[llvm] 2843e74 - [X86][SSE] Cleanup tests based on feedback from D81547
Simon Pilgrim via llvm-commits
llvm-commits at lists.llvm.org
Wed Jun 10 08:48:39 PDT 2020
Author: Simon Pilgrim
Date: 2020-06-10T16:38:33+01:00
New Revision: 2843e7495b7b1a57c629e17c1a3866eba8fb3ff6
URL: https://github.com/llvm/llvm-project/commit/2843e7495b7b1a57c629e17c1a3866eba8fb3ff6
DIFF: https://github.com/llvm/llvm-project/commit/2843e7495b7b1a57c629e17c1a3866eba8fb3ff6.diff
LOG: [X86][SSE] Cleanup tests based on feedback from D81547
Simplify tests to return i1 bool, removing superfluous function calls.
Use nounwind to remove cfi
Added:
Modified:
llvm/test/CodeGen/X86/pr45378.ll
Removed:
################################################################################
diff --git a/llvm/test/CodeGen/X86/pr45378.ll b/llvm/test/CodeGen/X86/pr45378.ll
index f6dcf28bff0c..b33d4c0ab22a 100644
--- a/llvm/test/CodeGen/X86/pr45378.ll
+++ b/llvm/test/CodeGen/X86/pr45378.ll
@@ -7,157 +7,75 @@
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=avx512bw | FileCheck %s --check-prefixes=CHECK,AVX,AVX512,AVX512BW
declare i64 @llvm.experimental.vector.reduce.or.v2i64(<2 x i64>)
-declare void @TrapFunc(i64)
-define void @parseHeaders(i64 * %ptr) {
-; SSE2-LABEL: parseHeaders:
-; SSE2: # %bb.0:
-; SSE2-NEXT: pushq %rax
-; SSE2-NEXT: .cfi_def_cfa_offset 16
-; SSE2-NEXT: movdqu (%rdi), %xmm1
-; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm1[2,3,0,1]
-; SSE2-NEXT: por %xmm0, %xmm1
-; SSE2-NEXT: movq %xmm1, %rax
-; SSE2-NEXT: testq %rax, %rax
-; SSE2-NEXT: je .LBB0_2
-; SSE2-NEXT: # %bb.1: # %trap
-; SSE2-NEXT: movq %xmm0, %rdi
-; SSE2-NEXT: callq TrapFunc
-; SSE2-NEXT: .LBB0_2: # %ret
-; SSE2-NEXT: popq %rax
-; SSE2-NEXT: .cfi_def_cfa_offset 8
-; SSE2-NEXT: retq
-;
-; SSE41-LABEL: parseHeaders:
-; SSE41: # %bb.0:
-; SSE41-NEXT: pushq %rax
-; SSE41-NEXT: .cfi_def_cfa_offset 16
-; SSE41-NEXT: movdqu (%rdi), %xmm0
-; SSE41-NEXT: pshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; SSE41-NEXT: por %xmm0, %xmm1
-; SSE41-NEXT: movq %xmm1, %rax
-; SSE41-NEXT: testq %rax, %rax
-; SSE41-NEXT: je .LBB0_2
-; SSE41-NEXT: # %bb.1: # %trap
-; SSE41-NEXT: pextrq $1, %xmm0, %rdi
-; SSE41-NEXT: callq TrapFunc
-; SSE41-NEXT: .LBB0_2: # %ret
-; SSE41-NEXT: popq %rax
-; SSE41-NEXT: .cfi_def_cfa_offset 8
-; SSE41-NEXT: retq
+define i1 @parseHeaders(i64 * %ptr) nounwind {
+; SSE-LABEL: parseHeaders:
+; SSE: # %bb.0:
+; SSE-NEXT: movdqu (%rdi), %xmm0
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
+; SSE-NEXT: por %xmm0, %xmm1
+; SSE-NEXT: movq %xmm1, %rax
+; SSE-NEXT: testq %rax, %rax
+; SSE-NEXT: sete %al
+; SSE-NEXT: retq
;
; AVX1-LABEL: parseHeaders:
; AVX1: # %bb.0:
-; AVX1-NEXT: pushq %rax
-; AVX1-NEXT: .cfi_def_cfa_offset 16
; AVX1-NEXT: vmovdqu (%rdi), %xmm0
; AVX1-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; AVX1-NEXT: vpor %xmm1, %xmm0, %xmm1
-; AVX1-NEXT: vmovq %xmm1, %rax
+; AVX1-NEXT: vpor %xmm1, %xmm0, %xmm0
+; AVX1-NEXT: vmovq %xmm0, %rax
; AVX1-NEXT: testq %rax, %rax
-; AVX1-NEXT: je .LBB0_2
-; AVX1-NEXT: # %bb.1: # %trap
-; AVX1-NEXT: vpextrq $1, %xmm0, %rdi
-; AVX1-NEXT: callq TrapFunc
-; AVX1-NEXT: .LBB0_2: # %ret
-; AVX1-NEXT: popq %rax
-; AVX1-NEXT: .cfi_def_cfa_offset 8
+; AVX1-NEXT: sete %al
; AVX1-NEXT: retq
;
; AVX2-LABEL: parseHeaders:
; AVX2: # %bb.0:
-; AVX2-NEXT: pushq %rax
-; AVX2-NEXT: .cfi_def_cfa_offset 16
-; AVX2-NEXT: vmovdqu (%rdi), %xmm0
-; AVX2-NEXT: vpbroadcastq 8(%rdi), %xmm1
-; AVX2-NEXT: vpor %xmm1, %xmm0, %xmm1
-; AVX2-NEXT: vmovq %xmm1, %rax
+; AVX2-NEXT: vpbroadcastq 8(%rdi), %xmm0
+; AVX2-NEXT: vpor (%rdi), %xmm0, %xmm0
+; AVX2-NEXT: vmovq %xmm0, %rax
; AVX2-NEXT: testq %rax, %rax
-; AVX2-NEXT: je .LBB0_2
-; AVX2-NEXT: # %bb.1: # %trap
-; AVX2-NEXT: vpextrq $1, %xmm0, %rdi
-; AVX2-NEXT: callq TrapFunc
-; AVX2-NEXT: .LBB0_2: # %ret
-; AVX2-NEXT: popq %rax
-; AVX2-NEXT: .cfi_def_cfa_offset 8
+; AVX2-NEXT: sete %al
; AVX2-NEXT: retq
;
; AVX512-LABEL: parseHeaders:
; AVX512: # %bb.0:
-; AVX512-NEXT: pushq %rax
-; AVX512-NEXT: .cfi_def_cfa_offset 16
-; AVX512-NEXT: vmovdqu (%rdi), %xmm0
-; AVX512-NEXT: vpbroadcastq 8(%rdi), %xmm1
-; AVX512-NEXT: vpor %xmm1, %xmm0, %xmm1
-; AVX512-NEXT: vmovq %xmm1, %rax
+; AVX512-NEXT: vpbroadcastq 8(%rdi), %xmm0
+; AVX512-NEXT: vpor (%rdi), %xmm0, %xmm0
+; AVX512-NEXT: vmovq %xmm0, %rax
; AVX512-NEXT: testq %rax, %rax
-; AVX512-NEXT: je .LBB0_2
-; AVX512-NEXT: # %bb.1: # %trap
-; AVX512-NEXT: vpextrq $1, %xmm0, %rdi
-; AVX512-NEXT: callq TrapFunc
-; AVX512-NEXT: .LBB0_2: # %ret
-; AVX512-NEXT: popq %rax
-; AVX512-NEXT: .cfi_def_cfa_offset 8
+; AVX512-NEXT: sete %al
; AVX512-NEXT: retq
%vptr = bitcast i64 * %ptr to <2 x i64> *
%vload = load <2 x i64>, <2 x i64> * %vptr, align 8
%vreduce = call i64 @llvm.experimental.vector.reduce.or.v2i64(<2 x i64> %vload)
%vcheck = icmp eq i64 %vreduce, 0
- br i1 %vcheck, label %ret, label %trap
-trap:
- %v2 = extractelement <2 x i64> %vload, i32 1
- call void @TrapFunc(i64 %v2)
- ret void
-ret:
- ret void
+ ret i1 %vcheck
}
-define void @parseHeaders2_scalar_or(i64 * %ptr) {
+define i1 @parseHeaders2_scalar_or(i64 * %ptr) nounwind {
; SSE2-LABEL: parseHeaders2_scalar_or:
; SSE2: # %bb.0:
-; SSE2-NEXT: pushq %rax
-; SSE2-NEXT: .cfi_def_cfa_offset 16
; SSE2-NEXT: movdqu (%rdi), %xmm0
; SSE2-NEXT: movq %xmm0, %rax
; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,3,0,1]
-; SSE2-NEXT: movq %xmm0, %rdi
-; SSE2-NEXT: orq %rdi, %rax
-; SSE2-NEXT: je .LBB1_2
-; SSE2-NEXT: # %bb.1: # %trap
-; SSE2-NEXT: callq TrapFunc
-; SSE2-NEXT: .LBB1_2: # %ret
-; SSE2-NEXT: popq %rax
-; SSE2-NEXT: .cfi_def_cfa_offset 8
+; SSE2-NEXT: movq %xmm0, %rcx
+; SSE2-NEXT: orq %rax, %rcx
+; SSE2-NEXT: sete %al
; SSE2-NEXT: retq
;
; SSE41-LABEL: parseHeaders2_scalar_or:
; SSE41: # %bb.0:
-; SSE41-NEXT: pushq %rax
-; SSE41-NEXT: .cfi_def_cfa_offset 16
; SSE41-NEXT: movdqu (%rdi), %xmm0
; SSE41-NEXT: ptest %xmm0, %xmm0
-; SSE41-NEXT: je .LBB1_2
-; SSE41-NEXT: # %bb.1: # %trap
-; SSE41-NEXT: pextrq $1, %xmm0, %rdi
-; SSE41-NEXT: callq TrapFunc
-; SSE41-NEXT: .LBB1_2: # %ret
-; SSE41-NEXT: popq %rax
-; SSE41-NEXT: .cfi_def_cfa_offset 8
+; SSE41-NEXT: sete %al
; SSE41-NEXT: retq
;
; AVX-LABEL: parseHeaders2_scalar_or:
; AVX: # %bb.0:
-; AVX-NEXT: pushq %rax
-; AVX-NEXT: .cfi_def_cfa_offset 16
; AVX-NEXT: vmovdqu (%rdi), %xmm0
; AVX-NEXT: vptest %xmm0, %xmm0
-; AVX-NEXT: je .LBB1_2
-; AVX-NEXT: # %bb.1: # %trap
-; AVX-NEXT: vpextrq $1, %xmm0, %rdi
-; AVX-NEXT: callq TrapFunc
-; AVX-NEXT: .LBB1_2: # %ret
-; AVX-NEXT: popq %rax
-; AVX-NEXT: .cfi_def_cfa_offset 8
+; AVX-NEXT: sete %al
; AVX-NEXT: retq
%vptr = bitcast i64 * %ptr to <2 x i64> *
%vload = load <2 x i64>, <2 x i64> * %vptr, align 8
@@ -165,62 +83,36 @@ define void @parseHeaders2_scalar_or(i64 * %ptr) {
%v2 = extractelement <2 x i64> %vload, i32 1
%vreduce = or i64 %v1, %v2
%vcheck = icmp eq i64 %vreduce, 0
- br i1 %vcheck, label %ret, label %trap
-trap:
- call void @TrapFunc(i64 %v2)
- ret void
-ret:
- ret void
+ ret i1 %vcheck
}
-define void @parseHeaders2_scalar_and(i64 * %ptr) {
+define i1 @parseHeaders2_scalar_and(i64 * %ptr) nounwind {
; SSE2-LABEL: parseHeaders2_scalar_and:
; SSE2: # %bb.0:
-; SSE2-NEXT: pushq %rax
-; SSE2-NEXT: .cfi_def_cfa_offset 16
; SSE2-NEXT: movdqu (%rdi), %xmm0
; SSE2-NEXT: movq %xmm0, %rax
; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,3,0,1]
-; SSE2-NEXT: movq %xmm0, %rdi
-; SSE2-NEXT: testq %rdi, %rax
-; SSE2-NEXT: je .LBB2_2
-; SSE2-NEXT: # %bb.1: # %trap
-; SSE2-NEXT: callq TrapFunc
-; SSE2-NEXT: .LBB2_2: # %ret
-; SSE2-NEXT: popq %rax
-; SSE2-NEXT: .cfi_def_cfa_offset 8
+; SSE2-NEXT: movq %xmm0, %rcx
+; SSE2-NEXT: testq %rcx, %rax
+; SSE2-NEXT: sete %al
; SSE2-NEXT: retq
;
; SSE41-LABEL: parseHeaders2_scalar_and:
; SSE41: # %bb.0:
-; SSE41-NEXT: pushq %rax
-; SSE41-NEXT: .cfi_def_cfa_offset 16
; SSE41-NEXT: movdqu (%rdi), %xmm0
; SSE41-NEXT: movq %xmm0, %rax
-; SSE41-NEXT: pextrq $1, %xmm0, %rdi
-; SSE41-NEXT: testq %rdi, %rax
-; SSE41-NEXT: je .LBB2_2
-; SSE41-NEXT: # %bb.1: # %trap
-; SSE41-NEXT: callq TrapFunc
-; SSE41-NEXT: .LBB2_2: # %ret
-; SSE41-NEXT: popq %rax
-; SSE41-NEXT: .cfi_def_cfa_offset 8
+; SSE41-NEXT: pextrq $1, %xmm0, %rcx
+; SSE41-NEXT: testq %rcx, %rax
+; SSE41-NEXT: sete %al
; SSE41-NEXT: retq
;
; AVX-LABEL: parseHeaders2_scalar_and:
; AVX: # %bb.0:
-; AVX-NEXT: pushq %rax
-; AVX-NEXT: .cfi_def_cfa_offset 16
; AVX-NEXT: vmovdqu (%rdi), %xmm0
; AVX-NEXT: vmovq %xmm0, %rax
-; AVX-NEXT: vpextrq $1, %xmm0, %rdi
-; AVX-NEXT: testq %rdi, %rax
-; AVX-NEXT: je .LBB2_2
-; AVX-NEXT: # %bb.1: # %trap
-; AVX-NEXT: callq TrapFunc
-; AVX-NEXT: .LBB2_2: # %ret
-; AVX-NEXT: popq %rax
-; AVX-NEXT: .cfi_def_cfa_offset 8
+; AVX-NEXT: vpextrq $1, %xmm0, %rcx
+; AVX-NEXT: testq %rcx, %rax
+; AVX-NEXT: sete %al
; AVX-NEXT: retq
%vptr = bitcast i64 * %ptr to <2 x i64> *
%vload = load <2 x i64>, <2 x i64> * %vptr, align 8
@@ -228,10 +120,5 @@ define void @parseHeaders2_scalar_and(i64 * %ptr) {
%v2 = extractelement <2 x i64> %vload, i32 1
%vreduce = and i64 %v1, %v2
%vcheck = icmp eq i64 %vreduce, 0
- br i1 %vcheck, label %ret, label %trap
-trap:
- call void @TrapFunc(i64 %v2)
- ret void
-ret:
- ret void
+ ret i1 %vcheck
}
More information about the llvm-commits
mailing list