Propagation of profile samples through the CFG.
This adds a propagation heuristic to convert instruction samples
into branch weights. It implements a similar heuristic to the one
implemented by Dehao Chen on GCC.
The propagation proceeds in 3 phases:
1- Assignment of block weights. All the basic blocks in the function
are initial assigned the same weight as their most frequently
executed instruction.
2- Creation of equivalence classes. Since samples may be missing from
blocks, we can fill in the gaps by setting the weights of all the
blocks in the same equivalence class to the same weight. To compute
the concept of equivalence, we use dominance and loop information.
Two blocks B1 and B2 are in the same equivalence class if B1
dominates B2, B2 post-dominates B1 and both are in the same loop.
3- Propagation of block weights into edges. This uses a simple
propagation heuristic. The following rules are applied to every
block B in the CFG:
- If B has a single predecessor/successor, then the weight
of that edge is the weight of the block.
- If all the edges are known except one, and the weight of the
block is already known, the weight of the unknown edge will
be the weight of the block minus the sum of all the known
edges. If the sum of all the known edges is larger than B's weight,
we set the unknown edge weight to zero.
- If there is a self-referential edge, and the weight of the block is
known, the weight for that edge is set to the weight of the block
minus the weight of the other incoming edges to that block (if
known).
Since this propagation is not guaranteed to finalize for every CFG, we
only allow it to proceed for a limited number of iterations (controlled
by -sample-profile-max-propagate-iterations). It currently uses the same
GCC default of 100.
Before propagation starts, the pass builds (for each block) a list of
unique predecessors and successors. This is necessary to handle
identical edges in multiway branches. Since we visit all blocks and all
edges of the CFG, it is cleaner to build these lists once at the start
of the pass.
Finally, the patch fixes the computation of relative line locations.
The profiler emits lines relative to the function header. To discover
it, we traverse the compilation unit looking for the subprogram
corresponding to the function. The line number of that subprogram is the
line where the function begins. That becomes line zero for all the
relative locations.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@198972 91177308-0d34-0410-b5e6-96231b3b80d8
2014-01-10 23:23:46 +00:00
|
|
|
; RUN: opt < %s -sample-profile -sample-profile-file=%S/Inputs/propagate.prof | opt -analyze -branch-prob | FileCheck %s
|
|
|
|
|
|
|
|
|
|
; Original C++ code for this test case:
|
|
|
|
|
;
|
|
|
|
|
; #include <stdio.h>
|
|
|
|
|
;
|
|
|
|
|
; long foo(int x, int y, long N) {
|
|
|
|
|
; if (x < y) {
|
|
|
|
|
; return y - x;
|
|
|
|
|
; } else {
|
|
|
|
|
; for (long i = 0; i < N; i++) {
|
|
|
|
|
; if (i > N / 3)
|
|
|
|
|
; x--;
|
|
|
|
|
; if (i > N / 4) {
|
|
|
|
|
; y++;
|
|
|
|
|
; x += 3;
|
|
|
|
|
; } else {
|
|
|
|
|
; for (unsigned j = 0; j < i; j++) {
|
|
|
|
|
; x += j;
|
|
|
|
|
; y -= 3;
|
|
|
|
|
; }
|
|
|
|
|
; }
|
|
|
|
|
; }
|
|
|
|
|
; }
|
|
|
|
|
; return y * x;
|
|
|
|
|
; }
|
|
|
|
|
;
|
|
|
|
|
; int main() {
|
|
|
|
|
; int x = 5678;
|
|
|
|
|
; int y = 1234;
|
|
|
|
|
; long N = 999999;
|
|
|
|
|
; printf("foo(%d, %d, %ld) = %ld\n", x, y, N, foo(x, y, N));
|
|
|
|
|
; return 0;
|
|
|
|
|
; }
|
|
|
|
|
|
|
|
|
|
; ModuleID = 'propagate.cc'
|
|
|
|
|
target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
|
|
|
|
|
target triple = "x86_64-unknown-linux-gnu"
|
|
|
|
|
|
|
|
|
|
@.str = private unnamed_addr constant [24 x i8] c"foo(%d, %d, %ld) = %ld\0A\00", align 1
|
|
|
|
|
|
|
|
|
|
; Function Attrs: nounwind uwtable
|
|
|
|
|
define i64 @_Z3fooiil(i32 %x, i32 %y, i64 %N) #0 {
|
|
|
|
|
entry:
|
|
|
|
|
%retval = alloca i64, align 8
|
|
|
|
|
%x.addr = alloca i32, align 4
|
|
|
|
|
%y.addr = alloca i32, align 4
|
|
|
|
|
%N.addr = alloca i64, align 8
|
|
|
|
|
%i = alloca i64, align 8
|
|
|
|
|
%j = alloca i32, align 4
|
|
|
|
|
store i32 %x, i32* %x.addr, align 4
|
|
|
|
|
store i32 %y, i32* %y.addr, align 4
|
|
|
|
|
store i64 %N, i64* %N.addr, align 8
|
|
|
|
|
%0 = load i32* %x.addr, align 4, !dbg !11
|
|
|
|
|
%1 = load i32* %y.addr, align 4, !dbg !11
|
|
|
|
|
%cmp = icmp slt i32 %0, %1, !dbg !11
|
|
|
|
|
br i1 %cmp, label %if.then, label %if.else, !dbg !11
|
|
|
|
|
|
|
|
|
|
if.then: ; preds = %entry
|
|
|
|
|
%2 = load i32* %y.addr, align 4, !dbg !13
|
|
|
|
|
%3 = load i32* %x.addr, align 4, !dbg !13
|
|
|
|
|
%sub = sub nsw i32 %2, %3, !dbg !13
|
|
|
|
|
%conv = sext i32 %sub to i64, !dbg !13
|
|
|
|
|
store i64 %conv, i64* %retval, !dbg !13
|
|
|
|
|
br label %return, !dbg !13
|
|
|
|
|
|
|
|
|
|
if.else: ; preds = %entry
|
|
|
|
|
store i64 0, i64* %i, align 8, !dbg !15
|
|
|
|
|
br label %for.cond, !dbg !15
|
|
|
|
|
|
|
|
|
|
for.cond: ; preds = %for.inc16, %if.else
|
|
|
|
|
%4 = load i64* %i, align 8, !dbg !15
|
|
|
|
|
%5 = load i64* %N.addr, align 8, !dbg !15
|
|
|
|
|
%cmp1 = icmp slt i64 %4, %5, !dbg !15
|
|
|
|
|
br i1 %cmp1, label %for.body, label %for.end18, !dbg !15
|
|
|
|
|
; CHECK: edge for.cond -> for.body probability is 10 / 11 = 90.9091% [HOT edge]
|
|
|
|
|
; CHECK: edge for.cond -> for.end18 probability is 1 / 11 = 9.09091%
|
|
|
|
|
|
|
|
|
|
for.body: ; preds = %for.cond
|
|
|
|
|
%6 = load i64* %i, align 8, !dbg !18
|
|
|
|
|
%7 = load i64* %N.addr, align 8, !dbg !18
|
|
|
|
|
%div = sdiv i64 %7, 3, !dbg !18
|
|
|
|
|
%cmp2 = icmp sgt i64 %6, %div, !dbg !18
|
|
|
|
|
br i1 %cmp2, label %if.then3, label %if.end, !dbg !18
|
|
|
|
|
; CHECK: edge for.body -> if.then3 probability is 1 / 5 = 20%
|
|
|
|
|
; CHECK: edge for.body -> if.end probability is 4 / 5 = 80%
|
|
|
|
|
|
|
|
|
|
if.then3: ; preds = %for.body
|
|
|
|
|
%8 = load i32* %x.addr, align 4, !dbg !21
|
|
|
|
|
%dec = add nsw i32 %8, -1, !dbg !21
|
|
|
|
|
store i32 %dec, i32* %x.addr, align 4, !dbg !21
|
|
|
|
|
br label %if.end, !dbg !21
|
|
|
|
|
|
|
|
|
|
if.end: ; preds = %if.then3, %for.body
|
|
|
|
|
%9 = load i64* %i, align 8, !dbg !22
|
|
|
|
|
%10 = load i64* %N.addr, align 8, !dbg !22
|
|
|
|
|
%div4 = sdiv i64 %10, 4, !dbg !22
|
|
|
|
|
%cmp5 = icmp sgt i64 %9, %div4, !dbg !22
|
|
|
|
|
br i1 %cmp5, label %if.then6, label %if.else7, !dbg !22
|
|
|
|
|
; CHECK: edge if.end -> if.then6 probability is 3 / 6342 = 0.0473037%
|
|
|
|
|
; CHECK: edge if.end -> if.else7 probability is 6339 / 6342 = 99.9527% [HOT edge]
|
|
|
|
|
|
|
|
|
|
if.then6: ; preds = %if.end
|
|
|
|
|
%11 = load i32* %y.addr, align 4, !dbg !24
|
|
|
|
|
%inc = add nsw i32 %11, 1, !dbg !24
|
|
|
|
|
store i32 %inc, i32* %y.addr, align 4, !dbg !24
|
|
|
|
|
%12 = load i32* %x.addr, align 4, !dbg !26
|
|
|
|
|
%add = add nsw i32 %12, 3, !dbg !26
|
|
|
|
|
store i32 %add, i32* %x.addr, align 4, !dbg !26
|
|
|
|
|
br label %if.end15, !dbg !27
|
|
|
|
|
|
|
|
|
|
if.else7: ; preds = %if.end
|
|
|
|
|
store i32 0, i32* %j, align 4, !dbg !28
|
|
|
|
|
br label %for.cond8, !dbg !28
|
|
|
|
|
|
|
|
|
|
for.cond8: ; preds = %for.inc, %if.else7
|
|
|
|
|
%13 = load i32* %j, align 4, !dbg !28
|
|
|
|
|
%conv9 = zext i32 %13 to i64, !dbg !28
|
|
|
|
|
%14 = load i64* %i, align 8, !dbg !28
|
|
|
|
|
%cmp10 = icmp slt i64 %conv9, %14, !dbg !28
|
|
|
|
|
br i1 %cmp10, label %for.body11, label %for.end, !dbg !28
|
|
|
|
|
; CHECK: edge for.cond8 -> for.body11 probability is 16191 / 16192 = 99.9938% [HOT edge]
|
|
|
|
|
; CHECK: edge for.cond8 -> for.end probability is 1 / 16192 = 0.00617589%
|
|
|
|
|
|
|
|
|
|
for.body11: ; preds = %for.cond8
|
|
|
|
|
%15 = load i32* %j, align 4, !dbg !31
|
|
|
|
|
%16 = load i32* %x.addr, align 4, !dbg !31
|
|
|
|
|
%add12 = add i32 %16, %15, !dbg !31
|
|
|
|
|
store i32 %add12, i32* %x.addr, align 4, !dbg !31
|
|
|
|
|
%17 = load i32* %y.addr, align 4, !dbg !33
|
|
|
|
|
%sub13 = sub nsw i32 %17, 3, !dbg !33
|
|
|
|
|
store i32 %sub13, i32* %y.addr, align 4, !dbg !33
|
|
|
|
|
br label %for.inc, !dbg !34
|
|
|
|
|
|
|
|
|
|
for.inc: ; preds = %for.body11
|
|
|
|
|
%18 = load i32* %j, align 4, !dbg !28
|
|
|
|
|
%inc14 = add i32 %18, 1, !dbg !28
|
|
|
|
|
store i32 %inc14, i32* %j, align 4, !dbg !28
|
|
|
|
|
br label %for.cond8, !dbg !28
|
|
|
|
|
|
|
|
|
|
for.end: ; preds = %for.cond8
|
|
|
|
|
br label %if.end15
|
|
|
|
|
|
|
|
|
|
if.end15: ; preds = %for.end, %if.then6
|
|
|
|
|
br label %for.inc16, !dbg !35
|
|
|
|
|
|
|
|
|
|
for.inc16: ; preds = %if.end15
|
|
|
|
|
%19 = load i64* %i, align 8, !dbg !15
|
|
|
|
|
%inc17 = add nsw i64 %19, 1, !dbg !15
|
|
|
|
|
store i64 %inc17, i64* %i, align 8, !dbg !15
|
|
|
|
|
br label %for.cond, !dbg !15
|
|
|
|
|
|
|
|
|
|
for.end18: ; preds = %for.cond
|
|
|
|
|
br label %if.end19
|
|
|
|
|
|
|
|
|
|
if.end19: ; preds = %for.end18
|
|
|
|
|
%20 = load i32* %y.addr, align 4, !dbg !36
|
|
|
|
|
%21 = load i32* %x.addr, align 4, !dbg !36
|
|
|
|
|
%mul = mul nsw i32 %20, %21, !dbg !36
|
|
|
|
|
%conv20 = sext i32 %mul to i64, !dbg !36
|
|
|
|
|
store i64 %conv20, i64* %retval, !dbg !36
|
|
|
|
|
br label %return, !dbg !36
|
|
|
|
|
|
|
|
|
|
return: ; preds = %if.end19, %if.then
|
|
|
|
|
%22 = load i64* %retval, !dbg !37
|
|
|
|
|
ret i64 %22, !dbg !37
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
; Function Attrs: uwtable
|
|
|
|
|
define i32 @main() #1 {
|
|
|
|
|
entry:
|
|
|
|
|
%retval = alloca i32, align 4
|
|
|
|
|
%x = alloca i32, align 4
|
|
|
|
|
%y = alloca i32, align 4
|
|
|
|
|
%N = alloca i64, align 8
|
|
|
|
|
store i32 0, i32* %retval
|
|
|
|
|
store i32 5678, i32* %x, align 4, !dbg !38
|
|
|
|
|
store i32 1234, i32* %y, align 4, !dbg !39
|
|
|
|
|
store i64 999999, i64* %N, align 8, !dbg !40
|
|
|
|
|
%0 = load i32* %x, align 4, !dbg !41
|
|
|
|
|
%1 = load i32* %y, align 4, !dbg !41
|
|
|
|
|
%2 = load i64* %N, align 8, !dbg !41
|
|
|
|
|
%3 = load i32* %x, align 4, !dbg !41
|
|
|
|
|
%4 = load i32* %y, align 4, !dbg !41
|
|
|
|
|
%5 = load i64* %N, align 8, !dbg !41
|
|
|
|
|
%call = call i64 @_Z3fooiil(i32 %3, i32 %4, i64 %5), !dbg !41
|
|
|
|
|
%call1 = call i32 (i8*, ...)* @printf(i8* getelementptr inbounds ([24 x i8]* @.str, i32 0, i32 0), i32 %0, i32 %1, i64 %2, i64 %call), !dbg !41
|
|
|
|
|
ret i32 0, !dbg !42
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
declare i32 @printf(i8*, ...) #2
|
|
|
|
|
|
|
|
|
|
attributes #0 = { nounwind uwtable "less-precise-fpmad"="false" "no-frame-pointer-elim"="true" "no-frame-pointer-elim-non-leaf" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
|
|
|
|
|
attributes #1 = { uwtable "less-precise-fpmad"="false" "no-frame-pointer-elim"="true" "no-frame-pointer-elim-non-leaf" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
|
|
|
|
|
attributes #2 = { "less-precise-fpmad"="false" "no-frame-pointer-elim"="true" "no-frame-pointer-elim-non-leaf" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
|
|
|
|
|
|
|
|
|
|
!llvm.dbg.cu = !{!0}
|
|
|
|
|
!llvm.module.flags = !{!8, !9}
|
|
|
|
|
!llvm.ident = !{!10}
|
|
|
|
|
|
|
|
|
|
!0 = metadata !{i32 786449, metadata !1, i32 4, metadata !"clang version 3.5 ", i1 false, metadata !"", i32 0, metadata !2, metadata !2, metadata !3, metadata !2, metadata !2, metadata !""} ; [ DW_TAG_compile_unit ] [propagate.cc] [DW_LANG_C_plus_plus]
|
|
|
|
|
!1 = metadata !{metadata !"propagate.cc", metadata !"."}
|
|
|
|
|
!2 = metadata !{i32 0}
|
|
|
|
|
!3 = metadata !{metadata !4, metadata !7}
|
|
|
|
|
!4 = metadata !{i32 786478, metadata !1, metadata !5, metadata !"foo", metadata !"foo", metadata !"", i32 3, metadata !6, i1 false, i1 true, i32 0, i32 0, null, i32 256, i1 false, i64 (i32, i32, i64)* @_Z3fooiil, null, null, metadata !2, i32 3} ; [ DW_TAG_subprogram ] [line 3] [def] [foo]
|
|
|
|
|
!5 = metadata !{i32 786473, metadata !1} ; [ DW_TAG_file_type ] [propagate.cc]
|
|
|
|
|
!6 = metadata !{i32 786453, i32 0, null, metadata !"", i32 0, i64 0, i64 0, i64 0, i32 0, null, metadata !2, i32 0, null, null, null} ; [ DW_TAG_subroutine_type ] [line 0, size 0, align 0, offset 0] [from ]
|
|
|
|
|
!7 = metadata !{i32 786478, metadata !1, metadata !5, metadata !"main", metadata !"main", metadata !"", i32 24, metadata !6, i1 false, i1 true, i32 0, i32 0, null, i32 256, i1 false, i32 ()* @main, null, null, metadata !2, i32 24} ; [ DW_TAG_subprogram ] [line 24] [def] [main]
|
|
|
|
|
!8 = metadata !{i32 2, metadata !"Dwarf Version", i32 4}
|
|
|
|
|
!9 = metadata !{i32 1, metadata !"Debug Info Version", i32 1}
|
|
|
|
|
!10 = metadata !{metadata !"clang version 3.5 "}
|
|
|
|
|
!11 = metadata !{i32 4, i32 0, metadata !12, null}
|
2014-03-10 22:41:28 +00:00
|
|
|
!12 = metadata !{i32 786443, metadata !1, metadata !4, i32 4, i32 0, i32 0, i32 0} ; [ DW_TAG_lexical_block ] [propagate.cc]
|
Propagation of profile samples through the CFG.
This adds a propagation heuristic to convert instruction samples
into branch weights. It implements a similar heuristic to the one
implemented by Dehao Chen on GCC.
The propagation proceeds in 3 phases:
1- Assignment of block weights. All the basic blocks in the function
are initial assigned the same weight as their most frequently
executed instruction.
2- Creation of equivalence classes. Since samples may be missing from
blocks, we can fill in the gaps by setting the weights of all the
blocks in the same equivalence class to the same weight. To compute
the concept of equivalence, we use dominance and loop information.
Two blocks B1 and B2 are in the same equivalence class if B1
dominates B2, B2 post-dominates B1 and both are in the same loop.
3- Propagation of block weights into edges. This uses a simple
propagation heuristic. The following rules are applied to every
block B in the CFG:
- If B has a single predecessor/successor, then the weight
of that edge is the weight of the block.
- If all the edges are known except one, and the weight of the
block is already known, the weight of the unknown edge will
be the weight of the block minus the sum of all the known
edges. If the sum of all the known edges is larger than B's weight,
we set the unknown edge weight to zero.
- If there is a self-referential edge, and the weight of the block is
known, the weight for that edge is set to the weight of the block
minus the weight of the other incoming edges to that block (if
known).
Since this propagation is not guaranteed to finalize for every CFG, we
only allow it to proceed for a limited number of iterations (controlled
by -sample-profile-max-propagate-iterations). It currently uses the same
GCC default of 100.
Before propagation starts, the pass builds (for each block) a list of
unique predecessors and successors. This is necessary to handle
identical edges in multiway branches. Since we visit all blocks and all
edges of the CFG, it is cleaner to build these lists once at the start
of the pass.
Finally, the patch fixes the computation of relative line locations.
The profiler emits lines relative to the function header. To discover
it, we traverse the compilation unit looking for the subprogram
corresponding to the function. The line number of that subprogram is the
line where the function begins. That becomes line zero for all the
relative locations.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@198972 91177308-0d34-0410-b5e6-96231b3b80d8
2014-01-10 23:23:46 +00:00
|
|
|
!13 = metadata !{i32 5, i32 0, metadata !14, null}
|
2014-03-10 22:41:28 +00:00
|
|
|
!14 = metadata !{i32 786443, metadata !1, metadata !12, i32 4, i32 0, i32 0, i32 1} ; [ DW_TAG_lexical_block ] [propagate.cc]
|
Propagation of profile samples through the CFG.
This adds a propagation heuristic to convert instruction samples
into branch weights. It implements a similar heuristic to the one
implemented by Dehao Chen on GCC.
The propagation proceeds in 3 phases:
1- Assignment of block weights. All the basic blocks in the function
are initial assigned the same weight as their most frequently
executed instruction.
2- Creation of equivalence classes. Since samples may be missing from
blocks, we can fill in the gaps by setting the weights of all the
blocks in the same equivalence class to the same weight. To compute
the concept of equivalence, we use dominance and loop information.
Two blocks B1 and B2 are in the same equivalence class if B1
dominates B2, B2 post-dominates B1 and both are in the same loop.
3- Propagation of block weights into edges. This uses a simple
propagation heuristic. The following rules are applied to every
block B in the CFG:
- If B has a single predecessor/successor, then the weight
of that edge is the weight of the block.
- If all the edges are known except one, and the weight of the
block is already known, the weight of the unknown edge will
be the weight of the block minus the sum of all the known
edges. If the sum of all the known edges is larger than B's weight,
we set the unknown edge weight to zero.
- If there is a self-referential edge, and the weight of the block is
known, the weight for that edge is set to the weight of the block
minus the weight of the other incoming edges to that block (if
known).
Since this propagation is not guaranteed to finalize for every CFG, we
only allow it to proceed for a limited number of iterations (controlled
by -sample-profile-max-propagate-iterations). It currently uses the same
GCC default of 100.
Before propagation starts, the pass builds (for each block) a list of
unique predecessors and successors. This is necessary to handle
identical edges in multiway branches. Since we visit all blocks and all
edges of the CFG, it is cleaner to build these lists once at the start
of the pass.
Finally, the patch fixes the computation of relative line locations.
The profiler emits lines relative to the function header. To discover
it, we traverse the compilation unit looking for the subprogram
corresponding to the function. The line number of that subprogram is the
line where the function begins. That becomes line zero for all the
relative locations.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@198972 91177308-0d34-0410-b5e6-96231b3b80d8
2014-01-10 23:23:46 +00:00
|
|
|
!15 = metadata !{i32 7, i32 0, metadata !16, null}
|
2014-03-10 22:41:28 +00:00
|
|
|
!16 = metadata !{i32 786443, metadata !1, metadata !17, i32 7, i32 0, i32 0, i32 3} ; [ DW_TAG_lexical_block ] [propagate.cc]
|
|
|
|
|
!17 = metadata !{i32 786443, metadata !1, metadata !12, i32 6, i32 0, i32 0, i32 2} ; [ DW_TAG_lexical_block ] [propagate.cc]
|
Propagation of profile samples through the CFG.
This adds a propagation heuristic to convert instruction samples
into branch weights. It implements a similar heuristic to the one
implemented by Dehao Chen on GCC.
The propagation proceeds in 3 phases:
1- Assignment of block weights. All the basic blocks in the function
are initial assigned the same weight as their most frequently
executed instruction.
2- Creation of equivalence classes. Since samples may be missing from
blocks, we can fill in the gaps by setting the weights of all the
blocks in the same equivalence class to the same weight. To compute
the concept of equivalence, we use dominance and loop information.
Two blocks B1 and B2 are in the same equivalence class if B1
dominates B2, B2 post-dominates B1 and both are in the same loop.
3- Propagation of block weights into edges. This uses a simple
propagation heuristic. The following rules are applied to every
block B in the CFG:
- If B has a single predecessor/successor, then the weight
of that edge is the weight of the block.
- If all the edges are known except one, and the weight of the
block is already known, the weight of the unknown edge will
be the weight of the block minus the sum of all the known
edges. If the sum of all the known edges is larger than B's weight,
we set the unknown edge weight to zero.
- If there is a self-referential edge, and the weight of the block is
known, the weight for that edge is set to the weight of the block
minus the weight of the other incoming edges to that block (if
known).
Since this propagation is not guaranteed to finalize for every CFG, we
only allow it to proceed for a limited number of iterations (controlled
by -sample-profile-max-propagate-iterations). It currently uses the same
GCC default of 100.
Before propagation starts, the pass builds (for each block) a list of
unique predecessors and successors. This is necessary to handle
identical edges in multiway branches. Since we visit all blocks and all
edges of the CFG, it is cleaner to build these lists once at the start
of the pass.
Finally, the patch fixes the computation of relative line locations.
The profiler emits lines relative to the function header. To discover
it, we traverse the compilation unit looking for the subprogram
corresponding to the function. The line number of that subprogram is the
line where the function begins. That becomes line zero for all the
relative locations.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@198972 91177308-0d34-0410-b5e6-96231b3b80d8
2014-01-10 23:23:46 +00:00
|
|
|
!18 = metadata !{i32 8, i32 0, metadata !19, null} ; [ DW_TAG_imported_declaration ]
|
2014-03-10 22:41:28 +00:00
|
|
|
!19 = metadata !{i32 786443, metadata !1, metadata !20, i32 8, i32 0, i32 0, i32 5} ; [ DW_TAG_lexical_block ] [propagate.cc]
|
|
|
|
|
!20 = metadata !{i32 786443, metadata !1, metadata !16, i32 7, i32 0, i32 0, i32 4} ; [ DW_TAG_lexical_block ] [propagate.cc]
|
Propagation of profile samples through the CFG.
This adds a propagation heuristic to convert instruction samples
into branch weights. It implements a similar heuristic to the one
implemented by Dehao Chen on GCC.
The propagation proceeds in 3 phases:
1- Assignment of block weights. All the basic blocks in the function
are initial assigned the same weight as their most frequently
executed instruction.
2- Creation of equivalence classes. Since samples may be missing from
blocks, we can fill in the gaps by setting the weights of all the
blocks in the same equivalence class to the same weight. To compute
the concept of equivalence, we use dominance and loop information.
Two blocks B1 and B2 are in the same equivalence class if B1
dominates B2, B2 post-dominates B1 and both are in the same loop.
3- Propagation of block weights into edges. This uses a simple
propagation heuristic. The following rules are applied to every
block B in the CFG:
- If B has a single predecessor/successor, then the weight
of that edge is the weight of the block.
- If all the edges are known except one, and the weight of the
block is already known, the weight of the unknown edge will
be the weight of the block minus the sum of all the known
edges. If the sum of all the known edges is larger than B's weight,
we set the unknown edge weight to zero.
- If there is a self-referential edge, and the weight of the block is
known, the weight for that edge is set to the weight of the block
minus the weight of the other incoming edges to that block (if
known).
Since this propagation is not guaranteed to finalize for every CFG, we
only allow it to proceed for a limited number of iterations (controlled
by -sample-profile-max-propagate-iterations). It currently uses the same
GCC default of 100.
Before propagation starts, the pass builds (for each block) a list of
unique predecessors and successors. This is necessary to handle
identical edges in multiway branches. Since we visit all blocks and all
edges of the CFG, it is cleaner to build these lists once at the start
of the pass.
Finally, the patch fixes the computation of relative line locations.
The profiler emits lines relative to the function header. To discover
it, we traverse the compilation unit looking for the subprogram
corresponding to the function. The line number of that subprogram is the
line where the function begins. That becomes line zero for all the
relative locations.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@198972 91177308-0d34-0410-b5e6-96231b3b80d8
2014-01-10 23:23:46 +00:00
|
|
|
!21 = metadata !{i32 9, i32 0, metadata !19, null}
|
|
|
|
|
!22 = metadata !{i32 10, i32 0, metadata !23, null}
|
2014-03-10 22:41:28 +00:00
|
|
|
!23 = metadata !{i32 786443, metadata !1, metadata !20, i32 10, i32 0, i32 0, i32 6} ; [ DW_TAG_lexical_block ] [propagate.cc]
|
Propagation of profile samples through the CFG.
This adds a propagation heuristic to convert instruction samples
into branch weights. It implements a similar heuristic to the one
implemented by Dehao Chen on GCC.
The propagation proceeds in 3 phases:
1- Assignment of block weights. All the basic blocks in the function
are initial assigned the same weight as their most frequently
executed instruction.
2- Creation of equivalence classes. Since samples may be missing from
blocks, we can fill in the gaps by setting the weights of all the
blocks in the same equivalence class to the same weight. To compute
the concept of equivalence, we use dominance and loop information.
Two blocks B1 and B2 are in the same equivalence class if B1
dominates B2, B2 post-dominates B1 and both are in the same loop.
3- Propagation of block weights into edges. This uses a simple
propagation heuristic. The following rules are applied to every
block B in the CFG:
- If B has a single predecessor/successor, then the weight
of that edge is the weight of the block.
- If all the edges are known except one, and the weight of the
block is already known, the weight of the unknown edge will
be the weight of the block minus the sum of all the known
edges. If the sum of all the known edges is larger than B's weight,
we set the unknown edge weight to zero.
- If there is a self-referential edge, and the weight of the block is
known, the weight for that edge is set to the weight of the block
minus the weight of the other incoming edges to that block (if
known).
Since this propagation is not guaranteed to finalize for every CFG, we
only allow it to proceed for a limited number of iterations (controlled
by -sample-profile-max-propagate-iterations). It currently uses the same
GCC default of 100.
Before propagation starts, the pass builds (for each block) a list of
unique predecessors and successors. This is necessary to handle
identical edges in multiway branches. Since we visit all blocks and all
edges of the CFG, it is cleaner to build these lists once at the start
of the pass.
Finally, the patch fixes the computation of relative line locations.
The profiler emits lines relative to the function header. To discover
it, we traverse the compilation unit looking for the subprogram
corresponding to the function. The line number of that subprogram is the
line where the function begins. That becomes line zero for all the
relative locations.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@198972 91177308-0d34-0410-b5e6-96231b3b80d8
2014-01-10 23:23:46 +00:00
|
|
|
!24 = metadata !{i32 11, i32 0, metadata !25, null}
|
2014-03-10 22:41:28 +00:00
|
|
|
!25 = metadata !{i32 786443, metadata !1, metadata !23, i32 10, i32 0, i32 0, i32 7} ; [ DW_TAG_lexical_block ] [propagate.cc]
|
Propagation of profile samples through the CFG.
This adds a propagation heuristic to convert instruction samples
into branch weights. It implements a similar heuristic to the one
implemented by Dehao Chen on GCC.
The propagation proceeds in 3 phases:
1- Assignment of block weights. All the basic blocks in the function
are initial assigned the same weight as their most frequently
executed instruction.
2- Creation of equivalence classes. Since samples may be missing from
blocks, we can fill in the gaps by setting the weights of all the
blocks in the same equivalence class to the same weight. To compute
the concept of equivalence, we use dominance and loop information.
Two blocks B1 and B2 are in the same equivalence class if B1
dominates B2, B2 post-dominates B1 and both are in the same loop.
3- Propagation of block weights into edges. This uses a simple
propagation heuristic. The following rules are applied to every
block B in the CFG:
- If B has a single predecessor/successor, then the weight
of that edge is the weight of the block.
- If all the edges are known except one, and the weight of the
block is already known, the weight of the unknown edge will
be the weight of the block minus the sum of all the known
edges. If the sum of all the known edges is larger than B's weight,
we set the unknown edge weight to zero.
- If there is a self-referential edge, and the weight of the block is
known, the weight for that edge is set to the weight of the block
minus the weight of the other incoming edges to that block (if
known).
Since this propagation is not guaranteed to finalize for every CFG, we
only allow it to proceed for a limited number of iterations (controlled
by -sample-profile-max-propagate-iterations). It currently uses the same
GCC default of 100.
Before propagation starts, the pass builds (for each block) a list of
unique predecessors and successors. This is necessary to handle
identical edges in multiway branches. Since we visit all blocks and all
edges of the CFG, it is cleaner to build these lists once at the start
of the pass.
Finally, the patch fixes the computation of relative line locations.
The profiler emits lines relative to the function header. To discover
it, we traverse the compilation unit looking for the subprogram
corresponding to the function. The line number of that subprogram is the
line where the function begins. That becomes line zero for all the
relative locations.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@198972 91177308-0d34-0410-b5e6-96231b3b80d8
2014-01-10 23:23:46 +00:00
|
|
|
!26 = metadata !{i32 12, i32 0, metadata !25, null}
|
|
|
|
|
!27 = metadata !{i32 13, i32 0, metadata !25, null}
|
|
|
|
|
!28 = metadata !{i32 14, i32 0, metadata !29, null}
|
2014-03-10 22:41:28 +00:00
|
|
|
!29 = metadata !{i32 786443, metadata !1, metadata !30, i32 14, i32 0, i32 0, i32 9} ; [ DW_TAG_lexical_block ] [propagate.cc]
|
|
|
|
|
!30 = metadata !{i32 786443, metadata !1, metadata !23, i32 13, i32 0, i32 0, i32 8} ; [ DW_TAG_lexical_block ] [propagate.cc]
|
Propagation of profile samples through the CFG.
This adds a propagation heuristic to convert instruction samples
into branch weights. It implements a similar heuristic to the one
implemented by Dehao Chen on GCC.
The propagation proceeds in 3 phases:
1- Assignment of block weights. All the basic blocks in the function
are initial assigned the same weight as their most frequently
executed instruction.
2- Creation of equivalence classes. Since samples may be missing from
blocks, we can fill in the gaps by setting the weights of all the
blocks in the same equivalence class to the same weight. To compute
the concept of equivalence, we use dominance and loop information.
Two blocks B1 and B2 are in the same equivalence class if B1
dominates B2, B2 post-dominates B1 and both are in the same loop.
3- Propagation of block weights into edges. This uses a simple
propagation heuristic. The following rules are applied to every
block B in the CFG:
- If B has a single predecessor/successor, then the weight
of that edge is the weight of the block.
- If all the edges are known except one, and the weight of the
block is already known, the weight of the unknown edge will
be the weight of the block minus the sum of all the known
edges. If the sum of all the known edges is larger than B's weight,
we set the unknown edge weight to zero.
- If there is a self-referential edge, and the weight of the block is
known, the weight for that edge is set to the weight of the block
minus the weight of the other incoming edges to that block (if
known).
Since this propagation is not guaranteed to finalize for every CFG, we
only allow it to proceed for a limited number of iterations (controlled
by -sample-profile-max-propagate-iterations). It currently uses the same
GCC default of 100.
Before propagation starts, the pass builds (for each block) a list of
unique predecessors and successors. This is necessary to handle
identical edges in multiway branches. Since we visit all blocks and all
edges of the CFG, it is cleaner to build these lists once at the start
of the pass.
Finally, the patch fixes the computation of relative line locations.
The profiler emits lines relative to the function header. To discover
it, we traverse the compilation unit looking for the subprogram
corresponding to the function. The line number of that subprogram is the
line where the function begins. That becomes line zero for all the
relative locations.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@198972 91177308-0d34-0410-b5e6-96231b3b80d8
2014-01-10 23:23:46 +00:00
|
|
|
!31 = metadata !{i32 15, i32 0, metadata !32, null}
|
2014-03-10 22:41:28 +00:00
|
|
|
!32 = metadata !{i32 786443, metadata !1, metadata !29, i32 14, i32 0, i32 0, i32 10} ; [ DW_TAG_lexical_block ] [propagate.cc]
|
Propagation of profile samples through the CFG.
This adds a propagation heuristic to convert instruction samples
into branch weights. It implements a similar heuristic to the one
implemented by Dehao Chen on GCC.
The propagation proceeds in 3 phases:
1- Assignment of block weights. All the basic blocks in the function
are initial assigned the same weight as their most frequently
executed instruction.
2- Creation of equivalence classes. Since samples may be missing from
blocks, we can fill in the gaps by setting the weights of all the
blocks in the same equivalence class to the same weight. To compute
the concept of equivalence, we use dominance and loop information.
Two blocks B1 and B2 are in the same equivalence class if B1
dominates B2, B2 post-dominates B1 and both are in the same loop.
3- Propagation of block weights into edges. This uses a simple
propagation heuristic. The following rules are applied to every
block B in the CFG:
- If B has a single predecessor/successor, then the weight
of that edge is the weight of the block.
- If all the edges are known except one, and the weight of the
block is already known, the weight of the unknown edge will
be the weight of the block minus the sum of all the known
edges. If the sum of all the known edges is larger than B's weight,
we set the unknown edge weight to zero.
- If there is a self-referential edge, and the weight of the block is
known, the weight for that edge is set to the weight of the block
minus the weight of the other incoming edges to that block (if
known).
Since this propagation is not guaranteed to finalize for every CFG, we
only allow it to proceed for a limited number of iterations (controlled
by -sample-profile-max-propagate-iterations). It currently uses the same
GCC default of 100.
Before propagation starts, the pass builds (for each block) a list of
unique predecessors and successors. This is necessary to handle
identical edges in multiway branches. Since we visit all blocks and all
edges of the CFG, it is cleaner to build these lists once at the start
of the pass.
Finally, the patch fixes the computation of relative line locations.
The profiler emits lines relative to the function header. To discover
it, we traverse the compilation unit looking for the subprogram
corresponding to the function. The line number of that subprogram is the
line where the function begins. That becomes line zero for all the
relative locations.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@198972 91177308-0d34-0410-b5e6-96231b3b80d8
2014-01-10 23:23:46 +00:00
|
|
|
!33 = metadata !{i32 16, i32 0, metadata !32, null}
|
|
|
|
|
!34 = metadata !{i32 17, i32 0, metadata !32, null}
|
|
|
|
|
!35 = metadata !{i32 19, i32 0, metadata !20, null}
|
|
|
|
|
!36 = metadata !{i32 21, i32 0, metadata !4, null}
|
|
|
|
|
!37 = metadata !{i32 22, i32 0, metadata !4, null}
|
|
|
|
|
!38 = metadata !{i32 25, i32 0, metadata !7, null}
|
|
|
|
|
!39 = metadata !{i32 26, i32 0, metadata !7, null}
|
|
|
|
|
!40 = metadata !{i32 27, i32 0, metadata !7, null}
|
|
|
|
|
!41 = metadata !{i32 28, i32 0, metadata !7, null}
|
|
|
|
|
!42 = metadata !{i32 29, i32 0, metadata !7, null}
|