6502/llvm-6502
1
0
Fork 0
mirror of https://github.com/c64scene-ar/llvm-6502.git synced 2024-12-27 13:30:05 +00:00
Code Issues Projects Releases Wiki Activity
7a37166a7a
llvm-6502/test/Transforms/GVN/rle.ll

690 lines
17 KiB
LLVM
Raw Normal View History

Fix GVN creating bitcast between address spaces git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@193710 91177308-0d34-0410-b5e6-96231b3b80d8
2013-10-30 19:05:41 +00:00
; RUN: opt < %s -default-data-layout="e-p:32:32:32-p1:16:16:16-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-f32:32:32-f64:32:64-n8:16:32" -basicaa -gvn -S -die | FileCheck %s
; RUN: opt < %s -default-data-layout="E-p:32:32:32-p1:16:16:16-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-f32:32:32-f64:64:64-n32" -basicaa -gvn -S -die | FileCheck %s
enhance GVN to forward substitute a stored value to a load (and load -> load) when the base pointers must alias but when they are different types. This occurs very very frequently in 176.gcc and other code that uses bitfields a lot. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@82399 91177308-0d34-0410-b5e6-96231b3b80d8
2009-09-20 19:03:47 +00:00
;; Trivial RLE test.
define i32 @test0(i32 %V, i32* %P) {
store i32 %V, i32* %P
%A = load i32* %P
ret 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 git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@186268 91177308-0d34-0410-b5e6-96231b3b80d8
2013-07-14 01:42:54 +00:00
; CHECK-LABEL: @test0(
enhance GVN to forward substitute a stored value to a load (and load -> load) when the base pointers must alias but when they are different types. This occurs very very frequently in 176.gcc and other code that uses bitfields a lot. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@82399 91177308-0d34-0410-b5e6-96231b3b80d8
2009-09-20 19:03:47 +00:00
; CHECK: ret i32 %V
}
fix PR5016, a crash I introduced in GVN handing first class arrays and structs, which cannot be bitcast to integers. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@82460 91177308-0d34-0410-b5e6-96231b3b80d8
2009-09-21 17:24:04 +00:00
;;===----------------------------------------------------------------------===;;
;; Tests for crashers
;;===----------------------------------------------------------------------===;;
;; PR5016
define i8 @crash0({i32, i32} %A, {i32, i32}* %P) {
store {i32, i32} %A, {i32, i32}* %P
%X = bitcast {i32, i32}* %P to i8*
%Y = load i8* %X
ret i8 %Y
}
Fix crasher in GVN due to my recent capture tracking changes. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@145047 91177308-0d34-0410-b5e6-96231b3b80d8
2011-11-21 19:42:56 +00:00
;; No PR filed, crashed in CaptureTracker.
declare void @helper()
define void @crash1() {
tail call void @llvm.memcpy.p0i8.p0i8.i64(i8* undef, i8* undef, i64 undef, i32 1, i1 false) nounwind
%tmp = load i8* bitcast (void ()* @helper to i8*)
%x = icmp eq i8 %tmp, 15
ret void
}
fix PR5016, a crash I introduced in GVN handing first class arrays and structs, which cannot be bitcast to integers. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@82460 91177308-0d34-0410-b5e6-96231b3b80d8
2009-09-21 17:24:04 +00:00
enhance GVN to forward substitute a stored value to a load (and load -> load) when the base pointers must alias but when they are different types. This occurs very very frequently in 176.gcc and other code that uses bitfields a lot. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@82399 91177308-0d34-0410-b5e6-96231b3b80d8
2009-09-20 19:03:47 +00:00
;;===----------------------------------------------------------------------===;;
;; Store -> Load and Load -> Load forwarding where src and dst are different
;; types, but where the base pointer is a must alias.
;;===----------------------------------------------------------------------===;;
;; i32 -> f32 forwarding.
define float @coerce_mustalias1(i32 %V, i32* %P) {
store i32 %V, i32* %P
%P2 = bitcast i32* %P to float*
%A = load float* %P2
ret float %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 $NAME fi done git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@186268 91177308-0d34-0410-b5e6-96231b3b80d8
2013-07-14 01:42:54 +00:00
; CHECK-LABEL: @coerce_mustalias1(
enhance GVN to forward substitute a stored value to a load (and load -> load) when the base pointers must alias but when they are different types. This occurs very very frequently in 176.gcc and other code that uses bitfields a lot. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@82399 91177308-0d34-0410-b5e6-96231b3b80d8
2009-09-20 19:03:47 +00:00
; CHECK-NOT: load
; CHECK: ret float
}
;; i32* -> float forwarding.
define float @coerce_mustalias2(i32* %V, i32** %P) {
store i32* %V, i32** %P
%P2 = bitcast i32** %P to float*
%A = load float* %P2
ret float %A
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$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 $NAME fi done git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@186268 91177308-0d34-0410-b5e6-96231b3b80d8
2013-07-14 01:42:54 +00:00
; CHECK-LABEL: @coerce_mustalias2(
enhance GVN to forward substitute a stored value to a load (and load -> load) when the base pointers must alias but when they are different types. This occurs very very frequently in 176.gcc and other code that uses bitfields a lot. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@82399 91177308-0d34-0410-b5e6-96231b3b80d8
2009-09-20 19:03:47 +00:00
; CHECK-NOT: load
; CHECK: ret float
}
;; float -> i32* forwarding.
define i32* @coerce_mustalias3(float %V, float* %P) {
store float %V, float* %P
%P2 = bitcast float* %P to i32**
%A = load i32** %P2
ret i32* %A
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$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 git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@186268 91177308-0d34-0410-b5e6-96231b3b80d8
2013-07-14 01:42:54 +00:00
; CHECK-LABEL: @coerce_mustalias3(
enhance GVN to forward substitute a stored value to a load (and load -> load) when the base pointers must alias but when they are different types. This occurs very very frequently in 176.gcc and other code that uses bitfields a lot. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@82399 91177308-0d34-0410-b5e6-96231b3b80d8
2009-09-20 19:03:47 +00:00
; CHECK-NOT: load
; CHECK: ret i32*
}
;; i32 -> f32 load forwarding.
define float @coerce_mustalias4(i32* %P, i1 %cond) {
%A = load i32* %P
Move CoerceAvailableValueToLoadType earlier in GVN.cpp. Hook it up so that nonlocal and partially redundant loads can use it as well. The testcase shows examples of craziness this can handle. This triggers *many* times in 176.gcc. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@82403 91177308-0d34-0410-b5e6-96231b3b80d8
2009-09-20 20:09:34 +00:00
enhance GVN to forward substitute a stored value to a load (and load -> load) when the base pointers must alias but when they are different types. This occurs very very frequently in 176.gcc and other code that uses bitfields a lot. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@82399 91177308-0d34-0410-b5e6-96231b3b80d8
2009-09-20 19:03:47 +00:00
%P2 = bitcast i32* %P to float*
%B = load float* %P2
Move CoerceAvailableValueToLoadType earlier in GVN.cpp. Hook it up so that nonlocal and partially redundant loads can use it as well. The testcase shows examples of craziness this can handle. This triggers *many* times in 176.gcc. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@82403 91177308-0d34-0410-b5e6-96231b3b80d8
2009-09-20 20:09:34 +00:00
br i1 %cond, label %T, label %F
T:
enhance GVN to forward substitute a stored value to a load (and load -> load) when the base pointers must alias but when they are different types. This occurs very very frequently in 176.gcc and other code that uses bitfields a lot. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@82399 91177308-0d34-0410-b5e6-96231b3b80d8
2009-09-20 19:03:47 +00:00
ret float %B
F:
%X = bitcast i32 %A to float
ret float %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 $NAME fi done git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@186268 91177308-0d34-0410-b5e6-96231b3b80d8
2013-07-14 01:42:54 +00:00
; CHECK-LABEL: @coerce_mustalias4(
enhance GVN to forward substitute a stored value to a load (and load -> load) when the base pointers must alias but when they are different types. This occurs very very frequently in 176.gcc and other code that uses bitfields a lot. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@82399 91177308-0d34-0410-b5e6-96231b3b80d8
2009-09-20 19:03:47 +00:00
; CHECK: %A = load i32* %P
; CHECK-NOT: load
; CHECK: ret float
; CHECK: F:
}
;; i32 -> i8 forwarding
define i8 @coerce_mustalias5(i32 %V, i32* %P) {
store i32 %V, i32* %P
%P2 = bitcast i32* %P to i8*
%A = load i8* %P2
ret i8 %A
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$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 $NAME fi done git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@186268 91177308-0d34-0410-b5e6-96231b3b80d8
2013-07-14 01:42:54 +00:00
; CHECK-LABEL: @coerce_mustalias5(
enhance GVN to forward substitute a stored value to a load (and load -> load) when the base pointers must alias but when they are different types. This occurs very very frequently in 176.gcc and other code that uses bitfields a lot. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@82399 91177308-0d34-0410-b5e6-96231b3b80d8
2009-09-20 19:03:47 +00:00
; CHECK-NOT: load
; CHECK: ret i8
}
;; i64 -> float forwarding
define float @coerce_mustalias6(i64 %V, i64* %P) {
store i64 %V, i64* %P
%P2 = bitcast i64* %P to float*
%A = load float* %P2
ret float %A
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$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 $NAME fi done git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@186268 91177308-0d34-0410-b5e6-96231b3b80d8
2013-07-14 01:42:54 +00:00
; CHECK-LABEL: @coerce_mustalias6(
enhance GVN to forward substitute a stored value to a load (and load -> load) when the base pointers must alias but when they are different types. This occurs very very frequently in 176.gcc and other code that uses bitfields a lot. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@82399 91177308-0d34-0410-b5e6-96231b3b80d8
2009-09-20 19:03:47 +00:00
; CHECK-NOT: load
; CHECK: ret float
}
;; i64 -> i8* (32-bit) forwarding
define i8* @coerce_mustalias7(i64 %V, i64* %P) {
store i64 %V, i64* %P
%P2 = bitcast i64* %P to i8**
%A = load i8** %P2
ret i8* %A
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$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 git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@186268 91177308-0d34-0410-b5e6-96231b3b80d8
2013-07-14 01:42:54 +00:00
; CHECK-LABEL: @coerce_mustalias7(
enhance GVN to forward substitute a stored value to a load (and load -> load) when the base pointers must alias but when they are different types. This occurs very very frequently in 176.gcc and other code that uses bitfields a lot. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@82399 91177308-0d34-0410-b5e6-96231b3b80d8
2009-09-20 19:03:47 +00:00
; CHECK-NOT: load
; CHECK: ret i8*
}
Handle forwarding local memsets to loads. For example, we optimize this: short x(short *A) { memset(A, 1, sizeof(*A)*100); return A[42]; } to 'return 257' instead of doing the load. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@90695 91177308-0d34-0410-b5e6-96231b3b80d8
2009-12-06 01:57:02 +00:00
; memset -> i16 forwarding.
define signext i16 @memset_to_i16_local(i16* %A) nounwind ssp {
entry:
%conv = bitcast i16* %A to i8*
rip out a ton of intrinsic modernization logic from AutoUpgrade.cpp, which is for pre-2.9 bitcode files. We keep x86 unaligned loads, movnt, crc32, and the target indep prefetch change. As usual, updating the testsuite is a PITA. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133337 91177308-0d34-0410-b5e6-96231b3b80d8
2011-06-18 06:05:24 +00:00
tail call void @llvm.memset.p0i8.i64(i8* %conv, i8 1, i64 200, i32 1, i1 false)
Handle forwarding local memsets to loads. For example, we optimize this: short x(short *A) { memset(A, 1, sizeof(*A)*100); return A[42]; } to 'return 257' instead of doing the load. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@90695 91177308-0d34-0410-b5e6-96231b3b80d8
2009-12-06 01:57:02 +00:00
%arrayidx = getelementptr inbounds i16* %A, i64 42
%tmp2 = load i16* %arrayidx
ret i16 %tmp2
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 $NAME fi done git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@186268 91177308-0d34-0410-b5e6-96231b3b80d8
2013-07-14 01:42:54 +00:00
; CHECK-LABEL: @memset_to_i16_local(
Handle forwarding local memsets to loads. For example, we optimize this: short x(short *A) { memset(A, 1, sizeof(*A)*100); return A[42]; } to 'return 257' instead of doing the load. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@90695 91177308-0d34-0410-b5e6-96231b3b80d8
2009-12-06 01:57:02 +00:00
; CHECK-NOT: load
; CHECK: ret i16 257
}
; memset -> float forwarding.
define float @memset_to_float_local(float* %A, i8 %Val) nounwind ssp {
entry:
%conv = bitcast float* %A to i8* ; <i8*> [#uses=1]
rip out a ton of intrinsic modernization logic from AutoUpgrade.cpp, which is for pre-2.9 bitcode files. We keep x86 unaligned loads, movnt, crc32, and the target indep prefetch change. As usual, updating the testsuite is a PITA. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133337 91177308-0d34-0410-b5e6-96231b3b80d8
2011-06-18 06:05:24 +00:00
tail call void @llvm.memset.p0i8.i64(i8* %conv, i8 %Val, i64 400, i32 1, i1 false)
Handle forwarding local memsets to loads. For example, we optimize this: short x(short *A) { memset(A, 1, sizeof(*A)*100); return A[42]; } to 'return 257' instead of doing the load. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@90695 91177308-0d34-0410-b5e6-96231b3b80d8
2009-12-06 01:57:02 +00:00
%arrayidx = getelementptr inbounds float* %A, i64 42 ; <float*> [#uses=1]
%tmp2 = load float* %arrayidx ; <float> [#uses=1]
ret float %tmp2
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$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 git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@186268 91177308-0d34-0410-b5e6-96231b3b80d8
2013-07-14 01:42:54 +00:00
; CHECK-LABEL: @memset_to_float_local(
Handle forwarding local memsets to loads. For example, we optimize this: short x(short *A) { memset(A, 1, sizeof(*A)*100); return A[42]; } to 'return 257' instead of doing the load. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@90695 91177308-0d34-0410-b5e6-96231b3b80d8
2009-12-06 01:57:02 +00:00
; CHECK-NOT: load
; CHECK: zext
; CHECK-NEXT: shl
; CHECK-NEXT: or
; CHECK-NEXT: shl
; CHECK-NEXT: or
; CHECK-NEXT: bitcast
; CHECK-NEXT: ret float
}
add support for forwarding mem intrinsic values to non-local loads. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@90697 91177308-0d34-0410-b5e6-96231b3b80d8
2009-12-06 04:54:31 +00:00
;; non-local memset -> i16 load forwarding.
define i16 @memset_to_i16_nonlocal0(i16* %P, i1 %cond) {
%P3 = bitcast i16* %P to i8*
br i1 %cond, label %T, label %F
T:
rip out a ton of intrinsic modernization logic from AutoUpgrade.cpp, which is for pre-2.9 bitcode files. We keep x86 unaligned loads, movnt, crc32, and the target indep prefetch change. As usual, updating the testsuite is a PITA. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133337 91177308-0d34-0410-b5e6-96231b3b80d8
2011-06-18 06:05:24 +00:00
tail call void @llvm.memset.p0i8.i64(i8* %P3, i8 1, i64 400, i32 1, i1 false)
add support for forwarding mem intrinsic values to non-local loads. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@90697 91177308-0d34-0410-b5e6-96231b3b80d8
2009-12-06 04:54:31 +00:00
br label %Cont
F:
rip out a ton of intrinsic modernization logic from AutoUpgrade.cpp, which is for pre-2.9 bitcode files. We keep x86 unaligned loads, movnt, crc32, and the target indep prefetch change. As usual, updating the testsuite is a PITA. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133337 91177308-0d34-0410-b5e6-96231b3b80d8
2011-06-18 06:05:24 +00:00
tail call void @llvm.memset.p0i8.i64(i8* %P3, i8 2, i64 400, i32 1, i1 false)
add support for forwarding mem intrinsic values to non-local loads. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@90697 91177308-0d34-0410-b5e6-96231b3b80d8
2009-12-06 04:54:31 +00:00
br label %Cont
Cont:
%P2 = getelementptr i16* %P, i32 4
%A = load i16* %P2
ret i16 %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 git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@186268 91177308-0d34-0410-b5e6-96231b3b80d8
2013-07-14 01:42:54 +00:00
; CHECK-LABEL: @memset_to_i16_nonlocal0(
add support for forwarding mem intrinsic values to non-local loads. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@90697 91177308-0d34-0410-b5e6-96231b3b80d8
2009-12-06 04:54:31 +00:00
; CHECK: Cont:
; CHECK-NEXT: %A = phi i16 [ 514, %F ], [ 257, %T ]
; CHECK-NOT: load
; CHECK: ret i16 %A
}
constant fold loads from memcpy's from global constants. This is important because clang lowers nontrivial automatic struct/array inits to memcpy from a global array. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@90698 91177308-0d34-0410-b5e6-96231b3b80d8
2009-12-06 05:29:56 +00:00
@GCst = constant {i32, float, i32 } { i32 42, float 14., i32 97 }
Fix GVN creating bitcast between address spaces git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@193710 91177308-0d34-0410-b5e6-96231b3b80d8
2013-10-30 19:05:41 +00:00
@GCst_as1 = addrspace(1) constant {i32, float, i32 } { i32 42, float 14., i32 97 }
constant fold loads from memcpy's from global constants. This is important because clang lowers nontrivial automatic struct/array inits to memcpy from a global array. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@90698 91177308-0d34-0410-b5e6-96231b3b80d8
2009-12-06 05:29:56 +00:00
; memset -> float forwarding.
define float @memcpy_to_float_local(float* %A) nounwind ssp {
entry:
%conv = bitcast float* %A to i8* ; <i8*> [#uses=1]
rip out a ton of intrinsic modernization logic from AutoUpgrade.cpp, which is for pre-2.9 bitcode files. We keep x86 unaligned loads, movnt, crc32, and the target indep prefetch change. As usual, updating the testsuite is a PITA. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133337 91177308-0d34-0410-b5e6-96231b3b80d8
2011-06-18 06:05:24 +00:00
tail call void @llvm.memcpy.p0i8.p0i8.i64(i8* %conv, i8* bitcast ({i32, float, i32 }* @GCst to i8*), i64 12, i32 1, i1 false)
constant fold loads from memcpy's from global constants. This is important because clang lowers nontrivial automatic struct/array inits to memcpy from a global array. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@90698 91177308-0d34-0410-b5e6-96231b3b80d8
2009-12-06 05:29:56 +00:00
%arrayidx = getelementptr inbounds float* %A, i64 1 ; <float*> [#uses=1]
%tmp2 = load float* %arrayidx ; <float> [#uses=1]
ret float %tmp2
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 git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@186268 91177308-0d34-0410-b5e6-96231b3b80d8
2013-07-14 01:42:54 +00:00
; CHECK-LABEL: @memcpy_to_float_local(
constant fold loads from memcpy's from global constants. This is important because clang lowers nontrivial automatic struct/array inits to memcpy from a global array. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@90698 91177308-0d34-0410-b5e6-96231b3b80d8
2009-12-06 05:29:56 +00:00
; CHECK-NOT: load
; CHECK: ret float 1.400000e+01
}
Fix GVN creating bitcast between address spaces git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@193710 91177308-0d34-0410-b5e6-96231b3b80d8
2013-10-30 19:05:41 +00:00
; memcpy from address space 1
define float @memcpy_to_float_local_as1(float* %A) nounwind ssp {
entry:
%conv = bitcast float* %A to i8* ; <i8*> [#uses=1]
tail call void @llvm.memcpy.p0i8.p1i8.i64(i8* %conv, i8 addrspace(1)* bitcast ({i32, float, i32 } addrspace(1)* @GCst_as1 to i8 addrspace(1)*), i64 12, i32 1, i1 false)
%arrayidx = getelementptr inbounds float* %A, i64 1 ; <float*> [#uses=1]
%tmp2 = load float* %arrayidx ; <float> [#uses=1]
ret float %tmp2
; CHECK-LABEL: @memcpy_to_float_local_as1(
; CHECK-NOT: load
; CHECK: ret float 1.400000e+01
}
Handle forwarding local memsets to loads. For example, we optimize this: short x(short *A) { memset(A, 1, sizeof(*A)*100); return A[42]; } to 'return 257' instead of doing the load. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@90695 91177308-0d34-0410-b5e6-96231b3b80d8
2009-12-06 01:57:02 +00:00
Move CoerceAvailableValueToLoadType earlier in GVN.cpp. Hook it up so that nonlocal and partially redundant loads can use it as well. The testcase shows examples of craziness this can handle. This triggers *many* times in 176.gcc. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@82403 91177308-0d34-0410-b5e6-96231b3b80d8
2009-09-20 20:09:34 +00:00
;; non-local i32/float -> i8 load forwarding.
define i8 @coerce_mustalias_nonlocal0(i32* %P, i1 %cond) {
%P2 = bitcast i32* %P to float*
%P3 = bitcast i32* %P to i8*
br i1 %cond, label %T, label %F
T:
store i32 42, i32* %P
br label %Cont
F:
store float 1.0, float* %P2
br label %Cont
Cont:
%A = load i8* %P3
ret i8 %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 git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@186268 91177308-0d34-0410-b5e6-96231b3b80d8
2013-07-14 01:42:54 +00:00
; CHECK-LABEL: @coerce_mustalias_nonlocal0(
Move CoerceAvailableValueToLoadType earlier in GVN.cpp. Hook it up so that nonlocal and partially redundant loads can use it as well. The testcase shows examples of craziness this can handle. This triggers *many* times in 176.gcc. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@82403 91177308-0d34-0410-b5e6-96231b3b80d8
2009-09-20 20:09:34 +00:00
; CHECK: Cont:
; CHECK: %A = phi i8 [
; CHECK-NOT: load
; CHECK: ret i8 %A
}
add support for forwarding mem intrinsic values to non-local loads. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@90697 91177308-0d34-0410-b5e6-96231b3b80d8
2009-12-06 04:54:31 +00:00
Improve GVN to be able to forward substitute a small load from a piece of a large store when both are in the same block. This allows clang to compile the testcase in PR4216 to this code: _test_bitfield: movl 4(%esp), %eax movl %eax, %ecx andl $-65536, %ecx orl $32962, %eax andl $40186, %eax orl %ecx, %eax ret This is not ideal, but is a whole lot better than the code produced by llvm-gcc: _test_bitfield: movw $-32574, %ax orw 4(%esp), %ax andw $-25350, %ax movw %ax, 4(%esp) movw 7(%esp), %cx shlw $8, %cx movzbl 6(%esp), %edx orw %cx, %dx movzwl %dx, %ecx shll $16, %ecx movzwl %ax, %eax orl %ecx, %eax ret and dramatically better than that produced by gcc 4.2: _test_bitfield: pushl %ebx call L3 "L00000000001$pb": L3: popl %ebx movl 8(%esp), %eax leal 0(,%eax,4), %edx sarb $7, %dl movl %eax, %ecx andl $7168, %ecx andl $-7201, %ebx movzbl %dl, %edx andl $1, %edx sall $5, %edx orl %ecx, %ebx orl %edx, %ebx andl $24, %eax andl $-58336, %ebx orl %eax, %ebx orl $32962, %ebx movl %ebx, %eax popl %ebx ret git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@82439 91177308-0d34-0410-b5e6-96231b3b80d8
2009-09-21 05:57:11 +00:00
;; non-local i32/float -> i8 load forwarding. This also tests that the "P3"
;; bitcast equivalence can be properly phi translated.
define i8 @coerce_mustalias_nonlocal1(i32* %P, i1 %cond) {
%P2 = bitcast i32* %P to float*
br i1 %cond, label %T, label %F
T:
store i32 42, i32* %P
br label %Cont
F:
store float 1.0, float* %P2
br label %Cont
Cont:
%P3 = bitcast i32* %P to i8*
%A = load i8* %P3
ret i8 %A
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$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 git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@186268 91177308-0d34-0410-b5e6-96231b3b80d8
2013-07-14 01:42:54 +00:00
; CHECK-LABEL: @coerce_mustalias_nonlocal1(
unHECKify test. It was fixed by Chris in 2009. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@170017 91177308-0d34-0410-b5e6-96231b3b80d8
2012-12-12 20:43:00 +00:00
; CHECK: Cont:
; CHECK: %A = phi i8 [
; CHECK-NOT: load
; CHECK: ret i8 %A
Improve GVN to be able to forward substitute a small load from a piece of a large store when both are in the same block. This allows clang to compile the testcase in PR4216 to this code: _test_bitfield: movl 4(%esp), %eax movl %eax, %ecx andl $-65536, %ecx orl $32962, %eax andl $40186, %eax orl %ecx, %eax ret This is not ideal, but is a whole lot better than the code produced by llvm-gcc: _test_bitfield: movw $-32574, %ax orw 4(%esp), %ax andw $-25350, %ax movw %ax, 4(%esp) movw 7(%esp), %cx shlw $8, %cx movzbl 6(%esp), %edx orw %cx, %dx movzwl %dx, %ecx shll $16, %ecx movzwl %ax, %eax orl %ecx, %eax ret and dramatically better than that produced by gcc 4.2: _test_bitfield: pushl %ebx call L3 "L00000000001$pb": L3: popl %ebx movl 8(%esp), %eax leal 0(,%eax,4), %edx sarb $7, %dl movl %eax, %ecx andl $7168, %ecx andl $-7201, %ebx movzbl %dl, %edx andl $1, %edx sall $5, %edx orl %ecx, %ebx orl %edx, %ebx andl $24, %eax andl $-58336, %ebx orl %eax, %ebx orl $32962, %ebx movl %ebx, %eax popl %ebx ret git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@82439 91177308-0d34-0410-b5e6-96231b3b80d8
2009-09-21 05:57:11 +00:00
}
Move CoerceAvailableValueToLoadType earlier in GVN.cpp. Hook it up so that nonlocal and partially redundant loads can use it as well. The testcase shows examples of craziness this can handle. This triggers *many* times in 176.gcc. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@82403 91177308-0d34-0410-b5e6-96231b3b80d8
2009-09-20 20:09:34 +00:00
;; non-local i32 -> i8 partial redundancy load forwarding.
define i8 @coerce_mustalias_pre0(i32* %P, i1 %cond) {
%P3 = bitcast i32* %P to i8*
br i1 %cond, label %T, label %F
T:
store i32 42, i32* %P
br label %Cont
F:
br label %Cont
Cont:
%A = load i8* %P3
ret i8 %A
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$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 git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@186268 91177308-0d34-0410-b5e6-96231b3b80d8
2013-07-14 01:42:54 +00:00
; CHECK-LABEL: @coerce_mustalias_pre0(
Move CoerceAvailableValueToLoadType earlier in GVN.cpp. Hook it up so that nonlocal and partially redundant loads can use it as well. The testcase shows examples of craziness this can handle. This triggers *many* times in 176.gcc. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@82403 91177308-0d34-0410-b5e6-96231b3b80d8
2009-09-20 20:09:34 +00:00
; CHECK: F:
; CHECK: load i8* %P3
; CHECK: Cont:
; CHECK: %A = phi i8 [
; CHECK-NOT: load
; CHECK: ret i8 %A
}
Improve GVN to be able to forward substitute a small load from a piece of a large store when both are in the same block. This allows clang to compile the testcase in PR4216 to this code: _test_bitfield: movl 4(%esp), %eax movl %eax, %ecx andl $-65536, %ecx orl $32962, %eax andl $40186, %eax orl %ecx, %eax ret This is not ideal, but is a whole lot better than the code produced by llvm-gcc: _test_bitfield: movw $-32574, %ax orw 4(%esp), %ax andw $-25350, %ax movw %ax, 4(%esp) movw 7(%esp), %cx shlw $8, %cx movzbl 6(%esp), %edx orw %cx, %dx movzwl %dx, %ecx shll $16, %ecx movzwl %ax, %eax orl %ecx, %eax ret and dramatically better than that produced by gcc 4.2: _test_bitfield: pushl %ebx call L3 "L00000000001$pb": L3: popl %ebx movl 8(%esp), %eax leal 0(,%eax,4), %edx sarb $7, %dl movl %eax, %ecx andl $7168, %ecx andl $-7201, %ebx movzbl %dl, %edx andl $1, %edx sall $5, %edx orl %ecx, %ebx orl %edx, %ebx andl $24, %eax andl $-58336, %ebx orl %eax, %ebx orl $32962, %ebx movl %ebx, %eax popl %ebx ret git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@82439 91177308-0d34-0410-b5e6-96231b3b80d8
2009-09-21 05:57:11 +00:00
;;===----------------------------------------------------------------------===;;
;; Store -> Load and Load -> Load forwarding where src and dst are different
;; types, and the reload is an offset from the store pointer.
;;===----------------------------------------------------------------------===;;
enable non-local analysis and PRE of large store -> little load. This doesn't kick in too much because of phi translation issues, but this can be resolved in the future. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@82447 91177308-0d34-0410-b5e6-96231b3b80d8
2009-09-21 06:48:08 +00:00
;; i32 -> i8 forwarding.
add pr# git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@82440 91177308-0d34-0410-b5e6-96231b3b80d8
2009-09-21 05:57:47 +00:00
;; PR4216
Improve GVN to be able to forward substitute a small load from a piece of a large store when both are in the same block. This allows clang to compile the testcase in PR4216 to this code: _test_bitfield: movl 4(%esp), %eax movl %eax, %ecx andl $-65536, %ecx orl $32962, %eax andl $40186, %eax orl %ecx, %eax ret This is not ideal, but is a whole lot better than the code produced by llvm-gcc: _test_bitfield: movw $-32574, %ax orw 4(%esp), %ax andw $-25350, %ax movw %ax, 4(%esp) movw 7(%esp), %cx shlw $8, %cx movzbl 6(%esp), %edx orw %cx, %dx movzwl %dx, %ecx shll $16, %ecx movzwl %ax, %eax orl %ecx, %eax ret and dramatically better than that produced by gcc 4.2: _test_bitfield: pushl %ebx call L3 "L00000000001$pb": L3: popl %ebx movl 8(%esp), %eax leal 0(,%eax,4), %edx sarb $7, %dl movl %eax, %ecx andl $7168, %ecx andl $-7201, %ebx movzbl %dl, %edx andl $1, %edx sall $5, %edx orl %ecx, %ebx orl %edx, %ebx andl $24, %eax andl $-58336, %ebx orl %eax, %ebx orl $32962, %ebx movl %ebx, %eax popl %ebx ret git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@82439 91177308-0d34-0410-b5e6-96231b3b80d8
2009-09-21 05:57:11 +00:00
define i8 @coerce_offset0(i32 %V, i32* %P) {
store i32 %V, i32* %P
%P2 = bitcast i32* %P to i8*
%P3 = getelementptr i8* %P2, i32 2
%A = load i8* %P3
ret i8 %A
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$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 git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@186268 91177308-0d34-0410-b5e6-96231b3b80d8
2013-07-14 01:42:54 +00:00
; CHECK-LABEL: @coerce_offset0(
Improve GVN to be able to forward substitute a small load from a piece of a large store when both are in the same block. This allows clang to compile the testcase in PR4216 to this code: _test_bitfield: movl 4(%esp), %eax movl %eax, %ecx andl $-65536, %ecx orl $32962, %eax andl $40186, %eax orl %ecx, %eax ret This is not ideal, but is a whole lot better than the code produced by llvm-gcc: _test_bitfield: movw $-32574, %ax orw 4(%esp), %ax andw $-25350, %ax movw %ax, 4(%esp) movw 7(%esp), %cx shlw $8, %cx movzbl 6(%esp), %edx orw %cx, %dx movzwl %dx, %ecx shll $16, %ecx movzwl %ax, %eax orl %ecx, %eax ret and dramatically better than that produced by gcc 4.2: _test_bitfield: pushl %ebx call L3 "L00000000001$pb": L3: popl %ebx movl 8(%esp), %eax leal 0(,%eax,4), %edx sarb $7, %dl movl %eax, %ecx andl $7168, %ecx andl $-7201, %ebx movzbl %dl, %edx andl $1, %edx sall $5, %edx orl %ecx, %ebx orl %edx, %ebx andl $24, %eax andl $-58336, %ebx orl %eax, %ebx orl $32962, %ebx movl %ebx, %eax popl %ebx ret git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@82439 91177308-0d34-0410-b5e6-96231b3b80d8
2009-09-21 05:57:11 +00:00
; CHECK-NOT: load
; CHECK: ret i8
}
enable non-local analysis and PRE of large store -> little load. This doesn't kick in too much because of phi translation issues, but this can be resolved in the future. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@82447 91177308-0d34-0410-b5e6-96231b3b80d8
2009-09-21 06:48:08 +00:00
;; non-local i32/float -> i8 load forwarding.
define i8 @coerce_offset_nonlocal0(i32* %P, i1 %cond) {
%P2 = bitcast i32* %P to float*
%P3 = bitcast i32* %P to i8*
%P4 = getelementptr i8* %P3, i32 2
br i1 %cond, label %T, label %F
T:
llvm/ConstantFolding.cpp: Make ReadDataFromGlobal() and FoldReinterpretLoadFromConstPtr() Big-endian-aware. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@167595 91177308-0d34-0410-b5e6-96231b3b80d8
2012-11-08 20:34:25 +00:00
store i32 57005, i32* %P
enable non-local analysis and PRE of large store -> little load. This doesn't kick in too much because of phi translation issues, but this can be resolved in the future. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@82447 91177308-0d34-0410-b5e6-96231b3b80d8
2009-09-21 06:48:08 +00:00
br label %Cont
F:
store float 1.0, float* %P2
br label %Cont
Cont:
%A = load i8* %P4
ret i8 %A
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$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 $NAME fi done git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@186268 91177308-0d34-0410-b5e6-96231b3b80d8
2013-07-14 01:42:54 +00:00
; CHECK-LABEL: @coerce_offset_nonlocal0(
enable non-local analysis and PRE of large store -> little load. This doesn't kick in too much because of phi translation issues, but this can be resolved in the future. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@82447 91177308-0d34-0410-b5e6-96231b3b80d8
2009-09-21 06:48:08 +00:00
; CHECK: Cont:
; CHECK: %A = phi i8 [
; CHECK-NOT: load
; CHECK: ret i8 %A
}
;; non-local i32 -> i8 partial redundancy load forwarding.
define i8 @coerce_offset_pre0(i32* %P, i1 %cond) {
%P3 = bitcast i32* %P to i8*
%P4 = getelementptr i8* %P3, i32 2
br i1 %cond, label %T, label %F
T:
store i32 42, i32* %P
br label %Cont
F:
br label %Cont
Cont:
%A = load i8* %P4
ret i8 %A
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$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 git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@186268 91177308-0d34-0410-b5e6-96231b3b80d8
2013-07-14 01:42:54 +00:00
; CHECK-LABEL: @coerce_offset_pre0(
enable non-local analysis and PRE of large store -> little load. This doesn't kick in too much because of phi translation issues, but this can be resolved in the future. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@82447 91177308-0d34-0410-b5e6-96231b3b80d8
2009-09-21 06:48:08 +00:00
; CHECK: F:
; CHECK: load i8* %P4
; CHECK: Cont:
; CHECK: %A = phi i8 [
; CHECK-NOT: load
; CHECK: ret i8 %A
}
Improve GVN to be able to forward substitute a small load from a piece of a large store when both are in the same block. This allows clang to compile the testcase in PR4216 to this code: _test_bitfield: movl 4(%esp), %eax movl %eax, %ecx andl $-65536, %ecx orl $32962, %eax andl $40186, %eax orl %ecx, %eax ret This is not ideal, but is a whole lot better than the code produced by llvm-gcc: _test_bitfield: movw $-32574, %ax orw 4(%esp), %ax andw $-25350, %ax movw %ax, 4(%esp) movw 7(%esp), %cx shlw $8, %cx movzbl 6(%esp), %edx orw %cx, %dx movzwl %dx, %ecx shll $16, %ecx movzwl %ax, %eax orl %ecx, %eax ret and dramatically better than that produced by gcc 4.2: _test_bitfield: pushl %ebx call L3 "L00000000001$pb": L3: popl %ebx movl 8(%esp), %eax leal 0(,%eax,4), %edx sarb $7, %dl movl %eax, %ecx andl $7168, %ecx andl $-7201, %ebx movzbl %dl, %edx andl $1, %edx sall $5, %edx orl %ecx, %ebx orl %edx, %ebx andl $24, %eax andl $-58336, %ebx orl %eax, %ebx orl $32962, %ebx movl %ebx, %eax popl %ebx ret git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@82439 91177308-0d34-0410-b5e6-96231b3b80d8
2009-09-21 05:57:11 +00:00
Fix PR17952. The symptom is that an assertion is triggered. The assertion was added by me to detect the situation when value is propagated from dead blocks. (We can certainly get rid of assertion; it is safe to do so, because propagating value from dead block to alive join node is certainly ok.) The root cause of this bug is : edge-splitting is conducted on the fly, the edge being split could be a dead edge, therefore the block that split the critial edge needs to be flagged "dead" as well. There are 3 ways to fix this bug: 1) Get rid of the assertion as I mentioned eariler 2) When an dead edge is split, flag the inserted block "dead". 3) proactively split the critical edges connecting dead and live blocks when new dead blocks are revealed. This fix go for 3) with additional 2 LOC. Testing case was added by Rafael the other day. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@194424 91177308-0d34-0410-b5e6-96231b3b80d8
2013-11-11 22:00:23 +00:00
define i32 @chained_load(i32** %p, i32 %x, i32 %y) {
merge two tests. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@90691 91177308-0d34-0410-b5e6-96231b3b80d8
2009-12-06 01:47:24 +00:00
block1:
remove a trivial test, make some other tests less trivial. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@131846 91177308-0d34-0410-b5e6-96231b3b80d8
2011-05-22 07:02:43 +00:00
%A = alloca i32*
merge two tests. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@90691 91177308-0d34-0410-b5e6-96231b3b80d8
2009-12-06 01:47:24 +00:00
%z = load i32** %p
remove a trivial test, make some other tests less trivial. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@131846 91177308-0d34-0410-b5e6-96231b3b80d8
2011-05-22 07:02:43 +00:00
store i32* %z, i32** %A
Fix PR17952. The symptom is that an assertion is triggered. The assertion was added by me to detect the situation when value is propagated from dead blocks. (We can certainly get rid of assertion; it is safe to do so, because propagating value from dead block to alive join node is certainly ok.) The root cause of this bug is : edge-splitting is conducted on the fly, the edge being split could be a dead edge, therefore the block that split the critial edge needs to be flagged "dead" as well. There are 3 ways to fix this bug: 1) Get rid of the assertion as I mentioned eariler 2) When an dead edge is split, flag the inserted block "dead". 3) proactively split the critical edges connecting dead and live blocks when new dead blocks are revealed. This fix go for 3) with additional 2 LOC. Testing case was added by Rafael the other day. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@194424 91177308-0d34-0410-b5e6-96231b3b80d8
2013-11-11 22:00:23 +00:00
%cmp = icmp eq i32 %x, %y
br i1 %cmp, label %block2, label %block3
merge two tests. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@90691 91177308-0d34-0410-b5e6-96231b3b80d8
2009-12-06 01:47:24 +00:00
block2:
%a = load i32** %p
br label %block4
block3:
%b = load i32** %p
br label %block4
block4:
%c = load i32** %p
%d = load i32* %c
ret i32 %d
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 git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@186268 91177308-0d34-0410-b5e6-96231b3b80d8
2013-07-14 01:42:54 +00:00
; CHECK-LABEL: @chained_load(
merge two tests. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@90691 91177308-0d34-0410-b5e6-96231b3b80d8
2009-12-06 01:47:24 +00:00
; CHECK: %z = load i32** %p
; CHECK-NOT: load
; CHECK: %d = load i32* %z
; CHECK-NEXT: ret i32 %d
}
Improve GVN to be able to forward substitute a small load from a piece of a large store when both are in the same block. This allows clang to compile the testcase in PR4216 to this code: _test_bitfield: movl 4(%esp), %eax movl %eax, %ecx andl $-65536, %ecx orl $32962, %eax andl $40186, %eax orl %ecx, %eax ret This is not ideal, but is a whole lot better than the code produced by llvm-gcc: _test_bitfield: movw $-32574, %ax orw 4(%esp), %ax andw $-25350, %ax movw %ax, 4(%esp) movw 7(%esp), %cx shlw $8, %cx movzbl 6(%esp), %edx orw %cx, %dx movzwl %dx, %ecx shll $16, %ecx movzwl %ax, %eax orl %ecx, %eax ret and dramatically better than that produced by gcc 4.2: _test_bitfield: pushl %ebx call L3 "L00000000001$pb": L3: popl %ebx movl 8(%esp), %eax leal 0(,%eax,4), %edx sarb $7, %dl movl %eax, %ecx andl $7168, %ecx andl $-7201, %ebx movzbl %dl, %edx andl $1, %edx sall $5, %edx orl %ecx, %ebx orl %edx, %ebx andl $24, %eax andl $-58336, %ebx orl %eax, %ebx orl $32962, %ebx movl %ebx, %eax popl %ebx ret git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@82439 91177308-0d34-0410-b5e6-96231b3b80d8
2009-09-21 05:57:11 +00:00
Small and carefully crafted testcase showing a miscompilation by GVN that I'm working on. This is manifesting as a miscompile of 255.vortex on some targets. No check lines yet because it fails. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@90520 91177308-0d34-0410-b5e6-96231b3b80d8
2009-12-04 02:12:12 +00:00
declare i1 @cond() readonly
declare i1 @cond2() readonly
define i32 @phi_trans2() {
Update Transforms tests to use CHECK-LABEL for easier debugging. No functionality change. This update was done with the following bash script: find test/Transforms -name "*.ll" | \ while read NAME; do echo "$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 git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@186268 91177308-0d34-0410-b5e6-96231b3b80d8
2013-07-14 01:42:54 +00:00
; CHECK-LABEL: @phi_trans2(
Small and carefully crafted testcase showing a miscompilation by GVN that I'm working on. This is manifesting as a miscompile of 255.vortex on some targets. No check lines yet because it fails. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@90520 91177308-0d34-0410-b5e6-96231b3b80d8
2009-12-04 02:12:12 +00:00
entry:
%P = alloca i32, i32 400
br label %F1
F1:
%A = phi i32 [1, %entry], [2, %F]
%cond2 = call i1 @cond()
br i1 %cond2, label %T1, label %TY
T1:
%P2 = getelementptr i32* %P, i32 %A
%x = load i32* %P2
%cond = call i1 @cond2()
br i1 %cond, label %TX, label %F
F:
%P3 = getelementptr i32* %P, i32 2
store i32 17, i32* %P3
store i32 42, i32* %P2 ; Provides "P[A]".
br label %F1
TX:
Switch GVN and memdep to use PHITransAddr, which correctly handles phi translation of complex expressions like &A[i+1]. This has the following benefits: 1. The phi translation logic is all contained in its own class with a strong interface and verification that it is self consistent. 2. The logic is more correct than before. Previously, if intermediate expressions got PHI translated, we'd miss the update and scan for the wrong pointers in predecessor blocks. @phi_trans2 is a testcase for this. 3. We have a lot less code in memdep. We can handle phi translation across blocks of things like @phi_trans3, which is pretty insane :). This patch should fix the miscompiles of 255.vortex, and I tested it with a bootstrap of llvm-gcc, llvm-test and dejagnu of course. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@90926 91177308-0d34-0410-b5e6-96231b3b80d8
2009-12-09 01:59:31 +00:00
; This load should not be compiled to 'ret i32 42'. An overly clever
; implementation of GVN would see that we're returning 17 if the loop
; executes once or 42 if it executes more than that, but we'd have to do
; loop restructuring to expose this, and GVN shouldn't do this sort of CFG
; transformation.
; CHECK: TX:
; CHECK: ret i32 %x
ret i32 %x
Small and carefully crafted testcase showing a miscompilation by GVN that I'm working on. This is manifesting as a miscompile of 255.vortex on some targets. No check lines yet because it fails. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@90520 91177308-0d34-0410-b5e6-96231b3b80d8
2009-12-04 02:12:12 +00:00
TY:
ret i32 0
}
Fix PR17952. The symptom is that an assertion is triggered. The assertion was added by me to detect the situation when value is propagated from dead blocks. (We can certainly get rid of assertion; it is safe to do so, because propagating value from dead block to alive join node is certainly ok.) The root cause of this bug is : edge-splitting is conducted on the fly, the edge being split could be a dead edge, therefore the block that split the critial edge needs to be flagged "dead" as well. There are 3 ways to fix this bug: 1) Get rid of the assertion as I mentioned eariler 2) When an dead edge is split, flag the inserted block "dead". 3) proactively split the critical edges connecting dead and live blocks when new dead blocks are revealed. This fix go for 3) with additional 2 LOC. Testing case was added by Rafael the other day. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@194424 91177308-0d34-0410-b5e6-96231b3b80d8
2013-11-11 22:00:23 +00:00
define i32 @phi_trans3(i32* %p, 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 git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@186268 91177308-0d34-0410-b5e6-96231b3b80d8
2013-07-14 01:42:54 +00:00
; CHECK-LABEL: @phi_trans3(
Switch GVN and memdep to use PHITransAddr, which correctly handles phi translation of complex expressions like &A[i+1]. This has the following benefits: 1. The phi translation logic is all contained in its own class with a strong interface and verification that it is self consistent. 2. The logic is more correct than before. Previously, if intermediate expressions got PHI translated, we'd miss the update and scan for the wrong pointers in predecessor blocks. @phi_trans2 is a testcase for this. 3. We have a lot less code in memdep. We can handle phi translation across blocks of things like @phi_trans3, which is pretty insane :). This patch should fix the miscompiles of 255.vortex, and I tested it with a bootstrap of llvm-gcc, llvm-test and dejagnu of course. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@90926 91177308-0d34-0410-b5e6-96231b3b80d8
2009-12-09 01:59:31 +00:00
block1:
Fix PR17952. The symptom is that an assertion is triggered. The assertion was added by me to detect the situation when value is propagated from dead blocks. (We can certainly get rid of assertion; it is safe to do so, because propagating value from dead block to alive join node is certainly ok.) The root cause of this bug is : edge-splitting is conducted on the fly, the edge being split could be a dead edge, therefore the block that split the critial edge needs to be flagged "dead" as well. There are 3 ways to fix this bug: 1) Get rid of the assertion as I mentioned eariler 2) When an dead edge is split, flag the inserted block "dead". 3) proactively split the critical edges connecting dead and live blocks when new dead blocks are revealed. This fix go for 3) with additional 2 LOC. Testing case was added by Rafael the other day. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@194424 91177308-0d34-0410-b5e6-96231b3b80d8
2013-11-11 22:00:23 +00:00
%cmpxy = icmp eq i32 %x, %y
br i1 %cmpxy, label %block2, label %block3
Switch GVN and memdep to use PHITransAddr, which correctly handles phi translation of complex expressions like &A[i+1]. This has the following benefits: 1. The phi translation logic is all contained in its own class with a strong interface and verification that it is self consistent. 2. The logic is more correct than before. Previously, if intermediate expressions got PHI translated, we'd miss the update and scan for the wrong pointers in predecessor blocks. @phi_trans2 is a testcase for this. 3. We have a lot less code in memdep. We can handle phi translation across blocks of things like @phi_trans3, which is pretty insane :). This patch should fix the miscompiles of 255.vortex, and I tested it with a bootstrap of llvm-gcc, llvm-test and dejagnu of course. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@90926 91177308-0d34-0410-b5e6-96231b3b80d8
2009-12-09 01:59:31 +00:00
block2:
store i32 87, i32* %p
br label %block4
block3:
%p2 = getelementptr i32* %p, i32 43
store i32 97, i32* %p2
br label %block4
block4:
%A = phi i32 [-1, %block2], [42, %block3]
Fix PR17952. The symptom is that an assertion is triggered. The assertion was added by me to detect the situation when value is propagated from dead blocks. (We can certainly get rid of assertion; it is safe to do so, because propagating value from dead block to alive join node is certainly ok.) The root cause of this bug is : edge-splitting is conducted on the fly, the edge being split could be a dead edge, therefore the block that split the critial edge needs to be flagged "dead" as well. There are 3 ways to fix this bug: 1) Get rid of the assertion as I mentioned eariler 2) When an dead edge is split, flag the inserted block "dead". 3) proactively split the critical edges connecting dead and live blocks when new dead blocks are revealed. This fix go for 3) with additional 2 LOC. Testing case was added by Rafael the other day. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@194424 91177308-0d34-0410-b5e6-96231b3b80d8
2013-11-11 22:00:23 +00:00
br i1 %cmpxy, label %block5, label %exit
Switch GVN and memdep to use PHITransAddr, which correctly handles phi translation of complex expressions like &A[i+1]. This has the following benefits: 1. The phi translation logic is all contained in its own class with a strong interface and verification that it is self consistent. 2. The logic is more correct than before. Previously, if intermediate expressions got PHI translated, we'd miss the update and scan for the wrong pointers in predecessor blocks. @phi_trans2 is a testcase for this. 3. We have a lot less code in memdep. We can handle phi translation across blocks of things like @phi_trans3, which is pretty insane :). This patch should fix the miscompiles of 255.vortex, and I tested it with a bootstrap of llvm-gcc, llvm-test and dejagnu of course. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@90926 91177308-0d34-0410-b5e6-96231b3b80d8
2009-12-09 01:59:31 +00:00
; CHECK: block4:
; CHECK-NEXT: %D = phi i32 [ 87, %block2 ], [ 97, %block3 ]
; CHECK-NOT: load
block5:
%B = add i32 %A, 1
Fix PR17952. The symptom is that an assertion is triggered. The assertion was added by me to detect the situation when value is propagated from dead blocks. (We can certainly get rid of assertion; it is safe to do so, because propagating value from dead block to alive join node is certainly ok.) The root cause of this bug is : edge-splitting is conducted on the fly, the edge being split could be a dead edge, therefore the block that split the critial edge needs to be flagged "dead" as well. There are 3 ways to fix this bug: 1) Get rid of the assertion as I mentioned eariler 2) When an dead edge is split, flag the inserted block "dead". 3) proactively split the critical edges connecting dead and live blocks when new dead blocks are revealed. This fix go for 3) with additional 2 LOC. Testing case was added by Rafael the other day. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@194424 91177308-0d34-0410-b5e6-96231b3b80d8
2013-11-11 22:00:23 +00:00
br i1 %cmpxy, label %block6, label %exit
Switch GVN and memdep to use PHITransAddr, which correctly handles phi translation of complex expressions like &A[i+1]. This has the following benefits: 1. The phi translation logic is all contained in its own class with a strong interface and verification that it is self consistent. 2. The logic is more correct than before. Previously, if intermediate expressions got PHI translated, we'd miss the update and scan for the wrong pointers in predecessor blocks. @phi_trans2 is a testcase for this. 3. We have a lot less code in memdep. We can handle phi translation across blocks of things like @phi_trans3, which is pretty insane :). This patch should fix the miscompiles of 255.vortex, and I tested it with a bootstrap of llvm-gcc, llvm-test and dejagnu of course. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@90926 91177308-0d34-0410-b5e6-96231b3b80d8
2009-12-09 01:59:31 +00:00
block6:
%C = getelementptr i32* %p, i32 %B
Fix PR17952. The symptom is that an assertion is triggered. The assertion was added by me to detect the situation when value is propagated from dead blocks. (We can certainly get rid of assertion; it is safe to do so, because propagating value from dead block to alive join node is certainly ok.) The root cause of this bug is : edge-splitting is conducted on the fly, the edge being split could be a dead edge, therefore the block that split the critial edge needs to be flagged "dead" as well. There are 3 ways to fix this bug: 1) Get rid of the assertion as I mentioned eariler 2) When an dead edge is split, flag the inserted block "dead". 3) proactively split the critical edges connecting dead and live blocks when new dead blocks are revealed. This fix go for 3) with additional 2 LOC. Testing case was added by Rafael the other day. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@194424 91177308-0d34-0410-b5e6-96231b3b80d8
2013-11-11 22:00:23 +00:00
br i1 %cmpxy, label %block7, label %exit
Switch GVN and memdep to use PHITransAddr, which correctly handles phi translation of complex expressions like &A[i+1]. This has the following benefits: 1. The phi translation logic is all contained in its own class with a strong interface and verification that it is self consistent. 2. The logic is more correct than before. Previously, if intermediate expressions got PHI translated, we'd miss the update and scan for the wrong pointers in predecessor blocks. @phi_trans2 is a testcase for this. 3. We have a lot less code in memdep. We can handle phi translation across blocks of things like @phi_trans3, which is pretty insane :). This patch should fix the miscompiles of 255.vortex, and I tested it with a bootstrap of llvm-gcc, llvm-test and dejagnu of course. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@90926 91177308-0d34-0410-b5e6-96231b3b80d8
2009-12-09 01:59:31 +00:00
block7:
%D = load i32* %C
ret i32 %D
; CHECK: block7:
; CHECK-NEXT: ret i32 %D
exit:
ret i32 -1
}
Small and carefully crafted testcase showing a miscompilation by GVN that I'm working on. This is manifesting as a miscompile of 255.vortex on some targets. No check lines yet because it fails. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@90520 91177308-0d34-0410-b5e6-96231b3b80d8
2009-12-04 02:12:12 +00:00
the code in GVN that tries to forward large loads to small stores is not phi translating, thus it miscompiles really crazy testcases. This is from inspection, I haven't seen this in the wild. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@90930 91177308-0d34-0410-b5e6-96231b3b80d8
2009-12-09 02:43:05 +00:00
define i8 @phi_trans4(i8* %p) {
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 git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@186268 91177308-0d34-0410-b5e6-96231b3b80d8
2013-07-14 01:42:54 +00:00
; CHECK-LABEL: @phi_trans4(
the code in GVN that tries to forward large loads to small stores is not phi translating, thus it miscompiles really crazy testcases. This is from inspection, I haven't seen this in the wild. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@90930 91177308-0d34-0410-b5e6-96231b3b80d8
2009-12-09 02:43:05 +00:00
entry:
fix hte last remaining known (by me) phi translation bug. When we reanalyze clobbers to forward pieces of large stores to small loads, we need to consider the properly phi translated pointer in the store block. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@90978 91177308-0d34-0410-b5e6-96231b3b80d8
2009-12-09 18:21:46 +00:00
%X3 = getelementptr i8* %p, i32 192
store i8 192, i8* %X3
the code in GVN that tries to forward large loads to small stores is not phi translating, thus it miscompiles really crazy testcases. This is from inspection, I haven't seen this in the wild. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@90930 91177308-0d34-0410-b5e6-96231b3b80d8
2009-12-09 02:43:05 +00:00
%X = getelementptr i8* %p, i32 4
%Y = load i8* %X
br label %loop
loop:
%i = phi i32 [4, %entry], [192, %loop]
%X2 = getelementptr i8* %p, i32 %i
fix hte last remaining known (by me) phi translation bug. When we reanalyze clobbers to forward pieces of large stores to small loads, we need to consider the properly phi translated pointer in the store block. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@90978 91177308-0d34-0410-b5e6-96231b3b80d8
2009-12-09 18:21:46 +00:00
%Y2 = load i8* %X2
; CHECK: loop:
; CHECK-NEXT: %Y2 = phi i8 [ %Y, %entry ], [ 0, %loop ]
; CHECK-NOT: load i8
the code in GVN that tries to forward large loads to small stores is not phi translating, thus it miscompiles really crazy testcases. This is from inspection, I haven't seen this in the wild. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@90930 91177308-0d34-0410-b5e6-96231b3b80d8
2009-12-09 02:43:05 +00:00
%cond = call i1 @cond2()
fix hte last remaining known (by me) phi translation bug. When we reanalyze clobbers to forward pieces of large stores to small loads, we need to consider the properly phi translated pointer in the store block. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@90978 91177308-0d34-0410-b5e6-96231b3b80d8
2009-12-09 18:21:46 +00:00
%Z = bitcast i8 *%X3 to i32*
the code in GVN that tries to forward large loads to small stores is not phi translating, thus it miscompiles really crazy testcases. This is from inspection, I haven't seen this in the wild. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@90930 91177308-0d34-0410-b5e6-96231b3b80d8
2009-12-09 02:43:05 +00:00
store i32 0, i32* %Z
br i1 %cond, label %loop, label %out
out:
%R = add i8 %Y, %Y2
ret i8 %R
}
define i8 @phi_trans5(i8* %p) {
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 git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@186268 91177308-0d34-0410-b5e6-96231b3b80d8
2013-07-14 01:42:54 +00:00
; CHECK-LABEL: @phi_trans5(
the code in GVN that tries to forward large loads to small stores is not phi translating, thus it miscompiles really crazy testcases. This is from inspection, I haven't seen this in the wild. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@90930 91177308-0d34-0410-b5e6-96231b3b80d8
2009-12-09 02:43:05 +00:00
entry:
fix hte last remaining known (by me) phi translation bug. When we reanalyze clobbers to forward pieces of large stores to small loads, we need to consider the properly phi translated pointer in the store block. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@90978 91177308-0d34-0410-b5e6-96231b3b80d8
2009-12-09 18:21:46 +00:00
the code in GVN that tries to forward large loads to small stores is not phi translating, thus it miscompiles really crazy testcases. This is from inspection, I haven't seen this in the wild. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@90930 91177308-0d34-0410-b5e6-96231b3b80d8
2009-12-09 02:43:05 +00:00
%X4 = getelementptr i8* %p, i32 2
store i8 19, i8* %X4
%X = getelementptr i8* %p, i32 4
%Y = load i8* %X
br label %loop
loop:
%i = phi i32 [4, %entry], [3, %cont]
%X2 = getelementptr i8* %p, i32 %i
fix hte last remaining known (by me) phi translation bug. When we reanalyze clobbers to forward pieces of large stores to small loads, we need to consider the properly phi translated pointer in the store block. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@90978 91177308-0d34-0410-b5e6-96231b3b80d8
2009-12-09 18:21:46 +00:00
%Y2 = load i8* %X2 ; Ensure this load is not being incorrectly replaced.
the code in GVN that tries to forward large loads to small stores is not phi translating, thus it miscompiles really crazy testcases. This is from inspection, I haven't seen this in the wild. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@90930 91177308-0d34-0410-b5e6-96231b3b80d8
2009-12-09 02:43:05 +00:00
%cond = call i1 @cond2()
br i1 %cond, label %cont, label %out
cont:
%Z = getelementptr i8* %X2, i32 -1
%Z2 = bitcast i8 *%Z to i32*
store i32 50462976, i32* %Z2 ;; (1 << 8) | (2 << 16) | (3 << 24)
fix hte last remaining known (by me) phi translation bug. When we reanalyze clobbers to forward pieces of large stores to small loads, we need to consider the properly phi translated pointer in the store block. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@90978 91177308-0d34-0410-b5e6-96231b3b80d8
2009-12-09 18:21:46 +00:00
; CHECK: store i32
; CHECK-NEXT: getelementptr i8* %p, i32 3
; CHECK-NEXT: load i8*
the code in GVN that tries to forward large loads to small stores is not phi translating, thus it miscompiles really crazy testcases. This is from inspection, I haven't seen this in the wild. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@90930 91177308-0d34-0410-b5e6-96231b3b80d8
2009-12-09 02:43:05 +00:00
br label %loop
out:
%R = add i8 %Y, %Y2
ret i8 %R
}
fix PR6642, GVN forwarding from memset to load of the base of the memset. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@99488 91177308-0d34-0410-b5e6-96231b3b80d8
2010-03-25 05:58:19 +00:00
; PR6642
define i32 @memset_to_load() nounwind readnone {
entry:
%x = alloca [256 x i32], align 4 ; <[256 x i32]*> [#uses=2]
%tmp = bitcast [256 x i32]* %x to i8* ; <i8*> [#uses=1]
rip out a ton of intrinsic modernization logic from AutoUpgrade.cpp, which is for pre-2.9 bitcode files. We keep x86 unaligned loads, movnt, crc32, and the target indep prefetch change. As usual, updating the testsuite is a PITA. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133337 91177308-0d34-0410-b5e6-96231b3b80d8
2011-06-18 06:05:24 +00:00
call void @llvm.memset.p0i8.i64(i8* %tmp, i8 0, i64 1024, i32 4, i1 false)
fix PR6642, GVN forwarding from memset to load of the base of the memset. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@99488 91177308-0d34-0410-b5e6-96231b3b80d8
2010-03-25 05:58:19 +00:00
%arraydecay = getelementptr inbounds [256 x i32]* %x, i32 0, i32 0 ; <i32*>
%tmp1 = load i32* %arraydecay ; <i32> [#uses=1]
ret i32 %tmp1
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 git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@186268 91177308-0d34-0410-b5e6-96231b3b80d8
2013-07-14 01:42:54 +00:00
; CHECK-LABEL: @memset_to_load(
fix PR6642, GVN forwarding from memset to load of the base of the memset. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@99488 91177308-0d34-0410-b5e6-96231b3b80d8
2010-03-25 05:58:19 +00:00
; CHECK: ret i32 0
}
the code in GVN that tries to forward large loads to small stores is not phi translating, thus it miscompiles really crazy testcases. This is from inspection, I haven't seen this in the wild. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@90930 91177308-0d34-0410-b5e6-96231b3b80d8
2009-12-09 02:43:05 +00:00
Enhance MemDep: When alias analysis returns a partial alias result, return it as a clobber. This allows GVN to do smart things. Enhance GVN to be smart about the case when a small load is clobbered by a larger overlapping load. In this case, forward the value. This allows us to compile stuff like this: int test(void *P) { int tmp = *(unsigned int*)P; return tmp+*((unsigned char*)P+1); } into: _test: ## @test movl (%rdi), %ecx movzbl %ch, %eax addl %ecx, %eax ret which has one load. We already handled the case where the smaller load was from a must-aliased base pointer. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@130180 91177308-0d34-0410-b5e6-96231b3b80d8
2011-04-26 01:21:15 +00:00
;;===----------------------------------------------------------------------===;;
;; Load -> Load forwarding in partial alias case.
;;===----------------------------------------------------------------------===;;
define i32 @load_load_partial_alias(i8* %P) nounwind ssp {
entry:
%0 = bitcast i8* %P to i32*
%tmp2 = load i32* %0
%add.ptr = getelementptr inbounds i8* %P, i64 1
%tmp5 = load i8* %add.ptr
%conv = zext i8 %tmp5 to i32
%add = add nsw i32 %tmp2, %conv
ret i32 %add
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 git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@186268 91177308-0d34-0410-b5e6-96231b3b80d8
2013-07-14 01:42:54 +00:00
; TEMPORARILYDISABLED-LABEL: @load_load_partial_alias(
Disable the main feature of 130180, the elimination of loads that are redundant with partially-aliasing loads. When computing what portion of a clobbering load value is needed, it doesn't consider phi-translation which may have occurred between the clobbing load and the redundant load. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@132631 91177308-0d34-0410-b5e6-96231b3b80d8
2011-06-04 06:48:50 +00:00
; TEMPORARILYDISABLED: load i32*
; TEMPORARILYDISABLED-NOT: load
; TEMPORARILYDISABLED: lshr i32 {{.*}}, 8
; TEMPORARILYDISABLED-NOT: load
; TEMPORARILYDISABLED: trunc i32 {{.*}} to i8
; TEMPORARILYDISABLED-NOT: load
; TEMPORARILYDISABLED: ret i32
Enhance MemDep: When alias analysis returns a partial alias result, return it as a clobber. This allows GVN to do smart things. Enhance GVN to be smart about the case when a small load is clobbered by a larger overlapping load. In this case, forward the value. This allows us to compile stuff like this: int test(void *P) { int tmp = *(unsigned int*)P; return tmp+*((unsigned char*)P+1); } into: _test: ## @test movl (%rdi), %ecx movzbl %ch, %eax addl %ecx, %eax ret which has one load. We already handled the case where the smaller load was from a must-aliased base pointer. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@130180 91177308-0d34-0410-b5e6-96231b3b80d8
2011-04-26 01:21:15 +00:00
}
Improve the bail-out predicate to really only kick in when phi translation fails. We were bailing out in some cases that would cause us to miss GVN'ing some non-local cases away. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@130206 91177308-0d34-0410-b5e6-96231b3b80d8
2011-04-26 17:41:02 +00:00
; Cross block partial alias case.
define i32 @load_load_partial_alias_cross_block(i8* %P) nounwind ssp {
entry:
%xx = bitcast i8* %P to i32*
%x1 = load i32* %xx, align 4
%cmp = icmp eq i32 %x1, 127
br i1 %cmp, label %land.lhs.true, label %if.end
land.lhs.true: ; preds = %entry
%arrayidx4 = getelementptr inbounds i8* %P, i64 1
%tmp5 = load i8* %arrayidx4, align 1
%conv6 = zext i8 %tmp5 to i32
ret i32 %conv6
if.end:
ret i32 52
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 git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@186268 91177308-0d34-0410-b5e6-96231b3b80d8
2013-07-14 01:42:54 +00:00
; TEMPORARILY_DISABLED-LABEL: @load_load_partial_alias_cross_block(
Disable the main feature of 130180, the elimination of loads that are redundant with partially-aliasing loads. When computing what portion of a clobbering load value is needed, it doesn't consider phi-translation which may have occurred between the clobbing load and the redundant load. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@132631 91177308-0d34-0410-b5e6-96231b3b80d8
2011-06-04 06:48:50 +00:00
; TEMPORARILY_DISABLED: land.lhs.true:
; TEMPORARILY_DISABLED-NOT: load i8
; TEMPORARILY_DISABLED: ret i32 %conv6
Improve the bail-out predicate to really only kick in when phi translation fails. We were bailing out in some cases that would cause us to miss GVN'ing some non-local cases away. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@130206 91177308-0d34-0410-b5e6-96231b3b80d8
2011-04-26 17:41:02 +00:00
}
teach GVN to widen integer loads when they are overaligned, when doing an wider load would allow elimination of subsequent loads, and when the wider load is still a native integer type. This eliminates a ton of loads on various benchmarks involving struct fields, though it is somewhat hobbled by clang not being very aggressive about field alignment. This is yet another step along the way towards resolving PR6627. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@130390 91177308-0d34-0410-b5e6-96231b3b80d8
2011-04-28 07:29:08 +00:00
;;===----------------------------------------------------------------------===;;
;; Load Widening
;;===----------------------------------------------------------------------===;;
final step needed to resolve PR6627, which allows us to flatten the code down to a nice and tidy: %x1 = load i32* %0, align 4 %1 = icmp eq i32 %x1, 1179403647 br i1 %1, label %if.then, label %if.end instead of doing lots of loads and branches. May the FreeBSD bootloader long fit in its allocated space. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@130416 91177308-0d34-0410-b5e6-96231b3b80d8
2011-04-28 18:15:47 +00:00
%widening1 = type { i32, i8, i8, i8, i8 }
teach GVN to widen integer loads when they are overaligned, when doing an wider load would allow elimination of subsequent loads, and when the wider load is still a native integer type. This eliminates a ton of loads on various benchmarks involving struct fields, though it is somewhat hobbled by clang not being very aggressive about field alignment. This is yet another step along the way towards resolving PR6627. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@130390 91177308-0d34-0410-b5e6-96231b3b80d8
2011-04-28 07:29:08 +00:00
@f = global %widening1 zeroinitializer, align 4
final step needed to resolve PR6627, which allows us to flatten the code down to a nice and tidy: %x1 = load i32* %0, align 4 %1 = icmp eq i32 %x1, 1179403647 br i1 %1, label %if.then, label %if.end instead of doing lots of loads and branches. May the FreeBSD bootloader long fit in its allocated space. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@130416 91177308-0d34-0410-b5e6-96231b3b80d8
2011-04-28 18:15:47 +00:00
define i32 @test_widening1(i8* %P) nounwind ssp noredzone {
teach GVN to widen integer loads when they are overaligned, when doing an wider load would allow elimination of subsequent loads, and when the wider load is still a native integer type. This eliminates a ton of loads on various benchmarks involving struct fields, though it is somewhat hobbled by clang not being very aggressive about field alignment. This is yet another step along the way towards resolving PR6627. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@130390 91177308-0d34-0410-b5e6-96231b3b80d8
2011-04-28 07:29:08 +00:00
entry:
%tmp = load i8* getelementptr inbounds (%widening1* @f, i64 0, i32 1), align 4
%conv = zext i8 %tmp to i32
%tmp1 = load i8* getelementptr inbounds (%widening1* @f, i64 0, i32 2), align 1
%conv2 = zext i8 %tmp1 to i32
%add = add nsw i32 %conv, %conv2
ret i32 %add
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 git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@186268 91177308-0d34-0410-b5e6-96231b3b80d8
2013-07-14 01:42:54 +00:00
; CHECK-LABEL: @test_widening1(
teach GVN to widen integer loads when they are overaligned, when doing an wider load would allow elimination of subsequent loads, and when the wider load is still a native integer type. This eliminates a ton of loads on various benchmarks involving struct fields, though it is somewhat hobbled by clang not being very aggressive about field alignment. This is yet another step along the way towards resolving PR6627. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@130390 91177308-0d34-0410-b5e6-96231b3b80d8
2011-04-28 07:29:08 +00:00
; CHECK-NOT: load
; CHECK: load i16*
; CHECK-NOT: load
Fix broken check lines. I really need to find a way to automate this, but I can't come up with a regex that has no false positives while handling tricky cases like custom check prefixes. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@162097 91177308-0d34-0410-b5e6-96231b3b80d8
2012-08-17 12:28:26 +00:00
; CHECK: ret i32
teach GVN to widen integer loads when they are overaligned, when doing an wider load would allow elimination of subsequent loads, and when the wider load is still a native integer type. This eliminates a ton of loads on various benchmarks involving struct fields, though it is somewhat hobbled by clang not being very aggressive about field alignment. This is yet another step along the way towards resolving PR6627. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@130390 91177308-0d34-0410-b5e6-96231b3b80d8
2011-04-28 07:29:08 +00:00
}
final step needed to resolve PR6627, which allows us to flatten the code down to a nice and tidy: %x1 = load i32* %0, align 4 %1 = icmp eq i32 %x1, 1179403647 br i1 %1, label %if.then, label %if.end instead of doing lots of loads and branches. May the FreeBSD bootloader long fit in its allocated space. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@130416 91177308-0d34-0410-b5e6-96231b3b80d8
2011-04-28 18:15:47 +00:00
define i32 @test_widening2() nounwind ssp noredzone {
entry:
%tmp = load i8* getelementptr inbounds (%widening1* @f, i64 0, i32 1), align 4
%conv = zext i8 %tmp to i32
%tmp1 = load i8* getelementptr inbounds (%widening1* @f, i64 0, i32 2), align 1
%conv2 = zext i8 %tmp1 to i32
%add = add nsw i32 %conv, %conv2
%tmp2 = load i8* getelementptr inbounds (%widening1* @f, i64 0, i32 3), align 2
%conv3 = zext i8 %tmp2 to i32
%add2 = add nsw i32 %add, %conv3
%tmp3 = load i8* getelementptr inbounds (%widening1* @f, i64 0, i32 4), align 1
%conv4 = zext i8 %tmp3 to i32
%add3 = add nsw i32 %add2, %conv3
ret i32 %add3
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 git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@186268 91177308-0d34-0410-b5e6-96231b3b80d8
2013-07-14 01:42:54 +00:00
; CHECK-LABEL: @test_widening2(
final step needed to resolve PR6627, which allows us to flatten the code down to a nice and tidy: %x1 = load i32* %0, align 4 %1 = icmp eq i32 %x1, 1179403647 br i1 %1, label %if.then, label %if.end instead of doing lots of loads and branches. May the FreeBSD bootloader long fit in its allocated space. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@130416 91177308-0d34-0410-b5e6-96231b3b80d8
2011-04-28 18:15:47 +00:00
; CHECK-NOT: load
; CHECK: load i32*
; CHECK-NOT: load
Fix broken check lines. I really need to find a way to automate this, but I can't come up with a regex that has no false positives while handling tricky cases like custom check prefixes. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@162097 91177308-0d34-0410-b5e6-96231b3b80d8
2012-08-17 12:28:26 +00:00
; CHECK: ret i32
final step needed to resolve PR6627, which allows us to flatten the code down to a nice and tidy: %x1 = load i32* %0, align 4 %1 = icmp eq i32 %x1, 1179403647 br i1 %1, label %if.then, label %if.end instead of doing lots of loads and branches. May the FreeBSD bootloader long fit in its allocated space. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@130416 91177308-0d34-0410-b5e6-96231b3b80d8
2011-04-28 18:15:47 +00:00
}
Enhance MemDep: When alias analysis returns a partial alias result, return it as a clobber. This allows GVN to do smart things. Enhance GVN to be smart about the case when a small load is clobbered by a larger overlapping load. In this case, forward the value. This allows us to compile stuff like this: int test(void *P) { int tmp = *(unsigned int*)P; return tmp+*((unsigned char*)P+1); } into: _test: ## @test movl (%rdi), %ecx movzbl %ch, %eax addl %ecx, %eax ret which has one load. We already handled the case where the smaller load was from a must-aliased base pointer. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@130180 91177308-0d34-0410-b5e6-96231b3b80d8
2011-04-26 01:21:15 +00:00
rip out a ton of intrinsic modernization logic from AutoUpgrade.cpp, which is for pre-2.9 bitcode files. We keep x86 unaligned loads, movnt, crc32, and the target indep prefetch change. As usual, updating the testsuite is a PITA. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133337 91177308-0d34-0410-b5e6-96231b3b80d8
2011-06-18 06:05:24 +00:00
declare void @llvm.memset.p0i8.i64(i8* nocapture, i8, i64, i32, i1) nounwind
declare void @llvm.memcpy.p0i8.p0i8.i64(i8* nocapture, i8* nocapture, i64, i32, i1) nounwind
Fix GVN creating bitcast between address spaces git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@193710 91177308-0d34-0410-b5e6-96231b3b80d8
2013-10-30 19:05:41 +00:00
declare void @llvm.memcpy.p0i8.p1i8.i64(i8* nocapture, i8 addrspace(1)* nocapture, i64, i32, i1) nounwind
rip out a ton of intrinsic modernization logic from AutoUpgrade.cpp, which is for pre-2.9 bitcode files. We keep x86 unaligned loads, movnt, crc32, and the target indep prefetch change. As usual, updating the testsuite is a PITA. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@133337 91177308-0d34-0410-b5e6-96231b3b80d8
2011-06-18 06:05:24 +00:00
Refactor capture tracking (which already had a couple flags for whether returns and stores capture) to permit the caller to see each capture point and decide whether to continue looking. Use this inside memdep to do an analysis that basicaa won't do. This lets us solve another devirtualization case, fixing PR8908! git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@144580 91177308-0d34-0410-b5e6-96231b3b80d8
2011-11-14 22:49:42 +00:00
;;===----------------------------------------------------------------------===;;
;; Load -> Store dependency which isn't interfered with by a call that happens
;; before the pointer was captured.
;;===----------------------------------------------------------------------===;;
%class.X = type { [8 x i8] }
@_ZTV1X = weak_odr constant [5 x i8*] zeroinitializer
@_ZTV1Y = weak_odr constant [5 x i8*] zeroinitializer
declare void @use()
declare void @use3(i8***, i8**)
; PR8908
define void @test_escape1() nounwind {
%x = alloca i8**, align 8
store i8** getelementptr inbounds ([5 x i8*]* @_ZTV1X, i64 0, i64 2), i8*** %x, align 8
call void @use() nounwind
%DEAD = load i8*** %x, align 8
call void @use3(i8*** %x, i8** %DEAD) nounwind
ret void
; CHECK: test_escape1
; CHECK-NOT: DEAD
; CHECK: ret
}
Reference in New Issue Copy Permalink