diff options
Diffstat (limited to 'test')
| -rw-r--r-- | test/CodeGen/X86/iv-users-in-other-loops.ll | 9 | ||||
| -rw-r--r-- | test/CodeGen/X86/masked-iv-safe.ll | 7 | ||||
| -rw-r--r-- | test/CodeGen/X86/subreg-to-reg-5.ll | 7 | ||||
| -rw-r--r-- | test/Transforms/IndVarSimplify/2009-04-15-shorten-iv-vars-2.ll | 3 | ||||
| -rw-r--r-- | test/Transforms/IndVarSimplify/ada-loops.ll | 90 | ||||
| -rw-r--r-- | test/Transforms/IndVarSimplify/iv-zext.ll | 33 | ||||
| -rw-r--r-- | test/Transforms/IndVarSimplify/loop_evaluate_6.ll | 29 | ||||
| -rw-r--r-- | test/Transforms/LoopStrengthReduce/2009-04-28-no-reduce-mul.ll | 2 |
8 files changed, 166 insertions, 14 deletions
diff --git a/test/CodeGen/X86/iv-users-in-other-loops.ll b/test/CodeGen/X86/iv-users-in-other-loops.ll index d11b025..275feba 100644 --- a/test/CodeGen/X86/iv-users-in-other-loops.ll +++ b/test/CodeGen/X86/iv-users-in-other-loops.ll @@ -1,10 +1,11 @@ ; RUN: llvm-as < %s | llc -march=x86-64 -f -o %t ; RUN: grep inc %t | count 1 ; RUN: grep dec %t | count 2 -; RUN: grep addq %t | count 13 -; RUN: grep leaq %t | count 8 -; RUN: grep leal %t | count 4 -; RUN: grep movq %t | count 5 +; RUN: grep addq %t | count 8 +; RUN: grep addb %t | count 2 +; RUN: grep leaq %t | count 12 +; RUN: grep leal %t | count 2 +; RUN: grep movq %t | count 4 ; IV users in each of the loops from other loops shouldn't cause LSR ; to insert new induction variables. Previously it would create a diff --git a/test/CodeGen/X86/masked-iv-safe.ll b/test/CodeGen/X86/masked-iv-safe.ll index e9b80a4..e102535 100644 --- a/test/CodeGen/X86/masked-iv-safe.ll +++ b/test/CodeGen/X86/masked-iv-safe.ll @@ -3,14 +3,13 @@ ; RUN: not grep movz %t ; RUN: not grep sar %t ; RUN: not grep shl %t -; RUN: grep add %t | count 6 -; RUN: grep inc %t | count 2 -; RUN: grep dec %t | count 4 +; RUN: grep add %t | count 2 +; RUN: grep inc %t | count 4 +; RUN: grep dec %t | count 2 ; RUN: grep lea %t | count 2 ; Optimize away zext-inreg and sext-inreg on the loop induction ; variable using trip-count information. -; Also, the loop-reversal algorithm kicks in twice. define void @count_up(double* %d, i64 %n) nounwind { entry: diff --git a/test/CodeGen/X86/subreg-to-reg-5.ll b/test/CodeGen/X86/subreg-to-reg-5.ll index eee751a..81b262a 100644 --- a/test/CodeGen/X86/subreg-to-reg-5.ll +++ b/test/CodeGen/X86/subreg-to-reg-5.ll @@ -8,7 +8,8 @@ entry: bb2: ; preds = %bb3, %entry %B_addr.0.rec = phi i64 [ %indvar.next154, %bb3 ], [ 0, %entry ] ; <i64> [#uses=2] - br i1 false, label %bb3, label %bb4 + %z = icmp slt i64 %B_addr.0.rec, 20000 + br i1 %z, label %bb3, label %bb4 bb3: ; preds = %bb2 %indvar.next154 = add i64 %B_addr.0.rec, 1 ; <i64> [#uses=1] @@ -17,7 +18,7 @@ bb3: ; preds = %bb2 bb4: ; preds = %bb2 %B_addr.0 = getelementptr float* %B, i64 %B_addr.0.rec ; <float*> [#uses=1] %t1 = ptrtoint float* %B_addr.0 to i64 ; <i64> [#uses=1] - %t2 = and i64 %t1, 15 ; <i64> [#uses=1] + %t2 = and i64 %t1, 4294967295 ; <i64> [#uses=1] %t3 = icmp eq i64 %t2, 0 ; <i1> [#uses=1] br i1 %t3, label %bb5, label %bb10.preheader @@ -25,7 +26,7 @@ bb10.preheader: ; preds = %bb4 br label %bb9 bb5: ; preds = %bb4 - unreachable + ret float 7.0 bb9: ; preds = %bb10.preheader %t5 = getelementptr float* %B, i64 0 ; <float*> [#uses=1] diff --git a/test/Transforms/IndVarSimplify/2009-04-15-shorten-iv-vars-2.ll b/test/Transforms/IndVarSimplify/2009-04-15-shorten-iv-vars-2.ll index 5cc595e..4d26803 100644 --- a/test/Transforms/IndVarSimplify/2009-04-15-shorten-iv-vars-2.ll +++ b/test/Transforms/IndVarSimplify/2009-04-15-shorten-iv-vars-2.ll @@ -1,5 +1,4 @@ -; RUN: llvm-as < %s | opt -indvars | llvm-dis | not grep {sext} -; RUN: llvm-as < %s | opt -indvars | llvm-dis | not grep {zext} +; RUN: llvm-as < %s | opt -indvars -instcombine | llvm-dis | not grep {\[sz\]ext} ; ModuleID = '<stdin>' ;extern int *a, *b, *c, *d, *e, *f; /* 64 bit */ ;extern int K[256]; diff --git a/test/Transforms/IndVarSimplify/ada-loops.ll b/test/Transforms/IndVarSimplify/ada-loops.ll new file mode 100644 index 0000000..56325b3 --- /dev/null +++ b/test/Transforms/IndVarSimplify/ada-loops.ll @@ -0,0 +1,90 @@ +; RUN: llvm-as < %s | opt -indvars | llvm-dis > %t +; RUN: grep phi %t | count 4 +; RUN: grep {= phi i32} %t | count 4 +; RUN: not grep {sext i} %t +; RUN: not grep {zext i} %t +; RUN: not grep {trunc i} %t +; RUN: not grep {add i8} %t +; PR1301 + +; Do a bunch of analysis and prove that the loops can use an i32 trip +; count without casting. + +; ModuleID = 'ada.bc' +target datalayout = "e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-f32:32:32-f64:32:64-v64:64:64-v128:128:128-a0:0:64" +target triple = "i686-pc-linux-gnu" + +define void @kinds__sbytezero([256 x i32]* nocapture %a) nounwind { +bb.thread: + %tmp46 = getelementptr [256 x i32]* %a, i32 0, i32 0 ; <i32*> [#uses=1] + store i32 0, i32* %tmp46 + br label %bb + +bb: ; preds = %bb, %bb.thread + %i.0.reg2mem.0 = phi i8 [ -128, %bb.thread ], [ %tmp8, %bb ] ; <i8> [#uses=1] + %tmp8 = add i8 %i.0.reg2mem.0, 1 ; <i8> [#uses=3] + %tmp1 = sext i8 %tmp8 to i32 ; <i32> [#uses=1] + %tmp3 = add i32 %tmp1, 128 ; <i32> [#uses=1] + %tmp4 = getelementptr [256 x i32]* %a, i32 0, i32 %tmp3 ; <i32*> [#uses=1] + store i32 0, i32* %tmp4 + %0 = icmp eq i8 %tmp8, 127 ; <i1> [#uses=1] + br i1 %0, label %return, label %bb + +return: ; preds = %bb + ret void +} + +define void @kinds__ubytezero([256 x i32]* nocapture %a) nounwind { +bb.thread: + %tmp35 = getelementptr [256 x i32]* %a, i32 0, i32 0 ; <i32*> [#uses=1] + store i32 0, i32* %tmp35 + br label %bb + +bb: ; preds = %bb, %bb.thread + %i.0.reg2mem.0 = phi i8 [ 0, %bb.thread ], [ %tmp7, %bb ] ; <i8> [#uses=1] + %tmp7 = add i8 %i.0.reg2mem.0, 1 ; <i8> [#uses=3] + %tmp1 = zext i8 %tmp7 to i32 ; <i32> [#uses=1] + %tmp3 = getelementptr [256 x i32]* %a, i32 0, i32 %tmp1 ; <i32*> [#uses=1] + store i32 0, i32* %tmp3 + %0 = icmp eq i8 %tmp7, -1 ; <i1> [#uses=1] + br i1 %0, label %return, label %bb + +return: ; preds = %bb + ret void +} + +define void @kinds__srangezero([21 x i32]* nocapture %a) nounwind { +bb.thread: + br label %bb + +bb: ; preds = %bb, %bb.thread + %i.0.reg2mem.0 = phi i8 [ -10, %bb.thread ], [ %tmp7, %bb ] ; <i8> [#uses=2] + %tmp12 = sext i8 %i.0.reg2mem.0 to i32 ; <i32> [#uses=1] + %tmp4 = add i32 %tmp12, 10 ; <i32> [#uses=1] + %tmp5 = getelementptr [21 x i32]* %a, i32 0, i32 %tmp4 ; <i32*> [#uses=1] + store i32 0, i32* %tmp5 + %tmp7 = add i8 %i.0.reg2mem.0, 1 ; <i8> [#uses=2] + %0 = icmp sgt i8 %tmp7, 10 ; <i1> [#uses=1] + br i1 %0, label %return, label %bb + +return: ; preds = %bb + ret void +} + +define void @kinds__urangezero([21 x i32]* nocapture %a) nounwind { +bb.thread: + br label %bb + +bb: ; preds = %bb, %bb.thread + %i.0.reg2mem.0 = phi i8 [ 10, %bb.thread ], [ %tmp7, %bb ] ; <i8> [#uses=2] + %tmp12 = sext i8 %i.0.reg2mem.0 to i32 ; <i32> [#uses=1] + %tmp4 = add i32 %tmp12, -10 ; <i32> [#uses=1] + %tmp5 = getelementptr [21 x i32]* %a, i32 0, i32 %tmp4 ; <i32*> [#uses=1] + store i32 0, i32* %tmp5 + %tmp7 = add i8 %i.0.reg2mem.0, 1 ; <i8> [#uses=2] + %0 = icmp sgt i8 %tmp7, 30 ; <i1> [#uses=1] + br i1 %0, label %return, label %bb + +return: ; preds = %bb + ret void +} diff --git a/test/Transforms/IndVarSimplify/iv-zext.ll b/test/Transforms/IndVarSimplify/iv-zext.ll new file mode 100644 index 0000000..76d48de --- /dev/null +++ b/test/Transforms/IndVarSimplify/iv-zext.ll @@ -0,0 +1,33 @@ +; RUN: llvm-as < %s | opt -indvars | llvm-dis > %t +; RUN: not grep and %t +; RUN: not grep zext %t + +target datalayout = "-p:64:64:64" + +define void @foo(double* %d, i64 %n) nounwind { +entry: + br label %loop + +loop: + %indvar = phi i64 [ 0, %entry ], [ %indvar.next, %loop ] + %indvar.i8 = and i64 %indvar, 255 + %t0 = getelementptr double* %d, i64 %indvar.i8 + %t1 = load double* %t0 + %t2 = mul double %t1, 0.1 + store double %t2, double* %t0 + %indvar.i24 = and i64 %indvar, 16777215 + %t3 = getelementptr double* %d, i64 %indvar.i24 + %t4 = load double* %t3 + %t5 = mul double %t4, 2.3 + store double %t5, double* %t3 + %t6 = getelementptr double* %d, i64 %indvar + %t7 = load double* %t6 + %t8 = mul double %t7, 4.5 + store double %t8, double* %t6 + %indvar.next = add i64 %indvar, 1 + %exitcond = icmp eq i64 %indvar.next, 10 + br i1 %exitcond, label %return, label %loop + +return: + ret void +} diff --git a/test/Transforms/IndVarSimplify/loop_evaluate_6.ll b/test/Transforms/IndVarSimplify/loop_evaluate_6.ll new file mode 100644 index 0000000..35fbf52 --- /dev/null +++ b/test/Transforms/IndVarSimplify/loop_evaluate_6.ll @@ -0,0 +1,29 @@ +; RUN: llvm-as < %s | opt -indvars -loop-deletion | llvm-dis | grep phi | count 1 + +; Indvars should be able to evaluate this loop, allowing loop deletion +; to delete it. + +define i32 @test(i32 %x_offs) nounwind readnone { +entry: + %0 = icmp sgt i32 %x_offs, 4 ; <i1> [#uses=1] + br i1 %0, label %bb.nph, label %bb2 + +bb.nph: ; preds = %entry + br label %bb + +bb: ; preds = %bb1, %bb.nph + %x_offs_addr.01 = phi i32 [ %1, %bb1 ], [ %x_offs, %bb.nph ] ; <i32> [#uses=1] + %1 = add i32 %x_offs_addr.01, -4 ; <i32> [#uses=3] + br label %bb1 + +bb1: ; preds = %bb + %2 = icmp sgt i32 %1, 4 ; <i1> [#uses=1] + br i1 %2, label %bb, label %bb1.bb2_crit_edge + +bb1.bb2_crit_edge: ; preds = %bb1 + br label %bb2 + +bb2: ; preds = %bb1.bb2_crit_edge, %entry + %x_offs_addr.0.lcssa = phi i32 [ %1, %bb1.bb2_crit_edge ], [ %x_offs, %entry ] ; <i32> [#uses=1] + ret i32 %x_offs_addr.0.lcssa +} diff --git a/test/Transforms/LoopStrengthReduce/2009-04-28-no-reduce-mul.ll b/test/Transforms/LoopStrengthReduce/2009-04-28-no-reduce-mul.ll index 153a181..f873b3d 100644 --- a/test/Transforms/LoopStrengthReduce/2009-04-28-no-reduce-mul.ll +++ b/test/Transforms/LoopStrengthReduce/2009-04-28-no-reduce-mul.ll @@ -1,4 +1,4 @@ -; RUN: llvm-as < %s | opt -loop-reduce | llvm-dis | grep mul | count 3 +; RUN: llvm-as < %s | opt -loop-reduce | llvm-dis | grep {mul.*%lsr.iv} | count 2 ; The multiply in bb2 must not be reduced to an add, as the sext causes the ; %1 argument to become negative after a while. ; ModuleID = '<stdin>' |