1; RUN: llc -O3 -disable-peephole -mtriple=x86_64-unknown-unknown -mattr=+bmi < %s | FileCheck %s 2 3target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128" 4target triple = "x86_64-unknown-unknown" 5 6; Stack reload folding tests. 7; 8; By including a nop call with sideeffects we can force a partial register spill of the 9; relevant registers and check that the reload is correctly folded into the instruction. 10 11define i32 @stack_fold_andn_u32(i32 %a0, i32 %a1) { 12 ;CHECK-LABEL: stack_fold_andn_u32 13 ;CHECK: andnl {{-?[0-9]*}}(%rsp), %eax, %eax {{.*#+}} 4-byte Folded Reload 14 %1 = tail call i64 asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"() 15 %2 = xor i32 %a0, -1 16 %3 = and i32 %a1, %2 17 ret i32 %3 18} 19 20define i64 @stack_fold_andn_u64(i64 %a0, i64 %a1) { 21 ;CHECK-LABEL: stack_fold_andn_u64 22 ;CHECK: andnq {{-?[0-9]*}}(%rsp), %rax, %rax {{.*#+}} 8-byte Folded Reload 23 %1 = tail call i64 asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"() 24 %2 = xor i64 %a0, -1 25 %3 = and i64 %a1, %2 26 ret i64 %3 27} 28 29define i32 @stack_fold_bextr_u32(i32 %a0, i32 %a1) { 30 ;CHECK-LABEL: stack_fold_bextr_u32 31 ;CHECK: # %bb.0: 32 ;CHECK: bextrl %eax, {{-?[0-9]*}}(%rsp), %eax {{.*#+}} 4-byte Folded Reload 33 %1 = tail call i64 asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"() 34 %2 = tail call i32 @llvm.x86.bmi.bextr.32(i32 %a0, i32 %a1) 35 ret i32 %2 36} 37declare i32 @llvm.x86.bmi.bextr.32(i32, i32) 38 39define i64 @stack_fold_bextr_u64(i64 %a0, i64 %a1) { 40 ;CHECK-LABEL: stack_fold_bextr_u64 41 ;CHECK: # %bb.0: 42 ;CHECK: bextrq %rax, {{-?[0-9]*}}(%rsp), %rax {{.*#+}} 8-byte Folded Reload 43 %1 = tail call i64 asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"() 44 %2 = tail call i64 @llvm.x86.bmi.bextr.64(i64 %a0, i64 %a1) 45 ret i64 %2 46} 47declare i64 @llvm.x86.bmi.bextr.64(i64, i64) 48 49define i32 @stack_fold_blsi_u32(i32 %a0) { 50 ;CHECK-LABEL: stack_fold_blsi_u32 51 ;CHECK: blsil {{-?[0-9]*}}(%rsp), %eax {{.*#+}} 4-byte Folded Reload 52 %1 = tail call i64 asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"() 53 %2 = sub i32 0, %a0 54 %3 = and i32 %2, %a0 55 ret i32 %3 56} 57 58define i64 @stack_fold_blsi_u64(i64 %a0) { 59 ;CHECK-LABEL: stack_fold_blsi_u64 60 ;CHECK: blsiq {{-?[0-9]*}}(%rsp), %rax {{.*#+}} 8-byte Folded Reload 61 %1 = tail call i64 asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"() 62 %2 = sub i64 0, %a0 63 %3 = and i64 %2, %a0 64 ret i64 %3 65} 66 67define i32 @stack_fold_blsmsk_u32(i32 %a0) { 68 ;CHECK-LABEL: stack_fold_blsmsk_u32 69 ;CHECK: blsmskl {{-?[0-9]*}}(%rsp), %eax {{.*#+}} 4-byte Folded Reload 70 %1 = tail call i64 asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"() 71 %2 = sub i32 %a0, 1 72 %3 = xor i32 %2, %a0 73 ret i32 %3 74} 75 76define i64 @stack_fold_blsmsk_u64(i64 %a0) { 77 ;CHECK-LABEL: stack_fold_blsmsk_u64 78 ;CHECK: blsmskq {{-?[0-9]*}}(%rsp), %rax {{.*#+}} 8-byte Folded Reload 79 %1 = tail call i64 asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"() 80 %2 = sub i64 %a0, 1 81 %3 = xor i64 %2, %a0 82 ret i64 %3 83} 84 85define i32 @stack_fold_blsr_u32(i32 %a0) { 86 ;CHECK-LABEL: stack_fold_blsr_u32 87 ;CHECK: blsrl {{-?[0-9]*}}(%rsp), %eax {{.*#+}} 4-byte Folded Reload 88 %1 = tail call i64 asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"() 89 %2 = sub i32 %a0, 1 90 %3 = and i32 %2, %a0 91 ret i32 %3 92} 93 94define i64 @stack_fold_blsr_u64(i64 %a0) { 95 ;CHECK-LABEL: stack_fold_blsr_u64 96 ;CHECK: blsrq {{-?[0-9]*}}(%rsp), %rax {{.*#+}} 8-byte Folded Reload 97 %1 = tail call i64 asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"() 98 %2 = sub i64 %a0, 1 99 %3 = and i64 %2, %a0 100 ret i64 %3 101} 102 103;TODO stack_fold_tzcnt_u16 104 105define i32 @stack_fold_tzcnt_u32(i32 %a0) { 106 ;CHECK-LABEL: stack_fold_tzcnt_u32 107 ;CHECK: tzcntl {{-?[0-9]*}}(%rsp), %eax {{.*#+}} 4-byte Folded Reload 108 %1 = tail call i64 asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"() 109 %2 = tail call i32 @llvm.cttz.i32(i32 %a0, i1 0) 110 ret i32 %2 111} 112declare i32 @llvm.cttz.i32(i32, i1) 113 114define i64 @stack_fold_tzcnt_u64(i64 %a0) { 115 ;CHECK-LABEL: stack_fold_tzcnt_u64 116 ;CHECK: tzcntq {{-?[0-9]*}}(%rsp), %rax {{.*#+}} 8-byte Folded Reload 117 %1 = tail call i64 asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"() 118 %2 = tail call i64 @llvm.cttz.i64(i64 %a0, i1 0) 119 ret i64 %2 120} 121declare i64 @llvm.cttz.i64(i64, i1) 122