[llvm] fbb6375 - [AArch64] Add cost model tests for min/max intrinsics.
Florian Hahn via llvm-commits
llvm-commits at lists.llvm.org
Thu Oct 22 05:36:40 PDT 2020
Author: Florian Hahn
Date: 2020-10-22T13:28:04+01:00
New Revision: fbb6375db0be912451646f14a35231d18cd91adf
URL: https://github.com/llvm/llvm-project/commit/fbb6375db0be912451646f14a35231d18cd91adf
DIFF: https://github.com/llvm/llvm-project/commit/fbb6375db0be912451646f14a35231d18cd91adf.diff
LOG: [AArch64] Add cost model tests for min/max intrinsics.
Added:
llvm/test/Analysis/CostModel/AArch64/min-max.ll
Modified:
Removed:
################################################################################
diff --git a/llvm/test/Analysis/CostModel/AArch64/min-max.ll b/llvm/test/Analysis/CostModel/AArch64/min-max.ll
new file mode 100644
index 000000000000..49eb6d40b33a
--- /dev/null
+++ b/llvm/test/Analysis/CostModel/AArch64/min-max.ll
@@ -0,0 +1,399 @@
+; RUN: opt < %s -mtriple=aarch64--linux-gnu -cost-model -analyze | FileCheck %s --check-prefix=COST
+; RUN: llc < %s -mtriple=aarch64--linux-gnu | FileCheck %s --check-prefix=CODE
+
+; COST-LABEL: umin.v8i8
+; COST-NEXT: Cost Model: Found an estimated cost of 30 for instruction: %res = call <8 x i8> @llvm.umin.v8i8(<8 x i8> %v0, <8 x i8> %v1)
+
+; CODE-LABEL: umin.v8i8
+; CODE: bb.0
+; CODE-NEXT: umin v{{.*}}.8b, v{{.*}}.8b, v{{.*}}.8b
+; CODE-NEXT: ret
+
+declare <8 x i8> @llvm.umin.v8i8(<8 x i8>, <8 x i8>)
+define <8 x i8> @umin.v8i8(<8 x i8> %v0, <8 x i8> %v1) {
+ %res = call <8 x i8> @llvm.umin.v8i8(<8 x i8> %v0, <8 x i8> %v1)
+ ret <8 x i8> %res
+}
+
+; COST-LABEL: umin.v9i8
+; COST-NEXT: Cost Model: Found an estimated cost of 34 for instruction: %res = call <9 x i8> @llvm.umin.v9i8(<9 x i8> %v0, <9 x i8> %v1)
+
+; CODE-LABEL: umin.v9i8
+; CODE: bb.0
+; CODE-NEXT: umin v{{.*}}.16b, v{{.*}}.16b, v{{.*}}.16b
+; CODE-NEXT: ret
+
+declare <9 x i8> @llvm.umin.v9i8(<9 x i8>, <9 x i8>)
+define <9 x i8> @umin.v9i8(<9 x i8> %v0, <9 x i8> %v1) {
+ %res = call <9 x i8> @llvm.umin.v9i8(<9 x i8> %v0, <9 x i8> %v1)
+ ret <9 x i8> %res
+}
+
+; COST-LABEL: umin.v16i8
+; COST-NEXT: Cost Model: Found an estimated cost of 62 for instruction: %res = call <16 x i8> @llvm.umin.v16i8(<16 x i8> %v0, <16 x i8> %v1)
+
+; CODE-LABEL: umin.v16i8
+; CODE: bb.0
+; CODE-NEXT: umin v{{.*}}.16b, v{{.*}}.16b, v{{.*}}.16b
+; CODE-NEXT: ret
+
+declare <16 x i8> @llvm.umin.v16i8(<16 x i8>, <16 x i8>)
+define <16 x i8> @umin.v16i8(<16 x i8> %v0, <16 x i8> %v1) {
+ %res = call <16 x i8> @llvm.umin.v16i8(<16 x i8> %v0, <16 x i8> %v1)
+ ret <16 x i8> %res
+}
+
+; COST-LABEL: umin.v8i16
+; COST-NEXT: Cost Model: Found an estimated cost of 30 for instruction: %res = call <8 x i16> @llvm.umin.v8i16(<8 x i16> %v0, <8 x i16> %v1)
+
+; CODE-LABEL: umin.v8i16
+; CODE: bb.0
+; CODE-NEXT: umin v{{.*}}.8h, v{{.*}}.8h, v{{.*}}.8h
+; CODE-NEXT: ret
+
+declare <8 x i16> @llvm.umin.v8i16(<8 x i16>, <8 x i16>)
+define <8 x i16> @umin.v8i16(<8 x i16> %v0, <8 x i16> %v1) {
+ %res = call <8 x i16> @llvm.umin.v8i16(<8 x i16> %v0, <8 x i16> %v1)
+ ret <8 x i16> %res
+}
+
+; COST-LABEL: umin.v4i32
+; COST-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %res = call <4 x i32> @llvm.umin.v4i32(<4 x i32> %v0, <4 x i32> %v1)
+
+; CODE-LABEL: umin.v4i32
+; CODE: bb.0
+; CODE-NEXT: umin v{{.*}}.4s, v{{.*}}.4s, v{{.*}}.4s
+; CODE-NEXT: ret
+
+declare <4 x i32> @llvm.umin.v4i32(<4 x i32>, <4 x i32>)
+define <4 x i32> @umin.v4i32(<4 x i32> %v0, <4 x i32> %v1) {
+ %res = call <4 x i32> @llvm.umin.v4i32(<4 x i32> %v0, <4 x i32> %v1)
+ ret <4 x i32> %res
+}
+
+; COST-LABEL: umin.v8i32
+; COST-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %res = call <8 x i32> @llvm.umin.v8i32(<8 x i32> %v0, <8 x i32> %v1)
+
+; CODE-LABEL: umin.v8i32
+; CODE: bb.0
+; CODE-NEXT: umin v{{.*}}.4s, v{{.*}}.4s, v{{.*}}.4s
+; CODE-NEXT: umin v{{.*}}.4s, v{{.*}}.4s, v{{.*}}.4s
+; CODE-NEXT: ret
+
+declare <8 x i32> @llvm.umin.v8i32(<8 x i32>, <8 x i32>)
+define <8 x i32> @umin.v8i32(<8 x i32> %v0, <8 x i32> %v1) {
+ %res = call <8 x i32> @llvm.umin.v8i32(<8 x i32> %v0, <8 x i32> %v1)
+ ret <8 x i32> %res
+}
+
+; COST-LABEL: umin.v2i64
+; COST-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res = call <2 x i64> @llvm.umin.v2i64(<2 x i64> %v0, <2 x i64> %v1)
+
+; CODE-LABEL: umin.v2i64
+; CODE: bb.0
+; CODE: csel
+; CODE: csel
+
+declare <2 x i64> @llvm.umin.v2i64(<2 x i64>, <2 x i64>)
+define <2 x i64> @umin.v2i64(<2 x i64> %v0, <2 x i64> %v1) {
+ %res = call <2 x i64> @llvm.umin.v2i64(<2 x i64> %v0, <2 x i64> %v1)
+ ret <2 x i64> %res
+}
+
+; COST-LABEL: smin.v8i8
+; COST-NEXT: Cost Model: Found an estimated cost of 30 for instruction: %res = call <8 x i8> @llvm.smin.v8i8(<8 x i8> %v0, <8 x i8> %v1)
+
+; CODE-LABEL: smin.v8i8
+; CODE: bb.0
+; CODE-NEXT: smin v{{.*}}.8b, v{{.*}}.8b, v{{.*}}.8b
+; CODE-NEXT: ret
+
+declare <8 x i8> @llvm.smin.v8i8(<8 x i8>, <8 x i8>)
+define <8 x i8> @smin.v8i8(<8 x i8> %v0, <8 x i8> %v1) {
+ %res = call <8 x i8> @llvm.smin.v8i8(<8 x i8> %v0, <8 x i8> %v1)
+ ret <8 x i8> %res
+}
+
+; COST-LABEL: smin.v9i8
+; COST-NEXT: Cost Model: Found an estimated cost of 34 for instruction: %res = call <9 x i8> @llvm.smin.v9i8(<9 x i8> %v0, <9 x i8> %v1)
+
+; CODE-LABEL: smin.v9i8
+; CODE: bb.0
+; CODE-NEXT: smin v{{.*}}.16b, v{{.*}}.16b, v{{.*}}.16b
+; CODE-NEXT: ret
+
+declare <9 x i8> @llvm.smin.v9i8(<9 x i8>, <9 x i8>)
+define <9 x i8> @smin.v9i8(<9 x i8> %v0, <9 x i8> %v1) {
+ %res = call <9 x i8> @llvm.smin.v9i8(<9 x i8> %v0, <9 x i8> %v1)
+ ret <9 x i8> %res
+}
+
+; COST-LABEL: smin.v16i8
+; COST-NEXT: Cost Model: Found an estimated cost of 62 for instruction: %res = call <16 x i8> @llvm.smin.v16i8(<16 x i8> %v0, <16 x i8> %v1)
+
+; CODE-LABEL: smin.v16i8
+; CODE: bb.0
+; CODE-NEXT: smin v{{.*}}.16b, v{{.*}}.16b, v{{.*}}.16b
+; CODE-NEXT: ret
+
+declare <16 x i8> @llvm.smin.v16i8(<16 x i8>, <16 x i8>)
+define <16 x i8> @smin.v16i8(<16 x i8> %v0, <16 x i8> %v1) {
+ %res = call <16 x i8> @llvm.smin.v16i8(<16 x i8> %v0, <16 x i8> %v1)
+ ret <16 x i8> %res
+}
+
+; COST-LABEL: smin.v8i16
+; COST-NEXT: Cost Model: Found an estimated cost of 30 for instruction: %res = call <8 x i16> @llvm.smin.v8i16(<8 x i16> %v0, <8 x i16> %v1)
+
+; CODE-LABEL: smin.v8i16
+; CODE: bb.0
+; CODE-NEXT: smin v{{.*}}.8h, v{{.*}}.8h, v{{.*}}.8h
+; CODE-NEXT: ret
+
+declare <8 x i16> @llvm.smin.v8i16(<8 x i16>, <8 x i16>)
+define <8 x i16> @smin.v8i16(<8 x i16> %v0, <8 x i16> %v1) {
+ %res = call <8 x i16> @llvm.smin.v8i16(<8 x i16> %v0, <8 x i16> %v1)
+ ret <8 x i16> %res
+}
+
+; COST-LABEL: smin.v4i32
+; COST-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %res = call <4 x i32> @llvm.smin.v4i32(<4 x i32> %v0, <4 x i32> %v1)
+
+; CODE-LABEL: smin.v4i32
+; CODE: bb.0
+; CODE-NEXT: smin v{{.*}}.4s, v{{.*}}.4s, v{{.*}}.4s
+; CODE-NEXT: ret
+
+declare <4 x i32> @llvm.smin.v4i32(<4 x i32>, <4 x i32>)
+define <4 x i32> @smin.v4i32(<4 x i32> %v0, <4 x i32> %v1) {
+ %res = call <4 x i32> @llvm.smin.v4i32(<4 x i32> %v0, <4 x i32> %v1)
+ ret <4 x i32> %res
+}
+
+; COST-LABEL: smin.v8i32
+; COST-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %res = call <8 x i32> @llvm.smin.v8i32(<8 x i32> %v0, <8 x i32> %v1)
+
+; CODE-LABEL: smin.v8i32
+; CODE: bb.0
+; CODE-NEXT: smin v{{.*}}.4s, v{{.*}}.4s, v{{.*}}.4s
+; CODE-NEXT: smin v{{.*}}.4s, v{{.*}}.4s, v{{.*}}.4s
+; CODE-NEXT: ret
+
+declare <8 x i32> @llvm.smin.v8i32(<8 x i32>, <8 x i32>)
+define <8 x i32> @smin.v8i32(<8 x i32> %v0, <8 x i32> %v1) {
+ %res = call <8 x i32> @llvm.smin.v8i32(<8 x i32> %v0, <8 x i32> %v1)
+ ret <8 x i32> %res
+}
+
+; COST-LABEL: smin.v2i64
+; COST-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res = call <2 x i64> @llvm.smin.v2i64(<2 x i64> %v0, <2 x i64> %v1)
+
+; CODE-LABEL: smin.v2i64
+; CODE: bb.0
+; CODE: csel
+; CODE: csel
+
+declare <2 x i64> @llvm.smin.v2i64(<2 x i64>, <2 x i64>)
+define <2 x i64> @smin.v2i64(<2 x i64> %v0, <2 x i64> %v1) {
+ %res = call <2 x i64> @llvm.smin.v2i64(<2 x i64> %v0, <2 x i64> %v1)
+ ret <2 x i64> %res
+}
+
+; COST-LABEL: umax.v8i8
+; COST-NEXT: Cost Model: Found an estimated cost of 30 for instruction: %res = call <8 x i8> @llvm.umax.v8i8(<8 x i8> %v0, <8 x i8> %v1)
+
+; CODE-LABEL: umax.v8i8
+; CODE: bb.0
+; CODE-NEXT: umax v{{.*}}.8b, v{{.*}}.8b, v{{.*}}.8b
+; CODE-NEXT: ret
+
+declare <8 x i8> @llvm.umax.v8i8(<8 x i8>, <8 x i8>)
+define <8 x i8> @umax.v8i8(<8 x i8> %v0, <8 x i8> %v1) {
+ %res = call <8 x i8> @llvm.umax.v8i8(<8 x i8> %v0, <8 x i8> %v1)
+ ret <8 x i8> %res
+}
+
+; COST-LABEL: umax.v9i8
+; COST-NEXT: Cost Model: Found an estimated cost of 34 for instruction: %res = call <9 x i8> @llvm.umax.v9i8(<9 x i8> %v0, <9 x i8> %v1)
+
+; CODE-LABEL: umax.v9i8
+; CODE: bb.0
+; CODE-NEXT: umax v{{.*}}.16b, v{{.*}}.16b, v{{.*}}.16b
+; CODE-NEXT: ret
+
+declare <9 x i8> @llvm.umax.v9i8(<9 x i8>, <9 x i8>)
+define <9 x i8> @umax.v9i8(<9 x i8> %v0, <9 x i8> %v1) {
+ %res = call <9 x i8> @llvm.umax.v9i8(<9 x i8> %v0, <9 x i8> %v1)
+ ret <9 x i8> %res
+}
+
+
+; COST-LABEL: umax.v16i8
+; COST-NEXT: Cost Model: Found an estimated cost of 62 for instruction: %res = call <16 x i8> @llvm.umax.v16i8(<16 x i8> %v0, <16 x i8> %v1)
+
+; CODE-LABEL: umax.v16i8
+; CODE: bb.0
+; CODE-NEXT: umax v{{.*}}.16b, v{{.*}}.16b, v{{.*}}.16b
+; CODE-NEXT: ret
+
+declare <16 x i8> @llvm.umax.v16i8(<16 x i8>, <16 x i8>)
+define <16 x i8> @umax.v16i8(<16 x i8> %v0, <16 x i8> %v1) {
+ %res = call <16 x i8> @llvm.umax.v16i8(<16 x i8> %v0, <16 x i8> %v1)
+ ret <16 x i8> %res
+}
+
+; COST-LABEL: umax.v8i16
+; COST-NEXT: Cost Model: Found an estimated cost of 30 for instruction: %res = call <8 x i16> @llvm.umax.v8i16(<8 x i16> %v0, <8 x i16> %v1)
+
+; CODE-LABEL: umax.v8i16
+; CODE: bb.0
+; CODE-NEXT: umax v{{.*}}.8h, v{{.*}}.8h, v{{.*}}.8h
+; CODE-NEXT: ret
+
+declare <8 x i16> @llvm.umax.v8i16(<8 x i16>, <8 x i16>)
+define <8 x i16> @umax.v8i16(<8 x i16> %v0, <8 x i16> %v1) {
+ %res = call <8 x i16> @llvm.umax.v8i16(<8 x i16> %v0, <8 x i16> %v1)
+ ret <8 x i16> %res
+}
+
+; COST-LABEL: umax.v4i32
+; COST-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %res = call <4 x i32> @llvm.umax.v4i32(<4 x i32> %v0, <4 x i32> %v1)
+
+; CODE-LABEL: umax.v4i32
+; CODE: bb.0
+; CODE-NEXT: umax v{{.*}}.4s, v{{.*}}.4s, v{{.*}}.4s
+; CODE-NEXT: ret
+
+declare <4 x i32> @llvm.umax.v4i32(<4 x i32>, <4 x i32>)
+define <4 x i32> @umax.v4i32(<4 x i32> %v0, <4 x i32> %v1) {
+ %res = call <4 x i32> @llvm.umax.v4i32(<4 x i32> %v0, <4 x i32> %v1)
+ ret <4 x i32> %res
+}
+
+; COST-LABEL: umax.v8i32
+; COST-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %res = call <8 x i32> @llvm.umax.v8i32(<8 x i32> %v0, <8 x i32> %v1)
+
+; CODE-LABEL: umax.v8i32
+; CODE: bb.0
+; CODE-NEXT: umax v{{.*}}.4s, v{{.*}}.4s, v{{.*}}.4s
+; CODE-NEXT: umax v{{.*}}.4s, v{{.*}}.4s, v{{.*}}.4s
+; CODE-NEXT: ret
+
+declare <8 x i32> @llvm.umax.v8i32(<8 x i32>, <8 x i32>)
+define <8 x i32> @umax.v8i32(<8 x i32> %v0, <8 x i32> %v1) {
+ %res = call <8 x i32> @llvm.umax.v8i32(<8 x i32> %v0, <8 x i32> %v1)
+ ret <8 x i32> %res
+}
+
+; COST-LABEL: umax.v2i64
+; COST-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res = call <2 x i64> @llvm.umax.v2i64(<2 x i64> %v0, <2 x i64> %v1)
+
+; CODE-LABEL: umax.v2i64
+; CODE: bb.0
+; CODE: csel
+; CODE: csel
+
+declare <2 x i64> @llvm.umax.v2i64(<2 x i64>, <2 x i64>)
+define <2 x i64> @umax.v2i64(<2 x i64> %v0, <2 x i64> %v1) {
+ %res = call <2 x i64> @llvm.umax.v2i64(<2 x i64> %v0, <2 x i64> %v1)
+ ret <2 x i64> %res
+}
+
+; COST-LABEL: smax.v8i8
+; COST-NEXT: Cost Model: Found an estimated cost of 30 for instruction: %res = call <8 x i8> @llvm.smax.v8i8(<8 x i8> %v0, <8 x i8> %v1)
+
+; CODE-LABEL: smax.v8i8
+; CODE: bb.0
+; CODE-NEXT: smax v{{.*}}.8b, v{{.*}}.8b, v{{.*}}.8b
+; CODE-NEXT: ret
+
+declare <8 x i8> @llvm.smax.v8i8(<8 x i8>, <8 x i8>)
+define <8 x i8> @smax.v8i8(<8 x i8> %v0, <8 x i8> %v1) {
+ %res = call <8 x i8> @llvm.smax.v8i8(<8 x i8> %v0, <8 x i8> %v1)
+ ret <8 x i8> %res
+}
+
+; COST-LABEL: smax.v9i8
+; COST-NEXT: Cost Model: Found an estimated cost of 34 for instruction: %res = call <9 x i8> @llvm.smax.v9i8(<9 x i8> %v0, <9 x i8> %v1)
+
+; CODE-LABEL: smax.v9i8
+; CODE: bb.0
+; CODE-NEXT: smax v{{.*}}.16b, v{{.*}}.16b, v{{.*}}.16b
+; CODE-NEXT: ret
+
+declare <9 x i8> @llvm.smax.v9i8(<9 x i8>, <9 x i8>)
+define <9 x i8> @smax.v9i8(<9 x i8> %v0, <9 x i8> %v1) {
+ %res = call <9 x i8> @llvm.smax.v9i8(<9 x i8> %v0, <9 x i8> %v1)
+ ret <9 x i8> %res
+}
+
+; COST-LABEL: smax.v16i8
+; COST-NEXT: Cost Model: Found an estimated cost of 62 for instruction: %res = call <16 x i8> @llvm.smax.v16i8(<16 x i8> %v0, <16 x i8> %v1)
+
+; CODE-LABEL: smax.v16i8
+; CODE: bb.0
+; CODE-NEXT: smax v{{.*}}.16b, v{{.*}}.16b, v{{.*}}.16b
+; CODE-NEXT: ret
+
+declare <16 x i8> @llvm.smax.v16i8(<16 x i8>, <16 x i8>)
+define <16 x i8> @smax.v16i8(<16 x i8> %v0, <16 x i8> %v1) {
+ %res = call <16 x i8> @llvm.smax.v16i8(<16 x i8> %v0, <16 x i8> %v1)
+ ret <16 x i8> %res
+}
+
+; COST-LABEL: smax.v8i16
+; COST-NEXT: Cost Model: Found an estimated cost of 30 for instruction: %res = call <8 x i16> @llvm.smax.v8i16(<8 x i16> %v0, <8 x i16> %v1)
+
+; CODE-LABEL: smax.v8i16
+; CODE: bb.0
+; CODE-NEXT: smax v{{.*}}.8h, v{{.*}}.8h, v{{.*}}.8h
+; CODE-NEXT: ret
+
+declare <8 x i16> @llvm.smax.v8i16(<8 x i16>, <8 x i16>)
+define <8 x i16> @smax.v8i16(<8 x i16> %v0, <8 x i16> %v1) {
+ %res = call <8 x i16> @llvm.smax.v8i16(<8 x i16> %v0, <8 x i16> %v1)
+ ret <8 x i16> %res
+}
+
+; COST-LABEL: smax.v4i32
+; COST-NEXT: Cost Model: Found an estimated cost of 14 for instruction: %res = call <4 x i32> @llvm.smax.v4i32(<4 x i32> %v0, <4 x i32> %v1)
+
+; CODE-LABEL: smax.v4i32
+; CODE: bb.0
+; CODE-NEXT: smax v{{.*}}.4s, v{{.*}}.4s, v{{.*}}.4s
+; CODE-NEXT: ret
+
+declare <4 x i32> @llvm.smax.v4i32(<4 x i32>, <4 x i32>)
+define <4 x i32> @smax.v4i32(<4 x i32> %v0, <4 x i32> %v1) {
+ %res = call <4 x i32> @llvm.smax.v4i32(<4 x i32> %v0, <4 x i32> %v1)
+ ret <4 x i32> %res
+}
+
+; COST-LABEL: smax.v8i32
+; COST-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %res = call <8 x i32> @llvm.smax.v8i32(<8 x i32> %v0, <8 x i32> %v1)
+
+; CODE-LABEL: smax.v8i32
+; CODE: bb.0
+; CODE-NEXT: smax v{{.*}}.4s, v{{.*}}.4s, v{{.*}}.4s
+; CODE-NEXT: smax v{{.*}}.4s, v{{.*}}.4s, v{{.*}}.4s
+; CODE-NEXT: ret
+
+declare <8 x i32> @llvm.smax.v8i32(<8 x i32>, <8 x i32>)
+define <8 x i32> @smax.v8i32(<8 x i32> %v0, <8 x i32> %v1) {
+ %res = call <8 x i32> @llvm.smax.v8i32(<8 x i32> %v0, <8 x i32> %v1)
+ ret <8 x i32> %res
+}
+
+; COST-LABEL: smax.v2i64
+; COST-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %res = call <2 x i64> @llvm.smax.v2i64(<2 x i64> %v0, <2 x i64> %v1)
+
+; CODE-LABEL: smax.v2i64
+; CODE: bb.0
+; CODE: csel
+; CODE: csel
+
+declare <2 x i64> @llvm.smax.v2i64(<2 x i64>, <2 x i64>)
+define <2 x i64> @smax.v2i64(<2 x i64> %v0, <2 x i64> %v1) {
+ %res = call <2 x i64> @llvm.smax.v2i64(<2 x i64> %v0, <2 x i64> %v1)
+ ret <2 x i64> %res
+}
More information about the llvm-commits
mailing list