This is an archive of the discontinued LLVM Phabricator instance.

Differential D157416

[RISCV][GlobalISel] Legalize multiplication
ClosedPublic

Authored by nitinjohnraj on Aug 8 2023, 10:00 AM.

Details

Reviewers
craig.topper
arsenm
Commits
rG638865c8f939: [RISCV][GlobalISel] Legalize multiplication
Summary

Legalize multiplication with the +m, +zmmul extensions and without extensions. With extensions, we test for (s7, s8, s16, s32, s48, s64, s96) on rv32 and (s8, s15, s32, s64, s72, s128, s192) on rv64. Without extensions, test (s7, s8, s16, s32) on rv32 and (s8, s15, s16, s32, s64) on rv64. Does not yet work for the type which is 2 times XLen without extensions.

Diff Detail

Event Timeline

nitinjohnraj created this revision.Aug 8 2023, 10:00 AM
Herald added a project: Restricted Project. · View Herald TranscriptAug 8 2023, 10:00 AM
Herald added subscribers: jobnoorman, luke, VincentWu and 28 others. · View Herald Transcript
nitinjohnraj updated this revision to Diff 548292.Aug 8 2023, 11:16 AM

Fixed legalization for 2 x XLen without extensions

nitinjohnraj published this revision for review.Aug 8 2023, 1:57 PM
Herald added a project: Restricted Project. · View Herald TranscriptAug 8 2023, 1:57 PM
Herald added subscribers: llvm-commits, wangpc, eopXD and 2 others. · View Herald Transcript
Harbormaster completed remote builds in B251158: Diff 548292.Aug 8 2023, 3:44 PM
craig.topper added inline comments.Aug 8 2023, 11:23 PM
llvm/lib/Target/RISCV/GISel/RISCVLegalizerInfo.cpp
95

Why did we stray from the usual formatting here?

llvm/test/CodeGen/RISCV/GlobalISel/legalizer/rv32/legalize-mul-ext.mir
6

Can +m and +zmmul share the same check prefix?

11

CHECK-I is not a valid prefix for this test.

llvm/test/CodeGen/RISCV/GlobalISel/legalizer/rv64/legalize-mul-ext.mir
4

Check prefixes in this test have an I in them that the rv32 tests don't have. We should be consistent

nitinjohnraj marked 3 inline comments as done.Aug 10 2023, 9:09 AM
nitinjohnraj added inline comments.
llvm/lib/Target/RISCV/GISel/RISCVLegalizerInfo.cpp
95

clang-format does this. You can see similar formatting on line 71. Should I manually format it with // clang-format off tags?

nitinjohnraj updated this revision to Diff 549071.Aug 10 2023, 9:10 AM

Fixed tests to use same prefix

craig.topper added inline comments.Aug 10 2023, 9:37 AM
llvm/lib/Target/RISCV/GISel/RISCVLegalizerInfo.cpp
95

I support turning off clang-format here

craig.topper added inline comments.Aug 10 2023, 9:38 AM
llvm/lib/Target/RISCV/GISel/RISCVLegalizerInfo.cpp
95

Why do we need .lower() on this? We have instructions for G_SMULH and G_UMULH.

arsenm added inline comments.Aug 10 2023, 9:40 AM
llvm/lib/Target/RISCV/GISel/RISCVLegalizerInfo.cpp
95

I would assume you want some clamping before going direct to lower (although I wouldn't be surprised if there wasn't complete support for mulh lowering)

craig.topper added a child revision: D157422: [RISCV][GlobalISel] Legalize division and remainder.Aug 10 2023, 9:43 AM
Harbormaster completed remote builds in B251716: Diff 549071.Aug 10 2023, 2:39 PM
craig.topper added inline comments.Aug 10 2023, 10:58 PM
llvm/lib/Target/RISCV/GISel/RISCVLegalizerInfo.cpp
95

Ignore my comment. I think I missed the legalFor

tschuett added a subscriber: tschuett.Aug 11 2023, 2:03 AM
tschuett added inline comments.
llvm/lib/Target/RISCV/GISel/RISCVLegalizerInfo.cpp
90

This builder looks the same with and without the extension? Could you hoist it above the if?

craig.topper added inline comments.Aug 11 2023, 9:22 AM
llvm/lib/Target/RISCV/GISel/RISCVLegalizerInfo.cpp
90

Is the MULO part of this tested?

nitinjohnraj updated this revision to Diff 550367.Aug 15 2023, 9:40 AM

Fixed formatting and deleted legalization for G_(U/S)MULO. G_(U/S)MULO needs to be legalized in another patch with its own tests.

Harbormaster completed remote builds in B252666: Diff 550367.Aug 15 2023, 9:41 AM
craig.topper added inline comments.Aug 15 2023, 1:18 PM
llvm/lib/Target/RISCV/GISel/RISCVLegalizerInfo.cpp
101

I assume MULO here isn't needed/tested either?

nitinjohnraj updated this revision to Diff 550556.Aug 15 2023, 5:13 PM

Removed MULO rules missed in previous update

Harbormaster completed remote builds in B252809: Diff 550556.Aug 15 2023, 5:13 PM
craig.topper accepted this revision.Aug 15 2023, 6:25 PM

LGTM

This revision is now accepted and ready to land.Aug 15 2023, 6:25 PM
This revision was landed with ongoing or failed builds.Aug 17 2023, 1:00 PM
Closed by commit rG638865c8f939: [RISCV][GlobalISel] Legalize multiplication (authored by nitinjohnraj). · Explain Why
This revision was automatically updated to reflect the committed changes.
nitinjohnraj added a commit: rG638865c8f939: [RISCV][GlobalISel] Legalize multiplication.
Herald added a subscriber: sunshaoce. · View Herald TranscriptAug 17 2023, 1:00 PM

Revision Contents

PathSize
llvm/
lib/
Target/
RISCV/
GISel/
21 lines
test/
CodeGen/
RISCV/
GlobalISel/
legalizer/
rv32/
199 lines
124 lines
rv64/
219 lines
147 lines

Diff 551237

llvm/lib/Target/RISCV/GISel/RISCVLegalizerInfo.cpp

Show All 16 Lines
#include "llvm/IR/DerivedTypes.h" #include "llvm/IR/DerivedTypes.h"
#include "llvm/IR/Type.h" #include "llvm/IR/Type.h"
using namespace llvm; using namespace llvm;
RISCVLegalizerInfo::RISCVLegalizerInfo(const RISCVSubtarget &ST) { RISCVLegalizerInfo::RISCVLegalizerInfo(const RISCVSubtarget &ST) {
const unsigned XLen = ST.getXLen(); const unsigned XLen = ST.getXLen();
const LLT XLenLLT = LLT::scalar(XLen); const LLT XLenLLT = LLT::scalar(XLen);
const LLT DoubleXLenLLT = LLT::scalar(2 * XLen);
const LLT p0 = LLT::pointer(0, XLen); const LLT p0 = LLT::pointer(0, XLen);
using namespace TargetOpcode; using namespace TargetOpcode;
getActionDefinitionsBuilder({G_ADD, G_SUB, G_AND, G_OR, G_XOR}) getActionDefinitionsBuilder({G_ADD, G_SUB, G_AND, G_OR, G_XOR})
.legalFor({XLenLLT}) .legalFor({XLenLLT})
.widenScalarToNextPow2(0) .widenScalarToNextPow2(0)
.clampScalar(0, XLenLLT, XLenLLT); .clampScalar(0, XLenLLT, XLenLLT);
▲ Show 20 Lines • Show All 48 Lines • ▼ Show 20 Lines getActionDefinitionsBuilder(G_SELECT)
.clampScalar(1, XLenLLT, XLenLLT); .clampScalar(1, XLenLLT, XLenLLT);
getActionDefinitionsBuilder(G_BRCOND) getActionDefinitionsBuilder(G_BRCOND)
.legalFor({XLenLLT}) .legalFor({XLenLLT})
.minScalar(0, XLenLLT); .minScalar(0, XLenLLT);
getActionDefinitionsBuilder(G_PHI) getActionDefinitionsBuilder(G_PHI)
.legalFor({p0, XLenLLT}) .legalFor({p0, XLenLLT})
.widenScalarToNextPow2(0) .widenScalarToNextPow2(0)
tschuettUnsubmitted
Not Done
ReplyInline Actions

This builder looks the same with and without the extension? Could you hoist it above the if?

tschuett: This builder looks the same with and without the extension? Could you hoist it above the if?
craig.topperUnsubmitted
Not Done
ReplyInline Actions

Is the MULO part of this tested?

craig.topper: Is the MULO part of this tested?
.clampScalar(0, XLenLLT, XLenLLT); .clampScalar(0, XLenLLT, XLenLLT);
getActionDefinitionsBuilder(G_GLOBAL_VALUE) getActionDefinitionsBuilder(G_GLOBAL_VALUE)
.legalFor({p0}); .legalFor({p0});
craig.topperUnsubmitted
Not Done
ReplyInline Actions

Why did we stray from the usual formatting here?

craig.topper: Why did we stray from the usual formatting here?
nitinjohnrajAuthorUnsubmitted
Done
ReplyInline Actions

clang-format does this. You can see similar formatting on line 71. Should I manually format it with // clang-format off tags?

nitinjohnraj: clang-format does this. You can see similar formatting on line 71. Should I manually format it…
craig.topperUnsubmitted
Not Done
ReplyInline Actions

I support turning off clang-format here

craig.topper: I support turning off clang-format here
craig.topperUnsubmitted
Not Done
ReplyInline Actions

Why do we need .lower() on this? We have instructions for G_SMULH and G_UMULH.

craig.topper: Why do we need .lower() on this? We have instructions for G_SMULH and G_UMULH.
arsenmUnsubmitted
Not Done
ReplyInline Actions

I would assume you want some clamping before going direct to lower (although I wouldn't be surprised if there wasn't complete support for mulh lowering)

arsenm: I would assume you want some clamping before going direct to lower (although I wouldn't be…
craig.topperUnsubmitted
Not Done
ReplyInline Actions

Ignore my comment. I think I missed the legalFor

craig.topper: Ignore my comment. I think I missed the `legalFor`
if (ST.hasStdExtM() || ST.hasStdExtZmmul()) {
getActionDefinitionsBuilder(G_MUL)
.legalFor({XLenLLT})
.widenScalarToNextPow2(0)
.clampScalar(0, XLenLLT, XLenLLT);
craig.topperUnsubmitted
Not Done
ReplyInline Actions

I assume MULO here isn't needed/tested either?

craig.topper: I assume MULO here isn't needed/tested either?
// clang-format off
getActionDefinitionsBuilder({G_SMULH, G_UMULH})
.legalFor({XLenLLT})
.lower();
// clang-format on
} else {
getActionDefinitionsBuilder(G_MUL)
.libcallFor({XLenLLT, DoubleXLenLLT})
.widenScalarToNextPow2(0)
.clampScalar(0, XLenLLT, DoubleXLenLLT);
getActionDefinitionsBuilder({G_SMULH, G_UMULH}).lowerFor({XLenLLT});
}
getLegacyLegalizerInfo().computeTables(); getLegacyLegalizerInfo().computeTables();
} }

llvm/test/CodeGen/RISCV/GlobalISel/legalizer/rv32/legalize-mul-ext.mir

  • This file was added.
# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
# RUN: llc -mattr=+m -mtriple=riscv32 -run-pass=legalizer %s -o - \
# RUN: | FileCheck %s
# RUN: llc -mattr=+zmmul -mtriple=riscv32 -run-pass=legalizer %s -o - \
# RUN: | FileCheck %s
---
craig.topperUnsubmitted
Done
ReplyInline Actions

Can +m and +zmmul share the same check prefix?

craig.topper: Can +m and +zmmul share the same check prefix?
name: mul_i7
body: |
bb.0.entry:
; CHECK-LABEL: name: mul_i7
; CHECK: [[COPY:%[0-9]+]]:_(s32) = COPY $x10
craig.topperUnsubmitted
Done
ReplyInline Actions

CHECK-I is not a valid prefix for this test.

craig.topper: CHECK-I is not a valid prefix for this test.
; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $x11
; CHECK-NEXT: [[MUL:%[0-9]+]]:_(s32) = G_MUL [[COPY]], [[COPY1]]
; CHECK-NEXT: $x10 = COPY [[MUL]](s32)
; CHECK-NEXT: PseudoRET implicit $x10
%0:_(s32) = COPY $x10
%1:_(s32) = COPY $x11
%2:_(s7) = G_TRUNC %0(s32)
%3:_(s7) = G_TRUNC %1(s32)
%4:_(s7) = G_MUL %2, %3
%5:_(s32) = G_ANYEXT %4(s7)
$x10 = COPY %5(s32)
PseudoRET implicit $x10
...
---
name: mul_i8
body: |
bb.0.entry:
; CHECK-LABEL: name: mul_i8
; CHECK: [[COPY:%[0-9]+]]:_(s32) = COPY $x10
; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $x11
; CHECK-NEXT: [[MUL:%[0-9]+]]:_(s32) = G_MUL [[COPY]], [[COPY1]]
; CHECK-NEXT: $x10 = COPY [[MUL]](s32)
; CHECK-NEXT: PseudoRET implicit $x10
%0:_(s32) = COPY $x10
%1:_(s32) = COPY $x11
%2:_(s8) = G_TRUNC %0(s32)
%3:_(s8) = G_TRUNC %1(s32)
%4:_(s8) = G_MUL %2, %3
%5:_(s32) = G_ANYEXT %4(s8)
$x10 = COPY %5(s32)
PseudoRET implicit $x10
...
---
name: mul_i16
body: |
bb.0.entry:
; CHECK-LABEL: name: mul_i16
; CHECK: [[COPY:%[0-9]+]]:_(s32) = COPY $x10
; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $x11
; CHECK-NEXT: [[MUL:%[0-9]+]]:_(s32) = G_MUL [[COPY]], [[COPY1]]
; CHECK-NEXT: $x10 = COPY [[MUL]](s32)
; CHECK-NEXT: PseudoRET implicit $x10
%0:_(s32) = COPY $x10
%1:_(s32) = COPY $x11
%2:_(s16) = G_TRUNC %0(s32)
%3:_(s16) = G_TRUNC %1(s32)
%4:_(s16) = G_MUL %2, %3
%5:_(s32) = G_ANYEXT %4(s16)
$x10 = COPY %5(s32)
PseudoRET implicit $x10
...
---
name: mul_i32
body: |
bb.0.entry:
; CHECK-LABEL: name: mul_i32
; CHECK: [[COPY:%[0-9]+]]:_(s32) = COPY $x10
; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $x11
; CHECK-NEXT: [[MUL:%[0-9]+]]:_(s32) = G_MUL [[COPY]], [[COPY1]]
; CHECK-NEXT: $x10 = COPY [[MUL]](s32)
; CHECK-NEXT: PseudoRET implicit $x10
%0:_(s32) = COPY $x10
%1:_(s32) = COPY $x11
%2:_(s32) = G_MUL %0, %1
$x10 = COPY %2(s32)
PseudoRET implicit $x10
...
---
name: mul_i48
body: |
bb.0.entry:
; CHECK-LABEL: name: mul_i48
; CHECK: %xhi:_(s32) = COPY $x10
; CHECK-NEXT: %xlo:_(s32) = COPY $x11
; CHECK-NEXT: %yhi:_(s32) = COPY $x12
; CHECK-NEXT: %ylo:_(s32) = COPY $x13
; CHECK-NEXT: [[MUL:%[0-9]+]]:_(s32) = G_MUL %xhi, %yhi
; CHECK-NEXT: [[MUL1:%[0-9]+]]:_(s32) = G_MUL %xlo, %yhi
; CHECK-NEXT: [[MUL2:%[0-9]+]]:_(s32) = G_MUL %xhi, %ylo
; CHECK-NEXT: [[UMULH:%[0-9]+]]:_(s32) = G_UMULH %xhi, %yhi
; CHECK-NEXT: [[ADD:%[0-9]+]]:_(s32) = G_ADD [[MUL1]], [[MUL2]]
; CHECK-NEXT: [[ADD1:%[0-9]+]]:_(s32) = G_ADD [[ADD]], [[UMULH]]
; CHECK-NEXT: $x10 = COPY [[MUL]](s32)
; CHECK-NEXT: $x11 = COPY [[ADD1]](s32)
; CHECK-NEXT: PseudoRET implicit $x10, implicit $x11
%xhi:_(s32) = COPY $x10
%xlo:_(s32) = COPY $x11
%yhi:_(s32) = COPY $x12
%ylo:_(s32) = COPY $x13
%x0:_(s64) = G_MERGE_VALUES %xhi(s32), %xlo(s32)
%y0:_(s64) = G_MERGE_VALUES %yhi(s32), %ylo(s32)
%x:_(s48) = G_TRUNC %x0(s64)
%y:_(s48) = G_TRUNC %y0(s64)
%z:_(s48) = G_MUL %x, %y
%z0:_(s64) = G_ANYEXT %z(s48)
%zhi:_(s32), %zlo:_(s32) = G_UNMERGE_VALUES %z0(s64)
$x10 = COPY %zhi(s32)
$x11 = COPY %zlo(s32)
PseudoRET implicit $x10, implicit $x11
...
---
name: mul_i64
body: |
bb.0.entry:
; CHECK-LABEL: name: mul_i64
; CHECK: %hi1:_(s32) = COPY $x10
; CHECK-NEXT: %lo1:_(s32) = COPY $x11
; CHECK-NEXT: %hi2:_(s32) = COPY $x12
; CHECK-NEXT: %lo2:_(s32) = COPY $x13
; CHECK-NEXT: [[MUL:%[0-9]+]]:_(s32) = G_MUL %hi1, %hi2
; CHECK-NEXT: [[MUL1:%[0-9]+]]:_(s32) = G_MUL %lo1, %hi2
; CHECK-NEXT: [[MUL2:%[0-9]+]]:_(s32) = G_MUL %hi1, %lo2
; CHECK-NEXT: [[UMULH:%[0-9]+]]:_(s32) = G_UMULH %hi1, %hi2
; CHECK-NEXT: [[ADD:%[0-9]+]]:_(s32) = G_ADD [[MUL1]], [[MUL2]]
; CHECK-NEXT: [[ADD1:%[0-9]+]]:_(s32) = G_ADD [[ADD]], [[UMULH]]
; CHECK-NEXT: $x10 = COPY [[MUL]](s32)
; CHECK-NEXT: $x11 = COPY [[ADD1]](s32)
; CHECK-NEXT: PseudoRET implicit $x10, implicit $x11
%hi1:_(s32) = COPY $x10
%lo1:_(s32) = COPY $x11
%hi2:_(s32) = COPY $x12
%lo2:_(s32) = COPY $x13
%x1:_(s64) = G_MERGE_VALUES %hi1(s32), %lo1(s32)
%x2:_(s64) = G_MERGE_VALUES %hi2(s32), %lo2(s32)
%y:_(s64) = G_MUL %x1, %x2
%hiy:_(s32), %loy:_(s32) = G_UNMERGE_VALUES %y(s64)
$x10 = COPY %hiy(s32)
$x11 = COPY %loy(s32)
PseudoRET implicit $x10, implicit $x11
...
---
name: mul_i96
body: |
bb.0.entry:
; CHECK-LABEL: name: mul_i96
; CHECK: %hi1:_(s32) = COPY $x10
; CHECK-NEXT: %mid1:_(s32) = COPY $x11
; CHECK-NEXT: %lo1:_(s32) = COPY $x12
; CHECK-NEXT: %hi2:_(s32) = COPY $x13
; CHECK-NEXT: %mid2:_(s32) = COPY $x14
; CHECK-NEXT: %lo2:_(s32) = COPY $x15
; CHECK-NEXT: [[MUL:%[0-9]+]]:_(s32) = G_MUL %hi1, %hi2
; CHECK-NEXT: [[MUL1:%[0-9]+]]:_(s32) = G_MUL %mid1, %hi2
; CHECK-NEXT: [[MUL2:%[0-9]+]]:_(s32) = G_MUL %hi1, %mid2
; CHECK-NEXT: [[UMULH:%[0-9]+]]:_(s32) = G_UMULH %hi1, %hi2
; CHECK-NEXT: [[UADDO:%[0-9]+]]:_(s32), [[UADDO1:%[0-9]+]]:_(s32) = G_UADDO [[MUL1]], [[MUL2]]
; CHECK-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
; CHECK-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[UADDO1]], [[C]]
; CHECK-NEXT: [[UADDO2:%[0-9]+]]:_(s32), [[UADDO3:%[0-9]+]]:_(s32) = G_UADDO [[UADDO]], [[UMULH]]
; CHECK-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
; CHECK-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[UADDO3]], [[C1]]
; CHECK-NEXT: [[ADD:%[0-9]+]]:_(s32) = G_ADD [[AND]], [[AND1]]
; CHECK-NEXT: [[MUL3:%[0-9]+]]:_(s32) = G_MUL %lo1, %hi2
; CHECK-NEXT: [[MUL4:%[0-9]+]]:_(s32) = G_MUL %mid1, %mid2
; CHECK-NEXT: [[MUL5:%[0-9]+]]:_(s32) = G_MUL %hi1, %lo2
; CHECK-NEXT: [[UMULH1:%[0-9]+]]:_(s32) = G_UMULH %mid1, %hi2
; CHECK-NEXT: [[UMULH2:%[0-9]+]]:_(s32) = G_UMULH %hi1, %mid2
; CHECK-NEXT: [[UADDO4:%[0-9]+]]:_(s32), [[UADDO5:%[0-9]+]]:_(s32) = G_UADDO [[MUL3]], [[MUL4]]
; CHECK-NEXT: [[UADDO6:%[0-9]+]]:_(s32), [[UADDO7:%[0-9]+]]:_(s32) = G_UADDO [[UADDO4]], [[MUL5]]
; CHECK-NEXT: [[UADDO8:%[0-9]+]]:_(s32), [[UADDO9:%[0-9]+]]:_(s32) = G_UADDO [[UADDO6]], [[UMULH1]]
; CHECK-NEXT: [[UADDO10:%[0-9]+]]:_(s32), [[UADDO11:%[0-9]+]]:_(s32) = G_UADDO [[UADDO8]], [[UMULH2]]
; CHECK-NEXT: [[UADDO12:%[0-9]+]]:_(s32), [[UADDO13:%[0-9]+]]:_(s32) = G_UADDO [[UADDO10]], [[ADD]]
; CHECK-NEXT: $x10 = COPY [[MUL]](s32)
; CHECK-NEXT: $x11 = COPY [[UADDO2]](s32)
; CHECK-NEXT: $x12 = COPY [[UADDO12]](s32)
; CHECK-NEXT: PseudoRET implicit $x10, implicit $x11, implicit $x12
%hi1:_(s32) = COPY $x10
%mid1:_(s32) = COPY $x11
%lo1:_(s32) = COPY $x12
%hi2:_(s32) = COPY $x13
%mid2:_(s32) = COPY $x14
%lo2:_(s32) = COPY $x15
%x1:_(s96) = G_MERGE_VALUES %hi1(s32), %mid1(s32), %lo1(s32)
%x2:_(s96) = G_MERGE_VALUES %hi2(s32), %mid2(s32), %lo2(s32)
%y:_(s96) = G_MUL %x1, %x2
%hi:_(s32), %mid:_(s32), %lo:_(s32) = G_UNMERGE_VALUES %y(s96)
$x10 = COPY %hi(s32)
$x11 = COPY %mid(s32)
$x12 = COPY %lo(s32)
PseudoRET implicit $x10, implicit $x11, implicit $x12
...

llvm/test/CodeGen/RISCV/GlobalISel/legalizer/rv32/legalize-mul.mir

  • This file was added.
# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
# RUN: llc -mtriple=riscv32 -run-pass=legalizer %s -o - \
# RUN: | FileCheck %s
---
name: mul_i7
body: |
bb.0.entry:
; CHECK-LABEL: name: mul_i7
; CHECK: [[COPY:%[0-9]+]]:_(s32) = COPY $x10
; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $x11
; CHECK-NEXT: $x10 = COPY [[COPY]](s32)
; CHECK-NEXT: $x11 = COPY [[COPY1]](s32)
; CHECK-NEXT: PseudoCALL target-flags(riscv-call) &__mulsi3, implicit-def $x1, implicit $x10, implicit $x11, implicit-def $x10
; CHECK-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY $x10
; CHECK-NEXT: $x10 = COPY [[COPY2]](s32)
; CHECK-NEXT: PseudoRET implicit $x10
%0:_(s32) = COPY $x10
%1:_(s32) = COPY $x11
%2:_(s7) = G_TRUNC %0(s32)
%3:_(s7) = G_TRUNC %1(s32)
%4:_(s7) = G_MUL %2, %3
%5:_(s32) = G_ANYEXT %4(s7)
$x10 = COPY %5(s32)
PseudoRET implicit $x10
...
---
name: mul_i8
body: |
bb.0.entry:
; CHECK-LABEL: name: mul_i8
; CHECK: [[COPY:%[0-9]+]]:_(s32) = COPY $x10
; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $x11
; CHECK-NEXT: $x10 = COPY [[COPY]](s32)
; CHECK-NEXT: $x11 = COPY [[COPY1]](s32)
; CHECK-NEXT: PseudoCALL target-flags(riscv-call) &__mulsi3, implicit-def $x1, implicit $x10, implicit $x11, implicit-def $x10
; CHECK-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY $x10
; CHECK-NEXT: $x10 = COPY [[COPY2]](s32)
; CHECK-NEXT: PseudoRET implicit $x10
%0:_(s32) = COPY $x10
%1:_(s32) = COPY $x11
%2:_(s8) = G_TRUNC %0(s32)
%3:_(s8) = G_TRUNC %1(s32)
%4:_(s8) = G_MUL %2, %3
%5:_(s32) = G_ANYEXT %4(s8)
$x10 = COPY %5(s32)
PseudoRET implicit $x10
...
---
name: mul_i16
body: |
bb.0.entry:
; CHECK-LABEL: name: mul_i16
; CHECK: [[COPY:%[0-9]+]]:_(s32) = COPY $x10
; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $x11
; CHECK-NEXT: $x10 = COPY [[COPY]](s32)
; CHECK-NEXT: $x11 = COPY [[COPY1]](s32)
; CHECK-NEXT: PseudoCALL target-flags(riscv-call) &__mulsi3, implicit-def $x1, implicit $x10, implicit $x11, implicit-def $x10
; CHECK-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY $x10
; CHECK-NEXT: $x10 = COPY [[COPY2]](s32)
; CHECK-NEXT: PseudoRET implicit $x10
%0:_(s32) = COPY $x10
%1:_(s32) = COPY $x11
%2:_(s16) = G_TRUNC %0(s32)
%3:_(s16) = G_TRUNC %1(s32)
%4:_(s16) = G_MUL %2, %3
%5:_(s32) = G_ANYEXT %4(s16)
$x10 = COPY %5(s32)
PseudoRET implicit $x10
...
---
name: mul_i32
body: |
bb.0.entry:
; CHECK-LABEL: name: mul_i32
; CHECK: [[COPY:%[0-9]+]]:_(s32) = COPY $x10
; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $x11
; CHECK-NEXT: $x10 = COPY [[COPY]](s32)
; CHECK-NEXT: $x11 = COPY [[COPY1]](s32)
; CHECK-NEXT: PseudoCALL target-flags(riscv-call) &__mulsi3, implicit-def $x1, implicit $x10, implicit $x11, implicit-def $x10
; CHECK-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY $x10
; CHECK-NEXT: $x10 = COPY [[COPY2]](s32)
; CHECK-NEXT: PseudoRET implicit $x10
%0:_(s32) = COPY $x10
%1:_(s32) = COPY $x11
%2:_(s32) = G_MUL %0, %1
$x10 = COPY %2(s32)
PseudoRET implicit $x10
...
---
name: mul_i64
body: |
bb.0.entry:
; CHECK-LABEL: name: mul_i64
; CHECK: %hi1:_(s32) = COPY $x10
; CHECK-NEXT: %lo1:_(s32) = COPY $x11
; CHECK-NEXT: %hi2:_(s32) = COPY $x12
; CHECK-NEXT: %lo2:_(s32) = COPY $x13
; CHECK-NEXT: $x10 = COPY %hi1(s32)
; CHECK-NEXT: $x11 = COPY %lo1(s32)
; CHECK-NEXT: $x12 = COPY %hi2(s32)
; CHECK-NEXT: $x13 = COPY %lo2(s32)
; CHECK-NEXT: PseudoCALL target-flags(riscv-call) &__muldi3, implicit-def $x1, implicit $x10, implicit $x11, implicit $x12, implicit $x13, implicit-def $x10, implicit-def $x11
; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $x10
; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $x11
; CHECK-NEXT: $x10 = COPY [[COPY]](s32)
; CHECK-NEXT: $x11 = COPY [[COPY1]](s32)
; CHECK-NEXT: PseudoRET implicit $x10, implicit $x11
%hi1:_(s32) = COPY $x10
%lo1:_(s32) = COPY $x11
%hi2:_(s32) = COPY $x12
%lo2:_(s32) = COPY $x13
%x1:_(s64) = G_MERGE_VALUES %hi1(s32), %lo1(s32)
%x2:_(s64) = G_MERGE_VALUES %hi2(s32), %lo2(s32)
%y:_(s64) = G_MUL %x1, %x2
%hiy:_(s32), %loy:_(s32) = G_UNMERGE_VALUES %y(s64)
$x10 = COPY %hiy(s32)
$x11 = COPY %loy(s32)
PseudoRET implicit $x10, implicit $x11
...

llvm/test/CodeGen/RISCV/GlobalISel/legalizer/rv64/legalize-mul-ext.mir

  • This file was added.
# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
# RUN: llc -mattr=+m -mtriple=riscv64 -run-pass=legalizer %s -o - \
# RUN: | FileCheck %s
# RUN: llc -mattr=+zmmul -mtriple=riscv64 -run-pass=legalizer %s -o - \
craig.topperUnsubmitted
Done
ReplyInline Actions

Check prefixes in this test have an I in them that the rv32 tests don't have. We should be consistent

craig.topper: Check prefixes in this test have an `I` in them that the rv32 tests don't have. We should be…
# RUN: | FileCheck %s
---
name: mul_i8
body: |
bb.0.entry:
; CHECK-LABEL: name: mul_i8
; CHECK: [[COPY:%[0-9]+]]:_(s64) = COPY $x10
; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s64) = COPY $x11
; CHECK-NEXT: [[MUL:%[0-9]+]]:_(s64) = G_MUL [[COPY]], [[COPY1]]
; CHECK-NEXT: $x10 = COPY [[MUL]](s64)
; CHECK-NEXT: PseudoRET implicit $x10
%0:_(s64) = COPY $x10
%1:_(s64) = COPY $x11
%2:_(s8) = G_TRUNC %0(s64)
%3:_(s8) = G_TRUNC %1(s64)
%4:_(s8) = G_MUL %2, %3
%5:_(s64) = G_ANYEXT %4(s8)
$x10 = COPY %5(s64)
PseudoRET implicit $x10
...
---
name: mul_i15
body: |
bb.0.entry:
; CHECK-LABEL: name: mul_i15
; CHECK: [[COPY:%[0-9]+]]:_(s64) = COPY $x10
; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s64) = COPY $x11
; CHECK-NEXT: [[MUL:%[0-9]+]]:_(s64) = G_MUL [[COPY]], [[COPY1]]
; CHECK-NEXT: $x10 = COPY [[MUL]](s64)
; CHECK-NEXT: PseudoRET implicit $x10
%0:_(s64) = COPY $x10
%1:_(s64) = COPY $x11
%2:_(s15) = G_TRUNC %0(s64)
%3:_(s15) = G_TRUNC %1(s64)
%4:_(s15) = G_MUL %2, %3
%5:_(s64) = G_ANYEXT %4(s15)
$x10 = COPY %5(s64)
PseudoRET implicit $x10
...
---
name: mul_i16
body: |
bb.0.entry:
; CHECK-LABEL: name: mul_i16
; CHECK: [[COPY:%[0-9]+]]:_(s64) = COPY $x10
; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s64) = COPY $x11
; CHECK-NEXT: [[MUL:%[0-9]+]]:_(s64) = G_MUL [[COPY]], [[COPY1]]
; CHECK-NEXT: $x10 = COPY [[MUL]](s64)
; CHECK-NEXT: PseudoRET implicit $x10
%0:_(s64) = COPY $x10
%1:_(s64) = COPY $x11
%2:_(s16) = G_TRUNC %0(s64)
%3:_(s16) = G_TRUNC %1(s64)
%4:_(s16) = G_MUL %2, %3
%5:_(s64) = G_ANYEXT %4(s16)
$x10 = COPY %5(s64)
PseudoRET implicit $x10
...
---
name: mul_i32
body: |
bb.0.entry:
; CHECK-LABEL: name: mul_i32
; CHECK: [[COPY:%[0-9]+]]:_(s64) = COPY $x10
; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s64) = COPY $x11
; CHECK-NEXT: [[MUL:%[0-9]+]]:_(s64) = G_MUL [[COPY]], [[COPY1]]
; CHECK-NEXT: $x10 = COPY [[MUL]](s64)
; CHECK-NEXT: PseudoRET implicit $x10
%0:_(s64) = COPY $x10
%1:_(s64) = COPY $x11
%2:_(s32) = G_TRUNC %0(s64)
%3:_(s32) = G_TRUNC %1(s64)
%4:_(s32) = G_MUL %2, %3
%5:_(s64) = G_ANYEXT %4(s32)
$x10 = COPY %5(s64)
PseudoRET implicit $x10
...
---
name: mul_i64
body: |
bb.0.entry:
; CHECK-LABEL: name: mul_i64
; CHECK: [[COPY:%[0-9]+]]:_(s64) = COPY $x10
; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s64) = COPY $x11
; CHECK-NEXT: [[MUL:%[0-9]+]]:_(s64) = G_MUL [[COPY]], [[COPY1]]
; CHECK-NEXT: $x10 = COPY [[MUL]](s64)
; CHECK-NEXT: PseudoRET implicit $x10
%0:_(s64) = COPY $x10
%1:_(s64) = COPY $x11
%2:_(s64) = G_MUL %0, %1
$x10 = COPY %2(s64)
PseudoRET implicit $x10
...
---
name: mul_i72
body: |
bb.0.entry:
; CHECK-LABEL: name: mul_i72
; CHECK: %xhi:_(s64) = COPY $x10
; CHECK-NEXT: %xlo:_(s64) = COPY $x11
; CHECK-NEXT: %yhi:_(s64) = COPY $x12
; CHECK-NEXT: %ylo:_(s64) = COPY $x13
; CHECK-NEXT: [[MUL:%[0-9]+]]:_(s64) = G_MUL %xhi, %yhi
; CHECK-NEXT: [[MUL1:%[0-9]+]]:_(s64) = G_MUL %xlo, %yhi
; CHECK-NEXT: [[MUL2:%[0-9]+]]:_(s64) = G_MUL %xhi, %ylo
; CHECK-NEXT: [[UMULH:%[0-9]+]]:_(s64) = G_UMULH %xhi, %yhi
; CHECK-NEXT: [[ADD:%[0-9]+]]:_(s64) = G_ADD [[MUL1]], [[MUL2]]
; CHECK-NEXT: [[ADD1:%[0-9]+]]:_(s64) = G_ADD [[ADD]], [[UMULH]]
; CHECK-NEXT: $x10 = COPY [[MUL]](s64)
; CHECK-NEXT: $x11 = COPY [[ADD1]](s64)
; CHECK-NEXT: PseudoRET implicit $x10, implicit $x11
%xhi:_(s64) = COPY $x10
%xlo:_(s64) = COPY $x11
%yhi:_(s64) = COPY $x12
%ylo:_(s64) = COPY $x13
%x0:_(s128) = G_MERGE_VALUES %xhi(s64), %xlo(s64)
%y0:_(s128) = G_MERGE_VALUES %yhi(s64), %ylo(s64)
%x:_(s72) = G_TRUNC %x0(s128)
%y:_(s72) = G_TRUNC %y0(s128)
%z:_(s72) = G_MUL %x, %y
%z0:_(s128) = G_ANYEXT %z(s72)
%zhi:_(s64), %zlo:_(s64) = G_UNMERGE_VALUES %z0(s128)
$x10 = COPY %zhi(s64)
$x11 = COPY %zlo(s64)
PseudoRET implicit $x10, implicit $x11
...
---
name: mul_i128
body: |
bb.0.entry:
; CHECK-LABEL: name: mul_i128
; CHECK: %hi1:_(s64) = COPY $x10
; CHECK-NEXT: %lo1:_(s64) = COPY $x11
; CHECK-NEXT: %hi2:_(s64) = COPY $x12
; CHECK-NEXT: %lo2:_(s64) = COPY $x13
; CHECK-NEXT: [[MUL:%[0-9]+]]:_(s64) = G_MUL %hi1, %hi2
; CHECK-NEXT: [[MUL1:%[0-9]+]]:_(s64) = G_MUL %lo1, %hi2
; CHECK-NEXT: [[MUL2:%[0-9]+]]:_(s64) = G_MUL %hi1, %lo2
; CHECK-NEXT: [[UMULH:%[0-9]+]]:_(s64) = G_UMULH %hi1, %hi2
; CHECK-NEXT: [[ADD:%[0-9]+]]:_(s64) = G_ADD [[MUL1]], [[MUL2]]
; CHECK-NEXT: [[ADD1:%[0-9]+]]:_(s64) = G_ADD [[ADD]], [[UMULH]]
; CHECK-NEXT: $x10 = COPY [[MUL]](s64)
; CHECK-NEXT: $x11 = COPY [[ADD1]](s64)
; CHECK-NEXT: PseudoRET implicit $x10, implicit $x11
%hi1:_(s64) = COPY $x10
%lo1:_(s64) = COPY $x11
%hi2:_(s64) = COPY $x12
%lo2:_(s64) = COPY $x13
%x1:_(s128) = G_MERGE_VALUES %hi1(s64), %lo1(s64)
%x2:_(s128) = G_MERGE_VALUES %hi2(s64), %lo2(s64)
%y:_(s128) = G_MUL %x1, %x2
%hiy:_(s64), %loy:_(s64) = G_UNMERGE_VALUES %y(s128)
$x10 = COPY %hiy(s64)
$x11 = COPY %loy(s64)
PseudoRET implicit $x10, implicit $x11
...
---
name: mul_i192
body: |
bb.0.entry:
; CHECK-LABEL: name: mul_i192
; CHECK: %hi1:_(s64) = COPY $x10
; CHECK-NEXT: %mid1:_(s64) = COPY $x11
; CHECK-NEXT: %lo1:_(s64) = COPY $x12
; CHECK-NEXT: %hi2:_(s64) = COPY $x13
; CHECK-NEXT: %mid2:_(s64) = COPY $x14
; CHECK-NEXT: %lo2:_(s64) = COPY $x15
; CHECK-NEXT: [[MUL:%[0-9]+]]:_(s64) = G_MUL %hi1, %hi2
; CHECK-NEXT: [[MUL1:%[0-9]+]]:_(s64) = G_MUL %mid1, %hi2
; CHECK-NEXT: [[MUL2:%[0-9]+]]:_(s64) = G_MUL %hi1, %mid2
; CHECK-NEXT: [[UMULH:%[0-9]+]]:_(s64) = G_UMULH %hi1, %hi2
; CHECK-NEXT: [[UADDO:%[0-9]+]]:_(s64), [[UADDO1:%[0-9]+]]:_(s64) = G_UADDO [[MUL1]], [[MUL2]]
; CHECK-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
; CHECK-NEXT: [[AND:%[0-9]+]]:_(s64) = G_AND [[UADDO1]], [[C]]
; CHECK-NEXT: [[UADDO2:%[0-9]+]]:_(s64), [[UADDO3:%[0-9]+]]:_(s64) = G_UADDO [[UADDO]], [[UMULH]]
; CHECK-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
; CHECK-NEXT: [[AND1:%[0-9]+]]:_(s64) = G_AND [[UADDO3]], [[C1]]
; CHECK-NEXT: [[ADD:%[0-9]+]]:_(s64) = G_ADD [[AND]], [[AND1]]
; CHECK-NEXT: [[MUL3:%[0-9]+]]:_(s64) = G_MUL %lo1, %hi2
; CHECK-NEXT: [[MUL4:%[0-9]+]]:_(s64) = G_MUL %mid1, %mid2
; CHECK-NEXT: [[MUL5:%[0-9]+]]:_(s64) = G_MUL %hi1, %lo2
; CHECK-NEXT: [[UMULH1:%[0-9]+]]:_(s64) = G_UMULH %mid1, %hi2
; CHECK-NEXT: [[UMULH2:%[0-9]+]]:_(s64) = G_UMULH %hi1, %mid2
; CHECK-NEXT: [[UADDO4:%[0-9]+]]:_(s64), [[UADDO5:%[0-9]+]]:_(s64) = G_UADDO [[MUL3]], [[MUL4]]
; CHECK-NEXT: [[UADDO6:%[0-9]+]]:_(s64), [[UADDO7:%[0-9]+]]:_(s64) = G_UADDO [[UADDO4]], [[MUL5]]
; CHECK-NEXT: [[UADDO8:%[0-9]+]]:_(s64), [[UADDO9:%[0-9]+]]:_(s64) = G_UADDO [[UADDO6]], [[UMULH1]]
; CHECK-NEXT: [[UADDO10:%[0-9]+]]:_(s64), [[UADDO11:%[0-9]+]]:_(s64) = G_UADDO [[UADDO8]], [[UMULH2]]
; CHECK-NEXT: [[UADDO12:%[0-9]+]]:_(s64), [[UADDO13:%[0-9]+]]:_(s64) = G_UADDO [[UADDO10]], [[ADD]]
; CHECK-NEXT: $x10 = COPY [[MUL]](s64)
; CHECK-NEXT: $x11 = COPY [[UADDO2]](s64)
; CHECK-NEXT: $x12 = COPY [[UADDO12]](s64)
; CHECK-NEXT: PseudoRET implicit $x10, implicit $x11, implicit $x12
%hi1:_(s64) = COPY $x10
%mid1:_(s64) = COPY $x11
%lo1:_(s64) = COPY $x12
%hi2:_(s64) = COPY $x13
%mid2:_(s64) = COPY $x14
%lo2:_(s64) = COPY $x15
%x1:_(s192) = G_MERGE_VALUES %hi1(s64), %mid1(s64), %lo1(s64)
%x2:_(s192) = G_MERGE_VALUES %hi2(s64), %mid2(s64), %lo2(s64)
%y:_(s192) = G_MUL %x1, %x2
%hi:_(s64), %mid:_(s64), %lo:_(s64) = G_UNMERGE_VALUES %y(s192)
$x10 = COPY %hi(s64)
$x11 = COPY %mid(s64)
$x12 = COPY %lo(s64)
PseudoRET implicit $x10, implicit $x11, implicit $x12
...

llvm/test/CodeGen/RISCV/GlobalISel/legalizer/rv64/legalize-mul.mir

  • This file was added.
# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
# RUN: llc -mtriple=riscv64 -run-pass=legalizer %s -o - \
# RUN: | FileCheck %s
---
name: mul_i8
body: |
bb.0.entry:
; CHECK-LABEL: name: mul_i8
; CHECK: [[COPY:%[0-9]+]]:_(s64) = COPY $x10
; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s64) = COPY $x11
; CHECK-NEXT: $x10 = COPY [[COPY]](s64)
; CHECK-NEXT: $x11 = COPY [[COPY1]](s64)
; CHECK-NEXT: PseudoCALL target-flags(riscv-call) &__muldi3, implicit-def $x1, implicit $x10, implicit $x11, implicit-def $x10
; CHECK-NEXT: [[COPY2:%[0-9]+]]:_(s64) = COPY $x10
; CHECK-NEXT: $x10 = COPY [[COPY2]](s64)
; CHECK-NEXT: PseudoRET implicit $x10
%0:_(s64) = COPY $x10
%1:_(s64) = COPY $x11
%2:_(s8) = G_TRUNC %0(s64)
%3:_(s8) = G_TRUNC %1(s64)
%4:_(s8) = G_MUL %2, %3
%5:_(s64) = G_ANYEXT %4(s8)
$x10 = COPY %5(s64)
PseudoRET implicit $x10
...
---
name: mul_i15
body: |
bb.0.entry:
; CHECK-LABEL: name: mul_i15
; CHECK: [[COPY:%[0-9]+]]:_(s64) = COPY $x10
; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s64) = COPY $x11
; CHECK-NEXT: $x10 = COPY [[COPY]](s64)
; CHECK-NEXT: $x11 = COPY [[COPY1]](s64)
; CHECK-NEXT: PseudoCALL target-flags(riscv-call) &__muldi3, implicit-def $x1, implicit $x10, implicit $x11, implicit-def $x10
; CHECK-NEXT: [[COPY2:%[0-9]+]]:_(s64) = COPY $x10
; CHECK-NEXT: $x10 = COPY [[COPY2]](s64)
; CHECK-NEXT: PseudoRET implicit $x10
%0:_(s64) = COPY $x10
%1:_(s64) = COPY $x11
%2:_(s15) = G_TRUNC %0(s64)
%3:_(s15) = G_TRUNC %1(s64)
%4:_(s15) = G_MUL %2, %3
%5:_(s64) = G_ANYEXT %4(s15)
$x10 = COPY %5(s64)
PseudoRET implicit $x10
...
---
name: mul_i16
body: |
bb.0.entry:
; CHECK-LABEL: name: mul_i16
; CHECK: [[COPY:%[0-9]+]]:_(s64) = COPY $x10
; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s64) = COPY $x11
; CHECK-NEXT: $x10 = COPY [[COPY]](s64)
; CHECK-NEXT: $x11 = COPY [[COPY1]](s64)
; CHECK-NEXT: PseudoCALL target-flags(riscv-call) &__muldi3, implicit-def $x1, implicit $x10, implicit $x11, implicit-def $x10
; CHECK-NEXT: [[COPY2:%[0-9]+]]:_(s64) = COPY $x10
; CHECK-NEXT: $x10 = COPY [[COPY2]](s64)
; CHECK-NEXT: PseudoRET implicit $x10
%0:_(s64) = COPY $x10
%1:_(s64) = COPY $x11
%2:_(s16) = G_TRUNC %0(s64)
%3:_(s16) = G_TRUNC %1(s64)
%4:_(s16) = G_MUL %2, %3
%5:_(s64) = G_ANYEXT %4(s16)
$x10 = COPY %5(s64)
PseudoRET implicit $x10
...
---
name: mul_i32
body: |
bb.0.entry:
; CHECK-LABEL: name: mul_i32
; CHECK: [[COPY:%[0-9]+]]:_(s64) = COPY $x10
; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s64) = COPY $x11
; CHECK-NEXT: $x10 = COPY [[COPY]](s64)
; CHECK-NEXT: $x11 = COPY [[COPY1]](s64)
; CHECK-NEXT: PseudoCALL target-flags(riscv-call) &__muldi3, implicit-def $x1, implicit $x10, implicit $x11, implicit-def $x10
; CHECK-NEXT: [[COPY2:%[0-9]+]]:_(s64) = COPY $x10
; CHECK-NEXT: $x10 = COPY [[COPY2]](s64)
; CHECK-NEXT: PseudoRET implicit $x10
%0:_(s64) = COPY $x10
%1:_(s64) = COPY $x11
%2:_(s32) = G_TRUNC %0(s64)
%3:_(s32) = G_TRUNC %1(s64)
%4:_(s32) = G_MUL %2, %3
%5:_(s64) = G_ANYEXT %4(s32)
$x10 = COPY %5(s64)
PseudoRET implicit $x10
...
---
name: mul_i64
body: |
bb.0.entry:
; CHECK-LABEL: name: mul_i64
; CHECK: [[COPY:%[0-9]+]]:_(s64) = COPY $x10
; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s64) = COPY $x11
; CHECK-NEXT: $x10 = COPY [[COPY]](s64)
; CHECK-NEXT: $x11 = COPY [[COPY1]](s64)
; CHECK-NEXT: PseudoCALL target-flags(riscv-call) &__muldi3, implicit-def $x1, implicit $x10, implicit $x11, implicit-def $x10
; CHECK-NEXT: [[COPY2:%[0-9]+]]:_(s64) = COPY $x10
; CHECK-NEXT: $x10 = COPY [[COPY2]](s64)
; CHECK-NEXT: PseudoRET implicit $x10
%0:_(s64) = COPY $x10
%1:_(s64) = COPY $x11
%2:_(s64) = G_MUL %0, %1
$x10 = COPY %2(s64)
PseudoRET implicit $x10
...
---
name: mul_i128
body: |
bb.0.entry:
; CHECK-LABEL: name: mul_i128
; CHECK: %hi1:_(s64) = COPY $x10
; CHECK-NEXT: %lo1:_(s64) = COPY $x11
; CHECK-NEXT: %hi2:_(s64) = COPY $x12
; CHECK-NEXT: %lo2:_(s64) = COPY $x13
; CHECK-NEXT: $x10 = COPY %hi1(s64)
; CHECK-NEXT: $x11 = COPY %lo1(s64)
; CHECK-NEXT: $x12 = COPY %hi2(s64)
; CHECK-NEXT: $x13 = COPY %lo2(s64)
; CHECK-NEXT: PseudoCALL target-flags(riscv-call) &__multi3, implicit-def $x1, implicit $x10, implicit $x11, implicit $x12, implicit $x13, implicit-def $x10, implicit-def $x11
; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s64) = COPY $x10
; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s64) = COPY $x11
; CHECK-NEXT: $x10 = COPY [[COPY]](s64)
; CHECK-NEXT: $x11 = COPY [[COPY1]](s64)
; CHECK-NEXT: PseudoRET implicit $x10, implicit $x11
%hi1:_(s64) = COPY $x10
%lo1:_(s64) = COPY $x11
%hi2:_(s64) = COPY $x12
%lo2:_(s64) = COPY $x13
%x1:_(s128) = G_MERGE_VALUES %hi1(s64), %lo1(s64)
%x2:_(s128) = G_MERGE_VALUES %hi2(s64), %lo2(s64)
%y:_(s128) = G_MUL %x1, %x2
%hiy:_(s64), %loy:_(s64) = G_UNMERGE_VALUES %y(s128)
$x10 = COPY %hiy(s64)
$x11 = COPY %loy(s64)
PseudoRET implicit $x10, implicit $x11
...