RPCS3/llvm-mirror
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-11-25 12:12:47 +01:00
Code Issues Projects Releases Wiki Activity
4b01d392e0
llvm-mirror/test/Transforms/Inline/devirtualize.ll

183 lines
8.7 KiB
LLVM
Raw Normal View History

InstCombine: Don't shrink allocas when combining with a bitcast. When considering folding a bitcast of an alloca into the alloca itself, make sure we don't shrink the amount of memory being allocated, or things rapidly go sideways. rdar://13324424 llvm-svn: 176547
2013-03-06 06:44:53 +01:00
; RUN: opt -S -Os < %s | FileCheck %s
Implement rdar://6295824 and PR6724 with two tiny changes that can have a big effect :). The first is to enable the iterative SCC passmanager juice that kicks in when the scc passmgr detects that a function pass has devirtualized a call. In this case, it will rerun all the passes it manages on the SCC, up to the iteration count limit (4). This is useful because a function pass may devirualize a call, and we want the inliner to inline it, or pruneeh to infer stuff about it, etc. The second patch is to add *all* call sites to the DevirtualizedCalls list the inliner uses. This list is about to get renamed, but the jist of this is that the inliner now reconsiders *all* inlined call sites as candidates for further inlining. The intuition is this that in cases like this: f() { g(1); } g(int x) { h(x); } We analyze this bottom up, and may decide that it isn't profitable to inline H into G. Next step, we decide that it is profitable to inline G into F, and do so, which means that F now calls H. Even though the call from G -> H may not have been profitable to inline, the call from F -> H may be (in this case because a constant allows folding etc). In my spot checks, this doesn't have a big impact on code. For example, the LLC output for 252.eon grew from 0.02% (from 317252 to 317308) and 176.gcc actually shrunk by .3% (from 1525612 to 1520964 bytes). 252.eon never iterated in the SCC Passmgr, 176.gcc iterated at most 1 time. llvm-svn: 102823
2010-05-01 03:15:56 +02:00
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"
target triple = "x86_64-apple-darwin10.0.0"
; Simple devirt testcase, requires iteration between inliner and GVN.
; rdar://6295824
define i32 @foo(i32 ()** noalias %p, i64* noalias %q) nounwind ssp {
entry:
store i32 ()* @bar, i32 ()** %p
store i64 0, i64* %q
[opaque pointer type] Add textual IR support for explicit type parameter to load instruction Essentially the same as the GEP change in r230786. A similar migration script can be used to update test cases, though a few more test case improvements/changes were required this time around: (r229269-r229278) import fileinput import sys import re pat = re.compile(r"((?:=|:|^)\s*load (?:atomic )?(?:volatile )?(.*?))(| addrspace\(\d+\) *)\*($| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$)") for line in sys.stdin: sys.stdout.write(re.sub(pat, r"\1, \2\3*\4", line)) Reviewers: rafael, dexonsmith, grosser Differential Revision: http://reviews.llvm.org/D7649 llvm-svn: 230794
2015-02-27 22:17:42 +01:00
%tmp3 = load i32 ()*, i32 ()** %p ; <i32 ()*> [#uses=1]
Implement rdar://6295824 and PR6724 with two tiny changes that can have a big effect :). The first is to enable the iterative SCC passmanager juice that kicks in when the scc passmgr detects that a function pass has devirtualized a call. In this case, it will rerun all the passes it manages on the SCC, up to the iteration count limit (4). This is useful because a function pass may devirualize a call, and we want the inliner to inline it, or pruneeh to infer stuff about it, etc. The second patch is to add *all* call sites to the DevirtualizedCalls list the inliner uses. This list is about to get renamed, but the jist of this is that the inliner now reconsiders *all* inlined call sites as candidates for further inlining. The intuition is this that in cases like this: f() { g(1); } g(int x) { h(x); } We analyze this bottom up, and may decide that it isn't profitable to inline H into G. Next step, we decide that it is profitable to inline G into F, and do so, which means that F now calls H. Even though the call from G -> H may not have been profitable to inline, the call from F -> H may be (in this case because a constant allows folding etc). In my spot checks, this doesn't have a big impact on code. For example, the LLC output for 252.eon grew from 0.02% (from 317252 to 317308) and 176.gcc actually shrunk by .3% (from 1525612 to 1520964 bytes). 252.eon never iterated in the SCC Passmgr, 176.gcc iterated at most 1 time. llvm-svn: 102823
2010-05-01 03:15:56 +02:00
%call = call i32 %tmp3() ; <i32> [#uses=1]
%X = add i32 %call, 4
ret i32 %X
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_]*\):\( *\)@$FUNC\([( ]*\)\$/;\1\2-LABEL:\3@$FUNC(/g" $TEMP done mv $TEMP $NAME fi done llvm-svn: 186268
2013-07-14 03:42:54 +02:00
; CHECK-LABEL: @foo(
Implement rdar://6295824 and PR6724 with two tiny changes that can have a big effect :). The first is to enable the iterative SCC passmanager juice that kicks in when the scc passmgr detects that a function pass has devirtualized a call. In this case, it will rerun all the passes it manages on the SCC, up to the iteration count limit (4). This is useful because a function pass may devirualize a call, and we want the inliner to inline it, or pruneeh to infer stuff about it, etc. The second patch is to add *all* call sites to the DevirtualizedCalls list the inliner uses. This list is about to get renamed, but the jist of this is that the inliner now reconsiders *all* inlined call sites as candidates for further inlining. The intuition is this that in cases like this: f() { g(1); } g(int x) { h(x); } We analyze this bottom up, and may decide that it isn't profitable to inline H into G. Next step, we decide that it is profitable to inline G into F, and do so, which means that F now calls H. Even though the call from G -> H may not have been profitable to inline, the call from F -> H may be (in this case because a constant allows folding etc). In my spot checks, this doesn't have a big impact on code. For example, the LLC output for 252.eon grew from 0.02% (from 317252 to 317308) and 176.gcc actually shrunk by .3% (from 1525612 to 1520964 bytes). 252.eon never iterated in the SCC Passmgr, 176.gcc iterated at most 1 time. llvm-svn: 102823
2010-05-01 03:15:56 +02:00
; CHECK-NEXT: entry:
; CHECK-NEXT: store
; CHECK-NEXT: store
; CHECK-NEXT: ret i32 11
}
define internal i32 @bar() nounwind ssp {
entry:
ret i32 7
}
;; More complex devirt case, from PR6724
; CHECK: @_Z1gv()
; CHECK-NEXT: entry:
; CHECK-NEXT: ret i32 7
%0 = type { i8*, i8* }
%1 = type { i8*, i8*, i32, i32, i8*, i64, i8*, i64 }
%2 = type { i8*, i8*, i8* }
%struct.A = type { i8** }
%struct.B = type { i8** }
%struct.C = type { [16 x i8] }
%struct.D = type { [16 x i8] }
@_ZTV1D = linkonce_odr constant [6 x i8*] [i8* null, i8* bitcast (%2* @_ZTI1D to i8*), i8* bitcast (i32 (%struct.C*)* @_ZN1D1fEv to i8*), i8* inttoptr (i64 -8 to i8*), i8* bitcast (%2* @_ZTI1D to i8*), i8* bitcast (i32 (%struct.C*)* @_ZThn8_N1D1fEv to i8*)] ; <[6 x i8*]*> [#uses=2]
@_ZTVN10__cxxabiv120__si_class_type_infoE = external global i8* ; <i8**> [#uses=1]
@_ZTS1D = linkonce_odr constant [3 x i8] c"1D\00" ; <[3 x i8]*> [#uses=1]
@_ZTVN10__cxxabiv121__vmi_class_type_infoE = external global i8* ; <i8**> [#uses=1]
@_ZTS1C = linkonce_odr constant [3 x i8] c"1C\00" ; <[3 x i8]*> [#uses=1]
@_ZTVN10__cxxabiv117__class_type_infoE = external global i8* ; <i8**> [#uses=1]
@_ZTS1A = linkonce_odr constant [3 x i8] c"1A\00" ; <[3 x i8]*> [#uses=1]
[opaque pointer type] Add textual IR support for explicit type parameter to gep operator Similar to gep (r230786) and load (r230794) changes. Similar migration script can be used to update test cases, which successfully migrated all of LLVM and Polly, but about 4 test cases needed manually changes in Clang. (this script will read the contents of stdin and massage it into stdout - wrap it in the 'apply.sh' script shown in previous commits + xargs to apply it over a large set of test cases) import fileinput import sys import re rep = re.compile(r"(getelementptr(?:\s+inbounds)?\s*\()((<\d*\s+x\s+)?([^@]*?)(|\s*addrspace\(\d+\))\s*\*(?(3)>)\s*)(?=$|%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|zeroinitializer|<|\[\[[a-zA-Z]|\{\{)", re.MULTILINE | re.DOTALL) def conv(match): line = match.group(1) line += match.group(4) line += ", " line += match.group(2) return line line = sys.stdin.read() off = 0 for match in re.finditer(rep, line): sys.stdout.write(line[off:match.start()]) sys.stdout.write(conv(match)) off = match.end() sys.stdout.write(line[off:]) llvm-svn: 232184
2015-03-13 19:20:45 +01:00
@_ZTI1A = linkonce_odr constant %0 { i8* bitcast (i8** getelementptr inbounds (i8*, i8** @_ZTVN10__cxxabiv117__class_type_infoE, i64 2) to i8*), i8* getelementptr inbounds ([3 x i8], [3 x i8]* @_ZTS1A, i32 0, i32 0) } ; <%0*> [#uses=1]
Implement rdar://6295824 and PR6724 with two tiny changes that can have a big effect :). The first is to enable the iterative SCC passmanager juice that kicks in when the scc passmgr detects that a function pass has devirtualized a call. In this case, it will rerun all the passes it manages on the SCC, up to the iteration count limit (4). This is useful because a function pass may devirualize a call, and we want the inliner to inline it, or pruneeh to infer stuff about it, etc. The second patch is to add *all* call sites to the DevirtualizedCalls list the inliner uses. This list is about to get renamed, but the jist of this is that the inliner now reconsiders *all* inlined call sites as candidates for further inlining. The intuition is this that in cases like this: f() { g(1); } g(int x) { h(x); } We analyze this bottom up, and may decide that it isn't profitable to inline H into G. Next step, we decide that it is profitable to inline G into F, and do so, which means that F now calls H. Even though the call from G -> H may not have been profitable to inline, the call from F -> H may be (in this case because a constant allows folding etc). In my spot checks, this doesn't have a big impact on code. For example, the LLC output for 252.eon grew from 0.02% (from 317252 to 317308) and 176.gcc actually shrunk by .3% (from 1525612 to 1520964 bytes). 252.eon never iterated in the SCC Passmgr, 176.gcc iterated at most 1 time. llvm-svn: 102823
2010-05-01 03:15:56 +02:00
@_ZTS1B = linkonce_odr constant [3 x i8] c"1B\00" ; <[3 x i8]*> [#uses=1]
[opaque pointer type] Add textual IR support for explicit type parameter to gep operator Similar to gep (r230786) and load (r230794) changes. Similar migration script can be used to update test cases, which successfully migrated all of LLVM and Polly, but about 4 test cases needed manually changes in Clang. (this script will read the contents of stdin and massage it into stdout - wrap it in the 'apply.sh' script shown in previous commits + xargs to apply it over a large set of test cases) import fileinput import sys import re rep = re.compile(r"(getelementptr(?:\s+inbounds)?\s*\()((<\d*\s+x\s+)?([^@]*?)(|\s*addrspace\(\d+\))\s*\*(?(3)>)\s*)(?=$|%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|zeroinitializer|<|\[\[[a-zA-Z]|\{\{)", re.MULTILINE | re.DOTALL) def conv(match): line = match.group(1) line += match.group(4) line += ", " line += match.group(2) return line line = sys.stdin.read() off = 0 for match in re.finditer(rep, line): sys.stdout.write(line[off:match.start()]) sys.stdout.write(conv(match)) off = match.end() sys.stdout.write(line[off:]) llvm-svn: 232184
2015-03-13 19:20:45 +01:00
@_ZTI1B = linkonce_odr constant %0 { i8* bitcast (i8** getelementptr inbounds (i8*, i8** @_ZTVN10__cxxabiv117__class_type_infoE, i64 2) to i8*), i8* getelementptr inbounds ([3 x i8], [3 x i8]* @_ZTS1B, i32 0, i32 0) } ; <%0*> [#uses=1]
@_ZTI1C = linkonce_odr constant %1 { i8* bitcast (i8** getelementptr inbounds (i8*, i8** @_ZTVN10__cxxabiv121__vmi_class_type_infoE, i64 2) to i8*), i8* getelementptr inbounds ([3 x i8], [3 x i8]* @_ZTS1C, i32 0, i32 0), i32 0, i32 2, i8* bitcast (%0* @_ZTI1A to i8*), i64 2, i8* bitcast (%0* @_ZTI1B to i8*), i64 2050 } ; <%1*> [#uses=1]
@_ZTI1D = linkonce_odr constant %2 { i8* bitcast (i8** getelementptr inbounds (i8*, i8** @_ZTVN10__cxxabiv120__si_class_type_infoE, i64 2) to i8*), i8* getelementptr inbounds ([3 x i8], [3 x i8]* @_ZTS1D, i32 0, i32 0), i8* bitcast (%1* @_ZTI1C to i8*) } ; <%2*> [#uses=1]
Implement rdar://6295824 and PR6724 with two tiny changes that can have a big effect :). The first is to enable the iterative SCC passmanager juice that kicks in when the scc passmgr detects that a function pass has devirtualized a call. In this case, it will rerun all the passes it manages on the SCC, up to the iteration count limit (4). This is useful because a function pass may devirualize a call, and we want the inliner to inline it, or pruneeh to infer stuff about it, etc. The second patch is to add *all* call sites to the DevirtualizedCalls list the inliner uses. This list is about to get renamed, but the jist of this is that the inliner now reconsiders *all* inlined call sites as candidates for further inlining. The intuition is this that in cases like this: f() { g(1); } g(int x) { h(x); } We analyze this bottom up, and may decide that it isn't profitable to inline H into G. Next step, we decide that it is profitable to inline G into F, and do so, which means that F now calls H. Even though the call from G -> H may not have been profitable to inline, the call from F -> H may be (in this case because a constant allows folding etc). In my spot checks, this doesn't have a big impact on code. For example, the LLC output for 252.eon grew from 0.02% (from 317252 to 317308) and 176.gcc actually shrunk by .3% (from 1525612 to 1520964 bytes). 252.eon never iterated in the SCC Passmgr, 176.gcc iterated at most 1 time. llvm-svn: 102823
2010-05-01 03:15:56 +02:00
@_ZTV1C = linkonce_odr constant [6 x i8*] [i8* null, i8* bitcast (%1* @_ZTI1C to i8*), i8* bitcast (i32 (%struct.C*)* @_ZN1C1fEv to i8*), i8* inttoptr (i64 -8 to i8*), i8* bitcast (%1* @_ZTI1C to i8*), i8* bitcast (i32 (%struct.C*)* @_ZThn8_N1C1fEv to i8*)] ; <[6 x i8*]*> [#uses=2]
@_ZTV1B = linkonce_odr constant [3 x i8*] [i8* null, i8* bitcast (%0* @_ZTI1B to i8*), i8* bitcast (i32 (%struct.A*)* @_ZN1B1fEv to i8*)] ; <[3 x i8*]*> [#uses=1]
@_ZTV1A = linkonce_odr constant [3 x i8*] [i8* null, i8* bitcast (%0* @_ZTI1A to i8*), i8* bitcast (i32 (%struct.A*)* @_ZN1A1fEv to i8*)] ; <[3 x i8*]*> [#uses=1]
define i32 @_Z1gv() ssp {
entry:
%d = alloca %struct.C, align 8 ; <%struct.C*> [#uses=2]
call void @_ZN1DC1Ev(%struct.C* %d)
%call = call i32 @_Z1fP1D(%struct.C* %d) ; <i32> [#uses=1]
%X = add i32 %call, 3
ret i32 %X
}
define linkonce_odr void @_ZN1DC1Ev(%struct.C* %this) inlinehint ssp align 2 {
entry:
call void @_ZN1DC2Ev(%struct.C* %this)
ret void
}
define internal i32 @_Z1fP1D(%struct.C* %d) ssp {
entry:
%0 = icmp eq %struct.C* %d, null ; <i1> [#uses=1]
br i1 %0, label %cast.end, label %cast.notnull
cast.notnull: ; preds = %entry
%1 = bitcast %struct.C* %d to i8* ; <i8*> [#uses=1]
[opaque pointer type] Add textual IR support for explicit type parameter to getelementptr instruction One of several parallel first steps to remove the target type of pointers, replacing them with a single opaque pointer type. This adds an explicit type parameter to the gep instruction so that when the first parameter becomes an opaque pointer type, the type to gep through is still available to the instructions. * This doesn't modify gep operators, only instructions (operators will be handled separately) * Textual IR changes only. Bitcode (including upgrade) and changing the in-memory representation will be in separate changes. * geps of vectors are transformed as: getelementptr <4 x float*> %x, ... ->getelementptr float, <4 x float*> %x, ... Then, once the opaque pointer type is introduced, this will ultimately look like: getelementptr float, <4 x ptr> %x with the unambiguous interpretation that it is a vector of pointers to float. * address spaces remain on the pointer, not the type: getelementptr float addrspace(1)* %x ->getelementptr float, float addrspace(1)* %x Then, eventually: getelementptr float, ptr addrspace(1) %x Importantly, the massive amount of test case churn has been automated by same crappy python code. I had to manually update a few test cases that wouldn't fit the script's model (r228970,r229196,r229197,r229198). The python script just massages stdin and writes the result to stdout, I then wrapped that in a shell script to handle replacing files, then using the usual find+xargs to migrate all the files. update.py: import fileinput import sys import re ibrep = re.compile(r"(^.*?[^%\w]getelementptr inbounds )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))") normrep = re.compile( r"(^.*?[^%\w]getelementptr )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))") def conv(match, line): if not match: return line line = match.groups()[0] if len(match.groups()[5]) == 0: line += match.groups()[2] line += match.groups()[3] line += ", " line += match.groups()[1] line += "\n" return line for line in sys.stdin: if line.find("getelementptr ") == line.find("getelementptr inbounds"): if line.find("getelementptr inbounds") != line.find("getelementptr inbounds ("): line = conv(re.match(ibrep, line), line) elif line.find("getelementptr ") != line.find("getelementptr ("): line = conv(re.match(normrep, line), line) sys.stdout.write(line) apply.sh: for name in "$@" do python3 `dirname "$0"`/update.py < "$name" > "$name.tmp" && mv "$name.tmp" "$name" rm -f "$name.tmp" done The actual commands: From llvm/src: find test/ -name *.ll | xargs ./apply.sh From llvm/src/tools/clang: find test/ -name *.mm -o -name *.m -o -name *.cpp -o -name *.c | xargs -I '{}' ../../apply.sh "{}" From llvm/src/tools/polly: find test/ -name *.ll | xargs ./apply.sh After that, check-all (with llvm, clang, clang-tools-extra, lld, compiler-rt, and polly all checked out). The extra 'rm' in the apply.sh script is due to a few files in clang's test suite using interesting unicode stuff that my python script was throwing exceptions on. None of those files needed to be migrated, so it seemed sufficient to ignore those cases. Reviewers: rafael, dexonsmith, grosser Differential Revision: http://reviews.llvm.org/D7636 llvm-svn: 230786
2015-02-27 20:29:02 +01:00
%add.ptr = getelementptr i8, i8* %1, i64 8 ; <i8*> [#uses=1]
Implement rdar://6295824 and PR6724 with two tiny changes that can have a big effect :). The first is to enable the iterative SCC passmanager juice that kicks in when the scc passmgr detects that a function pass has devirtualized a call. In this case, it will rerun all the passes it manages on the SCC, up to the iteration count limit (4). This is useful because a function pass may devirualize a call, and we want the inliner to inline it, or pruneeh to infer stuff about it, etc. The second patch is to add *all* call sites to the DevirtualizedCalls list the inliner uses. This list is about to get renamed, but the jist of this is that the inliner now reconsiders *all* inlined call sites as candidates for further inlining. The intuition is this that in cases like this: f() { g(1); } g(int x) { h(x); } We analyze this bottom up, and may decide that it isn't profitable to inline H into G. Next step, we decide that it is profitable to inline G into F, and do so, which means that F now calls H. Even though the call from G -> H may not have been profitable to inline, the call from F -> H may be (in this case because a constant allows folding etc). In my spot checks, this doesn't have a big impact on code. For example, the LLC output for 252.eon grew from 0.02% (from 317252 to 317308) and 176.gcc actually shrunk by .3% (from 1525612 to 1520964 bytes). 252.eon never iterated in the SCC Passmgr, 176.gcc iterated at most 1 time. llvm-svn: 102823
2010-05-01 03:15:56 +02:00
%2 = bitcast i8* %add.ptr to %struct.A* ; <%struct.A*> [#uses=1]
br label %cast.end
cast.end: ; preds = %entry, %cast.notnull
%3 = phi %struct.A* [ %2, %cast.notnull ], [ null, %entry ] ; <%struct.A*> [#uses=2]
%4 = bitcast %struct.A* %3 to i32 (%struct.A*)*** ; <i32 (%struct.A*)***> [#uses=1]
[opaque pointer type] Add textual IR support for explicit type parameter to load instruction Essentially the same as the GEP change in r230786. A similar migration script can be used to update test cases, though a few more test case improvements/changes were required this time around: (r229269-r229278) import fileinput import sys import re pat = re.compile(r"((?:=|:|^)\s*load (?:atomic )?(?:volatile )?(.*?))(| addrspace\(\d+\) *)\*($| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$)") for line in sys.stdin: sys.stdout.write(re.sub(pat, r"\1, \2\3*\4", line)) Reviewers: rafael, dexonsmith, grosser Differential Revision: http://reviews.llvm.org/D7649 llvm-svn: 230794
2015-02-27 22:17:42 +01:00
%5 = load i32 (%struct.A*)**, i32 (%struct.A*)*** %4 ; <i32 (%struct.A*)**> [#uses=1]
[opaque pointer type] Add textual IR support for explicit type parameter to getelementptr instruction One of several parallel first steps to remove the target type of pointers, replacing them with a single opaque pointer type. This adds an explicit type parameter to the gep instruction so that when the first parameter becomes an opaque pointer type, the type to gep through is still available to the instructions. * This doesn't modify gep operators, only instructions (operators will be handled separately) * Textual IR changes only. Bitcode (including upgrade) and changing the in-memory representation will be in separate changes. * geps of vectors are transformed as: getelementptr <4 x float*> %x, ... ->getelementptr float, <4 x float*> %x, ... Then, once the opaque pointer type is introduced, this will ultimately look like: getelementptr float, <4 x ptr> %x with the unambiguous interpretation that it is a vector of pointers to float. * address spaces remain on the pointer, not the type: getelementptr float addrspace(1)* %x ->getelementptr float, float addrspace(1)* %x Then, eventually: getelementptr float, ptr addrspace(1) %x Importantly, the massive amount of test case churn has been automated by same crappy python code. I had to manually update a few test cases that wouldn't fit the script's model (r228970,r229196,r229197,r229198). The python script just massages stdin and writes the result to stdout, I then wrapped that in a shell script to handle replacing files, then using the usual find+xargs to migrate all the files. update.py: import fileinput import sys import re ibrep = re.compile(r"(^.*?[^%\w]getelementptr inbounds )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))") normrep = re.compile( r"(^.*?[^%\w]getelementptr )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))") def conv(match, line): if not match: return line line = match.groups()[0] if len(match.groups()[5]) == 0: line += match.groups()[2] line += match.groups()[3] line += ", " line += match.groups()[1] line += "\n" return line for line in sys.stdin: if line.find("getelementptr ") == line.find("getelementptr inbounds"): if line.find("getelementptr inbounds") != line.find("getelementptr inbounds ("): line = conv(re.match(ibrep, line), line) elif line.find("getelementptr ") != line.find("getelementptr ("): line = conv(re.match(normrep, line), line) sys.stdout.write(line) apply.sh: for name in "$@" do python3 `dirname "$0"`/update.py < "$name" > "$name.tmp" && mv "$name.tmp" "$name" rm -f "$name.tmp" done The actual commands: From llvm/src: find test/ -name *.ll | xargs ./apply.sh From llvm/src/tools/clang: find test/ -name *.mm -o -name *.m -o -name *.cpp -o -name *.c | xargs -I '{}' ../../apply.sh "{}" From llvm/src/tools/polly: find test/ -name *.ll | xargs ./apply.sh After that, check-all (with llvm, clang, clang-tools-extra, lld, compiler-rt, and polly all checked out). The extra 'rm' in the apply.sh script is due to a few files in clang's test suite using interesting unicode stuff that my python script was throwing exceptions on. None of those files needed to be migrated, so it seemed sufficient to ignore those cases. Reviewers: rafael, dexonsmith, grosser Differential Revision: http://reviews.llvm.org/D7636 llvm-svn: 230786
2015-02-27 20:29:02 +01:00
%vfn = getelementptr inbounds i32 (%struct.A*)*, i32 (%struct.A*)** %5, i64 0 ; <i32 (%struct.A*)**> [#uses=1]
[opaque pointer type] Add textual IR support for explicit type parameter to load instruction Essentially the same as the GEP change in r230786. A similar migration script can be used to update test cases, though a few more test case improvements/changes were required this time around: (r229269-r229278) import fileinput import sys import re pat = re.compile(r"((?:=|:|^)\s*load (?:atomic )?(?:volatile )?(.*?))(| addrspace\(\d+\) *)\*($| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$)") for line in sys.stdin: sys.stdout.write(re.sub(pat, r"\1, \2\3*\4", line)) Reviewers: rafael, dexonsmith, grosser Differential Revision: http://reviews.llvm.org/D7649 llvm-svn: 230794
2015-02-27 22:17:42 +01:00
%6 = load i32 (%struct.A*)*, i32 (%struct.A*)** %vfn ; <i32 (%struct.A*)*> [#uses=1]
Implement rdar://6295824 and PR6724 with two tiny changes that can have a big effect :). The first is to enable the iterative SCC passmanager juice that kicks in when the scc passmgr detects that a function pass has devirtualized a call. In this case, it will rerun all the passes it manages on the SCC, up to the iteration count limit (4). This is useful because a function pass may devirualize a call, and we want the inliner to inline it, or pruneeh to infer stuff about it, etc. The second patch is to add *all* call sites to the DevirtualizedCalls list the inliner uses. This list is about to get renamed, but the jist of this is that the inliner now reconsiders *all* inlined call sites as candidates for further inlining. The intuition is this that in cases like this: f() { g(1); } g(int x) { h(x); } We analyze this bottom up, and may decide that it isn't profitable to inline H into G. Next step, we decide that it is profitable to inline G into F, and do so, which means that F now calls H. Even though the call from G -> H may not have been profitable to inline, the call from F -> H may be (in this case because a constant allows folding etc). In my spot checks, this doesn't have a big impact on code. For example, the LLC output for 252.eon grew from 0.02% (from 317252 to 317308) and 176.gcc actually shrunk by .3% (from 1525612 to 1520964 bytes). 252.eon never iterated in the SCC Passmgr, 176.gcc iterated at most 1 time. llvm-svn: 102823
2010-05-01 03:15:56 +02:00
%call = call i32 %6(%struct.A* %3) ; <i32> [#uses=1]
ret i32 %call
}
define linkonce_odr i32 @_ZN1D1fEv(%struct.C* %this) ssp align 2 {
entry:
ret i32 4
}
define linkonce_odr i32 @_ZThn8_N1D1fEv(%struct.C* %this) {
entry:
%0 = bitcast %struct.C* %this to i8* ; <i8*> [#uses=1]
[opaque pointer type] Add textual IR support for explicit type parameter to getelementptr instruction One of several parallel first steps to remove the target type of pointers, replacing them with a single opaque pointer type. This adds an explicit type parameter to the gep instruction so that when the first parameter becomes an opaque pointer type, the type to gep through is still available to the instructions. * This doesn't modify gep operators, only instructions (operators will be handled separately) * Textual IR changes only. Bitcode (including upgrade) and changing the in-memory representation will be in separate changes. * geps of vectors are transformed as: getelementptr <4 x float*> %x, ... ->getelementptr float, <4 x float*> %x, ... Then, once the opaque pointer type is introduced, this will ultimately look like: getelementptr float, <4 x ptr> %x with the unambiguous interpretation that it is a vector of pointers to float. * address spaces remain on the pointer, not the type: getelementptr float addrspace(1)* %x ->getelementptr float, float addrspace(1)* %x Then, eventually: getelementptr float, ptr addrspace(1) %x Importantly, the massive amount of test case churn has been automated by same crappy python code. I had to manually update a few test cases that wouldn't fit the script's model (r228970,r229196,r229197,r229198). The python script just massages stdin and writes the result to stdout, I then wrapped that in a shell script to handle replacing files, then using the usual find+xargs to migrate all the files. update.py: import fileinput import sys import re ibrep = re.compile(r"(^.*?[^%\w]getelementptr inbounds )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))") normrep = re.compile( r"(^.*?[^%\w]getelementptr )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))") def conv(match, line): if not match: return line line = match.groups()[0] if len(match.groups()[5]) == 0: line += match.groups()[2] line += match.groups()[3] line += ", " line += match.groups()[1] line += "\n" return line for line in sys.stdin: if line.find("getelementptr ") == line.find("getelementptr inbounds"): if line.find("getelementptr inbounds") != line.find("getelementptr inbounds ("): line = conv(re.match(ibrep, line), line) elif line.find("getelementptr ") != line.find("getelementptr ("): line = conv(re.match(normrep, line), line) sys.stdout.write(line) apply.sh: for name in "$@" do python3 `dirname "$0"`/update.py < "$name" > "$name.tmp" && mv "$name.tmp" "$name" rm -f "$name.tmp" done The actual commands: From llvm/src: find test/ -name *.ll | xargs ./apply.sh From llvm/src/tools/clang: find test/ -name *.mm -o -name *.m -o -name *.cpp -o -name *.c | xargs -I '{}' ../../apply.sh "{}" From llvm/src/tools/polly: find test/ -name *.ll | xargs ./apply.sh After that, check-all (with llvm, clang, clang-tools-extra, lld, compiler-rt, and polly all checked out). The extra 'rm' in the apply.sh script is due to a few files in clang's test suite using interesting unicode stuff that my python script was throwing exceptions on. None of those files needed to be migrated, so it seemed sufficient to ignore those cases. Reviewers: rafael, dexonsmith, grosser Differential Revision: http://reviews.llvm.org/D7636 llvm-svn: 230786
2015-02-27 20:29:02 +01:00
%1 = getelementptr inbounds i8, i8* %0, i64 -8 ; <i8*> [#uses=1]
Implement rdar://6295824 and PR6724 with two tiny changes that can have a big effect :). The first is to enable the iterative SCC passmanager juice that kicks in when the scc passmgr detects that a function pass has devirtualized a call. In this case, it will rerun all the passes it manages on the SCC, up to the iteration count limit (4). This is useful because a function pass may devirualize a call, and we want the inliner to inline it, or pruneeh to infer stuff about it, etc. The second patch is to add *all* call sites to the DevirtualizedCalls list the inliner uses. This list is about to get renamed, but the jist of this is that the inliner now reconsiders *all* inlined call sites as candidates for further inlining. The intuition is this that in cases like this: f() { g(1); } g(int x) { h(x); } We analyze this bottom up, and may decide that it isn't profitable to inline H into G. Next step, we decide that it is profitable to inline G into F, and do so, which means that F now calls H. Even though the call from G -> H may not have been profitable to inline, the call from F -> H may be (in this case because a constant allows folding etc). In my spot checks, this doesn't have a big impact on code. For example, the LLC output for 252.eon grew from 0.02% (from 317252 to 317308) and 176.gcc actually shrunk by .3% (from 1525612 to 1520964 bytes). 252.eon never iterated in the SCC Passmgr, 176.gcc iterated at most 1 time. llvm-svn: 102823
2010-05-01 03:15:56 +02:00
%2 = bitcast i8* %1 to %struct.C* ; <%struct.C*> [#uses=1]
%call = call i32 @_ZN1D1fEv(%struct.C* %2) ; <i32> [#uses=1]
ret i32 %call
}
define linkonce_odr void @_ZN1DC2Ev(%struct.C* %this) inlinehint ssp align 2 {
entry:
call void @_ZN1CC2Ev(%struct.C* %this)
%0 = bitcast %struct.C* %this to i8* ; <i8*> [#uses=1]
[opaque pointer type] Add textual IR support for explicit type parameter to getelementptr instruction One of several parallel first steps to remove the target type of pointers, replacing them with a single opaque pointer type. This adds an explicit type parameter to the gep instruction so that when the first parameter becomes an opaque pointer type, the type to gep through is still available to the instructions. * This doesn't modify gep operators, only instructions (operators will be handled separately) * Textual IR changes only. Bitcode (including upgrade) and changing the in-memory representation will be in separate changes. * geps of vectors are transformed as: getelementptr <4 x float*> %x, ... ->getelementptr float, <4 x float*> %x, ... Then, once the opaque pointer type is introduced, this will ultimately look like: getelementptr float, <4 x ptr> %x with the unambiguous interpretation that it is a vector of pointers to float. * address spaces remain on the pointer, not the type: getelementptr float addrspace(1)* %x ->getelementptr float, float addrspace(1)* %x Then, eventually: getelementptr float, ptr addrspace(1) %x Importantly, the massive amount of test case churn has been automated by same crappy python code. I had to manually update a few test cases that wouldn't fit the script's model (r228970,r229196,r229197,r229198). The python script just massages stdin and writes the result to stdout, I then wrapped that in a shell script to handle replacing files, then using the usual find+xargs to migrate all the files. update.py: import fileinput import sys import re ibrep = re.compile(r"(^.*?[^%\w]getelementptr inbounds )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))") normrep = re.compile( r"(^.*?[^%\w]getelementptr )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))") def conv(match, line): if not match: return line line = match.groups()[0] if len(match.groups()[5]) == 0: line += match.groups()[2] line += match.groups()[3] line += ", " line += match.groups()[1] line += "\n" return line for line in sys.stdin: if line.find("getelementptr ") == line.find("getelementptr inbounds"): if line.find("getelementptr inbounds") != line.find("getelementptr inbounds ("): line = conv(re.match(ibrep, line), line) elif line.find("getelementptr ") != line.find("getelementptr ("): line = conv(re.match(normrep, line), line) sys.stdout.write(line) apply.sh: for name in "$@" do python3 `dirname "$0"`/update.py < "$name" > "$name.tmp" && mv "$name.tmp" "$name" rm -f "$name.tmp" done The actual commands: From llvm/src: find test/ -name *.ll | xargs ./apply.sh From llvm/src/tools/clang: find test/ -name *.mm -o -name *.m -o -name *.cpp -o -name *.c | xargs -I '{}' ../../apply.sh "{}" From llvm/src/tools/polly: find test/ -name *.ll | xargs ./apply.sh After that, check-all (with llvm, clang, clang-tools-extra, lld, compiler-rt, and polly all checked out). The extra 'rm' in the apply.sh script is due to a few files in clang's test suite using interesting unicode stuff that my python script was throwing exceptions on. None of those files needed to be migrated, so it seemed sufficient to ignore those cases. Reviewers: rafael, dexonsmith, grosser Differential Revision: http://reviews.llvm.org/D7636 llvm-svn: 230786
2015-02-27 20:29:02 +01:00
%1 = getelementptr inbounds i8, i8* %0, i64 0 ; <i8*> [#uses=1]
Implement rdar://6295824 and PR6724 with two tiny changes that can have a big effect :). The first is to enable the iterative SCC passmanager juice that kicks in when the scc passmgr detects that a function pass has devirtualized a call. In this case, it will rerun all the passes it manages on the SCC, up to the iteration count limit (4). This is useful because a function pass may devirualize a call, and we want the inliner to inline it, or pruneeh to infer stuff about it, etc. The second patch is to add *all* call sites to the DevirtualizedCalls list the inliner uses. This list is about to get renamed, but the jist of this is that the inliner now reconsiders *all* inlined call sites as candidates for further inlining. The intuition is this that in cases like this: f() { g(1); } g(int x) { h(x); } We analyze this bottom up, and may decide that it isn't profitable to inline H into G. Next step, we decide that it is profitable to inline G into F, and do so, which means that F now calls H. Even though the call from G -> H may not have been profitable to inline, the call from F -> H may be (in this case because a constant allows folding etc). In my spot checks, this doesn't have a big impact on code. For example, the LLC output for 252.eon grew from 0.02% (from 317252 to 317308) and 176.gcc actually shrunk by .3% (from 1525612 to 1520964 bytes). 252.eon never iterated in the SCC Passmgr, 176.gcc iterated at most 1 time. llvm-svn: 102823
2010-05-01 03:15:56 +02:00
%2 = bitcast i8* %1 to i8*** ; <i8***> [#uses=1]
[opaque pointer type] Add textual IR support for explicit type parameter to gep operator Similar to gep (r230786) and load (r230794) changes. Similar migration script can be used to update test cases, which successfully migrated all of LLVM and Polly, but about 4 test cases needed manually changes in Clang. (this script will read the contents of stdin and massage it into stdout - wrap it in the 'apply.sh' script shown in previous commits + xargs to apply it over a large set of test cases) import fileinput import sys import re rep = re.compile(r"(getelementptr(?:\s+inbounds)?\s*\()((<\d*\s+x\s+)?([^@]*?)(|\s*addrspace\(\d+\))\s*\*(?(3)>)\s*)(?=$|%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|zeroinitializer|<|\[\[[a-zA-Z]|\{\{)", re.MULTILINE | re.DOTALL) def conv(match): line = match.group(1) line += match.group(4) line += ", " line += match.group(2) return line line = sys.stdin.read() off = 0 for match in re.finditer(rep, line): sys.stdout.write(line[off:match.start()]) sys.stdout.write(conv(match)) off = match.end() sys.stdout.write(line[off:]) llvm-svn: 232184
2015-03-13 19:20:45 +01:00
store i8** getelementptr inbounds ([6 x i8*], [6 x i8*]* @_ZTV1D, i64 0, i64 2), i8*** %2
Implement rdar://6295824 and PR6724 with two tiny changes that can have a big effect :). The first is to enable the iterative SCC passmanager juice that kicks in when the scc passmgr detects that a function pass has devirtualized a call. In this case, it will rerun all the passes it manages on the SCC, up to the iteration count limit (4). This is useful because a function pass may devirualize a call, and we want the inliner to inline it, or pruneeh to infer stuff about it, etc. The second patch is to add *all* call sites to the DevirtualizedCalls list the inliner uses. This list is about to get renamed, but the jist of this is that the inliner now reconsiders *all* inlined call sites as candidates for further inlining. The intuition is this that in cases like this: f() { g(1); } g(int x) { h(x); } We analyze this bottom up, and may decide that it isn't profitable to inline H into G. Next step, we decide that it is profitable to inline G into F, and do so, which means that F now calls H. Even though the call from G -> H may not have been profitable to inline, the call from F -> H may be (in this case because a constant allows folding etc). In my spot checks, this doesn't have a big impact on code. For example, the LLC output for 252.eon grew from 0.02% (from 317252 to 317308) and 176.gcc actually shrunk by .3% (from 1525612 to 1520964 bytes). 252.eon never iterated in the SCC Passmgr, 176.gcc iterated at most 1 time. llvm-svn: 102823
2010-05-01 03:15:56 +02:00
%3 = bitcast %struct.C* %this to i8* ; <i8*> [#uses=1]
[opaque pointer type] Add textual IR support for explicit type parameter to getelementptr instruction One of several parallel first steps to remove the target type of pointers, replacing them with a single opaque pointer type. This adds an explicit type parameter to the gep instruction so that when the first parameter becomes an opaque pointer type, the type to gep through is still available to the instructions. * This doesn't modify gep operators, only instructions (operators will be handled separately) * Textual IR changes only. Bitcode (including upgrade) and changing the in-memory representation will be in separate changes. * geps of vectors are transformed as: getelementptr <4 x float*> %x, ... ->getelementptr float, <4 x float*> %x, ... Then, once the opaque pointer type is introduced, this will ultimately look like: getelementptr float, <4 x ptr> %x with the unambiguous interpretation that it is a vector of pointers to float. * address spaces remain on the pointer, not the type: getelementptr float addrspace(1)* %x ->getelementptr float, float addrspace(1)* %x Then, eventually: getelementptr float, ptr addrspace(1) %x Importantly, the massive amount of test case churn has been automated by same crappy python code. I had to manually update a few test cases that wouldn't fit the script's model (r228970,r229196,r229197,r229198). The python script just massages stdin and writes the result to stdout, I then wrapped that in a shell script to handle replacing files, then using the usual find+xargs to migrate all the files. update.py: import fileinput import sys import re ibrep = re.compile(r"(^.*?[^%\w]getelementptr inbounds )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))") normrep = re.compile( r"(^.*?[^%\w]getelementptr )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))") def conv(match, line): if not match: return line line = match.groups()[0] if len(match.groups()[5]) == 0: line += match.groups()[2] line += match.groups()[3] line += ", " line += match.groups()[1] line += "\n" return line for line in sys.stdin: if line.find("getelementptr ") == line.find("getelementptr inbounds"): if line.find("getelementptr inbounds") != line.find("getelementptr inbounds ("): line = conv(re.match(ibrep, line), line) elif line.find("getelementptr ") != line.find("getelementptr ("): line = conv(re.match(normrep, line), line) sys.stdout.write(line) apply.sh: for name in "$@" do python3 `dirname "$0"`/update.py < "$name" > "$name.tmp" && mv "$name.tmp" "$name" rm -f "$name.tmp" done The actual commands: From llvm/src: find test/ -name *.ll | xargs ./apply.sh From llvm/src/tools/clang: find test/ -name *.mm -o -name *.m -o -name *.cpp -o -name *.c | xargs -I '{}' ../../apply.sh "{}" From llvm/src/tools/polly: find test/ -name *.ll | xargs ./apply.sh After that, check-all (with llvm, clang, clang-tools-extra, lld, compiler-rt, and polly all checked out). The extra 'rm' in the apply.sh script is due to a few files in clang's test suite using interesting unicode stuff that my python script was throwing exceptions on. None of those files needed to be migrated, so it seemed sufficient to ignore those cases. Reviewers: rafael, dexonsmith, grosser Differential Revision: http://reviews.llvm.org/D7636 llvm-svn: 230786
2015-02-27 20:29:02 +01:00
%4 = getelementptr inbounds i8, i8* %3, i64 8 ; <i8*> [#uses=1]
Implement rdar://6295824 and PR6724 with two tiny changes that can have a big effect :). The first is to enable the iterative SCC passmanager juice that kicks in when the scc passmgr detects that a function pass has devirtualized a call. In this case, it will rerun all the passes it manages on the SCC, up to the iteration count limit (4). This is useful because a function pass may devirualize a call, and we want the inliner to inline it, or pruneeh to infer stuff about it, etc. The second patch is to add *all* call sites to the DevirtualizedCalls list the inliner uses. This list is about to get renamed, but the jist of this is that the inliner now reconsiders *all* inlined call sites as candidates for further inlining. The intuition is this that in cases like this: f() { g(1); } g(int x) { h(x); } We analyze this bottom up, and may decide that it isn't profitable to inline H into G. Next step, we decide that it is profitable to inline G into F, and do so, which means that F now calls H. Even though the call from G -> H may not have been profitable to inline, the call from F -> H may be (in this case because a constant allows folding etc). In my spot checks, this doesn't have a big impact on code. For example, the LLC output for 252.eon grew from 0.02% (from 317252 to 317308) and 176.gcc actually shrunk by .3% (from 1525612 to 1520964 bytes). 252.eon never iterated in the SCC Passmgr, 176.gcc iterated at most 1 time. llvm-svn: 102823
2010-05-01 03:15:56 +02:00
%5 = bitcast i8* %4 to i8*** ; <i8***> [#uses=1]
[opaque pointer type] Add textual IR support for explicit type parameter to gep operator Similar to gep (r230786) and load (r230794) changes. Similar migration script can be used to update test cases, which successfully migrated all of LLVM and Polly, but about 4 test cases needed manually changes in Clang. (this script will read the contents of stdin and massage it into stdout - wrap it in the 'apply.sh' script shown in previous commits + xargs to apply it over a large set of test cases) import fileinput import sys import re rep = re.compile(r"(getelementptr(?:\s+inbounds)?\s*\()((<\d*\s+x\s+)?([^@]*?)(|\s*addrspace\(\d+\))\s*\*(?(3)>)\s*)(?=$|%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|zeroinitializer|<|\[\[[a-zA-Z]|\{\{)", re.MULTILINE | re.DOTALL) def conv(match): line = match.group(1) line += match.group(4) line += ", " line += match.group(2) return line line = sys.stdin.read() off = 0 for match in re.finditer(rep, line): sys.stdout.write(line[off:match.start()]) sys.stdout.write(conv(match)) off = match.end() sys.stdout.write(line[off:]) llvm-svn: 232184
2015-03-13 19:20:45 +01:00
store i8** getelementptr inbounds ([6 x i8*], [6 x i8*]* @_ZTV1D, i64 0, i64 5), i8*** %5
Implement rdar://6295824 and PR6724 with two tiny changes that can have a big effect :). The first is to enable the iterative SCC passmanager juice that kicks in when the scc passmgr detects that a function pass has devirtualized a call. In this case, it will rerun all the passes it manages on the SCC, up to the iteration count limit (4). This is useful because a function pass may devirualize a call, and we want the inliner to inline it, or pruneeh to infer stuff about it, etc. The second patch is to add *all* call sites to the DevirtualizedCalls list the inliner uses. This list is about to get renamed, but the jist of this is that the inliner now reconsiders *all* inlined call sites as candidates for further inlining. The intuition is this that in cases like this: f() { g(1); } g(int x) { h(x); } We analyze this bottom up, and may decide that it isn't profitable to inline H into G. Next step, we decide that it is profitable to inline G into F, and do so, which means that F now calls H. Even though the call from G -> H may not have been profitable to inline, the call from F -> H may be (in this case because a constant allows folding etc). In my spot checks, this doesn't have a big impact on code. For example, the LLC output for 252.eon grew from 0.02% (from 317252 to 317308) and 176.gcc actually shrunk by .3% (from 1525612 to 1520964 bytes). 252.eon never iterated in the SCC Passmgr, 176.gcc iterated at most 1 time. llvm-svn: 102823
2010-05-01 03:15:56 +02:00
ret void
}
define linkonce_odr void @_ZN1CC2Ev(%struct.C* %this) inlinehint ssp align 2 {
entry:
%0 = bitcast %struct.C* %this to %struct.A* ; <%struct.A*> [#uses=1]
call void @_ZN1AC2Ev(%struct.A* %0)
%1 = bitcast %struct.C* %this to i8* ; <i8*> [#uses=1]
[opaque pointer type] Add textual IR support for explicit type parameter to getelementptr instruction One of several parallel first steps to remove the target type of pointers, replacing them with a single opaque pointer type. This adds an explicit type parameter to the gep instruction so that when the first parameter becomes an opaque pointer type, the type to gep through is still available to the instructions. * This doesn't modify gep operators, only instructions (operators will be handled separately) * Textual IR changes only. Bitcode (including upgrade) and changing the in-memory representation will be in separate changes. * geps of vectors are transformed as: getelementptr <4 x float*> %x, ... ->getelementptr float, <4 x float*> %x, ... Then, once the opaque pointer type is introduced, this will ultimately look like: getelementptr float, <4 x ptr> %x with the unambiguous interpretation that it is a vector of pointers to float. * address spaces remain on the pointer, not the type: getelementptr float addrspace(1)* %x ->getelementptr float, float addrspace(1)* %x Then, eventually: getelementptr float, ptr addrspace(1) %x Importantly, the massive amount of test case churn has been automated by same crappy python code. I had to manually update a few test cases that wouldn't fit the script's model (r228970,r229196,r229197,r229198). The python script just massages stdin and writes the result to stdout, I then wrapped that in a shell script to handle replacing files, then using the usual find+xargs to migrate all the files. update.py: import fileinput import sys import re ibrep = re.compile(r"(^.*?[^%\w]getelementptr inbounds )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))") normrep = re.compile( r"(^.*?[^%\w]getelementptr )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))") def conv(match, line): if not match: return line line = match.groups()[0] if len(match.groups()[5]) == 0: line += match.groups()[2] line += match.groups()[3] line += ", " line += match.groups()[1] line += "\n" return line for line in sys.stdin: if line.find("getelementptr ") == line.find("getelementptr inbounds"): if line.find("getelementptr inbounds") != line.find("getelementptr inbounds ("): line = conv(re.match(ibrep, line), line) elif line.find("getelementptr ") != line.find("getelementptr ("): line = conv(re.match(normrep, line), line) sys.stdout.write(line) apply.sh: for name in "$@" do python3 `dirname "$0"`/update.py < "$name" > "$name.tmp" && mv "$name.tmp" "$name" rm -f "$name.tmp" done The actual commands: From llvm/src: find test/ -name *.ll | xargs ./apply.sh From llvm/src/tools/clang: find test/ -name *.mm -o -name *.m -o -name *.cpp -o -name *.c | xargs -I '{}' ../../apply.sh "{}" From llvm/src/tools/polly: find test/ -name *.ll | xargs ./apply.sh After that, check-all (with llvm, clang, clang-tools-extra, lld, compiler-rt, and polly all checked out). The extra 'rm' in the apply.sh script is due to a few files in clang's test suite using interesting unicode stuff that my python script was throwing exceptions on. None of those files needed to be migrated, so it seemed sufficient to ignore those cases. Reviewers: rafael, dexonsmith, grosser Differential Revision: http://reviews.llvm.org/D7636 llvm-svn: 230786
2015-02-27 20:29:02 +01:00
%2 = getelementptr inbounds i8, i8* %1, i64 8 ; <i8*> [#uses=1]
Implement rdar://6295824 and PR6724 with two tiny changes that can have a big effect :). The first is to enable the iterative SCC passmanager juice that kicks in when the scc passmgr detects that a function pass has devirtualized a call. In this case, it will rerun all the passes it manages on the SCC, up to the iteration count limit (4). This is useful because a function pass may devirualize a call, and we want the inliner to inline it, or pruneeh to infer stuff about it, etc. The second patch is to add *all* call sites to the DevirtualizedCalls list the inliner uses. This list is about to get renamed, but the jist of this is that the inliner now reconsiders *all* inlined call sites as candidates for further inlining. The intuition is this that in cases like this: f() { g(1); } g(int x) { h(x); } We analyze this bottom up, and may decide that it isn't profitable to inline H into G. Next step, we decide that it is profitable to inline G into F, and do so, which means that F now calls H. Even though the call from G -> H may not have been profitable to inline, the call from F -> H may be (in this case because a constant allows folding etc). In my spot checks, this doesn't have a big impact on code. For example, the LLC output for 252.eon grew from 0.02% (from 317252 to 317308) and 176.gcc actually shrunk by .3% (from 1525612 to 1520964 bytes). 252.eon never iterated in the SCC Passmgr, 176.gcc iterated at most 1 time. llvm-svn: 102823
2010-05-01 03:15:56 +02:00
%3 = bitcast i8* %2 to %struct.A* ; <%struct.A*> [#uses=1]
call void @_ZN1BC2Ev(%struct.A* %3)
%4 = bitcast %struct.C* %this to i8* ; <i8*> [#uses=1]
[opaque pointer type] Add textual IR support for explicit type parameter to getelementptr instruction One of several parallel first steps to remove the target type of pointers, replacing them with a single opaque pointer type. This adds an explicit type parameter to the gep instruction so that when the first parameter becomes an opaque pointer type, the type to gep through is still available to the instructions. * This doesn't modify gep operators, only instructions (operators will be handled separately) * Textual IR changes only. Bitcode (including upgrade) and changing the in-memory representation will be in separate changes. * geps of vectors are transformed as: getelementptr <4 x float*> %x, ... ->getelementptr float, <4 x float*> %x, ... Then, once the opaque pointer type is introduced, this will ultimately look like: getelementptr float, <4 x ptr> %x with the unambiguous interpretation that it is a vector of pointers to float. * address spaces remain on the pointer, not the type: getelementptr float addrspace(1)* %x ->getelementptr float, float addrspace(1)* %x Then, eventually: getelementptr float, ptr addrspace(1) %x Importantly, the massive amount of test case churn has been automated by same crappy python code. I had to manually update a few test cases that wouldn't fit the script's model (r228970,r229196,r229197,r229198). The python script just massages stdin and writes the result to stdout, I then wrapped that in a shell script to handle replacing files, then using the usual find+xargs to migrate all the files. update.py: import fileinput import sys import re ibrep = re.compile(r"(^.*?[^%\w]getelementptr inbounds )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))") normrep = re.compile( r"(^.*?[^%\w]getelementptr )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))") def conv(match, line): if not match: return line line = match.groups()[0] if len(match.groups()[5]) == 0: line += match.groups()[2] line += match.groups()[3] line += ", " line += match.groups()[1] line += "\n" return line for line in sys.stdin: if line.find("getelementptr ") == line.find("getelementptr inbounds"): if line.find("getelementptr inbounds") != line.find("getelementptr inbounds ("): line = conv(re.match(ibrep, line), line) elif line.find("getelementptr ") != line.find("getelementptr ("): line = conv(re.match(normrep, line), line) sys.stdout.write(line) apply.sh: for name in "$@" do python3 `dirname "$0"`/update.py < "$name" > "$name.tmp" && mv "$name.tmp" "$name" rm -f "$name.tmp" done The actual commands: From llvm/src: find test/ -name *.ll | xargs ./apply.sh From llvm/src/tools/clang: find test/ -name *.mm -o -name *.m -o -name *.cpp -o -name *.c | xargs -I '{}' ../../apply.sh "{}" From llvm/src/tools/polly: find test/ -name *.ll | xargs ./apply.sh After that, check-all (with llvm, clang, clang-tools-extra, lld, compiler-rt, and polly all checked out). The extra 'rm' in the apply.sh script is due to a few files in clang's test suite using interesting unicode stuff that my python script was throwing exceptions on. None of those files needed to be migrated, so it seemed sufficient to ignore those cases. Reviewers: rafael, dexonsmith, grosser Differential Revision: http://reviews.llvm.org/D7636 llvm-svn: 230786
2015-02-27 20:29:02 +01:00
%5 = getelementptr inbounds i8, i8* %4, i64 0 ; <i8*> [#uses=1]
Implement rdar://6295824 and PR6724 with two tiny changes that can have a big effect :). The first is to enable the iterative SCC passmanager juice that kicks in when the scc passmgr detects that a function pass has devirtualized a call. In this case, it will rerun all the passes it manages on the SCC, up to the iteration count limit (4). This is useful because a function pass may devirualize a call, and we want the inliner to inline it, or pruneeh to infer stuff about it, etc. The second patch is to add *all* call sites to the DevirtualizedCalls list the inliner uses. This list is about to get renamed, but the jist of this is that the inliner now reconsiders *all* inlined call sites as candidates for further inlining. The intuition is this that in cases like this: f() { g(1); } g(int x) { h(x); } We analyze this bottom up, and may decide that it isn't profitable to inline H into G. Next step, we decide that it is profitable to inline G into F, and do so, which means that F now calls H. Even though the call from G -> H may not have been profitable to inline, the call from F -> H may be (in this case because a constant allows folding etc). In my spot checks, this doesn't have a big impact on code. For example, the LLC output for 252.eon grew from 0.02% (from 317252 to 317308) and 176.gcc actually shrunk by .3% (from 1525612 to 1520964 bytes). 252.eon never iterated in the SCC Passmgr, 176.gcc iterated at most 1 time. llvm-svn: 102823
2010-05-01 03:15:56 +02:00
%6 = bitcast i8* %5 to i8*** ; <i8***> [#uses=1]
[opaque pointer type] Add textual IR support for explicit type parameter to gep operator Similar to gep (r230786) and load (r230794) changes. Similar migration script can be used to update test cases, which successfully migrated all of LLVM and Polly, but about 4 test cases needed manually changes in Clang. (this script will read the contents of stdin and massage it into stdout - wrap it in the 'apply.sh' script shown in previous commits + xargs to apply it over a large set of test cases) import fileinput import sys import re rep = re.compile(r"(getelementptr(?:\s+inbounds)?\s*\()((<\d*\s+x\s+)?([^@]*?)(|\s*addrspace\(\d+\))\s*\*(?(3)>)\s*)(?=$|%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|zeroinitializer|<|\[\[[a-zA-Z]|\{\{)", re.MULTILINE | re.DOTALL) def conv(match): line = match.group(1) line += match.group(4) line += ", " line += match.group(2) return line line = sys.stdin.read() off = 0 for match in re.finditer(rep, line): sys.stdout.write(line[off:match.start()]) sys.stdout.write(conv(match)) off = match.end() sys.stdout.write(line[off:]) llvm-svn: 232184
2015-03-13 19:20:45 +01:00
store i8** getelementptr inbounds ([6 x i8*], [6 x i8*]* @_ZTV1C, i64 0, i64 2), i8*** %6
Implement rdar://6295824 and PR6724 with two tiny changes that can have a big effect :). The first is to enable the iterative SCC passmanager juice that kicks in when the scc passmgr detects that a function pass has devirtualized a call. In this case, it will rerun all the passes it manages on the SCC, up to the iteration count limit (4). This is useful because a function pass may devirualize a call, and we want the inliner to inline it, or pruneeh to infer stuff about it, etc. The second patch is to add *all* call sites to the DevirtualizedCalls list the inliner uses. This list is about to get renamed, but the jist of this is that the inliner now reconsiders *all* inlined call sites as candidates for further inlining. The intuition is this that in cases like this: f() { g(1); } g(int x) { h(x); } We analyze this bottom up, and may decide that it isn't profitable to inline H into G. Next step, we decide that it is profitable to inline G into F, and do so, which means that F now calls H. Even though the call from G -> H may not have been profitable to inline, the call from F -> H may be (in this case because a constant allows folding etc). In my spot checks, this doesn't have a big impact on code. For example, the LLC output for 252.eon grew from 0.02% (from 317252 to 317308) and 176.gcc actually shrunk by .3% (from 1525612 to 1520964 bytes). 252.eon never iterated in the SCC Passmgr, 176.gcc iterated at most 1 time. llvm-svn: 102823
2010-05-01 03:15:56 +02:00
%7 = bitcast %struct.C* %this to i8* ; <i8*> [#uses=1]
[opaque pointer type] Add textual IR support for explicit type parameter to getelementptr instruction One of several parallel first steps to remove the target type of pointers, replacing them with a single opaque pointer type. This adds an explicit type parameter to the gep instruction so that when the first parameter becomes an opaque pointer type, the type to gep through is still available to the instructions. * This doesn't modify gep operators, only instructions (operators will be handled separately) * Textual IR changes only. Bitcode (including upgrade) and changing the in-memory representation will be in separate changes. * geps of vectors are transformed as: getelementptr <4 x float*> %x, ... ->getelementptr float, <4 x float*> %x, ... Then, once the opaque pointer type is introduced, this will ultimately look like: getelementptr float, <4 x ptr> %x with the unambiguous interpretation that it is a vector of pointers to float. * address spaces remain on the pointer, not the type: getelementptr float addrspace(1)* %x ->getelementptr float, float addrspace(1)* %x Then, eventually: getelementptr float, ptr addrspace(1) %x Importantly, the massive amount of test case churn has been automated by same crappy python code. I had to manually update a few test cases that wouldn't fit the script's model (r228970,r229196,r229197,r229198). The python script just massages stdin and writes the result to stdout, I then wrapped that in a shell script to handle replacing files, then using the usual find+xargs to migrate all the files. update.py: import fileinput import sys import re ibrep = re.compile(r"(^.*?[^%\w]getelementptr inbounds )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))") normrep = re.compile( r"(^.*?[^%\w]getelementptr )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))") def conv(match, line): if not match: return line line = match.groups()[0] if len(match.groups()[5]) == 0: line += match.groups()[2] line += match.groups()[3] line += ", " line += match.groups()[1] line += "\n" return line for line in sys.stdin: if line.find("getelementptr ") == line.find("getelementptr inbounds"): if line.find("getelementptr inbounds") != line.find("getelementptr inbounds ("): line = conv(re.match(ibrep, line), line) elif line.find("getelementptr ") != line.find("getelementptr ("): line = conv(re.match(normrep, line), line) sys.stdout.write(line) apply.sh: for name in "$@" do python3 `dirname "$0"`/update.py < "$name" > "$name.tmp" && mv "$name.tmp" "$name" rm -f "$name.tmp" done The actual commands: From llvm/src: find test/ -name *.ll | xargs ./apply.sh From llvm/src/tools/clang: find test/ -name *.mm -o -name *.m -o -name *.cpp -o -name *.c | xargs -I '{}' ../../apply.sh "{}" From llvm/src/tools/polly: find test/ -name *.ll | xargs ./apply.sh After that, check-all (with llvm, clang, clang-tools-extra, lld, compiler-rt, and polly all checked out). The extra 'rm' in the apply.sh script is due to a few files in clang's test suite using interesting unicode stuff that my python script was throwing exceptions on. None of those files needed to be migrated, so it seemed sufficient to ignore those cases. Reviewers: rafael, dexonsmith, grosser Differential Revision: http://reviews.llvm.org/D7636 llvm-svn: 230786
2015-02-27 20:29:02 +01:00
%8 = getelementptr inbounds i8, i8* %7, i64 8 ; <i8*> [#uses=1]
Implement rdar://6295824 and PR6724 with two tiny changes that can have a big effect :). The first is to enable the iterative SCC passmanager juice that kicks in when the scc passmgr detects that a function pass has devirtualized a call. In this case, it will rerun all the passes it manages on the SCC, up to the iteration count limit (4). This is useful because a function pass may devirualize a call, and we want the inliner to inline it, or pruneeh to infer stuff about it, etc. The second patch is to add *all* call sites to the DevirtualizedCalls list the inliner uses. This list is about to get renamed, but the jist of this is that the inliner now reconsiders *all* inlined call sites as candidates for further inlining. The intuition is this that in cases like this: f() { g(1); } g(int x) { h(x); } We analyze this bottom up, and may decide that it isn't profitable to inline H into G. Next step, we decide that it is profitable to inline G into F, and do so, which means that F now calls H. Even though the call from G -> H may not have been profitable to inline, the call from F -> H may be (in this case because a constant allows folding etc). In my spot checks, this doesn't have a big impact on code. For example, the LLC output for 252.eon grew from 0.02% (from 317252 to 317308) and 176.gcc actually shrunk by .3% (from 1525612 to 1520964 bytes). 252.eon never iterated in the SCC Passmgr, 176.gcc iterated at most 1 time. llvm-svn: 102823
2010-05-01 03:15:56 +02:00
%9 = bitcast i8* %8 to i8*** ; <i8***> [#uses=1]
[opaque pointer type] Add textual IR support for explicit type parameter to gep operator Similar to gep (r230786) and load (r230794) changes. Similar migration script can be used to update test cases, which successfully migrated all of LLVM and Polly, but about 4 test cases needed manually changes in Clang. (this script will read the contents of stdin and massage it into stdout - wrap it in the 'apply.sh' script shown in previous commits + xargs to apply it over a large set of test cases) import fileinput import sys import re rep = re.compile(r"(getelementptr(?:\s+inbounds)?\s*\()((<\d*\s+x\s+)?([^@]*?)(|\s*addrspace\(\d+\))\s*\*(?(3)>)\s*)(?=$|%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|zeroinitializer|<|\[\[[a-zA-Z]|\{\{)", re.MULTILINE | re.DOTALL) def conv(match): line = match.group(1) line += match.group(4) line += ", " line += match.group(2) return line line = sys.stdin.read() off = 0 for match in re.finditer(rep, line): sys.stdout.write(line[off:match.start()]) sys.stdout.write(conv(match)) off = match.end() sys.stdout.write(line[off:]) llvm-svn: 232184
2015-03-13 19:20:45 +01:00
store i8** getelementptr inbounds ([6 x i8*], [6 x i8*]* @_ZTV1C, i64 0, i64 5), i8*** %9
Implement rdar://6295824 and PR6724 with two tiny changes that can have a big effect :). The first is to enable the iterative SCC passmanager juice that kicks in when the scc passmgr detects that a function pass has devirtualized a call. In this case, it will rerun all the passes it manages on the SCC, up to the iteration count limit (4). This is useful because a function pass may devirualize a call, and we want the inliner to inline it, or pruneeh to infer stuff about it, etc. The second patch is to add *all* call sites to the DevirtualizedCalls list the inliner uses. This list is about to get renamed, but the jist of this is that the inliner now reconsiders *all* inlined call sites as candidates for further inlining. The intuition is this that in cases like this: f() { g(1); } g(int x) { h(x); } We analyze this bottom up, and may decide that it isn't profitable to inline H into G. Next step, we decide that it is profitable to inline G into F, and do so, which means that F now calls H. Even though the call from G -> H may not have been profitable to inline, the call from F -> H may be (in this case because a constant allows folding etc). In my spot checks, this doesn't have a big impact on code. For example, the LLC output for 252.eon grew from 0.02% (from 317252 to 317308) and 176.gcc actually shrunk by .3% (from 1525612 to 1520964 bytes). 252.eon never iterated in the SCC Passmgr, 176.gcc iterated at most 1 time. llvm-svn: 102823
2010-05-01 03:15:56 +02:00
ret void
}
define linkonce_odr i32 @_ZN1C1fEv(%struct.C* %this) ssp align 2 {
entry:
ret i32 3
}
define linkonce_odr i32 @_ZThn8_N1C1fEv(%struct.C* %this) {
entry:
%0 = bitcast %struct.C* %this to i8* ; <i8*> [#uses=1]
[opaque pointer type] Add textual IR support for explicit type parameter to getelementptr instruction One of several parallel first steps to remove the target type of pointers, replacing them with a single opaque pointer type. This adds an explicit type parameter to the gep instruction so that when the first parameter becomes an opaque pointer type, the type to gep through is still available to the instructions. * This doesn't modify gep operators, only instructions (operators will be handled separately) * Textual IR changes only. Bitcode (including upgrade) and changing the in-memory representation will be in separate changes. * geps of vectors are transformed as: getelementptr <4 x float*> %x, ... ->getelementptr float, <4 x float*> %x, ... Then, once the opaque pointer type is introduced, this will ultimately look like: getelementptr float, <4 x ptr> %x with the unambiguous interpretation that it is a vector of pointers to float. * address spaces remain on the pointer, not the type: getelementptr float addrspace(1)* %x ->getelementptr float, float addrspace(1)* %x Then, eventually: getelementptr float, ptr addrspace(1) %x Importantly, the massive amount of test case churn has been automated by same crappy python code. I had to manually update a few test cases that wouldn't fit the script's model (r228970,r229196,r229197,r229198). The python script just massages stdin and writes the result to stdout, I then wrapped that in a shell script to handle replacing files, then using the usual find+xargs to migrate all the files. update.py: import fileinput import sys import re ibrep = re.compile(r"(^.*?[^%\w]getelementptr inbounds )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))") normrep = re.compile( r"(^.*?[^%\w]getelementptr )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))") def conv(match, line): if not match: return line line = match.groups()[0] if len(match.groups()[5]) == 0: line += match.groups()[2] line += match.groups()[3] line += ", " line += match.groups()[1] line += "\n" return line for line in sys.stdin: if line.find("getelementptr ") == line.find("getelementptr inbounds"): if line.find("getelementptr inbounds") != line.find("getelementptr inbounds ("): line = conv(re.match(ibrep, line), line) elif line.find("getelementptr ") != line.find("getelementptr ("): line = conv(re.match(normrep, line), line) sys.stdout.write(line) apply.sh: for name in "$@" do python3 `dirname "$0"`/update.py < "$name" > "$name.tmp" && mv "$name.tmp" "$name" rm -f "$name.tmp" done The actual commands: From llvm/src: find test/ -name *.ll | xargs ./apply.sh From llvm/src/tools/clang: find test/ -name *.mm -o -name *.m -o -name *.cpp -o -name *.c | xargs -I '{}' ../../apply.sh "{}" From llvm/src/tools/polly: find test/ -name *.ll | xargs ./apply.sh After that, check-all (with llvm, clang, clang-tools-extra, lld, compiler-rt, and polly all checked out). The extra 'rm' in the apply.sh script is due to a few files in clang's test suite using interesting unicode stuff that my python script was throwing exceptions on. None of those files needed to be migrated, so it seemed sufficient to ignore those cases. Reviewers: rafael, dexonsmith, grosser Differential Revision: http://reviews.llvm.org/D7636 llvm-svn: 230786
2015-02-27 20:29:02 +01:00
%1 = getelementptr inbounds i8, i8* %0, i64 -8 ; <i8*> [#uses=1]
Implement rdar://6295824 and PR6724 with two tiny changes that can have a big effect :). The first is to enable the iterative SCC passmanager juice that kicks in when the scc passmgr detects that a function pass has devirtualized a call. In this case, it will rerun all the passes it manages on the SCC, up to the iteration count limit (4). This is useful because a function pass may devirualize a call, and we want the inliner to inline it, or pruneeh to infer stuff about it, etc. The second patch is to add *all* call sites to the DevirtualizedCalls list the inliner uses. This list is about to get renamed, but the jist of this is that the inliner now reconsiders *all* inlined call sites as candidates for further inlining. The intuition is this that in cases like this: f() { g(1); } g(int x) { h(x); } We analyze this bottom up, and may decide that it isn't profitable to inline H into G. Next step, we decide that it is profitable to inline G into F, and do so, which means that F now calls H. Even though the call from G -> H may not have been profitable to inline, the call from F -> H may be (in this case because a constant allows folding etc). In my spot checks, this doesn't have a big impact on code. For example, the LLC output for 252.eon grew from 0.02% (from 317252 to 317308) and 176.gcc actually shrunk by .3% (from 1525612 to 1520964 bytes). 252.eon never iterated in the SCC Passmgr, 176.gcc iterated at most 1 time. llvm-svn: 102823
2010-05-01 03:15:56 +02:00
%2 = bitcast i8* %1 to %struct.C* ; <%struct.C*> [#uses=1]
%call = call i32 @_ZN1C1fEv(%struct.C* %2) ; <i32> [#uses=1]
ret i32 %call
}
define linkonce_odr void @_ZN1AC2Ev(%struct.A* %this) inlinehint ssp align 2 {
entry:
%0 = bitcast %struct.A* %this to i8* ; <i8*> [#uses=1]
[opaque pointer type] Add textual IR support for explicit type parameter to getelementptr instruction One of several parallel first steps to remove the target type of pointers, replacing them with a single opaque pointer type. This adds an explicit type parameter to the gep instruction so that when the first parameter becomes an opaque pointer type, the type to gep through is still available to the instructions. * This doesn't modify gep operators, only instructions (operators will be handled separately) * Textual IR changes only. Bitcode (including upgrade) and changing the in-memory representation will be in separate changes. * geps of vectors are transformed as: getelementptr <4 x float*> %x, ... ->getelementptr float, <4 x float*> %x, ... Then, once the opaque pointer type is introduced, this will ultimately look like: getelementptr float, <4 x ptr> %x with the unambiguous interpretation that it is a vector of pointers to float. * address spaces remain on the pointer, not the type: getelementptr float addrspace(1)* %x ->getelementptr float, float addrspace(1)* %x Then, eventually: getelementptr float, ptr addrspace(1) %x Importantly, the massive amount of test case churn has been automated by same crappy python code. I had to manually update a few test cases that wouldn't fit the script's model (r228970,r229196,r229197,r229198). The python script just massages stdin and writes the result to stdout, I then wrapped that in a shell script to handle replacing files, then using the usual find+xargs to migrate all the files. update.py: import fileinput import sys import re ibrep = re.compile(r"(^.*?[^%\w]getelementptr inbounds )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))") normrep = re.compile( r"(^.*?[^%\w]getelementptr )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))") def conv(match, line): if not match: return line line = match.groups()[0] if len(match.groups()[5]) == 0: line += match.groups()[2] line += match.groups()[3] line += ", " line += match.groups()[1] line += "\n" return line for line in sys.stdin: if line.find("getelementptr ") == line.find("getelementptr inbounds"): if line.find("getelementptr inbounds") != line.find("getelementptr inbounds ("): line = conv(re.match(ibrep, line), line) elif line.find("getelementptr ") != line.find("getelementptr ("): line = conv(re.match(normrep, line), line) sys.stdout.write(line) apply.sh: for name in "$@" do python3 `dirname "$0"`/update.py < "$name" > "$name.tmp" && mv "$name.tmp" "$name" rm -f "$name.tmp" done The actual commands: From llvm/src: find test/ -name *.ll | xargs ./apply.sh From llvm/src/tools/clang: find test/ -name *.mm -o -name *.m -o -name *.cpp -o -name *.c | xargs -I '{}' ../../apply.sh "{}" From llvm/src/tools/polly: find test/ -name *.ll | xargs ./apply.sh After that, check-all (with llvm, clang, clang-tools-extra, lld, compiler-rt, and polly all checked out). The extra 'rm' in the apply.sh script is due to a few files in clang's test suite using interesting unicode stuff that my python script was throwing exceptions on. None of those files needed to be migrated, so it seemed sufficient to ignore those cases. Reviewers: rafael, dexonsmith, grosser Differential Revision: http://reviews.llvm.org/D7636 llvm-svn: 230786
2015-02-27 20:29:02 +01:00
%1 = getelementptr inbounds i8, i8* %0, i64 0 ; <i8*> [#uses=1]
Implement rdar://6295824 and PR6724 with two tiny changes that can have a big effect :). The first is to enable the iterative SCC passmanager juice that kicks in when the scc passmgr detects that a function pass has devirtualized a call. In this case, it will rerun all the passes it manages on the SCC, up to the iteration count limit (4). This is useful because a function pass may devirualize a call, and we want the inliner to inline it, or pruneeh to infer stuff about it, etc. The second patch is to add *all* call sites to the DevirtualizedCalls list the inliner uses. This list is about to get renamed, but the jist of this is that the inliner now reconsiders *all* inlined call sites as candidates for further inlining. The intuition is this that in cases like this: f() { g(1); } g(int x) { h(x); } We analyze this bottom up, and may decide that it isn't profitable to inline H into G. Next step, we decide that it is profitable to inline G into F, and do so, which means that F now calls H. Even though the call from G -> H may not have been profitable to inline, the call from F -> H may be (in this case because a constant allows folding etc). In my spot checks, this doesn't have a big impact on code. For example, the LLC output for 252.eon grew from 0.02% (from 317252 to 317308) and 176.gcc actually shrunk by .3% (from 1525612 to 1520964 bytes). 252.eon never iterated in the SCC Passmgr, 176.gcc iterated at most 1 time. llvm-svn: 102823
2010-05-01 03:15:56 +02:00
%2 = bitcast i8* %1 to i8*** ; <i8***> [#uses=1]
[opaque pointer type] Add textual IR support for explicit type parameter to gep operator Similar to gep (r230786) and load (r230794) changes. Similar migration script can be used to update test cases, which successfully migrated all of LLVM and Polly, but about 4 test cases needed manually changes in Clang. (this script will read the contents of stdin and massage it into stdout - wrap it in the 'apply.sh' script shown in previous commits + xargs to apply it over a large set of test cases) import fileinput import sys import re rep = re.compile(r"(getelementptr(?:\s+inbounds)?\s*\()((<\d*\s+x\s+)?([^@]*?)(|\s*addrspace\(\d+\))\s*\*(?(3)>)\s*)(?=$|%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|zeroinitializer|<|\[\[[a-zA-Z]|\{\{)", re.MULTILINE | re.DOTALL) def conv(match): line = match.group(1) line += match.group(4) line += ", " line += match.group(2) return line line = sys.stdin.read() off = 0 for match in re.finditer(rep, line): sys.stdout.write(line[off:match.start()]) sys.stdout.write(conv(match)) off = match.end() sys.stdout.write(line[off:]) llvm-svn: 232184
2015-03-13 19:20:45 +01:00
store i8** getelementptr inbounds ([3 x i8*], [3 x i8*]* @_ZTV1A, i64 0, i64 2), i8*** %2
Implement rdar://6295824 and PR6724 with two tiny changes that can have a big effect :). The first is to enable the iterative SCC passmanager juice that kicks in when the scc passmgr detects that a function pass has devirtualized a call. In this case, it will rerun all the passes it manages on the SCC, up to the iteration count limit (4). This is useful because a function pass may devirualize a call, and we want the inliner to inline it, or pruneeh to infer stuff about it, etc. The second patch is to add *all* call sites to the DevirtualizedCalls list the inliner uses. This list is about to get renamed, but the jist of this is that the inliner now reconsiders *all* inlined call sites as candidates for further inlining. The intuition is this that in cases like this: f() { g(1); } g(int x) { h(x); } We analyze this bottom up, and may decide that it isn't profitable to inline H into G. Next step, we decide that it is profitable to inline G into F, and do so, which means that F now calls H. Even though the call from G -> H may not have been profitable to inline, the call from F -> H may be (in this case because a constant allows folding etc). In my spot checks, this doesn't have a big impact on code. For example, the LLC output for 252.eon grew from 0.02% (from 317252 to 317308) and 176.gcc actually shrunk by .3% (from 1525612 to 1520964 bytes). 252.eon never iterated in the SCC Passmgr, 176.gcc iterated at most 1 time. llvm-svn: 102823
2010-05-01 03:15:56 +02:00
ret void
}
define linkonce_odr void @_ZN1BC2Ev(%struct.A* %this) inlinehint ssp align 2 {
entry:
%0 = bitcast %struct.A* %this to i8* ; <i8*> [#uses=1]
[opaque pointer type] Add textual IR support for explicit type parameter to getelementptr instruction One of several parallel first steps to remove the target type of pointers, replacing them with a single opaque pointer type. This adds an explicit type parameter to the gep instruction so that when the first parameter becomes an opaque pointer type, the type to gep through is still available to the instructions. * This doesn't modify gep operators, only instructions (operators will be handled separately) * Textual IR changes only. Bitcode (including upgrade) and changing the in-memory representation will be in separate changes. * geps of vectors are transformed as: getelementptr <4 x float*> %x, ... ->getelementptr float, <4 x float*> %x, ... Then, once the opaque pointer type is introduced, this will ultimately look like: getelementptr float, <4 x ptr> %x with the unambiguous interpretation that it is a vector of pointers to float. * address spaces remain on the pointer, not the type: getelementptr float addrspace(1)* %x ->getelementptr float, float addrspace(1)* %x Then, eventually: getelementptr float, ptr addrspace(1) %x Importantly, the massive amount of test case churn has been automated by same crappy python code. I had to manually update a few test cases that wouldn't fit the script's model (r228970,r229196,r229197,r229198). The python script just massages stdin and writes the result to stdout, I then wrapped that in a shell script to handle replacing files, then using the usual find+xargs to migrate all the files. update.py: import fileinput import sys import re ibrep = re.compile(r"(^.*?[^%\w]getelementptr inbounds )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))") normrep = re.compile( r"(^.*?[^%\w]getelementptr )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))") def conv(match, line): if not match: return line line = match.groups()[0] if len(match.groups()[5]) == 0: line += match.groups()[2] line += match.groups()[3] line += ", " line += match.groups()[1] line += "\n" return line for line in sys.stdin: if line.find("getelementptr ") == line.find("getelementptr inbounds"): if line.find("getelementptr inbounds") != line.find("getelementptr inbounds ("): line = conv(re.match(ibrep, line), line) elif line.find("getelementptr ") != line.find("getelementptr ("): line = conv(re.match(normrep, line), line) sys.stdout.write(line) apply.sh: for name in "$@" do python3 `dirname "$0"`/update.py < "$name" > "$name.tmp" && mv "$name.tmp" "$name" rm -f "$name.tmp" done The actual commands: From llvm/src: find test/ -name *.ll | xargs ./apply.sh From llvm/src/tools/clang: find test/ -name *.mm -o -name *.m -o -name *.cpp -o -name *.c | xargs -I '{}' ../../apply.sh "{}" From llvm/src/tools/polly: find test/ -name *.ll | xargs ./apply.sh After that, check-all (with llvm, clang, clang-tools-extra, lld, compiler-rt, and polly all checked out). The extra 'rm' in the apply.sh script is due to a few files in clang's test suite using interesting unicode stuff that my python script was throwing exceptions on. None of those files needed to be migrated, so it seemed sufficient to ignore those cases. Reviewers: rafael, dexonsmith, grosser Differential Revision: http://reviews.llvm.org/D7636 llvm-svn: 230786
2015-02-27 20:29:02 +01:00
%1 = getelementptr inbounds i8, i8* %0, i64 0 ; <i8*> [#uses=1]
Implement rdar://6295824 and PR6724 with two tiny changes that can have a big effect :). The first is to enable the iterative SCC passmanager juice that kicks in when the scc passmgr detects that a function pass has devirtualized a call. In this case, it will rerun all the passes it manages on the SCC, up to the iteration count limit (4). This is useful because a function pass may devirualize a call, and we want the inliner to inline it, or pruneeh to infer stuff about it, etc. The second patch is to add *all* call sites to the DevirtualizedCalls list the inliner uses. This list is about to get renamed, but the jist of this is that the inliner now reconsiders *all* inlined call sites as candidates for further inlining. The intuition is this that in cases like this: f() { g(1); } g(int x) { h(x); } We analyze this bottom up, and may decide that it isn't profitable to inline H into G. Next step, we decide that it is profitable to inline G into F, and do so, which means that F now calls H. Even though the call from G -> H may not have been profitable to inline, the call from F -> H may be (in this case because a constant allows folding etc). In my spot checks, this doesn't have a big impact on code. For example, the LLC output for 252.eon grew from 0.02% (from 317252 to 317308) and 176.gcc actually shrunk by .3% (from 1525612 to 1520964 bytes). 252.eon never iterated in the SCC Passmgr, 176.gcc iterated at most 1 time. llvm-svn: 102823
2010-05-01 03:15:56 +02:00
%2 = bitcast i8* %1 to i8*** ; <i8***> [#uses=1]
[opaque pointer type] Add textual IR support for explicit type parameter to gep operator Similar to gep (r230786) and load (r230794) changes. Similar migration script can be used to update test cases, which successfully migrated all of LLVM and Polly, but about 4 test cases needed manually changes in Clang. (this script will read the contents of stdin and massage it into stdout - wrap it in the 'apply.sh' script shown in previous commits + xargs to apply it over a large set of test cases) import fileinput import sys import re rep = re.compile(r"(getelementptr(?:\s+inbounds)?\s*\()((<\d*\s+x\s+)?([^@]*?)(|\s*addrspace\(\d+\))\s*\*(?(3)>)\s*)(?=$|%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|zeroinitializer|<|\[\[[a-zA-Z]|\{\{)", re.MULTILINE | re.DOTALL) def conv(match): line = match.group(1) line += match.group(4) line += ", " line += match.group(2) return line line = sys.stdin.read() off = 0 for match in re.finditer(rep, line): sys.stdout.write(line[off:match.start()]) sys.stdout.write(conv(match)) off = match.end() sys.stdout.write(line[off:]) llvm-svn: 232184
2015-03-13 19:20:45 +01:00
store i8** getelementptr inbounds ([3 x i8*], [3 x i8*]* @_ZTV1B, i64 0, i64 2), i8*** %2
Implement rdar://6295824 and PR6724 with two tiny changes that can have a big effect :). The first is to enable the iterative SCC passmanager juice that kicks in when the scc passmgr detects that a function pass has devirtualized a call. In this case, it will rerun all the passes it manages on the SCC, up to the iteration count limit (4). This is useful because a function pass may devirualize a call, and we want the inliner to inline it, or pruneeh to infer stuff about it, etc. The second patch is to add *all* call sites to the DevirtualizedCalls list the inliner uses. This list is about to get renamed, but the jist of this is that the inliner now reconsiders *all* inlined call sites as candidates for further inlining. The intuition is this that in cases like this: f() { g(1); } g(int x) { h(x); } We analyze this bottom up, and may decide that it isn't profitable to inline H into G. Next step, we decide that it is profitable to inline G into F, and do so, which means that F now calls H. Even though the call from G -> H may not have been profitable to inline, the call from F -> H may be (in this case because a constant allows folding etc). In my spot checks, this doesn't have a big impact on code. For example, the LLC output for 252.eon grew from 0.02% (from 317252 to 317308) and 176.gcc actually shrunk by .3% (from 1525612 to 1520964 bytes). 252.eon never iterated in the SCC Passmgr, 176.gcc iterated at most 1 time. llvm-svn: 102823
2010-05-01 03:15:56 +02:00
ret void
}
define linkonce_odr i32 @_ZN1B1fEv(%struct.A* %this) ssp align 2 {
entry:
ret i32 2
}
define linkonce_odr i32 @_ZN1A1fEv(%struct.A* %this) ssp align 2 {
entry:
ret i32 1
}
Reference in New Issue Copy Permalink