RPCS3/llvm-mirror
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-11-25 04:02:41 +01:00
Code Issues Projects Releases Wiki Activity

[X86][BMI] Add BMI2 stack folding tests

Browse Source 
llvm-svn: 295470
This commit is contained in:
Simon Pilgrim 2017-02-17 18:00:43 +00:00
parent 4974f41ade
commit fb59aba2a3
1 changed files with 77 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

77
test/CodeGen/X86/stack-folding-bmi2.ll Normal file
View File

@ -0,0 +1,77 @@
; RUN: llc -O3 -disable-peephole -mtriple=x86_64-unknown-unknown -mattr=+bmi,+bmi2 < %s | FileCheck %s
target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
target triple = "x86_64-unknown-unknown"
; Stack reload folding tests.
;
; By including a nop call with sideeffects we can force a partial register spill of the
; relevant registers and check that the reload is correctly folded into the instruction.
define i32 @stack_fold_bzhi_u32(i32 %a0, i32 %a1) {
;CHECK-LABEL: stack_fold_bzhi_u32
;CHECK: bzhil %eax, {{-?[0-9]*}}(%rsp), %eax {{.*#+}} 4-byte Folded Reload
%1 = tail call i32 asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
%2 = tail call i32 @llvm.x86.bmi.bzhi.32(i32 %a0, i32 %a1)
ret i32 %2
}
declare i32 @llvm.x86.bmi.bzhi.32(i32, i32)
define i64 @stack_fold_bzhi_u64(i64 %a0, i64 %a1) {
;CHECK-LABEL: stack_fold_bzhi_u64
;CHECK: bzhiq %rax, {{-?[0-9]*}}(%rsp), %rax {{.*#+}} 8-byte Folded Reload
%1 = tail call i64 asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
%2 = tail call i64 @llvm.x86.bmi.bzhi.64(i64 %a0, i64 %a1)
ret i64 %2
}
declare i64 @llvm.x86.bmi.bzhi.64(i64, i64)
define i64 @stack_fold_mulx_u64(i64 %a0, i64 %a1, i64 *%a2) {
;CHECK-LABEL: stack_fold_mulx_u64
;CHECK: mulxq {{-?[0-9]*}}(%rsp), %rax, %rcx {{.*#+}} 8-byte Folded Reload
%1 = tail call i64 asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
%2 = zext i64 %a0 to i128
%3 = zext i64 %a1 to i128
%4 = mul i128 %2, %3
%5 = lshr i128 %4, 64
%6 = trunc i128 %4 to i64
%7 = trunc i128 %5 to i64
store i64 %7, i64 *%a2
ret i64 %6
}
define i32 @stack_fold_pdep_u32(i32 %a0, i32 %a1) {
;CHECK-LABEL: stack_fold_pdep_u32
;CHECK: pdepl {{-?[0-9]*}}(%rsp), %eax, %eax {{.*#+}} 4-byte Folded Reload
%1 = tail call i32 asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
%2 = tail call i32 @llvm.x86.bmi.pdep.32(i32 %a0, i32 %a1)
ret i32 %2
}
declare i32 @llvm.x86.bmi.pdep.32(i32, i32)
define i64 @stack_fold_pdep_u64(i64 %a0, i64 %a1) {
;CHECK-LABEL: stack_fold_pdep_u64
;CHECK: pdepq {{-?[0-9]*}}(%rsp), %rax, %rax {{.*#+}} 8-byte Folded Reload
%1 = tail call i64 asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
%2 = tail call i64 @llvm.x86.bmi.pdep.64(i64 %a0, i64 %a1)
ret i64 %2
}
declare i64 @llvm.x86.bmi.pdep.64(i64, i64)
define i32 @stack_fold_pext_u32(i32 %a0, i32 %a1) {
;CHECK-LABEL: stack_fold_pext_u32
;CHECK: pextl {{-?[0-9]*}}(%rsp), %eax, %eax {{.*#+}} 4-byte Folded Reload
%1 = tail call i32 asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
%2 = tail call i32 @llvm.x86.bmi.pext.32(i32 %a0, i32 %a1)
ret i32 %2
}
declare i32 @llvm.x86.bmi.pext.32(i32, i32)
define i64 @stack_fold_pext_u64(i64 %a0, i64 %a1) {
;CHECK-LABEL: stack_fold_pext_u64
;CHECK: pextq {{-?[0-9]*}}(%rsp), %rax, %rax {{.*#+}} 8-byte Folded Reload
%1 = tail call i64 asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
%2 = tail call i64 @llvm.x86.bmi.pext.64(i64 %a0, i64 %a1)
ret i64 %2
}
declare i64 @llvm.x86.bmi.pext.64(i64, i64)