Change the testcase tail-merge-after-mbp.ll to tail-merge-after-mbp.mir

[haicheng]

Summary

I found tail-merge-after-mbp.ll as a .ll test was fragile to the changes in preivous pipeline, so I like to change it to a .mir test and run it with only block-placement pass.

[gberry]

I believe a lot of the IR in this test case is superfluous (as well as some of the MIR property flags like hasStackMap, etc.). Could you try cutting it down to only the necessary IR and MIR to make it more robust against future changes?
I've added Matthias as a reviewer since he is more familiar with the current state of the MIR parser.

[gberry]

Sorry if I wasn't clear. I'm not suggesting you change what is being tested, just that you remove code from the test that doesn't impact it. For example, I believe you can get rid of all of the IR code if you also remove the MemoryOperands info from the MIR (the :: (load 8 from xxx) stuff).

[haicheng]

Block 3 4 7 9 in IR are not necessary. The generated MIR can be manually reduced as Geoff said. If you can wait till early next week, I can simplify this test case for you.

[wmi]

Remove some code from the test that doesn't impact it.

[MatzeB]

The question of reducing .mir files keeps popping up, I will add some tips and tricks to the llvm documentation.

[MatzeB]

In D32037#726609, @MatzeB wrote:

The question of reducing .mir files keeps popping up, I will add some tips and tricks to the llvm documentation.

I wrote some paragraphs about simplifying .mir tests: https://reviews.llvm.org/D32058

(We really should have tools for some things like dropping IR references, but until we do have tools there is at least a description of the process now).

[haicheng]

Simplified the mir test case a little bit.

[MatzeB]

This looks like you can remove the whole IR block (--- | to ... included).

In any case LGTM, thanks.

[haicheng]

Remove the whole IR module.

[wmi]

Haicheng, the testcase LGTM. Thanks for working on it!

test/CodeGen/X86/tail-merge-after-mbp.ll

-; RUN: llc -mtriple=x86_64-linux -o - %s | FileCheck %s
-
-%0 = type { %1, %3* }
-%1 = type { %2* }
-%2 = type { %2*, i8* }
-%3 = type { i32, i32 (i32, i32)* }
-
-
-declare i32 @Up(...) 
-declare i32 @f(i32, i32) 
-
-; check loop block_14 is not merged with block_21
-; check loop block_11 is not merged with block_18, block_25
-define i32 @foo(%0* nocapture readonly, i32, i1 %c, i8* %p1, %2** %p2) {
-; CHECK-LABEL: foo:
-; CHECK:     # %block_11
-; CHECK-NEXT:        movq    (%r14), %rax
-; CHECK-NEXT:        testq   %rax, %rax
-; CHECK-NEXT:        je      
-; CHECK-NEXT:# %block_14
-; CHECK-NEXT:        cmpq    $0, 8(%rax)
-; CHECK-NEXT:        jne    
-; CHECK-NEXT:# %block_18
-; CHECK-NEXT:        movq    (%r14), %rax
-; CHECK-NEXT:        testq   %rax, %rax
-; CHECK-NEXT:        je    
-; CHECK-NEXT:# %block_21
-; CHECK-NEXT:# =>This Inner Loop Header
-; CHECK-NEXT:        cmpq    $0, 8(%rax)
-; CHECK-NEXT:        jne  
-; CHECK-NEXT:# %block_25
-; CHECK-NEXT:#   in Loop
-; CHECK-NEXT:        movq    (%r14), %rax
-; CHECK-NEXT:        testq   %rax, %rax
-; CHECK-NEXT:        jne 
-  br i1 %c, label %block_34, label %block_3
-
-block_3:                                      ; preds = %2
-  br i1 %c, label %block_7, label %block_4
-
-block_4:                                      ; preds = %block_3
-  %a5 = tail call i32 @f(i32 undef, i32 undef)
-  %a6 = icmp eq i32 %a5, 0
-  br i1 %a6, label %block_7, label %block_34
-
-block_7:                                      ; preds = %block_4, %block_3
-  %a8 = icmp eq %2* null, null
-  br i1 %a8, label %block_34, label %block_9
-
-block_9:                                      ; preds = %block_7
-  %a10 = icmp eq i8* %p1, null
-  br i1 %a10, label %block_11, label %block_32
-
-block_11:                                     ; preds = %block_9
-  %a12 = load %2*, %2** %p2, align 8
-  %a13 = icmp eq %2* %a12, null
-  br i1 %a13, label %block_34, label %block_14
-
-block_14:                                     ; preds = %block_11
-  %a15 = getelementptr inbounds %2, %2* %a12, i64 0, i32 1
-  %a16 = load i8*, i8** %a15, align 8
-  %a17 = icmp eq i8* %a16, null
-  br i1 %a17, label %block_18, label %block_32
-
-block_18:                                     ; preds = %block_14
-  %a19 = load %2*, %2** %p2, align 8
-  %a20 = icmp eq %2* %a19, null
-  br i1 %a20, label %block_34, label %block_21
-
-block_21:                                     ; preds = %block_18
-  %a22 = getelementptr inbounds %2, %2* %a19, i64 0, i32 1
-  %a23 = load i8*, i8** %a22, align 8
-  %a24 = icmp eq i8* %a23, null
-  br i1 %a24, label %block_25, label %block_32
-
-block_25:                                     ; preds = %block_28, %block_21
-  %a26 = load %2*, %2** %p2, align 8
-  %a27 = icmp eq %2* %a26, null
-  br i1 %a27, label %block_34, label %block_28
-
-block_28:                                     ; preds = %block_25
-  %a29 = getelementptr inbounds %2, %2* %a26, i64 0, i32 1
-  %a30 = load i8*, i8** %a29, align 8
-  %a31 = icmp eq i8* %a30, null
-  br i1 %a31, label %block_25, label %block_32
-
-block_32:                                     ; preds = %block_28, %block_21, %block_14, %block_9
-  %a33 = tail call i32 (...) @Up()
-  br label %block_34
-
-block_34:                                     ; preds = %block_32, %block_25, %block_18, %block_11, %block_7, %block_4, %2
-  %a35 = phi i32 [ 0, %2 ], [ %a5, %block_4 ], [ 0, %block_7 ], [ 0, %block_11 ], [ 0, %block_32 ], [ 0, %block_18 ], [ 0, %block_25 ]
-  ret i32 %a35
-}

test/CodeGen/X86/tail-merge-after-mbp.mir

+# RUN: llc -mtriple=x86_64-linux -run-pass=block-placement -o - %s | FileCheck %s
+
+--- |
+  target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
+  target triple = "x86_64--linux"
+  
+  %0 = type { %1, %3* }
+  %1 = type { %2* }
+  %2 = type { %2*, i8* }
+  %3 = type { i32, i32 (i32, i32)* }
+  
+  declare i32 @Up(...)
+  
+  declare i32 @f(i32, i32)
+  
+  define i32 @foo(%0* nocapture readonly, i32, i1 %c, i8* %p1, %2** %p2) {
+    br i1 %c, label %block_34, label %block_3
+  
+  block_3:                                          ; preds = %2
+    br i1 %c, label %block_7, label %block_4
+  
+  block_4:                                          ; preds = %block_3
+    %a5 = tail call i32 @f(i32 undef, i32 undef)
+    %a6 = icmp eq i32 %a5, 0
+    br i1 %a6, label %block_7, label %block_34
+  
+  block_7:                                          ; preds = %block_4, %block_3
+    %a8 = icmp eq %2* null, null
+    br i1 %a8, label %block_34, label %block_9
+  
+  block_9:                                          ; preds = %block_7
+    %a10 = icmp eq i8* %p1, null
+    br i1 %a10, label %block_11, label %block_32
+  
+  block_11:                                         ; preds = %block_9
+    %a12 = load %2*, %2** %p2, align 8
+    %a13 = icmp eq %2* %a12, null
+    br i1 %a13, label %block_34, label %block_14
+  
+  block_14:                                         ; preds = %block_11
+    %a15 = getelementptr inbounds %2, %2* %a12, i64 0, i32 1
+    %a16 = load i8*, i8** %a15, align 8
+    %a17 = icmp eq i8* %a16, null
+    br i1 %a17, label %block_18, label %block_32
+  
+  block_18:                                         ; preds = %block_14
+    %a19 = load %2*, %2** %p2, align 8
+    %a20 = icmp eq %2* %a19, null
+    br i1 %a20, label %block_34, label %block_21
+  
+  block_21:                                         ; preds = %block_18
+    %a22 = getelementptr inbounds %2, %2* %a19, i64 0, i32 1
+    %a23 = load i8*, i8** %a22, align 8
+    %a24 = icmp eq i8* %a23, null
+    br i1 %a24, label %block_25.preheader, label %block_32
+  
+  block_25.preheader:                               ; preds = %block_21
+    br label %block_25
+  
+  block_25:                                         ; preds = %block_25.preheader, %block_28
+    %a26 = load %2*, %2** %p2, align 8
+    %a27 = icmp eq %2* %a26, null
+    br i1 %a27, label %block_34, label %block_28
+  
+  block_28:                                         ; preds = %block_25
+    %a29 = getelementptr inbounds %2, %2* %a26, i64 0, i32 1
+    %a30 = load i8*, i8** %a29, align 8
+    %a31 = icmp eq i8* %a30, null
+    br i1 %a31, label %block_25, label %block_32
+  
+  block_32:                                         ; preds = %block_28, %block_21, %block_14, %block_9
+    %a33 = tail call i32 (...) @Up()
+    br label %block_34
+  
+  block_34:                                         ; preds = %block_25, %block_32, %block_18, %block_11, %block_7, %block_4, %2
+    %a35 = phi i32 [ 0, %2 ], [ %a5, %block_4 ], [ 0, %block_7 ], [ 0, %block_11 ], [ 0, %block_32 ], [ 0, %block_18 ], [ 0, %block_25 ]
+    ret i32 %a35
+  }
+  
+  ; Function Attrs: nounwind
+  declare void @llvm.stackprotector(i8*, i8**) #0
+  
+  attributes #0 = { nounwind }
+
+...
+---
+# CHECK: name:            foo
+name:            foo
+alignment:       4
+tracksRegLiveness: true
+liveins:         
+  - { reg: '%edx' }
+  - { reg: '%rcx' }
+  - { reg: '%r8' }
+fixedStack:      
+  - { id: 0, type: spill-slot, offset: -32, size: 8, alignment: 16, callee-saved-register: '%rbx' }
+  - { id: 1, type: spill-slot, offset: -24, size: 8, alignment: 8, callee-saved-register: '%r14' }
+  - { id: 2, type: spill-slot, offset: -16, size: 8, alignment: 16, callee-saved-register: '%rbp' }
+
+# check loop block_14 is not merged with block_21
+# check loop block_11 is not merged with block_18, block_25
+#
+# CHECK: bb.9.block_11:
+# CHECK: MOV64rm
+# CHECK-NEXT: TEST64rr
+# CHECK-NEXT: JE_1
+# CHECK: bb.10.block_14:
+# CHECK: CMP64mi8
+# CHECK-NEXT: JNE_1
+# CHECK: bb.11.block_18:
+# CHECK: MOV64rm
+# CHECK-NEXT: TEST64rr
+# CHECK-NEXT: JE_1
+# CHECK: bb.12.block_21:
+# CHECK: CMP64mi8
+# CHECK-NEXT: JNE_1
+# CHECK: bb.13.block_25:
+# CHECK: MOV64rm
+# CHECK-NEXT: TEST64rr
+# CHECK-NEXT: JNE_1
+
+body:             |
+  bb.0 (%ir-block.2):
+    successors: %bb.1(0x40000000), %bb.2.block_3(0x40000000)
+    liveins: %edx, %rcx, %r8, %rbp, %r14, %rbx
+  
+    frame-setup PUSH64r killed %rbp, implicit-def %rsp, implicit %rsp
+    frame-setup PUSH64r killed %r14, implicit-def %rsp, implicit %rsp
+    frame-setup PUSH64r killed %rbx, implicit-def %rsp, implicit %rsp
+    %r14 = MOV64rr %r8
+    %rbx = MOV64rr %rcx
+    TEST8ri %dl, 1, implicit-def %eflags, implicit killed %edx
+    JE_1 %bb.2.block_3, implicit %eflags
+  
+  bb.1:
+    successors: %bb.16.block_34(0x80000000)
+  
+    %ebp = XOR32rr undef %ebp, undef %ebp, implicit-def dead %eflags
+    JMP_1 %bb.16.block_34
+  
+  bb.2.block_3:
+    successors: %bb.4.block_7(0x40000000), %bb.3.block_4(0x40000000)
+    liveins: %eflags, %rbx, %r14
+  
+    JNE_1 %bb.4.block_7, implicit %eflags
+  
+  bb.3.block_4:
+    successors: %bb.4.block_7(0x30000000), %bb.16.block_34(0x50000000)
+    liveins: %rbx, %r14
+  
+    CALL64pcrel32 @f, csr_64, implicit %rsp, implicit undef %edi, implicit undef %esi, implicit-def %rsp, implicit-def %eax
+    %ebp = MOV32rr %eax
+    TEST32rr %ebp, %ebp, implicit-def %eflags
+    JNE_1 %bb.16.block_34, implicit %eflags
+  
+  bb.4.block_7:
+    successors: %bb.5(0x30000000), %bb.6.block_9(0x50000000)
+    liveins: %rbx, %r14
+  
+    %al = MOV8ri 1
+    TEST8rr killed %al, %al, implicit-def %eflags
+    JE_1 %bb.6.block_9, implicit killed %eflags
+  
+  bb.5:
+    successors: %bb.16.block_34(0x80000000)
+  
+    %ebp = XOR32rr undef %ebp, undef %ebp, implicit-def dead %eflags
+    JMP_1 %bb.16.block_34
+  
+  bb.6.block_9:
+    successors: %bb.7.block_11(0x30000000), %bb.15.block_32(0x50000000)
+    liveins: %rbx, %r14
+  
+    TEST64rr killed %rbx, %rbx, implicit-def %eflags
+    JNE_1 %bb.15.block_32, implicit %eflags
+  
+  bb.7.block_11:
+    successors: %bb.8(0x30000000), %bb.9.block_14(0x50000000)
+    liveins: %r14
+  
+    %rax = MOV64rm %r14, 1, _, 0, _
+    TEST64rr %rax, %rax, implicit-def %eflags
+    JNE_1 %bb.9.block_14, implicit killed %eflags
+  
+  bb.8:
+    successors: %bb.16.block_34(0x80000000)
+  
+    %ebp = XOR32rr undef %ebp, undef %ebp, implicit-def dead %eflags
+    JMP_1 %bb.16.block_34
+  
+  bb.9.block_14:
+    successors: %bb.10.block_18(0x30000000), %bb.15.block_32(0x50000000)
+    liveins: %rax, %r14
+  
+    CMP64mi8 killed %rax, 1, _, 8, _, 0, implicit-def %eflags
+    JNE_1 %bb.15.block_32, implicit %eflags
+  
+  bb.10.block_18:
+    successors: %bb.11(0x30000000), %bb.12.block_21(0x50000000)
+    liveins: %r14
+  
+    %rax = MOV64rm %r14, 1, _, 0, _
+    TEST64rr %rax, %rax, implicit-def %eflags
+    JNE_1 %bb.12.block_21, implicit %eflags
+  
+  bb.11:
+    successors: %bb.16.block_34(0x80000000)
+  
+    %ebp = XOR32rr undef %ebp, undef %ebp, implicit-def dead %eflags
+    JMP_1 %bb.16.block_34
+  
+  bb.12.block_21:
+    successors: %bb.13.block_25(0x71555555), %bb.15.block_32(0x0eaaaaab)
+    liveins: %r14, %rax
+  
+    CMP64mi8 killed %rax, 1, _, 8, _, 0, implicit-def %eflags
+    JNE_1 %bb.15.block_32, implicit %eflags
+  
+  bb.13.block_25:
+    successors: %bb.14(0x04000000), %bb.12.block_21(0x7c000000)
+    liveins: %r14
+  
+    %rax = MOV64rm %r14, 1, _, 0, _
+    TEST64rr %rax, %rax, implicit-def %eflags
+    JNE_1 %bb.12.block_21, implicit %eflags
+  
+  bb.14:
+    successors: %bb.16.block_34(0x80000000)
+  
+    %ebp = XOR32rr undef %ebp, undef %ebp, implicit-def dead %eflags
+    JMP_1 %bb.16.block_34
+  
+  bb.15.block_32:
+    successors: %bb.16.block_34(0x80000000)
+  
+    %ebp = XOR32rr undef %ebp, undef %ebp, implicit-def dead %eflags
+    dead %eax = XOR32rr undef %eax, undef %eax, implicit-def dead %eflags, implicit-def %al
+    CALL64pcrel32 @Up, csr_64, implicit %rsp, implicit %al, implicit-def %rsp, implicit-def dead %eax
+  
+  bb.16.block_34:
+    liveins: %ebp
+  
+    %eax = MOV32rr killed %ebp
+    %rbx = POP64r implicit-def %rsp, implicit %rsp
+    %r14 = POP64r implicit-def %rsp, implicit %rsp
+    %rbp = POP64r implicit-def %rsp, implicit %rsp
+    RETQ %eax
+
+...