Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

InlineSpiller: Delete assert that implicit_def has no implicit operands #69087

Merged
merged 1 commit into from
Oct 19, 2023

Conversation

arsenm
Copy link
Contributor

@arsenm arsenm commented Oct 15, 2023

It's not a verifier enforced property that implicit_def may only have one operand. Fixes assertions after the coalescer implicit-defs to preserve super register liveness to arbitrary instructions.

For some reason I'm unable to reproduce this as a MIR test running only the allocator for the x86 test. Not sure it's worth keeping around.

@llvmbot
Copy link

llvmbot commented Oct 15, 2023

@llvm/pr-subscribers-llvm-regalloc
@llvm/pr-subscribers-backend-aarch64

@llvm/pr-subscribers-backend-x86

Author: Matt Arsenault (arsenm)

Changes

It's not a verifier enforced property that implicit_def may only have one operand. Fixes assertions after the coalescer implicit-defs to preserve super register liveness to arbitrary instructions.

For some reason I'm unable to reproduce this as a MIR test running only the allocator for the x86 test. Not sure it's worth keeping around.


Full diff: https://github.com/llvm/llvm-project/pull/69087.diff

3 Files Affected:

  • (modified) llvm/lib/CodeGen/InlineSpiller.cpp (+1-2)
  • (added) llvm/test/CodeGen/AArch64/implicit-def-with-impdef-greedy-assert.mir (+181)
  • (added) llvm/test/CodeGen/X86/inline-spiller-impdef-on-implicit-def-regression.ll (+181)
diff --git a/llvm/lib/CodeGen/InlineSpiller.cpp b/llvm/lib/CodeGen/InlineSpiller.cpp
index c62f3db9d321562..46fcc62e09e8a8c 100644
--- a/llvm/lib/CodeGen/InlineSpiller.cpp
+++ b/llvm/lib/CodeGen/InlineSpiller.cpp
@@ -1071,8 +1071,7 @@ void InlineSpiller::insertReload(Register NewVReg,
 static bool isRealSpill(const MachineInstr &Def) {
   if (!Def.isImplicitDef())
     return true;
-  assert(Def.getNumOperands() == 1 &&
-         "Implicit def with more than one definition");
+
   // We can say that the VReg defined by Def is undef, only if it is
   // fully defined by Def. Otherwise, some of the lanes may not be
   // undef and the value of the VReg matters.
diff --git a/llvm/test/CodeGen/AArch64/implicit-def-with-impdef-greedy-assert.mir b/llvm/test/CodeGen/AArch64/implicit-def-with-impdef-greedy-assert.mir
new file mode 100644
index 000000000000000..c79c951fdc15247
--- /dev/null
+++ b/llvm/test/CodeGen/AArch64/implicit-def-with-impdef-greedy-assert.mir
@@ -0,0 +1,181 @@
+# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py UTC_ARGS: --version 3
+# RUN: llc -mtriple=arm64-apple-ios -run-pass=greedy -o - %s | FileCheck %s
+
+---
+name:            widget
+tracksRegLiveness: true
+jumpTable:
+  kind:            label-difference32
+  entries:
+    - id:              0
+      blocks:          [ '%bb.9', '%bb.5', '%bb.2', '%bb.2', '%bb.2' ]
+body:             |
+  ; CHECK-LABEL: name: widget
+  ; CHECK: bb.0:
+  ; CHECK-NEXT:   successors: %bb.1(0x80000000)
+  ; CHECK-NEXT:   liveins: $w0, $w1, $x2, $x3, $x4, $w5, $w6
+  ; CHECK-NEXT: {{  $}}
+  ; CHECK-NEXT:   [[DEF:%[0-9]+]]:gpr64common = IMPLICIT_DEF
+  ; CHECK-NEXT:   [[DEF1:%[0-9]+]]:gpr64 = IMPLICIT_DEF
+  ; CHECK-NEXT:   [[DEF2:%[0-9]+]]:gpr32common = IMPLICIT_DEF
+  ; CHECK-NEXT:   [[DEF3:%[0-9]+]]:gpr32 = IMPLICIT_DEF
+  ; CHECK-NEXT:   [[DEF4:%[0-9]+]]:gpr32 = IMPLICIT_DEF
+  ; CHECK-NEXT:   [[DEF5:%[0-9]+]]:gpr64common = IMPLICIT_DEF
+  ; CHECK-NEXT:   [[DEF6:%[0-9]+]]:gpr32 = IMPLICIT_DEF
+  ; CHECK-NEXT:   undef [[DEF7:%[0-9]+]].sub_32:gpr64 = IMPLICIT_DEF
+  ; CHECK-NEXT:   [[DEF8:%[0-9]+]]:gpr64common = IMPLICIT_DEF
+  ; CHECK-NEXT:   [[DEF9:%[0-9]+]]:gpr32 = IMPLICIT_DEF
+  ; CHECK-NEXT:   dead [[DEF10:%[0-9]+]]:gpr64 = IMPLICIT_DEF
+  ; CHECK-NEXT:   undef [[DEF11:%[0-9]+]].sub_32:gpr64 = IMPLICIT_DEF implicit-def dead %11
+  ; CHECK-NEXT:   STRXui [[DEF11]], %stack.0, 0 :: (store (s64) into %stack.0)
+  ; CHECK-NEXT: {{  $}}
+  ; CHECK-NEXT: bb.1:
+  ; CHECK-NEXT:   successors: %bb.2(0x80000000)
+  ; CHECK-NEXT: {{  $}}
+  ; CHECK-NEXT: {{  $}}
+  ; CHECK-NEXT: bb.2:
+  ; CHECK-NEXT:   successors: %bb.3(0x0fbefbf0), %bb.4(0x70410410)
+  ; CHECK-NEXT: {{  $}}
+  ; CHECK-NEXT:   Bcc 8, %bb.3, implicit killed undef $nzcv
+  ; CHECK-NEXT:   B %bb.4
+  ; CHECK-NEXT: {{  $}}
+  ; CHECK-NEXT: bb.3:
+  ; CHECK-NEXT:   successors: %bb.11(0x00000000), %bb.2(0x80000000)
+  ; CHECK-NEXT: {{  $}}
+  ; CHECK-NEXT:   dead $wzr = SUBSWri [[DEF2]], 64, 0, implicit-def $nzcv
+  ; CHECK-NEXT:   Bcc 0, %bb.11, implicit killed undef $nzcv
+  ; CHECK-NEXT:   B %bb.2
+  ; CHECK-NEXT: {{  $}}
+  ; CHECK-NEXT: bb.4:
+  ; CHECK-NEXT:   successors: %bb.9(0x01288b01), %bb.5(0x01288b01), %bb.2(0x11f46a91), %bb.6(0x23e8d524), %bb.7(0x47d1aa49)
+  ; CHECK-NEXT: {{  $}}
+  ; CHECK-NEXT:   dead early-clobber %12:gpr64, dead early-clobber %13:gpr64sp = JumpTableDest32 [[DEF8]], [[DEF7]], %jump-table.0
+  ; CHECK-NEXT:   BR undef %18:gpr64
+  ; CHECK-NEXT: {{  $}}
+  ; CHECK-NEXT: bb.5:
+  ; CHECK-NEXT:   successors: %bb.8(0x80000000)
+  ; CHECK-NEXT: {{  $}}
+  ; CHECK-NEXT:   B %bb.8
+  ; CHECK-NEXT: {{  $}}
+  ; CHECK-NEXT: bb.6:
+  ; CHECK-NEXT:   successors: %bb.2(0x80000000)
+  ; CHECK-NEXT: {{  $}}
+  ; CHECK-NEXT:   STRWui [[DEF9]], [[DEF5]], 0 :: (store (s32))
+  ; CHECK-NEXT:   B %bb.2
+  ; CHECK-NEXT: {{  $}}
+  ; CHECK-NEXT: bb.7:
+  ; CHECK-NEXT:   successors: %bb.2(0x80000000)
+  ; CHECK-NEXT: {{  $}}
+  ; CHECK-NEXT:   STRWui $wzr, [[DEF]], 0 :: (store (s32))
+  ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def dead $sp, implicit $sp
+  ; CHECK-NEXT:   $w0 = COPY [[DEF4]]
+  ; CHECK-NEXT:   $x1 = COPY [[DEF1]]
+  ; CHECK-NEXT:   BL 0, csr_darwin_aarch64_aapcs, implicit-def dead $lr, implicit $sp, implicit-def $sp, implicit-def dead $w0
+  ; CHECK-NEXT:   ADJCALLSTACKUP 0, 0, implicit-def dead $sp, implicit $sp
+  ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def dead $sp, implicit $sp
+  ; CHECK-NEXT:   $w0 = COPY [[DEF6]]
+  ; CHECK-NEXT:   ADJCALLSTACKUP 0, 0, implicit-def dead $sp, implicit $sp
+  ; CHECK-NEXT:   B %bb.2
+  ; CHECK-NEXT: {{  $}}
+  ; CHECK-NEXT: bb.8:
+  ; CHECK-NEXT:   successors: %bb.8(0x80000000)
+  ; CHECK-NEXT: {{  $}}
+  ; CHECK-NEXT:   B %bb.8
+  ; CHECK-NEXT: {{  $}}
+  ; CHECK-NEXT: bb.9:
+  ; CHECK-NEXT:   successors: %bb.10(0x40000000), %bb.1(0x40000000)
+  ; CHECK-NEXT: {{  $}}
+  ; CHECK-NEXT:   undef [[DEF12:%[0-9]+]].sub_32:gpr64 = IMPLICIT_DEF
+  ; CHECK-NEXT:   STRXui [[DEF12]], %stack.0, 0 :: (store (s64) into %stack.0)
+  ; CHECK-NEXT:   TBZW [[DEF3]], 0, %bb.1
+  ; CHECK-NEXT:   B %bb.10
+  ; CHECK-NEXT: {{  $}}
+  ; CHECK-NEXT: bb.10:
+  ; CHECK-NEXT:   successors: %bb.1(0x80000000)
+  ; CHECK-NEXT: {{  $}}
+  ; CHECK-NEXT:   ADJCALLSTACKDOWN 32, 0, implicit-def dead $sp, implicit $sp
+  ; CHECK-NEXT:   ADJCALLSTACKUP 32, 0, implicit-def dead $sp, implicit $sp
+  ; CHECK-NEXT:   B %bb.1
+  ; CHECK-NEXT: {{  $}}
+  ; CHECK-NEXT: bb.11:
+  ; CHECK-NEXT:   [[LDRXui:%[0-9]+]]:gpr64 = LDRXui %stack.0, 0 :: (load (s64) from %stack.0)
+  ; CHECK-NEXT:   dead undef [[COPY:%[0-9]+]].sub_32:gpr64 = COPY [[LDRXui]].sub_32
+  ; CHECK-NEXT:   ADJCALLSTACKDOWN 8, 0, implicit-def dead $sp, implicit $sp
+  ; CHECK-NEXT:   ADJCALLSTACKUP 8, 0, implicit-def dead $sp, implicit $sp
+  bb.0:
+    liveins: $w0, $w1, $x2, $x3, $x4, $w5, $w6
+
+    %0:gpr64common = IMPLICIT_DEF
+    %1:gpr64 = IMPLICIT_DEF
+    %2:gpr32common = IMPLICIT_DEF
+    %3:gpr32 = IMPLICIT_DEF
+    %4:gpr32 = IMPLICIT_DEF
+    %5:gpr64common = IMPLICIT_DEF
+    %6:gpr32 = IMPLICIT_DEF
+    undef %7.sub_32:gpr64 = IMPLICIT_DEF
+    %8:gpr64common = IMPLICIT_DEF
+    %9:gpr32 = IMPLICIT_DEF
+    %10:gpr64 = IMPLICIT_DEF
+    undef %10.sub_32:gpr64 = IMPLICIT_DEF implicit-def %11:gpr64
+
+  bb.1:
+
+  bb.2:
+    successors: %bb.3(0x0fbefbf0), %bb.4(0x70410410)
+
+    Bcc 8, %bb.3, implicit killed undef $nzcv
+    B %bb.4
+
+  bb.3:
+    successors: %bb.11(0x00000000), %bb.2(0x80000000)
+
+    dead $wzr = SUBSWri %2, 64, 0, implicit-def $nzcv
+    Bcc 0, %bb.11, implicit killed undef $nzcv
+    B %bb.2
+
+  bb.4:
+    successors: %bb.9(0x01288b01), %bb.5(0x01288b01), %bb.2(0x11f46a91), %bb.6(0x23e8d524), %bb.7(0x47d1aa49)
+
+    early-clobber %12:gpr64, dead early-clobber %13:gpr64sp = JumpTableDest32 %8, %7, %jump-table.0
+    BR undef %10
+
+  bb.5:
+    B %bb.8
+
+  bb.6:
+    STRWui %9, %5, 0 :: (store (s32))
+    B %bb.2
+
+  bb.7:
+    STRWui $wzr, %0, 0 :: (store (s32))
+    ADJCALLSTACKDOWN 0, 0, implicit-def dead $sp, implicit $sp
+    $w0 = COPY %4
+    $x1 = COPY %1
+    BL 0, csr_darwin_aarch64_aapcs, implicit-def dead $lr, implicit $sp, implicit-def $sp, implicit-def dead $w0
+    ADJCALLSTACKUP 0, 0, implicit-def dead $sp, implicit $sp
+    ADJCALLSTACKDOWN 0, 0, implicit-def dead $sp, implicit $sp
+    $w0 = COPY %6
+    ADJCALLSTACKUP 0, 0, implicit-def dead $sp, implicit $sp
+    B %bb.2
+
+  bb.8:
+    B %bb.8
+
+  bb.9:
+    successors: %bb.10, %bb.1
+
+    undef %10.sub_32:gpr64 = IMPLICIT_DEF
+    TBZW %3, 0, %bb.1
+    B %bb.10
+
+  bb.10:
+    ADJCALLSTACKDOWN 32, 0, implicit-def dead $sp, implicit $sp
+    ADJCALLSTACKUP 32, 0, implicit-def dead $sp, implicit $sp
+    B %bb.1
+
+  bb.11:
+    undef %14.sub_32:gpr64 = COPY %10.sub_32
+    ADJCALLSTACKDOWN 8, 0, implicit-def dead $sp, implicit $sp
+    ADJCALLSTACKUP 8, 0, implicit-def dead $sp, implicit $sp
+
+...
diff --git a/llvm/test/CodeGen/X86/inline-spiller-impdef-on-implicit-def-regression.ll b/llvm/test/CodeGen/X86/inline-spiller-impdef-on-implicit-def-regression.ll
new file mode 100644
index 000000000000000..1449f9a700965a0
--- /dev/null
+++ b/llvm/test/CodeGen/X86/inline-spiller-impdef-on-implicit-def-regression.ll
@@ -0,0 +1,181 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 3
+; RUN: llc -mtriple=x86_64-unknown-linux-gnu < %s | FileCheck %s
+
+; Make sure there's no assert on an implicit-def with implicit operands
+; during register allocation.
+
+%struct.BlockContext = type { [32 x i8], [32 x i8], [2 x [32 x i8]], [32 x i8], [32 x i8], [32 x i8], [32 x i8], [32 x i8], [2 x [32 x i8]], [2 x [32 x i8]], [32 x i8], [32 x i8], [32 x i8], [32 x i8], [16 x i8], [32 x i8], [32 x i8] }
+%struct.CdfModeContext = type { [4 x [16 x i16]], [2 x [13 x [16 x i16]]], [9 x [16 x i16]], [5 x [4 x [16 x i16]]], [6 x [16 x i16]], [2 x [16 x i16]], [16 x i16], [2 x [13 x [8 x i16]]], [3 x [13 x [8 x i16]]], [8 x i16], [8 x [8 x i16]], [8 x i16], [8 x [8 x i16]], [3 x [8 x i16]], [2 x [7 x [8 x i16]]], [2 x [7 x [5 x [8 x i16]]]], [2 x [8 x [4 x i16]]], [4 x [3 x [4 x i16]]], [22 x [4 x i16]], [4 x i16], [5 x [4 x i16]], [4 x [4 x i16]], [4 x i16], [2 x i16], [2 x i16], [7 x [2 x i16]], [7 x [2 x i16]], [4 x [2 x i16]], [22 x [2 x i16]], [6 x [2 x i16]], [2 x [2 x i16]], [6 x [2 x i16]], [3 x [2 x i16]], [4 x [2 x i16]], [5 x [2 x i16]], [5 x [2 x i16]], [6 x [2 x i16]], [6 x [2 x i16]], [9 x [2 x i16]], [6 x [3 x [2 x i16]]], [3 x [3 x [2 x i16]]], [2 x [3 x [2 x i16]]], [3 x [3 x [2 x i16]]], [7 x [3 x [2 x i16]]], [3 x [2 x i16]], [3 x [2 x i16]], [3 x [2 x i16]], [22 x [2 x i16]], [7 x [3 x [2 x i16]]], [2 x [2 x i16]], [2 x i16], [8 x i8] }
+
+define i32 @decode_sb(ptr %t, i32 %bl, i32 %_msprop1966, i32 %sub.i, i64 %idxprom, i1 %cmp54) #0 {
+; CHECK-LABEL: decode_sb:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    pushq %rbp
+; CHECK-NEXT:    .cfi_def_cfa_offset 16
+; CHECK-NEXT:    .cfi_offset %rbp, -16
+; CHECK-NEXT:    movq %rsp, %rbp
+; CHECK-NEXT:    .cfi_def_cfa_register %rbp
+; CHECK-NEXT:    pushq %r15
+; CHECK-NEXT:    pushq %r14
+; CHECK-NEXT:    pushq %r13
+; CHECK-NEXT:    pushq %r12
+; CHECK-NEXT:    pushq %rbx
+; CHECK-NEXT:    subq $24, %rsp
+; CHECK-NEXT:    .cfi_offset %rbx, -56
+; CHECK-NEXT:    .cfi_offset %r12, -48
+; CHECK-NEXT:    .cfi_offset %r13, -40
+; CHECK-NEXT:    .cfi_offset %r14, -32
+; CHECK-NEXT:    .cfi_offset %r15, -24
+; CHECK-NEXT:    movl %r9d, %ebx
+; CHECK-NEXT:    movabsq $87960930222080, %r15 # imm = 0x500000000000
+; CHECK-NEXT:    movl 0, %r13d
+; CHECK-NEXT:    movl %esi, %r12d
+; CHECK-NEXT:    # implicit-def: $eax
+; CHECK-NEXT:    movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; CHECK-NEXT:    testb $1, %bl
+; CHECK-NEXT:    jne .LBB0_9
+; CHECK-NEXT:  # %bb.1: # %if.else
+; CHECK-NEXT:    movq %r8, %r14
+; CHECK-NEXT:    movl %ecx, %eax
+; CHECK-NEXT:    andl $1, %eax
+; CHECK-NEXT:    movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; CHECK-NEXT:    movzbl 544(%rax), %eax
+; CHECK-NEXT:    andl $1, %eax
+; CHECK-NEXT:    movl %r15d, %r9d
+; CHECK-NEXT:    andl $1, %r9d
+; CHECK-NEXT:    movl %r14d, %r10d
+; CHECK-NEXT:    andl $1, %r10d
+; CHECK-NEXT:    movl %esi, %r11d
+; CHECK-NEXT:    # kill: def $cl killed $cl killed $ecx
+; CHECK-NEXT:    shrl %cl, %r11d
+; CHECK-NEXT:    movabsq $17592186044416, %r8 # imm = 0x100000000000
+; CHECK-NEXT:    orq %r10, %r8
+; CHECK-NEXT:    andl $2, %r11d
+; CHECK-NEXT:    testb $1, %bl
+; CHECK-NEXT:    cmoveq %r9, %r8
+; CHECK-NEXT:    movl %edx, %ecx
+; CHECK-NEXT:    orq %rax, %rcx
+; CHECK-NEXT:    movq %r13, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; CHECK-NEXT:    orq $1, %r13
+; CHECK-NEXT:    orl %esi, %r11d
+; CHECK-NEXT:    movl $1, %edx
+; CHECK-NEXT:    je .LBB0_3
+; CHECK-NEXT:  # %bb.2: # %if.else
+; CHECK-NEXT:    movl (%r8), %edx
+; CHECK-NEXT:  .LBB0_3: # %if.else
+; CHECK-NEXT:    shlq $5, %rcx
+; CHECK-NEXT:    movq %r12, %rsi
+; CHECK-NEXT:    shlq $7, %rsi
+; CHECK-NEXT:    addq %rcx, %rsi
+; CHECK-NEXT:    addq $1248, %rsi # imm = 0x4E0
+; CHECK-NEXT:    movq %r13, 0
+; CHECK-NEXT:    movq %rdi, %r15
+; CHECK-NEXT:    movl %edx, (%rdi)
+; CHECK-NEXT:    xorl %eax, %eax
+; CHECK-NEXT:    xorl %edi, %edi
+; CHECK-NEXT:    xorl %edx, %edx
+; CHECK-NEXT:    callq *%rax
+; CHECK-NEXT:    xorq $1, %r14
+; CHECK-NEXT:    cmpl $0, (%r14)
+; CHECK-NEXT:    je .LBB0_6
+; CHECK-NEXT:  # %bb.4: # %if.else
+; CHECK-NEXT:    movb $1, %al
+; CHECK-NEXT:    testb %al, %al
+; CHECK-NEXT:    je .LBB0_5
+; CHECK-NEXT:  .LBB0_6: # %bb19
+; CHECK-NEXT:    testb $1, %bl
+; CHECK-NEXT:    movq %r15, %rdi
+; CHECK-NEXT:    movabsq $87960930222080, %r15 # imm = 0x500000000000
+; CHECK-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %r13 # 8-byte Reload
+; CHECK-NEXT:    je .LBB0_9
+; CHECK-NEXT:  # %bb.7: # %land.lhs.true56
+; CHECK-NEXT:    xorl %eax, %eax
+; CHECK-NEXT:    jmp .LBB0_8
+; CHECK-NEXT:  .LBB0_9: # %if.end69
+; CHECK-NEXT:    movl %r13d, 0
+; CHECK-NEXT:    xorl %eax, %eax
+; CHECK-NEXT:    xorl %esi, %esi
+; CHECK-NEXT:    xorl %edx, %edx
+; CHECK-NEXT:    xorl %ecx, %ecx
+; CHECK-NEXT:    xorl %r8d, %r8d
+; CHECK-NEXT:    callq *%rax
+; CHECK-NEXT:    xorq %r15, %r12
+; CHECK-NEXT:    movslq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 4-byte Folded Reload
+; CHECK-NEXT:    movzbl (%r12), %ecx
+; CHECK-NEXT:    movb %cl, 544(%rax)
+; CHECK-NEXT:    xorl %eax, %eax
+; CHECK-NEXT:  .LBB0_8: # %land.lhs.true56
+; CHECK-NEXT:    addq $24, %rsp
+; CHECK-NEXT:    popq %rbx
+; CHECK-NEXT:    popq %r12
+; CHECK-NEXT:    popq %r13
+; CHECK-NEXT:    popq %r14
+; CHECK-NEXT:    popq %r15
+; CHECK-NEXT:    popq %rbp
+; CHECK-NEXT:    .cfi_def_cfa %rsp, 8
+; CHECK-NEXT:    retq
+; CHECK-NEXT:  .LBB0_5: # %bb
+entry:
+  %i = load i32, ptr null, align 8
+  br i1 %cmp54, label %if.end69, label %if.else
+
+if.else:                                          ; preds = %entry
+  %shr18 = and i32 %sub.i, 1
+  %idxprom.i = zext i32 %shr18 to i64
+  %arrayidx.i = getelementptr %struct.BlockContext, ptr null, i64 0, i32 14, i64 %idxprom.i
+  %i1 = load i8, ptr %arrayidx.i, align 1
+  %conv.i = zext i8 %i1 to i32
+  %and.i = and i32 %conv.i, 1
+  %i2 = and i64 87960930222080, 1
+  %i3 = inttoptr i64 %i2 to ptr
+  %i4 = load i32, ptr %i3, align 4
+  %i5 = and i64 %idxprom, 1
+  %i6 = or i64 %i5, 17592186044416
+  %i7 = inttoptr i64 %i6 to ptr
+  %i8 = load i32, ptr %i7, align 4
+  %i9 = lshr i32 %bl, %sub.i
+  %i10 = and i32 %i9, 2
+  %i11 = or i32 %bl, %i10
+  %i12 = select i1 %cmp54, i32 %i8, i32 %i4
+  %add.i = or i32 %_msprop1966, %and.i
+  %idxprom4 = zext i32 %bl to i64
+  %idxprom24 = zext i32 %add.i to i64
+  %i13 = or i32 %i, 1
+  %i14 = zext i32 %i13 to i64
+  %.not2329 = icmp eq i32 %i11, 0
+  %i15 = select i1 %.not2329, i32 1, i32 %i12
+  %arrayidx25 = getelementptr %struct.CdfModeContext, ptr null, i64 0, i32 3, i64 %idxprom4, i64 %idxprom24
+  store i64 %i14, ptr null, align 8
+  store i32 %i15, ptr %t, align 4
+  %call53 = tail call i32 null(ptr null, ptr %arrayidx25, i64 0)
+  %i16 = xor i64 %idxprom, 1
+  %i17 = inttoptr i64 %i16 to ptr
+  %_msld1992 = load i32, ptr %i17, align 8
+  %i18 = icmp ne i32 %_msld1992, 0
+  %_msprop_icmp1993 = and i1 %i18, false
+  br i1 %_msprop_icmp1993, label %bb, label %bb19
+
+bb:                                               ; preds = %if.else
+  unreachable
+
+bb19:                                             ; preds = %if.else
+  br i1 %cmp54, label %land.lhs.true56, label %if.end69
+
+land.lhs.true56:                                  ; preds = %bb19
+  ret i32 0
+
+if.end69:                                         ; preds = %bb19, %entry
+  %bx8.011941201 = phi i32 [ %shr18, %bb19 ], [ undef, %entry ]
+  store i32 %i, ptr null, align 8
+  %call79 = tail call fastcc i32 null(ptr %t, i32 0, i32 0, i32 0, i32 0)
+  %idxprom666 = zext i32 %bl to i64
+  %i20 = xor i64 %idxprom666, 87960930222080
+  %idxprom675 = sext i32 %bx8.011941201 to i64
+  %arrayidx676 = getelementptr %struct.BlockContext, ptr null, i64 0, i32 14, i64 %idxprom675
+  %i21 = inttoptr i64 %i20 to ptr
+  %_msld1414 = load i8, ptr %i21, align 1
+  store i8 %_msld1414, ptr %arrayidx676, align 1
+  ret i32 0
+}
+
+attributes #0 = { "frame-pointer"="all" "target-cpu"="x86-64" }

@@ -1071,8 +1071,7 @@ void InlineSpiller::insertReload(Register NewVReg,
static bool isRealSpill(const MachineInstr &Def) {
if (!Def.isImplicitDef())
return true;
assert(Def.getNumOperands() == 1 &&
"Implicit def with more than one definition");

Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

If we really want to keep an assert here, we could check that we have only one explicit operand.
Don't think that's worth it though.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The verifier handles that case

It's not a verifier enforced property that implicit_def may only have one
operand. Fixes assertions after the coalescer implicit-defs to preserve
super register liveness to arbitrary instructions.

For some reason I'm unable to reproduce this as a MIR test running
only the allocator for the x86 test. Not sure it's worth keeping around.
@arsenm arsenm force-pushed the inline-spiller-delete-impdef-assert branch from 7941e06 to 6cf66ce Compare October 19, 2023 01:36
@arsenm arsenm merged commit 3e49ce6 into llvm:main Oct 19, 2023
3 checks passed
@arsenm arsenm deleted the inline-spiller-delete-impdef-assert branch October 19, 2023 15:51
Guzhu-AMD pushed a commit to GPUOpen-Drivers/llvm-project that referenced this pull request Oct 26, 2023
Local branch amd-gfx 10c5c64 Merged main:25002b7aeb2f into amd-gfx:8ae830866bb0
Remote branch main 3e49ce6 InlineSpiller: Delete assert that implicit_def has no implicit operands (llvm#69087)
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants