6502/llvm-6502
1
0
Fork 0
mirror of https://github.com/c64scene-ar/llvm-6502.git synced 2024-12-15 20:29:48 +00:00
Code Issues Projects Releases Wiki Activity
f6eb7e3175
llvm-6502/test/Transforms/SimplifyCFG/speculate-store.ll

109 lines
3.0 KiB
LLVM
Raw Normal View History

SimplifyCFG: If convert single conditional stores This resurrects r179957, but adds code that makes sure we don't touch atomic/volatile stores: This transformation will transform a conditional store with a preceeding uncondtional store to the same location: a[i] = may-alias with a[i] load if (cond) a[i] = Y into an unconditional store. a[i] = X may-alias with a[i] load tmp = cond ? Y : X; a[i] = tmp We assume that on average the cost of a mispredicted branch is going to be higher than the cost of a second store to the same location, and that the secondary benefits of creating a bigger basic block for other optimizations to work on outway the potential case where the branch would be correctly predicted and the cost of the executing the second store would be noticably reflected in performance. hmmer's execution time improves by 30% on an imac12,2 on ref data sets. With this change we are on par with gcc's performance (gcc also performs this transformation). There was a 1.2 % performance improvement on a ARM swift chip. Other tests in the test-suite+external seem to be mostly uninfluenced in my experiments: This optimization was triggered on 41 tests such that the executable was different before/after the patch. Only 1 out of the 40 tests (dealII) was reproducable below 100% (by about .4%). Given that hmmer benefits so much I believe this to be a fair trade off. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@180731 91177308-0d34-0410-b5e6-96231b3b80d8
2013-04-29 21:28:24 +00:00
; RUN: opt -simplifycfg -S < %s | FileCheck %s
define void @ifconvertstore(i32 %m, i32* %A, i32* %B, i32 %C, i32 %D) {
entry:
%arrayidx = getelementptr inbounds i32* %B, i64 0
%0 = load i32* %arrayidx, align 4
%add = add nsw i32 %0, %C
%arrayidx2 = getelementptr inbounds i32* %A, i64 0
; First store to the location.
store i32 %add, i32* %arrayidx2, align 4
%arrayidx4 = getelementptr inbounds i32* %B, i64 1
%1 = load i32* %arrayidx4, align 4
%add5 = add nsw i32 %1, %D
%cmp6 = icmp sgt i32 %add5, %C
br i1 %cmp6, label %if.then, label %ret.end
; Make sure we speculate stores like the following one. It is cheap compared to
; a mispredicated branch.
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: @ifconvertstore(
SimplifyCFG: If convert single conditional stores This resurrects r179957, but adds code that makes sure we don't touch atomic/volatile stores: This transformation will transform a conditional store with a preceeding uncondtional store to the same location: a[i] = may-alias with a[i] load if (cond) a[i] = Y into an unconditional store. a[i] = X may-alias with a[i] load tmp = cond ? Y : X; a[i] = tmp We assume that on average the cost of a mispredicted branch is going to be higher than the cost of a second store to the same location, and that the secondary benefits of creating a bigger basic block for other optimizations to work on outway the potential case where the branch would be correctly predicted and the cost of the executing the second store would be noticably reflected in performance. hmmer's execution time improves by 30% on an imac12,2 on ref data sets. With this change we are on par with gcc's performance (gcc also performs this transformation). There was a 1.2 % performance improvement on a ARM swift chip. Other tests in the test-suite+external seem to be mostly uninfluenced in my experiments: This optimization was triggered on 41 tests such that the executable was different before/after the patch. Only 1 out of the 40 tests (dealII) was reproducable below 100% (by about .4%). Given that hmmer benefits so much I believe this to be a fair trade off. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@180731 91177308-0d34-0410-b5e6-96231b3b80d8
2013-04-29 21:28:24 +00:00
; CHECK: %add5.add = select i1 %cmp6, i32 %add5, i32 %add
; CHECK: store i32 %add5.add, i32* %arrayidx2, align 4
if.then:
store i32 %add5, i32* %arrayidx2, align 4
br label %ret.end
ret.end:
ret void
}
define void @noifconvertstore1(i32 %m, i32* %A, i32* %B, i32 %C, i32 %D) {
entry:
%arrayidx = getelementptr inbounds i32* %B, i64 0
%0 = load i32* %arrayidx, align 4
%add = add nsw i32 %0, %C
%arrayidx2 = getelementptr inbounds i32* %A, i64 0
; Store to a different location.
store i32 %add, i32* %arrayidx, align 4
%arrayidx4 = getelementptr inbounds i32* %B, i64 1
%1 = load i32* %arrayidx4, align 4
%add5 = add nsw i32 %1, %D
%cmp6 = icmp sgt i32 %add5, %C
br i1 %cmp6, label %if.then, label %ret.end
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: @noifconvertstore1(
SimplifyCFG: If convert single conditional stores This resurrects r179957, but adds code that makes sure we don't touch atomic/volatile stores: This transformation will transform a conditional store with a preceeding uncondtional store to the same location: a[i] = may-alias with a[i] load if (cond) a[i] = Y into an unconditional store. a[i] = X may-alias with a[i] load tmp = cond ? Y : X; a[i] = tmp We assume that on average the cost of a mispredicted branch is going to be higher than the cost of a second store to the same location, and that the secondary benefits of creating a bigger basic block for other optimizations to work on outway the potential case where the branch would be correctly predicted and the cost of the executing the second store would be noticably reflected in performance. hmmer's execution time improves by 30% on an imac12,2 on ref data sets. With this change we are on par with gcc's performance (gcc also performs this transformation). There was a 1.2 % performance improvement on a ARM swift chip. Other tests in the test-suite+external seem to be mostly uninfluenced in my experiments: This optimization was triggered on 41 tests such that the executable was different before/after the patch. Only 1 out of the 40 tests (dealII) was reproducable below 100% (by about .4%). Given that hmmer benefits so much I believe this to be a fair trade off. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@180731 91177308-0d34-0410-b5e6-96231b3b80d8
2013-04-29 21:28:24 +00:00
; CHECK-NOT: select
if.then:
store i32 %add5, i32* %arrayidx2, align 4
br label %ret.end
ret.end:
ret void
}
declare void @unknown_fun()
define void @noifconvertstore2(i32 %m, i32* %A, i32* %B, i32 %C, i32 %D) {
entry:
%arrayidx = getelementptr inbounds i32* %B, i64 0
%0 = load i32* %arrayidx, align 4
%add = add nsw i32 %0, %C
%arrayidx2 = getelementptr inbounds i32* %A, i64 0
; First store to the location.
store i32 %add, i32* %arrayidx2, align 4
call void @unknown_fun()
%arrayidx4 = getelementptr inbounds i32* %B, i64 1
%1 = load i32* %arrayidx4, align 4
%add5 = add nsw i32 %1, %D
%cmp6 = icmp sgt i32 %add5, %C
br i1 %cmp6, label %if.then, label %ret.end
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: @noifconvertstore2(
SimplifyCFG: If convert single conditional stores This resurrects r179957, but adds code that makes sure we don't touch atomic/volatile stores: This transformation will transform a conditional store with a preceeding uncondtional store to the same location: a[i] = may-alias with a[i] load if (cond) a[i] = Y into an unconditional store. a[i] = X may-alias with a[i] load tmp = cond ? Y : X; a[i] = tmp We assume that on average the cost of a mispredicted branch is going to be higher than the cost of a second store to the same location, and that the secondary benefits of creating a bigger basic block for other optimizations to work on outway the potential case where the branch would be correctly predicted and the cost of the executing the second store would be noticably reflected in performance. hmmer's execution time improves by 30% on an imac12,2 on ref data sets. With this change we are on par with gcc's performance (gcc also performs this transformation). There was a 1.2 % performance improvement on a ARM swift chip. Other tests in the test-suite+external seem to be mostly uninfluenced in my experiments: This optimization was triggered on 41 tests such that the executable was different before/after the patch. Only 1 out of the 40 tests (dealII) was reproducable below 100% (by about .4%). Given that hmmer benefits so much I believe this to be a fair trade off. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@180731 91177308-0d34-0410-b5e6-96231b3b80d8
2013-04-29 21:28:24 +00:00
; CHECK-NOT: select
if.then:
store i32 %add5, i32* %arrayidx2, align 4
br label %ret.end
ret.end:
ret void
}
define void @noifconvertstore_volatile(i32 %m, i32* %A, i32* %B, i32 %C, i32 %D) {
entry:
%arrayidx = getelementptr inbounds i32* %B, i64 0
%0 = load i32* %arrayidx, align 4
%add = add nsw i32 %0, %C
%arrayidx2 = getelementptr inbounds i32* %A, i64 0
; First store to the location.
store i32 %add, i32* %arrayidx2, align 4
%arrayidx4 = getelementptr inbounds i32* %B, i64 1
%1 = load i32* %arrayidx4, align 4
%add5 = add nsw i32 %1, %D
%cmp6 = icmp sgt i32 %add5, %C
br i1 %cmp6, label %if.then, label %ret.end
; Make sure we don't speculate volatile stores.
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: @noifconvertstore_volatile(
SimplifyCFG: If convert single conditional stores This resurrects r179957, but adds code that makes sure we don't touch atomic/volatile stores: This transformation will transform a conditional store with a preceeding uncondtional store to the same location: a[i] = may-alias with a[i] load if (cond) a[i] = Y into an unconditional store. a[i] = X may-alias with a[i] load tmp = cond ? Y : X; a[i] = tmp We assume that on average the cost of a mispredicted branch is going to be higher than the cost of a second store to the same location, and that the secondary benefits of creating a bigger basic block for other optimizations to work on outway the potential case where the branch would be correctly predicted and the cost of the executing the second store would be noticably reflected in performance. hmmer's execution time improves by 30% on an imac12,2 on ref data sets. With this change we are on par with gcc's performance (gcc also performs this transformation). There was a 1.2 % performance improvement on a ARM swift chip. Other tests in the test-suite+external seem to be mostly uninfluenced in my experiments: This optimization was triggered on 41 tests such that the executable was different before/after the patch. Only 1 out of the 40 tests (dealII) was reproducable below 100% (by about .4%). Given that hmmer benefits so much I believe this to be a fair trade off. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@180731 91177308-0d34-0410-b5e6-96231b3b80d8
2013-04-29 21:28:24 +00:00
; CHECK-NOT: select
if.then:
store volatile i32 %add5, i32* %arrayidx2, align 4
br label %ret.end
ret.end:
ret void
}
Reference in New Issue Copy Permalink