[llvm-commits] [llvm] r73807 - in /llvm/trunk: lib/Analysis/ScalarEvolution.cpp test/Transforms/IndVarSimplify/ashr-tripcount.ll
Dan Gohman
gohman at apple.com
Fri Jun 19 17:35:32 PDT 2009
Author: djg
Date: Fri Jun 19 19:35:32 2009
New Revision: 73807
URL: http://llvm.org/viewvc/llvm-project?rev=73807&view=rev
Log:
Generalize isLoopGuardedByCond's checking to consider two
SCEVUnknowns with identical Instructions to be equal. This allows
it to analze cases such as the attached testcase, where the front-end
has cloned the loop controlling expression. Along with r73805, this
lets IndVarSimplify eliminate all the sign-extend casts in the
loop in the attached testcase.
Added:
llvm/trunk/test/Transforms/IndVarSimplify/ashr-tripcount.ll
Modified:
llvm/trunk/lib/Analysis/ScalarEvolution.cpp
Modified: llvm/trunk/lib/Analysis/ScalarEvolution.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Analysis/ScalarEvolution.cpp?rev=73807&r1=73806&r2=73807&view=diff
==============================================================================
--- llvm/trunk/lib/Analysis/ScalarEvolution.cpp (original)
+++ llvm/trunk/lib/Analysis/ScalarEvolution.cpp Fri Jun 19 19:35:32 2009
@@ -3655,6 +3655,29 @@
return 0;
}
+/// HasSameValue - SCEV structural equivalence is usually sufficient for
+/// testing whether two expressions are equal, however for the purposes of
+/// looking for a condition guarding a loop, it can be useful to be a little
+/// more general, since a front-end may have replicated the controlling
+/// expression.
+///
+static bool HasSameValue(const SCEVHandle &A, const SCEVHandle &B) {
+ // Quick check to see if they are the same SCEV.
+ if (A == B) return true;
+
+ // Otherwise, if they're both SCEVUnknown, it's possible that they hold
+ // two different instructions with the same value. Check for this case.
+ if (const SCEVUnknown *AU = dyn_cast<SCEVUnknown>(A))
+ if (const SCEVUnknown *BU = dyn_cast<SCEVUnknown>(B))
+ if (const Instruction *AI = dyn_cast<Instruction>(AU->getValue()))
+ if (const Instruction *BI = dyn_cast<Instruction>(BU->getValue()))
+ if (AI->isIdenticalTo(BI))
+ return true;
+
+ // Otherwise assume they may have a different value.
+ return false;
+}
+
/// isLoopGuardedByCond - Test whether entry to the loop is protected by
/// a conditional between LHS and RHS. This is used to help avoid max
/// expressions in loop trip counts.
@@ -3755,9 +3778,10 @@
SCEVHandle PreCondLHSSCEV = getSCEV(PreCondLHS);
SCEVHandle PreCondRHSSCEV = getSCEV(PreCondRHS);
- if ((LHS == PreCondLHSSCEV && RHS == PreCondRHSSCEV) ||
- (LHS == getNotSCEV(PreCondRHSSCEV) &&
- RHS == getNotSCEV(PreCondLHSSCEV)))
+ if ((HasSameValue(LHS, PreCondLHSSCEV) &&
+ HasSameValue(RHS, PreCondRHSSCEV)) ||
+ (HasSameValue(LHS, getNotSCEV(PreCondRHSSCEV)) &&
+ HasSameValue(RHS, getNotSCEV(PreCondLHSSCEV))))
return true;
}
Added: llvm/trunk/test/Transforms/IndVarSimplify/ashr-tripcount.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/IndVarSimplify/ashr-tripcount.ll?rev=73807&view=auto
==============================================================================
--- llvm/trunk/test/Transforms/IndVarSimplify/ashr-tripcount.ll (added)
+++ llvm/trunk/test/Transforms/IndVarSimplify/ashr-tripcount.ll Fri Jun 19 19:35:32 2009
@@ -0,0 +1,107 @@
+; RUN: llvm-as < %s | opt -indvars | llvm-dis > %t
+; RUN: grep sext %t | count 1
+
+; Indvars should be able to eliminate all of the sign extensions
+; inside the loop.
+
+target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128"
+ at pow_2_tab = external constant [0 x float] ; <[0 x float]*> [#uses=1]
+ at pow_2_025_tab = external constant [0 x float] ; <[0 x float]*> [#uses=1]
+ at i_pow_2_tab = external constant [0 x float] ; <[0 x float]*> [#uses=1]
+ at i_pow_2_025_tab = external constant [0 x float] ; <[0 x float]*> [#uses=1]
+
+define void @foo(i32 %gain, i32 %noOfLines, i32* %quaSpectrum, float* %iquaSpectrum, float* %pow4_3_tab_ptr) nounwind {
+entry:
+ %t0 = icmp slt i32 %gain, 0 ; <i1> [#uses=1]
+ br i1 %t0, label %bb1, label %bb2
+
+bb1: ; preds = %entry
+ %t1 = sub i32 0, %gain ; <i32> [#uses=1]
+ %t2 = sub i32 0, %gain ; <i32> [#uses=1]
+ br label %bb2
+
+bb2: ; preds = %bb1, %entry
+ %pow_2_tab.pn = phi [0 x float]* [ @i_pow_2_tab, %bb1 ], [ @pow_2_tab, %entry ] ; <[0 x float]*> [#uses=1]
+ %.pn3.in.in = phi i32 [ %t1, %bb1 ], [ %gain, %entry ] ; <i32> [#uses=1]
+ %pow_2_025_tab.pn = phi [0 x float]* [ @i_pow_2_025_tab, %bb1 ], [ @pow_2_025_tab, %entry ] ; <[0 x float]*> [#uses=1]
+ %.pn2.in.in = phi i32 [ %t2, %bb1 ], [ %gain, %entry ] ; <i32> [#uses=1]
+ %.pn3.in = ashr i32 %.pn3.in.in, 2 ; <i32> [#uses=1]
+ %.pn2.in = and i32 %.pn2.in.in, 3 ; <i32> [#uses=1]
+ %.pn3 = sext i32 %.pn3.in to i64 ; <i64> [#uses=1]
+ %.pn2 = zext i32 %.pn2.in to i64 ; <i64> [#uses=1]
+ %.pn.in = getelementptr [0 x float]* %pow_2_tab.pn, i64 0, i64 %.pn3 ; <float*> [#uses=1]
+ %.pn1.in = getelementptr [0 x float]* %pow_2_025_tab.pn, i64 0, i64 %.pn2 ; <float*> [#uses=1]
+ %.pn = load float* %.pn.in ; <float> [#uses=1]
+ %.pn1 = load float* %.pn1.in ; <float> [#uses=1]
+ %invQuantizer.0 = fmul float %.pn, %.pn1 ; <float> [#uses=4]
+ %t3 = ashr i32 %noOfLines, 2 ; <i32> [#uses=1]
+ %t4 = icmp sgt i32 %t3, 0 ; <i1> [#uses=1]
+ br i1 %t4, label %bb.nph, label %return
+
+bb.nph: ; preds = %bb2
+ %t5 = ashr i32 %noOfLines, 2 ; <i32> [#uses=1]
+ br label %bb3
+
+bb3: ; preds = %bb4, %bb.nph
+ %i.05 = phi i32 [ %t49, %bb4 ], [ 0, %bb.nph ] ; <i32> [#uses=9]
+ %k.04 = phi i32 [ %t48, %bb4 ], [ 0, %bb.nph ] ; <i32> [#uses=1]
+ %t6 = sext i32 %i.05 to i64 ; <i64> [#uses=1]
+ %t7 = getelementptr i32* %quaSpectrum, i64 %t6 ; <i32*> [#uses=1]
+ %t8 = load i32* %t7, align 4 ; <i32> [#uses=1]
+ %t9 = zext i32 %t8 to i64 ; <i64> [#uses=1]
+ %t10 = getelementptr float* %pow4_3_tab_ptr, i64 %t9 ; <float*> [#uses=1]
+ %t11 = load float* %t10, align 4 ; <float> [#uses=1]
+ %t12 = or i32 %i.05, 1 ; <i32> [#uses=1]
+ %t13 = sext i32 %t12 to i64 ; <i64> [#uses=1]
+ %t14 = getelementptr i32* %quaSpectrum, i64 %t13 ; <i32*> [#uses=1]
+ %t15 = load i32* %t14, align 4 ; <i32> [#uses=1]
+ %t16 = zext i32 %t15 to i64 ; <i64> [#uses=1]
+ %t17 = getelementptr float* %pow4_3_tab_ptr, i64 %t16 ; <float*> [#uses=1]
+ %t18 = load float* %t17, align 4 ; <float> [#uses=1]
+ %t19 = or i32 %i.05, 2 ; <i32> [#uses=1]
+ %t20 = sext i32 %t19 to i64 ; <i64> [#uses=1]
+ %t21 = getelementptr i32* %quaSpectrum, i64 %t20 ; <i32*> [#uses=1]
+ %t22 = load i32* %t21, align 4 ; <i32> [#uses=1]
+ %t23 = zext i32 %t22 to i64 ; <i64> [#uses=1]
+ %t24 = getelementptr float* %pow4_3_tab_ptr, i64 %t23 ; <float*> [#uses=1]
+ %t25 = load float* %t24, align 4 ; <float> [#uses=1]
+ %t26 = or i32 %i.05, 3 ; <i32> [#uses=1]
+ %t27 = sext i32 %t26 to i64 ; <i64> [#uses=1]
+ %t28 = getelementptr i32* %quaSpectrum, i64 %t27 ; <i32*> [#uses=1]
+ %t29 = load i32* %t28, align 4 ; <i32> [#uses=1]
+ %t30 = zext i32 %t29 to i64 ; <i64> [#uses=1]
+ %t31 = getelementptr float* %pow4_3_tab_ptr, i64 %t30 ; <float*> [#uses=1]
+ %t32 = load float* %t31, align 4 ; <float> [#uses=1]
+ %t33 = fmul float %t11, %invQuantizer.0 ; <float> [#uses=1]
+ %t34 = sext i32 %i.05 to i64 ; <i64> [#uses=1]
+ %t35 = getelementptr float* %iquaSpectrum, i64 %t34 ; <float*> [#uses=1]
+ store float %t33, float* %t35, align 4
+ %t36 = or i32 %i.05, 1 ; <i32> [#uses=1]
+ %t37 = fmul float %t18, %invQuantizer.0 ; <float> [#uses=1]
+ %t38 = sext i32 %t36 to i64 ; <i64> [#uses=1]
+ %t39 = getelementptr float* %iquaSpectrum, i64 %t38 ; <float*> [#uses=1]
+ store float %t37, float* %t39, align 4
+ %t40 = or i32 %i.05, 2 ; <i32> [#uses=1]
+ %t41 = fmul float %t25, %invQuantizer.0 ; <float> [#uses=1]
+ %t42 = sext i32 %t40 to i64 ; <i64> [#uses=1]
+ %t43 = getelementptr float* %iquaSpectrum, i64 %t42 ; <float*> [#uses=1]
+ store float %t41, float* %t43, align 4
+ %t44 = or i32 %i.05, 3 ; <i32> [#uses=1]
+ %t45 = fmul float %t32, %invQuantizer.0 ; <float> [#uses=1]
+ %t46 = sext i32 %t44 to i64 ; <i64> [#uses=1]
+ %t47 = getelementptr float* %iquaSpectrum, i64 %t46 ; <float*> [#uses=1]
+ store float %t45, float* %t47, align 4
+ %t48 = add i32 %k.04, 1 ; <i32> [#uses=2]
+ %t49 = add i32 %i.05, 4 ; <i32> [#uses=1]
+ br label %bb4
+
+bb4: ; preds = %bb3
+ %t50 = icmp sgt i32 %t5, %t48 ; <i1> [#uses=1]
+ br i1 %t50, label %bb3, label %bb4.return_crit_edge
+
+bb4.return_crit_edge: ; preds = %bb4
+ br label %return
+
+return: ; preds = %bb4.return_crit_edge, %bb2
+ ret void
+}
More information about the llvm-commits
mailing list