3ec2de2753
Reapply PR142507 with fix for test: add in the same x86_64-linux requirement as other tests as the stack ids are currently computed differently on big endian systems. This will be investigated separately. In order to allow selective reporting of context hinting during the LTO link, and in the future to allow selective more aggressive cloning, add an option to specify a minimum percent of the max cold size in the profile summary. Contexts that meet that threshold will get context size info metadata (and ThinLTO summary information) on the associated allocations. Specifying -memprof-report-hinted-sizes during the pre-LTO compile step will continue to cause all contexts to receive this metadata. But specifying -memprof-report-hinted-sizes only during the LTO link will cause only those that meet the new threshold and have the metadata to get reported. To support this, because the alloc info summary and associated bitcode requires the context size information to be in the same order as the other context information, 0s are inserted for contexts without this metadata. The bitcode writer uses a more compact format for the context ids to allow better compression of the 0s. As part of this change several helper methods are added to query whether metadata contains context size info on any or all contexts.
166 lines
8.4 KiB
Plaintext
166 lines
8.4 KiB
Plaintext
;; Test the -memprof-min-percent-max-cold-size flag for annotating only the
|
|
;; largest contexts with size info.
|
|
|
|
; REQUIRES: x86_64-linux
|
|
|
|
; RUN: split-file %s %t
|
|
|
|
;; Specify version 4 so that a summary is generated.
|
|
; RUN: llvm-profdata merge --memprof-version=4 %t/memprof_max_cold_threshold.yaml -o %t/memprof_max_cold_threshold.memprofdata
|
|
|
|
; RUN: llvm-profdata show --memory %t/memprof_max_cold_threshold.memprofdata | FileCheck %s --check-prefixes=SUMMARY
|
|
; SUMMARY: # MemProfSummary:
|
|
; SUMMARY: # Total contexts: 5
|
|
; SUMMARY: # Total cold contexts: 4
|
|
; SUMMARY: # Total hot contexts: 0
|
|
; SUMMARY: # Maximum cold context total size: 400
|
|
; SUMMARY: # Maximum warm context total size: 500
|
|
; SUMMARY: # Maximum hot context total size: 0
|
|
|
|
;; Test with various thresholds, starting with the baseline of no context size
|
|
;; info.
|
|
; RUN: opt < %t/memprof_max_cold_threshold.ll -passes='memprof-use<profile-filename=%t/memprof_max_cold_threshold.memprofdata>' -S | FileCheck %s --check-prefixes=ALL,NONE
|
|
; RUN: opt < %t/memprof_max_cold_threshold.ll -memprof-min-percent-max-cold-size=99 -passes='memprof-use<profile-filename=%t/memprof_max_cold_threshold.memprofdata>' -S | FileCheck %s --check-prefixes=ALL,THRESH99
|
|
; RUN: opt < %t/memprof_max_cold_threshold.ll -memprof-min-percent-max-cold-size=75 -passes='memprof-use<profile-filename=%t/memprof_max_cold_threshold.memprofdata>' -S | FileCheck %s --check-prefixes=ALL,THRESH75
|
|
; RUN: opt < %t/memprof_max_cold_threshold.ll -memprof-min-percent-max-cold-size=50 -passes='memprof-use<profile-filename=%t/memprof_max_cold_threshold.memprofdata>' -S | FileCheck %s --check-prefixes=ALL,THRESH50
|
|
|
|
;; Make sure serializing / deserializing through bitcode works
|
|
; RUN: opt < %t/memprof_max_cold_threshold.ll -memprof-min-percent-max-cold-size=75 -passes='memprof-use<profile-filename=%t/memprof_max_cold_threshold.memprofdata>' -o %t/out.bc
|
|
; RUN: llvm-dis %t/out.bc -o - | FileCheck %s --check-prefixes=ALL,THRESH75
|
|
|
|
;--- memprof_max_cold_threshold.yaml
|
|
---
|
|
HeapProfileRecords:
|
|
- GUID: _Z3foov
|
|
AllocSites:
|
|
- Callstack:
|
|
- { Function: _Z3foov, LineOffset: 0, Column: 22, IsInlineFrame: false }
|
|
- { Function: main, LineOffset: 2, Column: 5, IsInlineFrame: false }
|
|
MemInfoBlock:
|
|
# Smallest cold, 25% of largest cold size
|
|
TotalSize: 100
|
|
AllocCount: 1
|
|
TotalLifetimeAccessDensity: 1
|
|
TotalLifetime: 1000000
|
|
- Callstack:
|
|
- { Function: _Z3foov, LineOffset: 0, Column: 22, IsInlineFrame: false }
|
|
- { Function: main, LineOffset: 3, Column: 5, IsInlineFrame: false }
|
|
MemInfoBlock:
|
|
# Largest cold
|
|
TotalSize: 400
|
|
AllocCount: 1
|
|
TotalLifetimeAccessDensity: 1
|
|
TotalLifetime: 1000000
|
|
- Callstack:
|
|
- { Function: _Z3foov, LineOffset: 0, Column: 22, IsInlineFrame: false }
|
|
- { Function: main, LineOffset: 4, Column: 5, IsInlineFrame: false }
|
|
MemInfoBlock:
|
|
# Second largest cold, 75% of largest cold size
|
|
TotalSize: 300
|
|
AllocCount: 1
|
|
TotalLifetimeAccessDensity: 1
|
|
TotalLifetime: 1000000
|
|
- Callstack:
|
|
- { Function: _Z3foov, LineOffset: 0, Column: 22, IsInlineFrame: false }
|
|
- { Function: main, LineOffset: 5, Column: 5, IsInlineFrame: false }
|
|
MemInfoBlock:
|
|
# Second smallest cold, 50% of largest cold size
|
|
TotalSize: 200
|
|
AllocCount: 1
|
|
TotalLifetimeAccessDensity: 1
|
|
TotalLifetime: 1000000
|
|
- Callstack:
|
|
- { Function: _Z3foov, LineOffset: 0, Column: 22, IsInlineFrame: false }
|
|
- { Function: main, LineOffset: 6, Column: 5, IsInlineFrame: false }
|
|
MemInfoBlock:
|
|
# Largest context, which is non-cold (due to short lifetime)
|
|
TotalSize: 500
|
|
AllocCount: 1
|
|
TotalLifetimeAccessDensity: 1
|
|
TotalLifetime: 1
|
|
CallSites: []
|
|
...
|
|
;--- memprof_max_cold_threshold.ll
|
|
define dso_local ptr @_Z3foov() !dbg !4 {
|
|
entry:
|
|
%call = call ptr @_Znam(i64 4) #0, !dbg !5
|
|
; ALL: call ptr @_Znam(i64 4) {{.*}} !memprof ![[M1:[0-9]+]]
|
|
ret ptr %call
|
|
}
|
|
|
|
;; No MIBs get context size info without option.
|
|
; NONE: ![[M1]] = !{![[M2:[0-9]+]], ![[M3:[0-9]+]], ![[M4:[0-9]+]], ![[M5:[0-9]+]], ![[M6:[0-9]+]]}
|
|
; NONE: ![[M2]] = !{![[C2:[0-9]+]], !"cold"}
|
|
; NONE: ![[C2]] = !{i64 590523745590780990, i64 720385627691022109}
|
|
; NONE: ![[M3]] = !{![[C3:[0-9]+]], !"cold"}
|
|
; NONE: ![[C3]] = !{i64 590523745590780990, i64 8256520048276991898}
|
|
; NONE: ![[M4]] = !{![[C4:[0-9]+]], !"cold"}
|
|
; NONE: ![[C4]] = !{i64 590523745590780990, i64 -6953100768213558995}
|
|
; NONE: ![[M5]] = !{![[C5:[0-9]+]], !"cold"}
|
|
; NONE: ![[C5]] = !{i64 590523745590780990, i64 -6435117705372285425}
|
|
; NONE: ![[M6]] = !{![[C6:[0-9]+]], !"notcold"}
|
|
; NONE: ![[C6]] = !{i64 590523745590780990, i64 -2847840206325626610}
|
|
|
|
; THRESH99: ![[M1]] = !{![[M2:[0-9]+]], ![[M3:[0-9]+]], ![[M4:[0-9]+]], ![[M5:[0-9]+]], ![[M6:[0-9]+]]}
|
|
; THRESH99: ![[M2]] = !{![[C2:[0-9]+]], !"cold"}
|
|
; THRESH99: ![[C2]] = !{i64 590523745590780990, i64 720385627691022109}
|
|
;; MIB with size 400 is now included with threshold 99%
|
|
; THRESH99: ![[M3]] = !{![[C3:[0-9]+]], !"cold", ![[S3:[0-9]+]]}
|
|
; THRESH99: ![[C3]] = !{i64 590523745590780990, i64 8256520048276991898}
|
|
; THRESH99: ![[S3]] = !{i64 6509573709067523871, i64 400}
|
|
; THRESH99: ![[M4]] = !{![[C4:[0-9]+]], !"cold"}
|
|
; THRESH99: ![[C4]] = !{i64 590523745590780990, i64 -6953100768213558995}
|
|
; THRESH99: ![[M5]] = !{![[C5:[0-9]+]], !"cold"}
|
|
; THRESH99: ![[C5]] = !{i64 590523745590780990, i64 -6435117705372285425}
|
|
; THRESH99: ![[M6]] = !{![[C6:[0-9]+]], !"notcold"}
|
|
; THRESH99: ![[C6]] = !{i64 590523745590780990, i64 -2847840206325626610}
|
|
|
|
; THRESH75: ![[M1]] = !{![[M2:[0-9]+]], ![[M3:[0-9]+]], ![[M4:[0-9]+]], ![[M5:[0-9]+]], ![[M6:[0-9]+]]}
|
|
; THRESH75: ![[M2]] = !{![[C2:[0-9]+]], !"cold"}
|
|
; THRESH75: ![[C2]] = !{i64 590523745590780990, i64 720385627691022109}
|
|
; THRESH75: ![[M3]] = !{![[C3:[0-9]+]], !"cold", ![[S3:[0-9]+]]}
|
|
; THRESH75: ![[C3]] = !{i64 590523745590780990, i64 8256520048276991898}
|
|
; THRESH75: ![[S3]] = !{i64 6509573709067523871, i64 400}
|
|
; THRESH75: ![[M4]] = !{![[C4:[0-9]+]], !"cold"}
|
|
; THRESH75: ![[C4]] = !{i64 590523745590780990, i64 -6953100768213558995}
|
|
;; MIB with size 300 is now included with threshold 75%
|
|
; THRESH75: ![[M5]] = !{![[C5:[0-9]+]], !"cold", ![[S5:[0-9]+]]}
|
|
; THRESH75: ![[C5]] = !{i64 590523745590780990, i64 -6435117705372285425}
|
|
; THRESH75: ![[S5]] = !{i64 86248815258696712, i64 300}
|
|
; THRESH75: ![[M6]] = !{![[C6:[0-9]+]], !"notcold"}
|
|
; THRESH75: ![[C6]] = !{i64 590523745590780990, i64 -2847840206325626610}
|
|
|
|
; THRESH50: ![[M1]] = !{![[M2:[0-9]+]], ![[M3:[0-9]+]], ![[M4:[0-9]+]], ![[M5:[0-9]+]], ![[M6:[0-9]+]]}
|
|
;; The smallest never gets context size info.
|
|
; THRESH50: ![[M2]] = !{![[C2:[0-9]+]], !"cold"}
|
|
; THRESH50: ![[C2]] = !{i64 590523745590780990, i64 720385627691022109}
|
|
; THRESH50: ![[M3]] = !{![[C3:[0-9]+]], !"cold", ![[S3:[0-9]+]]}
|
|
; THRESH50: ![[C3]] = !{i64 590523745590780990, i64 8256520048276991898}
|
|
; THRESH50: ![[S3]] = !{i64 6509573709067523871, i64 400}
|
|
;; MIB with size 200 is now included with threshold 50%
|
|
; THRESH50: ![[M4]] = !{![[C4:[0-9]+]], !"cold", ![[S4:[0-9]+]]}
|
|
; THRESH50: ![[C4]] = !{i64 590523745590780990, i64 -6953100768213558995}
|
|
; THRESH50: ![[S4]] = !{i64 -2629930016428010893, i64 200}
|
|
; THRESH50: ![[M5]] = !{![[C5:[0-9]+]], !"cold", ![[S5:[0-9]+]]}
|
|
; THRESH50: ![[C5]] = !{i64 590523745590780990, i64 -6435117705372285425}
|
|
; THRESH50: ![[S5]] = !{i64 86248815258696712, i64 300}
|
|
;; Non cold context never gets context size info
|
|
; THRESH50: ![[M6]] = !{![[C6:[0-9]+]], !"notcold"}
|
|
; THRESH50: ![[C6]] = !{i64 590523745590780990, i64 -2847840206325626610}
|
|
|
|
declare ptr @_Znam(i64)
|
|
|
|
attributes #0 = { builtin allocsize(0) }
|
|
|
|
!llvm.module.flags = !{!2, !3}
|
|
|
|
!0 = distinct !DICompileUnit(language: DW_LANG_C_plus_plus_14, file: !1)
|
|
!1 = !DIFile(filename: "t", directory: "/")
|
|
!2 = !{i32 7, !"Dwarf Version", i32 5}
|
|
!3 = !{i32 2, !"Debug Info Version", i32 3}
|
|
!4 = distinct !DISubprogram(name: "foo", linkageName: "_Z3foov", scope: !1, file: !1, line: 1, unit: !0)
|
|
!5 = !DILocation(line: 1, column: 22, scope: !4)
|
|
!6 = !DILocation(line: 2, column: 22, scope: !4)
|
|
!7 = !DILocation(line: 3, column: 22, scope: !4)
|
|
!8 = !DILocation(line: 4, column: 22, scope: !4)
|