mirror of
https://github.com/RPCS3/llvm-mirror.git
synced 2024-10-29 23:12:55 +01:00
b401fd4c9e
Also avoid locals evicting locals just because they want a cheaper register. Problem: MI Sched knows exactly how many registers we have and assumes they can be colored. In cases where we have large blocks, usually from unrolled loops, greedy coloring fails. This is a source of "regressions" from the MI Scheduler on x86. I noticed this issue on x86 where we have long chains of two-address defs in the same live range. It's easy to see this in matrix multiplication benchmarks like IRSmk and even the unit test misched-matmul.ll. A fundamental difference between the LLVM register allocator and conventional graph coloring is that in our model a live range can't discover its neighbors, it can only verify its neighbors. That's why we initially went for greedy coloring and added eviction to deal with the hard cases. However, for singly defined and two-address live ranges, we can optimally color without visiting neighbors simply by processing the live ranges in instruction order. Other beneficial side effects: It is much easier to understand and debug regalloc for large blocks when the live ranges are allocated in order. Yes, global allocation is still very confusing, but it's nice to be able to comprehend what happened locally. Heuristics could be added to bias register assignment based on instruction locality (think late register pairing, banks...). Intuituvely this will make some test cases that are on the threshold of register pressure more stable. llvm-svn: 187139
59 lines
2.3 KiB
LLVM
59 lines
2.3 KiB
LLVM
; RUN: llc < %s -mtriple=x86_64-linux-gnu -mcpu=atom | FileCheck %s
|
|
; <rdar://problem/8006248>
|
|
|
|
; This randomly started passing after an unrelated change, if it fails again it
|
|
; might be worth looking at PR12324: misched bringup.
|
|
|
|
@llvm.used = appending global [1 x i8*] [i8* bitcast (void ([40 x i16]*, i32*, i16**, i64*)* @func to i8*)], section "llvm.metadata"
|
|
|
|
define void @func([40 x i16]* %a, i32* %b, i16** %c, i64* %d) nounwind {
|
|
entry:
|
|
%tmp103 = getelementptr inbounds [40 x i16]* %a, i64 0, i64 4
|
|
%tmp104 = load i16* %tmp103, align 2
|
|
%tmp105 = sext i16 %tmp104 to i32
|
|
%tmp106 = load i32* %b, align 4
|
|
%tmp107 = sub nsw i32 4, %tmp106
|
|
%tmp108 = load i16** %c, align 8
|
|
%tmp109 = sext i32 %tmp107 to i64
|
|
%tmp110 = getelementptr inbounds i16* %tmp108, i64 %tmp109
|
|
%tmp111 = load i16* %tmp110, align 1
|
|
%tmp112 = sext i16 %tmp111 to i32
|
|
%tmp = mul i32 355244649, %tmp112
|
|
%tmp1 = mul i32 %tmp, %tmp105
|
|
%tmp2 = add i32 %tmp1, 2138875574
|
|
%tmp3 = add i32 %tmp2, 1546991088
|
|
%tmp4 = mul i32 %tmp3, 2122487257
|
|
%tmp5 = icmp sge i32 %tmp4, 2138875574
|
|
%tmp6 = icmp slt i32 %tmp4, -8608074
|
|
%tmp7 = or i1 %tmp5, %tmp6
|
|
%outSign = select i1 %tmp7, i32 1, i32 -1
|
|
%tmp8 = icmp slt i32 %tmp4, 0
|
|
%tmp9 = icmp eq i32 %outSign, 1
|
|
%tmp10 = and i1 %tmp8, %tmp9
|
|
%tmp11 = sext i32 %tmp4 to i64
|
|
%tmp12 = add i64 %tmp11, 5089792279245435153
|
|
|
|
; CHECK: addl $2138875574, %e[[REGISTER_zext:[a-z0-9]+]]
|
|
; CHECK-NEXT: cmpl $-8608074, %e[[REGISTER_zext]]
|
|
; CHECK: movslq %e[[REGISTER_zext]], [[REGISTER_tmp:%r[a-z0-9]+]]
|
|
; CHECK: movq [[REGISTER_tmp]], [[REGISTER_sext:%r[a-z0-9]+]]
|
|
; CHECK-NOT: [[REGISTER_zext]]
|
|
; CHECK: subq %r[[REGISTER_zext]], [[REGISTER_sext]]
|
|
|
|
%tmp13 = sub i64 %tmp12, 2138875574
|
|
%tmp14 = zext i32 %tmp4 to i64
|
|
%tmp15 = sub i64 %tmp11, %tmp14
|
|
%tmp16 = select i1 %tmp10, i64 %tmp15, i64 0
|
|
%tmp17 = sub i64 %tmp13, %tmp16
|
|
%tmp18 = mul i64 %tmp17, 4540133155013554595
|
|
%tmp19 = sub i64 %tmp18, 5386586244038704851
|
|
%tmp20 = add i64 %tmp19, -1368057358110947217
|
|
%tmp21 = mul i64 %tmp20, -422037402840850817
|
|
%tmp115 = load i64* %d, align 8
|
|
%alphaX = mul i64 468858157810230901, %tmp21
|
|
%alphaXbetaY = add i64 %alphaX, %tmp115
|
|
%transformed = add i64 %alphaXbetaY, 9040145182981852475
|
|
store i64 %transformed, i64* %d, align 8
|
|
ret void
|
|
}
|