diff options
author | Jakob Stoklund Olesen <stoklund@2pi.dk> | 2012-04-23 17:39:52 +0000 |
---|---|---|
committer | Jakob Stoklund Olesen <stoklund@2pi.dk> | 2012-04-23 17:39:52 +0000 |
commit | 72847f3057a7d82be087d3aad7d13b7719409a32 (patch) | |
tree | eb43c8ccca87249cc68ffbd0827d5ed5d30ce757 /test/Transforms/InstCombine/apint-shift.ll | |
parent | 9dc06bd431c826f068a5217718e2171303782f7a (diff) | |
download | llvm-72847f3057a7d82be087d3aad7d13b7719409a32.tar.gz llvm-72847f3057a7d82be087d3aad7d13b7719409a32.tar.bz2 llvm-72847f3057a7d82be087d3aad7d13b7719409a32.tar.xz |
Reapply r155136 after fixing PR12599.
Original commit message:
Defer some shl transforms to DAGCombine.
The shl instruction is used to represent multiplication by a constant
power of two as well as bitwise left shifts. Some InstCombine
transformations would turn an shl instruction into a bit mask operation,
making it difficult for later analysis passes to recognize the
constsnt multiplication.
Disable those shl transformations, deferring them to DAGCombine time.
An 'shl X, C' instruction is now treated mostly the same was as 'mul X, C'.
These transformations are deferred:
(X >>? C) << C --> X & (-1 << C) (When X >> C has multiple uses)
(X >>? C1) << C2 --> X << (C2-C1) & (-1 << C2) (When C2 > C1)
(X >>? C1) << C2 --> X >>? (C1-C2) & (-1 << C2) (When C1 > C2)
The corresponding exact transformations are preserved, just like
div-exact + mul:
(X >>?,exact C) << C --> X
(X >>?,exact C1) << C2 --> X << (C2-C1)
(X >>?,exact C1) << C2 --> X >>?,exact (C1-C2)
The disabled transformations could also prevent the instruction selector
from recognizing rotate patterns in hash functions and cryptographic
primitives. I have a test case for that, but it is too fragile.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@155362 91177308-0d34-0410-b5e6-96231b3b80d8
Diffstat (limited to 'test/Transforms/InstCombine/apint-shift.ll')
-rw-r--r-- | test/Transforms/InstCombine/apint-shift.ll | 16 |
1 files changed, 13 insertions, 3 deletions
diff --git a/test/Transforms/InstCombine/apint-shift.ll b/test/Transforms/InstCombine/apint-shift.ll index 0ea73a058c..73f630ebfe 100644 --- a/test/Transforms/InstCombine/apint-shift.ll +++ b/test/Transforms/InstCombine/apint-shift.ll @@ -47,13 +47,21 @@ define i32 @test5a(i32 %A) { } ; CHECK: @test6 -; CHECK-NOT: sh +; CHECK: mul i55 %A, 6 define i55 @test6(i55 %A) { %B = shl i55 %A, 1 ; <i55> [#uses=1] %C = mul i55 %B, 3 ; <i55> [#uses=1] ret i55 %C } +; CHECK: @test6a +; CHECK: mul i55 %A, 6 +define i55 @test6a(i55 %A) { + %B = mul i55 %A, 3 ; <i55> [#uses=1] + %C = shl i55 %B, 1 ; <i55> [#uses=1] + ret i55 %C +} + ; CHECK: @test7 ; CHECK-NOT: sh define i29 @test7(i8 %X) { @@ -87,7 +95,8 @@ define i19 @test10(i19 %A) { } ; CHECK: @test11 -; CHECK-NOT: sh +; Don't hide the shl from scalar evolution. DAGCombine will get it. +; CHECK: shl define i23 @test11(i23 %A) { %a = mul i23 %A, 3 ; <i23> [#uses=1] %B = lshr i23 %a, 11 ; <i23> [#uses=1] @@ -104,7 +113,8 @@ define i47 @test12(i47 %A) { } ; CHECK: @test13 -; CHECK-NOT: sh +; Don't hide the shl from scalar evolution. DAGCombine will get it. +; CHECK: shl define i18 @test13(i18 %A) { %a = mul i18 %A, 3 ; <i18> [#uses=1] %B = ashr i18 %a, 8 ; <i18> [#uses=1] |