maxjhandsome
/
llvm-project
forked from OSchip/llvm-project
1
0
Fork 0
Code Issues Pull Requests Packages Releases Wiki Activity
llvm-project/llvm/test/CodeGen/AArch64/GlobalISel/legalize-load-store.mir

131 lines
3.8 KiB
Plaintext
Raw Normal View History

MIRTests: Remove unnecessary 2>&1 redirection llc mir output goes to stdout nowadays, so the 2>&1 is not necessary anymore for most tests. llvm-svn: 295859
2017-02-23 02:47:41 +08:00
# RUN: llc -O0 -run-pass=legalizer -global-isel %s -o - | FileCheck %s
GlobalISel: legalize 1-bit load/store and mark 8/16 bit variants legal on AArch64. llvm-svn: 279548
2016-08-24 02:20:09 +08:00
--- |
target datalayout = "e-m:o-i64:64-i128:128-n32:64-S128"
[GlobalISel] Get the AArch64 tests to work on Linux Mostly this just means changing the triple from aarch64-apple-ios to the generic aarch64--. Only one test needs more significant changes, but GlobalISel already does the right thing so it's ok to just change the checks. Differential Revision: https://reviews.llvm.org/D25532 llvm-svn: 284223
2016-10-14 18:19:40 +08:00
target triple = "aarch64--"
GlobalISel: legalize 1-bit load/store and mark 8/16 bit variants legal on AArch64. llvm-svn: 279548
2016-08-24 02:20:09 +08:00
define void @test_load(i8* %addr) {
entry:
ret void
}
define void @test_store(i8* %addr) {
entry:
ret void
}
...
---
name: test_load
registers:
- { id: 0, class: _ }
- { id: 1, class: _ }
- { id: 2, class: _ }
- { id: 3, class: _ }
- { id: 4, class: _ }
- { id: 5, class: _ }
GlobalISel: mark pointer stores as legal on AArch64. llvm-svn: 281448
2016-09-14 16:28:54 +08:00
- { id: 6, class: _ }
[AArch64][MachineLegalizer] Mark v2s32 G_LOAD as legal. Actually every 64-bit loads are legal, but right now the API does not offer a simple way to express that. llvm-svn: 283829
2016-10-11 08:21:08 +08:00
- { id: 7, class: _ }
GlobalISel: Implement narrowing for G_LOAD llvm-svn: 292461
2017-01-19 09:05:48 +08:00
- { id: 8, class: _ }
GlobalISel: legalize 1-bit load/store and mark 8/16 bit variants legal on AArch64. llvm-svn: 279548
2016-08-24 02:20:09 +08:00
body: |
bb.0.entry:
Followup on Proposal to move MIR physical register namespace to '$' sigil. Discussed here: http://lists.llvm.org/pipermail/llvm-dev/2018-January/120320.html In preparation for adding support for named vregs we are changing the sigil for physical registers in MIR to '$' from '%'. This will prevent name clashes of named physical register with named vregs. llvm-svn: 323922
2018-02-01 06:04:26 +08:00
liveins: $x0, $x1, $x2, $x3
GlobalISel: legalize 1-bit load/store and mark 8/16 bit variants legal on AArch64. llvm-svn: 279548
2016-08-24 02:20:09 +08:00
; CHECK-LABEL: name: test_load
Followup on Proposal to move MIR physical register namespace to '$' sigil. Discussed here: http://lists.llvm.org/pipermail/llvm-dev/2018-January/120320.html In preparation for adding support for named vregs we are changing the sigil for physical registers in MIR to '$' from '%'. This will prevent name clashes of named physical register with named vregs. llvm-svn: 323922
2018-02-01 06:04:26 +08:00
%0(p0) = COPY $x0
GlobalISel: legalize 1-bit load/store and mark 8/16 bit variants legal on AArch64. llvm-svn: 279548
2016-08-24 02:20:09 +08:00
GlobalISel: move type information to MachineRegisterInfo. We want each register to have a canonical type, which means the best place to store this is in MachineRegisterInfo rather than on every MachineInstr that happens to use or define that register. Most changes following from this are pretty simple (you need an MRI anyway if you're going to be doing any transformations, so just check the type there). But legalization doesn't really want to check redundant operands (when, for example, a G_ADD only ever has one type) so I've made use of MCInstrDesc's operand type field to encode these constraints and limit legalization's work. As an added bonus, more validation is possible, both in MachineVerifier and MachineIRBuilder (coming soon). llvm-svn: 281035
2016-09-09 19:46:34 +08:00
%1(s1) = G_LOAD %0 :: (load 1 from %ir.addr)
[GISel]: Rework legalization algorithm for better elimination of artifacts along with DCE Legalization Artifacts are all those insts that are there to make the type system happy. Currently, the target needs to say all combinations of extends and truncs are legal and there's no way of verifying that post legalization, we only have *truly* legal instructions. This patch changes roughly the legalization algorithm to process all illegal insts at one go, and then process all truncs/extends that were added to satisfy the type constraints separately trying to combine trivial cases until they converge. This has the added benefit that, the target legalizerinfo can only say which truncs and extends are okay and the artifact combiner would combine away other exts and truncs. Updated legalization algorithm to roughly the following pseudo code. WorkList Insts, Artifacts; collect_all_insts_and_artifacts(Insts, Artifacts); do { for (Inst in Insts) legalizeInstrStep(Inst, Insts, Artifacts); for (Artifact in Artifacts) tryCombineArtifact(Artifact, Insts, Artifacts); } while(!Insts.empty()); Also, wrote a simple wrapper equivalent to SetVector, except for erasing, it avoids moving all elements over by one and instead just nulls them out. llvm-svn: 318210
2017-11-15 06:42:19 +08:00
%9:_(s32) = G_ANYEXT %1
Followup on Proposal to move MIR physical register namespace to '$' sigil. Discussed here: http://lists.llvm.org/pipermail/llvm-dev/2018-January/120320.html In preparation for adding support for named vregs we are changing the sigil for physical registers in MIR to '$' from '%'. This will prevent name clashes of named physical register with named vregs. llvm-svn: 323922
2018-02-01 06:04:26 +08:00
$w0 = COPY %9
GlobalISel: legalize 1-bit load/store and mark 8/16 bit variants legal on AArch64. llvm-svn: 279548
2016-08-24 02:20:09 +08:00
MIR: Print the register class or bank in vreg defs This updates the MIRPrinter to include the regclass when printing virtual register defs, which is already valid syntax for the parser. That is, given 64 bit %0 and %1 in a "gpr" regbank, %1(s64) = COPY %0(s64) would now be written as %1:gpr(s64) = COPY %0(s64) While this change alone introduces a bit of redundancy with the registers block, it allows us to update the tests to be more concise and understandable and brings us closer to being able to remove the registers block completely. Note: We generally only print the class in defs, but there is one exception. If there are uses without any defs whatsoever, we'll print the class on all uses. I'm not completely convinced this comes up in meaningful machine IR, but for now the MIRParser and MachineVerifier both accept that kind of stuff, so we don't want to have a situation where we can print something we can't parse. llvm-svn: 316479
2017-10-25 02:04:54 +08:00
; CHECK: %2:_(s8) = G_LOAD %0(p0) :: (load 1 from %ir.addr)
GlobalISel: move type information to MachineRegisterInfo. We want each register to have a canonical type, which means the best place to store this is in MachineRegisterInfo rather than on every MachineInstr that happens to use or define that register. Most changes following from this are pretty simple (you need an MRI anyway if you're going to be doing any transformations, so just check the type there). But legalization doesn't really want to check redundant operands (when, for example, a G_ADD only ever has one type) so I've made use of MCInstrDesc's operand type field to encode these constraints and limit legalization's work. As an added bonus, more validation is possible, both in MachineVerifier and MachineIRBuilder (coming soon). llvm-svn: 281035
2016-09-09 19:46:34 +08:00
%2(s8) = G_LOAD %0 :: (load 1 from %ir.addr)
[GISel]: Rework legalization algorithm for better elimination of artifacts along with DCE Legalization Artifacts are all those insts that are there to make the type system happy. Currently, the target needs to say all combinations of extends and truncs are legal and there's no way of verifying that post legalization, we only have *truly* legal instructions. This patch changes roughly the legalization algorithm to process all illegal insts at one go, and then process all truncs/extends that were added to satisfy the type constraints separately trying to combine trivial cases until they converge. This has the added benefit that, the target legalizerinfo can only say which truncs and extends are okay and the artifact combiner would combine away other exts and truncs. Updated legalization algorithm to roughly the following pseudo code. WorkList Insts, Artifacts; collect_all_insts_and_artifacts(Insts, Artifacts); do { for (Inst in Insts) legalizeInstrStep(Inst, Insts, Artifacts); for (Artifact in Artifacts) tryCombineArtifact(Artifact, Insts, Artifacts); } while(!Insts.empty()); Also, wrote a simple wrapper equivalent to SetVector, except for erasing, it avoids moving all elements over by one and instead just nulls them out. llvm-svn: 318210
2017-11-15 06:42:19 +08:00
%10:_(s32) = G_ANYEXT %2
Followup on Proposal to move MIR physical register namespace to '$' sigil. Discussed here: http://lists.llvm.org/pipermail/llvm-dev/2018-January/120320.html In preparation for adding support for named vregs we are changing the sigil for physical registers in MIR to '$' from '%'. This will prevent name clashes of named physical register with named vregs. llvm-svn: 323922
2018-02-01 06:04:26 +08:00
$w0 = COPY %10
GlobalISel: legalize 1-bit load/store and mark 8/16 bit variants legal on AArch64. llvm-svn: 279548
2016-08-24 02:20:09 +08:00
MIR: Print the register class or bank in vreg defs This updates the MIRPrinter to include the regclass when printing virtual register defs, which is already valid syntax for the parser. That is, given 64 bit %0 and %1 in a "gpr" regbank, %1(s64) = COPY %0(s64) would now be written as %1:gpr(s64) = COPY %0(s64) While this change alone introduces a bit of redundancy with the registers block, it allows us to update the tests to be more concise and understandable and brings us closer to being able to remove the registers block completely. Note: We generally only print the class in defs, but there is one exception. If there are uses without any defs whatsoever, we'll print the class on all uses. I'm not completely convinced this comes up in meaningful machine IR, but for now the MIRParser and MachineVerifier both accept that kind of stuff, so we don't want to have a situation where we can print something we can't parse. llvm-svn: 316479
2017-10-25 02:04:54 +08:00
; CHECK: %3:_(s16) = G_LOAD %0(p0) :: (load 2 from %ir.addr)
GlobalISel: move type information to MachineRegisterInfo. We want each register to have a canonical type, which means the best place to store this is in MachineRegisterInfo rather than on every MachineInstr that happens to use or define that register. Most changes following from this are pretty simple (you need an MRI anyway if you're going to be doing any transformations, so just check the type there). But legalization doesn't really want to check redundant operands (when, for example, a G_ADD only ever has one type) so I've made use of MCInstrDesc's operand type field to encode these constraints and limit legalization's work. As an added bonus, more validation is possible, both in MachineVerifier and MachineIRBuilder (coming soon). llvm-svn: 281035
2016-09-09 19:46:34 +08:00
%3(s16) = G_LOAD %0 :: (load 2 from %ir.addr)
[GISel]: Rework legalization algorithm for better elimination of artifacts along with DCE Legalization Artifacts are all those insts that are there to make the type system happy. Currently, the target needs to say all combinations of extends and truncs are legal and there's no way of verifying that post legalization, we only have *truly* legal instructions. This patch changes roughly the legalization algorithm to process all illegal insts at one go, and then process all truncs/extends that were added to satisfy the type constraints separately trying to combine trivial cases until they converge. This has the added benefit that, the target legalizerinfo can only say which truncs and extends are okay and the artifact combiner would combine away other exts and truncs. Updated legalization algorithm to roughly the following pseudo code. WorkList Insts, Artifacts; collect_all_insts_and_artifacts(Insts, Artifacts); do { for (Inst in Insts) legalizeInstrStep(Inst, Insts, Artifacts); for (Artifact in Artifacts) tryCombineArtifact(Artifact, Insts, Artifacts); } while(!Insts.empty()); Also, wrote a simple wrapper equivalent to SetVector, except for erasing, it avoids moving all elements over by one and instead just nulls them out. llvm-svn: 318210
2017-11-15 06:42:19 +08:00
%11:_(s32) = G_ANYEXT %3
Followup on Proposal to move MIR physical register namespace to '$' sigil. Discussed here: http://lists.llvm.org/pipermail/llvm-dev/2018-January/120320.html In preparation for adding support for named vregs we are changing the sigil for physical registers in MIR to '$' from '%'. This will prevent name clashes of named physical register with named vregs. llvm-svn: 323922
2018-02-01 06:04:26 +08:00
$w0 = COPY %11
GlobalISel: legalize 1-bit load/store and mark 8/16 bit variants legal on AArch64. llvm-svn: 279548
2016-08-24 02:20:09 +08:00
MIR: Print the register class or bank in vreg defs This updates the MIRPrinter to include the regclass when printing virtual register defs, which is already valid syntax for the parser. That is, given 64 bit %0 and %1 in a "gpr" regbank, %1(s64) = COPY %0(s64) would now be written as %1:gpr(s64) = COPY %0(s64) While this change alone introduces a bit of redundancy with the registers block, it allows us to update the tests to be more concise and understandable and brings us closer to being able to remove the registers block completely. Note: We generally only print the class in defs, but there is one exception. If there are uses without any defs whatsoever, we'll print the class on all uses. I'm not completely convinced this comes up in meaningful machine IR, but for now the MIRParser and MachineVerifier both accept that kind of stuff, so we don't want to have a situation where we can print something we can't parse. llvm-svn: 316479
2017-10-25 02:04:54 +08:00
; CHECK: %4:_(s32) = G_LOAD %0(p0) :: (load 4 from %ir.addr)
GlobalISel: move type information to MachineRegisterInfo. We want each register to have a canonical type, which means the best place to store this is in MachineRegisterInfo rather than on every MachineInstr that happens to use or define that register. Most changes following from this are pretty simple (you need an MRI anyway if you're going to be doing any transformations, so just check the type there). But legalization doesn't really want to check redundant operands (when, for example, a G_ADD only ever has one type) so I've made use of MCInstrDesc's operand type field to encode these constraints and limit legalization's work. As an added bonus, more validation is possible, both in MachineVerifier and MachineIRBuilder (coming soon). llvm-svn: 281035
2016-09-09 19:46:34 +08:00
%4(s32) = G_LOAD %0 :: (load 4 from %ir.addr)
Followup on Proposal to move MIR physical register namespace to '$' sigil. Discussed here: http://lists.llvm.org/pipermail/llvm-dev/2018-January/120320.html In preparation for adding support for named vregs we are changing the sigil for physical registers in MIR to '$' from '%'. This will prevent name clashes of named physical register with named vregs. llvm-svn: 323922
2018-02-01 06:04:26 +08:00
$w0 = COPY %4
GlobalISel: legalize 1-bit load/store and mark 8/16 bit variants legal on AArch64. llvm-svn: 279548
2016-08-24 02:20:09 +08:00
MIR: Print the register class or bank in vreg defs This updates the MIRPrinter to include the regclass when printing virtual register defs, which is already valid syntax for the parser. That is, given 64 bit %0 and %1 in a "gpr" regbank, %1(s64) = COPY %0(s64) would now be written as %1:gpr(s64) = COPY %0(s64) While this change alone introduces a bit of redundancy with the registers block, it allows us to update the tests to be more concise and understandable and brings us closer to being able to remove the registers block completely. Note: We generally only print the class in defs, but there is one exception. If there are uses without any defs whatsoever, we'll print the class on all uses. I'm not completely convinced this comes up in meaningful machine IR, but for now the MIRParser and MachineVerifier both accept that kind of stuff, so we don't want to have a situation where we can print something we can't parse. llvm-svn: 316479
2017-10-25 02:04:54 +08:00
; CHECK: %5:_(s64) = G_LOAD %0(p0) :: (load 8 from %ir.addr)
GlobalISel: move type information to MachineRegisterInfo. We want each register to have a canonical type, which means the best place to store this is in MachineRegisterInfo rather than on every MachineInstr that happens to use or define that register. Most changes following from this are pretty simple (you need an MRI anyway if you're going to be doing any transformations, so just check the type there). But legalization doesn't really want to check redundant operands (when, for example, a G_ADD only ever has one type) so I've made use of MCInstrDesc's operand type field to encode these constraints and limit legalization's work. As an added bonus, more validation is possible, both in MachineVerifier and MachineIRBuilder (coming soon). llvm-svn: 281035
2016-09-09 19:46:34 +08:00
%5(s64) = G_LOAD %0 :: (load 8 from %ir.addr)
Followup on Proposal to move MIR physical register namespace to '$' sigil. Discussed here: http://lists.llvm.org/pipermail/llvm-dev/2018-January/120320.html In preparation for adding support for named vregs we are changing the sigil for physical registers in MIR to '$' from '%'. This will prevent name clashes of named physical register with named vregs. llvm-svn: 323922
2018-02-01 06:04:26 +08:00
$x0 = COPY %5
GlobalISel: mark pointer stores as legal on AArch64. llvm-svn: 281448
2016-09-14 16:28:54 +08:00
%6(p0) = G_LOAD %0(p0) :: (load 8 from %ir.addr)
[GISel]: Rework legalization algorithm for better elimination of artifacts along with DCE Legalization Artifacts are all those insts that are there to make the type system happy. Currently, the target needs to say all combinations of extends and truncs are legal and there's no way of verifying that post legalization, we only have *truly* legal instructions. This patch changes roughly the legalization algorithm to process all illegal insts at one go, and then process all truncs/extends that were added to satisfy the type constraints separately trying to combine trivial cases until they converge. This has the added benefit that, the target legalizerinfo can only say which truncs and extends are okay and the artifact combiner would combine away other exts and truncs. Updated legalization algorithm to roughly the following pseudo code. WorkList Insts, Artifacts; collect_all_insts_and_artifacts(Insts, Artifacts); do { for (Inst in Insts) legalizeInstrStep(Inst, Insts, Artifacts); for (Artifact in Artifacts) tryCombineArtifact(Artifact, Insts, Artifacts); } while(!Insts.empty()); Also, wrote a simple wrapper equivalent to SetVector, except for erasing, it avoids moving all elements over by one and instead just nulls them out. llvm-svn: 318210
2017-11-15 06:42:19 +08:00
%12:_(s64) = G_PTRTOINT %6
Followup on Proposal to move MIR physical register namespace to '$' sigil. Discussed here: http://lists.llvm.org/pipermail/llvm-dev/2018-January/120320.html In preparation for adding support for named vregs we are changing the sigil for physical registers in MIR to '$' from '%'. This will prevent name clashes of named physical register with named vregs. llvm-svn: 323922
2018-02-01 06:04:26 +08:00
$x0 = COPY %12
[AArch64][MachineLegalizer] Mark v2s32 G_LOAD as legal. Actually every 64-bit loads are legal, but right now the API does not offer a simple way to express that. llvm-svn: 283829
2016-10-11 08:21:08 +08:00
MIR: Print the register class or bank in vreg defs This updates the MIRPrinter to include the regclass when printing virtual register defs, which is already valid syntax for the parser. That is, given 64 bit %0 and %1 in a "gpr" regbank, %1(s64) = COPY %0(s64) would now be written as %1:gpr(s64) = COPY %0(s64) While this change alone introduces a bit of redundancy with the registers block, it allows us to update the tests to be more concise and understandable and brings us closer to being able to remove the registers block completely. Note: We generally only print the class in defs, but there is one exception. If there are uses without any defs whatsoever, we'll print the class on all uses. I'm not completely convinced this comes up in meaningful machine IR, but for now the MIRParser and MachineVerifier both accept that kind of stuff, so we don't want to have a situation where we can print something we can't parse. llvm-svn: 316479
2017-10-25 02:04:54 +08:00
; CHECK: %7:_(<2 x s32>) = G_LOAD %0(p0) :: (load 8 from %ir.addr)
[AArch64][MachineLegalizer] Mark v2s32 G_LOAD as legal. Actually every 64-bit loads are legal, but right now the API does not offer a simple way to express that. llvm-svn: 283829
2016-10-11 08:21:08 +08:00
%7(<2 x s32>) = G_LOAD %0(p0) :: (load 8 from %ir.addr)
[GISel]: Rework legalization algorithm for better elimination of artifacts along with DCE Legalization Artifacts are all those insts that are there to make the type system happy. Currently, the target needs to say all combinations of extends and truncs are legal and there's no way of verifying that post legalization, we only have *truly* legal instructions. This patch changes roughly the legalization algorithm to process all illegal insts at one go, and then process all truncs/extends that were added to satisfy the type constraints separately trying to combine trivial cases until they converge. This has the added benefit that, the target legalizerinfo can only say which truncs and extends are okay and the artifact combiner would combine away other exts and truncs. Updated legalization algorithm to roughly the following pseudo code. WorkList Insts, Artifacts; collect_all_insts_and_artifacts(Insts, Artifacts); do { for (Inst in Insts) legalizeInstrStep(Inst, Insts, Artifacts); for (Artifact in Artifacts) tryCombineArtifact(Artifact, Insts, Artifacts); } while(!Insts.empty()); Also, wrote a simple wrapper equivalent to SetVector, except for erasing, it avoids moving all elements over by one and instead just nulls them out. llvm-svn: 318210
2017-11-15 06:42:19 +08:00
%13:_(s64) = G_BITCAST %7
Followup on Proposal to move MIR physical register namespace to '$' sigil. Discussed here: http://lists.llvm.org/pipermail/llvm-dev/2018-January/120320.html In preparation for adding support for named vregs we are changing the sigil for physical registers in MIR to '$' from '%'. This will prevent name clashes of named physical register with named vregs. llvm-svn: 323922
2018-02-01 06:04:26 +08:00
$x0 = COPY %13
GlobalISel: Implement narrowing for G_LOAD llvm-svn: 292461
2017-01-19 09:05:48 +08:00
MIR: Print the register class or bank in vreg defs This updates the MIRPrinter to include the regclass when printing virtual register defs, which is already valid syntax for the parser. That is, given 64 bit %0 and %1 in a "gpr" regbank, %1(s64) = COPY %0(s64) would now be written as %1:gpr(s64) = COPY %0(s64) While this change alone introduces a bit of redundancy with the registers block, it allows us to update the tests to be more concise and understandable and brings us closer to being able to remove the registers block completely. Note: We generally only print the class in defs, but there is one exception. If there are uses without any defs whatsoever, we'll print the class on all uses. I'm not completely convinced this comes up in meaningful machine IR, but for now the MIRParser and MachineVerifier both accept that kind of stuff, so we don't want to have a situation where we can print something we can't parse. llvm-svn: 316479
2017-10-25 02:04:54 +08:00
; CHECK: [[LOAD0:%[0-9]+]]:_(s64) = G_LOAD %0(p0) :: (load 16 from %ir.addr)
; CHECK: [[OFFSET1:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
; CHECK: [[GEP1:%[0-9]+]]:_(p0) = G_GEP %0, [[OFFSET1]](s64)
; CHECK: [[LOAD1:%[0-9]+]]:_(s64) = G_LOAD [[GEP1]](p0) :: (load 16 from %ir.addr)
; CHECK: %8:_(s128) = G_MERGE_VALUES [[LOAD0]](s64), [[LOAD1]](s64)
GlobalISel: Implement narrowing for G_LOAD llvm-svn: 292461
2017-01-19 09:05:48 +08:00
%8(s128) = G_LOAD %0(p0) :: (load 16 from %ir.addr)
[GISel]: Rework legalization algorithm for better elimination of artifacts along with DCE Legalization Artifacts are all those insts that are there to make the type system happy. Currently, the target needs to say all combinations of extends and truncs are legal and there's no way of verifying that post legalization, we only have *truly* legal instructions. This patch changes roughly the legalization algorithm to process all illegal insts at one go, and then process all truncs/extends that were added to satisfy the type constraints separately trying to combine trivial cases until they converge. This has the added benefit that, the target legalizerinfo can only say which truncs and extends are okay and the artifact combiner would combine away other exts and truncs. Updated legalization algorithm to roughly the following pseudo code. WorkList Insts, Artifacts; collect_all_insts_and_artifacts(Insts, Artifacts); do { for (Inst in Insts) legalizeInstrStep(Inst, Insts, Artifacts); for (Artifact in Artifacts) tryCombineArtifact(Artifact, Insts, Artifacts); } while(!Insts.empty()); Also, wrote a simple wrapper equivalent to SetVector, except for erasing, it avoids moving all elements over by one and instead just nulls them out. llvm-svn: 318210
2017-11-15 06:42:19 +08:00
%14:_(s64) = G_TRUNC %8
Followup on Proposal to move MIR physical register namespace to '$' sigil. Discussed here: http://lists.llvm.org/pipermail/llvm-dev/2018-January/120320.html In preparation for adding support for named vregs we are changing the sigil for physical registers in MIR to '$' from '%'. This will prevent name clashes of named physical register with named vregs. llvm-svn: 323922
2018-02-01 06:04:26 +08:00
$x0 = COPY %14
GlobalISel: legalize 1-bit load/store and mark 8/16 bit variants legal on AArch64. llvm-svn: 279548
2016-08-24 02:20:09 +08:00
...
---
name: test_store
registers:
- { id: 0, class: _ }
- { id: 1, class: _ }
- { id: 2, class: _ }
- { id: 3, class: _ }
- { id: 4, class: _ }
GlobalISel: move type information to MachineRegisterInfo. We want each register to have a canonical type, which means the best place to store this is in MachineRegisterInfo rather than on every MachineInstr that happens to use or define that register. Most changes following from this are pretty simple (you need an MRI anyway if you're going to be doing any transformations, so just check the type there). But legalization doesn't really want to check redundant operands (when, for example, a G_ADD only ever has one type) so I've made use of MCInstrDesc's operand type field to encode these constraints and limit legalization's work. As an added bonus, more validation is possible, both in MachineVerifier and MachineIRBuilder (coming soon). llvm-svn: 281035
2016-09-09 19:46:34 +08:00
- { id: 5, class: _ }
GlobalISel: Implement narrowing for G_STORE Legalize stores of types that are too wide by breaking them up into sequences of smaller stores. llvm-svn: 292412
2017-01-19 01:29:54 +08:00
- { id: 6, class: _ }
- { id: 7, class: _ }
GlobalISel: legalize 1-bit load/store and mark 8/16 bit variants legal on AArch64. llvm-svn: 279548
2016-08-24 02:20:09 +08:00
body: |
bb.0.entry:
Followup on Proposal to move MIR physical register namespace to '$' sigil. Discussed here: http://lists.llvm.org/pipermail/llvm-dev/2018-January/120320.html In preparation for adding support for named vregs we are changing the sigil for physical registers in MIR to '$' from '%'. This will prevent name clashes of named physical register with named vregs. llvm-svn: 323922
2018-02-01 06:04:26 +08:00
liveins: $x0, $x1, $x2, $x3
GlobalISel: legalize 1-bit load/store and mark 8/16 bit variants legal on AArch64. llvm-svn: 279548
2016-08-24 02:20:09 +08:00
; CHECK-LABEL: name: test_store
Followup on Proposal to move MIR physical register namespace to '$' sigil. Discussed here: http://lists.llvm.org/pipermail/llvm-dev/2018-January/120320.html In preparation for adding support for named vregs we are changing the sigil for physical registers in MIR to '$' from '%'. This will prevent name clashes of named physical register with named vregs. llvm-svn: 323922
2018-02-01 06:04:26 +08:00
%0(p0) = COPY $x0
%1(s32) = COPY $w1
GlobalISel: legalize 1-bit load/store and mark 8/16 bit variants legal on AArch64. llvm-svn: 279548
2016-08-24 02:20:09 +08:00
MIR: Print the register class or bank in vreg defs This updates the MIRPrinter to include the regclass when printing virtual register defs, which is already valid syntax for the parser. That is, given 64 bit %0 and %1 in a "gpr" regbank, %1(s64) = COPY %0(s64) would now be written as %1:gpr(s64) = COPY %0(s64) While this change alone introduces a bit of redundancy with the registers block, it allows us to update the tests to be more concise and understandable and brings us closer to being able to remove the registers block completely. Note: We generally only print the class in defs, but there is one exception. If there are uses without any defs whatsoever, we'll print the class on all uses. I'm not completely convinced this comes up in meaningful machine IR, but for now the MIRParser and MachineVerifier both accept that kind of stuff, so we don't want to have a situation where we can print something we can't parse. llvm-svn: 316479
2017-10-25 02:04:54 +08:00
; CHECK: [[C1:%.*]]:_(s32) = G_CONSTANT i32 1
; CHECK: [[B:%.*]]:_(s32) = COPY %1(s32)
; CHECK: [[COPY_C1:%.*]]:_(s32) = COPY [[C1]]
; CHECK: [[AND:%.*]]:_(s32) = G_AND [[B]], [[COPY_C1]]
; CHECK: [[BIT8:%.*]]:_(s8) = G_TRUNC [[AND]]
[GISel]: Add a clean up combiner during legalization. Added a combiner which can clean up truncs/extends that are created in order to make the types work during legalization. Also moved the combineMerges to the LegalizeCombiner. https://reviews.llvm.org/D36880 llvm-svn: 312158
2017-08-31 03:32:59 +08:00
GlobalISel: disambiguate types when printing MIR Some generic instructions have multiple types. While in theory these always be discovered by inspecting the single definition of each generic vreg, in practice those definitions won't always be local and traipsing through a big function to find them will not be fun. So this changes MIRPrinter to print out the type of uses as well as defs, if they're known to be different or not known to be the same. On the parsing side, we're a little more flexible: provided each register is given a type in at least one place it's mentioned (and all types are consistent) we accept the MIR. This doesn't introduce ambiguity but makes writing tests manually a bit less painful. llvm-svn: 281204
2016-09-12 19:20:10 +08:00
; CHECK: G_STORE [[BIT8]](s8), %0(p0) :: (store 1 into %ir.addr)
GlobalISel: move type information to MachineRegisterInfo. We want each register to have a canonical type, which means the best place to store this is in MachineRegisterInfo rather than on every MachineInstr that happens to use or define that register. Most changes following from this are pretty simple (you need an MRI anyway if you're going to be doing any transformations, so just check the type there). But legalization doesn't really want to check redundant operands (when, for example, a G_ADD only ever has one type) so I've made use of MCInstrDesc's operand type field to encode these constraints and limit legalization's work. As an added bonus, more validation is possible, both in MachineVerifier and MachineIRBuilder (coming soon). llvm-svn: 281035
2016-09-09 19:46:34 +08:00
%2(s1) = G_TRUNC %1
G_STORE %2, %0 :: (store 1 into %ir.addr)
GlobalISel: legalize 1-bit load/store and mark 8/16 bit variants legal on AArch64. llvm-svn: 279548
2016-08-24 02:20:09 +08:00
GlobalISel: disambiguate types when printing MIR Some generic instructions have multiple types. While in theory these always be discovered by inspecting the single definition of each generic vreg, in practice those definitions won't always be local and traipsing through a big function to find them will not be fun. So this changes MIRPrinter to print out the type of uses as well as defs, if they're known to be different or not known to be the same. On the parsing side, we're a little more flexible: provided each register is given a type in at least one place it's mentioned (and all types are consistent) we accept the MIR. This doesn't introduce ambiguity but makes writing tests manually a bit less painful. llvm-svn: 281204
2016-09-12 19:20:10 +08:00
; CHECK: G_STORE %3(s8), %0(p0) :: (store 1 into %ir.addr)
GlobalISel: move type information to MachineRegisterInfo. We want each register to have a canonical type, which means the best place to store this is in MachineRegisterInfo rather than on every MachineInstr that happens to use or define that register. Most changes following from this are pretty simple (you need an MRI anyway if you're going to be doing any transformations, so just check the type there). But legalization doesn't really want to check redundant operands (when, for example, a G_ADD only ever has one type) so I've made use of MCInstrDesc's operand type field to encode these constraints and limit legalization's work. As an added bonus, more validation is possible, both in MachineVerifier and MachineIRBuilder (coming soon). llvm-svn: 281035
2016-09-09 19:46:34 +08:00
%3(s8) = G_TRUNC %1
G_STORE %3, %0 :: (store 1 into %ir.addr)
GlobalISel: legalize 1-bit load/store and mark 8/16 bit variants legal on AArch64. llvm-svn: 279548
2016-08-24 02:20:09 +08:00
GlobalISel: disambiguate types when printing MIR Some generic instructions have multiple types. While in theory these always be discovered by inspecting the single definition of each generic vreg, in practice those definitions won't always be local and traipsing through a big function to find them will not be fun. So this changes MIRPrinter to print out the type of uses as well as defs, if they're known to be different or not known to be the same. On the parsing side, we're a little more flexible: provided each register is given a type in at least one place it's mentioned (and all types are consistent) we accept the MIR. This doesn't introduce ambiguity but makes writing tests manually a bit less painful. llvm-svn: 281204
2016-09-12 19:20:10 +08:00
; CHECK: G_STORE %4(s16), %0(p0) :: (store 2 into %ir.addr)
GlobalISel: move type information to MachineRegisterInfo. We want each register to have a canonical type, which means the best place to store this is in MachineRegisterInfo rather than on every MachineInstr that happens to use or define that register. Most changes following from this are pretty simple (you need an MRI anyway if you're going to be doing any transformations, so just check the type there). But legalization doesn't really want to check redundant operands (when, for example, a G_ADD only ever has one type) so I've made use of MCInstrDesc's operand type field to encode these constraints and limit legalization's work. As an added bonus, more validation is possible, both in MachineVerifier and MachineIRBuilder (coming soon). llvm-svn: 281035
2016-09-09 19:46:34 +08:00
%4(s16) = G_TRUNC %1
G_STORE %4, %0 :: (store 2 into %ir.addr)
GlobalISel: legalize 1-bit load/store and mark 8/16 bit variants legal on AArch64. llvm-svn: 279548
2016-08-24 02:20:09 +08:00
GlobalISel: disambiguate types when printing MIR Some generic instructions have multiple types. While in theory these always be discovered by inspecting the single definition of each generic vreg, in practice those definitions won't always be local and traipsing through a big function to find them will not be fun. So this changes MIRPrinter to print out the type of uses as well as defs, if they're known to be different or not known to be the same. On the parsing side, we're a little more flexible: provided each register is given a type in at least one place it's mentioned (and all types are consistent) we accept the MIR. This doesn't introduce ambiguity but makes writing tests manually a bit less painful. llvm-svn: 281204
2016-09-12 19:20:10 +08:00
; CHECK: G_STORE %1(s32), %0(p0) :: (store 4 into %ir.addr)
GlobalISel: move type information to MachineRegisterInfo. We want each register to have a canonical type, which means the best place to store this is in MachineRegisterInfo rather than on every MachineInstr that happens to use or define that register. Most changes following from this are pretty simple (you need an MRI anyway if you're going to be doing any transformations, so just check the type there). But legalization doesn't really want to check redundant operands (when, for example, a G_ADD only ever has one type) so I've made use of MCInstrDesc's operand type field to encode these constraints and limit legalization's work. As an added bonus, more validation is possible, both in MachineVerifier and MachineIRBuilder (coming soon). llvm-svn: 281035
2016-09-09 19:46:34 +08:00
G_STORE %1, %0 :: (store 4 into %ir.addr)
GlobalISel: legalize 1-bit load/store and mark 8/16 bit variants legal on AArch64. llvm-svn: 279548
2016-08-24 02:20:09 +08:00
GlobalISel: disambiguate types when printing MIR Some generic instructions have multiple types. While in theory these always be discovered by inspecting the single definition of each generic vreg, in practice those definitions won't always be local and traipsing through a big function to find them will not be fun. So this changes MIRPrinter to print out the type of uses as well as defs, if they're known to be different or not known to be the same. On the parsing side, we're a little more flexible: provided each register is given a type in at least one place it's mentioned (and all types are consistent) we accept the MIR. This doesn't introduce ambiguity but makes writing tests manually a bit less painful. llvm-svn: 281204
2016-09-12 19:20:10 +08:00
; CHECK: G_STORE %5(s64), %0(p0) :: (store 8 into %ir.addr)
%5(s64) = G_PTRTOINT %0(p0)
GlobalISel: move type information to MachineRegisterInfo. We want each register to have a canonical type, which means the best place to store this is in MachineRegisterInfo rather than on every MachineInstr that happens to use or define that register. Most changes following from this are pretty simple (you need an MRI anyway if you're going to be doing any transformations, so just check the type there). But legalization doesn't really want to check redundant operands (when, for example, a G_ADD only ever has one type) so I've made use of MCInstrDesc's operand type field to encode these constraints and limit legalization's work. As an added bonus, more validation is possible, both in MachineVerifier and MachineIRBuilder (coming soon). llvm-svn: 281035
2016-09-09 19:46:34 +08:00
G_STORE %5, %0 :: (store 8 into %ir.addr)
GlobalISel: mark pointer stores as legal on AArch64. llvm-svn: 281448
2016-09-14 16:28:54 +08:00
; CHECK: G_STORE %0(p0), %0(p0) :: (store 8 into %ir.addr)
G_STORE %0(p0), %0(p0) :: (store 8 into %ir.addr)
GlobalISel: Implement narrowing for G_STORE Legalize stores of types that are too wide by breaking them up into sequences of smaller stores. llvm-svn: 292412
2017-01-19 01:29:54 +08:00
[globalisel][legalizer] G_LOAD/G_STORE NarrowScalar should not emit G_GEP x, 0. Summary: When legalizing G_LOAD/G_STORE using NarrowScalar, we should avoid emitting %0 = G_CONSTANT ty 0 %1 = G_GEP %x, %0 since it's cheaper to not emit the redundant instructions than it is to fold them away later. Reviewers: qcolombet, t.p.northover, ab, rovka, aditya_nandakumar, kristof.beyls Reviewed By: qcolombet Subscribers: javed.absar, llvm-commits, igorb Differential Revision: https://reviews.llvm.org/D32746 llvm-svn: 305340
2017-06-14 07:42:32 +08:00
; CHECK: G_STORE %5(s64), %0(p0) :: (store 16 into %ir.addr)
MIR: Print the register class or bank in vreg defs This updates the MIRPrinter to include the regclass when printing virtual register defs, which is already valid syntax for the parser. That is, given 64 bit %0 and %1 in a "gpr" regbank, %1(s64) = COPY %0(s64) would now be written as %1:gpr(s64) = COPY %0(s64) While this change alone introduces a bit of redundancy with the registers block, it allows us to update the tests to be more concise and understandable and brings us closer to being able to remove the registers block completely. Note: We generally only print the class in defs, but there is one exception. If there are uses without any defs whatsoever, we'll print the class on all uses. I'm not completely convinced this comes up in meaningful machine IR, but for now the MIRParser and MachineVerifier both accept that kind of stuff, so we don't want to have a situation where we can print something we can't parse. llvm-svn: 316479
2017-10-25 02:04:54 +08:00
; CHECK: [[OFFSET1:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
; CHECK: [[GEP1:%[0-9]+]]:_(p0) = G_GEP %0, [[OFFSET1]](s64)
GlobalISel: Implement narrowing for G_STORE Legalize stores of types that are too wide by breaking them up into sequences of smaller stores. llvm-svn: 292412
2017-01-19 01:29:54 +08:00
; CHECK: G_STORE %6(s64), [[GEP1]](p0) :: (store 16 into %ir.addr)
%6(s64) = G_PTRTOINT %0(p0)
GlobalISel: add merge/unmerge nodes for legalization. These are simplified variants of the current G_SEQUENCE and G_EXTRACT, which assume the individual parts will be contiguous, homogeneous, and occupy the entirity of the larger register. This makes reasoning about them much easer since you only have to look at the first register being merged and the result to know what the instruction is doing. I intend to gradually replace all uses of the more complicated sequence/extract with these (or single-element insert/extracts), and then remove the older variants. For now we start with legalization. llvm-svn: 296921
2017-03-04 06:46:09 +08:00
%7(s128) = G_MERGE_VALUES %5, %6
GlobalISel: Implement narrowing for G_STORE Legalize stores of types that are too wide by breaking them up into sequences of smaller stores. llvm-svn: 292412
2017-01-19 01:29:54 +08:00
G_STORE %7, %0 :: (store 16 into %ir.addr)
GlobalISel: legalize 1-bit load/store and mark 8/16 bit variants legal on AArch64. llvm-svn: 279548
2016-08-24 02:20:09 +08:00
...