2013-05-06 16:17:29 +00:00
|
|
|
; Test the handling of the frame pointer (%r11).
|
|
|
|
|
;
|
|
|
|
|
; RUN: llc < %s -mtriple=s390x-linux-gnu -disable-fp-elim | FileCheck %s
|
|
|
|
|
|
|
|
|
|
; We should always initialise %r11 when FP elimination is disabled.
|
|
|
|
|
; We don't need to allocate any more than the caller-provided 160-byte
|
|
|
|
|
; area though.
|
|
|
|
|
define i32 @f1(i32 %x) {
|
2013-07-14 06:24:09 +00:00
|
|
|
; CHECK-LABEL: f1:
|
2013-05-06 16:17:29 +00:00
|
|
|
; CHECK: stmg %r11, %r15, 88(%r15)
|
|
|
|
|
; CHECK: .cfi_offset %r11, -72
|
|
|
|
|
; CHECK: .cfi_offset %r15, -40
|
|
|
|
|
; CHECK-NOT: ag
|
|
|
|
|
; CHECK: lgr %r11, %r15
|
|
|
|
|
; CHECK: .cfi_def_cfa_register %r11
|
|
|
|
|
; CHECK: lmg %r11, %r15, 88(%r11)
|
|
|
|
|
; CHECK: br %r14
|
|
|
|
|
%y = add i32 %x, 1
|
|
|
|
|
ret i32 %y
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
; Make sure that frame accesses after the initial allocation are relative
|
|
|
|
|
; to %r11 rather than %r15.
|
|
|
|
|
define void @f2(i64 %x) {
|
2013-07-14 06:24:09 +00:00
|
|
|
; CHECK-LABEL: f2:
|
2013-05-06 16:17:29 +00:00
|
|
|
; CHECK: stmg %r11, %r15, 88(%r15)
|
|
|
|
|
; CHECK: .cfi_offset %r11, -72
|
|
|
|
|
; CHECK: .cfi_offset %r15, -40
|
|
|
|
|
; CHECK: aghi %r15, -168
|
|
|
|
|
; CHECK: .cfi_def_cfa_offset 328
|
|
|
|
|
; CHECK: lgr %r11, %r15
|
|
|
|
|
; CHECK: .cfi_def_cfa_register %r11
|
|
|
|
|
; CHECK: stg %r2, 160(%r11)
|
|
|
|
|
; CHECK: lmg %r11, %r15, 256(%r11)
|
|
|
|
|
; CHECK: br %r14
|
|
|
|
|
%y = alloca i64, align 8
|
|
|
|
|
store volatile i64 %x, i64* %y
|
|
|
|
|
ret void
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
; This function should require all GPRs but no other spill slots.
|
|
|
|
|
; It shouldn't need to allocate its own frame.
|
|
|
|
|
define void @f3(i32 *%ptr) {
|
2013-07-14 06:24:09 +00:00
|
|
|
; CHECK-LABEL: f3:
|
2013-05-06 16:17:29 +00:00
|
|
|
; CHECK: stmg %r6, %r15, 48(%r15)
|
|
|
|
|
; CHECK-NOT: %r15
|
|
|
|
|
; CHECK-NOT: %r11
|
|
|
|
|
; CHECK: .cfi_offset %r6, -112
|
|
|
|
|
; CHECK: .cfi_offset %r7, -104
|
|
|
|
|
; CHECK: .cfi_offset %r8, -96
|
|
|
|
|
; CHECK: .cfi_offset %r9, -88
|
|
|
|
|
; CHECK: .cfi_offset %r10, -80
|
|
|
|
|
; CHECK: .cfi_offset %r11, -72
|
|
|
|
|
; CHECK: .cfi_offset %r12, -64
|
|
|
|
|
; CHECK: .cfi_offset %r13, -56
|
|
|
|
|
; CHECK: .cfi_offset %r14, -48
|
|
|
|
|
; CHECK: .cfi_offset %r15, -40
|
|
|
|
|
; CHECK-NOT: ag
|
|
|
|
|
; CHECK: lgr %r11, %r15
|
|
|
|
|
; CHECK: .cfi_def_cfa_register %r11
|
|
|
|
|
; ...main function body...
|
|
|
|
|
; CHECK-NOT: %r15
|
|
|
|
|
; CHECK-NOT: %r11
|
|
|
|
|
; CHECK: st {{.*}}, 4(%r2)
|
|
|
|
|
; CHECK: lmg %r6, %r15, 48(%r11)
|
|
|
|
|
; CHECK: br %r14
|
2015-02-27 21:17:42 +00:00
|
|
|
%l0 = load volatile i32 , i32 *%ptr
|
|
|
|
|
%l1 = load volatile i32 , i32 *%ptr
|
|
|
|
|
%l3 = load volatile i32 , i32 *%ptr
|
|
|
|
|
%l4 = load volatile i32 , i32 *%ptr
|
|
|
|
|
%l5 = load volatile i32 , i32 *%ptr
|
|
|
|
|
%l6 = load volatile i32 , i32 *%ptr
|
|
|
|
|
%l7 = load volatile i32 , i32 *%ptr
|
|
|
|
|
%l8 = load volatile i32 , i32 *%ptr
|
|
|
|
|
%l9 = load volatile i32 , i32 *%ptr
|
|
|
|
|
%l10 = load volatile i32 , i32 *%ptr
|
|
|
|
|
%l12 = load volatile i32 , i32 *%ptr
|
|
|
|
|
%l13 = load volatile i32 , i32 *%ptr
|
|
|
|
|
%l14 = load volatile i32 , i32 *%ptr
|
2013-05-06 16:17:29 +00:00
|
|
|
%add0 = add i32 %l0, %l0
|
|
|
|
|
%add1 = add i32 %l1, %add0
|
|
|
|
|
%add3 = add i32 %l3, %add1
|
|
|
|
|
%add4 = add i32 %l4, %add3
|
|
|
|
|
%add5 = add i32 %l5, %add4
|
|
|
|
|
%add6 = add i32 %l6, %add5
|
|
|
|
|
%add7 = add i32 %l7, %add6
|
|
|
|
|
%add8 = add i32 %l8, %add7
|
|
|
|
|
%add9 = add i32 %l9, %add8
|
|
|
|
|
%add10 = add i32 %l10, %add9
|
|
|
|
|
%add12 = add i32 %l12, %add10
|
|
|
|
|
%add13 = add i32 %l13, %add12
|
|
|
|
|
%add14 = add i32 %l14, %add13
|
|
|
|
|
store volatile i32 %add0, i32 *%ptr
|
|
|
|
|
store volatile i32 %add1, i32 *%ptr
|
|
|
|
|
store volatile i32 %add3, i32 *%ptr
|
|
|
|
|
store volatile i32 %add4, i32 *%ptr
|
|
|
|
|
store volatile i32 %add5, i32 *%ptr
|
|
|
|
|
store volatile i32 %add6, i32 *%ptr
|
|
|
|
|
store volatile i32 %add7, i32 *%ptr
|
|
|
|
|
store volatile i32 %add8, i32 *%ptr
|
|
|
|
|
store volatile i32 %add9, i32 *%ptr
|
|
|
|
|
store volatile i32 %add10, i32 *%ptr
|
|
|
|
|
store volatile i32 %add12, i32 *%ptr
|
|
|
|
|
store volatile i32 %add13, i32 *%ptr
|
[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
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@230786 91177308-0d34-0410-b5e6-96231b3b80d8
2015-02-27 19:29:02 +00:00
|
|
|
%final = getelementptr i32, i32 *%ptr, i32 1
|
2013-05-06 16:17:29 +00:00
|
|
|
store volatile i32 %add14, i32 *%final
|
|
|
|
|
ret void
|
|
|
|
|
}
|
|
|
|
|
|
2013-07-05 13:11:52 +00:00
|
|
|
; The largest frame for which the LMG is in range. This frame has two
|
|
|
|
|
; emergency spill slots at 160(%r11), so create a frame of size 524192
|
|
|
|
|
; by allocating (524192 - 176) / 8 = 65502 doublewords.
|
2013-05-06 16:17:29 +00:00
|
|
|
define void @f4(i64 %x) {
|
2013-07-14 06:24:09 +00:00
|
|
|
; CHECK-LABEL: f4:
|
2013-05-06 16:17:29 +00:00
|
|
|
; CHECK: stmg %r11, %r15, 88(%r15)
|
|
|
|
|
; CHECK: .cfi_offset %r11, -72
|
|
|
|
|
; CHECK: .cfi_offset %r15, -40
|
|
|
|
|
; CHECK: agfi %r15, -524192
|
|
|
|
|
; CHECK: .cfi_def_cfa_offset 524352
|
|
|
|
|
; CHECK: lgr %r11, %r15
|
|
|
|
|
; CHECK: .cfi_def_cfa_register %r11
|
2013-07-05 13:11:52 +00:00
|
|
|
; CHECK: stg %r2, 176(%r11)
|
2013-05-06 16:17:29 +00:00
|
|
|
; CHECK-NOT: ag
|
|
|
|
|
; CHECK: lmg %r11, %r15, 524280(%r11)
|
|
|
|
|
; CHECK: br %r14
|
2013-07-05 13:11:52 +00:00
|
|
|
%y = alloca [65502 x i64], align 8
|
[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
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@230786 91177308-0d34-0410-b5e6-96231b3b80d8
2015-02-27 19:29:02 +00:00
|
|
|
%ptr = getelementptr inbounds [65502 x i64], [65502 x i64]* %y, i64 0, i64 0
|
2013-05-06 16:17:29 +00:00
|
|
|
store volatile i64 %x, i64* %ptr
|
|
|
|
|
ret void
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
; The next frame size larger than f4.
|
|
|
|
|
define void @f5(i64 %x) {
|
2013-07-14 06:24:09 +00:00
|
|
|
; CHECK-LABEL: f5:
|
2013-05-06 16:17:29 +00:00
|
|
|
; CHECK: stmg %r11, %r15, 88(%r15)
|
|
|
|
|
; CHECK: .cfi_offset %r11, -72
|
|
|
|
|
; CHECK: .cfi_offset %r15, -40
|
|
|
|
|
; CHECK: agfi %r15, -524200
|
|
|
|
|
; CHECK: .cfi_def_cfa_offset 524360
|
|
|
|
|
; CHECK: lgr %r11, %r15
|
|
|
|
|
; CHECK: .cfi_def_cfa_register %r11
|
2013-07-05 13:11:52 +00:00
|
|
|
; CHECK: stg %r2, 176(%r11)
|
2013-05-06 16:17:29 +00:00
|
|
|
; CHECK: aghi %r11, 8
|
|
|
|
|
; CHECK: lmg %r11, %r15, 524280(%r11)
|
|
|
|
|
; CHECK: br %r14
|
2013-07-05 13:11:52 +00:00
|
|
|
%y = alloca [65503 x i64], align 8
|
[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
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@230786 91177308-0d34-0410-b5e6-96231b3b80d8
2015-02-27 19:29:02 +00:00
|
|
|
%ptr = getelementptr inbounds [65503 x i64], [65503 x i64]* %y, i64 0, i64 0
|
2013-05-06 16:17:29 +00:00
|
|
|
store volatile i64 %x, i64* %ptr
|
|
|
|
|
ret void
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
; The tests above establish that %r11 is handled like %r15 for LMG.
|
|
|
|
|
; Rely on the %r15-based tests in frame-08.ll for other cases.
|
