RPCS3/llvm-mirror
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-10-29 23:12:55 +01:00
Code Issues Projects Releases Wiki Activity
6e615bab35
llvm-mirror/test/Transforms/InstSimplify/compare.ll

938 lines
20 KiB
LLVM
Raw Normal View History

At -O123 the early-cse pass is run before instcombine has run. According to my auto-simplier the transform most missed by early-cse is (zext X) != 0 -> X != 0. This patch adds this transform and some related logic to InstructionSimplify and removes some of the logic from instcombine (unfortunately not all because there are several situations in which instcombine can improve things by making new instructions, whereas instsimplify is not allowed to do this). At -O2 this often results in more than 15% more simplifications by early-cse, and results in hundreds of lines of bitcode being eliminated from the testsuite. I did see some small negative effects in the testsuite, for example a few additional instructions in three programs. One program, 483.xalancbmk, got an additional 35 instructions, which seems to be due to a function getting an additional instruction and then being inlined all over the place. llvm-svn: 123911
2011-01-20 14:21:55 +01:00
; RUN: opt < %s -instsimplify -S | FileCheck %s
target datalayout = "p:32:32"
define i1 @ptrtoint() {
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_]*\):\( *\)@$FUNC\([( ]*\)\$/;\1\2-LABEL:\3@$FUNC(/g" $TEMP done mv $TEMP $NAME fi done llvm-svn: 186268
2013-07-14 03:42:54 +02:00
; CHECK-LABEL: @ptrtoint(
At -O123 the early-cse pass is run before instcombine has run. According to my auto-simplier the transform most missed by early-cse is (zext X) != 0 -> X != 0. This patch adds this transform and some related logic to InstructionSimplify and removes some of the logic from instcombine (unfortunately not all because there are several situations in which instcombine can improve things by making new instructions, whereas instsimplify is not allowed to do this). At -O2 this often results in more than 15% more simplifications by early-cse, and results in hundreds of lines of bitcode being eliminated from the testsuite. I did see some small negative effects in the testsuite, for example a few additional instructions in three programs. One program, 483.xalancbmk, got an additional 35 instructions, which seems to be due to a function getting an additional instruction and then being inlined all over the place. llvm-svn: 123911
2011-01-20 14:21:55 +01:00
%a = alloca i8
%tmp = ptrtoint i8* %a to i32
%r = icmp eq i32 %tmp, 0
ret i1 %r
; CHECK: ret i1 false
}
InstSimplify: Ignore pointer casts when constant folding compares between pointers. llvm-svn: 150690
2012-02-16 14:49:39 +01:00
define i1 @bitcast() {
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_]*\):\( *\)@$FUNC\([( ]*\)\$/;\1\2-LABEL:\3@$FUNC(/g" $TEMP done mv $TEMP $NAME fi done llvm-svn: 186268
2013-07-14 03:42:54 +02:00
; CHECK-LABEL: @bitcast(
InstSimplify: Ignore pointer casts when constant folding compares between pointers. llvm-svn: 150690
2012-02-16 14:49:39 +01:00
%a = alloca i32
%b = alloca i64
%x = bitcast i32* %a to i8*
%y = bitcast i64* %b to i8*
%cmp = icmp eq i8* %x, %y
ret i1 %cmp
; CHECK-NEXT: ret i1 false
}
define i1 @gep() {
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_]*\):\( *\)@$FUNC\([( ]*\)\$/;\1\2-LABEL:\3@$FUNC(/g" $TEMP done mv $TEMP $NAME fi done llvm-svn: 186268
2013-07-14 03:42:54 +02:00
; CHECK-LABEL: @gep(
InstSimplify: Ignore pointer casts when constant folding compares between pointers. llvm-svn: 150690
2012-02-16 14:49:39 +01:00
%a = alloca [3 x i8], align 8
%x = getelementptr inbounds [3 x i8]* %a, i32 0, i32 0
%cmp = icmp eq i8* %x, null
ret i1 %cmp
; CHECK-NEXT: ret i1 false
}
Fix a rather nasty regression from r150690: LHS != RHS does not imply LHS->stripPointerCasts() != RHS->stripPointerCasts(). llvm-svn: 150863
2012-02-18 04:29:25 +01:00
define i1 @gep2() {
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_]*\):\( *\)@$FUNC\([( ]*\)\$/;\1\2-LABEL:\3@$FUNC(/g" $TEMP done mv $TEMP $NAME fi done llvm-svn: 186268
2013-07-14 03:42:54 +02:00
; CHECK-LABEL: @gep2(
Fix a rather nasty regression from r150690: LHS != RHS does not imply LHS->stripPointerCasts() != RHS->stripPointerCasts(). llvm-svn: 150863
2012-02-18 04:29:25 +01:00
%a = alloca [3 x i8], align 8
%x = getelementptr inbounds [3 x i8]* %a, i32 0, i32 0
%y = getelementptr inbounds [3 x i8]* %a, i32 0, i32 0
%cmp = icmp eq i8* %x, %y
ret i1 %cmp
; CHECK-NEXT: ret i1 true
}
Reinstate the optimization from r151449 with a fix to not turn 'gep %x' into 'gep null' when the icmp predicate is unsigned (or is signed without inbounds). llvm-svn: 151467
2012-02-26 03:09:49 +01:00
; PR11238
%gept = type { i32, i32 }
@gepy = global %gept zeroinitializer, align 8
@gepz = extern_weak global %gept
define i1 @gep3() {
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_]*\):\( *\)@$FUNC\([( ]*\)\$/;\1\2-LABEL:\3@$FUNC(/g" $TEMP done mv $TEMP $NAME fi done llvm-svn: 186268
2013-07-14 03:42:54 +02:00
; CHECK-LABEL: @gep3(
Reinstate the optimization from r151449 with a fix to not turn 'gep %x' into 'gep null' when the icmp predicate is unsigned (or is signed without inbounds). llvm-svn: 151467
2012-02-26 03:09:49 +01:00
%x = alloca %gept, align 8
%a = getelementptr %gept* %x, i64 0, i32 0
%b = getelementptr %gept* %x, i64 0, i32 1
%equal = icmp eq i32* %a, %b
ret i1 %equal
; CHECK-NEXT: ret i1 false
}
define i1 @gep4() {
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_]*\):\( *\)@$FUNC\([( ]*\)\$/;\1\2-LABEL:\3@$FUNC(/g" $TEMP done mv $TEMP $NAME fi done llvm-svn: 186268
2013-07-14 03:42:54 +02:00
; CHECK-LABEL: @gep4(
Reinstate the optimization from r151449 with a fix to not turn 'gep %x' into 'gep null' when the icmp predicate is unsigned (or is signed without inbounds). llvm-svn: 151467
2012-02-26 03:09:49 +01:00
%x = alloca %gept, align 8
%a = getelementptr %gept* @gepy, i64 0, i32 0
%b = getelementptr %gept* @gepy, i64 0, i32 1
%equal = icmp eq i32* %a, %b
ret i1 %equal
; CHECK-NEXT: ret i1 false
}
define i1 @gep5() {
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_]*\):\( *\)@$FUNC\([( ]*\)\$/;\1\2-LABEL:\3@$FUNC(/g" $TEMP done mv $TEMP $NAME fi done llvm-svn: 186268
2013-07-14 03:42:54 +02:00
; CHECK-LABEL: @gep5(
Reinstate the optimization from r151449 with a fix to not turn 'gep %x' into 'gep null' when the icmp predicate is unsigned (or is signed without inbounds). llvm-svn: 151467
2012-02-26 03:09:49 +01:00
%x = alloca %gept, align 8
%a = getelementptr inbounds %gept* %x, i64 0, i32 1
%b = getelementptr %gept* @gepy, i64 0, i32 0
%equal = icmp eq i32* %a, %b
ret i1 %equal
; CHECK-NEXT: ret i1 false
}
define i1 @gep6(%gept* %x) {
; Same as @gep3 but potentially null.
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_]*\):\( *\)@$FUNC\([( ]*\)\$/;\1\2-LABEL:\3@$FUNC(/g" $TEMP done mv $TEMP $NAME fi done llvm-svn: 186268
2013-07-14 03:42:54 +02:00
; CHECK-LABEL: @gep6(
Reinstate the optimization from r151449 with a fix to not turn 'gep %x' into 'gep null' when the icmp predicate is unsigned (or is signed without inbounds). llvm-svn: 151467
2012-02-26 03:09:49 +01:00
%a = getelementptr %gept* %x, i64 0, i32 0
%b = getelementptr %gept* %x, i64 0, i32 1
%equal = icmp eq i32* %a, %b
ret i1 %equal
; CHECK-NEXT: ret i1 false
}
define i1 @gep7(%gept* %x) {
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_]*\):\( *\)@$FUNC\([( ]*\)\$/;\1\2-LABEL:\3@$FUNC(/g" $TEMP done mv $TEMP $NAME fi done llvm-svn: 186268
2013-07-14 03:42:54 +02:00
; CHECK-LABEL: @gep7(
Reinstate the optimization from r151449 with a fix to not turn 'gep %x' into 'gep null' when the icmp predicate is unsigned (or is signed without inbounds). llvm-svn: 151467
2012-02-26 03:09:49 +01:00
%a = getelementptr %gept* %x, i64 0, i32 0
%b = getelementptr %gept* @gepz, i64 0, i32 0
%equal = icmp eq i32* %a, %b
ret i1 %equal
; CHECK: ret i1 %equal
}
define i1 @gep8(%gept* %x) {
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_]*\):\( *\)@$FUNC\([( ]*\)\$/;\1\2-LABEL:\3@$FUNC(/g" $TEMP done mv $TEMP $NAME fi done llvm-svn: 186268
2013-07-14 03:42:54 +02:00
; CHECK-LABEL: @gep8(
Reinstate the optimization from r151449 with a fix to not turn 'gep %x' into 'gep null' when the icmp predicate is unsigned (or is signed without inbounds). llvm-svn: 151467
2012-02-26 03:09:49 +01:00
%a = getelementptr %gept* %x, i32 1
%b = getelementptr %gept* %x, i32 -1
%equal = icmp ugt %gept* %a, %b
ret i1 %equal
; CHECK: ret i1 %equal
}
Teach instsimplify how to simplify comparisons of pointers which are constant-offsets of a common base using the generic GEP-walking logic I added for computing pointer differences in the same situation. llvm-svn: 153419
2012-03-25 23:28:14 +02:00
define i1 @gep9(i8* %ptr) {
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_]*\):\( *\)@$FUNC\([( ]*\)\$/;\1\2-LABEL:\3@$FUNC(/g" $TEMP done mv $TEMP $NAME fi done llvm-svn: 186268
2013-07-14 03:42:54 +02:00
; CHECK-LABEL: @gep9(
Teach instsimplify how to simplify comparisons of pointers which are constant-offsets of a common base using the generic GEP-walking logic I added for computing pointer differences in the same situation. llvm-svn: 153419
2012-03-25 23:28:14 +02:00
; CHECK-NOT: ret
; CHECK: ret i1 true
entry:
%first1 = getelementptr inbounds i8* %ptr, i32 0
%first2 = getelementptr inbounds i8* %first1, i32 1
%first3 = getelementptr inbounds i8* %first2, i32 2
%first4 = getelementptr inbounds i8* %first3, i32 4
%last1 = getelementptr inbounds i8* %first2, i32 48
%last2 = getelementptr inbounds i8* %last1, i32 8
%last3 = getelementptr inbounds i8* %last2, i32 -4
%last4 = getelementptr inbounds i8* %last3, i32 -4
%first.int = ptrtoint i8* %first4 to i32
%last.int = ptrtoint i8* %last4 to i32
%cmp = icmp ne i32 %last.int, %first.int
ret i1 %cmp
}
define i1 @gep10(i8* %ptr) {
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_]*\):\( *\)@$FUNC\([( ]*\)\$/;\1\2-LABEL:\3@$FUNC(/g" $TEMP done mv $TEMP $NAME fi done llvm-svn: 186268
2013-07-14 03:42:54 +02:00
; CHECK-LABEL: @gep10(
Teach instsimplify how to simplify comparisons of pointers which are constant-offsets of a common base using the generic GEP-walking logic I added for computing pointer differences in the same situation. llvm-svn: 153419
2012-03-25 23:28:14 +02:00
; CHECK-NOT: ret
; CHECK: ret i1 true
entry:
%first1 = getelementptr inbounds i8* %ptr, i32 -2
%first2 = getelementptr inbounds i8* %first1, i32 44
%last1 = getelementptr inbounds i8* %ptr, i32 48
%last2 = getelementptr inbounds i8* %last1, i32 -6
%first.int = ptrtoint i8* %first2 to i32
%last.int = ptrtoint i8* %last2 to i32
%cmp = icmp eq i32 %last.int, %first.int
ret i1 %cmp
}
define i1 @gep11(i8* %ptr) {
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_]*\):\( *\)@$FUNC\([( ]*\)\$/;\1\2-LABEL:\3@$FUNC(/g" $TEMP done mv $TEMP $NAME fi done llvm-svn: 186268
2013-07-14 03:42:54 +02:00
; CHECK-LABEL: @gep11(
Teach instsimplify how to simplify comparisons of pointers which are constant-offsets of a common base using the generic GEP-walking logic I added for computing pointer differences in the same situation. llvm-svn: 153419
2012-03-25 23:28:14 +02:00
; CHECK-NOT: ret
; CHECK: ret i1 true
entry:
%first1 = getelementptr inbounds i8* %ptr, i32 -2
%last1 = getelementptr inbounds i8* %ptr, i32 48
%last2 = getelementptr inbounds i8* %last1, i32 -6
%cmp = icmp ult i8* %first1, %last2
ret i1 %cmp
}
define i1 @gep12(i8* %ptr) {
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_]*\):\( *\)@$FUNC\([( ]*\)\$/;\1\2-LABEL:\3@$FUNC(/g" $TEMP done mv $TEMP $NAME fi done llvm-svn: 186268
2013-07-14 03:42:54 +02:00
; CHECK-LABEL: @gep12(
Teach instsimplify how to simplify comparisons of pointers which are constant-offsets of a common base using the generic GEP-walking logic I added for computing pointer differences in the same situation. llvm-svn: 153419
2012-03-25 23:28:14 +02:00
; CHECK-NOT: ret
; CHECK: ret i1 %cmp
entry:
%first1 = getelementptr inbounds i8* %ptr, i32 -2
%last1 = getelementptr inbounds i8* %ptr, i32 48
%last2 = getelementptr inbounds i8* %last1, i32 -6
%cmp = icmp slt i8* %first1, %last2
ret i1 %cmp
}
Add support to ValueTracking for determining that a pointer is non-null by virtue of inbounds GEPs that preclude a null pointer. This is a very common pattern in the code generated by std::vector and other standard library routines which use allocators that test for null pervasively. This is one step closer to teaching Clang+LLVM to be able to produce an empty function for: void f() { std::vector<int> v; v.push_back(1); v.push_back(2); v.push_back(3); v.push_back(4); } Which is related to getting them to completely fold SmallVector push_back sequences into constants when inlining and other optimizations make that a possibility. llvm-svn: 169573
2012-12-07 03:08:58 +01:00
define i1 @gep13(i8* %ptr) {
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_]*\):\( *\)@$FUNC\([( ]*\)\$/;\1\2-LABEL:\3@$FUNC(/g" $TEMP done mv $TEMP $NAME fi done llvm-svn: 186268
2013-07-14 03:42:54 +02:00
; CHECK-LABEL: @gep13(
Add support to ValueTracking for determining that a pointer is non-null by virtue of inbounds GEPs that preclude a null pointer. This is a very common pattern in the code generated by std::vector and other standard library routines which use allocators that test for null pervasively. This is one step closer to teaching Clang+LLVM to be able to produce an empty function for: void f() { std::vector<int> v; v.push_back(1); v.push_back(2); v.push_back(3); v.push_back(4); } Which is related to getting them to completely fold SmallVector push_back sequences into constants when inlining and other optimizations make that a possibility. llvm-svn: 169573
2012-12-07 03:08:58 +01:00
; We can prove this GEP is non-null because it is inbounds.
%x = getelementptr inbounds i8* %ptr, i32 1
%cmp = icmp eq i8* %x, null
ret i1 %cmp
; CHECK-NEXT: ret i1 false
}
define i1 @gep14({ {}, i8 }* %ptr) {
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_]*\):\( *\)@$FUNC\([( ]*\)\$/;\1\2-LABEL:\3@$FUNC(/g" $TEMP done mv $TEMP $NAME fi done llvm-svn: 186268
2013-07-14 03:42:54 +02:00
; CHECK-LABEL: @gep14(
Add support to ValueTracking for determining that a pointer is non-null by virtue of inbounds GEPs that preclude a null pointer. This is a very common pattern in the code generated by std::vector and other standard library routines which use allocators that test for null pervasively. This is one step closer to teaching Clang+LLVM to be able to produce an empty function for: void f() { std::vector<int> v; v.push_back(1); v.push_back(2); v.push_back(3); v.push_back(4); } Which is related to getting them to completely fold SmallVector push_back sequences into constants when inlining and other optimizations make that a possibility. llvm-svn: 169573
2012-12-07 03:08:58 +01:00
; We can't simplify this because the offset of one in the GEP actually doesn't
; move the pointer.
%x = getelementptr inbounds { {}, i8 }* %ptr, i32 0, i32 1
%cmp = icmp eq i8* %x, null
ret i1 %cmp
; CHECK-NOT: ret i1 false
}
define i1 @gep15({ {}, [4 x {i8, i8}]}* %ptr, i32 %y) {
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_]*\):\( *\)@$FUNC\([( ]*\)\$/;\1\2-LABEL:\3@$FUNC(/g" $TEMP done mv $TEMP $NAME fi done llvm-svn: 186268
2013-07-14 03:42:54 +02:00
; CHECK-LABEL: @gep15(
Add support to ValueTracking for determining that a pointer is non-null by virtue of inbounds GEPs that preclude a null pointer. This is a very common pattern in the code generated by std::vector and other standard library routines which use allocators that test for null pervasively. This is one step closer to teaching Clang+LLVM to be able to produce an empty function for: void f() { std::vector<int> v; v.push_back(1); v.push_back(2); v.push_back(3); v.push_back(4); } Which is related to getting them to completely fold SmallVector push_back sequences into constants when inlining and other optimizations make that a possibility. llvm-svn: 169573
2012-12-07 03:08:58 +01:00
; We can prove this GEP is non-null even though there is a user value, as we
; would necessarily violate inbounds on one side or the other.
%x = getelementptr inbounds { {}, [4 x {i8, i8}]}* %ptr, i32 0, i32 1, i32 %y, i32 1
%cmp = icmp eq i8* %x, null
ret i1 %cmp
; CHECK-NEXT: ret i1 false
}
define i1 @gep16(i8* %ptr, i32 %a) {
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_]*\):\( *\)@$FUNC\([( ]*\)\$/;\1\2-LABEL:\3@$FUNC(/g" $TEMP done mv $TEMP $NAME fi done llvm-svn: 186268
2013-07-14 03:42:54 +02:00
; CHECK-LABEL: @gep16(
Add support to ValueTracking for determining that a pointer is non-null by virtue of inbounds GEPs that preclude a null pointer. This is a very common pattern in the code generated by std::vector and other standard library routines which use allocators that test for null pervasively. This is one step closer to teaching Clang+LLVM to be able to produce an empty function for: void f() { std::vector<int> v; v.push_back(1); v.push_back(2); v.push_back(3); v.push_back(4); } Which is related to getting them to completely fold SmallVector push_back sequences into constants when inlining and other optimizations make that a possibility. llvm-svn: 169573
2012-12-07 03:08:58 +01:00
; We can prove this GEP is non-null because it is inbounds and because we know
; %b is non-zero even though we don't know its value.
%b = or i32 %a, 1
%x = getelementptr inbounds i8* %ptr, i32 %b
%cmp = icmp eq i8* %x, null
ret i1 %cmp
; CHECK-NEXT: ret i1 false
}
Teach instsimplify how to simplify comparisons of pointers which are constant-offsets of a common base using the generic GEP-walking logic I added for computing pointer differences in the same situation. llvm-svn: 153419
2012-03-25 23:28:14 +02:00
At -O123 the early-cse pass is run before instcombine has run. According to my auto-simplier the transform most missed by early-cse is (zext X) != 0 -> X != 0. This patch adds this transform and some related logic to InstructionSimplify and removes some of the logic from instcombine (unfortunately not all because there are several situations in which instcombine can improve things by making new instructions, whereas instsimplify is not allowed to do this). At -O2 this often results in more than 15% more simplifications by early-cse, and results in hundreds of lines of bitcode being eliminated from the testsuite. I did see some small negative effects in the testsuite, for example a few additional instructions in three programs. One program, 483.xalancbmk, got an additional 35 instructions, which seems to be due to a function getting an additional instruction and then being inlined all over the place. llvm-svn: 123911
2011-01-20 14:21:55 +01:00
define i1 @zext(i32 %x) {
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_]*\):\( *\)@$FUNC\([( ]*\)\$/;\1\2-LABEL:\3@$FUNC(/g" $TEMP done mv $TEMP $NAME fi done llvm-svn: 186268
2013-07-14 03:42:54 +02:00
; CHECK-LABEL: @zext(
At -O123 the early-cse pass is run before instcombine has run. According to my auto-simplier the transform most missed by early-cse is (zext X) != 0 -> X != 0. This patch adds this transform and some related logic to InstructionSimplify and removes some of the logic from instcombine (unfortunately not all because there are several situations in which instcombine can improve things by making new instructions, whereas instsimplify is not allowed to do this). At -O2 this often results in more than 15% more simplifications by early-cse, and results in hundreds of lines of bitcode being eliminated from the testsuite. I did see some small negative effects in the testsuite, for example a few additional instructions in three programs. One program, 483.xalancbmk, got an additional 35 instructions, which seems to be due to a function getting an additional instruction and then being inlined all over the place. llvm-svn: 123911
2011-01-20 14:21:55 +01:00
%e1 = zext i32 %x to i64
%e2 = zext i32 %x to i64
%r = icmp eq i64 %e1, %e2
ret i1 %r
; CHECK: ret i1 true
}
define i1 @zext2(i1 %x) {
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_]*\):\( *\)@$FUNC\([( ]*\)\$/;\1\2-LABEL:\3@$FUNC(/g" $TEMP done mv $TEMP $NAME fi done llvm-svn: 186268
2013-07-14 03:42:54 +02:00
; CHECK-LABEL: @zext2(
At -O123 the early-cse pass is run before instcombine has run. According to my auto-simplier the transform most missed by early-cse is (zext X) != 0 -> X != 0. This patch adds this transform and some related logic to InstructionSimplify and removes some of the logic from instcombine (unfortunately not all because there are several situations in which instcombine can improve things by making new instructions, whereas instsimplify is not allowed to do this). At -O2 this often results in more than 15% more simplifications by early-cse, and results in hundreds of lines of bitcode being eliminated from the testsuite. I did see some small negative effects in the testsuite, for example a few additional instructions in three programs. One program, 483.xalancbmk, got an additional 35 instructions, which seems to be due to a function getting an additional instruction and then being inlined all over the place. llvm-svn: 123911
2011-01-20 14:21:55 +01:00
%e = zext i1 %x to i32
%c = icmp ne i32 %e, 0
ret i1 %c
; CHECK: ret i1 %x
}
According to my auto-simplifier the most common missed simplifications in optimized code are: (non-negative number)+(power-of-two) != 0 -> true and (x | 1) != 0 -> true Instcombine knows about the second one of course, but only does it if X|1 has only one use. These fire thousands of times in the testsuite. llvm-svn: 124183
2011-01-25 10:38:29 +01:00
define i1 @zext3() {
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_]*\):\( *\)@$FUNC\([( ]*\)\$/;\1\2-LABEL:\3@$FUNC(/g" $TEMP done mv $TEMP $NAME fi done llvm-svn: 186268
2013-07-14 03:42:54 +02:00
; CHECK-LABEL: @zext3(
According to my auto-simplifier the most common missed simplifications in optimized code are: (non-negative number)+(power-of-two) != 0 -> true and (x | 1) != 0 -> true Instcombine knows about the second one of course, but only does it if X|1 has only one use. These fire thousands of times in the testsuite. llvm-svn: 124183
2011-01-25 10:38:29 +01:00
%e = zext i1 1 to i32
%c = icmp ne i32 %e, 0
ret i1 %c
; CHECK: ret i1 true
}
At -O123 the early-cse pass is run before instcombine has run. According to my auto-simplier the transform most missed by early-cse is (zext X) != 0 -> X != 0. This patch adds this transform and some related logic to InstructionSimplify and removes some of the logic from instcombine (unfortunately not all because there are several situations in which instcombine can improve things by making new instructions, whereas instsimplify is not allowed to do this). At -O2 this often results in more than 15% more simplifications by early-cse, and results in hundreds of lines of bitcode being eliminated from the testsuite. I did see some small negative effects in the testsuite, for example a few additional instructions in three programs. One program, 483.xalancbmk, got an additional 35 instructions, which seems to be due to a function getting an additional instruction and then being inlined all over the place. llvm-svn: 123911
2011-01-20 14:21:55 +01:00
define i1 @sext(i32 %x) {
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_]*\):\( *\)@$FUNC\([( ]*\)\$/;\1\2-LABEL:\3@$FUNC(/g" $TEMP done mv $TEMP $NAME fi done llvm-svn: 186268
2013-07-14 03:42:54 +02:00
; CHECK-LABEL: @sext(
At -O123 the early-cse pass is run before instcombine has run. According to my auto-simplier the transform most missed by early-cse is (zext X) != 0 -> X != 0. This patch adds this transform and some related logic to InstructionSimplify and removes some of the logic from instcombine (unfortunately not all because there are several situations in which instcombine can improve things by making new instructions, whereas instsimplify is not allowed to do this). At -O2 this often results in more than 15% more simplifications by early-cse, and results in hundreds of lines of bitcode being eliminated from the testsuite. I did see some small negative effects in the testsuite, for example a few additional instructions in three programs. One program, 483.xalancbmk, got an additional 35 instructions, which seems to be due to a function getting an additional instruction and then being inlined all over the place. llvm-svn: 123911
2011-01-20 14:21:55 +01:00
%e1 = sext i32 %x to i64
%e2 = sext i32 %x to i64
%r = icmp eq i64 %e1, %e2
ret i1 %r
; CHECK: ret i1 true
}
define i1 @sext2(i1 %x) {
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_]*\):\( *\)@$FUNC\([( ]*\)\$/;\1\2-LABEL:\3@$FUNC(/g" $TEMP done mv $TEMP $NAME fi done llvm-svn: 186268
2013-07-14 03:42:54 +02:00
; CHECK-LABEL: @sext2(
At -O123 the early-cse pass is run before instcombine has run. According to my auto-simplier the transform most missed by early-cse is (zext X) != 0 -> X != 0. This patch adds this transform and some related logic to InstructionSimplify and removes some of the logic from instcombine (unfortunately not all because there are several situations in which instcombine can improve things by making new instructions, whereas instsimplify is not allowed to do this). At -O2 this often results in more than 15% more simplifications by early-cse, and results in hundreds of lines of bitcode being eliminated from the testsuite. I did see some small negative effects in the testsuite, for example a few additional instructions in three programs. One program, 483.xalancbmk, got an additional 35 instructions, which seems to be due to a function getting an additional instruction and then being inlined all over the place. llvm-svn: 123911
2011-01-20 14:21:55 +01:00
%e = sext i1 %x to i32
%c = icmp ne i32 %e, 0
ret i1 %c
; CHECK: ret i1 %x
}
According to my auto-simplifier the most common missed simplifications in optimized code are: (non-negative number)+(power-of-two) != 0 -> true and (x | 1) != 0 -> true Instcombine knows about the second one of course, but only does it if X|1 has only one use. These fire thousands of times in the testsuite. llvm-svn: 124183
2011-01-25 10:38:29 +01:00
define i1 @sext3() {
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_]*\):\( *\)@$FUNC\([( ]*\)\$/;\1\2-LABEL:\3@$FUNC(/g" $TEMP done mv $TEMP $NAME fi done llvm-svn: 186268
2013-07-14 03:42:54 +02:00
; CHECK-LABEL: @sext3(
According to my auto-simplifier the most common missed simplifications in optimized code are: (non-negative number)+(power-of-two) != 0 -> true and (x | 1) != 0 -> true Instcombine knows about the second one of course, but only does it if X|1 has only one use. These fire thousands of times in the testsuite. llvm-svn: 124183
2011-01-25 10:38:29 +01:00
%e = sext i1 1 to i32
%c = icmp ne i32 %e, 0
ret i1 %c
; CHECK: ret i1 true
}
define i1 @add(i32 %x, i32 %y) {
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_]*\):\( *\)@$FUNC\([( ]*\)\$/;\1\2-LABEL:\3@$FUNC(/g" $TEMP done mv $TEMP $NAME fi done llvm-svn: 186268
2013-07-14 03:42:54 +02:00
; CHECK-LABEL: @add(
According to my auto-simplifier the most common missed simplifications in optimized code are: (non-negative number)+(power-of-two) != 0 -> true and (x | 1) != 0 -> true Instcombine knows about the second one of course, but only does it if X|1 has only one use. These fire thousands of times in the testsuite. llvm-svn: 124183
2011-01-25 10:38:29 +01:00
%l = lshr i32 %x, 1
In which I discover that zero+zero is zero, d'oh! llvm-svn: 124188
2011-01-25 16:14:15 +01:00
%q = lshr i32 %y, 1
%r = or i32 %q, 1
According to my auto-simplifier the most common missed simplifications in optimized code are: (non-negative number)+(power-of-two) != 0 -> true and (x | 1) != 0 -> true Instcombine knows about the second one of course, but only does it if X|1 has only one use. These fire thousands of times in the testsuite. llvm-svn: 124183
2011-01-25 10:38:29 +01:00
%s = add i32 %l, %r
%c = icmp eq i32 %s, 0
ret i1 %c
In which I discover that zero+zero is zero, d'oh! llvm-svn: 124188
2011-01-25 16:14:15 +01:00
; CHECK: ret i1 false
According to my auto-simplifier the most common missed simplifications in optimized code are: (non-negative number)+(power-of-two) != 0 -> true and (x | 1) != 0 -> true Instcombine knows about the second one of course, but only does it if X|1 has only one use. These fire thousands of times in the testsuite. llvm-svn: 124183
2011-01-25 10:38:29 +01:00
}
define i1 @add2(i8 %x, i8 %y) {
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_]*\):\( *\)@$FUNC\([( ]*\)\$/;\1\2-LABEL:\3@$FUNC(/g" $TEMP done mv $TEMP $NAME fi done llvm-svn: 186268
2013-07-14 03:42:54 +02:00
; CHECK-LABEL: @add2(
According to my auto-simplifier the most common missed simplifications in optimized code are: (non-negative number)+(power-of-two) != 0 -> true and (x | 1) != 0 -> true Instcombine knows about the second one of course, but only does it if X|1 has only one use. These fire thousands of times in the testsuite. llvm-svn: 124183
2011-01-25 10:38:29 +01:00
%l = or i8 %x, 128
%r = or i8 %y, 129
%s = add i8 %l, %r
%c = icmp eq i8 %s, 0
ret i1 %c
; CHECK: ret i1 false
}
In which I discover that zero+zero is zero, d'oh! llvm-svn: 124188
2011-01-25 16:14:15 +01:00
define i1 @add3(i8 %x, i8 %y) {
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_]*\):\( *\)@$FUNC\([( ]*\)\$/;\1\2-LABEL:\3@$FUNC(/g" $TEMP done mv $TEMP $NAME fi done llvm-svn: 186268
2013-07-14 03:42:54 +02:00
; CHECK-LABEL: @add3(
In which I discover that zero+zero is zero, d'oh! llvm-svn: 124188
2011-01-25 16:14:15 +01:00
%l = zext i8 %x to i32
%r = zext i8 %y to i32
%s = add i32 %l, %r
%c = icmp eq i32 %s, 0
ret i1 %c
; CHECK: ret i1 %c
}
Teach instsimplify that X+Y>=X+Z is the same as Y>=Z if neither side overflows, plus some variations of this. According to my auto-simplifier this occurs a lot but usually in combination with max/min idioms. Because max/min aren't handled yet this unfortunately doesn't have much effect in the testsuite. llvm-svn: 125462
2011-02-13 18:15:40 +01:00
define i1 @add4(i32 %x, i32 %y) {
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_]*\):\( *\)@$FUNC\([( ]*\)\$/;\1\2-LABEL:\3@$FUNC(/g" $TEMP done mv $TEMP $NAME fi done llvm-svn: 186268
2013-07-14 03:42:54 +02:00
; CHECK-LABEL: @add4(
Teach instsimplify that X+Y>=X+Z is the same as Y>=Z if neither side overflows, plus some variations of this. According to my auto-simplifier this occurs a lot but usually in combination with max/min idioms. Because max/min aren't handled yet this unfortunately doesn't have much effect in the testsuite. llvm-svn: 125462
2011-02-13 18:15:40 +01:00
%z = add nsw i32 %y, 1
%s1 = add nsw i32 %x, %y
%s2 = add nsw i32 %x, %z
%c = icmp slt i32 %s1, %s2
ret i1 %c
; CHECK: ret i1 true
}
define i1 @add5(i32 %x, i32 %y) {
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_]*\):\( *\)@$FUNC\([( ]*\)\$/;\1\2-LABEL:\3@$FUNC(/g" $TEMP done mv $TEMP $NAME fi done llvm-svn: 186268
2013-07-14 03:42:54 +02:00
; CHECK-LABEL: @add5(
Teach instsimplify that X+Y>=X+Z is the same as Y>=Z if neither side overflows, plus some variations of this. According to my auto-simplifier this occurs a lot but usually in combination with max/min idioms. Because max/min aren't handled yet this unfortunately doesn't have much effect in the testsuite. llvm-svn: 125462
2011-02-13 18:15:40 +01:00
%z = add nuw i32 %y, 1
%s1 = add nuw i32 %x, %z
%s2 = add nuw i32 %x, %y
%c = icmp ugt i32 %s1, %s2
ret i1 %c
; CHECK: ret i1 true
}
InstructionSimplify should be able to simplify A+B==B+A to 'true' but wasn't due to the same logic bug that caused PR14361. llvm-svn: 168186
2012-11-16 20:41:26 +01:00
define i1 @add6(i64 %A, i64 %B) {
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_]*\):\( *\)@$FUNC\([( ]*\)\$/;\1\2-LABEL:\3@$FUNC(/g" $TEMP done mv $TEMP $NAME fi done llvm-svn: 186268
2013-07-14 03:42:54 +02:00
; CHECK-LABEL: @add6(
InstructionSimplify should be able to simplify A+B==B+A to 'true' but wasn't due to the same logic bug that caused PR14361. llvm-svn: 168186
2012-11-16 20:41:26 +01:00
%s1 = add i64 %A, %B
%s2 = add i64 %B, %A
%cmp = icmp eq i64 %s1, %s2
ret i1 %cmp
; CHECK: ret i1 true
}
According to my auto-simplifier the most common missed simplifications in optimized code are: (non-negative number)+(power-of-two) != 0 -> true and (x | 1) != 0 -> true Instcombine knows about the second one of course, but only does it if X|1 has only one use. These fire thousands of times in the testsuite. llvm-svn: 124183
2011-01-25 10:38:29 +01:00
define i1 @addpowtwo(i32 %x, i32 %y) {
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_]*\):\( *\)@$FUNC\([( ]*\)\$/;\1\2-LABEL:\3@$FUNC(/g" $TEMP done mv $TEMP $NAME fi done llvm-svn: 186268
2013-07-14 03:42:54 +02:00
; CHECK-LABEL: @addpowtwo(
According to my auto-simplifier the most common missed simplifications in optimized code are: (non-negative number)+(power-of-two) != 0 -> true and (x | 1) != 0 -> true Instcombine knows about the second one of course, but only does it if X|1 has only one use. These fire thousands of times in the testsuite. llvm-svn: 124183
2011-01-25 10:38:29 +01:00
%l = lshr i32 %x, 1
%r = shl i32 1, %y
%s = add i32 %l, %r
%c = icmp eq i32 %s, 0
ret i1 %c
; CHECK: ret i1 false
}
define i1 @or(i32 %x) {
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_]*\):\( *\)@$FUNC\([( ]*\)\$/;\1\2-LABEL:\3@$FUNC(/g" $TEMP done mv $TEMP $NAME fi done llvm-svn: 186268
2013-07-14 03:42:54 +02:00
; CHECK-LABEL: @or(
According to my auto-simplifier the most common missed simplifications in optimized code are: (non-negative number)+(power-of-two) != 0 -> true and (x | 1) != 0 -> true Instcombine knows about the second one of course, but only does it if X|1 has only one use. These fire thousands of times in the testsuite. llvm-svn: 124183
2011-01-25 10:38:29 +01:00
%o = or i32 %x, 1
%c = icmp eq i32 %o, 0
ret i1 %c
; CHECK: ret i1 false
}
Fix typo: should have been testing that X was odd, not V. llvm-svn: 124533
2011-01-29 14:27:00 +01:00
define i1 @shl(i32 %x) {
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_]*\):\( *\)@$FUNC\([( ]*\)\$/;\1\2-LABEL:\3@$FUNC(/g" $TEMP done mv $TEMP $NAME fi done llvm-svn: 186268
2013-07-14 03:42:54 +02:00
; CHECK-LABEL: @shl(
Fix typo: should have been testing that X was odd, not V. llvm-svn: 124533
2011-01-29 14:27:00 +01:00
%s = shl i32 1, %x
%c = icmp eq i32 %s, 0
ret i1 %c
; CHECK: ret i1 false
}
Teach instruction simplify to use constant ranges to solve problems of the form "icmp pred %X, CI" and a number of examples where "%X = binop %Y, CI2". Some of these cases (div and rem) used to make it through opt -O2, but the others are probably now making code elsewhere redundant (probably instcombine). llvm-svn: 126988
2011-03-04 08:00:57 +01:00
define i1 @lshr1(i32 %x) {
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_]*\):\( *\)@$FUNC\([( ]*\)\$/;\1\2-LABEL:\3@$FUNC(/g" $TEMP done mv $TEMP $NAME fi done llvm-svn: 186268
2013-07-14 03:42:54 +02:00
; CHECK-LABEL: @lshr1(
Fix typo: should have been testing that X was odd, not V. llvm-svn: 124533
2011-01-29 14:27:00 +01:00
%s = lshr i32 -1, %x
%c = icmp eq i32 %s, 0
ret i1 %c
; CHECK: ret i1 false
}
Teach instruction simplify to use constant ranges to solve problems of the form "icmp pred %X, CI" and a number of examples where "%X = binop %Y, CI2". Some of these cases (div and rem) used to make it through opt -O2, but the others are probably now making code elsewhere redundant (probably instcombine). llvm-svn: 126988
2011-03-04 08:00:57 +01:00
define i1 @lshr2(i32 %x) {
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_]*\):\( *\)@$FUNC\([( ]*\)\$/;\1\2-LABEL:\3@$FUNC(/g" $TEMP done mv $TEMP $NAME fi done llvm-svn: 186268
2013-07-14 03:42:54 +02:00
; CHECK-LABEL: @lshr2(
Teach instruction simplify to use constant ranges to solve problems of the form "icmp pred %X, CI" and a number of examples where "%X = binop %Y, CI2". Some of these cases (div and rem) used to make it through opt -O2, but the others are probably now making code elsewhere redundant (probably instcombine). llvm-svn: 126988
2011-03-04 08:00:57 +01:00
%s = lshr i32 %x, 30
%c = icmp ugt i32 %s, 8
ret i1 %c
; CHECK: ret i1 false
}
InstSimplify: X >> X -> 0 llvm-svn: 185973
2013-07-10 00:01:22 +02:00
define i1 @lshr3(i32 %x) {
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_]*\):\( *\)@$FUNC\([( ]*\)\$/;\1\2-LABEL:\3@$FUNC(/g" $TEMP done mv $TEMP $NAME fi done llvm-svn: 186268
2013-07-14 03:42:54 +02:00
; CHECK-LABEL: @lshr3(
InstSimplify: X >> X -> 0 llvm-svn: 185973
2013-07-10 00:01:22 +02:00
%s = lshr i32 %x, %x
%c = icmp eq i32 %s, 0
ret i1 %c
; CHECK: ret i1 true
}
Teach instruction simplify to use constant ranges to solve problems of the form "icmp pred %X, CI" and a number of examples where "%X = binop %Y, CI2". Some of these cases (div and rem) used to make it through opt -O2, but the others are probably now making code elsewhere redundant (probably instcombine). llvm-svn: 126988
2011-03-04 08:00:57 +01:00
define i1 @ashr1(i32 %x) {
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_]*\):\( *\)@$FUNC\([( ]*\)\$/;\1\2-LABEL:\3@$FUNC(/g" $TEMP done mv $TEMP $NAME fi done llvm-svn: 186268
2013-07-14 03:42:54 +02:00
; CHECK-LABEL: @ashr1(
Fix typo: should have been testing that X was odd, not V. llvm-svn: 124533
2011-01-29 14:27:00 +01:00
%s = ashr i32 -1, %x
%c = icmp eq i32 %s, 0
ret i1 %c
; CHECK: ret i1 false
}
Improve threading of comparisons over select instructions (spotted by my auto-simplifier). This has a big impact on Ada code, but not much else. Unfortunately the impact is mostly negative! This is due to PR9004 (aka SCCP failing to resolve conditional branch conditions in the destination blocks of the branch), in which simple correlated expressions are not resolved but complicated ones are, so simplifying has a bad effect! llvm-svn: 124788
2011-02-03 10:37:39 +01:00
Teach instruction simplify to use constant ranges to solve problems of the form "icmp pred %X, CI" and a number of examples where "%X = binop %Y, CI2". Some of these cases (div and rem) used to make it through opt -O2, but the others are probably now making code elsewhere redundant (probably instcombine). llvm-svn: 126988
2011-03-04 08:00:57 +01:00
define i1 @ashr2(i32 %x) {
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_]*\):\( *\)@$FUNC\([( ]*\)\$/;\1\2-LABEL:\3@$FUNC(/g" $TEMP done mv $TEMP $NAME fi done llvm-svn: 186268
2013-07-14 03:42:54 +02:00
; CHECK-LABEL: @ashr2(
Teach instruction simplify to use constant ranges to solve problems of the form "icmp pred %X, CI" and a number of examples where "%X = binop %Y, CI2". Some of these cases (div and rem) used to make it through opt -O2, but the others are probably now making code elsewhere redundant (probably instcombine). llvm-svn: 126988
2011-03-04 08:00:57 +01:00
%s = ashr i32 %x, 30
%c = icmp slt i32 %s, -5
ret i1 %c
; CHECK: ret i1 false
}
InstSimplify: X >> X -> 0 llvm-svn: 185973
2013-07-10 00:01:22 +02:00
define i1 @ashr3(i32 %x) {
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_]*\):\( *\)@$FUNC\([( ]*\)\$/;\1\2-LABEL:\3@$FUNC(/g" $TEMP done mv $TEMP $NAME fi done llvm-svn: 186268
2013-07-14 03:42:54 +02:00
; CHECK-LABEL: @ashr3(
InstSimplify: X >> X -> 0 llvm-svn: 185973
2013-07-10 00:01:22 +02:00
%s = ashr i32 %x, %x
%c = icmp eq i32 %s, 0
ret i1 %c
; CHECK: ret i1 true
}
Improve threading of comparisons over select instructions (spotted by my auto-simplifier). This has a big impact on Ada code, but not much else. Unfortunately the impact is mostly negative! This is due to PR9004 (aka SCCP failing to resolve conditional branch conditions in the destination blocks of the branch), in which simple correlated expressions are not resolved but complicated ones are, so simplifying has a bad effect! llvm-svn: 124788
2011-02-03 10:37:39 +01:00
define i1 @select1(i1 %cond) {
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_]*\):\( *\)@$FUNC\([( ]*\)\$/;\1\2-LABEL:\3@$FUNC(/g" $TEMP done mv $TEMP $NAME fi done llvm-svn: 186268
2013-07-14 03:42:54 +02:00
; CHECK-LABEL: @select1(
Improve threading of comparisons over select instructions (spotted by my auto-simplifier). This has a big impact on Ada code, but not much else. Unfortunately the impact is mostly negative! This is due to PR9004 (aka SCCP failing to resolve conditional branch conditions in the destination blocks of the branch), in which simple correlated expressions are not resolved but complicated ones are, so simplifying has a bad effect! llvm-svn: 124788
2011-02-03 10:37:39 +01:00
%s = select i1 %cond, i32 1, i32 0
%c = icmp eq i32 %s, 1
ret i1 %c
; CHECK: ret i1 %cond
}
define i1 @select2(i1 %cond) {
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_]*\):\( *\)@$FUNC\([( ]*\)\$/;\1\2-LABEL:\3@$FUNC(/g" $TEMP done mv $TEMP $NAME fi done llvm-svn: 186268
2013-07-14 03:42:54 +02:00
; CHECK-LABEL: @select2(
Improve threading of comparisons over select instructions (spotted by my auto-simplifier). This has a big impact on Ada code, but not much else. Unfortunately the impact is mostly negative! This is due to PR9004 (aka SCCP failing to resolve conditional branch conditions in the destination blocks of the branch), in which simple correlated expressions are not resolved but complicated ones are, so simplifying has a bad effect! llvm-svn: 124788
2011-02-03 10:37:39 +01:00
%x = zext i1 %cond to i32
%s = select i1 %cond, i32 %x, i32 0
%c = icmp ne i32 %s, 0
ret i1 %c
; CHECK: ret i1 %cond
}
define i1 @select3(i1 %cond) {
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_]*\):\( *\)@$FUNC\([( ]*\)\$/;\1\2-LABEL:\3@$FUNC(/g" $TEMP done mv $TEMP $NAME fi done llvm-svn: 186268
2013-07-14 03:42:54 +02:00
; CHECK-LABEL: @select3(
Improve threading of comparisons over select instructions (spotted by my auto-simplifier). This has a big impact on Ada code, but not much else. Unfortunately the impact is mostly negative! This is due to PR9004 (aka SCCP failing to resolve conditional branch conditions in the destination blocks of the branch), in which simple correlated expressions are not resolved but complicated ones are, so simplifying has a bad effect! llvm-svn: 124788
2011-02-03 10:37:39 +01:00
%x = zext i1 %cond to i32
%s = select i1 %cond, i32 1, i32 %x
%c = icmp ne i32 %s, 0
ret i1 %c
; CHECK: ret i1 %cond
}
define i1 @select4(i1 %cond) {
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_]*\):\( *\)@$FUNC\([( ]*\)\$/;\1\2-LABEL:\3@$FUNC(/g" $TEMP done mv $TEMP $NAME fi done llvm-svn: 186268
2013-07-14 03:42:54 +02:00
; CHECK-LABEL: @select4(
Improve threading of comparisons over select instructions (spotted by my auto-simplifier). This has a big impact on Ada code, but not much else. Unfortunately the impact is mostly negative! This is due to PR9004 (aka SCCP failing to resolve conditional branch conditions in the destination blocks of the branch), in which simple correlated expressions are not resolved but complicated ones are, so simplifying has a bad effect! llvm-svn: 124788
2011-02-03 10:37:39 +01:00
%invert = xor i1 %cond, 1
%s = select i1 %invert, i32 0, i32 1
%c = icmp ne i32 %s, 0
ret i1 %c
; CHECK: ret i1 %cond
}
Optimize "icmp pred (urem X, Y), Y" --> true/false depending on pred. There's more work to do here, "icmp ult (urem X, 10), 11" doesn't optimize away yet. Fixes example 3 from PR9343! llvm-svn: 126741
2011-03-01 09:15:50 +01:00
Reapply commit 143214 with a fix: m_ICmp doesn't match conditions with the given predicate, it matches any condition and returns the predicate - d'oh! Original commit message: The expression icmp eq (select (icmp eq x, 0), 1, x), 0 folds to false. Spotted by my super-optimizer in 186.crafty and 450.soplex. We really need a proper infrastructure for handling generalizations of this kind of thing (which occur a lot), however this case is so simple that I decided to go ahead and implement it directly. llvm-svn: 143318
2011-10-30 20:56:36 +01:00
define i1 @select5(i32 %x) {
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_]*\):\( *\)@$FUNC\([( ]*\)\$/;\1\2-LABEL:\3@$FUNC(/g" $TEMP done mv $TEMP $NAME fi done llvm-svn: 186268
2013-07-14 03:42:54 +02:00
; CHECK-LABEL: @select5(
Reapply commit 143214 with a fix: m_ICmp doesn't match conditions with the given predicate, it matches any condition and returns the predicate - d'oh! Original commit message: The expression icmp eq (select (icmp eq x, 0), 1, x), 0 folds to false. Spotted by my super-optimizer in 186.crafty and 450.soplex. We really need a proper infrastructure for handling generalizations of this kind of thing (which occur a lot), however this case is so simple that I decided to go ahead and implement it directly. llvm-svn: 143318
2011-10-30 20:56:36 +01:00
%c = icmp eq i32 %x, 0
%s = select i1 %c, i32 1, i32 %x
%c2 = icmp eq i32 %s, 0
ret i1 %c2
; CHECK: ret i1 false
}
define i1 @select6(i32 %x) {
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_]*\):\( *\)@$FUNC\([( ]*\)\$/;\1\2-LABEL:\3@$FUNC(/g" $TEMP done mv $TEMP $NAME fi done llvm-svn: 186268
2013-07-14 03:42:54 +02:00
; CHECK-LABEL: @select6(
Reapply commit 143214 with a fix: m_ICmp doesn't match conditions with the given predicate, it matches any condition and returns the predicate - d'oh! Original commit message: The expression icmp eq (select (icmp eq x, 0), 1, x), 0 folds to false. Spotted by my super-optimizer in 186.crafty and 450.soplex. We really need a proper infrastructure for handling generalizations of this kind of thing (which occur a lot), however this case is so simple that I decided to go ahead and implement it directly. llvm-svn: 143318
2011-10-30 20:56:36 +01:00
%c = icmp sgt i32 %x, 0
%s = select i1 %c, i32 %x, i32 4
%c2 = icmp eq i32 %s, 0
ret i1 %c2
; CHECK: ret i1 %c2
}
Optimize "icmp pred (urem X, Y), Y" --> true/false depending on pred. There's more work to do here, "icmp ult (urem X, 10), 11" doesn't optimize away yet. Fixes example 3 from PR9343! llvm-svn: 126741
2011-03-01 09:15:50 +01:00
define i1 @urem1(i32 %X, i32 %Y) {
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_]*\):\( *\)@$FUNC\([( ]*\)\$/;\1\2-LABEL:\3@$FUNC(/g" $TEMP done mv $TEMP $NAME fi done llvm-svn: 186268
2013-07-14 03:42:54 +02:00
; CHECK-LABEL: @urem1(
Optimize "icmp pred (urem X, Y), Y" --> true/false depending on pred. There's more work to do here, "icmp ult (urem X, 10), 11" doesn't optimize away yet. Fixes example 3 from PR9343! llvm-svn: 126741
2011-03-01 09:15:50 +01:00
%A = urem i32 %X, %Y
%B = icmp ult i32 %A, %Y
ret i1 %B
; CHECK: ret i1 true
}
define i1 @urem2(i32 %X, i32 %Y) {
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_]*\):\( *\)@$FUNC\([( ]*\)\$/;\1\2-LABEL:\3@$FUNC(/g" $TEMP done mv $TEMP $NAME fi done llvm-svn: 186268
2013-07-14 03:42:54 +02:00
; CHECK-LABEL: @urem2(
Optimize "icmp pred (urem X, Y), Y" --> true/false depending on pred. There's more work to do here, "icmp ult (urem X, 10), 11" doesn't optimize away yet. Fixes example 3 from PR9343! llvm-svn: 126741
2011-03-01 09:15:50 +01:00
%A = urem i32 %X, %Y
%B = icmp eq i32 %A, %Y
ret i1 %B
Fix mistyped CHECK lines. llvm-svn: 127366
2011-03-09 23:07:31 +01:00
; CHECK: ret i1 false
Optimize "icmp pred (urem X, Y), Y" --> true/false depending on pred. There's more work to do here, "icmp ult (urem X, 10), 11" doesn't optimize away yet. Fixes example 3 from PR9343! llvm-svn: 126741
2011-03-01 09:15:50 +01:00
}
Teach instruction simplify to use constant ranges to solve problems of the form "icmp pred %X, CI" and a number of examples where "%X = binop %Y, CI2". Some of these cases (div and rem) used to make it through opt -O2, but the others are probably now making code elsewhere redundant (probably instcombine). llvm-svn: 126988
2011-03-04 08:00:57 +01:00
define i1 @urem3(i32 %X) {
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_]*\):\( *\)@$FUNC\([( ]*\)\$/;\1\2-LABEL:\3@$FUNC(/g" $TEMP done mv $TEMP $NAME fi done llvm-svn: 186268
2013-07-14 03:42:54 +02:00
; CHECK-LABEL: @urem3(
Teach instruction simplify to use constant ranges to solve problems of the form "icmp pred %X, CI" and a number of examples where "%X = binop %Y, CI2". Some of these cases (div and rem) used to make it through opt -O2, but the others are probably now making code elsewhere redundant (probably instcombine). llvm-svn: 126988
2011-03-04 08:00:57 +01:00
%A = urem i32 %X, 10
%B = icmp ult i32 %A, 15
ret i1 %B
; CHECK: ret i1 true
}
define i1 @urem4(i32 %X) {
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_]*\):\( *\)@$FUNC\([( ]*\)\$/;\1\2-LABEL:\3@$FUNC(/g" $TEMP done mv $TEMP $NAME fi done llvm-svn: 186268
2013-07-14 03:42:54 +02:00
; CHECK-LABEL: @urem4(
Teach instruction simplify to use constant ranges to solve problems of the form "icmp pred %X, CI" and a number of examples where "%X = binop %Y, CI2". Some of these cases (div and rem) used to make it through opt -O2, but the others are probably now making code elsewhere redundant (probably instcombine). llvm-svn: 126988
2011-03-04 08:00:57 +01:00
%A = urem i32 %X, 15
%B = icmp ult i32 %A, 10
ret i1 %B
; CHECK: ret i1 %B
}
Fold "icmp pred (srem X, Y), Y" like we do for urem. Handle signed comparisons in the urem case, though not the other way around. This is enough to get #3 from PR9343! llvm-svn: 126991
2011-03-04 11:06:52 +01:00
define i1 @urem5(i16 %X, i32 %Y) {
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_]*\):\( *\)@$FUNC\([( ]*\)\$/;\1\2-LABEL:\3@$FUNC(/g" $TEMP done mv $TEMP $NAME fi done llvm-svn: 186268
2013-07-14 03:42:54 +02:00
; CHECK-LABEL: @urem5(
Fold "icmp pred (srem X, Y), Y" like we do for urem. Handle signed comparisons in the urem case, though not the other way around. This is enough to get #3 from PR9343! llvm-svn: 126991
2011-03-04 11:06:52 +01:00
%A = zext i16 %X to i32
%B = urem i32 %A, %Y
%C = icmp slt i32 %B, %Y
ret i1 %C
Fix logic error optimizing "icmp pred (urem X, Y), Y" where pred is signed. Fixes PR16605. llvm-svn: 186229
2013-07-13 01:42:57 +02:00
; CHECK-NOT: ret i1 true
Fold "icmp pred (srem X, Y), Y" like we do for urem. Handle signed comparisons in the urem case, though not the other way around. This is enough to get #3 from PR9343! llvm-svn: 126991
2011-03-04 11:06:52 +01:00
}
Add another micro-optimization. Apologies for the lack of refactoring, but I gave up when I realized I couldn't come up with a good name for what the refactored function would be, to describe what it does. This is PR9343 test12, which is test3 with arguments reordered. Whoops! llvm-svn: 127318
2011-03-09 07:26:03 +01:00
define i1 @urem6(i32 %X, i32 %Y) {
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_]*\):\( *\)@$FUNC\([( ]*\)\$/;\1\2-LABEL:\3@$FUNC(/g" $TEMP done mv $TEMP $NAME fi done llvm-svn: 186268
2013-07-14 03:42:54 +02:00
; CHECK-LABEL: @urem6(
Add another micro-optimization. Apologies for the lack of refactoring, but I gave up when I realized I couldn't come up with a good name for what the refactored function would be, to describe what it does. This is PR9343 test12, which is test3 with arguments reordered. Whoops! llvm-svn: 127318
2011-03-09 07:26:03 +01:00
%A = urem i32 %X, %Y
%B = icmp ugt i32 %Y, %A
ret i1 %B
; CHECK: ret i1 true
}
Fix logic error optimizing "icmp pred (urem X, Y), Y" where pred is signed. Fixes PR16605. llvm-svn: 186229
2013-07-13 01:42:57 +02:00
define i1 @urem7(i32 %X) {
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_]*\):\( *\)@$FUNC\([( ]*\)\$/;\1\2-LABEL:\3@$FUNC(/g" $TEMP done mv $TEMP $NAME fi done llvm-svn: 186268
2013-07-14 03:42:54 +02:00
; CHECK-LABEL: @urem7(
Fix logic error optimizing "icmp pred (urem X, Y), Y" where pred is signed. Fixes PR16605. llvm-svn: 186229
2013-07-13 01:42:57 +02:00
%A = urem i32 1, %X
%B = icmp sgt i32 %A, %X
ret i1 %B
; CHECK-NOT: ret i1 false
}
Teach instruction simplify to use constant ranges to solve problems of the form "icmp pred %X, CI" and a number of examples where "%X = binop %Y, CI2". Some of these cases (div and rem) used to make it through opt -O2, but the others are probably now making code elsewhere redundant (probably instcombine). llvm-svn: 126988
2011-03-04 08:00:57 +01:00
define i1 @srem1(i32 %X) {
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_]*\):\( *\)@$FUNC\([( ]*\)\$/;\1\2-LABEL:\3@$FUNC(/g" $TEMP done mv $TEMP $NAME fi done llvm-svn: 186268
2013-07-14 03:42:54 +02:00
; CHECK-LABEL: @srem1(
Teach instruction simplify to use constant ranges to solve problems of the form "icmp pred %X, CI" and a number of examples where "%X = binop %Y, CI2". Some of these cases (div and rem) used to make it through opt -O2, but the others are probably now making code elsewhere redundant (probably instcombine). llvm-svn: 126988
2011-03-04 08:00:57 +01:00
%A = srem i32 %X, -5
%B = icmp sgt i32 %A, 5
ret i1 %B
; CHECK: ret i1 false
}
Teach ComputeMaskedBits about nsw on add. I don't think there's anything we can do with nuw here, but sub and mul should be given similar treatment. Fixes PR9343 #15! llvm-svn: 127463
2011-03-11 10:00:19 +01:00
; PR9343 #15
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_]*\):\( *\)@$FUNC\([( ]*\)\$/;\1\2-LABEL:\3@$FUNC(/g" $TEMP done mv $TEMP $NAME fi done llvm-svn: 186268
2013-07-14 03:42:54 +02:00
; CHECK-LABEL: @srem2(
Teach ComputeMaskedBits about nsw on add. I don't think there's anything we can do with nuw here, but sub and mul should be given similar treatment. Fixes PR9343 #15! llvm-svn: 127463
2011-03-11 10:00:19 +01:00
; CHECK: ret i1 false
define i1 @srem2(i16 %X, i32 %Y) {
%A = zext i16 %X to i32
%B = add nsw i32 %A, 1
%C = srem i32 %B, %Y
%D = icmp slt i32 %C, 0
ret i1 %D
}
Teach ComputeMaskedBits about sub nsw. llvm-svn: 127548
2011-03-12 18:18:11 +01:00
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_]*\):\( *\)@$FUNC\([( ]*\)\$/;\1\2-LABEL:\3@$FUNC(/g" $TEMP done mv $TEMP $NAME fi done llvm-svn: 186268
2013-07-14 03:42:54 +02:00
; CHECK-LABEL: @srem3(
Teach ComputeMaskedBits about sub nsw. llvm-svn: 127548
2011-03-12 18:18:11 +01:00
; CHECK-NEXT: ret i1 false
define i1 @srem3(i16 %X, i32 %Y) {
%A = zext i16 %X to i32
%B = or i32 2147483648, %A
%C = sub nsw i32 1, %B
%D = srem i32 %C, %Y
%E = icmp slt i32 %D, 0
ret i1 %E
}
Teach instruction simplify to use constant ranges to solve problems of the form "icmp pred %X, CI" and a number of examples where "%X = binop %Y, CI2". Some of these cases (div and rem) used to make it through opt -O2, but the others are probably now making code elsewhere redundant (probably instcombine). llvm-svn: 126988
2011-03-04 08:00:57 +01:00
define i1 @udiv1(i32 %X) {
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_]*\):\( *\)@$FUNC\([( ]*\)\$/;\1\2-LABEL:\3@$FUNC(/g" $TEMP done mv $TEMP $NAME fi done llvm-svn: 186268
2013-07-14 03:42:54 +02:00
; CHECK-LABEL: @udiv1(
Teach instruction simplify to use constant ranges to solve problems of the form "icmp pred %X, CI" and a number of examples where "%X = binop %Y, CI2". Some of these cases (div and rem) used to make it through opt -O2, but the others are probably now making code elsewhere redundant (probably instcombine). llvm-svn: 126988
2011-03-04 08:00:57 +01:00
%A = udiv i32 %X, 1000000
%B = icmp ult i32 %A, 5000
ret i1 %B
; CHECK: ret i1 true
}
Thread comparisons over udiv/sdiv/ashr/lshr exact and lshr nuw/nsw whenever possible. This goes into instcombine and instsimplify because instsimplify doesn't need to check hasOneUse since it returns (almost exclusively) constants. This fixes PR9343 #4 #5 and #8! llvm-svn: 127064
2011-03-05 06:19:11 +01:00
define i1 @udiv2(i32 %X, i32 %Y, i32 %Z) {
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_]*\):\( *\)@$FUNC\([( ]*\)\$/;\1\2-LABEL:\3@$FUNC(/g" $TEMP done mv $TEMP $NAME fi done llvm-svn: 186268
2013-07-14 03:42:54 +02:00
; CHECK-LABEL: @udiv2(
Thread comparisons over udiv/sdiv/ashr/lshr exact and lshr nuw/nsw whenever possible. This goes into instcombine and instsimplify because instsimplify doesn't need to check hasOneUse since it returns (almost exclusively) constants. This fixes PR9343 #4 #5 and #8! llvm-svn: 127064
2011-03-05 06:19:11 +01:00
%A = udiv exact i32 10, %Z
%B = udiv exact i32 20, %Z
%C = icmp ult i32 %A, %B
ret i1 %C
; CHECK: ret i1 true
}
Fold icmp ugt (udiv X, Y), X to false. Spotted by my super-optimizer in 186.crafty. llvm-svn: 143209
2011-10-28 20:17:44 +02:00
define i1 @udiv3(i32 %X, i32 %Y) {
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_]*\):\( *\)@$FUNC\([( ]*\)\$/;\1\2-LABEL:\3@$FUNC(/g" $TEMP done mv $TEMP $NAME fi done llvm-svn: 186268
2013-07-14 03:42:54 +02:00
; CHECK-LABEL: @udiv3(
Fold icmp ugt (udiv X, Y), X to false. Spotted by my super-optimizer in 186.crafty. llvm-svn: 143209
2011-10-28 20:17:44 +02:00
%A = udiv i32 %X, %Y
%C = icmp ugt i32 %A, %X
ret i1 %C
; CHECK: ret i1 false
}
define i1 @udiv4(i32 %X, i32 %Y) {
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_]*\):\( *\)@$FUNC\([( ]*\)\$/;\1\2-LABEL:\3@$FUNC(/g" $TEMP done mv $TEMP $NAME fi done llvm-svn: 186268
2013-07-14 03:42:54 +02:00
; CHECK-LABEL: @udiv4(
Fold icmp ugt (udiv X, Y), X to false. Spotted by my super-optimizer in 186.crafty. llvm-svn: 143209
2011-10-28 20:17:44 +02:00
%A = udiv i32 %X, %Y
%C = icmp ule i32 %A, %X
ret i1 %C
; CHECK: ret i1 true
}
define i1 @udiv5(i32 %X) {
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_]*\):\( *\)@$FUNC\([( ]*\)\$/;\1\2-LABEL:\3@$FUNC(/g" $TEMP done mv $TEMP $NAME fi done llvm-svn: 186268
2013-07-14 03:42:54 +02:00
; CHECK-LABEL: @udiv5(
Fold icmp ugt (udiv X, Y), X to false. Spotted by my super-optimizer in 186.crafty. llvm-svn: 143209
2011-10-28 20:17:44 +02:00
%A = udiv i32 123, %X
%C = icmp ugt i32 %A, 124
ret i1 %C
; CHECK: ret i1 false
}
Fix code to match comment. Fixes PR11340, a regression from r143209. llvm-svn: 144121
2011-11-08 22:08:02 +01:00
; PR11340
define i1 @udiv6(i32 %X) nounwind {
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_]*\):\( *\)@$FUNC\([( ]*\)\$/;\1\2-LABEL:\3@$FUNC(/g" $TEMP done mv $TEMP $NAME fi done llvm-svn: 186268
2013-07-14 03:42:54 +02:00
; CHECK-LABEL: @udiv6(
Fix code to match comment. Fixes PR11340, a regression from r143209. llvm-svn: 144121
2011-11-08 22:08:02 +01:00
%A = udiv i32 1, %X
%C = icmp eq i32 %A, 0
ret i1 %C
; CHECK: ret i1 %C
}
Teach instruction simplify to use constant ranges to solve problems of the form "icmp pred %X, CI" and a number of examples where "%X = binop %Y, CI2". Some of these cases (div and rem) used to make it through opt -O2, but the others are probably now making code elsewhere redundant (probably instcombine). llvm-svn: 126988
2011-03-04 08:00:57 +01:00
define i1 @sdiv1(i32 %X) {
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_]*\):\( *\)@$FUNC\([( ]*\)\$/;\1\2-LABEL:\3@$FUNC(/g" $TEMP done mv $TEMP $NAME fi done llvm-svn: 186268
2013-07-14 03:42:54 +02:00
; CHECK-LABEL: @sdiv1(
Teach instruction simplify to use constant ranges to solve problems of the form "icmp pred %X, CI" and a number of examples where "%X = binop %Y, CI2". Some of these cases (div and rem) used to make it through opt -O2, but the others are probably now making code elsewhere redundant (probably instcombine). llvm-svn: 126988
2011-03-04 08:00:57 +01:00
%A = sdiv i32 %X, 1000000
%B = icmp slt i32 %A, 3000
ret i1 %B
; CHECK: ret i1 true
}
define i1 @or1(i32 %X) {
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_]*\):\( *\)@$FUNC\([( ]*\)\$/;\1\2-LABEL:\3@$FUNC(/g" $TEMP done mv $TEMP $NAME fi done llvm-svn: 186268
2013-07-14 03:42:54 +02:00
; CHECK-LABEL: @or1(
Teach instruction simplify to use constant ranges to solve problems of the form "icmp pred %X, CI" and a number of examples where "%X = binop %Y, CI2". Some of these cases (div and rem) used to make it through opt -O2, but the others are probably now making code elsewhere redundant (probably instcombine). llvm-svn: 126988
2011-03-04 08:00:57 +01:00
%A = or i32 %X, 62
%B = icmp ult i32 %A, 50
ret i1 %B
; CHECK: ret i1 false
}
define i1 @and1(i32 %X) {
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_]*\):\( *\)@$FUNC\([( ]*\)\$/;\1\2-LABEL:\3@$FUNC(/g" $TEMP done mv $TEMP $NAME fi done llvm-svn: 186268
2013-07-14 03:42:54 +02:00
; CHECK-LABEL: @and1(
Teach instruction simplify to use constant ranges to solve problems of the form "icmp pred %X, CI" and a number of examples where "%X = binop %Y, CI2". Some of these cases (div and rem) used to make it through opt -O2, but the others are probably now making code elsewhere redundant (probably instcombine). llvm-svn: 126988
2011-03-04 08:00:57 +01:00
%A = and i32 %X, 62
%B = icmp ugt i32 %A, 70
ret i1 %B
; CHECK: ret i1 false
}
Reapply commit 143028 with a fix: the problem was casting a ConstantExpr Mul using BinaryOperator (which only works for instructions) when it should have been a cast to OverflowingBinaryOperator (which also works for constants). While there, correct a few other dubious looking uses of BinaryOperator. Thanks to Chad Rosier for the testcase. Original commit message: My super-optimizer noticed that we weren't folding this expression to true: (x *nsw x) sgt 0, where x = (y | 1). This occurs in 464.h264ref. llvm-svn: 143125
2011-10-27 21:16:21 +02:00
define i1 @mul1(i32 %X) {
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_]*\):\( *\)@$FUNC\([( ]*\)\$/;\1\2-LABEL:\3@$FUNC(/g" $TEMP done mv $TEMP $NAME fi done llvm-svn: 186268
2013-07-14 03:42:54 +02:00
; CHECK-LABEL: @mul1(
Reapply commit 143028 with a fix: the problem was casting a ConstantExpr Mul using BinaryOperator (which only works for instructions) when it should have been a cast to OverflowingBinaryOperator (which also works for constants). While there, correct a few other dubious looking uses of BinaryOperator. Thanks to Chad Rosier for the testcase. Original commit message: My super-optimizer noticed that we weren't folding this expression to true: (x *nsw x) sgt 0, where x = (y | 1). This occurs in 464.h264ref. llvm-svn: 143125
2011-10-27 21:16:21 +02:00
; Square of a non-zero number is non-zero if there is no overflow.
%Y = or i32 %X, 1
%M = mul nuw i32 %Y, %Y
%C = icmp eq i32 %M, 0
ret i1 %C
; CHECK: ret i1 false
}
define i1 @mul2(i32 %X) {
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_]*\):\( *\)@$FUNC\([( ]*\)\$/;\1\2-LABEL:\3@$FUNC(/g" $TEMP done mv $TEMP $NAME fi done llvm-svn: 186268
2013-07-14 03:42:54 +02:00
; CHECK-LABEL: @mul2(
Reapply commit 143028 with a fix: the problem was casting a ConstantExpr Mul using BinaryOperator (which only works for instructions) when it should have been a cast to OverflowingBinaryOperator (which also works for constants). While there, correct a few other dubious looking uses of BinaryOperator. Thanks to Chad Rosier for the testcase. Original commit message: My super-optimizer noticed that we weren't folding this expression to true: (x *nsw x) sgt 0, where x = (y | 1). This occurs in 464.h264ref. llvm-svn: 143125
2011-10-27 21:16:21 +02:00
; Square of a non-zero number is positive if there is no signed overflow.
%Y = or i32 %X, 1
%M = mul nsw i32 %Y, %Y
%C = icmp sgt i32 %M, 0
ret i1 %C
; CHECK: ret i1 true
}
define i1 @mul3(i32 %X, i32 %Y) {
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_]*\):\( *\)@$FUNC\([( ]*\)\$/;\1\2-LABEL:\3@$FUNC(/g" $TEMP done mv $TEMP $NAME fi done llvm-svn: 186268
2013-07-14 03:42:54 +02:00
; CHECK-LABEL: @mul3(
Reapply commit 143028 with a fix: the problem was casting a ConstantExpr Mul using BinaryOperator (which only works for instructions) when it should have been a cast to OverflowingBinaryOperator (which also works for constants). While there, correct a few other dubious looking uses of BinaryOperator. Thanks to Chad Rosier for the testcase. Original commit message: My super-optimizer noticed that we weren't folding this expression to true: (x *nsw x) sgt 0, where x = (y | 1). This occurs in 464.h264ref. llvm-svn: 143125
2011-10-27 21:16:21 +02:00
; Product of non-negative numbers is non-negative if there is no signed overflow.
%XX = mul nsw i32 %X, %X
%YY = mul nsw i32 %Y, %Y
%M = mul nsw i32 %XX, %YY
%C = icmp sge i32 %M, 0
ret i1 %C
; CHECK: ret i1 true
}
Revert commit 149912 (lattner) and add a testcase that shows the problem (which is that patterns no longer match for vectors of booleans, because you only get ConstantDataVector when the vector element type is i8, i16, etc, not when it is i1). Original commit message: Remove some dead code and tidy things up now that vectors use ConstantDataVector instead of always using ConstantVector. llvm-svn: 150246
2012-02-10 15:26:42 +01:00
define <2 x i1> @vectorselect1(<2 x i1> %cond) {
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_]*\):\( *\)@$FUNC\([( ]*\)\$/;\1\2-LABEL:\3@$FUNC(/g" $TEMP done mv $TEMP $NAME fi done llvm-svn: 186268
2013-07-14 03:42:54 +02:00
; CHECK-LABEL: @vectorselect1(
Revert commit 149912 (lattner) and add a testcase that shows the problem (which is that patterns no longer match for vectors of booleans, because you only get ConstantDataVector when the vector element type is i8, i16, etc, not when it is i1). Original commit message: Remove some dead code and tidy things up now that vectors use ConstantDataVector instead of always using ConstantVector. llvm-svn: 150246
2012-02-10 15:26:42 +01:00
%invert = xor <2 x i1> %cond, <i1 1, i1 1>
%s = select <2 x i1> %invert, <2 x i32> <i32 0, i32 0>, <2 x i32> <i32 1, i32 1>
%c = icmp ne <2 x i32> %s, <i32 0, i32 0>
ret <2 x i1> %c
; CHECK: ret <2 x i1> %cond
}
Fix PR11948: the result type of an icmp may be a vector of boolean - don't assume it is a boolean. llvm-svn: 150247
2012-02-10 15:31:24 +01:00
fold comparisons of gep'd alloca points with null to false, implementing PR12013. We now compile the testcase to: __Z4testv: ## @_Z4testv ## BB#0: ## %_ZN4llvm15SmallVectorImplIiE9push_backERKi.exit pushq %rbx subq $64, %rsp leaq 32(%rsp), %rbx movq %rbx, (%rsp) leaq 64(%rsp), %rax movq %rax, 16(%rsp) movl $1, 32(%rsp) leaq 36(%rsp), %rax movq %rax, 8(%rsp) leaq (%rsp), %rdi callq __Z1gRN4llvm11SmallVectorIiLj8EEE movq (%rsp), %rdi cmpq %rbx, %rdi je LBB0_2 ## BB#1: callq _free LBB0_2: ## %_ZN4llvm11SmallVectorIiLj8EED1Ev.exit addq $64, %rsp popq %rbx ret instead of: __Z4testv: ## @_Z4testv ## BB#0: pushq %rbx subq $64, %rsp xorl %eax, %eax leaq (%rsp), %rbx addq $32, %rbx movq %rbx, (%rsp) movq %rbx, 8(%rsp) leaq 64(%rsp), %rcx movq %rcx, 16(%rsp) je LBB0_2 ## BB#1: movl $1, 32(%rsp) movq %rbx, %rax LBB0_2: ## %_ZN4llvm15SmallVectorImplIiE9push_backERKi.exit addq $4, %rax movq %rax, 8(%rsp) leaq (%rsp), %rdi callq __Z1gRN4llvm11SmallVectorIiLj8EEE movq (%rsp), %rdi cmpq %rbx, %rdi je LBB0_4 ## BB#3: callq _free LBB0_4: ## %_ZN4llvm11SmallVectorIiLj8EED1Ev.exit addq $64, %rsp popq %rbx ret This doesn't shrink clang noticably though. llvm-svn: 150944
2012-02-20 01:42:49 +01:00
; PR11948
define <2 x i1> @vectorselectcrash(i32 %arg1) {
Fix PR11948: the result type of an icmp may be a vector of boolean - don't assume it is a boolean. llvm-svn: 150247
2012-02-10 15:31:24 +01:00
%tobool40 = icmp ne i32 %arg1, 0
%cond43 = select i1 %tobool40, <2 x i16> <i16 -5, i16 66>, <2 x i16> <i16 46, i16 1>
%cmp45 = icmp ugt <2 x i16> %cond43, <i16 73, i16 21>
ret <2 x i1> %cmp45
}
fold comparisons of gep'd alloca points with null to false, implementing PR12013. We now compile the testcase to: __Z4testv: ## @_Z4testv ## BB#0: ## %_ZN4llvm15SmallVectorImplIiE9push_backERKi.exit pushq %rbx subq $64, %rsp leaq 32(%rsp), %rbx movq %rbx, (%rsp) leaq 64(%rsp), %rax movq %rax, 16(%rsp) movl $1, 32(%rsp) leaq 36(%rsp), %rax movq %rax, 8(%rsp) leaq (%rsp), %rdi callq __Z1gRN4llvm11SmallVectorIiLj8EEE movq (%rsp), %rdi cmpq %rbx, %rdi je LBB0_2 ## BB#1: callq _free LBB0_2: ## %_ZN4llvm11SmallVectorIiLj8EED1Ev.exit addq $64, %rsp popq %rbx ret instead of: __Z4testv: ## @_Z4testv ## BB#0: pushq %rbx subq $64, %rsp xorl %eax, %eax leaq (%rsp), %rbx addq $32, %rbx movq %rbx, (%rsp) movq %rbx, 8(%rsp) leaq 64(%rsp), %rcx movq %rcx, 16(%rsp) je LBB0_2 ## BB#1: movl $1, 32(%rsp) movq %rbx, %rax LBB0_2: ## %_ZN4llvm15SmallVectorImplIiE9push_backERKi.exit addq $4, %rax movq %rax, 8(%rsp) leaq (%rsp), %rdi callq __Z1gRN4llvm11SmallVectorIiLj8EEE movq (%rsp), %rdi cmpq %rbx, %rdi je LBB0_4 ## BB#3: callq _free LBB0_4: ## %_ZN4llvm11SmallVectorIiLj8EED1Ev.exit addq $64, %rsp popq %rbx ret This doesn't shrink clang noticably though. llvm-svn: 150944
2012-02-20 01:42:49 +01:00
; PR12013
define i1 @alloca_compare(i64 %idx) {
%sv = alloca { i32, i32, [124 x i32] }
%1 = getelementptr inbounds { i32, i32, [124 x i32] }* %sv, i32 0, i32 2, i64 %idx
%2 = icmp eq i32* %1, null
ret i1 %2
; CHECK: alloca_compare
; CHECK: ret i1 false
}
fix PR12075, a regression in a recent transform I added. In unreachable code, gep chains can be infinite. Just like "stripPointerCasts", use a set to keep track of visited instructions so we don't recurse infinitely. llvm-svn: 151383
2012-02-24 20:01:58 +01:00
; PR12075
define i1 @infinite_gep() {
ret i1 1
unreachableblock:
%X = getelementptr i32 *%X, i32 1
%Y = icmp eq i32* %X, null
ret i1 %Y
}
An alloca can be equal to an argument. It can't *alias* an alloca, but it could be equal, since there's nothing preventing a caller from correctly predicting the stack location of an alloca. llvm-svn: 174119
2013-02-01 00:49:33 +01:00
; It's not valid to fold a comparison of an argument with an alloca, even though
; that's tempting. An argument can't *alias* an alloca, however the aliasing rule
; relies on restrictions against guessing an object's address and dereferencing.
; There are no restrictions against guessing an object's address and comparing.
define i1 @alloca_argument_compare(i64* %arg) {
%alloc = alloca i64
%cmp = icmp eq i64* %arg, %alloc
ret i1 %cmp
; CHECK: alloca_argument_compare
; CHECK: ret i1 %cmp
}
Rewrite instsimplify's handling if icmp on pointer values to remove the remaining use of AliasAnalysis concepts such as isIdentifiedObject to prove pointer inequality. @external_compare in test/Transforms/InstSimplify/compare.ll shows a simple case where a noalias argument can be equal to a global variable address, and while AliasAnalysis can get away with saying that these pointers don't alias, instsimplify cannot say that they are not equal. llvm-svn: 174122
2013-02-01 01:11:13 +01:00
; As above, but with the operands reversed.
define i1 @alloca_argument_compare_swapped(i64* %arg) {
%alloc = alloca i64
%cmp = icmp eq i64* %alloc, %arg
ret i1 %cmp
; CHECK: alloca_argument_compare_swapped
; CHECK: ret i1 %cmp
}
; Don't assume that a noalias argument isn't equal to a global variable's
; address. This is an example where AliasAnalysis' NoAlias concept is
; different from actual pointer inequality.
@y = external global i32
define zeroext i1 @external_compare(i32* noalias %x) {
%cmp = icmp eq i32* %x, @y
ret i1 %cmp
; CHECK: external_compare
; CHECK: ret i1 %cmp
}
Check whether a pointer is non-null (isKnownNonNull) in isKnownNonZero. This handles the case where we have an inbounds GEP with alloca as the pointer. This fixes the regression in PR12750 and rdar://13286434. Note that we can also fix this by handling some GEP cases in isKnownNonNull. llvm-svn: 177321
2013-03-18 22:23:25 +01:00
define i1 @alloca_gep(i64 %a, i64 %b) {
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_]*\):\( *\)@$FUNC\([( ]*\)\$/;\1\2-LABEL:\3@$FUNC(/g" $TEMP done mv $TEMP $NAME fi done llvm-svn: 186268
2013-07-14 03:42:54 +02:00
; CHECK-LABEL: @alloca_gep(
Check whether a pointer is non-null (isKnownNonNull) in isKnownNonZero. This handles the case where we have an inbounds GEP with alloca as the pointer. This fixes the regression in PR12750 and rdar://13286434. Note that we can also fix this by handling some GEP cases in isKnownNonNull. llvm-svn: 177321
2013-03-18 22:23:25 +01:00
; We can prove this GEP is non-null because it is inbounds and the pointer
; is non-null.
%strs = alloca [1000 x [1001 x i8]], align 16
%x = getelementptr inbounds [1000 x [1001 x i8]]* %strs, i64 0, i64 %a, i64 %b
%cmp = icmp eq i8* %x, null
ret i1 %cmp
; CHECK-NEXT: ret i1 false
}
InstSimplify: Fold equality comparisons between non-inbounds GEPs. Overflow doesn't affect the correctness of equalities. Computing this is cheap, we just reuse the computation for the inbounds case and try to peel of more non-inbounds GEPs. This pattern is unlikely to ever appear in code generated by Clang, but SCEV occasionally produces it. llvm-svn: 191200
2013-09-23 16:16:38 +02:00
define i1 @non_inbounds_gep_compare(i64* %a) {
; CHECK-LABEL: @non_inbounds_gep_compare(
; Equality compares with non-inbounds GEPs can be folded.
%x = getelementptr i64* %a, i64 42
Expand test case a bit. llvm-svn: 191205
2013-09-23 16:41:35 +02:00
%y = getelementptr inbounds i64* %x, i64 -42
%z = getelementptr i64* %a, i64 -42
%w = getelementptr inbounds i64* %z, i64 42
%cmp = icmp eq i64* %y, %w
InstSimplify: Fold equality comparisons between non-inbounds GEPs. Overflow doesn't affect the correctness of equalities. Computing this is cheap, we just reuse the computation for the inbounds case and try to peel of more non-inbounds GEPs. This pattern is unlikely to ever appear in code generated by Clang, but SCEV occasionally produces it. llvm-svn: 191200
2013-09-23 16:16:38 +02:00
ret i1 %cmp
Expand test case a bit. llvm-svn: 191205
2013-09-23 16:41:35 +02:00
; CHECK-NEXT: ret i1 true
InstSimplify: Fold equality comparisons between non-inbounds GEPs. Overflow doesn't affect the correctness of equalities. Computing this is cheap, we just reuse the computation for the inbounds case and try to peel of more non-inbounds GEPs. This pattern is unlikely to ever appear in code generated by Clang, but SCEV occasionally produces it. llvm-svn: 191200
2013-09-23 16:16:38 +02:00
}
Add a test that large offsets on GEPs on 32 bits targets are handled correctly. llvm-svn: 191628
2013-09-28 23:27:49 +02:00
define i1 @non_inbounds_gep_compare2(i64* %a) {
; CHECK-LABEL: @non_inbounds_gep_compare2(
; Equality compares with non-inbounds GEPs can be folded.
%x = getelementptr i64* %a, i64 4294967297
%y = getelementptr i64* %a, i64 1
%cmp = icmp eq i64* %y, %y
ret i1 %cmp
; CHECK-NEXT: ret i1 true
}
InstSimplify: Make shift, select and GEP simplifications vector-aware. llvm-svn: 200016
2014-01-24 18:09:53 +01:00
define <4 x i8> @vectorselectfold(<4 x i8> %a, <4 x i8> %b) {
%false = icmp ne <4 x i8> zeroinitializer, zeroinitializer
%sel = select <4 x i1> %false, <4 x i8> %a, <4 x i8> %b
ret <4 x i8> %sel
; CHECK-LABEL: @vectorselectfold
; CHECK-NEXT: ret <4 x i8> %b
}
define <4 x i8> @vectorselectfold2(<4 x i8> %a, <4 x i8> %b) {
%true = icmp eq <4 x i8> zeroinitializer, zeroinitializer
%sel = select <4 x i1> %true, <4 x i8> %a, <4 x i8> %b
ret <4 x i8> %sel
; CHECK-LABEL: @vectorselectfold
; CHECK-NEXT: ret <4 x i8> %a
}
InstSimplify: Optimize signed icmp of -(zext V) Summary: We know that -(zext V) will always be <= zero, simplify signed icmps that have these. Uncovered using http://www.cs.utah.edu/~regehr/souper/ Reviewers: nicholas Subscribers: llvm-commits Differential Revision: http://reviews.llvm.org/D3754 llvm-svn: 208809
2014-05-14 22:16:28 +02:00
define i1 @compare_always_true_slt(i16 %a) {
%1 = zext i16 %a to i32
%2 = sub nsw i32 0, %1
%3 = icmp slt i32 %2, 1
ret i1 %3
; CHECK-LABEL: @compare_always_true_slt
; CHECK-NEXT: ret i1 true
}
define i1 @compare_always_true_sle(i16 %a) {
%1 = zext i16 %a to i32
%2 = sub nsw i32 0, %1
%3 = icmp sle i32 %2, 0
ret i1 %3
; CHECK-LABEL: @compare_always_true_sle
; CHECK-NEXT: ret i1 true
}
define i1 @compare_always_false_sgt(i16 %a) {
%1 = zext i16 %a to i32
%2 = sub nsw i32 0, %1
%3 = icmp sgt i32 %2, 0
ret i1 %3
; CHECK-LABEL: @compare_always_false_sgt
; CHECK-NEXT: ret i1 false
}
define i1 @compare_always_false_sge(i16 %a) {
%1 = zext i16 %a to i32
%2 = sub nsw i32 0, %1
%3 = icmp sge i32 %2, 1
ret i1 %3
; CHECK-LABEL: @compare_always_false_sge
; CHECK-NEXT: ret i1 false
}
define i1 @compare_always_false_eq(i16 %a) {
%1 = zext i16 %a to i32
%2 = sub nsw i32 0, %1
%3 = icmp eq i32 %2, 1
ret i1 %3
; CHECK-LABEL: @compare_always_false_eq
; CHECK-NEXT: ret i1 false
}
define i1 @compare_always_false_ne(i16 %a) {
%1 = zext i16 %a to i32
%2 = sub nsw i32 0, %1
%3 = icmp ne i32 %2, 1
ret i1 %3
; CHECK-LABEL: @compare_always_false_ne
; CHECK-NEXT: ret i1 true
}
InstSimplify: Optimize using dividend in sdiv Summary: The dividend in an sdiv tells us the largest and smallest possible results. Use this fact to optimize comparisons against an sdiv with a constant dividend. Reviewers: nicholas Subscribers: llvm-commits Differential Revision: http://reviews.llvm.org/D3795 llvm-svn: 208999
2014-05-16 18:57:04 +02:00
define i1 @compare_dividend(i32 %a) {
%div = sdiv i32 2, %a
%cmp = icmp eq i32 %div, 3
ret i1 %cmp
; CHECK-LABEL: @compare_dividend
; CHECK-NEXT: ret i1 false
}
InstSimplify: Improve handling of ashr/lshr Summary: Analyze the range of values produced by ashr/lshr cst, %V when it is being used in an icmp. Reviewers: nicholas Subscribers: llvm-commits Differential Revision: http://reviews.llvm.org/D3774 llvm-svn: 209000
2014-05-16 19:14:03 +02:00
define i1 @lshr_ugt_false(i32 %a) {
%shr = lshr i32 1, %a
%cmp = icmp ugt i32 %shr, 1
ret i1 %cmp
; CHECK-LABEL: @lshr_ugt_false
; CHECK-NEXT: ret i1 false
}
define i1 @exact_lshr_ugt_false(i32 %a) {
%shr = lshr exact i32 30, %a
%cmp = icmp ult i32 %shr, 15
ret i1 %cmp
; CHECK-LABEL: @exact_lshr_ugt_false
; CHECK-NEXT: ret i1 false
}
define i1 @lshr_sgt_false(i32 %a) {
%shr = lshr i32 1, %a
%cmp = icmp sgt i32 %shr, 1
ret i1 %cmp
; CHECK-LABEL: @lshr_sgt_false
; CHECK-NEXT: ret i1 false
}
define i1 @ashr_sgt_false(i32 %a) {
%shr = ashr i32 -30, %a
%cmp = icmp sgt i32 %shr, -1
ret i1 %cmp
; CHECK-LABEL: @ashr_sgt_false
; CHECK-NEXT: ret i1 false
}
define i1 @exact_ashr_sgt_false(i32 %a) {
%shr = ashr exact i32 -30, %a
%cmp = icmp sgt i32 %shr, -15
ret i1 %cmp
; CHECK-LABEL: @exact_ashr_sgt_false
; CHECK-NEXT: ret i1 false
}
Teach isKnownNonNull that a nonnull return is not null. Add a test for this case as well as the case of a nonnull attribute (already handled but not tested). llvm-svn: 209193
2014-05-20 07:13:21 +02:00
define i1 @nonnull_arg(i32* nonnull %i) {
%cmp = icmp eq i32* %i, null
ret i1 %cmp
; CHECK-LABEL: @nonnull_arg
; CHECK: ret i1 false
}
declare nonnull i32* @returns_nonnull_helper()
define i1 @returns_nonnull() {
%call = call nonnull i32* @returns_nonnull_helper()
%cmp = icmp eq i32* %call, null
ret i1 %cmp
; CHECK-LABEL: @returns_nonnull
; CHECK: ret i1 false
}
Make instsimplify's analysis of icmp eq/ne use computeKnownBits to determine whether the icmp is always true or false. Patch by Suyog Sarda! llvm-svn: 211251
2014-06-19 05:35:49 +02:00
; If a bit is known to be zero for A and known to be one for B,
; then A and B cannot be equal.
define i1 @icmp_eq_const(i32 %a) nounwind {
%b = mul nsw i32 %a, -2
%c = icmp eq i32 %b, 1
ret i1 %c
; CHECK-LABEL: @icmp_eq_const
; CHECK-NEXT: ret i1 false
}
define i1 @icmp_ne_const(i32 %a) nounwind {
%b = mul nsw i32 %a, -2
%c = icmp ne i32 %b, 1
ret i1 %c
; CHECK-LABEL: @icmp_ne_const
; CHECK-NEXT: ret i1 true
}
InstSimplify: Fix a bug when INT_MIN is in a sdiv When INT_MIN is the numerator in a sdiv, we would not properly handle overflow when calculating the bounds of possible values; abs(INT_MIN) is not a meaningful number. Instead, check and handle INT_MIN by reasoning that the largest value is INT_MIN/-2 and the smallest value is INT_MIN. This fixes PR20199. llvm-svn: 212307
2014-07-04 02:23:39 +02:00
define i1 @icmp_sdiv_int_min(i32 %a) {
%div = sdiv i32 -2147483648, %a
%cmp = icmp ne i32 %div, -1073741824
ret i1 %cmp
; CHECK-LABEL: @icmp_sdiv_int_min
; CHECK-NEXT: [[DIV:%.*]] = sdiv i32 -2147483648, %a
; CHECK-NEXT: [[CMP:%.*]] = icmp ne i32 [[DIV]], -1073741824
; CHECK-NEXT: ret i1 [[CMP]]
}
InstSimplify: The upper bound of X / C was missing a rounding step Summary: When calculating the upper bound of X / -8589934592, we would perform the following calculation: Floor[INT_MAX / 8589934592] However, flooring the result would make us wrongly come to the conclusion that 1073741824 was not in the set of possible values. Instead, use the ceiling of the result. Reviewers: nicholas Subscribers: llvm-commits Differential Revision: http://reviews.llvm.org/D4502 llvm-svn: 212976
2014-07-14 21:49:57 +02:00
define i1 @icmp_sdiv_pr20288(i64 %a) {
%div = sdiv i64 %a, -8589934592
%cmp = icmp ne i64 %div, 1073741824
ret i1 %cmp
; CHECK-LABEL: @icmp_sdiv_pr20288
; CHECK-NEXT: [[DIV:%.*]] = sdiv i64 %a, -8589934592
; CHECK-NEXT: [[CMP:%.*]] = icmp ne i64 [[DIV]], 1073741824
; CHECK-NEXT: ret i1 [[CMP]]
}
InstSimplify: Correct sdiv x / -1 Determining the bounds of x/ -1 would start off with us dividing it by INT_MIN. Suffice to say, this would not work very well. Instead, handle it upfront by checking for -1 and mapping it to the range: [INT_MIN + 1, INT_MAX. This means that the result of our division can be any value other than INT_MIN. llvm-svn: 212981
2014-07-14 22:38:45 +02:00
define i1 @icmp_sdiv_neg1(i64 %a) {
%div = sdiv i64 %call, -1
%cmp = icmp ne i64 %div, 1073741824
ret i1 %cmp
; CHECK-LABEL: @icmp_sdiv_neg1
; CHECK-NEXT: [[DIV:%.*]] = sdiv i64 %a, -1
; CHECK-NEXT: [[CMP:%.*]] = icmp ne i64 [[DIV]], 1073741824
; CHECK-NEXT: ret i1 [[CMP]]
}
Reference in New Issue Copy Permalink