2009-11-27 07:32:59 +08:00
|
|
|
; RUN: opt < %s -gvn -S | FileCheck %s
|
|
|
|
|
|
Implement initial support for PHI translation in memdep. This means that
memdep keeps track of how PHIs affect the pointer in dep queries, which
allows it to eliminate the load in cases like rle-phi-translate.ll, which
basically end up being:
BB1:
X = load P
br BB3
BB2:
Y = load Q
br BB3
BB3:
R = phi [P] [Q]
load R
turning "load R" into a phi of X/Y. In addition to additional exposed
opportunities, this makes memdep safe in many cases that it wasn't before
(which is required for load PRE) and also makes it substantially more
efficient. For example, consider:
bb1: // has many predecessors.
P = some_operator()
load P
In this example, previously memdep would scan all the predecessors of BB1
to see if they had something that would mustalias P. In some cases (e.g.
test/Transforms/GVN/rle-must-alias.ll) it would actually find them and end
up eliminating something. In many other cases though, it would scan and not
find anything useful. MemDep now stops at a block if the pointer is defined
in that block and cannot be phi translated to predecessors. This causes it
to miss the (rare) cases like rle-must-alias.ll, but makes it faster by not
scanning tons of stuff that is unlikely to be useful. For example, this
speeds up GVN as a whole from 3.928s to 2.448s (60%)!. IMO, scalar GVN
should be enhanced to simplify the rle-must-alias pointer base anyway, which
would allow the loads to be eliminated.
In the future, this should be enhanced to phi translate through geps and
bitcasts as well (as indicated by FIXMEs) making memdep even more powerful.
llvm-svn: 61022
2008-12-15 11:35:32 +08:00
|
|
|
target datalayout = "e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-f32:32:32-f64:32:64-v64:64:64-v128:128:128-a0:0:64-f80:128:128"
|
|
|
|
|
target triple = "i386-apple-darwin7"
|
|
|
|
|
|
2009-11-27 07:32:59 +08:00
|
|
|
define i32 @test1(i32* %b, i32* %c) nounwind {
|
2013-07-14 09:42:54 +08:00
|
|
|
; CHECK-LABEL: @test1(
|
Implement initial support for PHI translation in memdep. This means that
memdep keeps track of how PHIs affect the pointer in dep queries, which
allows it to eliminate the load in cases like rle-phi-translate.ll, which
basically end up being:
BB1:
X = load P
br BB3
BB2:
Y = load Q
br BB3
BB3:
R = phi [P] [Q]
load R
turning "load R" into a phi of X/Y. In addition to additional exposed
opportunities, this makes memdep safe in many cases that it wasn't before
(which is required for load PRE) and also makes it substantially more
efficient. For example, consider:
bb1: // has many predecessors.
P = some_operator()
load P
In this example, previously memdep would scan all the predecessors of BB1
to see if they had something that would mustalias P. In some cases (e.g.
test/Transforms/GVN/rle-must-alias.ll) it would actually find them and end
up eliminating something. In many other cases though, it would scan and not
find anything useful. MemDep now stops at a block if the pointer is defined
in that block and cannot be phi translated to predecessors. This causes it
to miss the (rare) cases like rle-must-alias.ll, but makes it faster by not
scanning tons of stuff that is unlikely to be useful. For example, this
speeds up GVN as a whole from 3.928s to 2.448s (60%)!. IMO, scalar GVN
should be enhanced to simplify the rle-must-alias pointer base anyway, which
would allow the loads to be eliminated.
In the future, this should be enhanced to phi translate through geps and
bitcasts as well (as indicated by FIXMEs) making memdep even more powerful.
llvm-svn: 61022
2008-12-15 11:35:32 +08:00
|
|
|
entry:
|
2009-11-27 07:32:59 +08:00
|
|
|
%g = alloca i32
|
|
|
|
|
%t1 = icmp eq i32* %b, null
|
Implement initial support for PHI translation in memdep. This means that
memdep keeps track of how PHIs affect the pointer in dep queries, which
allows it to eliminate the load in cases like rle-phi-translate.ll, which
basically end up being:
BB1:
X = load P
br BB3
BB2:
Y = load Q
br BB3
BB3:
R = phi [P] [Q]
load R
turning "load R" into a phi of X/Y. In addition to additional exposed
opportunities, this makes memdep safe in many cases that it wasn't before
(which is required for load PRE) and also makes it substantially more
efficient. For example, consider:
bb1: // has many predecessors.
P = some_operator()
load P
In this example, previously memdep would scan all the predecessors of BB1
to see if they had something that would mustalias P. In some cases (e.g.
test/Transforms/GVN/rle-must-alias.ll) it would actually find them and end
up eliminating something. In many other cases though, it would scan and not
find anything useful. MemDep now stops at a block if the pointer is defined
in that block and cannot be phi translated to predecessors. This causes it
to miss the (rare) cases like rle-must-alias.ll, but makes it faster by not
scanning tons of stuff that is unlikely to be useful. For example, this
speeds up GVN as a whole from 3.928s to 2.448s (60%)!. IMO, scalar GVN
should be enhanced to simplify the rle-must-alias pointer base anyway, which
would allow the loads to be eliminated.
In the future, this should be enhanced to phi translate through geps and
bitcasts as well (as indicated by FIXMEs) making memdep even more powerful.
llvm-svn: 61022
2008-12-15 11:35:32 +08:00
|
|
|
br i1 %t1, label %bb, label %bb1
|
|
|
|
|
|
2009-11-27 07:32:59 +08:00
|
|
|
bb:
|
2015-02-28 05:17:42 +08:00
|
|
|
%t2 = load i32, i32* %c, align 4
|
2009-11-27 07:32:59 +08:00
|
|
|
%t3 = add i32 %t2, 1
|
Implement initial support for PHI translation in memdep. This means that
memdep keeps track of how PHIs affect the pointer in dep queries, which
allows it to eliminate the load in cases like rle-phi-translate.ll, which
basically end up being:
BB1:
X = load P
br BB3
BB2:
Y = load Q
br BB3
BB3:
R = phi [P] [Q]
load R
turning "load R" into a phi of X/Y. In addition to additional exposed
opportunities, this makes memdep safe in many cases that it wasn't before
(which is required for load PRE) and also makes it substantially more
efficient. For example, consider:
bb1: // has many predecessors.
P = some_operator()
load P
In this example, previously memdep would scan all the predecessors of BB1
to see if they had something that would mustalias P. In some cases (e.g.
test/Transforms/GVN/rle-must-alias.ll) it would actually find them and end
up eliminating something. In many other cases though, it would scan and not
find anything useful. MemDep now stops at a block if the pointer is defined
in that block and cannot be phi translated to predecessors. This causes it
to miss the (rare) cases like rle-must-alias.ll, but makes it faster by not
scanning tons of stuff that is unlikely to be useful. For example, this
speeds up GVN as a whole from 3.928s to 2.448s (60%)!. IMO, scalar GVN
should be enhanced to simplify the rle-must-alias pointer base anyway, which
would allow the loads to be eliminated.
In the future, this should be enhanced to phi translate through geps and
bitcasts as well (as indicated by FIXMEs) making memdep even more powerful.
llvm-svn: 61022
2008-12-15 11:35:32 +08:00
|
|
|
store i32 %t3, i32* %g, align 4
|
|
|
|
|
br label %bb2
|
|
|
|
|
|
|
|
|
|
bb1: ; preds = %entry
|
2015-02-28 05:17:42 +08:00
|
|
|
%t5 = load i32, i32* %b, align 4
|
2009-11-27 07:32:59 +08:00
|
|
|
%t6 = add i32 %t5, 1
|
Implement initial support for PHI translation in memdep. This means that
memdep keeps track of how PHIs affect the pointer in dep queries, which
allows it to eliminate the load in cases like rle-phi-translate.ll, which
basically end up being:
BB1:
X = load P
br BB3
BB2:
Y = load Q
br BB3
BB3:
R = phi [P] [Q]
load R
turning "load R" into a phi of X/Y. In addition to additional exposed
opportunities, this makes memdep safe in many cases that it wasn't before
(which is required for load PRE) and also makes it substantially more
efficient. For example, consider:
bb1: // has many predecessors.
P = some_operator()
load P
In this example, previously memdep would scan all the predecessors of BB1
to see if they had something that would mustalias P. In some cases (e.g.
test/Transforms/GVN/rle-must-alias.ll) it would actually find them and end
up eliminating something. In many other cases though, it would scan and not
find anything useful. MemDep now stops at a block if the pointer is defined
in that block and cannot be phi translated to predecessors. This causes it
to miss the (rare) cases like rle-must-alias.ll, but makes it faster by not
scanning tons of stuff that is unlikely to be useful. For example, this
speeds up GVN as a whole from 3.928s to 2.448s (60%)!. IMO, scalar GVN
should be enhanced to simplify the rle-must-alias pointer base anyway, which
would allow the loads to be eliminated.
In the future, this should be enhanced to phi translate through geps and
bitcasts as well (as indicated by FIXMEs) making memdep even more powerful.
llvm-svn: 61022
2008-12-15 11:35:32 +08:00
|
|
|
store i32 %t6, i32* %g, align 4
|
|
|
|
|
br label %bb2
|
|
|
|
|
|
|
|
|
|
bb2: ; preds = %bb1, %bb
|
2009-11-27 07:32:59 +08:00
|
|
|
%c_addr.0 = phi i32* [ %g, %bb1 ], [ %c, %bb ]
|
|
|
|
|
%b_addr.0 = phi i32* [ %b, %bb1 ], [ %g, %bb ]
|
2015-02-28 05:17:42 +08:00
|
|
|
%cv = load i32, i32* %c_addr.0, align 4
|
|
|
|
|
%bv = load i32, i32* %b_addr.0, align 4
|
2009-11-27 07:32:59 +08:00
|
|
|
; CHECK: %bv = phi i32
|
|
|
|
|
; CHECK: %cv = phi i32
|
|
|
|
|
; CHECK-NOT: load
|
|
|
|
|
; CHECK: ret i32
|
|
|
|
|
%ret = add i32 %cv, %bv
|
Implement initial support for PHI translation in memdep. This means that
memdep keeps track of how PHIs affect the pointer in dep queries, which
allows it to eliminate the load in cases like rle-phi-translate.ll, which
basically end up being:
BB1:
X = load P
br BB3
BB2:
Y = load Q
br BB3
BB3:
R = phi [P] [Q]
load R
turning "load R" into a phi of X/Y. In addition to additional exposed
opportunities, this makes memdep safe in many cases that it wasn't before
(which is required for load PRE) and also makes it substantially more
efficient. For example, consider:
bb1: // has many predecessors.
P = some_operator()
load P
In this example, previously memdep would scan all the predecessors of BB1
to see if they had something that would mustalias P. In some cases (e.g.
test/Transforms/GVN/rle-must-alias.ll) it would actually find them and end
up eliminating something. In many other cases though, it would scan and not
find anything useful. MemDep now stops at a block if the pointer is defined
in that block and cannot be phi translated to predecessors. This causes it
to miss the (rare) cases like rle-must-alias.ll, but makes it faster by not
scanning tons of stuff that is unlikely to be useful. For example, this
speeds up GVN as a whole from 3.928s to 2.448s (60%)!. IMO, scalar GVN
should be enhanced to simplify the rle-must-alias pointer base anyway, which
would allow the loads to be eliminated.
In the future, this should be enhanced to phi translate through geps and
bitcasts as well (as indicated by FIXMEs) making memdep even more powerful.
llvm-svn: 61022
2008-12-15 11:35:32 +08:00
|
|
|
ret i32 %ret
|
|
|
|
|
}
|
|
|
|
|
|
2009-11-27 07:41:07 +08:00
|
|
|
define i8 @test2(i1 %cond, i32* %b, i32* %c) nounwind {
|
2013-07-14 09:42:54 +08:00
|
|
|
; CHECK-LABEL: @test2(
|
2009-11-27 07:41:07 +08:00
|
|
|
entry:
|
2009-12-20 04:44:43 +08:00
|
|
|
br i1 %cond, label %bb, label %bb1
|
2009-11-27 07:41:07 +08:00
|
|
|
|
|
|
|
|
bb:
|
|
|
|
|
%b1 = bitcast i32* %b to i8*
|
|
|
|
|
store i8 4, i8* %b1
|
2009-12-20 04:44:43 +08:00
|
|
|
br label %bb2
|
2009-11-27 07:41:07 +08:00
|
|
|
|
|
|
|
|
bb1:
|
|
|
|
|
%c1 = bitcast i32* %c to i8*
|
|
|
|
|
store i8 92, i8* %c1
|
2009-12-20 04:44:43 +08:00
|
|
|
br label %bb2
|
2009-11-27 07:41:07 +08:00
|
|
|
|
|
|
|
|
bb2:
|
2009-12-20 04:44:43 +08:00
|
|
|
%d = phi i32* [ %c, %bb1 ], [ %b, %bb ]
|
2009-11-27 07:41:07 +08:00
|
|
|
%d1 = bitcast i32* %d to i8*
|
2015-02-28 05:17:42 +08:00
|
|
|
%dv = load i8, i8* %d1
|
2009-11-27 08:07:37 +08:00
|
|
|
; CHECK: %dv = phi i8 [ 92, %bb1 ], [ 4, %bb ]
|
2009-11-27 07:41:07 +08:00
|
|
|
; CHECK-NOT: load
|
|
|
|
|
; CHECK: ret i8 %dv
|
2009-12-20 04:44:43 +08:00
|
|
|
ret i8 %dv
|
2009-11-27 07:41:07 +08:00
|
|
|
}
|
|
|
|
|
|
2009-11-27 08:07:37 +08:00
|
|
|
define i32 @test3(i1 %cond, i32* %b, i32* %c) nounwind {
|
2013-07-14 09:42:54 +08:00
|
|
|
; CHECK-LABEL: @test3(
|
2009-11-27 08:07:37 +08:00
|
|
|
entry:
|
2009-12-20 04:44:43 +08:00
|
|
|
br i1 %cond, label %bb, label %bb1
|
2009-11-27 08:07:37 +08:00
|
|
|
|
|
|
|
|
bb:
|
[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
2015-02-28 03:29:02 +08:00
|
|
|
%b1 = getelementptr i32, i32* %b, i32 17
|
2009-11-27 08:07:37 +08:00
|
|
|
store i32 4, i32* %b1
|
2009-12-20 04:44:43 +08:00
|
|
|
br label %bb2
|
2009-11-27 08:07:37 +08:00
|
|
|
|
|
|
|
|
bb1:
|
[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
2015-02-28 03:29:02 +08:00
|
|
|
%c1 = getelementptr i32, i32* %c, i32 7
|
2009-11-27 08:07:37 +08:00
|
|
|
store i32 82, i32* %c1
|
2009-12-20 04:44:43 +08:00
|
|
|
br label %bb2
|
2009-11-27 08:07:37 +08:00
|
|
|
|
|
|
|
|
bb2:
|
2009-12-20 04:44:43 +08:00
|
|
|
%d = phi i32* [ %c, %bb1 ], [ %b, %bb ]
|
|
|
|
|
%i = phi i32 [ 7, %bb1 ], [ 17, %bb ]
|
[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
2015-02-28 03:29:02 +08:00
|
|
|
%d1 = getelementptr i32, i32* %d, i32 %i
|
2015-02-28 05:17:42 +08:00
|
|
|
%dv = load i32, i32* %d1
|
2009-11-27 14:31:14 +08:00
|
|
|
; CHECK: %dv = phi i32 [ 82, %bb1 ], [ 4, %bb ]
|
|
|
|
|
; CHECK-NOT: load
|
|
|
|
|
; CHECK: ret i32 %dv
|
2009-12-20 04:44:43 +08:00
|
|
|
ret i32 %dv
|
2009-11-27 08:07:37 +08:00
|
|
|
}
|
|
|
|
|
|
teach phi translation of GEPs to simplify geps like 'gep x, 0'.
This allows us to compile the example from PR5313 into:
LBB1_2: ## %bb
incl %ecx
movb %al, (%rsi)
movslq %ecx, %rax
movb (%rdi,%rax), %al
testb %al, %al
jne LBB1_2
instead of:
LBB1_2: ## %bb
movslq %eax, %rcx
incl %eax
movb (%rdi,%rcx), %cl
movb %cl, (%rsi)
movslq %eax, %rcx
cmpb $0, (%rdi,%rcx)
jne LBB1_2
llvm-svn: 89981
2009-11-27 08:34:38 +08:00
|
|
|
; PR5313
|
|
|
|
|
define i32 @test4(i1 %cond, i32* %b, i32* %c) nounwind {
|
2013-07-14 09:42:54 +08:00
|
|
|
; CHECK-LABEL: @test4(
|
teach phi translation of GEPs to simplify geps like 'gep x, 0'.
This allows us to compile the example from PR5313 into:
LBB1_2: ## %bb
incl %ecx
movb %al, (%rsi)
movslq %ecx, %rax
movb (%rdi,%rax), %al
testb %al, %al
jne LBB1_2
instead of:
LBB1_2: ## %bb
movslq %eax, %rcx
incl %eax
movb (%rdi,%rcx), %cl
movb %cl, (%rsi)
movslq %eax, %rcx
cmpb $0, (%rdi,%rcx)
jne LBB1_2
llvm-svn: 89981
2009-11-27 08:34:38 +08:00
|
|
|
entry:
|
2009-12-20 04:44:43 +08:00
|
|
|
br i1 %cond, label %bb, label %bb1
|
teach phi translation of GEPs to simplify geps like 'gep x, 0'.
This allows us to compile the example from PR5313 into:
LBB1_2: ## %bb
incl %ecx
movb %al, (%rsi)
movslq %ecx, %rax
movb (%rdi,%rax), %al
testb %al, %al
jne LBB1_2
instead of:
LBB1_2: ## %bb
movslq %eax, %rcx
incl %eax
movb (%rdi,%rcx), %cl
movb %cl, (%rsi)
movslq %eax, %rcx
cmpb $0, (%rdi,%rcx)
jne LBB1_2
llvm-svn: 89981
2009-11-27 08:34:38 +08:00
|
|
|
|
|
|
|
|
bb:
|
|
|
|
|
store i32 4, i32* %b
|
2009-12-20 04:44:43 +08:00
|
|
|
br label %bb2
|
teach phi translation of GEPs to simplify geps like 'gep x, 0'.
This allows us to compile the example from PR5313 into:
LBB1_2: ## %bb
incl %ecx
movb %al, (%rsi)
movslq %ecx, %rax
movb (%rdi,%rax), %al
testb %al, %al
jne LBB1_2
instead of:
LBB1_2: ## %bb
movslq %eax, %rcx
incl %eax
movb (%rdi,%rcx), %cl
movb %cl, (%rsi)
movslq %eax, %rcx
cmpb $0, (%rdi,%rcx)
jne LBB1_2
llvm-svn: 89981
2009-11-27 08:34:38 +08:00
|
|
|
|
|
|
|
|
bb1:
|
[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
2015-02-28 03:29:02 +08:00
|
|
|
%c1 = getelementptr i32, i32* %c, i32 7
|
teach phi translation of GEPs to simplify geps like 'gep x, 0'.
This allows us to compile the example from PR5313 into:
LBB1_2: ## %bb
incl %ecx
movb %al, (%rsi)
movslq %ecx, %rax
movb (%rdi,%rax), %al
testb %al, %al
jne LBB1_2
instead of:
LBB1_2: ## %bb
movslq %eax, %rcx
incl %eax
movb (%rdi,%rcx), %cl
movb %cl, (%rsi)
movslq %eax, %rcx
cmpb $0, (%rdi,%rcx)
jne LBB1_2
llvm-svn: 89981
2009-11-27 08:34:38 +08:00
|
|
|
store i32 82, i32* %c1
|
2009-12-20 04:44:43 +08:00
|
|
|
br label %bb2
|
teach phi translation of GEPs to simplify geps like 'gep x, 0'.
This allows us to compile the example from PR5313 into:
LBB1_2: ## %bb
incl %ecx
movb %al, (%rsi)
movslq %ecx, %rax
movb (%rdi,%rax), %al
testb %al, %al
jne LBB1_2
instead of:
LBB1_2: ## %bb
movslq %eax, %rcx
incl %eax
movb (%rdi,%rcx), %cl
movb %cl, (%rsi)
movslq %eax, %rcx
cmpb $0, (%rdi,%rcx)
jne LBB1_2
llvm-svn: 89981
2009-11-27 08:34:38 +08:00
|
|
|
|
|
|
|
|
bb2:
|
2009-12-20 04:44:43 +08:00
|
|
|
%d = phi i32* [ %c, %bb1 ], [ %b, %bb ]
|
|
|
|
|
%i = phi i32 [ 7, %bb1 ], [ 0, %bb ]
|
[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
2015-02-28 03:29:02 +08:00
|
|
|
%d1 = getelementptr i32, i32* %d, i32 %i
|
2015-02-28 05:17:42 +08:00
|
|
|
%dv = load i32, i32* %d1
|
2009-11-27 14:31:14 +08:00
|
|
|
; CHECK: %dv = phi i32 [ 82, %bb1 ], [ 4, %bb ]
|
|
|
|
|
; CHECK-NOT: load
|
|
|
|
|
; CHECK: ret i32 %dv
|
2009-12-20 04:44:43 +08:00
|
|
|
ret i32 %dv
|
teach phi translation of GEPs to simplify geps like 'gep x, 0'.
This allows us to compile the example from PR5313 into:
LBB1_2: ## %bb
incl %ecx
movb %al, (%rsi)
movslq %ecx, %rax
movb (%rdi,%rax), %al
testb %al, %al
jne LBB1_2
instead of:
LBB1_2: ## %bb
movslq %eax, %rcx
incl %eax
movb (%rdi,%rcx), %cl
movb %cl, (%rsi)
movslq %eax, %rcx
cmpb $0, (%rdi,%rcx)
jne LBB1_2
llvm-svn: 89981
2009-11-27 08:34:38 +08:00
|
|
|
}
|
|
|
|
|
|
2009-12-20 05:29:22 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
; void test5(int N, double* G) {
|
|
|
|
|
; for (long j = 1; j < 1000; j++)
|
|
|
|
|
; G[j] = G[j] + G[j-1];
|
|
|
|
|
; }
|
|
|
|
|
;
|
|
|
|
|
; Should compile into one load in the loop.
|
|
|
|
|
define void @test5(i32 %N, double* nocapture %G) nounwind ssp {
|
2013-07-14 09:42:54 +08:00
|
|
|
; CHECK-LABEL: @test5(
|
2009-12-20 05:29:22 +08:00
|
|
|
bb.nph:
|
|
|
|
|
br label %for.body
|
|
|
|
|
|
|
|
|
|
for.body:
|
|
|
|
|
%indvar = phi i64 [ 0, %bb.nph ], [ %tmp, %for.body ]
|
[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
2015-02-28 03:29:02 +08:00
|
|
|
%arrayidx6 = getelementptr double, double* %G, i64 %indvar
|
2009-12-20 05:29:22 +08:00
|
|
|
%tmp = add i64 %indvar, 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
2015-02-28 03:29:02 +08:00
|
|
|
%arrayidx = getelementptr double, double* %G, i64 %tmp
|
2015-02-28 05:17:42 +08:00
|
|
|
%tmp3 = load double, double* %arrayidx
|
|
|
|
|
%tmp7 = load double, double* %arrayidx6
|
2009-12-20 05:29:22 +08:00
|
|
|
%add = fadd double %tmp3, %tmp7
|
|
|
|
|
store double %add, double* %arrayidx
|
|
|
|
|
%exitcond = icmp eq i64 %tmp, 999
|
|
|
|
|
br i1 %exitcond, label %for.end, label %for.body
|
|
|
|
|
; CHECK: for.body:
|
|
|
|
|
; CHECK: phi double
|
|
|
|
|
; CHECK: load double
|
|
|
|
|
; CHECK-NOT: load double
|
|
|
|
|
; CHECK: br i1
|
|
|
|
|
for.end:
|
|
|
|
|
ret void
|
|
|
|
|
}
|