summaryrefslogtreecommitdiff
path: root/llvm/test/CodeGen/LoongArch/lsx/fdiv-reciprocal-estimate.ll
diff options
context:
space:
mode:
Diffstat (limited to 'llvm/test/CodeGen/LoongArch/lsx/fdiv-reciprocal-estimate.ll')
-rw-r--r--llvm/test/CodeGen/LoongArch/lsx/fdiv-reciprocal-estimate.ll135
1 files changed, 87 insertions, 48 deletions
diff --git a/llvm/test/CodeGen/LoongArch/lsx/fdiv-reciprocal-estimate.ll b/llvm/test/CodeGen/LoongArch/lsx/fdiv-reciprocal-estimate.ll
index 21dbbf310ad8..58e16d37ae27 100644
--- a/llvm/test/CodeGen/LoongArch/lsx/fdiv-reciprocal-estimate.ll
+++ b/llvm/test/CodeGen/LoongArch/lsx/fdiv-reciprocal-estimate.ll
@@ -1,6 +1,8 @@
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 4
-; RUN: llc --mtriple=loongarch64 --mattr=+lsx,-frecipe < %s | FileCheck %s --check-prefix=FAULT
-; RUN: llc --mtriple=loongarch64 --mattr=+lsx,+frecipe < %s | FileCheck %s
+; RUN: llc --mtriple=loongarch32 --mattr=+32s,+lsx,-frecipe < %s | FileCheck %s --check-prefixes=FAULT,FAULT-LA32
+; RUN: llc --mtriple=loongarch32 --mattr=+32s,+lsx,+frecipe < %s | FileCheck %s --check-prefixes=CHECK,LA32
+; RUN: llc --mtriple=loongarch64 --mattr=+lsx,-frecipe < %s | FileCheck %s --check-prefixes=FAULT,FAULT-LA64
+; RUN: llc --mtriple=loongarch64 --mattr=+lsx,+frecipe < %s | FileCheck %s --check-prefixes=CHECK,LA64
define void @fdiv_v4f32(ptr %res, ptr %a0, ptr %a1) nounwind {
; FAULT-LABEL: fdiv_v4f32:
@@ -13,13 +15,13 @@ define void @fdiv_v4f32(ptr %res, ptr %a0, ptr %a1) nounwind {
;
; CHECK-LABEL: fdiv_v4f32:
; CHECK: # %bb.0: # %entry
-; CHECK-NEXT: vld $vr0, $a2, 0
-; CHECK-NEXT: vld $vr1, $a1, 0
-; CHECK-NEXT: vfrecipe.s $vr2, $vr0
-; CHECK-NEXT: vfmul.s $vr3, $vr1, $vr2
-; CHECK-NEXT: vfnmsub.s $vr0, $vr0, $vr3, $vr1
-; CHECK-NEXT: vfmadd.s $vr0, $vr2, $vr0, $vr3
-; CHECK-NEXT: vst $vr0, $a0, 0
+; CHECK-NEXT: vld $vr0, $a2, 0
+; CHECK-NEXT: vld $vr1, $a1, 0
+; CHECK-NEXT: vfrecipe.s $vr2, $vr0
+; CHECK-NEXT: vfmul.s $vr3, $vr1, $vr2
+; CHECK-NEXT: vfnmsub.s $vr0, $vr0, $vr3, $vr1
+; CHECK-NEXT: vfmadd.s $vr0, $vr2, $vr0, $vr3
+; CHECK-NEXT: vst $vr0, $a0, 0
; CHECK-NEXT: ret
entry:
%v0 = load <4 x float>, ptr %a0
@@ -38,20 +40,35 @@ define void @fdiv_v2f64(ptr %res, ptr %a0, ptr %a1) nounwind {
; FAULT-NEXT: vst $vr0, $a0, 0
; FAULT-NEXT: ret
;
-; CHECK-LABEL: fdiv_v2f64:
-; CHECK: # %bb.0: # %entry
-; CHECK-NEXT: vld $vr0, $a2, 0
-; CHECK-NEXT: vld $vr1, $a1, 0
-; CHECK-NEXT: lu52i.d $a1, $zero, -1025
-; CHECK-NEXT: vreplgr2vr.d $vr2, $a1
-; CHECK-NEXT: vfrecipe.d $vr3, $vr0
-; CHECK-NEXT: vfmadd.d $vr2, $vr0, $vr3, $vr2
-; CHECK-NEXT: vfnmsub.d $vr2, $vr2, $vr3, $vr3
-; CHECK-NEXT: vfmul.d $vr3, $vr1, $vr2
-; CHECK-NEXT: vfnmsub.d $vr0, $vr0, $vr3, $vr1
-; CHECK-NEXT: vfmadd.d $vr0, $vr2, $vr0, $vr3
-; CHECK-NEXT: vst $vr0, $a0, 0
-; CHECK-NEXT: ret
+; LA32-LABEL: fdiv_v2f64:
+; LA32: # %bb.0: # %entry
+; LA32-NEXT: pcalau12i $a3, %pc_hi20(.LCPI1_0)
+; LA32-NEXT: vld $vr0, $a2, 0
+; LA32-NEXT: vld $vr1, $a3, %pc_lo12(.LCPI1_0)
+; LA32-NEXT: vld $vr2, $a1, 0
+; LA32-NEXT: vfrecipe.d $vr3, $vr0
+; LA32-NEXT: vfmadd.d $vr1, $vr0, $vr3, $vr1
+; LA32-NEXT: vfnmsub.d $vr1, $vr1, $vr3, $vr3
+; LA32-NEXT: vfmul.d $vr3, $vr2, $vr1
+; LA32-NEXT: vfnmsub.d $vr0, $vr0, $vr3, $vr2
+; LA32-NEXT: vfmadd.d $vr0, $vr1, $vr0, $vr3
+; LA32-NEXT: vst $vr0, $a0, 0
+; LA32-NEXT: ret
+;
+; LA64-LABEL: fdiv_v2f64:
+; LA64: # %bb.0: # %entry
+; LA64-NEXT: vld $vr0, $a2, 0
+; LA64-NEXT: vld $vr1, $a1, 0
+; LA64-NEXT: lu52i.d $a1, $zero, -1025
+; LA64-NEXT: vreplgr2vr.d $vr2, $a1
+; LA64-NEXT: vfrecipe.d $vr3, $vr0
+; LA64-NEXT: vfmadd.d $vr2, $vr0, $vr3, $vr2
+; LA64-NEXT: vfnmsub.d $vr2, $vr2, $vr3, $vr3
+; LA64-NEXT: vfmul.d $vr3, $vr1, $vr2
+; LA64-NEXT: vfnmsub.d $vr0, $vr0, $vr3, $vr1
+; LA64-NEXT: vfmadd.d $vr0, $vr2, $vr0, $vr3
+; LA64-NEXT: vst $vr0, $a0, 0
+; LA64-NEXT: ret
entry:
%v0 = load <2 x double>, ptr %a0
%v1 = load <2 x double>, ptr %a1
@@ -71,13 +88,13 @@ define void @one_fdiv_v4f32(ptr %res, ptr %a0) nounwind {
;
; CHECK-LABEL: one_fdiv_v4f32:
; CHECK: # %bb.0: # %entry
-; CHECK-NEXT: vld $vr0, $a1, 0
-; CHECK-NEXT: vfrecipe.s $vr1, $vr0
-; CHECK-NEXT: lu12i.w $a1, -264192
-; CHECK-NEXT: vreplgr2vr.w $vr2, $a1
-; CHECK-NEXT: vfmadd.s $vr0, $vr0, $vr1, $vr2
-; CHECK-NEXT: vfnmsub.s $vr0, $vr0, $vr1, $vr1
-; CHECK-NEXT: vst $vr0, $a0, 0
+; CHECK-NEXT: vld $vr0, $a1, 0
+; CHECK-NEXT: vfrecipe.s $vr1, $vr0
+; CHECK-NEXT: lu12i.w $a1, -264192
+; CHECK-NEXT: vreplgr2vr.w $vr2, $a1
+; CHECK-NEXT: vfmadd.s $vr0, $vr0, $vr1, $vr2
+; CHECK-NEXT: vfnmsub.s $vr0, $vr0, $vr1, $vr1
+; CHECK-NEXT: vst $vr0, $a0, 0
; CHECK-NEXT: ret
entry:
%v0 = load <4 x float>, ptr %a0
@@ -87,25 +104,47 @@ entry:
}
define void @one_fdiv_v2f64(ptr %res, ptr %a0) nounwind {
-; FAULT-LABEL: one_fdiv_v2f64:
-; FAULT: # %bb.0: # %entry
-; FAULT-NEXT: vld $vr0, $a1, 0
-; FAULT-NEXT: vfrecip.d $vr0, $vr0
-; FAULT-NEXT: vst $vr0, $a0, 0
-; FAULT-NEXT: ret
+; FAULT-LA32-LABEL: one_fdiv_v2f64:
+; FAULT-LA32: # %bb.0: # %entry
+; FAULT-LA32-NEXT: vld $vr0, $a1, 0
+; FAULT-LA32-NEXT: pcalau12i $a1, %pc_hi20(.LCPI3_0)
+; FAULT-LA32-NEXT: vld $vr1, $a1, %pc_lo12(.LCPI3_0)
+; FAULT-LA32-NEXT: vfdiv.d $vr0, $vr1, $vr0
+; FAULT-LA32-NEXT: vst $vr0, $a0, 0
+; FAULT-LA32-NEXT: ret
;
-; CHECK-LABEL: one_fdiv_v2f64:
-; CHECK: # %bb.0: # %entry
-; CHECK-NEXT: vld $vr0, $a1, 0
-; CHECK-NEXT: vfrecipe.d $vr1, $vr0
-; CHECK-NEXT: lu52i.d $a1, $zero, 1023
-; CHECK-NEXT: vreplgr2vr.d $vr2, $a1
-; CHECK-NEXT: vfnmsub.d $vr3, $vr0, $vr1, $vr2
-; CHECK-NEXT: vfmadd.d $vr1, $vr1, $vr3, $vr1
-; CHECK-NEXT: vfnmsub.d $vr0, $vr0, $vr1, $vr2
-; CHECK-NEXT: vfmadd.d $vr0, $vr1, $vr0, $vr1
-; CHECK-NEXT: vst $vr0, $a0, 0
-; CHECK-NEXT: ret
+; LA32-LABEL: one_fdiv_v2f64:
+; LA32: # %bb.0: # %entry
+; LA32-NEXT: vld $vr0, $a1, 0
+; LA32-NEXT: pcalau12i $a1, %pc_hi20(.LCPI3_0)
+; LA32-NEXT: vld $vr1, $a1, %pc_lo12(.LCPI3_0)
+; LA32-NEXT: vfrecipe.d $vr2, $vr0
+; LA32-NEXT: vfnmsub.d $vr3, $vr0, $vr2, $vr1
+; LA32-NEXT: vfmadd.d $vr2, $vr2, $vr3, $vr2
+; LA32-NEXT: vfnmsub.d $vr0, $vr0, $vr2, $vr1
+; LA32-NEXT: vfmadd.d $vr0, $vr2, $vr0, $vr2
+; LA32-NEXT: vst $vr0, $a0, 0
+; LA32-NEXT: ret
+;
+; FAULT-LA64-LABEL: one_fdiv_v2f64:
+; FAULT-LA64: # %bb.0: # %entry
+; FAULT-LA64-NEXT: vld $vr0, $a1, 0
+; FAULT-LA64-NEXT: vfrecip.d $vr0, $vr0
+; FAULT-LA64-NEXT: vst $vr0, $a0, 0
+; FAULT-LA64-NEXT: ret
+;
+; LA64-LABEL: one_fdiv_v2f64:
+; LA64: # %bb.0: # %entry
+; LA64-NEXT: vld $vr0, $a1, 0
+; LA64-NEXT: vfrecipe.d $vr1, $vr0
+; LA64-NEXT: lu52i.d $a1, $zero, 1023
+; LA64-NEXT: vreplgr2vr.d $vr2, $a1
+; LA64-NEXT: vfnmsub.d $vr3, $vr0, $vr1, $vr2
+; LA64-NEXT: vfmadd.d $vr1, $vr1, $vr3, $vr1
+; LA64-NEXT: vfnmsub.d $vr0, $vr0, $vr1, $vr2
+; LA64-NEXT: vfmadd.d $vr0, $vr1, $vr0, $vr1
+; LA64-NEXT: vst $vr0, $a0, 0
+; LA64-NEXT: ret
entry:
%v0 = load <2 x double>, ptr %a0
%div = fdiv fast <2 x double> <double 1.0, double 1.0>, %v0
generated by cgit v1.2.3 (git 2.25.1) at 2026-05-31 16:57:08 +0000