e6bf48d110
As of revea222be0d, LLVMs assembler will actually try to honour the "fill value" part of p2align directives. X86 printed these as 0x90, which isn't actually what it wanted: we want multi-byte nops for .text padding. Compiling via a textual assembly file produces single-byte nop padding sinceea222be0dbut the built-in assembler will produce multi-byte nops. This divergent behaviour is undesirable. To fix: don't set the byte padding field for x86, which allows the assembler to pick multi-byte nops. Test that we get the same multi-byte padding when compiled via textual assembly or directly to object file. Added same-align-bytes-with-llasm-llobj.ll to that effect, updated numerous other tests to not contain check-lines for the explicit padding.
57 lines
1.6 KiB
LLVM
57 lines
1.6 KiB
LLVM
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
|
|
; RUN: llc < %s -mtriple=x86_64-unknown | FileCheck %s
|
|
|
|
; With optimization at O2 we actually get the legalized function optimized
|
|
; away through legalization and stack coloring, but check that we do all of
|
|
; that here and don't crash during legalization.
|
|
|
|
; Original program:
|
|
; typedef enum { A, B, C, D } P;
|
|
; struct { P x[2]; } a;
|
|
|
|
; void fn2();
|
|
; void fn1() {
|
|
; int b;
|
|
; unsigned c;
|
|
; for (;; c++) {
|
|
; fn2();
|
|
; unsigned n;
|
|
; for (; c; c++) {
|
|
; b = a.x[c] == A || a.x[c] == B || a.x[c] == D;
|
|
; if (b) n++;
|
|
; }
|
|
; if (n)
|
|
; for (;;)
|
|
; ;
|
|
; }
|
|
; }
|
|
|
|
define void @fn1(i64 %a0) {
|
|
; CHECK-LABEL: fn1:
|
|
; CHECK: # %bb.0: # %for.cond
|
|
; CHECK-NEXT: .p2align 4
|
|
; CHECK-NEXT: .LBB0_1: # %vector.body
|
|
; CHECK-NEXT: # =>This Inner Loop Header: Depth=1
|
|
; CHECK-NEXT: cmpq $8, %rdi
|
|
; CHECK-NEXT: jne .LBB0_1
|
|
; CHECK-NEXT: # %bb.2: # %middle.block
|
|
; CHECK-NEXT: retq
|
|
for.cond:
|
|
br label %vector.body
|
|
|
|
vector.body: ; preds = %vector.body, %for.cond
|
|
%x42 = bitcast <4 x i4> zeroinitializer to i16
|
|
%x43 = icmp ne i16 %x42, 0
|
|
%x44 = select i1 %x43, i32 undef, i32 0
|
|
%x72 = bitcast <4 x i1> zeroinitializer to i4
|
|
%x73 = icmp ne i4 %x72, 0
|
|
%x74 = select i1 %x73, i32 %x44, i32 undef
|
|
%x84 = select i1 undef, i32 undef, i32 %x74
|
|
%x88 = icmp eq i64 %a0, 8
|
|
br i1 %x88, label %middle.block, label %vector.body
|
|
|
|
middle.block: ; preds = %vector.body
|
|
%0 = select i1 undef, i32 undef, i32 %x84
|
|
ret void
|
|
}
|