mirror of
https://github.com/RPCS3/llvm-mirror.git
synced 2024-11-23 03:02:36 +01:00
0d99339102
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
127 lines
3.7 KiB
LLVM
127 lines
3.7 KiB
LLVM
; Test the allocation of frames in cases where we do not need to save
|
|
; registers in the prologue.
|
|
;
|
|
; RUN: llc < %s -mtriple=s390x-linux-gnu | FileCheck %s
|
|
|
|
declare void @foo(i32 *)
|
|
|
|
; The CFA offset is 160 (the caller-allocated part of the frame) + 168.
|
|
define void @f1(i64 %x) {
|
|
; CHECK-LABEL: f1:
|
|
; CHECK: aghi %r15, -168
|
|
; CHECK: .cfi_def_cfa_offset 328
|
|
; CHECK: stg %r2, 160(%r15)
|
|
; CHECK: aghi %r15, 168
|
|
; CHECK: br %r14
|
|
%y = alloca i64, align 8
|
|
store volatile i64 %x, i64* %y
|
|
ret void
|
|
}
|
|
|
|
; Check frames of size 32760, which is the largest size that can be both
|
|
; allocated and freed using AGHI. This size is big enough to require
|
|
; two emergency spill slots at 160(%r15), for instructions with unsigned
|
|
; 12-bit offsets that end up being out of range. Fill the remaining
|
|
; 32760 - 176 bytes by allocating (32760 - 176) / 8 = 4073 doublewords.
|
|
define void @f2(i64 %x) {
|
|
; CHECK-LABEL: f2:
|
|
; CHECK: aghi %r15, -32760
|
|
; CHECK: .cfi_def_cfa_offset 32920
|
|
; CHECK: stg %r2, 176(%r15)
|
|
; CHECK: aghi %r15, 32760
|
|
; CHECK: br %r14
|
|
%y = alloca [4073 x i64], align 8
|
|
%ptr = getelementptr inbounds [4073 x i64], [4073 x i64]* %y, i64 0, i64 0
|
|
store volatile i64 %x, i64* %ptr
|
|
ret void
|
|
}
|
|
|
|
; Allocate one more doubleword. This is the one frame size that we can
|
|
; allocate using AGHI but must free using AGFI.
|
|
define void @f3(i64 %x) {
|
|
; CHECK-LABEL: f3:
|
|
; CHECK: aghi %r15, -32768
|
|
; CHECK: .cfi_def_cfa_offset 32928
|
|
; CHECK: stg %r2, 176(%r15)
|
|
; CHECK: agfi %r15, 32768
|
|
; CHECK: br %r14
|
|
%y = alloca [4074 x i64], align 8
|
|
%ptr = getelementptr inbounds [4074 x i64], [4074 x i64]* %y, i64 0, i64 0
|
|
store volatile i64 %x, i64* %ptr
|
|
ret void
|
|
}
|
|
|
|
; Allocate another doubleword on top of that. The allocation and free
|
|
; must both use AGFI.
|
|
define void @f4(i64 %x) {
|
|
; CHECK-LABEL: f4:
|
|
; CHECK: agfi %r15, -32776
|
|
; CHECK: .cfi_def_cfa_offset 32936
|
|
; CHECK: stg %r2, 176(%r15)
|
|
; CHECK: agfi %r15, 32776
|
|
; CHECK: br %r14
|
|
%y = alloca [4075 x i64], align 8
|
|
%ptr = getelementptr inbounds [4075 x i64], [4075 x i64]* %y, i64 0, i64 0
|
|
store volatile i64 %x, i64* %ptr
|
|
ret void
|
|
}
|
|
|
|
; The largest size that can be both allocated and freed using AGFI.
|
|
; At this point the frame is too big to represent properly in the CFI.
|
|
define void @f5(i64 %x) {
|
|
; CHECK-LABEL: f5:
|
|
; CHECK: agfi %r15, -2147483640
|
|
; CHECK: stg %r2, 176(%r15)
|
|
; CHECK: agfi %r15, 2147483640
|
|
; CHECK: br %r14
|
|
%y = alloca [268435433 x i64], align 8
|
|
%ptr = getelementptr inbounds [268435433 x i64], [268435433 x i64]* %y, i64 0, i64 0
|
|
store volatile i64 %x, i64* %ptr
|
|
ret void
|
|
}
|
|
|
|
; The only frame size that can be allocated using a single AGFI but which
|
|
; must be freed using two instructions.
|
|
define void @f6(i64 %x) {
|
|
; CHECK-LABEL: f6:
|
|
; CHECK: agfi %r15, -2147483648
|
|
; CHECK: stg %r2, 176(%r15)
|
|
; CHECK: agfi %r15, 2147483640
|
|
; CHECK: aghi %r15, 8
|
|
; CHECK: br %r14
|
|
%y = alloca [268435434 x i64], align 8
|
|
%ptr = getelementptr inbounds [268435434 x i64], [268435434 x i64]* %y, i64 0, i64 0
|
|
store volatile i64 %x, i64* %ptr
|
|
ret void
|
|
}
|
|
|
|
; The smallest frame size that needs two instructions to both allocate
|
|
; and free the frame.
|
|
define void @f7(i64 %x) {
|
|
; CHECK-LABEL: f7:
|
|
; CHECK: agfi %r15, -2147483648
|
|
; CHECK: aghi %r15, -8
|
|
; CHECK: stg %r2, 176(%r15)
|
|
; CHECK: agfi %r15, 2147483640
|
|
; CHECK: aghi %r15, 16
|
|
; CHECK: br %r14
|
|
%y = alloca [268435435 x i64], align 8
|
|
%ptr = getelementptr inbounds [268435435 x i64], [268435435 x i64]* %y, i64 0, i64 0
|
|
store volatile i64 %x, i64* %ptr
|
|
ret void
|
|
}
|
|
|
|
; Make sure that LA can be rematerialized.
|
|
define void @f8() {
|
|
; CHECK-LABEL: f8:
|
|
; CHECK: la %r2, 164(%r15)
|
|
; CHECK: brasl %r14, foo@PLT
|
|
; CHECK: la %r2, 164(%r15)
|
|
; CHECK: brasl %r14, foo@PLT
|
|
; CHECK: br %r14
|
|
%ptr = alloca i32
|
|
call void @foo(i32 *%ptr)
|
|
call void @foo(i32 *%ptr)
|
|
ret void
|
|
}
|