2013-11-30 20:39:27 +01:00
|
|
|
; RUN: llc -mcpu=ppc64 -O0 -disable-fp-elim -fast-isel=false < %s | FileCheck %s
|
2012-09-19 18:18:23 +02:00
|
|
|
|
2012-09-19 17:42:13 +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-v128:128:128-n32:64"
|
|
|
|
|
target triple = "powerpc64-unknown-linux-gnu"
|
|
|
|
|
|
|
|
|
|
%struct.s1 = type { i8 }
|
|
|
|
|
%struct.s2 = type { i16 }
|
|
|
|
|
%struct.s4 = type { i32 }
|
|
|
|
|
%struct.t1 = type { i8 }
|
|
|
|
|
%struct.t3 = type <{ i16, i8 }>
|
|
|
|
|
%struct.t5 = type <{ i32, i8 }>
|
|
|
|
|
%struct.t6 = type <{ i32, i16 }>
|
|
|
|
|
%struct.t7 = type <{ i32, i16, i8 }>
|
|
|
|
|
%struct.s3 = type { i16, i8 }
|
|
|
|
|
%struct.s5 = type { i32, i8 }
|
|
|
|
|
%struct.s6 = type { i32, i16 }
|
|
|
|
|
%struct.s7 = type { i32, i16, i8 }
|
|
|
|
|
%struct.t2 = type <{ i16 }>
|
|
|
|
|
%struct.t4 = type <{ i32 }>
|
|
|
|
|
|
|
|
|
|
@caller1.p1 = private unnamed_addr constant %struct.s1 { i8 1 }, align 1
|
|
|
|
|
@caller1.p2 = private unnamed_addr constant %struct.s2 { i16 2 }, align 2
|
|
|
|
|
@caller1.p3 = private unnamed_addr constant { i16, i8, i8 } { i16 4, i8 8, i8 undef }, align 2
|
|
|
|
|
@caller1.p4 = private unnamed_addr constant %struct.s4 { i32 16 }, align 4
|
|
|
|
|
@caller1.p5 = private unnamed_addr constant { i32, i8, [3 x i8] } { i32 32, i8 64, [3 x i8] undef }, align 4
|
|
|
|
|
@caller1.p6 = private unnamed_addr constant { i32, i16, [2 x i8] } { i32 128, i16 256, [2 x i8] undef }, align 4
|
|
|
|
|
@caller1.p7 = private unnamed_addr constant { i32, i16, i8, i8 } { i32 512, i16 1024, i8 -3, i8 undef }, align 4
|
|
|
|
|
@caller2.p1 = private unnamed_addr constant %struct.t1 { i8 1 }, align 1
|
|
|
|
|
@caller2.p2 = private unnamed_addr constant { i16 } { i16 2 }, align 1
|
|
|
|
|
@caller2.p3 = private unnamed_addr constant %struct.t3 <{ i16 4, i8 8 }>, align 1
|
|
|
|
|
@caller2.p4 = private unnamed_addr constant { i32 } { i32 16 }, align 1
|
|
|
|
|
@caller2.p5 = private unnamed_addr constant %struct.t5 <{ i32 32, i8 64 }>, align 1
|
|
|
|
|
@caller2.p6 = private unnamed_addr constant %struct.t6 <{ i32 128, i16 256 }>, align 1
|
|
|
|
|
@caller2.p7 = private unnamed_addr constant %struct.t7 <{ i32 512, i16 1024, i8 -3 }>, align 1
|
|
|
|
|
|
|
|
|
|
define i32 @caller1() nounwind {
|
|
|
|
|
entry:
|
|
|
|
|
%p1 = alloca %struct.s1, align 1
|
|
|
|
|
%p2 = alloca %struct.s2, align 2
|
|
|
|
|
%p3 = alloca %struct.s3, align 2
|
|
|
|
|
%p4 = alloca %struct.s4, align 4
|
|
|
|
|
%p5 = alloca %struct.s5, align 4
|
|
|
|
|
%p6 = alloca %struct.s6, align 4
|
|
|
|
|
%p7 = alloca %struct.s7, align 4
|
|
|
|
|
%0 = bitcast %struct.s1* %p1 to i8*
|
2015-11-19 06:56:52 +01:00
|
|
|
call void @llvm.memcpy.p0i8.p0i8.i64(i8* %0, i8* getelementptr inbounds (%struct.s1, %struct.s1* @caller1.p1, i32 0, i32 0), i64 1, i32 1, i1 false)
|
2012-09-19 17:42:13 +02:00
|
|
|
%1 = bitcast %struct.s2* %p2 to i8*
|
2015-11-19 06:56:52 +01:00
|
|
|
call void @llvm.memcpy.p0i8.p0i8.i64(i8* %1, i8* bitcast (%struct.s2* @caller1.p2 to i8*), i64 2, i32 2, i1 false)
|
2012-09-19 17:42:13 +02:00
|
|
|
%2 = bitcast %struct.s3* %p3 to i8*
|
2015-11-19 06:56:52 +01:00
|
|
|
call void @llvm.memcpy.p0i8.p0i8.i64(i8* %2, i8* bitcast ({ i16, i8, i8 }* @caller1.p3 to i8*), i64 4, i32 2, i1 false)
|
2012-09-19 17:42:13 +02:00
|
|
|
%3 = bitcast %struct.s4* %p4 to i8*
|
2015-11-19 06:56:52 +01:00
|
|
|
call void @llvm.memcpy.p0i8.p0i8.i64(i8* %3, i8* bitcast (%struct.s4* @caller1.p4 to i8*), i64 4, i32 4, i1 false)
|
2012-09-19 17:42:13 +02:00
|
|
|
%4 = bitcast %struct.s5* %p5 to i8*
|
2015-11-19 06:56:52 +01:00
|
|
|
call void @llvm.memcpy.p0i8.p0i8.i64(i8* %4, i8* bitcast ({ i32, i8, [3 x i8] }* @caller1.p5 to i8*), i64 8, i32 4, i1 false)
|
2012-09-19 17:42:13 +02:00
|
|
|
%5 = bitcast %struct.s6* %p6 to i8*
|
2015-11-19 06:56:52 +01:00
|
|
|
call void @llvm.memcpy.p0i8.p0i8.i64(i8* %5, i8* bitcast ({ i32, i16, [2 x i8] }* @caller1.p6 to i8*), i64 8, i32 4, i1 false)
|
2012-09-19 17:42:13 +02:00
|
|
|
%6 = bitcast %struct.s7* %p7 to i8*
|
2015-11-19 06:56:52 +01:00
|
|
|
call void @llvm.memcpy.p0i8.p0i8.i64(i8* %6, i8* bitcast ({ i32, i16, i8, i8 }* @caller1.p7 to i8*), i64 8, i32 4, i1 false)
|
2012-09-19 17:42:13 +02:00
|
|
|
%call = call i32 @callee1(%struct.s1* byval %p1, %struct.s2* byval %p2, %struct.s3* byval %p3, %struct.s4* byval %p4, %struct.s5* byval %p5, %struct.s6* byval %p6, %struct.s7* byval %p7)
|
|
|
|
|
ret i32 %call
|
|
|
|
|
|
2013-03-14 20:34:32 +01:00
|
|
|
; CHECK: ld 9, 112(31)
|
|
|
|
|
; CHECK: ld 8, 120(31)
|
|
|
|
|
; CHECK: ld 7, 128(31)
|
|
|
|
|
; CHECK: lwz 6, 136(31)
|
|
|
|
|
; CHECK: lwz 5, 144(31)
|
|
|
|
|
; CHECK: lhz 4, 152(31)
|
|
|
|
|
; CHECK: lbz 3, 160(31)
|
2012-09-19 17:42:13 +02:00
|
|
|
}
|
|
|
|
|
|
2015-11-19 06:56:52 +01:00
|
|
|
declare void @llvm.memcpy.p0i8.p0i8.i64(i8* nocapture, i8* nocapture, i64, i32, i1) nounwind
|
2012-09-19 17:42:13 +02:00
|
|
|
|
|
|
|
|
define internal i32 @callee1(%struct.s1* byval %v1, %struct.s2* byval %v2, %struct.s3* byval %v3, %struct.s4* byval %v4, %struct.s5* byval %v5, %struct.s6* byval %v6, %struct.s7* byval %v7) nounwind {
|
|
|
|
|
entry:
|
[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
|
|
|
%a = getelementptr inbounds %struct.s1, %struct.s1* %v1, i32 0, i32 0
|
2015-02-27 22:17:42 +01:00
|
|
|
%0 = load i8, i8* %a, align 1
|
2012-09-19 17:42:13 +02:00
|
|
|
%conv = zext i8 %0 to i32
|
[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
|
|
|
%a1 = getelementptr inbounds %struct.s2, %struct.s2* %v2, i32 0, i32 0
|
2015-02-27 22:17:42 +01:00
|
|
|
%1 = load i16, i16* %a1, align 2
|
2012-09-19 17:42:13 +02:00
|
|
|
%conv2 = sext i16 %1 to i32
|
|
|
|
|
%add = add nsw i32 %conv, %conv2
|
[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
|
|
|
%a3 = getelementptr inbounds %struct.s3, %struct.s3* %v3, i32 0, i32 0
|
2015-02-27 22:17:42 +01:00
|
|
|
%2 = load i16, i16* %a3, align 2
|
2012-09-19 17:42:13 +02:00
|
|
|
%conv4 = sext i16 %2 to i32
|
|
|
|
|
%add5 = add nsw i32 %add, %conv4
|
[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
|
|
|
%a6 = getelementptr inbounds %struct.s4, %struct.s4* %v4, i32 0, i32 0
|
2015-02-27 22:17:42 +01:00
|
|
|
%3 = load i32, i32* %a6, align 4
|
2012-09-19 17:42:13 +02:00
|
|
|
%add7 = add nsw i32 %add5, %3
|
[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
|
|
|
%a8 = getelementptr inbounds %struct.s5, %struct.s5* %v5, i32 0, i32 0
|
2015-02-27 22:17:42 +01:00
|
|
|
%4 = load i32, i32* %a8, align 4
|
2012-09-19 17:42:13 +02:00
|
|
|
%add9 = add nsw i32 %add7, %4
|
[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
|
|
|
%a10 = getelementptr inbounds %struct.s6, %struct.s6* %v6, i32 0, i32 0
|
2015-02-27 22:17:42 +01:00
|
|
|
%5 = load i32, i32* %a10, align 4
|
2012-09-19 17:42:13 +02:00
|
|
|
%add11 = add nsw i32 %add9, %5
|
[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
|
|
|
%a12 = getelementptr inbounds %struct.s7, %struct.s7* %v7, i32 0, i32 0
|
2015-02-27 22:17:42 +01:00
|
|
|
%6 = load i32, i32* %a12, align 4
|
2012-09-19 17:42:13 +02:00
|
|
|
%add13 = add nsw i32 %add11, %6
|
|
|
|
|
ret i32 %add13
|
|
|
|
|
|
|
|
|
|
; CHECK: std 9, 96(1)
|
|
|
|
|
; CHECK: std 8, 88(1)
|
|
|
|
|
; CHECK: std 7, 80(1)
|
This patch addresses PR13949.
For the PowerPC 64-bit ELF Linux ABI, aggregates of size less than 8
bytes are to be passed in the low-order bits ("right-adjusted") of the
doubleword register or memory slot assigned to them. A previous patch
addressed this for aggregates passed in registers. However, small
aggregates passed in the overflow portion of the parameter save area are
still being passed left-adjusted.
The fix is made in PPCTargetLowering::LowerCall_Darwin_Or_64SVR4 on the
caller side, and in PPCTargetLowering::LowerFormalArguments_64SVR4 on
the callee side. The main fix on the callee side simply extends
existing logic for 1- and 2-byte objects to 1- through 7-byte objects,
and correcting a constant left over from 32-bit code. There is also a
fix to a bogus calculation of the offset to the following argument in
the parameter save area.
On the caller side, again a constant left over from 32-bit code is
fixed. Additionally, some code for 1, 2, and 4-byte objects is
duplicated to handle the 3, 5, 6, and 7-byte objects for SVR4 only. The
LowerCall_Darwin_Or_64SVR4 logic is getting fairly convoluted trying to
handle both ABIs, and I propose to separate this into two functions in a
future patch, at which time the duplication can be removed.
The patch adds a new test (structsinmem.ll) to demonstrate correct
passing of structures of all seven sizes. Eight dummy parameters are
used to force these structures to be in the overflow portion of the
parameter save area.
As a side effect, this corrects the case when aggregates passed in
registers are saved into the first eight doublewords of the parameter
save area: Previously they were stored left-justified, and now are
properly stored right-justified. This requires changing the expected
output of existing test case structsinregs.ll.
llvm-svn: 166022
2012-10-16 15:30:53 +02:00
|
|
|
; CHECK: stw 6, 76(1)
|
|
|
|
|
; CHECK: stw 5, 68(1)
|
|
|
|
|
; CHECK: sth 4, 62(1)
|
|
|
|
|
; CHECK: stb 3, 55(1)
|
|
|
|
|
; CHECK: lha {{[0-9]+}}, 62(1)
|
|
|
|
|
; CHECK: lha {{[0-9]+}}, 68(1)
|
2013-03-27 03:40:14 +01:00
|
|
|
; CHECK: lbz {{[0-9]+}}, 55(1)
|
This patch addresses PR13949.
For the PowerPC 64-bit ELF Linux ABI, aggregates of size less than 8
bytes are to be passed in the low-order bits ("right-adjusted") of the
doubleword register or memory slot assigned to them. A previous patch
addressed this for aggregates passed in registers. However, small
aggregates passed in the overflow portion of the parameter save area are
still being passed left-adjusted.
The fix is made in PPCTargetLowering::LowerCall_Darwin_Or_64SVR4 on the
caller side, and in PPCTargetLowering::LowerFormalArguments_64SVR4 on
the callee side. The main fix on the callee side simply extends
existing logic for 1- and 2-byte objects to 1- through 7-byte objects,
and correcting a constant left over from 32-bit code. There is also a
fix to a bogus calculation of the offset to the following argument in
the parameter save area.
On the caller side, again a constant left over from 32-bit code is
fixed. Additionally, some code for 1, 2, and 4-byte objects is
duplicated to handle the 3, 5, 6, and 7-byte objects for SVR4 only. The
LowerCall_Darwin_Or_64SVR4 logic is getting fairly convoluted trying to
handle both ABIs, and I propose to separate this into two functions in a
future patch, at which time the duplication can be removed.
The patch adds a new test (structsinmem.ll) to demonstrate correct
passing of structures of all seven sizes. Eight dummy parameters are
used to force these structures to be in the overflow portion of the
parameter save area.
As a side effect, this corrects the case when aggregates passed in
registers are saved into the first eight doublewords of the parameter
save area: Previously they were stored left-justified, and now are
properly stored right-justified. This requires changing the expected
output of existing test case structsinregs.ll.
llvm-svn: 166022
2012-10-16 15:30:53 +02:00
|
|
|
; CHECK: lwz {{[0-9]+}}, 76(1)
|
2012-09-19 17:42:13 +02:00
|
|
|
; CHECK: lwz {{[0-9]+}}, 80(1)
|
|
|
|
|
; CHECK: lwz {{[0-9]+}}, 88(1)
|
|
|
|
|
; CHECK: lwz {{[0-9]+}}, 96(1)
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
define i32 @caller2() nounwind {
|
|
|
|
|
entry:
|
|
|
|
|
%p1 = alloca %struct.t1, align 1
|
|
|
|
|
%p2 = alloca %struct.t2, align 1
|
|
|
|
|
%p3 = alloca %struct.t3, align 1
|
|
|
|
|
%p4 = alloca %struct.t4, align 1
|
|
|
|
|
%p5 = alloca %struct.t5, align 1
|
|
|
|
|
%p6 = alloca %struct.t6, align 1
|
|
|
|
|
%p7 = alloca %struct.t7, align 1
|
|
|
|
|
%0 = bitcast %struct.t1* %p1 to i8*
|
2015-11-19 06:56:52 +01:00
|
|
|
call void @llvm.memcpy.p0i8.p0i8.i64(i8* %0, i8* getelementptr inbounds (%struct.t1, %struct.t1* @caller2.p1, i32 0, i32 0), i64 1, i32 1, i1 false)
|
2012-09-19 17:42:13 +02:00
|
|
|
%1 = bitcast %struct.t2* %p2 to i8*
|
2015-11-19 06:56:52 +01:00
|
|
|
call void @llvm.memcpy.p0i8.p0i8.i64(i8* %1, i8* bitcast ({ i16 }* @caller2.p2 to i8*), i64 2, i32 1, i1 false)
|
2012-09-19 17:42:13 +02:00
|
|
|
%2 = bitcast %struct.t3* %p3 to i8*
|
2015-11-19 06:56:52 +01:00
|
|
|
call void @llvm.memcpy.p0i8.p0i8.i64(i8* %2, i8* bitcast (%struct.t3* @caller2.p3 to i8*), i64 3, i32 1, i1 false)
|
2012-09-19 17:42:13 +02:00
|
|
|
%3 = bitcast %struct.t4* %p4 to i8*
|
2015-11-19 06:56:52 +01:00
|
|
|
call void @llvm.memcpy.p0i8.p0i8.i64(i8* %3, i8* bitcast ({ i32 }* @caller2.p4 to i8*), i64 4, i32 1, i1 false)
|
2012-09-19 17:42:13 +02:00
|
|
|
%4 = bitcast %struct.t5* %p5 to i8*
|
2015-11-19 06:56:52 +01:00
|
|
|
call void @llvm.memcpy.p0i8.p0i8.i64(i8* %4, i8* bitcast (%struct.t5* @caller2.p5 to i8*), i64 5, i32 1, i1 false)
|
2012-09-19 17:42:13 +02:00
|
|
|
%5 = bitcast %struct.t6* %p6 to i8*
|
2015-11-19 06:56:52 +01:00
|
|
|
call void @llvm.memcpy.p0i8.p0i8.i64(i8* %5, i8* bitcast (%struct.t6* @caller2.p6 to i8*), i64 6, i32 1, i1 false)
|
2012-09-19 17:42:13 +02:00
|
|
|
%6 = bitcast %struct.t7* %p7 to i8*
|
2015-11-19 06:56:52 +01:00
|
|
|
call void @llvm.memcpy.p0i8.p0i8.i64(i8* %6, i8* bitcast (%struct.t7* @caller2.p7 to i8*), i64 7, i32 1, i1 false)
|
2012-09-19 17:42:13 +02:00
|
|
|
%call = call i32 @callee2(%struct.t1* byval %p1, %struct.t2* byval %p2, %struct.t3* byval %p3, %struct.t4* byval %p4, %struct.t5* byval %p5, %struct.t6* byval %p6, %struct.t7* byval %p7)
|
|
|
|
|
ret i32 %call
|
|
|
|
|
|
|
|
|
|
; CHECK: stb {{[0-9]+}}, 71(1)
|
|
|
|
|
; CHECK: sth {{[0-9]+}}, 69(1)
|
|
|
|
|
; CHECK: stb {{[0-9]+}}, 87(1)
|
|
|
|
|
; CHECK: stw {{[0-9]+}}, 83(1)
|
|
|
|
|
; CHECK: sth {{[0-9]+}}, 94(1)
|
|
|
|
|
; CHECK: stw {{[0-9]+}}, 90(1)
|
|
|
|
|
; CHECK: stb {{[0-9]+}}, 103(1)
|
|
|
|
|
; CHECK: sth {{[0-9]+}}, 101(1)
|
|
|
|
|
; CHECK: stw {{[0-9]+}}, 97(1)
|
|
|
|
|
; CHECK: ld 9, 96(1)
|
|
|
|
|
; CHECK: ld 8, 88(1)
|
|
|
|
|
; CHECK: ld 7, 80(1)
|
2013-03-14 20:34:32 +01:00
|
|
|
; CHECK: lwz 6, 136(31)
|
2012-09-19 17:42:13 +02:00
|
|
|
; CHECK: ld 5, 64(1)
|
2013-03-14 20:34:32 +01:00
|
|
|
; CHECK: lhz 4, 152(31)
|
|
|
|
|
; CHECK: lbz 3, 160(31)
|
2012-09-19 17:42:13 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
define internal i32 @callee2(%struct.t1* byval %v1, %struct.t2* byval %v2, %struct.t3* byval %v3, %struct.t4* byval %v4, %struct.t5* byval %v5, %struct.t6* byval %v6, %struct.t7* byval %v7) nounwind {
|
|
|
|
|
entry:
|
[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
|
|
|
%a = getelementptr inbounds %struct.t1, %struct.t1* %v1, i32 0, i32 0
|
2015-02-27 22:17:42 +01:00
|
|
|
%0 = load i8, i8* %a, align 1
|
2012-09-19 17:42:13 +02:00
|
|
|
%conv = zext i8 %0 to i32
|
[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
|
|
|
%a1 = getelementptr inbounds %struct.t2, %struct.t2* %v2, i32 0, i32 0
|
2015-02-27 22:17:42 +01:00
|
|
|
%1 = load i16, i16* %a1, align 1
|
2012-09-19 17:42:13 +02:00
|
|
|
%conv2 = sext i16 %1 to i32
|
|
|
|
|
%add = add nsw i32 %conv, %conv2
|
[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
|
|
|
%a3 = getelementptr inbounds %struct.t3, %struct.t3* %v3, i32 0, i32 0
|
2015-02-27 22:17:42 +01:00
|
|
|
%2 = load i16, i16* %a3, align 1
|
2012-09-19 17:42:13 +02:00
|
|
|
%conv4 = sext i16 %2 to i32
|
|
|
|
|
%add5 = add nsw i32 %add, %conv4
|
[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
|
|
|
%a6 = getelementptr inbounds %struct.t4, %struct.t4* %v4, i32 0, i32 0
|
2015-02-27 22:17:42 +01:00
|
|
|
%3 = load i32, i32* %a6, align 1
|
2012-09-19 17:42:13 +02:00
|
|
|
%add7 = add nsw i32 %add5, %3
|
[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
|
|
|
%a8 = getelementptr inbounds %struct.t5, %struct.t5* %v5, i32 0, i32 0
|
2015-02-27 22:17:42 +01:00
|
|
|
%4 = load i32, i32* %a8, align 1
|
2012-09-19 17:42:13 +02:00
|
|
|
%add9 = add nsw i32 %add7, %4
|
[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
|
|
|
%a10 = getelementptr inbounds %struct.t6, %struct.t6* %v6, i32 0, i32 0
|
2015-02-27 22:17:42 +01:00
|
|
|
%5 = load i32, i32* %a10, align 1
|
2012-09-19 17:42:13 +02:00
|
|
|
%add11 = add nsw i32 %add9, %5
|
[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
|
|
|
%a12 = getelementptr inbounds %struct.t7, %struct.t7* %v7, i32 0, i32 0
|
2015-02-27 22:17:42 +01:00
|
|
|
%6 = load i32, i32* %a12, align 1
|
2012-09-19 17:42:13 +02:00
|
|
|
%add13 = add nsw i32 %add11, %6
|
|
|
|
|
ret i32 %add13
|
|
|
|
|
|
|
|
|
|
; CHECK: std 9, 96(1)
|
|
|
|
|
; CHECK: std 8, 88(1)
|
|
|
|
|
; CHECK: std 7, 80(1)
|
This patch addresses a PPC64 ELF issue with passing parameters consisting of
structs having size 3, 5, 6, or 7. Such a struct must be passed and received
as right-justified within its register or memory slot. The problem is only
present for structs that are passed in registers.
Previously, as part of a patch handling all structs of size less than 8, I
added logic to rotate the incoming register so that the struct was left-
justified prior to storing the whole register. This was incorrect because
the address of the parameter had already been adjusted earlier to point to
the right-adjusted value in the storage slot. Essentially I had accidentally
accounted for the right-adjustment twice.
In this patch, I removed the incorrect logic and reorganized the code to make
the flow clearer.
The removal of the rotates changes the expected code generation, so test case
structsinregs.ll has been modified to reflect this. I also added a new test
case, jaggedstructs.ll, to demonstrate that structs of these sizes can now
be properly received and passed.
I've built and tested the code on powerpc64-unknown-linux-gnu with no new
regressions. I also ran the GCC compatibility test suite and verified that
earlier problems with these structs are now resolved, with no new regressions.
llvm-svn: 166680
2012-10-25 15:38:09 +02:00
|
|
|
; CHECK: stw 6, 76(1)
|
2012-09-19 17:42:13 +02:00
|
|
|
; CHECK: std 5, 64(1)
|
This patch addresses a PPC64 ELF issue with passing parameters consisting of
structs having size 3, 5, 6, or 7. Such a struct must be passed and received
as right-justified within its register or memory slot. The problem is only
present for structs that are passed in registers.
Previously, as part of a patch handling all structs of size less than 8, I
added logic to rotate the incoming register so that the struct was left-
justified prior to storing the whole register. This was incorrect because
the address of the parameter had already been adjusted earlier to point to
the right-adjusted value in the storage slot. Essentially I had accidentally
accounted for the right-adjustment twice.
In this patch, I removed the incorrect logic and reorganized the code to make
the flow clearer.
The removal of the rotates changes the expected code generation, so test case
structsinregs.ll has been modified to reflect this. I also added a new test
case, jaggedstructs.ll, to demonstrate that structs of these sizes can now
be properly received and passed.
I've built and tested the code on powerpc64-unknown-linux-gnu with no new
regressions. I also ran the GCC compatibility test suite and verified that
earlier problems with these structs are now resolved, with no new regressions.
llvm-svn: 166680
2012-10-25 15:38:09 +02:00
|
|
|
; CHECK: sth 4, 62(1)
|
|
|
|
|
; CHECK: stb 3, 55(1)
|
This patch addresses PR13949.
For the PowerPC 64-bit ELF Linux ABI, aggregates of size less than 8
bytes are to be passed in the low-order bits ("right-adjusted") of the
doubleword register or memory slot assigned to them. A previous patch
addressed this for aggregates passed in registers. However, small
aggregates passed in the overflow portion of the parameter save area are
still being passed left-adjusted.
The fix is made in PPCTargetLowering::LowerCall_Darwin_Or_64SVR4 on the
caller side, and in PPCTargetLowering::LowerFormalArguments_64SVR4 on
the callee side. The main fix on the callee side simply extends
existing logic for 1- and 2-byte objects to 1- through 7-byte objects,
and correcting a constant left over from 32-bit code. There is also a
fix to a bogus calculation of the offset to the following argument in
the parameter save area.
On the caller side, again a constant left over from 32-bit code is
fixed. Additionally, some code for 1, 2, and 4-byte objects is
duplicated to handle the 3, 5, 6, and 7-byte objects for SVR4 only. The
LowerCall_Darwin_Or_64SVR4 logic is getting fairly convoluted trying to
handle both ABIs, and I propose to separate this into two functions in a
future patch, at which time the duplication can be removed.
The patch adds a new test (structsinmem.ll) to demonstrate correct
passing of structures of all seven sizes. Eight dummy parameters are
used to force these structures to be in the overflow portion of the
parameter save area.
As a side effect, this corrects the case when aggregates passed in
registers are saved into the first eight doublewords of the parameter
save area: Previously they were stored left-justified, and now are
properly stored right-justified. This requires changing the expected
output of existing test case structsinregs.ll.
llvm-svn: 166022
2012-10-16 15:30:53 +02:00
|
|
|
; CHECK: lha {{[0-9]+}}, 62(1)
|
2013-03-18 23:12:04 +01:00
|
|
|
; CHECK: lha {{[0-9]+}}, 69(1)
|
2013-03-27 03:40:14 +01:00
|
|
|
; CHECK: lbz {{[0-9]+}}, 55(1)
|
This patch addresses PR13949.
For the PowerPC 64-bit ELF Linux ABI, aggregates of size less than 8
bytes are to be passed in the low-order bits ("right-adjusted") of the
doubleword register or memory slot assigned to them. A previous patch
addressed this for aggregates passed in registers. However, small
aggregates passed in the overflow portion of the parameter save area are
still being passed left-adjusted.
The fix is made in PPCTargetLowering::LowerCall_Darwin_Or_64SVR4 on the
caller side, and in PPCTargetLowering::LowerFormalArguments_64SVR4 on
the callee side. The main fix on the callee side simply extends
existing logic for 1- and 2-byte objects to 1- through 7-byte objects,
and correcting a constant left over from 32-bit code. There is also a
fix to a bogus calculation of the offset to the following argument in
the parameter save area.
On the caller side, again a constant left over from 32-bit code is
fixed. Additionally, some code for 1, 2, and 4-byte objects is
duplicated to handle the 3, 5, 6, and 7-byte objects for SVR4 only. The
LowerCall_Darwin_Or_64SVR4 logic is getting fairly convoluted trying to
handle both ABIs, and I propose to separate this into two functions in a
future patch, at which time the duplication can be removed.
The patch adds a new test (structsinmem.ll) to demonstrate correct
passing of structures of all seven sizes. Eight dummy parameters are
used to force these structures to be in the overflow portion of the
parameter save area.
As a side effect, this corrects the case when aggregates passed in
registers are saved into the first eight doublewords of the parameter
save area: Previously they were stored left-justified, and now are
properly stored right-justified. This requires changing the expected
output of existing test case structsinregs.ll.
llvm-svn: 166022
2012-10-16 15:30:53 +02:00
|
|
|
; CHECK: lwz {{[0-9]+}}, 76(1)
|
2013-03-18 23:12:04 +01:00
|
|
|
; CHECK: lwz {{[0-9]+}}, 83(1)
|
|
|
|
|
; CHECK: lwz {{[0-9]+}}, 90(1)
|
|
|
|
|
; CHECK: lwz {{[0-9]+}}, 97(1)
|
2012-09-19 17:42:13 +02:00
|
|
|
}
|
