aboutsummaryrefslogtreecommitdiffstats
path: root/test/CodeGen/AArch64/fp16-v4-instructions.ll
diff options
context:
space:
mode:
authorPirama Arumuga Nainar <pirama@google.com>2015-05-20 12:55:36 -0700
committerPirama Arumuga Nainar <pirama@google.com>2015-05-26 14:28:33 -0700
commit560a17f4e3f8019a108abb7848a4735c59a43be0 (patch)
tree89b942fa3484ef229429f490a5e8219e741af0d1 /test/CodeGen/AArch64/fp16-v4-instructions.ll
parent2c3e0051c31c3f5b2328b447eadf1cf9c4427442 (diff)
downloadexternal_llvm-560a17f4e3f8019a108abb7848a4735c59a43be0.zip
external_llvm-560a17f4e3f8019a108abb7848a4735c59a43be0.tar.gz
external_llvm-560a17f4e3f8019a108abb7848a4735c59a43be0.tar.bz2
Update aosp/master LLVM with patches for fp16
Cherry-pick LLVM revisions r235191, r235215, r235220, r235341, r235363, r235530, r235609, r235610, r237004 r235191 has a required bug-fix and the rest are all related to fp16. Change-Id: I7fe8da5ffd8f2c06150885a54769abd18c3a04c6 (cherry picked from commit a18e6af1712fd41c4a705a19ad71f6e9ac7a4e68)
Diffstat (limited to 'test/CodeGen/AArch64/fp16-v4-instructions.ll')
-rw-r--r--test/CodeGen/AArch64/fp16-v4-instructions.ll92
1 files changed, 91 insertions, 1 deletions
diff --git a/test/CodeGen/AArch64/fp16-v4-instructions.ll b/test/CodeGen/AArch64/fp16-v4-instructions.ll
index 6db4e97..0dbda15 100644
--- a/test/CodeGen/AArch64/fp16-v4-instructions.ll
+++ b/test/CodeGen/AArch64/fp16-v4-instructions.ll
@@ -1,4 +1,4 @@
-; RUN: llc < %s -mtriple=aarch64-none-eabi | FileCheck %s
+; RUN: llc < %s -asm-verbose=false -mtriple=aarch64-none-eabi | FileCheck %s
define <4 x half> @add_h(<4 x half> %a, <4 x half> %b) {
entry:
@@ -129,3 +129,93 @@ define <4 x i16> @bitcast_h_to_i(float, <4 x half> %a) {
%2 = bitcast <4 x half> %a to <4 x i16>
ret <4 x i16> %2
}
+
+
+define <4 x half> @sitofp_i8(<4 x i8> %a) #0 {
+; CHECK-LABEL: sitofp_i8:
+; CHECK-NEXT: shl [[OP1:v[0-9]+\.4h]], v0.4h, #8
+; CHECK-NEXT: sshr [[OP2:v[0-9]+\.4h]], [[OP1]], #8
+; CHECK-NEXT: sshll [[OP3:v[0-9]+\.4s]], [[OP2]], #0
+; CHECK-NEXT: scvtf [[OP4:v[0-9]+\.4s]], [[OP3]]
+; CHECK-NEXT: fcvtn v0.4h, [[OP4]]
+; CHECK-NEXT: ret
+ %1 = sitofp <4 x i8> %a to <4 x half>
+ ret <4 x half> %1
+}
+
+
+define <4 x half> @sitofp_i16(<4 x i16> %a) #0 {
+; CHECK-LABEL: sitofp_i16:
+; CHECK-NEXT: sshll [[OP1:v[0-9]+\.4s]], v0.4h, #0
+; CHECK-NEXT: scvtf [[OP2:v[0-9]+\.4s]], [[OP1]]
+; CHECK-NEXT: fcvtn v0.4h, [[OP2]]
+; CHECK-NEXT: ret
+ %1 = sitofp <4 x i16> %a to <4 x half>
+ ret <4 x half> %1
+}
+
+
+define <4 x half> @sitofp_i32(<4 x i32> %a) #0 {
+; CHECK-LABEL: sitofp_i32:
+; CHECK-NEXT: scvtf [[OP1:v[0-9]+\.4s]], v0.4s
+; CHECK-NEXT: fcvtn v0.4h, [[OP1]]
+ %1 = sitofp <4 x i32> %a to <4 x half>
+ ret <4 x half> %1
+}
+
+
+define <4 x half> @sitofp_i64(<4 x i64> %a) #0 {
+; CHECK-LABEL: sitofp_i64:
+; CHECK-DAG: scvtf [[OP1:v[0-9]+\.2d]], v0.2d
+; CHECK-DAG: scvtf [[OP2:v[0-9]+\.2d]], v1.2d
+; CHECK-DAG: fcvtn [[OP3:v[0-9]+]].2s, [[OP1]]
+; CHECK-NEXT: fcvtn2 [[OP3]].4s, [[OP2]]
+; CHECK-NEXT: fcvtn v0.4h, [[OP3]].4s
+ %1 = sitofp <4 x i64> %a to <4 x half>
+ ret <4 x half> %1
+}
+
+define <4 x half> @uitofp_i8(<4 x i8> %a) #0 {
+; CHECK-LABEL: uitofp_i8:
+; CHECK-NEXT: bic v0.4h, #0xff, lsl #8
+; CHECK-NEXT: ushll [[OP1:v[0-9]+\.4s]], v0.4h, #0
+; CHECK-NEXT: ucvtf [[OP2:v[0-9]+\.4s]], [[OP1]]
+; CHECK-NEXT: fcvtn v0.4h, [[OP2]]
+; CHECK-NEXT: ret
+ %1 = uitofp <4 x i8> %a to <4 x half>
+ ret <4 x half> %1
+}
+
+
+define <4 x half> @uitofp_i16(<4 x i16> %a) #0 {
+; CHECK-LABEL: uitofp_i16:
+; CHECK-NEXT: ushll [[OP1:v[0-9]+\.4s]], v0.4h, #0
+; CHECK-NEXT: ucvtf [[OP2:v[0-9]+\.4s]], [[OP1]]
+; CHECK-NEXT: fcvtn v0.4h, [[OP2]]
+; CHECK-NEXT: ret
+ %1 = uitofp <4 x i16> %a to <4 x half>
+ ret <4 x half> %1
+}
+
+
+define <4 x half> @uitofp_i32(<4 x i32> %a) #0 {
+; CHECK-LABEL: uitofp_i32:
+; CHECK-NEXT: ucvtf [[OP1:v[0-9]+\.4s]], v0.4s
+; CHECK-NEXT: fcvtn v0.4h, [[OP1]]
+ %1 = uitofp <4 x i32> %a to <4 x half>
+ ret <4 x half> %1
+}
+
+
+define <4 x half> @uitofp_i64(<4 x i64> %a) #0 {
+; CHECK-LABEL: uitofp_i64:
+; CHECK-DAG: ucvtf [[OP1:v[0-9]+\.2d]], v0.2d
+; CHECK-DAG: ucvtf [[OP2:v[0-9]+\.2d]], v1.2d
+; CHECK-DAG: fcvtn [[OP3:v[0-9]+]].2s, [[OP1]]
+; CHECK-NEXT: fcvtn2 [[OP3]].4s, [[OP2]]
+; CHECK-NEXT: fcvtn v0.4h, [[OP3]].4s
+ %1 = uitofp <4 x i64> %a to <4 x half>
+ ret <4 x half> %1
+}
+
+attributes #0 = { nounwind }