mirror of
https://github.com/RPCS3/llvm-mirror.git
synced 2024-11-24 11:42:57 +01:00
bccf832eec
Summary: Related to https://bugs.llvm.org/show_bug.cgi?id=37793, https://reviews.llvm.org/D46760#1127287 We'd like to do this canonicalization https://rise4fun.com/Alive/Gmc But it is currently restricted by rL155136 / rL155362, which says: ``` // This is a constant shift of a constant shift. Be careful about hiding // shl instructions behind bit masks. They are used to represent multiplies // by a constant, and it is important that simple arithmetic expressions // are still recognizable by scalar evolution. // // The transforms applied to shl are very similar to the transforms applied // to mul by constant. We can be more aggressive about optimizing right // shifts. // // Combinations of right and left shifts will still be optimized in // DAGCombine where scalar evolution no longer applies. ``` I think these tests show that for *constants*, SCEV has no issues with that canonicalization. Reviewers: mkazantsev, spatel, efriedma, sanjoy Reviewed By: mkazantsev Subscribers: sanjoy, javed.absar, llvm-commits, stoklund, bixia Differential Revision: https://reviews.llvm.org/D48229 llvm-svn: 335101
94 lines
3.5 KiB
LLVM
94 lines
3.5 KiB
LLVM
; NOTE: Assertions have been autogenerated by utils/update_analyze_test_checks.py
|
|
; RUN: opt -S -analyze -scalar-evolution < %s | FileCheck %s
|
|
|
|
; These testcases aren't *identical* but they have the same/similar meaning.
|
|
|
|
; The obvious case.
|
|
define i32 @mul(i32 %val, i32 %num) nounwind {
|
|
; CHECK-LABEL: 'mul'
|
|
; CHECK-NEXT: Classifying expressions for: @mul
|
|
; CHECK-NEXT: %tmp1 = mul i32 %val, %num
|
|
; CHECK-NEXT: --> (%val * %num) U: full-set S: full-set
|
|
; CHECK-NEXT: %tmp2 = udiv i32 %tmp1, %num
|
|
; CHECK-NEXT: --> ((%val * %num) /u %num) U: full-set S: full-set
|
|
; CHECK-NEXT: Determining loop execution counts for: @mul
|
|
;
|
|
%tmp1 = mul i32 %val, %num
|
|
%tmp2 = udiv i32 %tmp1, %num
|
|
ret i32 %tmp2
|
|
}
|
|
|
|
; Or, it could be any number of equivalent patterns with mask:
|
|
; a) x & (1 << nbits) - 1
|
|
; b) x & ~(-1 << nbits)
|
|
; c) x & (-1 >> (32 - y))
|
|
; d) x << (32 - y) >> (32 - y)
|
|
|
|
define i32 @mask_a(i32 %val, i32 %numlowbits) nounwind {
|
|
; CHECK-LABEL: 'mask_a'
|
|
; CHECK-NEXT: Classifying expressions for: @mask_a
|
|
; CHECK-NEXT: %onebit = shl i32 1, %numlowbits
|
|
; CHECK-NEXT: --> %onebit U: full-set S: full-set
|
|
; CHECK-NEXT: %mask = add nsw i32 %onebit, -1
|
|
; CHECK-NEXT: --> (-1 + %onebit) U: full-set S: full-set
|
|
; CHECK-NEXT: %masked = and i32 %mask, %val
|
|
; CHECK-NEXT: --> %masked U: full-set S: full-set
|
|
; CHECK-NEXT: Determining loop execution counts for: @mask_a
|
|
;
|
|
%onebit = shl i32 1, %numlowbits
|
|
%mask = add nsw i32 %onebit, -1
|
|
%masked = and i32 %mask, %val
|
|
ret i32 %masked
|
|
}
|
|
|
|
define i32 @mask_b(i32 %val, i32 %numlowbits) nounwind {
|
|
; CHECK-LABEL: 'mask_b'
|
|
; CHECK-NEXT: Classifying expressions for: @mask_b
|
|
; CHECK-NEXT: %notmask = shl i32 -1, %numlowbits
|
|
; CHECK-NEXT: --> %notmask U: full-set S: full-set
|
|
; CHECK-NEXT: %mask = xor i32 %notmask, -1
|
|
; CHECK-NEXT: --> (-1 + (-1 * %notmask)) U: full-set S: full-set
|
|
; CHECK-NEXT: %masked = and i32 %mask, %val
|
|
; CHECK-NEXT: --> %masked U: full-set S: full-set
|
|
; CHECK-NEXT: Determining loop execution counts for: @mask_b
|
|
;
|
|
%notmask = shl i32 -1, %numlowbits
|
|
%mask = xor i32 %notmask, -1
|
|
%masked = and i32 %mask, %val
|
|
ret i32 %masked
|
|
}
|
|
|
|
define i32 @mask_c(i32 %val, i32 %numlowbits) nounwind {
|
|
; CHECK-LABEL: 'mask_c'
|
|
; CHECK-NEXT: Classifying expressions for: @mask_c
|
|
; CHECK-NEXT: %numhighbits = sub i32 32, %numlowbits
|
|
; CHECK-NEXT: --> (32 + (-1 * %numlowbits)) U: full-set S: full-set
|
|
; CHECK-NEXT: %mask = lshr i32 -1, %numhighbits
|
|
; CHECK-NEXT: --> %mask U: full-set S: full-set
|
|
; CHECK-NEXT: %masked = and i32 %mask, %val
|
|
; CHECK-NEXT: --> %masked U: full-set S: full-set
|
|
; CHECK-NEXT: Determining loop execution counts for: @mask_c
|
|
;
|
|
%numhighbits = sub i32 32, %numlowbits
|
|
%mask = lshr i32 -1, %numhighbits
|
|
%masked = and i32 %mask, %val
|
|
ret i32 %masked
|
|
}
|
|
|
|
define i32 @mask_d(i32 %val, i32 %numlowbits) nounwind {
|
|
; CHECK-LABEL: 'mask_d'
|
|
; CHECK-NEXT: Classifying expressions for: @mask_d
|
|
; CHECK-NEXT: %numhighbits = sub i32 32, %numlowbits
|
|
; CHECK-NEXT: --> (32 + (-1 * %numlowbits)) U: full-set S: full-set
|
|
; CHECK-NEXT: %highbitscleared = shl i32 %val, %numhighbits
|
|
; CHECK-NEXT: --> %highbitscleared U: full-set S: full-set
|
|
; CHECK-NEXT: %masked = lshr i32 %highbitscleared, %numhighbits
|
|
; CHECK-NEXT: --> %masked U: full-set S: full-set
|
|
; CHECK-NEXT: Determining loop execution counts for: @mask_d
|
|
;
|
|
%numhighbits = sub i32 32, %numlowbits
|
|
%highbitscleared = shl i32 %val, %numhighbits
|
|
%masked = lshr i32 %highbitscleared, %numhighbits
|
|
ret i32 %masked
|
|
}
|