mirror of
https://github.com/c64scene-ar/llvm-6502.git
synced 2024-12-24 06:30:19 +00:00
9379557478
This patch just uses a peephole test for "add; compare; branch" sequences within a single block. The IR optimizers already convert loops to decrement-and-branch-on-nonzero form in some cases, so even this simplistic test triggers many times during a clang bootstrap and projects/test-suite run. It looks like there are still cases where we need to more strongly prefer branches on nonzero though. E.g. I saw a case where a loop that started out with a check for 0 ended up with a check for -1. I'll try to look at that sometime. I ended up adding the Reference class because MachineInstr::readsRegister() doesn't check for subregisters (by design, as far as I could tell). git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@187723 91177308-0d34-0410-b5e6-96231b3b80d8
69 lines
2.0 KiB
Python
69 lines
2.0 KiB
Python
# Test 32-bit BRANCH RELATIVE ON COUNT in cases where some branches are out
|
|
# of range.
|
|
# RUN: python %s | llc -mtriple=s390x-linux-gnu | FileCheck %s
|
|
|
|
# Construct:
|
|
#
|
|
# loopN:
|
|
# load of countN
|
|
# ...
|
|
# loop0:
|
|
# 0xffd8 bytes, from MVIY instructions
|
|
# conditional branch to main
|
|
# after0:
|
|
# ...
|
|
# decrement of countN
|
|
# conditional branch to loopN
|
|
# afterN:
|
|
#
|
|
# Each load occupies 4 bytes. Each decrement and branch occupies 4
|
|
# bytes if BRCT can be used, otherwise it occupies 10 bytes (AHI + BRCL).
|
|
# This means that loop 6 contains 5 * 4 + 0xffd8 + 5 * 4 == 0x10000 bytes
|
|
# and is therefore (just) in range. Loop 7 is out of range.
|
|
#
|
|
# CHECK: brct {{%r[0-9]+}}
|
|
# CHECK: brct {{%r[0-9]+}}
|
|
# CHECK: brct {{%r[0-9]+}}
|
|
# CHECK: brct {{%r[0-9]+}}
|
|
# CHECK: brct {{%r[0-9]+}}
|
|
# CHECK: brct {{%r[0-9]+}}
|
|
# CHECK: ahi {{%r[0-9]+}}, -1
|
|
# CHECK: jglh
|
|
# CHECK: ahi {{%r[0-9]+}}, -1
|
|
# CHECK: jglh
|
|
|
|
branch_blocks = 8
|
|
main_size = 0xffd8
|
|
|
|
print 'define void @f1(i8 *%base, i32 *%counts) {'
|
|
print 'entry:'
|
|
|
|
for i in xrange(branch_blocks - 1, -1, -1):
|
|
print ' %%countptr%d = getelementptr i32 *%%counts, i64 %d' % (i, i)
|
|
print ' %%initcount%d = load i32 *%%countptr%d' % (i, i)
|
|
print ' br label %%loop%d' % i
|
|
|
|
print 'loop%d:' % i
|
|
block1 = 'entry' if i == branch_blocks - 1 else 'loop%d' % (i + 1)
|
|
block2 = 'loop0' if i == 0 else 'after%d' % (i - 1)
|
|
print (' %%count%d = phi i32 [ %%initcount%d, %%%s ],'
|
|
' [ %%nextcount%d, %%%s ]' % (i, i, block1, i, block2))
|
|
|
|
a, b = 1, 1
|
|
for i in xrange(0, main_size, 6):
|
|
a, b = b, a + b
|
|
offset = 4096 + b % 500000
|
|
value = a % 256
|
|
print ' %%ptr%d = getelementptr i8 *%%base, i64 %d' % (i, offset)
|
|
print ' store volatile i8 %d, i8 *%%ptr%d' % (value, i)
|
|
|
|
for i in xrange(branch_blocks):
|
|
print ' %%nextcount%d = add i32 %%count%d, -1' % (i, i)
|
|
print ' %%test%d = icmp ne i32 %%nextcount%d, 0' % (i, i)
|
|
print ' br i1 %%test%d, label %%loop%d, label %%after%d' % (i, i, i)
|
|
print ''
|
|
print 'after%d:' % i
|
|
|
|
print ' ret void'
|
|
print '}'
|