[llvm] e311cdd - [NFC][X86][LV] Add costmodel test coverage for interleaved i8 load/store stride=6
Roman Lebedev via llvm-commits
llvm-commits at lists.llvm.org
Sun Oct 3 04:35:39 PDT 2021
Author: Roman Lebedev
Date: 2021-10-03T14:33:59+03:00
New Revision: e311cdd18d21e8aae6ff54bd0cab47bf1bcc4325
URL: https://github.com/llvm/llvm-project/commit/e311cdd18d21e8aae6ff54bd0cab47bf1bcc4325
DIFF: https://github.com/llvm/llvm-project/commit/e311cdd18d21e8aae6ff54bd0cab47bf1bcc4325.diff
LOG: [NFC][X86][LV] Add costmodel test coverage for interleaved i8 load/store stride=6
Added:
llvm/test/Analysis/CostModel/X86/interleaved-load-i8-stride-6.ll
llvm/test/Analysis/CostModel/X86/interleaved-store-i8-stride-6.ll
Modified:
Removed:
################################################################################
diff --git a/llvm/test/Analysis/CostModel/X86/interleaved-load-i8-stride-6.ll b/llvm/test/Analysis/CostModel/X86/interleaved-load-i8-stride-6.ll
new file mode 100644
index 0000000000000..2528f87d92dce
--- /dev/null
+++ b/llvm/test/Analysis/CostModel/X86/interleaved-load-i8-stride-6.ll
@@ -0,0 +1,88 @@
+; RUN: opt -loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+sse2 --debug-only=loop-vectorize < %s 2>&1 | FileCheck %s --check-prefixes=CHECK,SSE2
+; RUN: opt -loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx --debug-only=loop-vectorize < %s 2>&1 | FileCheck %s --check-prefixes=CHECK,AVX1
+; RUN: opt -loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx2 --debug-only=loop-vectorize < %s 2>&1 | FileCheck %s --check-prefixes=CHECK,AVX2
+; RUN: opt -loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx512bw,+avx512vl --debug-only=loop-vectorize < %s 2>&1 | FileCheck %s --check-prefixes=CHECK,AVX512
+; REQUIRES: asserts
+
+target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-unknown-linux-gnu"
+
+ at A = global [1024 x i8] zeroinitializer, align 128
+ at B = global [1024 x i8] zeroinitializer, align 128
+
+; CHECK: LV: Checking a loop in "test"
+;
+; SSE2: LV: Found an estimated cost of 1 for VF 1 For instruction: %v0 = load i8, i8* %in0, align 1
+; SSE2: LV: Found an estimated cost of 47 for VF 2 For instruction: %v0 = load i8, i8* %in0, align 1
+; SSE2: LV: Found an estimated cost of 90 for VF 4 For instruction: %v0 = load i8, i8* %in0, align 1
+; SSE2: LV: Found an estimated cost of 186 for VF 8 For instruction: %v0 = load i8, i8* %in0, align 1
+; SSE2: LV: Found an estimated cost of 378 for VF 16 For instruction: %v0 = load i8, i8* %in0, align 1
+;
+; AVX1: LV: Found an estimated cost of 1 for VF 1 For instruction: %v0 = load i8, i8* %in0, align 1
+; AVX1: LV: Found an estimated cost of 27 for VF 2 For instruction: %v0 = load i8, i8* %in0, align 1
+; AVX1: LV: Found an estimated cost of 59 for VF 4 For instruction: %v0 = load i8, i8* %in0, align 1
+; AVX1: LV: Found an estimated cost of 114 for VF 8 For instruction: %v0 = load i8, i8* %in0, align 1
+; AVX1: LV: Found an estimated cost of 243 for VF 16 For instruction: %v0 = load i8, i8* %in0, align 1
+; AVX1: LV: Found an estimated cost of 498 for VF 32 For instruction: %v0 = load i8, i8* %in0, align 1
+;
+; AVX2: LV: Found an estimated cost of 1 for VF 1 For instruction: %v0 = load i8, i8* %in0, align 1
+; AVX2: LV: Found an estimated cost of 27 for VF 2 For instruction: %v0 = load i8, i8* %in0, align 1
+; AVX2: LV: Found an estimated cost of 59 for VF 4 For instruction: %v0 = load i8, i8* %in0, align 1
+; AVX2: LV: Found an estimated cost of 114 for VF 8 For instruction: %v0 = load i8, i8* %in0, align 1
+; AVX2: LV: Found an estimated cost of 243 for VF 16 For instruction: %v0 = load i8, i8* %in0, align 1
+; AVX2: LV: Found an estimated cost of 498 for VF 32 For instruction: %v0 = load i8, i8* %in0, align 1
+;
+; AVX512: LV: Found an estimated cost of 1 for VF 1 For instruction: %v0 = load i8, i8* %in0, align 1
+; AVX512: LV: Found an estimated cost of 7 for VF 2 For instruction: %v0 = load i8, i8* %in0, align 1
+; AVX512: LV: Found an estimated cost of 25 for VF 4 For instruction: %v0 = load i8, i8* %in0, align 1
+; AVX512: LV: Found an estimated cost of 49 for VF 8 For instruction: %v0 = load i8, i8* %in0, align 1
+; AVX512: LV: Found an estimated cost of 119 for VF 16 For instruction: %v0 = load i8, i8* %in0, align 1
+; AVX512: LV: Found an estimated cost of 237 for VF 32 For instruction: %v0 = load i8, i8* %in0, align 1
+; AVX512: LV: Found an estimated cost of 591 for VF 64 For instruction: %v0 = load i8, i8* %in0, align 1
+;
+; CHECK-NOT: LV: Found an estimated cost of {{[0-9]+}} for VF {{[0-9]+}} For instruction: %v0 = load i8, i8* %in0, align 1
+
+define void @test() {
+entry:
+ br label %for.body
+
+for.body:
+ %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]
+
+ %iv.0 = add nuw nsw i64 %iv, 0
+ %iv.1 = add nuw nsw i64 %iv, 1
+ %iv.2 = add nuw nsw i64 %iv, 2
+ %iv.3 = add nuw nsw i64 %iv, 3
+ %iv.4 = add nuw nsw i64 %iv, 4
+ %iv.5 = add nuw nsw i64 %iv, 5
+
+ %in0 = getelementptr inbounds [1024 x i8], [1024 x i8]* @A, i64 0, i64 %iv.0
+ %in1 = getelementptr inbounds [1024 x i8], [1024 x i8]* @A, i64 0, i64 %iv.1
+ %in2 = getelementptr inbounds [1024 x i8], [1024 x i8]* @A, i64 0, i64 %iv.2
+ %in3 = getelementptr inbounds [1024 x i8], [1024 x i8]* @A, i64 0, i64 %iv.3
+ %in4 = getelementptr inbounds [1024 x i8], [1024 x i8]* @A, i64 0, i64 %iv.4
+ %in5 = getelementptr inbounds [1024 x i8], [1024 x i8]* @A, i64 0, i64 %iv.5
+
+ %v0 = load i8, i8* %in0
+ %v1 = load i8, i8* %in1
+ %v2 = load i8, i8* %in2
+ %v3 = load i8, i8* %in3
+ %v4 = load i8, i8* %in4
+ %v5 = load i8, i8* %in5
+
+ %reduce.add.0 = add i8 %v0, %v1
+ %reduce.add.1 = add i8 %reduce.add.0, %v2
+ %reduce.add.2 = add i8 %reduce.add.1, %v3
+ %reduce.add.3 = add i8 %reduce.add.2, %v4
+ %reduce.add.4 = add i8 %reduce.add.3, %v5
+
+ %out = getelementptr inbounds [1024 x i8], [1024 x i8]* @B, i64 0, i64 %iv.0
+ store i8 %reduce.add.4, i8* %out
+
+ %iv.next = add nuw nsw i64 %iv.0, 6
+ %cmp = icmp ult i64 %iv.next, 1024
+ br i1 %cmp, label %for.body, label %for.cond.cleanup
+
+for.cond.cleanup:
+ ret void
+}
diff --git a/llvm/test/Analysis/CostModel/X86/interleaved-store-i8-stride-6.ll b/llvm/test/Analysis/CostModel/X86/interleaved-store-i8-stride-6.ll
new file mode 100644
index 0000000000000..b6793e7a0c58c
--- /dev/null
+++ b/llvm/test/Analysis/CostModel/X86/interleaved-store-i8-stride-6.ll
@@ -0,0 +1,89 @@
+; RUN: opt -loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+sse2 --debug-only=loop-vectorize < %s 2>&1 | FileCheck %s --check-prefixes=CHECK,SSE2
+; RUN: opt -loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx --debug-only=loop-vectorize < %s 2>&1 | FileCheck %s --check-prefixes=CHECK,AVX1
+; RUN: opt -loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx2 --debug-only=loop-vectorize < %s 2>&1 | FileCheck %s --check-prefixes=CHECK,AVX2
+; RUN: opt -loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx512bw,+avx512vl --debug-only=loop-vectorize < %s 2>&1 | FileCheck %s --check-prefixes=CHECK,AVX512
+; REQUIRES: asserts
+
+target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-unknown-linux-gnu"
+
+ at A = global [1024 x i8] zeroinitializer, align 128
+ at B = global [1024 x i8] zeroinitializer, align 128
+
+; CHECK: LV: Checking a loop in "test"
+;
+; SSE2: LV: Found an estimated cost of 1 for VF 1 For instruction: store i8 %v5, i8* %out5, align 1
+; SSE2: LV: Found an estimated cost of 49 for VF 2 For instruction: store i8 %v5, i8* %out5, align 1
+; SSE2: LV: Found an estimated cost of 98 for VF 4 For instruction: store i8 %v5, i8* %out5, align 1
+; SSE2: LV: Found an estimated cost of 201 for VF 8 For instruction: store i8 %v5, i8* %out5, align 1
+; SSE2: LV: Found an estimated cost of 408 for VF 16 For instruction: store i8 %v5, i8* %out5, align 1
+;
+; AVX1: LV: Found an estimated cost of 1 for VF 1 For instruction: store i8 %v5, i8* %out5, align 1
+; AVX1: LV: Found an estimated cost of 27 for VF 2 For instruction: store i8 %v5, i8* %out5, align 1
+; AVX1: LV: Found an estimated cost of 54 for VF 4 For instruction: store i8 %v5, i8* %out5, align 1
+; AVX1: LV: Found an estimated cost of 101 for VF 8 For instruction: store i8 %v5, i8* %out5, align 1
+; AVX1: LV: Found an estimated cost of 201 for VF 16 For instruction: store i8 %v5, i8* %out5, align 1
+; AVX1: LV: Found an estimated cost of 498 for VF 32 For instruction: store i8 %v5, i8* %out5, align 1
+;
+; AVX2: LV: Found an estimated cost of 1 for VF 1 For instruction: store i8 %v5, i8* %out5, align 1
+; AVX2: LV: Found an estimated cost of 27 for VF 2 For instruction: store i8 %v5, i8* %out5, align 1
+; AVX2: LV: Found an estimated cost of 54 for VF 4 For instruction: store i8 %v5, i8* %out5, align 1
+; AVX2: LV: Found an estimated cost of 101 for VF 8 For instruction: store i8 %v5, i8* %out5, align 1
+; AVX2: LV: Found an estimated cost of 201 for VF 16 For instruction: store i8 %v5, i8* %out5, align 1
+; AVX2: LV: Found an estimated cost of 498 for VF 32 For instruction: store i8 %v5, i8* %out5, align 1
+;
+; AVX512: LV: Found an estimated cost of 1 for VF 1 For instruction: store i8 %v5, i8* %out5, align 1
+; AVX512: LV: Found an estimated cost of 18 for VF 2 For instruction: store i8 %v5, i8* %out5, align 1
+; AVX512: LV: Found an estimated cost of 38 for VF 4 For instruction: store i8 %v5, i8* %out5, align 1
+; AVX512: LV: Found an estimated cost of 98 for VF 8 For instruction: store i8 %v5, i8* %out5, align 1
+; AVX512: LV: Found an estimated cost of 197 for VF 16 For instruction: store i8 %v5, i8* %out5, align 1
+; AVX512: LV: Found an estimated cost of 295 for VF 32 For instruction: store i8 %v5, i8* %out5, align 1
+; AVX512: LV: Found an estimated cost of 591 for VF 64 For instruction: store i8 %v5, i8* %out5, align 1
+;
+; CHECK-NOT: LV: Found an estimated cost of {{[0-9]+}} for VF {{[0-9]+}} For instruction: store i8 %v5, i8* %out5, align 1
+
+define void @test() {
+entry:
+ br label %for.body
+
+for.body:
+ %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]
+
+ %iv.0 = add nuw nsw i64 %iv, 0
+ %iv.1 = add nuw nsw i64 %iv, 1
+ %iv.2 = add nuw nsw i64 %iv, 2
+ %iv.3 = add nuw nsw i64 %iv, 3
+ %iv.4 = add nuw nsw i64 %iv, 4
+ %iv.5 = add nuw nsw i64 %iv, 5
+
+ %in = getelementptr inbounds [1024 x i8], [1024 x i8]* @A, i64 0, i64 %iv.0
+ %v = load i8, i8* %in
+
+ %v0 = add i8 %v, 0
+ %v1 = add i8 %v, 1
+ %v2 = add i8 %v, 2
+ %v3 = add i8 %v, 3
+ %v4 = add i8 %v, 4
+ %v5 = add i8 %v, 5
+
+ %out0 = getelementptr inbounds [1024 x i8], [1024 x i8]* @B, i64 0, i64 %iv.0
+ %out1 = getelementptr inbounds [1024 x i8], [1024 x i8]* @B, i64 0, i64 %iv.1
+ %out2 = getelementptr inbounds [1024 x i8], [1024 x i8]* @B, i64 0, i64 %iv.2
+ %out3 = getelementptr inbounds [1024 x i8], [1024 x i8]* @B, i64 0, i64 %iv.3
+ %out4 = getelementptr inbounds [1024 x i8], [1024 x i8]* @B, i64 0, i64 %iv.4
+ %out5 = getelementptr inbounds [1024 x i8], [1024 x i8]* @B, i64 0, i64 %iv.5
+
+ store i8 %v0, i8* %out0
+ store i8 %v1, i8* %out1
+ store i8 %v2, i8* %out2
+ store i8 %v3, i8* %out3
+ store i8 %v4, i8* %out4
+ store i8 %v5, i8* %out5
+
+ %iv.next = add nuw nsw i64 %iv.0, 6
+ %cmp = icmp ult i64 %iv.next, 1024
+ br i1 %cmp, label %for.body, label %for.cond.cleanup
+
+for.cond.cleanup:
+ ret void
+}
More information about the llvm-commits
mailing list