llvm-6502/test/CodeGen/ARM/vst1.ll

141 lines
4.6 KiB
LLVM
Raw Normal View History

; RUN: llc -mtriple=arm-eabi -mattr=+neon %s -o - | FileCheck %s
define void @vst1i8(i8* %A, <8 x i8>* %B) nounwind {
;CHECK-LABEL: vst1i8:
;Check the alignment value. Max for this instruction is 64 bits:
;CHECK: vst1.8 {d16}, [r0:64]
%tmp1 = load <8 x i8>, <8 x i8>* %B
call void @llvm.arm.neon.vst1.v8i8(i8* %A, <8 x i8> %tmp1, i32 16)
ret void
}
define void @vst1i16(i16* %A, <4 x i16>* %B) nounwind {
;CHECK-LABEL: vst1i16:
;CHECK: vst1.16
%tmp0 = bitcast i16* %A to i8*
%tmp1 = load <4 x i16>, <4 x i16>* %B
call void @llvm.arm.neon.vst1.v4i16(i8* %tmp0, <4 x i16> %tmp1, i32 1)
ret void
}
define void @vst1i32(i32* %A, <2 x i32>* %B) nounwind {
;CHECK-LABEL: vst1i32:
;CHECK: vst1.32
%tmp0 = bitcast i32* %A to i8*
%tmp1 = load <2 x i32>, <2 x i32>* %B
call void @llvm.arm.neon.vst1.v2i32(i8* %tmp0, <2 x i32> %tmp1, i32 1)
ret void
}
define void @vst1f(float* %A, <2 x float>* %B) nounwind {
;CHECK-LABEL: vst1f:
;CHECK: vst1.32
%tmp0 = bitcast float* %A to i8*
%tmp1 = load <2 x float>, <2 x float>* %B
call void @llvm.arm.neon.vst1.v2f32(i8* %tmp0, <2 x float> %tmp1, i32 1)
ret void
}
;Check for a post-increment updating store.
define void @vst1f_update(float** %ptr, <2 x float>* %B) nounwind {
;CHECK-LABEL: vst1f_update:
;CHECK: vst1.32 {d16}, [r1]!
%A = load float*, float** %ptr
%tmp0 = bitcast float* %A to i8*
%tmp1 = load <2 x float>, <2 x float>* %B
call void @llvm.arm.neon.vst1.v2f32(i8* %tmp0, <2 x float> %tmp1, i32 1)
[opaque pointer type] Add textual IR support for explicit type parameter to getelementptr instruction One of several parallel first steps to remove the target type of pointers, replacing them with a single opaque pointer type. This adds an explicit type parameter to the gep instruction so that when the first parameter becomes an opaque pointer type, the type to gep through is still available to the instructions. * This doesn't modify gep operators, only instructions (operators will be handled separately) * Textual IR changes only. Bitcode (including upgrade) and changing the in-memory representation will be in separate changes. * geps of vectors are transformed as: getelementptr <4 x float*> %x, ... ->getelementptr float, <4 x float*> %x, ... Then, once the opaque pointer type is introduced, this will ultimately look like: getelementptr float, <4 x ptr> %x with the unambiguous interpretation that it is a vector of pointers to float. * address spaces remain on the pointer, not the type: getelementptr float addrspace(1)* %x ->getelementptr float, float addrspace(1)* %x Then, eventually: getelementptr float, ptr addrspace(1) %x Importantly, the massive amount of test case churn has been automated by same crappy python code. I had to manually update a few test cases that wouldn't fit the script's model (r228970,r229196,r229197,r229198). The python script just massages stdin and writes the result to stdout, I then wrapped that in a shell script to handle replacing files, then using the usual find+xargs to migrate all the files. update.py: import fileinput import sys import re ibrep = re.compile(r"(^.*?[^%\w]getelementptr inbounds )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))") normrep = re.compile( r"(^.*?[^%\w]getelementptr )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))") def conv(match, line): if not match: return line line = match.groups()[0] if len(match.groups()[5]) == 0: line += match.groups()[2] line += match.groups()[3] line += ", " line += match.groups()[1] line += "\n" return line for line in sys.stdin: if line.find("getelementptr ") == line.find("getelementptr inbounds"): if line.find("getelementptr inbounds") != line.find("getelementptr inbounds ("): line = conv(re.match(ibrep, line), line) elif line.find("getelementptr ") != line.find("getelementptr ("): line = conv(re.match(normrep, line), line) sys.stdout.write(line) apply.sh: for name in "$@" do python3 `dirname "$0"`/update.py < "$name" > "$name.tmp" && mv "$name.tmp" "$name" rm -f "$name.tmp" done The actual commands: From llvm/src: find test/ -name *.ll | xargs ./apply.sh From llvm/src/tools/clang: find test/ -name *.mm -o -name *.m -o -name *.cpp -o -name *.c | xargs -I '{}' ../../apply.sh "{}" From llvm/src/tools/polly: find test/ -name *.ll | xargs ./apply.sh After that, check-all (with llvm, clang, clang-tools-extra, lld, compiler-rt, and polly all checked out). The extra 'rm' in the apply.sh script is due to a few files in clang's test suite using interesting unicode stuff that my python script was throwing exceptions on. None of those files needed to be migrated, so it seemed sufficient to ignore those cases. Reviewers: rafael, dexonsmith, grosser Differential Revision: http://reviews.llvm.org/D7636 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
%tmp2 = getelementptr float, float* %A, i32 2
store float* %tmp2, float** %ptr
ret void
}
define void @vst1i64(i64* %A, <1 x i64>* %B) nounwind {
;CHECK-LABEL: vst1i64:
;CHECK: vst1.64
%tmp0 = bitcast i64* %A to i8*
%tmp1 = load <1 x i64>, <1 x i64>* %B
call void @llvm.arm.neon.vst1.v1i64(i8* %tmp0, <1 x i64> %tmp1, i32 1)
ret void
}
define void @vst1Qi8(i8* %A, <16 x i8>* %B) nounwind {
;CHECK-LABEL: vst1Qi8:
;Check the alignment value. Max for this instruction is 128 bits:
;CHECK: vst1.8 {d16, d17}, [r0:64]
%tmp1 = load <16 x i8>, <16 x i8>* %B
call void @llvm.arm.neon.vst1.v16i8(i8* %A, <16 x i8> %tmp1, i32 8)
ret void
}
define void @vst1Qi16(i16* %A, <8 x i16>* %B) nounwind {
;CHECK-LABEL: vst1Qi16:
;Check the alignment value. Max for this instruction is 128 bits:
;CHECK: vst1.16 {d16, d17}, [r0:128]
%tmp0 = bitcast i16* %A to i8*
%tmp1 = load <8 x i16>, <8 x i16>* %B
call void @llvm.arm.neon.vst1.v8i16(i8* %tmp0, <8 x i16> %tmp1, i32 32)
ret void
}
;Check for a post-increment updating store with register increment.
define void @vst1Qi16_update(i16** %ptr, <8 x i16>* %B, i32 %inc) nounwind {
;CHECK-LABEL: vst1Qi16_update:
;CHECK: vst1.16 {d16, d17}, [r1:64], r2
%A = load i16*, i16** %ptr
%tmp0 = bitcast i16* %A to i8*
%tmp1 = load <8 x i16>, <8 x i16>* %B
call void @llvm.arm.neon.vst1.v8i16(i8* %tmp0, <8 x i16> %tmp1, i32 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
%tmp2 = getelementptr i16, i16* %A, i32 %inc
store i16* %tmp2, i16** %ptr
ret void
}
define void @vst1Qi32(i32* %A, <4 x i32>* %B) nounwind {
;CHECK-LABEL: vst1Qi32:
;CHECK: vst1.32
%tmp0 = bitcast i32* %A to i8*
%tmp1 = load <4 x i32>, <4 x i32>* %B
call void @llvm.arm.neon.vst1.v4i32(i8* %tmp0, <4 x i32> %tmp1, i32 1)
ret void
}
define void @vst1Qf(float* %A, <4 x float>* %B) nounwind {
;CHECK-LABEL: vst1Qf:
;CHECK: vst1.32
%tmp0 = bitcast float* %A to i8*
%tmp1 = load <4 x float>, <4 x float>* %B
call void @llvm.arm.neon.vst1.v4f32(i8* %tmp0, <4 x float> %tmp1, i32 1)
ret void
}
define void @vst1Qi64(i64* %A, <2 x i64>* %B) nounwind {
;CHECK-LABEL: vst1Qi64:
;CHECK: vst1.64
%tmp0 = bitcast i64* %A to i8*
%tmp1 = load <2 x i64>, <2 x i64>* %B
call void @llvm.arm.neon.vst1.v2i64(i8* %tmp0, <2 x i64> %tmp1, i32 1)
ret void
}
define void @vst1Qf64(double* %A, <2 x double>* %B) nounwind {
;CHECK-LABEL: vst1Qf64:
;CHECK: vst1.64
%tmp0 = bitcast double* %A to i8*
%tmp1 = load <2 x double>, <2 x double>* %B
call void @llvm.arm.neon.vst1.v2f64(i8* %tmp0, <2 x double> %tmp1, i32 1)
ret void
}
declare void @llvm.arm.neon.vst1.v8i8(i8*, <8 x i8>, i32) nounwind
declare void @llvm.arm.neon.vst1.v4i16(i8*, <4 x i16>, i32) nounwind
declare void @llvm.arm.neon.vst1.v2i32(i8*, <2 x i32>, i32) nounwind
declare void @llvm.arm.neon.vst1.v2f32(i8*, <2 x float>, i32) nounwind
declare void @llvm.arm.neon.vst1.v1i64(i8*, <1 x i64>, i32) nounwind
declare void @llvm.arm.neon.vst1.v16i8(i8*, <16 x i8>, i32) nounwind
declare void @llvm.arm.neon.vst1.v8i16(i8*, <8 x i16>, i32) nounwind
declare void @llvm.arm.neon.vst1.v4i32(i8*, <4 x i32>, i32) nounwind
declare void @llvm.arm.neon.vst1.v4f32(i8*, <4 x float>, i32) nounwind
declare void @llvm.arm.neon.vst1.v2i64(i8*, <2 x i64>, i32) nounwind
declare void @llvm.arm.neon.vst1.v2f64(i8*, <2 x double>, i32) nounwind