0b93e90971
This reverts commit 23a5073600.
Although this patch achieved better codegen in most cases, it is really
important to accurately describe the cost of instructions. So I revert it.
93 lines
3.4 KiB
LLVM
93 lines
3.4 KiB
LLVM
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
|
|
; RUN: llc -mtriple=riscv32 -verify-machineinstrs < %s | FileCheck %s
|
|
; RUN: llc -mtriple=riscv64 -verify-machineinstrs < %s | FileCheck %s
|
|
|
|
@x = global [6 x i32] [i32 0, i32 1, i32 2, i32 3, i32 4, i32 5], align 4
|
|
@check = global [6 x i32] [i32 0, i32 1, i32 2, i32 3, i32 4, i32 5], align 4
|
|
|
|
; This test case checks whether the base address of an array is repeatedly
|
|
; rematerialised within a unrolled loop.
|
|
define signext i32 @unroll_loop_cse() {
|
|
; CHECK-LABEL: unroll_loop_cse:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: lui a1, %hi(x)
|
|
; CHECK-NEXT: lw a3, %lo(x)(a1)
|
|
; CHECK-NEXT: lui a2, %hi(check)
|
|
; CHECK-NEXT: lw a4, %lo(check)(a2)
|
|
; CHECK-NEXT: li a0, 1
|
|
; CHECK-NEXT: bne a3, a4, .LBB0_6
|
|
; CHECK-NEXT: # %bb.1:
|
|
; CHECK-NEXT: addi a1, a1, %lo(x)
|
|
; CHECK-NEXT: lw a1, 4(a1)
|
|
; CHECK-NEXT: addi a2, a2, %lo(check)
|
|
; CHECK-NEXT: lw a2, 4(a2)
|
|
; CHECK-NEXT: bne a1, a2, .LBB0_6
|
|
; CHECK-NEXT: # %bb.2:
|
|
; CHECK-NEXT: lui a1, %hi(x)
|
|
; CHECK-NEXT: addi a1, a1, %lo(x)
|
|
; CHECK-NEXT: lw a3, 8(a1)
|
|
; CHECK-NEXT: lui a2, %hi(check)
|
|
; CHECK-NEXT: addi a2, a2, %lo(check)
|
|
; CHECK-NEXT: lw a4, 8(a2)
|
|
; CHECK-NEXT: bne a3, a4, .LBB0_6
|
|
; CHECK-NEXT: # %bb.3:
|
|
; CHECK-NEXT: lw a1, 12(a1)
|
|
; CHECK-NEXT: lw a2, 12(a2)
|
|
; CHECK-NEXT: bne a1, a2, .LBB0_6
|
|
; CHECK-NEXT: # %bb.4:
|
|
; CHECK-NEXT: lui a1, %hi(x)
|
|
; CHECK-NEXT: addi a1, a1, %lo(x)
|
|
; CHECK-NEXT: lw a3, 16(a1)
|
|
; CHECK-NEXT: lui a2, %hi(check)
|
|
; CHECK-NEXT: addi a2, a2, %lo(check)
|
|
; CHECK-NEXT: lw a4, 16(a2)
|
|
; CHECK-NEXT: bne a3, a4, .LBB0_6
|
|
; CHECK-NEXT: # %bb.5:
|
|
; CHECK-NEXT: lw a0, 20(a1)
|
|
; CHECK-NEXT: lw a1, 20(a2)
|
|
; CHECK-NEXT: xor a0, a0, a1
|
|
; CHECK-NEXT: snez a0, a0
|
|
; CHECK-NEXT: .LBB0_6:
|
|
; CHECK-NEXT: ret
|
|
%1 = load i32, i32* getelementptr inbounds ([6 x i32], [6 x i32]* @x, i64 0, i64 0), align 4
|
|
%2 = load i32, i32* getelementptr inbounds ([6 x i32], [6 x i32]* @check, i64 0, i64 0), align 4
|
|
%3 = icmp eq i32 %1, %2
|
|
br i1 %3, label %4, label %25
|
|
|
|
4:
|
|
%5 = load i32, i32* getelementptr inbounds ([6 x i32], [6 x i32]* @x, i64 0, i64 1), align 4
|
|
%6 = load i32, i32* getelementptr inbounds ([6 x i32], [6 x i32]* @check, i64 0, i64 1), align 4
|
|
%7 = icmp eq i32 %5, %6
|
|
br i1 %7, label %8, label %25
|
|
|
|
8:
|
|
%9 = load i32, i32* getelementptr inbounds ([6 x i32], [6 x i32]* @x, i64 0, i64 2), align 4
|
|
%10 = load i32, i32* getelementptr inbounds ([6 x i32], [6 x i32]* @check, i64 0, i64 2), align 4
|
|
%11 = icmp eq i32 %9, %10
|
|
br i1 %11, label %12, label %25
|
|
|
|
12:
|
|
%13 = load i32, i32* getelementptr inbounds ([6 x i32], [6 x i32]* @x, i64 0, i64 3), align 4
|
|
%14 = load i32, i32* getelementptr inbounds ([6 x i32], [6 x i32]* @check, i64 0, i64 3), align 4
|
|
%15 = icmp eq i32 %13, %14
|
|
br i1 %15, label %16, label %25
|
|
|
|
16:
|
|
%17 = load i32, i32* getelementptr inbounds ([6 x i32], [6 x i32]* @x, i64 0, i64 4), align 4
|
|
%18 = load i32, i32* getelementptr inbounds ([6 x i32], [6 x i32]* @check, i64 0, i64 4), align 4
|
|
%19 = icmp eq i32 %17, %18
|
|
br i1 %19, label %20, label %25
|
|
|
|
20:
|
|
%21 = load i32, i32* getelementptr inbounds ([6 x i32], [6 x i32]* @x, i64 0, i64 5), align 4
|
|
%22 = load i32, i32* getelementptr inbounds ([6 x i32], [6 x i32]* @check, i64 0, i64 5), align 4
|
|
%23 = icmp ne i32 %21, %22
|
|
%24 = zext i1 %23 to i32
|
|
br label %25
|
|
|
|
25:
|
|
%26 = phi i32 [ 1, %0 ], [ 1, %4 ], [ 1, %8 ], [ 1, %12 ], [ 1, %16 ], [ %24, %20 ]
|
|
ret i32 %26
|
|
}
|
|
|