[X86] ptest.ll - add test coverage for #144861 load chains

This commit is contained in:
Simon Pilgrim 2025-06-20 11:07:20 +01:00
parent 95bd05d7ca
commit a5fa5bd2a8

View File

@ -1,8 +1,8 @@
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+sse2,-avx | FileCheck %s --check-prefix=SSE2 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+sse2,-avx | FileCheck %s --check-prefixes=CHECK,SSE2
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+sse4.1,-avx | FileCheck %s --check-prefix=SSE41 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+sse4.1,-avx | FileCheck %s --check-prefixes=CHECK,SSE41
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx,-avx2 | FileCheck %s --check-prefixes=AVX,AVX1 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx,-avx2 | FileCheck %s --check-prefixes=CHECK,AVX,AVX1
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512vl,+avx512dq,+avx512bw | FileCheck %s --check-prefixes=AVX,AVX512 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512vl,+avx512dq,+avx512bw | FileCheck %s --check-prefixes=CHECK,AVX,AVX512
define i32 @veccond128(<4 x i32> %input) { define i32 @veccond128(<4 x i32> %input) {
; SSE2-LABEL: veccond128: ; SSE2-LABEL: veccond128:
@ -388,3 +388,163 @@ define i32 @vecsel512(<16 x i32> %input, i32 %a, i32 %b) {
%t2 = select i1 %t1, i32 %a, i32 %b %t2 = select i1 %t1, i32 %a, i32 %b
ret i32 %t2 ret i32 %t2
} }
define i1 @vecmp_load64x2(ptr %p0) {
; CHECK-LABEL: vecmp_load64x2:
; CHECK: # %bb.0:
; CHECK-NEXT: movq (%rdi), %rax
; CHECK-NEXT: orq 8(%rdi), %rax
; CHECK-NEXT: sete %al
; CHECK-NEXT: retq
%p1 = getelementptr i8, ptr %p0, i64 8
%i0 = load i64, ptr %p0, align 1
%i1 = load i64, ptr %p1, align 1
%or = or i64 %i0, %i1
%ne = icmp ne i64 %or, 0
%zx = zext i1 %ne to i32
%eq = icmp eq i32 %zx, 0
ret i1 %eq
}
define i1 @vecmp_load64x4(ptr %p0) {
; CHECK-LABEL: vecmp_load64x4:
; CHECK: # %bb.0:
; CHECK-NEXT: movq (%rdi), %rax
; CHECK-NEXT: movq 8(%rdi), %rcx
; CHECK-NEXT: orq 16(%rdi), %rax
; CHECK-NEXT: orq 24(%rdi), %rcx
; CHECK-NEXT: orq %rax, %rcx
; CHECK-NEXT: sete %al
; CHECK-NEXT: retq
%p1 = getelementptr i8, ptr %p0, i64 8
%p2 = getelementptr i8, ptr %p0, i64 16
%p3 = getelementptr i8, ptr %p0, i64 24
%i0 = load i64, ptr %p0, align 1
%i1 = load i64, ptr %p1, align 1
%i2 = load i64, ptr %p2, align 1
%i3 = load i64, ptr %p3, align 1
%or02 = or i64 %i0, %i2
%or13 = or i64 %i1, %i3
%or = or i64 %or02, %or13
%ne = icmp ne i64 %or, 0
%zx = zext i1 %ne to i32
%eq = icmp eq i32 %zx, 0
ret i1 %eq
}
define i1 @vecmp_load128x2(ptr %p0) {
; CHECK-LABEL: vecmp_load128x2:
; CHECK: # %bb.0:
; CHECK-NEXT: movq (%rdi), %rax
; CHECK-NEXT: movq 8(%rdi), %rcx
; CHECK-NEXT: orq 24(%rdi), %rcx
; CHECK-NEXT: orq 16(%rdi), %rax
; CHECK-NEXT: orq %rcx, %rax
; CHECK-NEXT: sete %al
; CHECK-NEXT: retq
%p1 = getelementptr i8, ptr %p0, i64 16
%i0 = load i128, ptr %p0, align 1
%i1 = load i128, ptr %p1, align 1
%or = or i128 %i0, %i1
%ne = icmp ne i128 %or, 0
%zx = zext i1 %ne to i32
%eq = icmp eq i32 %zx, 0
ret i1 %eq
}
define i1 @vecmp_load128x4(ptr %p0) {
; CHECK-LABEL: vecmp_load128x4:
; CHECK: # %bb.0:
; CHECK-NEXT: movq (%rdi), %rax
; CHECK-NEXT: movq 8(%rdi), %rcx
; CHECK-NEXT: movq 24(%rdi), %rdx
; CHECK-NEXT: movq 16(%rdi), %rsi
; CHECK-NEXT: orq 32(%rdi), %rax
; CHECK-NEXT: orq 40(%rdi), %rcx
; CHECK-NEXT: orq 48(%rdi), %rsi
; CHECK-NEXT: orq %rax, %rsi
; CHECK-NEXT: orq 56(%rdi), %rdx
; CHECK-NEXT: orq %rcx, %rdx
; CHECK-NEXT: orq %rsi, %rdx
; CHECK-NEXT: sete %al
; CHECK-NEXT: retq
%p1 = getelementptr i8, ptr %p0, i64 16
%p2 = getelementptr i8, ptr %p0, i64 32
%p3 = getelementptr i8, ptr %p0, i64 48
%i0 = load i128, ptr %p0, align 1
%i1 = load i128, ptr %p1, align 1
%i2 = load i128, ptr %p2, align 1
%i3 = load i128, ptr %p3, align 1
%or02 = or i128 %i0, %i2
%or13 = or i128 %i1, %i3
%or = or i128 %or02, %or13
%ne = icmp ne i128 %or, 0
%zx = zext i1 %ne to i32
%eq = icmp eq i32 %zx, 0
ret i1 %eq
}
; PR144861
define i1 @vecmp_load256x2(ptr %p0) {
; CHECK-LABEL: vecmp_load256x2:
; CHECK: # %bb.0:
; CHECK-NEXT: movq 24(%rdi), %rax
; CHECK-NEXT: movq (%rdi), %rcx
; CHECK-NEXT: movq 8(%rdi), %rdx
; CHECK-NEXT: movq 16(%rdi), %rsi
; CHECK-NEXT: orq 48(%rdi), %rsi
; CHECK-NEXT: orq 32(%rdi), %rcx
; CHECK-NEXT: orq %rsi, %rcx
; CHECK-NEXT: orq 56(%rdi), %rax
; CHECK-NEXT: orq 40(%rdi), %rdx
; CHECK-NEXT: orq %rax, %rdx
; CHECK-NEXT: orq %rcx, %rdx
; CHECK-NEXT: sete %al
; CHECK-NEXT: retq
%p1 = getelementptr i8, ptr %p0, i64 32
%i0 = load i256, ptr %p0, align 1
%i1 = load i256, ptr %p1, align 1
%or = or i256 %i0, %i1
%ne = icmp ne i256 %or, 0
%zx = zext i1 %ne to i32
%eq = icmp eq i32 %zx, 0
ret i1 %eq
}
define i1 @vecmp_load512x2(ptr %p0) {
; CHECK-LABEL: vecmp_load512x2:
; CHECK: # %bb.0:
; CHECK-NEXT: movq 24(%rdi), %rax
; CHECK-NEXT: movq 56(%rdi), %rdx
; CHECK-NEXT: movq 40(%rdi), %rsi
; CHECK-NEXT: movq 16(%rdi), %rcx
; CHECK-NEXT: movq 48(%rdi), %r8
; CHECK-NEXT: movq (%rdi), %r9
; CHECK-NEXT: movq 8(%rdi), %r10
; CHECK-NEXT: movq 32(%rdi), %r11
; CHECK-NEXT: orq 96(%rdi), %r11
; CHECK-NEXT: orq 64(%rdi), %r9
; CHECK-NEXT: orq %r11, %r9
; CHECK-NEXT: orq 112(%rdi), %r8
; CHECK-NEXT: orq 80(%rdi), %rcx
; CHECK-NEXT: orq %r8, %rcx
; CHECK-NEXT: orq %r9, %rcx
; CHECK-NEXT: orq 104(%rdi), %rsi
; CHECK-NEXT: orq 72(%rdi), %r10
; CHECK-NEXT: orq %rsi, %r10
; CHECK-NEXT: orq 120(%rdi), %rdx
; CHECK-NEXT: orq 88(%rdi), %rax
; CHECK-NEXT: orq %rdx, %rax
; CHECK-NEXT: orq %r10, %rax
; CHECK-NEXT: orq %rcx, %rax
; CHECK-NEXT: sete %al
; CHECK-NEXT: retq
%p1 = getelementptr i8, ptr %p0, i64 64
%i0 = load i512, ptr %p0, align 1
%i1 = load i512, ptr %p1, align 1
%or = or i512 %i0, %i1
%ne = icmp ne i512 %or, 0
%zx = zext i1 %ne to i32
%eq = icmp eq i32 %zx, 0
ret i1 %eq
}