followed by a single MVC to handle the excess (if any). This loop will also be needed in future when support for variable lengths is added. Because the same tablegen classes are used to define MVC and CLC, the patch also has the side-effect of defining a pseudo loop instruction for CLC. That instruction isn't used yet (and wouldn't be handled correctly if it were). I'm planning to use it soon though. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@189331 91177308-0d34-0410-b5e6-96231b3b80d8
2013-08-27 09:54:29 +00:00
SDTCisVT<2, i64>]>;
def SDT_ZMemMemLoop : SDTypeProfile<0, 4,
[SDTCisPtrTy<0>,
SDTCisPtrTy<1>,
SDTCisVT<2, i64>,
SDTCisVT<3, i64>]>;
[SystemZ] Use CLST to implement strcmp git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@188544 91177308-0d34-0410-b5e6-96231b3b80d8
2013-08-16 11:21:54 +00:00
def SDT_ZString : SDTypeProfile<1, 3,
[SDTCisPtrTy<0>,
SDTCisPtrTy<1>,
SDTCisPtrTy<2>,
SDTCisVT<3, i32>]>;
[SystemZ] Use CLC and IPM to implement memcmp For now this is restricted to fixed-length comparisons with a length in the range [1, 256], as for memcpy() and MVC. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@188163 91177308-0d34-0410-b5e6-96231b3b80d8
2013-08-12 10:28:10 +00:00
def SDT_ZI32Intrinsic : SDTypeProfile<1, 0, [SDTCisVT<0, i32>]>;
[SystemZ] Add basic prefetch support Just the instructions and intrinsics for now. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@189100 91177308-0d34-0410-b5e6-96231b3b80d8
2013-08-23 11:36:42 +00:00
def SDT_ZPrefetch : SDTypeProfile<0, 2,
[SDTCisVT<0, i8>,
SDTCisPtrTy<1>]>;
[SystemZ] Add back end This adds the actual lib/Target/SystemZ target files necessary to implement the SystemZ target. Note that at this point, the target cannot yet be built since the configure bits are missing. Those will be provided shortly by a follow-on patch. This version of the patch incorporates feedback from reviews by Chris Lattner and Anton Korobeynikov. Thanks to all reviewers! Patch by Richard Sandiford. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@181203 91177308-0d34-0410-b5e6-96231b3b80d8
2013-05-06 16:15:19 +00:00
//===----------------------------------------------------------------------===//
// Node definitions
//===----------------------------------------------------------------------===//
// These are target-independent nodes, but have target-specific formats.
def callseq_start : SDNode<"ISD::CALLSEQ_START", SDT_CallSeqStart,
[SDNPHasChain, SDNPSideEffect, SDNPOutGlue]>;
def callseq_end : SDNode<"ISD::CALLSEQ_END", SDT_CallSeqEnd,
[SDNPHasChain, SDNPSideEffect, SDNPOptInGlue,
SDNPOutGlue]>;
// Nodes for SystemZISD::*. See SystemZISelLowering.h for more details.
def z_retflag : SDNode<"SystemZISD::RET_FLAG", SDTNone,
[SDNPHasChain, SDNPOptInGlue, SDNPVariadic]>;
def z_call : SDNode<"SystemZISD::CALL", SDT_ZCall,
[SDNPHasChain, SDNPOutGlue, SDNPOptInGlue,
SDNPVariadic]>;
[SystemZ] Add support for sibling calls This first cut is pretty conservative. The final argument register (R6) is call-saved, so we would need to make sure that the R6 argument to a sibling call is the same as the R6 argument to the calling function, which seems worth keeping as a separate patch. Saying that integer truncations are free means that we no longer use the extending instructions LGF and LLGF for spills in int-conv-09.ll and int-conv-10.ll. Instead we treat the registers as 64 bits wide and truncate them to 32-bits where necessary. I think it's unlikely we'd use LGF and LLGF for spills in other situations for the same reason, so I'm removing the tests rather than replacing them. The associated code is generic and applies to many more instructions than just LGF and LLGF, so there is no corresponding code removal. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@188669 91177308-0d34-0410-b5e6-96231b3b80d8
2013-08-19 12:42:31 +00:00
def z_sibcall : SDNode<"SystemZISD::SIBCALL", SDT_ZCall,
[SDNPHasChain, SDNPOutGlue, SDNPOptInGlue,
SDNPVariadic]>;
[SystemZ] Add back end This adds the actual lib/Target/SystemZ target files necessary to implement the SystemZ target. Note that at this point, the target cannot yet be built since the configure bits are missing. Those will be provided shortly by a follow-on patch. This version of the patch incorporates feedback from reviews by Chris Lattner and Anton Korobeynikov. Thanks to all reviewers! Patch by Richard Sandiford. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@181203 91177308-0d34-0410-b5e6-96231b3b80d8
2013-05-06 16:15:19 +00:00
def z_pcrel_wrapper : SDNode<"SystemZISD::PCREL_WRAPPER", SDT_ZWrapPtr, []>;
[SystemZ] Improve handling of PC-relative addresses The backend previously folded offsets into PC-relative addresses whereever possible. That's the right thing to do when the address can be used directly in a PC-relative memory reference (using things like LRL). But if we have a register-based memory reference and need to load the PC-relative address separately, it's better to use an anchor point that could be shared with other accesses to the same area of the variable. Fixes a FIXME. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@191524 91177308-0d34-0410-b5e6-96231b3b80d8
2013-09-27 15:14:04 +00:00
def z_pcrel_offset : SDNode<"SystemZISD::PCREL_OFFSET",
SDT_ZWrapOffset, []>;
[SystemZ] Extend integer absolute selection This patch makes more use of LPGFR and LNGFR. It builds on top of the LTGFR selection from r197234. Most of the tests are motivated by what InstCombine would produce. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@197236 91177308-0d34-0410-b5e6-96231b3b80d8
2013-12-13 15:35:00 +00:00
def z_iabs : SDNode<"SystemZISD::IABS", SDTIntUnaryOp, []>;
[SystemZ] Tweak integer comparison code The architecture has many comparison instructions, including some that extend one of the operands. The signed comparison instructions use sign extensions and the unsigned comparison instructions use zero extensions. In cases where we had a free choice between signed or unsigned comparisons, we were trying to decide at lowering time which would best fit the available instructions, taking things like extension type into account. The code to do that was getting increasingly hairy and was also making some bad decisions. E.g. when comparing the result of two LLCs, it is better to use CR rather than CLR, since CR can be fused with a branch while CLR can't. This patch removes the lowering code and instead adds an operand to integer comparisons to say whether signed comparison is required, whether unsigned comparison is required, or whether either is OK. We can then leave the choice of instruction up to the normal isel code. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@190138 91177308-0d34-0410-b5e6-96231b3b80d8
2013-09-06 11:51:39 +00:00
def z_icmp : SDNode<"SystemZISD::ICMP", SDT_ZICmp, [SDNPOutGlue]>;
def z_fcmp : SDNode<"SystemZISD::FCMP", SDT_ZCmp, [SDNPOutGlue]>;
[SystemZ] Add TM and TMY The main complication here is that TM and TMY (the memory forms) set CC differently from the register forms. When the tested bits contain some 0s and some 1s, the register forms set CC to 1 or 2 based on the value the uppermost bit. The memory forms instead set CC to 1 regardless of the uppermost bit. Until now, I've tried to make it so that a branch never tests for an impossible CC value. E.g. NR only sets CC to 0 or 1, so branches on the result will only test for 0 or 1. Originally I'd tried to do the same thing for TM and TMY by using custom matching code in ISelDAGToDAG. That ended up being very ugly though, and would have meant duplicating some of the chain checks that the common isel code does. I've therefore gone for the simpler alternative of adding an extra operand to the TM DAG opcode to say whether a memory form would be OK. This means that the inverse of a "TM;JE" is "TM;JNE" rather than the more precise "TM;JNLE", just like the inverse of "TMLL;JE" is "TMLL;JNE". I suppose that's arguably less confusing though... git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@190400 91177308-0d34-0410-b5e6-96231b3b80d8
2013-09-10 10:20:32 +00:00
def z_tm : SDNode<"SystemZISD::TM", SDT_ZICmp, [SDNPOutGlue]>;
[SystemZ] Add back end This adds the actual lib/Target/SystemZ target files necessary to implement the SystemZ target. Note that at this point, the target cannot yet be built since the configure bits are missing. Those will be provided shortly by a follow-on patch. This version of the patch incorporates feedback from reviews by Chris Lattner and Anton Korobeynikov. Thanks to all reviewers! Patch by Richard Sandiford. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@181203 91177308-0d34-0410-b5e6-96231b3b80d8
2013-05-06 16:15:19 +00:00
def z_br_ccmask : SDNode<"SystemZISD::BR_CCMASK", SDT_ZBRCCMask,
[SDNPHasChain, SDNPInGlue]>;
def z_select_ccmask : SDNode<"SystemZISD::SELECT_CCMASK", SDT_ZSelectCCMask,
[SDNPInGlue]>;
def z_adjdynalloc : SDNode<"SystemZISD::ADJDYNALLOC", SDT_ZAdjDynAlloc>;
def z_extract_access : SDNode<"SystemZISD::EXTRACT_ACCESS",
SDT_ZExtractAccess>;
def z_umul_lohi64 : SDNode<"SystemZISD::UMUL_LOHI64", SDT_ZGR128Binary64>;
[SystemZ] Use DSGFR over DSGR in more cases Fixes some cases where we were using full 64-bit division for (sdiv i32, i32) and (sdiv i64, i32). The "32" in "SDIVREM32" just refers to the second operand. The first operand of all *DIVREM*s is a GR128. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@185435 91177308-0d34-0410-b5e6-96231b3b80d8
2013-07-02 15:40:22 +00:00
def z_sdivrem32 : SDNode<"SystemZISD::SDIVREM32", SDT_ZGR128Binary32>;
[SystemZ] Add back end This adds the actual lib/Target/SystemZ target files necessary to implement the SystemZ target. Note that at this point, the target cannot yet be built since the configure bits are missing. Those will be provided shortly by a follow-on patch. This version of the patch incorporates feedback from reviews by Chris Lattner and Anton Korobeynikov. Thanks to all reviewers! Patch by Richard Sandiford. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@181203 91177308-0d34-0410-b5e6-96231b3b80d8
2013-05-06 16:15:19 +00:00
def z_sdivrem64 : SDNode<"SystemZISD::SDIVREM64", SDT_ZGR128Binary64>;
def z_udivrem32 : SDNode<"SystemZISD::UDIVREM32", SDT_ZGR128Binary32>;
def z_udivrem64 : SDNode<"SystemZISD::UDIVREM64", SDT_ZGR128Binary64>;
Add TargetLowering::prepareVolatileOrAtomicLoad One unusual feature of the z architecture is that the result of a previous load can be reused indefinitely for subsequent loads, even if a cache-coherent store to that location is performed by another CPU. A special serializing instruction must be used if you want to force a load to be reattempted. Since volatile loads are not supposed to be omitted in this way, we should insert a serializing instruction before each such load. The same goes for atomic loads. The patch implements this at the IR->DAG boundary, in a similar way to atomic fences. It is a no-op for targets other than SystemZ. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@196905 91177308-0d34-0410-b5e6-96231b3b80d8
2013-12-10 10:36:34 +00:00
def z_serialize : SDNode<"SystemZISD::SERIALIZE", SDTNone,
[SDNPHasChain, SDNPMayStore]>;
[SystemZ] Add back end This adds the actual lib/Target/SystemZ target files necessary to implement the SystemZ target. Note that at this point, the target cannot yet be built since the configure bits are missing. Those will be provided shortly by a follow-on patch. This version of the patch incorporates feedback from reviews by Chris Lattner and Anton Korobeynikov. Thanks to all reviewers! Patch by Richard Sandiford. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@181203 91177308-0d34-0410-b5e6-96231b3b80d8
2013-05-06 16:15:19 +00:00
class AtomicWOp<string name, SDTypeProfile profile = SDT_ZAtomicLoadBinaryW>
: SDNode<"SystemZISD::"##name, profile,
[SDNPHasChain, SDNPMayStore, SDNPMayLoad, SDNPMemOperand]>;
def z_atomic_swapw : AtomicWOp<"ATOMIC_SWAPW">;
def z_atomic_loadw_add : AtomicWOp<"ATOMIC_LOADW_ADD">;
def z_atomic_loadw_sub : AtomicWOp<"ATOMIC_LOADW_SUB">;
def z_atomic_loadw_and : AtomicWOp<"ATOMIC_LOADW_AND">;
def z_atomic_loadw_or : AtomicWOp<"ATOMIC_LOADW_OR">;
def z_atomic_loadw_xor : AtomicWOp<"ATOMIC_LOADW_XOR">;
def z_atomic_loadw_nand : AtomicWOp<"ATOMIC_LOADW_NAND">;
def z_atomic_loadw_min : AtomicWOp<"ATOMIC_LOADW_MIN">;
def z_atomic_loadw_max : AtomicWOp<"ATOMIC_LOADW_MAX">;
def z_atomic_loadw_umin : AtomicWOp<"ATOMIC_LOADW_UMIN">;
def z_atomic_loadw_umax : AtomicWOp<"ATOMIC_LOADW_UMAX">;
def z_atomic_cmp_swapw : AtomicWOp<"ATOMIC_CMP_SWAPW", SDT_ZAtomicCmpSwapW>;
[SystemZ] Add a definition of the CLC instruction git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@188162 91177308-0d34-0410-b5e6-96231b3b80d8
2013-08-12 10:17:33 +00:00
def z_mvc : SDNode<"SystemZISD::MVC", SDT_ZMemMemLength,
[SystemZ] Use MVC for memcpy Use MVC for memcpy in cases where a single MVC is enough. Using MVC is a win for longer copies too, but I'll leave that for later. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@185802 91177308-0d34-0410-b5e6-96231b3b80d8
2013-07-08 09:35:23 +00:00
[SDNPHasChain, SDNPMayStore, SDNPMayLoad]>;
[SystemZ] Extend memcpy and memset support to all constant lengths Lengths up to a certain threshold (currently 6 * 256) use a series of MVCs. Lengths above that threshold use a loop to handle X*256 bytes followed by a single MVC to handle the excess (if any). This loop will also be needed in future when support for variable lengths is added. Because the same tablegen classes are used to define MVC and CLC, the patch also has the side-effect of defining a pseudo loop instruction for CLC. That instruction isn't used yet (and wouldn't be handled correctly if it were). I'm planning to use it soon though. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@189331 91177308-0d34-0410-b5e6-96231b3b80d8
2013-08-27 09:54:29 +00:00
def z_mvc_loop : SDNode<"SystemZISD::MVC_LOOP", SDT_ZMemMemLoop,
[SDNPHasChain, SDNPMayStore, SDNPMayLoad]>;
[SystemZ] Add NC, OC and XC For now these are just used to handle scalar ANDs, ORs and XORs in which all operands are memory. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@190041 91177308-0d34-0410-b5e6-96231b3b80d8
2013-09-05 10:36:45 +00:00
def z_nc : SDNode<"SystemZISD::NC", SDT_ZMemMemLength,
[SDNPHasChain, SDNPMayStore, SDNPMayLoad]>;
def z_nc_loop : SDNode<"SystemZISD::NC_LOOP", SDT_ZMemMemLoop,
[SDNPHasChain, SDNPMayStore, SDNPMayLoad]>;
def z_oc : SDNode<"SystemZISD::OC", SDT_ZMemMemLength,
[SDNPHasChain, SDNPMayStore, SDNPMayLoad]>;
def z_oc_loop : SDNode<"SystemZISD::OC_LOOP", SDT_ZMemMemLoop,
[SDNPHasChain, SDNPMayStore, SDNPMayLoad]>;
def z_xc : SDNode<"SystemZISD::XC", SDT_ZMemMemLength,
[SDNPHasChain, SDNPMayStore, SDNPMayLoad]>;
def z_xc_loop : SDNode<"SystemZISD::XC_LOOP", SDT_ZMemMemLoop,
[SDNPHasChain, SDNPMayStore, SDNPMayLoad]>;
[SystemZ] Add a definition of the CLC instruction git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@188162 91177308-0d34-0410-b5e6-96231b3b80d8
2013-08-12 10:17:33 +00:00
def z_clc : SDNode<"SystemZISD::CLC", SDT_ZMemMemLength,
[SystemZ] Use CLC and IPM to implement memcmp For now this is restricted to fixed-length comparisons with a length in the range [1, 256], as for memcpy() and MVC. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@188163 91177308-0d34-0410-b5e6-96231b3b80d8
2013-08-12 10:28:10 +00:00
[SDNPHasChain, SDNPOutGlue, SDNPMayLoad]>;
[SystemZ] Extend memcpy and memset support to all constant lengths Lengths up to a certain threshold (currently 6 * 256) use a series of MVCs. Lengths above that threshold use a loop to handle X*256 bytes followed by a single MVC to handle the excess (if any). This loop will also be needed in future when support for variable lengths is added. Because the same tablegen classes are used to define MVC and CLC, the patch also has the side-effect of defining a pseudo loop instruction for CLC. That instruction isn't used yet (and wouldn't be handled correctly if it were). I'm planning to use it soon though. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@189331 91177308-0d34-0410-b5e6-96231b3b80d8
2013-08-27 09:54:29 +00:00
def z_clc_loop : SDNode<"SystemZISD::CLC_LOOP", SDT_ZMemMemLoop,
[SDNPHasChain, SDNPOutGlue, SDNPMayLoad]>;
[SystemZ] Use CLST to implement strcmp git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@188544 91177308-0d34-0410-b5e6-96231b3b80d8
2013-08-16 11:21:54 +00:00
def z_strcmp : SDNode<"SystemZISD::STRCMP", SDT_ZString,
[SDNPHasChain, SDNPOutGlue, SDNPMayLoad]>;
[SystemZ] Use MVST to implement strcpy and stpcpy git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@188546 91177308-0d34-0410-b5e6-96231b3b80d8
2013-08-16 11:29:37 +00:00
def z_stpcpy : SDNode<"SystemZISD::STPCPY", SDT_ZString,
[SDNPHasChain, SDNPMayStore, SDNPMayLoad]>;
[SystemZ] Use SRST to implement strlen and strnlen It would also make sense to use it for memchr; I'm working on that now. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@188547 91177308-0d34-0410-b5e6-96231b3b80d8
2013-08-16 11:41:43 +00:00
def z_search_string : SDNode<"SystemZISD::SEARCH_STRING", SDT_ZString,
[SDNPHasChain, SDNPOutGlue, SDNPMayLoad]>;
[SystemZ] Use CLC and IPM to implement memcmp For now this is restricted to fixed-length comparisons with a length in the range [1, 256], as for memcpy() and MVC. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@188163 91177308-0d34-0410-b5e6-96231b3b80d8
2013-08-12 10:28:10 +00:00
def z_ipm : SDNode<"SystemZISD::IPM", SDT_ZI32Intrinsic,
[SDNPInGlue]>;
[SystemZ] Add basic prefetch support Just the instructions and intrinsics for now. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@189100 91177308-0d34-0410-b5e6-96231b3b80d8
2013-08-23 11:36:42 +00:00
def z_prefetch : SDNode<"SystemZISD::PREFETCH", SDT_ZPrefetch,
[SDNPHasChain, SDNPMayLoad, SDNPMayStore,
SDNPMemOperand]>;
[SystemZ] Use MVC for memcpy Use MVC for memcpy in cases where a single MVC is enough. Using MVC is a win for longer copies too, but I'll leave that for later. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@185802 91177308-0d34-0410-b5e6-96231b3b80d8
2013-07-08 09:35:23 +00:00
[SystemZ] Add back end This adds the actual lib/Target/SystemZ target files necessary to implement the SystemZ target. Note that at this point, the target cannot yet be built since the configure bits are missing. Those will be provided shortly by a follow-on patch. This version of the patch incorporates feedback from reviews by Chris Lattner and Anton Korobeynikov. Thanks to all reviewers! Patch by Richard Sandiford. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@181203 91177308-0d34-0410-b5e6-96231b3b80d8
2013-05-06 16:15:19 +00:00
//===----------------------------------------------------------------------===//
// Pattern fragments
//===----------------------------------------------------------------------===//
[SystemZ] Tweak integer comparison code The architecture has many comparison instructions, including some that extend one of the operands. The signed comparison instructions use sign extensions and the unsigned comparison instructions use zero extensions. In cases where we had a free choice between signed or unsigned comparisons, we were trying to decide at lowering time which would best fit the available instructions, taking things like extension type into account. The code to do that was getting increasingly hairy and was also making some bad decisions. E.g. when comparing the result of two LLCs, it is better to use CR rather than CLR, since CR can be fused with a branch while CLR can't. This patch removes the lowering code and instead adds an operand to integer comparisons to say whether signed comparison is required, whether unsigned comparison is required, or whether either is OK. We can then leave the choice of instruction up to the normal isel code. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@190138 91177308-0d34-0410-b5e6-96231b3b80d8
2013-09-06 11:51:39 +00:00
// Signed and unsigned comparisons.
def z_scmp : PatFrag<(ops node:$a, node:$b), (z_icmp node:$a, node:$b, imm), [{
unsigned Type = cast<ConstantSDNode>(N->getOperand(2))->getZExtValue();
return Type != SystemZICMP::UnsignedOnly;
}]>;
def z_ucmp : PatFrag<(ops node:$a, node:$b), (z_icmp node:$a, node:$b, imm), [{
unsigned Type = cast<ConstantSDNode>(N->getOperand(2))->getZExtValue();
return Type != SystemZICMP::SignedOnly;
}]>;
[SystemZ] Add TM and TMY The main complication here is that TM and TMY (the memory forms) set CC differently from the register forms. When the tested bits contain some 0s and some 1s, the register forms set CC to 1 or 2 based on the value the uppermost bit. The memory forms instead set CC to 1 regardless of the uppermost bit. Until now, I've tried to make it so that a branch never tests for an impossible CC value. E.g. NR only sets CC to 0 or 1, so branches on the result will only test for 0 or 1. Originally I'd tried to do the same thing for TM and TMY by using custom matching code in ISelDAGToDAG. That ended up being very ugly though, and would have meant duplicating some of the chain checks that the common isel code does. I've therefore gone for the simpler alternative of adding an extra operand to the TM DAG opcode to say whether a memory form would be OK. This means that the inverse of a "TM;JE" is "TM;JNE" rather than the more precise "TM;JNLE", just like the inverse of "TMLL;JE" is "TMLL;JNE". I suppose that's arguably less confusing though... git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@190400 91177308-0d34-0410-b5e6-96231b3b80d8
2013-09-10 10:20:32 +00:00
// Register- and memory-based TEST UNDER MASK.
def z_tm_reg : PatFrag<(ops node:$a, node:$b), (z_tm node:$a, node:$b, imm)>;
def z_tm_mem : PatFrag<(ops node:$a, node:$b), (z_tm node:$a, node:$b, 0)>;
[SystemZ] Add back end This adds the actual lib/Target/SystemZ target files necessary to implement the SystemZ target. Note that at this point, the target cannot yet be built since the configure bits are missing. Those will be provided shortly by a follow-on patch. This version of the patch incorporates feedback from reviews by Chris Lattner and Anton Korobeynikov. Thanks to all reviewers! Patch by Richard Sandiford. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@181203 91177308-0d34-0410-b5e6-96231b3b80d8
2013-05-06 16:15:19 +00:00
// Register sign-extend operations. Sub-32-bit values are represented as i32s.
def sext8 : PatFrag<(ops node:$src), (sext_inreg node:$src, i8)>;
def sext16 : PatFrag<(ops node:$src), (sext_inreg node:$src, i16)>;
def sext32 : PatFrag<(ops node:$src), (sext (i32 node:$src))>;
// Register zero-extend operations. Sub-32-bit values are represented as i32s.
def zext8 : PatFrag<(ops node:$src), (and node:$src, 0xff)>;
def zext16 : PatFrag<(ops node:$src), (and node:$src, 0xffff)>;
def zext32 : PatFrag<(ops node:$src), (zext (i32 node:$src))>;
// Typed floating-point loads.
def loadf32 : PatFrag<(ops node:$src), (f32 (load node:$src))>;
def loadf64 : PatFrag<(ops node:$src), (f64 (load node:$src))>;
[SystemZ] Improve extload handling The port originally had special patterns for extload, mapping them to the same instructions as sextload. It seemed neater to have patterns that match "an extension that is allowed to be signed" and "an extension that is allowed to be unsigned". This was originally meant to be a clean-up, but it does improve the handling of promoted integers a little, as shown by args-06.ll. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@190777 91177308-0d34-0410-b5e6-96231b3b80d8
2013-09-16 09:03:10 +00:00
// Extending loads in which the extension type can be signed.
def asextload : PatFrag<(ops node:$ptr), (unindexedload node:$ptr), [{
unsigned Type = cast<LoadSDNode>(N)->getExtensionType();
return Type == ISD::EXTLOAD || Type == ISD::SEXTLOAD;
}]>;
def asextloadi8 : PatFrag<(ops node:$ptr), (asextload node:$ptr), [{
return cast<LoadSDNode>(N)->getMemoryVT() == MVT::i8;
}]>;
def asextloadi16 : PatFrag<(ops node:$ptr), (asextload node:$ptr), [{
return cast<LoadSDNode>(N)->getMemoryVT() == MVT::i16;
}]>;
def asextloadi32 : PatFrag<(ops node:$ptr), (asextload node:$ptr), [{
return cast<LoadSDNode>(N)->getMemoryVT() == MVT::i32;
}]>;
// Extending loads in which the extension type can be unsigned.
def azextload : PatFrag<(ops node:$ptr), (unindexedload node:$ptr), [{
unsigned Type = cast<LoadSDNode>(N)->getExtensionType();
return Type == ISD::EXTLOAD || Type == ISD::ZEXTLOAD;
}]>;
def azextloadi8 : PatFrag<(ops node:$ptr), (azextload node:$ptr), [{
return cast<LoadSDNode>(N)->getMemoryVT() == MVT::i8;
}]>;
def azextloadi16 : PatFrag<(ops node:$ptr), (azextload node:$ptr), [{
return cast<LoadSDNode>(N)->getMemoryVT() == MVT::i16;
}]>;
def azextloadi32 : PatFrag<(ops node:$ptr), (azextload node:$ptr), [{
return cast<LoadSDNode>(N)->getMemoryVT() == MVT::i32;
}]>;
[SystemZ] Add conditional store patterns Add pseudo conditional store instructions, so that we use: branch foo: store foo: instead of: load branch foo: move foo: store z196 has real 32-bit and 64-bit conditional stores, but we don't use any z196 instructions yet. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@185065 91177308-0d34-0410-b5e6-96231b3b80d8
2013-06-27 09:27:40 +00:00
// Extending loads in which the extension type doesn't matter.
def anyextload : PatFrag<(ops node:$ptr), (unindexedload node:$ptr), [{
return cast<LoadSDNode>(N)->getExtensionType() != ISD::NON_EXTLOAD;
}]>;
def anyextloadi8 : PatFrag<(ops node:$ptr), (anyextload node:$ptr), [{
return cast<LoadSDNode>(N)->getMemoryVT() == MVT::i8;
}]>;
def anyextloadi16 : PatFrag<(ops node:$ptr), (anyextload node:$ptr), [{
return cast<LoadSDNode>(N)->getMemoryVT() == MVT::i16;
}]>;
def anyextloadi32 : PatFrag<(ops node:$ptr), (anyextload node:$ptr), [{
return cast<LoadSDNode>(N)->getMemoryVT() == MVT::i32;
}]>;
[SystemZ] Add back end This adds the actual lib/Target/SystemZ target files necessary to implement the SystemZ target. Note that at this point, the target cannot yet be built since the configure bits are missing. Those will be provided shortly by a follow-on patch. This version of the patch incorporates feedback from reviews by Chris Lattner and Anton Korobeynikov. Thanks to all reviewers! Patch by Richard Sandiford. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@181203 91177308-0d34-0410-b5e6-96231b3b80d8
2013-05-06 16:15:19 +00:00
// Aligned loads.
class AlignedLoad<SDPatternOperator load>
: PatFrag<(ops node:$addr), (load node:$addr), [{
[SystemZ] Use "auto" for cast results No functional change intended. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@203106 91177308-0d34-0410-b5e6-96231b3b80d8
2014-03-06 11:22:58 +00:00
auto *Load = cast<LoadSDNode>(N);
[SystemZ] Add back end This adds the actual lib/Target/SystemZ target files necessary to implement the SystemZ target. Note that at this point, the target cannot yet be built since the configure bits are missing. Those will be provided shortly by a follow-on patch. This version of the patch incorporates feedback from reviews by Chris Lattner and Anton Korobeynikov. Thanks to all reviewers! Patch by Richard Sandiford. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@181203 91177308-0d34-0410-b5e6-96231b3b80d8
2013-05-06 16:15:19 +00:00
return Load->getAlignment() >= Load->getMemoryVT().getStoreSize();
}]>;
[SystemZ] Improve extload handling The port originally had special patterns for extload, mapping them to the same instructions as sextload. It seemed neater to have patterns that match "an extension that is allowed to be signed" and "an extension that is allowed to be unsigned". This was originally meant to be a clean-up, but it does improve the handling of promoted integers a little, as shown by args-06.ll. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@190777 91177308-0d34-0410-b5e6-96231b3b80d8
2013-09-16 09:03:10 +00:00
def aligned_load : AlignedLoad<load>;
def aligned_asextloadi16 : AlignedLoad<asextloadi16>;
def aligned_asextloadi32 : AlignedLoad<asextloadi32>;
def aligned_azextloadi16 : AlignedLoad<azextloadi16>;
def aligned_azextloadi32 : AlignedLoad<azextloadi32>;
[SystemZ] Add back end This adds the actual lib/Target/SystemZ target files necessary to implement the SystemZ target. Note that at this point, the target cannot yet be built since the configure bits are missing. Those will be provided shortly by a follow-on patch. This version of the patch incorporates feedback from reviews by Chris Lattner and Anton Korobeynikov. Thanks to all reviewers! Patch by Richard Sandiford. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@181203 91177308-0d34-0410-b5e6-96231b3b80d8
2013-05-06 16:15:19 +00:00
// Aligned stores.
class AlignedStore<SDPatternOperator store>
: PatFrag<(ops node:$src, node:$addr), (store node:$src, node:$addr), [{
[SystemZ] Use "auto" for cast results No functional change intended. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@203106 91177308-0d34-0410-b5e6-96231b3b80d8
2014-03-06 11:22:58 +00:00
auto *Store = cast<StoreSDNode>(N);
[SystemZ] Add back end This adds the actual lib/Target/SystemZ target files necessary to implement the SystemZ target. Note that at this point, the target cannot yet be built since the configure bits are missing. Those will be provided shortly by a follow-on patch. This version of the patch incorporates feedback from reviews by Chris Lattner and Anton Korobeynikov. Thanks to all reviewers! Patch by Richard Sandiford. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@181203 91177308-0d34-0410-b5e6-96231b3b80d8
2013-05-06 16:15:19 +00:00
return Store->getAlignment() >= Store->getMemoryVT().getStoreSize();
}]>;
def aligned_store : AlignedStore<store>;
def aligned_truncstorei16 : AlignedStore<truncstorei16>;
def aligned_truncstorei32 : AlignedStore<truncstorei32>;
[SystemZ] Don't use LOAD and STORE REVERSED for volatile accesses Unlike most -- hopefully "all other", but I'm still checking -- memory instructions we support, LOAD REVERSED and STORE REVERSED may access the memory location several times. This means that they are not suitable for volatile loads and stores. This patch is a prerequisite for better atomic load and store support. The same principle applies there: almost all memory instructions we support are inherently atomic ("block concurrent"), but LOAD REVERSED and STORE REVERSED are exceptions. Other instructions continue to allow volatile operands. I will add positive "allows volatile" tests at the same time as the "allows atomic load or store" tests. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@183002 91177308-0d34-0410-b5e6-96231b3b80d8
2013-05-31 13:25:22 +00:00
// Non-volatile loads. Used for instructions that might access the storage
// location multiple times.
class NonvolatileLoad<SDPatternOperator load>
: PatFrag<(ops node:$addr), (load node:$addr), [{
[SystemZ] Use "auto" for cast results No functional change intended. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@203106 91177308-0d34-0410-b5e6-96231b3b80d8
2014-03-06 11:22:58 +00:00
auto *Load = cast<LoadSDNode>(N);
[SystemZ] Don't use LOAD and STORE REVERSED for volatile accesses Unlike most -- hopefully "all other", but I'm still checking -- memory instructions we support, LOAD REVERSED and STORE REVERSED may access the memory location several times. This means that they are not suitable for volatile loads and stores. This patch is a prerequisite for better atomic load and store support. The same principle applies there: almost all memory instructions we support are inherently atomic ("block concurrent"), but LOAD REVERSED and STORE REVERSED are exceptions. Other instructions continue to allow volatile operands. I will add positive "allows volatile" tests at the same time as the "allows atomic load or store" tests. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@183002 91177308-0d34-0410-b5e6-96231b3b80d8
2013-05-31 13:25:22 +00:00
return !Load->isVolatile();
}]>;
[SystemZ] Add conditional store patterns Add pseudo conditional store instructions, so that we use: branch foo: store foo: instead of: load branch foo: move foo: store z196 has real 32-bit and 64-bit conditional stores, but we don't use any z196 instructions yet. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@185065 91177308-0d34-0410-b5e6-96231b3b80d8
2013-06-27 09:27:40 +00:00
def nonvolatile_load : NonvolatileLoad<load>;
def nonvolatile_anyextloadi8 : NonvolatileLoad<anyextloadi8>;
def nonvolatile_anyextloadi16 : NonvolatileLoad<anyextloadi16>;
def nonvolatile_anyextloadi32 : NonvolatileLoad<anyextloadi32>;
[SystemZ] Don't use LOAD and STORE REVERSED for volatile accesses Unlike most -- hopefully "all other", but I'm still checking -- memory instructions we support, LOAD REVERSED and STORE REVERSED may access the memory location several times. This means that they are not suitable for volatile loads and stores. This patch is a prerequisite for better atomic load and store support. The same principle applies there: almost all memory instructions we support are inherently atomic ("block concurrent"), but LOAD REVERSED and STORE REVERSED are exceptions. Other instructions continue to allow volatile operands. I will add positive "allows volatile" tests at the same time as the "allows atomic load or store" tests. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@183002 91177308-0d34-0410-b5e6-96231b3b80d8
2013-05-31 13:25:22 +00:00
// Non-volatile stores.
class NonvolatileStore<SDPatternOperator store>
: PatFrag<(ops node:$src, node:$addr), (store node:$src, node:$addr), [{
[SystemZ] Use "auto" for cast results No functional change intended. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@203106 91177308-0d34-0410-b5e6-96231b3b80d8
2014-03-06 11:22:58 +00:00
auto *Store = cast<StoreSDNode>(N);
[SystemZ] Don't use LOAD and STORE REVERSED for volatile accesses Unlike most -- hopefully "all other", but I'm still checking -- memory instructions we support, LOAD REVERSED and STORE REVERSED may access the memory location several times. This means that they are not suitable for volatile loads and stores. This patch is a prerequisite for better atomic load and store support. The same principle applies there: almost all memory instructions we support are inherently atomic ("block concurrent"), but LOAD REVERSED and STORE REVERSED are exceptions. Other instructions continue to allow volatile operands. I will add positive "allows volatile" tests at the same time as the "allows atomic load or store" tests. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@183002 91177308-0d34-0410-b5e6-96231b3b80d8
2013-05-31 13:25:22 +00:00
return !Store->isVolatile();
}]>;
[SystemZ] Add conditional store patterns Add pseudo conditional store instructions, so that we use: branch foo: store foo: instead of: load branch foo: move foo: store z196 has real 32-bit and 64-bit conditional stores, but we don't use any z196 instructions yet. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@185065 91177308-0d34-0410-b5e6-96231b3b80d8
2013-06-27 09:27:40 +00:00
def nonvolatile_store : NonvolatileStore<store>;
def nonvolatile_truncstorei8 : NonvolatileStore<truncstorei8>;
def nonvolatile_truncstorei16 : NonvolatileStore<truncstorei16>;
def nonvolatile_truncstorei32 : NonvolatileStore<truncstorei32>;
[SystemZ] Don't use LOAD and STORE REVERSED for volatile accesses Unlike most -- hopefully "all other", but I'm still checking -- memory instructions we support, LOAD REVERSED and STORE REVERSED may access the memory location several times. This means that they are not suitable for volatile loads and stores. This patch is a prerequisite for better atomic load and store support. The same principle applies there: almost all memory instructions we support are inherently atomic ("block concurrent"), but LOAD REVERSED and STORE REVERSED are exceptions. Other instructions continue to allow volatile operands. I will add positive "allows volatile" tests at the same time as the "allows atomic load or store" tests. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@183002 91177308-0d34-0410-b5e6-96231b3b80d8
2013-05-31 13:25:22 +00:00
[SystemZ] Add NC, OC and XC For now these are just used to handle scalar ANDs, ORs and XORs in which all operands are memory. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@190041 91177308-0d34-0410-b5e6-96231b3b80d8
2013-09-05 10:36:45 +00:00
// A store of a load that can be implemented using MVC.
def mvc_store : PatFrag<(ops node:$value, node:$addr),
(unindexedstore node:$value, node:$addr),
[{ return storeLoadCanUseMVC(N); }]>;
// Binary read-modify-write operations on memory in which the other
// operand is also memory and for which block operations like NC can
// be used. There are two patterns for each operator, depending on
// which operand contains the "other" load.
multiclass block_op<SDPatternOperator operator> {
def "1" : PatFrag<(ops node:$value, node:$addr),
(unindexedstore (operator node:$value,
(unindexedload node:$addr)),
node:$addr),
[{ return storeLoadCanUseBlockBinary(N, 0); }]>;
def "2" : PatFrag<(ops node:$value, node:$addr),
(unindexedstore (operator (unindexedload node:$addr),
node:$value),
node:$addr),
[{ return storeLoadCanUseBlockBinary(N, 1); }]>;
}
defm block_and : block_op<and>;
defm block_or : block_op<or>;
defm block_xor : block_op<xor>;
[SystemZ] Add back end This adds the actual lib/Target/SystemZ target files necessary to implement the SystemZ target. Note that at this point, the target cannot yet be built since the configure bits are missing. Those will be provided shortly by a follow-on patch. This version of the patch incorporates feedback from reviews by Chris Lattner and Anton Korobeynikov. Thanks to all reviewers! Patch by Richard Sandiford. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@181203 91177308-0d34-0410-b5e6-96231b3b80d8
2013-05-06 16:15:19 +00:00
// Insertions.
def inserti8 : PatFrag<(ops node:$src1, node:$src2),
(or (and node:$src1, -256), node:$src2)>;
def insertll : PatFrag<(ops node:$src1, node:$src2),
(or (and node:$src1, 0xffffffffffff0000), node:$src2)>;
def insertlh : PatFrag<(ops node:$src1, node:$src2),
(or (and node:$src1, 0xffffffff0000ffff), node:$src2)>;
def inserthl : PatFrag<(ops node:$src1, node:$src2),
(or (and node:$src1, 0xffff0000ffffffff), node:$src2)>;
def inserthh : PatFrag<(ops node:$src1, node:$src2),
(or (and node:$src1, 0x0000ffffffffffff), node:$src2)>;
def insertlf : PatFrag<(ops node:$src1, node:$src2),
(or (and node:$src1, 0xffffffff00000000), node:$src2)>;
def inserthf : PatFrag<(ops node:$src1, node:$src2),
(or (and node:$src1, 0x00000000ffffffff), node:$src2)>;
// ORs that can be treated as insertions.
def or_as_inserti8 : PatFrag<(ops node:$src1, node:$src2),
(or node:$src1, node:$src2), [{
unsigned BitWidth = N->getValueType(0).getScalarType().getSizeInBits();
return CurDAG->MaskedValueIsZero(N->getOperand(0),
APInt::getLowBitsSet(BitWidth, 8));
}]>;
// ORs that can be treated as reversed insertions.
def or_as_revinserti8 : PatFrag<(ops node:$src1, node:$src2),
(or node:$src1, node:$src2), [{
unsigned BitWidth = N->getValueType(0).getScalarType().getSizeInBits();
return CurDAG->MaskedValueIsZero(N->getOperand(1),
APInt::getLowBitsSet(BitWidth, 8));
}]>;
[SystemZ] Extend integer absolute selection This patch makes more use of LPGFR and LNGFR. It builds on top of the LTGFR selection from r197234. Most of the tests are motivated by what InstCombine would produce. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@197236 91177308-0d34-0410-b5e6-96231b3b80d8
2013-12-13 15:35:00 +00:00
// Negative integer absolute.
def z_inegabs : PatFrag<(ops node:$src), (ineg (z_iabs node:$src))>;
[SystemZ] Add integer absolute (load positive) git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@188670 91177308-0d34-0410-b5e6-96231b3b80d8
2013-08-19 12:48:54 +00:00
// Integer absolute, matching the canonical form generated by DAGCombiner.
def z_iabs32 : PatFrag<(ops node:$src),
(xor (add node:$src, (sra node:$src, (i32 31))),
(sra node:$src, (i32 31)))>;
def z_iabs64 : PatFrag<(ops node:$src),
(xor (add node:$src, (sra node:$src, (i32 63))),
(sra node:$src, (i32 63)))>;
[SystemZ] Add negative integer absolute (load negative) For now this matches the equivalent of (neg (abs ...)), which did hit a few times in projects/test-suite. We should probably also match cases where absolute-like selects are used with reversed arguments. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@188671 91177308-0d34-0410-b5e6-96231b3b80d8
2013-08-19 12:56:58 +00:00
def z_inegabs32 : PatFrag<(ops node:$src), (ineg (z_iabs32 node:$src))>;
def z_inegabs64 : PatFrag<(ops node:$src), (ineg (z_iabs64 node:$src))>;
[SystemZ] Add integer absolute (load positive) git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@188670 91177308-0d34-0410-b5e6-96231b3b80d8
2013-08-19 12:48:54 +00:00
[SystemZ] Add back end This adds the actual lib/Target/SystemZ target files necessary to implement the SystemZ target. Note that at this point, the target cannot yet be built since the configure bits are missing. Those will be provided shortly by a follow-on patch. This version of the patch incorporates feedback from reviews by Chris Lattner and Anton Korobeynikov. Thanks to all reviewers! Patch by Richard Sandiford. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@181203 91177308-0d34-0410-b5e6-96231b3b80d8
2013-05-06 16:15:19 +00:00
// Fused multiply-add and multiply-subtract, but with the order of the
// operands matching SystemZ's MA and MS instructions.
def z_fma : PatFrag<(ops node:$src1, node:$src2, node:$src3),
(fma node:$src2, node:$src3, node:$src1)>;
def z_fms : PatFrag<(ops node:$src1, node:$src2, node:$src3),
(fma node:$src2, node:$src3, (fneg node:$src1))>;
// Floating-point negative absolute.
def fnabs : PatFrag<(ops node:$ptr), (fneg (fabs node:$ptr))>;
// Create a unary operator that loads from memory and then performs
// the given operation on it.
[SystemZ] Don't use LOAD and STORE REVERSED for volatile accesses Unlike most -- hopefully "all other", but I'm still checking -- memory instructions we support, LOAD REVERSED and STORE REVERSED may access the memory location several times. This means that they are not suitable for volatile loads and stores. This patch is a prerequisite for better atomic load and store support. The same principle applies there: almost all memory instructions we support are inherently atomic ("block concurrent"), but LOAD REVERSED and STORE REVERSED are exceptions. Other instructions continue to allow volatile operands. I will add positive "allows volatile" tests at the same time as the "allows atomic load or store" tests. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@183002 91177308-0d34-0410-b5e6-96231b3b80d8
2013-05-31 13:25:22 +00:00
class loadu<SDPatternOperator operator, SDPatternOperator load = load>
[SystemZ] Add back end This adds the actual lib/Target/SystemZ target files necessary to implement the SystemZ target. Note that at this point, the target cannot yet be built since the configure bits are missing. Those will be provided shortly by a follow-on patch. This version of the patch incorporates feedback from reviews by Chris Lattner and Anton Korobeynikov. Thanks to all reviewers! Patch by Richard Sandiford. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@181203 91177308-0d34-0410-b5e6-96231b3b80d8
2013-05-06 16:15:19 +00:00
: PatFrag<(ops node:$addr), (operator (load node:$addr))>;
// Create a store operator that performs the given unary operation
// on the value before storing it.
[SystemZ] Don't use LOAD and STORE REVERSED for volatile accesses Unlike most -- hopefully "all other", but I'm still checking -- memory instructions we support, LOAD REVERSED and STORE REVERSED may access the memory location several times. This means that they are not suitable for volatile loads and stores. This patch is a prerequisite for better atomic load and store support. The same principle applies there: almost all memory instructions we support are inherently atomic ("block concurrent"), but LOAD REVERSED and STORE REVERSED are exceptions. Other instructions continue to allow volatile operands. I will add positive "allows volatile" tests at the same time as the "allows atomic load or store" tests. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@183002 91177308-0d34-0410-b5e6-96231b3b80d8
2013-05-31 13:25:22 +00:00
class storeu<SDPatternOperator operator, SDPatternOperator store = store>
[SystemZ] Add back end This adds the actual lib/Target/SystemZ target files necessary to implement the SystemZ target. Note that at this point, the target cannot yet be built since the configure bits are missing. Those will be provided shortly by a follow-on patch. This version of the patch incorporates feedback from reviews by Chris Lattner and Anton Korobeynikov. Thanks to all reviewers! Patch by Richard Sandiford. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@181203 91177308-0d34-0410-b5e6-96231b3b80d8
2013-05-06 16:15:19 +00:00
: PatFrag<(ops node:$value, node:$addr),
(store (operator node:$value), node:$addr)>;
Reference in New Issue Copy Permalink