6502/llvm-6502
1
0
Fork 0
mirror of https://github.com/c64scene-ar/llvm-6502.git synced 2024-12-15 04:30:12 +00:00
Code Issues Projects Releases Wiki Activity
11ef9456a8
llvm-6502/test/CodeGen/X86/sink-hoist.ll

175 lines
7.2 KiB
LLVM
Raw Normal View History

Enable SandyBridgeModel for all modern Intel P6 descendants. All Intel CPUs since Yonah look a lot alike, at least at the granularity of the scheduling models. We can add more accurate models for processors that aren't Sandy Bridge if required. Haswell will probably need its own. The Atom processor and anything based on NetBurst is completely different. So are the non-Intel chips. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@178080 91177308-0d34-0410-b5e6-96231b3b80d8
2013-03-26 22:19:12 +00:00
; RUN: llc < %s -march=x86-64 -asm-verbose=false -mtriple=x86_64-unknown-linux-gnu -mcpu=nehalem -post-RA-scheduler=true -schedmodel=false | FileCheck %s
Unbreak MachineLICM for instructions that reference RIP on x86-64 too. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@82825 91177308-0d34-0410-b5e6-96231b3b80d8
2009-09-25 23:58:45 +00:00
; Currently, floating-point selects are lowered to CFG triangles.
; This means that one side of the select is always unconditionally
; evaluated, however with MachineSink we can sink the other side so
; that it's conditionally evaluated.
Mass update to CodeGen tests to use CHECK-LABEL for labels corresponding to function definitions for more informative error messages. No functionality change and all updated tests passed locally. This update was done with the following bash script: find test/CodeGen -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc.*debug" $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 sed -i '' "s/;\(.*\)-LABEL-LABEL:/;\1-LABEL:/" $TEMP sed -i '' "s/;\(.*\)-NEXT-LABEL:/;\1-NEXT:/" $TEMP sed -i '' "s/;\(.*\)-NOT-LABEL:/;\1-NOT:/" $TEMP sed -i '' "s/;\(.*\)-DAG-LABEL:/;\1-DAG:/" $TEMP mv $TEMP $NAME fi done git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@186280 91177308-0d34-0410-b5e6-96231b3b80d8
2013-07-14 06:24:09 +00:00
; CHECK-LABEL: foo:
Turn on post-alloc scheduling for x86. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@84431 91177308-0d34-0410-b5e6-96231b3b80d8
2009-10-18 19:57:27 +00:00
; CHECK-NEXT: testb $1, %dil
Flip the new block-placement pass to be on by default. This is mostly to test the waters. I'd like to get results from FNT build bots and other bots running on non-x86 platforms. This feature has been pretty heavily tested over the last few months by me, and it fixes several of the execution time regressions caused by the inlining work by preventing inlining decisions from radically impacting block layout. I've seen very large improvements in yacr2 and ackermann benchmarks, along with the expected noise across all of the benchmark suite whenever code layout changes. I've analyzed all of the regressions and fixed them, or found them to be impossible to fix. See my email to llvmdev for more details. I'd like for this to be in 3.1 as it complements the inliner changes, but if any failures are showing up or anyone has concerns, it is just a flag flip and so can be easily turned off. I'm switching it on tonight to try and get at least one run through various folks' performance suites in case SPEC or something else has serious issues with it. I'll watch bots and revert if anything shows up. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@154816 91177308-0d34-0410-b5e6-96231b3b80d8
2012-04-16 13:49:17 +00:00
; CHECK-NEXT: jne
Turn on post-alloc scheduling for x86. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@84431 91177308-0d34-0410-b5e6-96231b3b80d8
2009-10-18 19:57:27 +00:00
; CHECK-NEXT: divsd
Flip the new block-placement pass to be on by default. This is mostly to test the waters. I'd like to get results from FNT build bots and other bots running on non-x86 platforms. This feature has been pretty heavily tested over the last few months by me, and it fixes several of the execution time regressions caused by the inlining work by preventing inlining decisions from radically impacting block layout. I've seen very large improvements in yacr2 and ackermann benchmarks, along with the expected noise across all of the benchmark suite whenever code layout changes. I've analyzed all of the regressions and fixed them, or found them to be impossible to fix. See my email to llvmdev for more details. I'd like for this to be in 3.1 as it complements the inliner changes, but if any failures are showing up or anyone has concerns, it is just a flag flip and so can be easily turned off. I'm switching it on tonight to try and get at least one run through various folks' performance suites in case SPEC or something else has serious issues with it. I'll watch bots and revert if anything shows up. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@154816 91177308-0d34-0410-b5e6-96231b3b80d8
2012-04-16 13:49:17 +00:00
; CHECK-NEXT: movaps
Enable machine sinking critical edge splitting. e.g. define double @foo(double %x, double %y, i1 %c) nounwind { %a = fdiv double %x, 3.2 %z = select i1 %c, double %a, double %y ret double %z } Was: _foo: divsd LCPI0_0(%rip), %xmm0 testb $1, %dil jne LBB0_2 movaps %xmm1, %xmm0 LBB0_2: ret Now: _foo: testb $1, %dil je LBB0_2 divsd LCPI0_0(%rip), %xmm0 ret LBB0_2: movaps %xmm1, %xmm0 ret This avoids the divsd when early exit is taken. rdar://8454886 git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@114372 91177308-0d34-0410-b5e6-96231b3b80d8
2010-09-20 22:52:00 +00:00
; CHECK-NEXT: ret
; CHECK: divsd
Unbreak MachineLICM for instructions that reference RIP on x86-64 too. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@82825 91177308-0d34-0410-b5e6-96231b3b80d8
2009-09-25 23:58:45 +00:00
define double @foo(double %x, double %y, i1 %c) nounwind {
%a = fdiv double %x, 3.2
%b = fdiv double %y, 3.3
%z = select i1 %c, double %a, double %b
ret double %z
}
Enable machine sinking critical edge splitting. e.g. define double @foo(double %x, double %y, i1 %c) nounwind { %a = fdiv double %x, 3.2 %z = select i1 %c, double %a, double %y ret double %z } Was: _foo: divsd LCPI0_0(%rip), %xmm0 testb $1, %dil jne LBB0_2 movaps %xmm1, %xmm0 LBB0_2: ret Now: _foo: testb $1, %dil je LBB0_2 divsd LCPI0_0(%rip), %xmm0 ret LBB0_2: movaps %xmm1, %xmm0 ret This avoids the divsd when early exit is taken. rdar://8454886 git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@114372 91177308-0d34-0410-b5e6-96231b3b80d8
2010-09-20 22:52:00 +00:00
; Make sure the critical edge is broken so the divsd is sunken below
; the conditional branch.
; rdar://8454886
Mass update to CodeGen tests to use CHECK-LABEL for labels corresponding to function definitions for more informative error messages. No functionality change and all updated tests passed locally. This update was done with the following bash script: find test/CodeGen -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc.*debug" $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 sed -i '' "s/;\(.*\)-LABEL-LABEL:/;\1-LABEL:/" $TEMP sed -i '' "s/;\(.*\)-NEXT-LABEL:/;\1-NEXT:/" $TEMP sed -i '' "s/;\(.*\)-NOT-LABEL:/;\1-NOT:/" $TEMP sed -i '' "s/;\(.*\)-DAG-LABEL:/;\1-DAG:/" $TEMP mv $TEMP $NAME fi done git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@186280 91177308-0d34-0410-b5e6-96231b3b80d8
2013-07-14 06:24:09 +00:00
; CHECK-LABEL: split:
Enable machine sinking critical edge splitting. e.g. define double @foo(double %x, double %y, i1 %c) nounwind { %a = fdiv double %x, 3.2 %z = select i1 %c, double %a, double %y ret double %z } Was: _foo: divsd LCPI0_0(%rip), %xmm0 testb $1, %dil jne LBB0_2 movaps %xmm1, %xmm0 LBB0_2: ret Now: _foo: testb $1, %dil je LBB0_2 divsd LCPI0_0(%rip), %xmm0 ret LBB0_2: movaps %xmm1, %xmm0 ret This avoids the divsd when early exit is taken. rdar://8454886 git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@114372 91177308-0d34-0410-b5e6-96231b3b80d8
2010-09-20 22:52:00 +00:00
; CHECK-NEXT: testb $1, %dil
Enable MI Sched for x86. This changes the SelectionDAG scheduling preference to source order. Soon, the SelectionDAG scheduler can be bypassed saving a nice chunk of compile time. Performance differences that result from this change are often a consequence of register coalescing. The register coalescer is far from perfect. Bugs can be filed for deficiencies. On x86 SandyBridge/Haswell, the source order schedule is often preserved, particularly for small blocks. Register pressure is generally improved over the SD scheduler's ILP mode. However, we are still able to handle large blocks that require latency hiding, unlike the SD scheduler's BURR mode. MI scheduler also attempts to discover the critical path in single-block loops and adjust heuristics accordingly. The MI scheduler relies on the new machine model. This is currently unimplemented for AVX, so we may not be generating the best code yet. Unit tests are updated so they don't depend on SD scheduling heuristics. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@192750 91177308-0d34-0410-b5e6-96231b3b80d8
2013-10-15 23:33:07 +00:00
; CHECK-NEXT: je
Flip the new block-placement pass to be on by default. This is mostly to test the waters. I'd like to get results from FNT build bots and other bots running on non-x86 platforms. This feature has been pretty heavily tested over the last few months by me, and it fixes several of the execution time regressions caused by the inlining work by preventing inlining decisions from radically impacting block layout. I've seen very large improvements in yacr2 and ackermann benchmarks, along with the expected noise across all of the benchmark suite whenever code layout changes. I've analyzed all of the regressions and fixed them, or found them to be impossible to fix. See my email to llvmdev for more details. I'd like for this to be in 3.1 as it complements the inliner changes, but if any failures are showing up or anyone has concerns, it is just a flag flip and so can be easily turned off. I'm switching it on tonight to try and get at least one run through various folks' performance suites in case SPEC or something else has serious issues with it. I'll watch bots and revert if anything shows up. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@154816 91177308-0d34-0410-b5e6-96231b3b80d8
2012-04-16 13:49:17 +00:00
; CHECK: divsd
Enable MI Sched for x86. This changes the SelectionDAG scheduling preference to source order. Soon, the SelectionDAG scheduler can be bypassed saving a nice chunk of compile time. Performance differences that result from this change are often a consequence of register coalescing. The register coalescer is far from perfect. Bugs can be filed for deficiencies. On x86 SandyBridge/Haswell, the source order schedule is often preserved, particularly for small blocks. Register pressure is generally improved over the SD scheduler's ILP mode. However, we are still able to handle large blocks that require latency hiding, unlike the SD scheduler's BURR mode. MI scheduler also attempts to discover the critical path in single-block loops and adjust heuristics accordingly. The MI scheduler relies on the new machine model. This is currently unimplemented for AVX, so we may not be generating the best code yet. Unit tests are updated so they don't depend on SD scheduling heuristics. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@192750 91177308-0d34-0410-b5e6-96231b3b80d8
2013-10-15 23:33:07 +00:00
; CHECK: movaps
; CHECK: ret
Enable machine sinking critical edge splitting. e.g. define double @foo(double %x, double %y, i1 %c) nounwind { %a = fdiv double %x, 3.2 %z = select i1 %c, double %a, double %y ret double %z } Was: _foo: divsd LCPI0_0(%rip), %xmm0 testb $1, %dil jne LBB0_2 movaps %xmm1, %xmm0 LBB0_2: ret Now: _foo: testb $1, %dil je LBB0_2 divsd LCPI0_0(%rip), %xmm0 ret LBB0_2: movaps %xmm1, %xmm0 ret This avoids the divsd when early exit is taken. rdar://8454886 git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@114372 91177308-0d34-0410-b5e6-96231b3b80d8
2010-09-20 22:52:00 +00:00
define double @split(double %x, double %y, i1 %c) nounwind {
%a = fdiv double %x, 3.2
%z = select i1 %c, double %a, double %y
ret double %z
}
Unbreak MachineLICM for instructions that reference RIP on x86-64 too. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@82825 91177308-0d34-0410-b5e6-96231b3b80d8
2009-09-25 23:58:45 +00:00
; Hoist floating-point constant-pool loads out of loops.
Mass update to CodeGen tests to use CHECK-LABEL for labels corresponding to function definitions for more informative error messages. No functionality change and all updated tests passed locally. This update was done with the following bash script: find test/CodeGen -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc.*debug" $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 sed -i '' "s/;\(.*\)-LABEL-LABEL:/;\1-LABEL:/" $TEMP sed -i '' "s/;\(.*\)-NEXT-LABEL:/;\1-NEXT:/" $TEMP sed -i '' "s/;\(.*\)-NOT-LABEL:/;\1-NOT:/" $TEMP sed -i '' "s/;\(.*\)-DAG-LABEL:/;\1-DAG:/" $TEMP mv $TEMP $NAME fi done git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@186280 91177308-0d34-0410-b5e6-96231b3b80d8
2013-07-14 06:24:09 +00:00
; CHECK-LABEL: bar:
Unbreak MachineLICM for instructions that reference RIP on x86-64 too. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@82825 91177308-0d34-0410-b5e6-96231b3b80d8
2009-09-25 23:58:45 +00:00
; CHECK: movsd
; CHECK: align
define void @bar(double* nocapture %p, i64 %n) nounwind {
entry:
%0 = icmp sgt i64 %n, 0
br i1 %0, label %bb, label %return
bb:
%i.03 = phi i64 [ 0, %entry ], [ %3, %bb ]
%scevgep = getelementptr double* %p, i64 %i.03
%1 = load double* %scevgep, align 8
%2 = fdiv double 3.200000e+00, %1
store double %2, double* %scevgep, align 8
%3 = add nsw i64 %i.03, 1
%exitcond = icmp eq i64 %3, %n
br i1 %exitcond, label %return, label %bb
return:
ret void
}
Mark dead physregdefs dead immediately. This helps MachineSink and MachineLICM and other things which run before LiveVariables is run. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@85360 91177308-0d34-0410-b5e6-96231b3b80d8
2009-10-28 01:13:53 +00:00
; Sink instructions with dead EFLAGS defs.
Machine sink could potentially sink instructions into a block where the physical registers it defines then interfere with an existing preg live range. For instance, if we had something like these machine instructions: BB#0 ... = imul ... EFLAGS<imp-def,dead> test ..., EFLAGS<imp-def> jcc BB#2 EFLAGS<imp-use> BB#1 ... ; fallthrough to BB#2 BB#2 ... ; No code that defines EFLAGS jcc ... EFLAGS<imp-use> Machine sink will come along, see that imul implicitly defines EFLAGS, but because it's "dead", it assumes that it can move imul into BB#2. But when it does, imul's "dead" imp-def of EFLAGS is raised from the dead (a zombie) and messes up the condition code for the jump (and pretty much anything else which relies upon it being correct). The solution is to know which pregs are live going into a basic block. However, that information isn't calculated at this point. Nor does the LiveVariables pass take into account non-allocatable physical registers. In lieu of this, we do a *very* conservative pass through the basic block to determine if a preg is live coming out of it. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@105387 91177308-0d34-0410-b5e6-96231b3b80d8
2010-06-03 07:54:20 +00:00
; FIXME: Unfail the zzz test if we can correctly mark pregs with the kill flag.
Enable MI Sched for x86. This changes the SelectionDAG scheduling preference to source order. Soon, the SelectionDAG scheduler can be bypassed saving a nice chunk of compile time. Performance differences that result from this change are often a consequence of register coalescing. The register coalescer is far from perfect. Bugs can be filed for deficiencies. On x86 SandyBridge/Haswell, the source order schedule is often preserved, particularly for small blocks. Register pressure is generally improved over the SD scheduler's ILP mode. However, we are still able to handle large blocks that require latency hiding, unlike the SD scheduler's BURR mode. MI scheduler also attempts to discover the critical path in single-block loops and adjust heuristics accordingly. The MI scheduler relies on the new machine model. This is currently unimplemented for AVX, so we may not be generating the best code yet. Unit tests are updated so they don't depend on SD scheduling heuristics. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@192750 91177308-0d34-0410-b5e6-96231b3b80d8
2013-10-15 23:33:07 +00:00
;
Machine sink could potentially sink instructions into a block where the physical registers it defines then interfere with an existing preg live range. For instance, if we had something like these machine instructions: BB#0 ... = imul ... EFLAGS<imp-def,dead> test ..., EFLAGS<imp-def> jcc BB#2 EFLAGS<imp-use> BB#1 ... ; fallthrough to BB#2 BB#2 ... ; No code that defines EFLAGS jcc ... EFLAGS<imp-use> Machine sink will come along, see that imul implicitly defines EFLAGS, but because it's "dead", it assumes that it can move imul into BB#2. But when it does, imul's "dead" imp-def of EFLAGS is raised from the dead (a zombie) and messes up the condition code for the jump (and pretty much anything else which relies upon it being correct). The solution is to know which pregs are live going into a basic block. However, that information isn't calculated at this point. Nor does the LiveVariables pass take into account non-allocatable physical registers. In lieu of this, we do a *very* conservative pass through the basic block to determine if a preg is live coming out of it. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@105387 91177308-0d34-0410-b5e6-96231b3b80d8
2010-06-03 07:54:20 +00:00
; See <rdar://problem/8030636>. This test isn't valid after we made machine
; sinking more conservative about sinking instructions that define a preg into a
; block when we don't know if the preg is killed within the current block.
Mark dead physregdefs dead immediately. This helps MachineSink and MachineLICM and other things which run before LiveVariables is run. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@85360 91177308-0d34-0410-b5e6-96231b3b80d8
2009-10-28 01:13:53 +00:00
Machine sink could potentially sink instructions into a block where the physical registers it defines then interfere with an existing preg live range. For instance, if we had something like these machine instructions: BB#0 ... = imul ... EFLAGS<imp-def,dead> test ..., EFLAGS<imp-def> jcc BB#2 EFLAGS<imp-use> BB#1 ... ; fallthrough to BB#2 BB#2 ... ; No code that defines EFLAGS jcc ... EFLAGS<imp-use> Machine sink will come along, see that imul implicitly defines EFLAGS, but because it's "dead", it assumes that it can move imul into BB#2. But when it does, imul's "dead" imp-def of EFLAGS is raised from the dead (a zombie) and messes up the condition code for the jump (and pretty much anything else which relies upon it being correct). The solution is to know which pregs are live going into a basic block. However, that information isn't calculated at this point. Nor does the LiveVariables pass take into account non-allocatable physical registers. In lieu of this, we do a *very* conservative pass through the basic block to determine if a preg is live coming out of it. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@105387 91177308-0d34-0410-b5e6-96231b3b80d8
2010-06-03 07:54:20 +00:00
; FIXMEHECK: zzz:
; FIXMEHECK: je
; FIXMEHECK-NEXT: orb
; define zeroext i8 @zzz(i8 zeroext %a, i8 zeroext %b) nounwind readnone {
; entry:
; %tmp = zext i8 %a to i32 ; <i32> [#uses=1]
; %tmp2 = icmp eq i8 %a, 0 ; <i1> [#uses=1]
; %tmp3 = or i8 %b, -128 ; <i8> [#uses=1]
; %tmp4 = and i8 %b, 127 ; <i8> [#uses=1]
; %b_addr.0 = select i1 %tmp2, i8 %tmp4, i8 %tmp3 ; <i8> [#uses=1]
; ret i8 %b_addr.0
; }
Teach MachineLICM to unfold loads from constant memory from otherwise unhoistable instructions in order to allow the loads to be hoisted. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@85364 91177308-0d34-0410-b5e6-96231b3b80d8
2009-10-28 03:21:57 +00:00
Revert r141569 and r141576. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@141594 91177308-0d34-0410-b5e6-96231b3b80d8
2011-10-10 23:18:02 +00:00
; Codegen should hoist and CSE these constants.
Mass update to CodeGen tests to use CHECK-LABEL for labels corresponding to function definitions for more informative error messages. No functionality change and all updated tests passed locally. This update was done with the following bash script: find test/CodeGen -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc.*debug" $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 sed -i '' "s/;\(.*\)-LABEL-LABEL:/;\1-LABEL:/" $TEMP sed -i '' "s/;\(.*\)-NEXT-LABEL:/;\1-NEXT:/" $TEMP sed -i '' "s/;\(.*\)-NOT-LABEL:/;\1-NOT:/" $TEMP sed -i '' "s/;\(.*\)-DAG-LABEL:/;\1-DAG:/" $TEMP mv $TEMP $NAME fi done git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@186280 91177308-0d34-0410-b5e6-96231b3b80d8
2013-07-14 06:24:09 +00:00
; CHECK-LABEL: vv:
Revert r141569 and r141576. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@141594 91177308-0d34-0410-b5e6-96231b3b80d8
2011-10-10 23:18:02 +00:00
; CHECK: LCPI3_0(%rip), %xmm0
; CHECK: LCPI3_1(%rip), %xmm1
; CHECK: LCPI3_2(%rip), %xmm2
; CHECK: align
; CHECK-NOT: LCPI
; CHECK: ret
@_minusZero.6007 = internal constant <4 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00> ; <<4 x float>*> [#uses=0]
@twoTo23.6008 = internal constant <4 x float> <float 8.388608e+06, float 8.388608e+06, float 8.388608e+06, float 8.388608e+06> ; <<4 x float>*> [#uses=0]
define void @vv(float* %y, float* %x, i32* %n) nounwind ssp {
entry:
br label %bb60
bb: ; preds = %bb60
Add dominance check for the instruction being hoisted. For example, MachineLICM should not hoist a load that is not guaranteed to be executed. Radar 10254254. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@141689 91177308-0d34-0410-b5e6-96231b3b80d8
2011-10-11 18:09:58 +00:00
%i.0 = phi i32 [ 0, %bb60 ] ; <i32> [#uses=2]
Revert r141569 and r141576. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@141594 91177308-0d34-0410-b5e6-96231b3b80d8
2011-10-10 23:18:02 +00:00
%0 = bitcast float* %x_addr.0 to <4 x float>* ; <<4 x float>*> [#uses=1]
%1 = load <4 x float>* %0, align 16 ; <<4 x float>> [#uses=4]
%tmp20 = bitcast <4 x float> %1 to <4 x i32> ; <<4 x i32>> [#uses=1]
%tmp22 = and <4 x i32> %tmp20, <i32 2147483647, i32 2147483647, i32 2147483647, i32 2147483647> ; <<4 x i32>> [#uses=1]
%tmp23 = bitcast <4 x i32> %tmp22 to <4 x float> ; <<4 x float>> [#uses=1]
%tmp25 = bitcast <4 x float> %1 to <4 x i32> ; <<4 x i32>> [#uses=1]
%tmp27 = and <4 x i32> %tmp25, <i32 -2147483648, i32 -2147483648, i32 -2147483648, i32 -2147483648> ; <<4 x i32>> [#uses=2]
%tmp30 = call <4 x float> @llvm.x86.sse.cmp.ps(<4 x float> %tmp23, <4 x float> <float 8.388608e+06, float 8.388608e+06, float 8.388608e+06, float 8.388608e+06>, i8 5) ; <<4 x float>> [#uses=1]
%tmp34 = bitcast <4 x float> %tmp30 to <4 x i32> ; <<4 x i32>> [#uses=1]
%tmp36 = xor <4 x i32> %tmp34, <i32 -1, i32 -1, i32 -1, i32 -1> ; <<4 x i32>> [#uses=1]
%tmp37 = and <4 x i32> %tmp36, <i32 1258291200, i32 1258291200, i32 1258291200, i32 1258291200> ; <<4 x i32>> [#uses=1]
%tmp42 = or <4 x i32> %tmp37, %tmp27 ; <<4 x i32>> [#uses=1]
%tmp43 = bitcast <4 x i32> %tmp42 to <4 x float> ; <<4 x float>> [#uses=2]
%tmp45 = fadd <4 x float> %1, %tmp43 ; <<4 x float>> [#uses=1]
%tmp47 = fsub <4 x float> %tmp45, %tmp43 ; <<4 x float>> [#uses=2]
%tmp49 = call <4 x float> @llvm.x86.sse.cmp.ps(<4 x float> %1, <4 x float> %tmp47, i8 1) ; <<4 x float>> [#uses=1]
%2 = bitcast <4 x float> %tmp49 to <4 x i32> ; <<4 x i32>> [#uses=1]
%3 = call <4 x float> @llvm.x86.sse2.cvtdq2ps(<4 x i32> %2) nounwind readnone ; <<4 x float>> [#uses=1]
%tmp53 = fadd <4 x float> %tmp47, %3 ; <<4 x float>> [#uses=1]
%tmp55 = bitcast <4 x float> %tmp53 to <4 x i32> ; <<4 x i32>> [#uses=1]
%tmp57 = or <4 x i32> %tmp55, %tmp27 ; <<4 x i32>> [#uses=1]
%tmp58 = bitcast <4 x i32> %tmp57 to <4 x float> ; <<4 x float>> [#uses=1]
%4 = bitcast float* %y_addr.0 to <4 x float>* ; <<4 x float>*> [#uses=1]
store <4 x float> %tmp58, <4 x float>* %4, align 16
%5 = getelementptr float* %x_addr.0, i64 4 ; <float*> [#uses=1]
%6 = getelementptr float* %y_addr.0, i64 4 ; <float*> [#uses=1]
%7 = add i32 %i.0, 4 ; <i32> [#uses=1]
Add dominance check for the instruction being hoisted. For example, MachineLICM should not hoist a load that is not guaranteed to be executed. Radar 10254254. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@141689 91177308-0d34-0410-b5e6-96231b3b80d8
2011-10-11 18:09:58 +00:00
%8 = load i32* %n, align 4 ; <i32> [#uses=1]
%9 = icmp sgt i32 %8, %7 ; <i1> [#uses=1]
br i1 %9, label %bb60, label %return
Revert r141569 and r141576. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@141594 91177308-0d34-0410-b5e6-96231b3b80d8
2011-10-10 23:18:02 +00:00
bb60: ; preds = %bb, %entry
%x_addr.0 = phi float* [ %x, %entry ], [ %5, %bb ] ; <float*> [#uses=2]
%y_addr.0 = phi float* [ %y, %entry ], [ %6, %bb ] ; <float*> [#uses=2]
Add dominance check for the instruction being hoisted. For example, MachineLICM should not hoist a load that is not guaranteed to be executed. Radar 10254254. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@141689 91177308-0d34-0410-b5e6-96231b3b80d8
2011-10-11 18:09:58 +00:00
br label %bb
Revert r141569 and r141576. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@141594 91177308-0d34-0410-b5e6-96231b3b80d8
2011-10-10 23:18:02 +00:00
return: ; preds = %bb60
ret void
}
Teach MachineLICM to unfold loads from constant memory from otherwise unhoistable instructions in order to allow the loads to be hoisted. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@85364 91177308-0d34-0410-b5e6-96231b3b80d8
2009-10-28 03:21:57 +00:00
declare <4 x float> @llvm.x86.sse.cmp.ps(<4 x float>, <4 x float>, i8) nounwind readnone
declare <4 x float> @llvm.x86.sse2.cvtdq2ps(<4 x i32>) nounwind readnone
Fix MachineLICM to use the correct virtual register class when unfolding loads for hoisting. getOpcodeAfterMemoryUnfold returns the opcode of the original operation without the load, not the load itself, MachineLICM needs to know the operand index in order to get the correct register class. Extend getOpcodeAfterMemoryUnfold to return this information. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@85622 91177308-0d34-0410-b5e6-96231b3b80d8
2009-10-30 22:18:41 +00:00
; CodeGen should use the correct register class when extracting
; a load from a zero-extending load for hoisting.
Mass update to CodeGen tests to use CHECK-LABEL for labels corresponding to function definitions for more informative error messages. No functionality change and all updated tests passed locally. This update was done with the following bash script: find test/CodeGen -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc.*debug" $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 sed -i '' "s/;\(.*\)-LABEL-LABEL:/;\1-LABEL:/" $TEMP sed -i '' "s/;\(.*\)-NEXT-LABEL:/;\1-NEXT:/" $TEMP sed -i '' "s/;\(.*\)-NOT-LABEL:/;\1-NOT:/" $TEMP sed -i '' "s/;\(.*\)-DAG-LABEL:/;\1-DAG:/" $TEMP mv $TEMP $NAME fi done git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@186280 91177308-0d34-0410-b5e6-96231b3b80d8
2013-07-14 06:24:09 +00:00
; CHECK-LABEL: default_get_pch_validity:
Fix MachineLICM to use the correct virtual register class when unfolding loads for hoisting. getOpcodeAfterMemoryUnfold returns the opcode of the original operation without the load, not the load itself, MachineLICM needs to know the operand index in order to get the correct register class. Extend getOpcodeAfterMemoryUnfold to return this information. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@85622 91177308-0d34-0410-b5e6-96231b3b80d8
2009-10-30 22:18:41 +00:00
; CHECK: movl cl_options_count(%rip), %ecx
@cl_options_count = external constant i32 ; <i32*> [#uses=2]
define void @default_get_pch_validity() nounwind {
entry:
%tmp4 = load i32* @cl_options_count, align 4 ; <i32> [#uses=1]
%tmp5 = icmp eq i32 %tmp4, 0 ; <i1> [#uses=1]
br i1 %tmp5, label %bb6, label %bb2
bb2: ; preds = %bb2, %entry
%i.019 = phi i64 [ 0, %entry ], [ %tmp25, %bb2 ] ; <i64> [#uses=1]
%tmp25 = add i64 %i.019, 1 ; <i64> [#uses=2]
%tmp11 = load i32* @cl_options_count, align 4 ; <i32> [#uses=1]
%tmp12 = zext i32 %tmp11 to i64 ; <i64> [#uses=1]
%tmp13 = icmp ugt i64 %tmp12, %tmp25 ; <i1> [#uses=1]
br i1 %tmp13, label %bb2, label %bb6
bb6: ; preds = %bb2, %entry
ret void
}
Reference in New Issue Copy Permalink