Home | History | Annotate | Download | only in LoopUnroll
      1 ; In this test we check how heuristics for complete unrolling work. We have
      2 ; three knobs:
      3 ;  1) -unroll-threshold
      4 ;  3) -unroll-percent-dynamic-cost-saved-threshold and
      5 ;  2) -unroll-dynamic-cost-savings-discount
      6 ;
      7 ; They control loop-unrolling according to the following rules:
      8 ;  * If size of unrolled loop exceeds the absoulte threshold, we don't unroll
      9 ;    this loop under any circumstances.
     10 ;  * If size of unrolled loop is below the '-unroll-threshold', then we'll
     11 ;    consider this loop as a very small one, and completely unroll it.
     12 ;  * If a loop size is between these two tresholds, we only do complete unroll
     13 ;    it if estimated number of potentially optimized instructions is high (we
     14 ;    specify the minimal percent of such instructions).
     15 
     16 ; In this particular test-case, complete unrolling will allow later
     17 ; optimizations to remove ~55% of the instructions, the loop body size is 9,
     18 ; and unrolled size is 65.
     19 
     20 ; RUN: opt < %s -S -loop-unroll -unroll-max-iteration-count-to-analyze=1000 -unroll-threshold=10  -unroll-percent-dynamic-cost-saved-threshold=20 -unroll-dynamic-cost-savings-discount=0 | FileCheck %s -check-prefix=TEST1
     21 ; RUN: opt < %s -S -loop-unroll -unroll-max-iteration-count-to-analyze=1000 -unroll-threshold=10  -unroll-percent-dynamic-cost-saved-threshold=20 -unroll-dynamic-cost-savings-discount=90 | FileCheck %s -check-prefix=TEST2
     22 ; RUN: opt < %s -S -loop-unroll -unroll-max-iteration-count-to-analyze=1000 -unroll-threshold=10  -unroll-percent-dynamic-cost-saved-threshold=80 -unroll-dynamic-cost-savings-discount=90 | FileCheck %s -check-prefix=TEST3
     23 ; RUN: opt < %s -S -loop-unroll -unroll-max-iteration-count-to-analyze=1000 -unroll-threshold=100 -unroll-percent-dynamic-cost-saved-threshold=80 -unroll-dynamic-cost-savings-discount=0 | FileCheck %s -check-prefix=TEST4
     24 
     25 ; If the absolute threshold is too low, or if we can't optimize away requested
     26 ; percent of instructions, we shouldn't unroll:
     27 ; TEST1: %array_const_idx = getelementptr inbounds [9 x i32], [9 x i32]* @known_constant, i64 0, i64 %iv
     28 ; TEST3: %array_const_idx = getelementptr inbounds [9 x i32], [9 x i32]* @known_constant, i64 0, i64 %iv
     29 
     30 ; Otherwise, we should:
     31 ; TEST2-NOT: %array_const_idx = getelementptr inbounds [9 x i32], [9 x i32]* @known_constant, i64 0, i64 %iv
     32 
     33 ; Also, we should unroll if the 'unroll-threshold' is big enough:
     34 ; TEST4-NOT: %array_const_idx = getelementptr inbounds [9 x i32], [9 x i32]* @known_constant, i64 0, i64 %iv
     35 
     36 ; And check that we don't crash when we're not allowed to do any analysis.
     37 ; RUN: opt < %s -loop-unroll -unroll-max-iteration-count-to-analyze=0 -disable-output
     38 target datalayout = "e-m:o-i64:64-f80:128-n8:16:32:64-S128"
     39 
     40 @known_constant = internal unnamed_addr constant [9 x i32] [i32 0, i32 -1, i32 0, i32 -1, i32 5, i32 -1, i32 0, i32 -1, i32 0], align 16
     41 
     42 define i32 @foo(i32* noalias nocapture readonly %src) {
     43 entry:
     44   br label %loop
     45 
     46 loop:                                                ; preds = %loop, %entry
     47   %iv = phi i64 [ 0, %entry ], [ %inc, %loop ]
     48   %r  = phi i32 [ 0, %entry ], [ %add, %loop ]
     49   %arrayidx = getelementptr inbounds i32, i32* %src, i64 %iv
     50   %src_element = load i32, i32* %arrayidx, align 4
     51   %array_const_idx = getelementptr inbounds [9 x i32], [9 x i32]* @known_constant, i64 0, i64 %iv
     52   %const_array_element = load i32, i32* %array_const_idx, align 4
     53   %mul = mul nsw i32 %src_element, %const_array_element
     54   %add = add nsw i32 %mul, %r
     55   %inc = add nuw nsw i64 %iv, 1
     56   %exitcond86.i = icmp eq i64 %inc, 9
     57   br i1 %exitcond86.i, label %loop.end, label %loop
     58 
     59 loop.end:                                            ; preds = %loop
     60   %r.lcssa = phi i32 [ %r, %loop ]
     61   ret i32 %r.lcssa
     62 }
     63