mirror of
https://github.com/c64scene-ar/llvm-6502.git
synced 2025-02-10 20:33:15 +00:00
cc1e1707b8
Rewrite the shared implementation of BlockFrequencyInfo and
MachineBlockFrequencyInfo entirely.
The old implementation had a fundamental flaw: precision losses from
nested loops (or very wide branches) compounded past loop exits (and
convergence points).
The @nested_loops testcase at the end of
test/Analysis/BlockFrequencyAnalysis/basic.ll is motivating. This
function has three nested loops, with branch weights in the loop headers
of 1:4000 (exit:continue). The old analysis gives non-sensical results:
Printing analysis 'Block Frequency Analysis' for function 'nested_loops':
---- Block Freqs ----
entry = 1.0
for.cond1.preheader = 1.00103
for.cond4.preheader = 5.5222
for.body6 = 18095.19995
for.inc8 = 4.52264
for.inc11 = 0.00109
for.end13 = 0.0
The new analysis gives correct results:
Printing analysis 'Block Frequency Analysis' for function 'nested_loops':
block-frequency-info: nested_loops
- entry: float = 1.0, int = 8
- for.cond1.preheader: float = 4001.0, int = 32007
- for.cond4.preheader: float = 16008001.0, int = 128064007
- for.body6: float = 64048012001.0, int = 512384096007
- for.inc8: float = 16008001.0, int = 128064007
- for.inc11: float = 4001.0, int = 32007
- for.end13: float = 1.0, int = 8
Most importantly, the frequency leaving each loop matches the frequency
entering it.
The new algorithm leverages BlockMass and PositiveFloat to maintain
precision, separates "probability mass distribution" from "loop
scaling", and uses dithering to eliminate probability mass loss. I have
unit tests for these types out of tree, but it was decided in the review
to make the classes private to BlockFrequencyInfoImpl, and try to shrink
them (or remove them entirely) in follow-up commits.
The new algorithm should generally have a complexity advantage over the
old. The previous algorithm was quadratic in the worst case. The new
algorithm is still worst-case quadratic in the presence of irreducible
control flow, but it's linear without it.
The key difference between the old algorithm and the new is that control
flow within a loop is evaluated separately from control flow outside,
limiting propagation of precision problems and allowing loop scale to be
calculated independently of mass distribution. Loops are visited
bottom-up, their loop scales are calculated, and they are replaced by
pseudo-nodes. Mass is then distributed through the function, which is
now a DAG. Finally, loops are revisited top-down to multiply through
the loop scales and the masses distributed to pseudo nodes.
There are some remaining flaws.
- Irreducible control flow isn't modelled correctly. LoopInfo and
MachineLoopInfo ignore irreducible edges, so this algorithm will
fail to scale accordingly. There's a note in the class
documentation about how to get closer. See also the comments in
test/Analysis/BlockFrequencyInfo/irreducible.ll.
- Loop scale is limited to 4096 per loop (2^12) to avoid exhausting
the 64-bit integer precision used downstream.
- The "bias" calculation proposed on llvmdev is *not* incorporated
here. This will be added in a follow-up commit, once comments from
this review have been handled.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206548 91177308-0d34-0410-b5e6-96231b3b80d8
362 lines
8.7 KiB
LLVM
362 lines
8.7 KiB
LLVM
; RUN: llc < %s -march=xcore | FileCheck %s
|
|
; RUN: llc < %s -march=xcore -disable-fp-elim | FileCheck %s -check-prefix=CHECKFP
|
|
|
|
declare i8* @llvm.frameaddress(i32) nounwind readnone
|
|
declare i8* @llvm.returnaddress(i32) nounwind
|
|
declare i8* @llvm.eh.dwarf.cfa(i32) nounwind
|
|
declare void @llvm.eh.return.i32(i32, i8*) nounwind
|
|
declare void @llvm.eh.unwind.init() nounwind
|
|
|
|
define i8* @FA0() nounwind {
|
|
entry:
|
|
; CHECK-LABEL: FA0
|
|
; CHECK: ldaw r0, sp[0]
|
|
; CHECK-NEXT: retsp 0
|
|
%0 = call i8* @llvm.frameaddress(i32 0)
|
|
ret i8* %0
|
|
}
|
|
|
|
define i8* @FA1() nounwind {
|
|
entry:
|
|
; CHECK-LABEL: FA1
|
|
; CHECK: entsp 100
|
|
; CHECK-NEXT: ldaw r0, sp[0]
|
|
; CHECK-NEXT: retsp 100
|
|
%0 = alloca [100 x i32]
|
|
%1 = call i8* @llvm.frameaddress(i32 0)
|
|
ret i8* %1
|
|
}
|
|
|
|
define i8* @RA0() nounwind {
|
|
entry:
|
|
; CHECK-LABEL: RA0
|
|
; CHECK: stw lr, sp[0]
|
|
; CHECK-NEXT: ldw r0, sp[0]
|
|
; CHECK-NEXT: ldw lr, sp[0]
|
|
; CHECK-NEXT: retsp 0
|
|
%0 = call i8* @llvm.returnaddress(i32 0)
|
|
ret i8* %0
|
|
}
|
|
|
|
define i8* @RA1() nounwind {
|
|
entry:
|
|
; CHECK-LABEL: RA1
|
|
; CHECK: entsp 100
|
|
; CHECK-NEXT: ldw r0, sp[100]
|
|
; CHECK-NEXT: retsp 100
|
|
%0 = alloca [100 x i32]
|
|
%1 = call i8* @llvm.returnaddress(i32 0)
|
|
ret i8* %1
|
|
}
|
|
|
|
; test FRAME_TO_ARGS_OFFSET lowering
|
|
define i8* @FTAO0() nounwind {
|
|
entry:
|
|
; CHECK-LABEL: FTAO0
|
|
; CHECK: ldc r0, 0
|
|
; CHECK-NEXT: ldaw r1, sp[0]
|
|
; CHECK-NEXT: add r0, r1, r0
|
|
; CHECK-NEXT: retsp 0
|
|
%0 = call i8* @llvm.eh.dwarf.cfa(i32 0)
|
|
ret i8* %0
|
|
}
|
|
|
|
define i8* @FTAO1() nounwind {
|
|
entry:
|
|
; CHECK-LABEL: FTAO1
|
|
; CHECK: entsp 100
|
|
; CHECK-NEXT: ldc r0, 400
|
|
; CHECK-NEXT: ldaw r1, sp[0]
|
|
; CHECK-NEXT: add r0, r1, r0
|
|
; CHECK-NEXT: retsp 100
|
|
%0 = alloca [100 x i32]
|
|
%1 = call i8* @llvm.eh.dwarf.cfa(i32 0)
|
|
ret i8* %1
|
|
}
|
|
|
|
define i8* @EH0(i32 %offset, i8* %handler) {
|
|
entry:
|
|
; CHECK-LABEL: EH0
|
|
; CHECK: entsp 2
|
|
; CHECK: .cfi_def_cfa_offset 8
|
|
; CHECK: .cfi_offset 15, 0
|
|
; CHECK: .cfi_offset 1, -8
|
|
; CHECK: .cfi_offset 0, -4
|
|
; CHECK: ldc r2, 8
|
|
; CHECK-NEXT: ldaw r3, sp[0]
|
|
; CHECK-NEXT: add r2, r3, r2
|
|
; CHECK-NEXT: add r2, r2, r0
|
|
; CHECK-NEXT: mov r3, r1
|
|
; CHECK-NEXT: ldw r1, sp[0]
|
|
; CHECK-NEXT: ldw r0, sp[1]
|
|
; CHECK-NEXT: set sp, r2
|
|
; CHECK-NEXT: bau r3
|
|
call void @llvm.eh.return.i32(i32 %offset, i8* %handler)
|
|
unreachable
|
|
}
|
|
|
|
declare void @foo(...)
|
|
define i8* @EH1(i32 %offset, i8* %handler) {
|
|
entry:
|
|
; CHECK-LABEL: EH1
|
|
; CHECK: entsp 5
|
|
; CHECK: .cfi_def_cfa_offset 20
|
|
; CHECK: .cfi_offset 15, 0
|
|
; CHECK: .cfi_offset 1, -16
|
|
; CHECK: .cfi_offset 0, -12
|
|
; CHECK: stw r4, sp[4]
|
|
; CHECK: .cfi_offset 4, -4
|
|
; CHECK: stw r5, sp[3]
|
|
; CHECK: .cfi_offset 5, -8
|
|
; CHECK: mov r4, r1
|
|
; CHECK-NEXT: mov r5, r0
|
|
; CHECK-NEXT: bl foo
|
|
; CHECK-NEXT: ldc r0, 20
|
|
; CHECK-NEXT: ldaw r1, sp[0]
|
|
; CHECK-NEXT: add r0, r1, r0
|
|
; CHECK-NEXT: add r2, r0, r5
|
|
; CHECK-NEXT: mov r3, r4
|
|
; CHECK-NEXT: ldw r5, sp[3]
|
|
; CHECK-NEXT: ldw r4, sp[4]
|
|
; CHECK-NEXT: ldw r1, sp[1]
|
|
; CHECK-NEXT: ldw r0, sp[2]
|
|
; CHECK-NEXT: set sp, r2
|
|
; CHECK-NEXT: bau r3
|
|
call void (...)* @foo()
|
|
call void @llvm.eh.return.i32(i32 %offset, i8* %handler)
|
|
unreachable
|
|
}
|
|
|
|
@offset = external constant i32
|
|
@handler = external constant i8
|
|
define i8* @EH2(i32 %r0, i32 %r1, i32 %r2, i32 %r3) {
|
|
entry:
|
|
; CHECK-LABEL: EH2
|
|
; CHECK: entsp 3
|
|
; CHECK: bl foo
|
|
; CHECK-NEXT: ldw r0, dp[offset]
|
|
; CHECK-NEXT: ldc r1, 12
|
|
; CHECK-NEXT: ldaw r2, sp[0]
|
|
; CHECK-NEXT: add r1, r2, r1
|
|
; CHECK-NEXT: add r2, r1, r0
|
|
; CHECK-NEXT: ldaw r3, dp[handler]
|
|
; CHECK-NEXT: ldw r1, sp[1]
|
|
; CHECK-NEXT: ldw r0, sp[2]
|
|
; CHECK-NEXT: set sp, r2
|
|
; CHECK-NEXT: bau r3
|
|
call void (...)* @foo()
|
|
%0 = load i32* @offset
|
|
call void @llvm.eh.return.i32(i32 %0, i8* @handler)
|
|
unreachable
|
|
}
|
|
|
|
|
|
; FP: spill FP+SR+R0:1+R4:9 = entsp 2+2+6
|
|
; But we dont actually spill or restore R0:1
|
|
; CHECKFP-LABEL: Unwind0:
|
|
; CHECKFP: entsp 10
|
|
; CHECKFP: stw r10, sp[1]
|
|
; CHECKFP: ldaw r10, sp[0]
|
|
; CHECKFP: stw r4, r10[9]
|
|
; CHECKFP: stw r5, r10[8]
|
|
; CHECKFP: stw r6, r10[7]
|
|
; CHECKFP: stw r7, r10[6]
|
|
; CHECKFP: stw r8, r10[5]
|
|
; CHECKFP: stw r9, r10[4]
|
|
; CHECKFP: ldw r9, r10[4]
|
|
; CHECKFP: ldw r8, r10[5]
|
|
; CHECKFP: ldw r7, r10[6]
|
|
; CHECKFP: ldw r6, r10[7]
|
|
; CHECKFP: ldw r5, r10[8]
|
|
; CHECKFP: ldw r4, r10[9]
|
|
; CHECKFP: set sp, r10
|
|
; CHECKFP: ldw r10, sp[1]
|
|
; CHECKFP: retsp 10
|
|
;
|
|
; !FP: spill R0:1+R4:10 = entsp 2+7
|
|
; But we dont actually spill or restore R0:1
|
|
; CHECK-LABEL: Unwind0:
|
|
; CHECK: entsp 9
|
|
; CHECK: stw r4, sp[8]
|
|
; CHECK: stw r5, sp[7]
|
|
; CHECK: stw r6, sp[6]
|
|
; CHECK: stw r7, sp[5]
|
|
; CHECK: stw r8, sp[4]
|
|
; CHECK: stw r9, sp[3]
|
|
; CHECK: stw r10, sp[2]
|
|
; CHECK: ldw r10, sp[2]
|
|
; CHECK: ldw r9, sp[3]
|
|
; CHECK: ldw r8, sp[4]
|
|
; CHECK: ldw r7, sp[5]
|
|
; CHECK: ldw r6, sp[6]
|
|
; CHECK: ldw r5, sp[7]
|
|
; CHECK: ldw r4, sp[8]
|
|
; CHECK: retsp 9
|
|
define void @Unwind0() {
|
|
call void @llvm.eh.unwind.init()
|
|
ret void
|
|
}
|
|
|
|
|
|
; FP: spill FP+SR+R0:1+R4:9+LR = entsp 2+2+6 + extsp 1
|
|
; But we dont actually spill or restore R0:1
|
|
; CHECKFP-LABEL: Unwind1:
|
|
; CHECKFP: entsp 10
|
|
; CHECKFP: stw r10, sp[1]
|
|
; CHECKFP: ldaw r10, sp[0]
|
|
; CHECKFP: stw r4, r10[9]
|
|
; CHECKFP: stw r5, r10[8]
|
|
; CHECKFP: stw r6, r10[7]
|
|
; CHECKFP: stw r7, r10[6]
|
|
; CHECKFP: stw r8, r10[5]
|
|
; CHECKFP: stw r9, r10[4]
|
|
; CHECKFP: extsp 1
|
|
; CHECKFP: bl foo
|
|
; CHECKFP: ldaw sp, sp[1]
|
|
; CHECKFP: ldw r9, r10[4]
|
|
; CHECKFP: ldw r8, r10[5]
|
|
; CHECKFP: ldw r7, r10[6]
|
|
; CHECKFP: ldw r6, r10[7]
|
|
; CHECKFP: ldw r5, r10[8]
|
|
; CHECKFP: ldw r4, r10[9]
|
|
; CHECKFP: set sp, r10
|
|
; CHECKFP: ldw r10, sp[1]
|
|
; CHECKFP: retsp 10
|
|
;
|
|
; !FP: spill R0:1+R4:10+LR = entsp 2+7+1
|
|
; But we dont actually spill or restore R0:1
|
|
; CHECK-LABEL: Unwind1:
|
|
; CHECK: entsp 10
|
|
; CHECK: stw r4, sp[9]
|
|
; CHECK: stw r5, sp[8]
|
|
; CHECK: stw r6, sp[7]
|
|
; CHECK: stw r7, sp[6]
|
|
; CHECK: stw r8, sp[5]
|
|
; CHECK: stw r9, sp[4]
|
|
; CHECK: stw r10, sp[3]
|
|
; CHECK: bl foo
|
|
; CHECK: ldw r10, sp[3]
|
|
; CHECK: ldw r9, sp[4]
|
|
; CHECK: ldw r8, sp[5]
|
|
; CHECK: ldw r7, sp[6]
|
|
; CHECK: ldw r6, sp[7]
|
|
; CHECK: ldw r5, sp[8]
|
|
; CHECK: ldw r4, sp[9]
|
|
; CHECK: retsp 10
|
|
define void @Unwind1() {
|
|
call void (...)* @foo()
|
|
call void @llvm.eh.unwind.init()
|
|
ret void
|
|
}
|
|
|
|
; FP: spill FP+SR+R0:1+R4:9 = entsp 2+2+6
|
|
; We dont spill R0:1
|
|
; We only restore R0:1 during eh.return
|
|
; CHECKFP-LABEL: UnwindEH:
|
|
; CHECKFP: entsp 10
|
|
; CHECKFP: .cfi_def_cfa_offset 40
|
|
; CHECKFP: .cfi_offset 15, 0
|
|
; CHECKFP: stw r10, sp[1]
|
|
; CHECKFP: .cfi_offset 10, -36
|
|
; CHECKFP: ldaw r10, sp[0]
|
|
; CHECKFP: .cfi_def_cfa_register 10
|
|
; CHECKFP: .cfi_offset 1, -32
|
|
; CHECKFP: .cfi_offset 0, -28
|
|
; CHECKFP: stw r4, r10[9]
|
|
; CHECKFP: .cfi_offset 4, -4
|
|
; CHECKFP: stw r5, r10[8]
|
|
; CHECKFP: .cfi_offset 5, -8
|
|
; CHECKFP: stw r6, r10[7]
|
|
; CHECKFP: .cfi_offset 6, -12
|
|
; CHECKFP: stw r7, r10[6]
|
|
; CHECKFP: .cfi_offset 7, -16
|
|
; CHECKFP: stw r8, r10[5]
|
|
; CHECKFP: .cfi_offset 8, -20
|
|
; CHECKFP: stw r9, r10[4]
|
|
; CHECKFP: .cfi_offset 9, -24
|
|
; CHECKFP: bt r0, .LBB{{[0-9_]+}}
|
|
; CHECKFP: ldw r9, r10[4]
|
|
; CHECKFP-NEXT: ldw r8, r10[5]
|
|
; CHECKFP-NEXT: ldw r7, r10[6]
|
|
; CHECKFP-NEXT: ldw r6, r10[7]
|
|
; CHECKFP-NEXT: ldw r5, r10[8]
|
|
; CHECKFP-NEXT: ldw r4, r10[9]
|
|
; CHECKFP-NEXT: set sp, r10
|
|
; CHECKFP-NEXT: ldw r10, sp[1]
|
|
; CHECKFP-NEXT: retsp 10
|
|
; CHECKFP: .LBB{{[0-9_]+}}
|
|
; CHECKFP-NEXT: ldc r2, 40
|
|
; CHECKFP-NEXT: add r2, r10, r2
|
|
; CHECKFP-NEXT: add r2, r2, r0
|
|
; CHECKFP-NEXT: mov r3, r1
|
|
; CHECKFP-NEXT: ldw r9, r10[4]
|
|
; CHECKFP-NEXT: ldw r8, r10[5]
|
|
; CHECKFP-NEXT: ldw r7, r10[6]
|
|
; CHECKFP-NEXT: ldw r6, r10[7]
|
|
; CHECKFP-NEXT: ldw r5, r10[8]
|
|
; CHECKFP-NEXT: ldw r4, r10[9]
|
|
; CHECKFP-NEXT: ldw r1, sp[2]
|
|
; CHECKFP-NEXT: ldw r0, sp[3]
|
|
; CHECKFP-NEXT: set sp, r2
|
|
; CHECKFP-NEXT: bau r3
|
|
;
|
|
; !FP: spill R0:1+R4:10 = entsp 2+7
|
|
; We dont spill R0:1
|
|
; We only restore R0:1 during eh.return
|
|
; CHECK-LABEL: UnwindEH:
|
|
; CHECK: entsp 9
|
|
; CHECK: .cfi_def_cfa_offset 36
|
|
; CHECK: .cfi_offset 15, 0
|
|
; CHECK: .cfi_offset 1, -36
|
|
; CHECK: .cfi_offset 0, -32
|
|
; CHECK: stw r4, sp[8]
|
|
; CHECK: .cfi_offset 4, -4
|
|
; CHECK: stw r5, sp[7]
|
|
; CHECK: .cfi_offset 5, -8
|
|
; CHECK: stw r6, sp[6]
|
|
; CHECK: .cfi_offset 6, -12
|
|
; CHECK: stw r7, sp[5]
|
|
; CHECK: .cfi_offset 7, -16
|
|
; CHECK: stw r8, sp[4]
|
|
; CHECK: .cfi_offset 8, -20
|
|
; CHECK: stw r9, sp[3]
|
|
; CHECK: .cfi_offset 9, -24
|
|
; CHECK: stw r10, sp[2]
|
|
; CHECK: .cfi_offset 10, -28
|
|
; CHECK: bt r0, .LBB{{[0-9_]+}}
|
|
; CHECK: ldw r10, sp[2]
|
|
; CHECK-NEXT: ldw r9, sp[3]
|
|
; CHECK-NEXT: ldw r8, sp[4]
|
|
; CHECK-NEXT: ldw r7, sp[5]
|
|
; CHECK-NEXT: ldw r6, sp[6]
|
|
; CHECK-NEXT: ldw r5, sp[7]
|
|
; CHECK-NEXT: ldw r4, sp[8]
|
|
; CHECK-NEXT: retsp 9
|
|
; CHECK: .LBB{{[0-9_]+}}
|
|
; CHECK-NEXT: ldc r2, 36
|
|
; CHECK-NEXT: ldaw r3, sp[0]
|
|
; CHECK-NEXT: add r2, r3, r2
|
|
; CHECK-NEXT: add r2, r2, r0
|
|
; CHECK-NEXT: mov r3, r1
|
|
; CHECK-NEXT: ldw r10, sp[2]
|
|
; CHECK-NEXT: ldw r9, sp[3]
|
|
; CHECK-NEXT: ldw r8, sp[4]
|
|
; CHECK-NEXT: ldw r7, sp[5]
|
|
; CHECK-NEXT: ldw r6, sp[6]
|
|
; CHECK-NEXT: ldw r5, sp[7]
|
|
; CHECK-NEXT: ldw r4, sp[8]
|
|
; CHECK-NEXT: ldw r1, sp[0]
|
|
; CHECK-NEXT: ldw r0, sp[1]
|
|
; CHECK-NEXT: set sp, r2
|
|
; CHECK-NEXT: bau r3
|
|
define void @UnwindEH(i32 %offset, i8* %handler) {
|
|
call void @llvm.eh.unwind.init()
|
|
%cmp = icmp eq i32 %offset, 0
|
|
br i1 %cmp, label %normal, label %eh
|
|
eh:
|
|
call void @llvm.eh.return.i32(i32 %offset, i8* %handler)
|
|
unreachable
|
|
normal:
|
|
ret void
|
|
}
|