Subzero: Randomly insert nops.

tests_lit/llvm2ice_tests/nop-insertion.ll

Index: tests_lit/llvm2ice_tests/nop-insertion.ll
diff --git a/tests_lit/llvm2ice_tests/nop-insertion.ll b/tests_lit/llvm2ice_tests/nop-insertion.ll
new file mode 100644
index 0000000000000000000000000000000000000000..2a48d25b5ff9c29cd2ea36fd8e89fab836228fdb
--- /dev/null
+++ b/tests_lit/llvm2ice_tests/nop-insertion.ll
@@ -0,0 +1,95 @@
+; This is a smoke test of nop insertion.
+
+; RUN: %llvm2ice -rng-seed=1 -nop-insertion -nop-insertion-percentage=50 \
+; RUN:    -max-nops-per-instruction=1 %s | FileCheck %s --check-prefix=PROB50
+; RUN: %llvm2ice -rng-seed=1 -nop-insertion -nop-insertion-percentage=90 \
+; RUN:    -max-nops-per-instruction=1 %s | FileCheck %s --check-prefix=PROB90
+; RUN: %llvm2ice -rng-seed=1 -nop-insertion -nop-insertion-percentage=50 \
+; RUN:    -max-nops-per-instruction=2 %s | FileCheck %s --check-prefix=MAXNOPS2
+
+define <4 x i32> @mul_v4i32(<4 x i32> %a, <4 x i32> %b) {
+entry:
+  %res = mul <4 x i32> %a, %b
+  ret <4 x i32> %res
+; PROB50-LABEL: mul_v4i32:
+; PROB50: sub esp, 60
+; PROB50: nop # variant = 4
+; PROB50: movups xmmword ptr [esp+32], xmm0
+; PROB50: movups xmmword ptr [esp+16], xmm1
+; PROB50: nop # variant = 0
+; PROB50: movups xmm0, xmmword ptr [esp+32]
+; PROB50: nop # variant = 4
+; PROB50: pshufd xmm1, xmmword ptr [esp+32], 49
+; PROB50: pshufd xmm2, xmmword ptr [esp+16], 49
+; PROB50: pmuludq xmm0, xmmword ptr [esp+16]
+; PROB50: pmuludq xmm1, xmm2
+; PROB50: nop # variant = 0
+; PROB50: shufps xmm0, xmm1, 136
+; PROB50: pshufd xmm3, xmm0, 216
+; PROB50: nop # variant = 2
+; PROB50: movups xmmword ptr [esp], xmm3
+; PROB50: movups xmm0, xmmword ptr [esp]
+; PROB50: add esp, 60
+; PROB50: nop # variant = 0
+; PROB50: ret
+; PROB50: nop # variant = 3
+
+; PROB90-LABEL: mul_v4i32:
+; PROB90: sub esp, 60
+; PROB90: nop # variant = 4
+; PROB90: movups xmmword ptr [esp+32], xmm0
+; PROB90: nop # variant = 3
+; PROB90: movups xmmword ptr [esp+16], xmm1
+; PROB90: nop # variant = 2
+; PROB90: movups xmm0, xmmword ptr [esp+32]
+; PROB90: nop # variant = 3
+; PROB90: pshufd xmm1, xmmword ptr [esp+32], 49
+; PROB90: nop # variant = 4
+; PROB90: pshufd xmm2, xmmword ptr [esp+16], 49
+; PROB90: nop # variant = 0
+; PROB90: pmuludq xmm0, xmmword ptr [esp+16]
+; PROB90: nop # variant = 2
+; PROB90: pmuludq xmm1, xmm2
+; PROB90: nop # variant = 3
+; PROB90: shufps xmm0, xmm1, 136
+; PROB90: nop # variant = 4
+; PROB90: pshufd xmm3, xmm0, 216
+; PROB90: nop # variant = 2
+; PROB90: movups xmmword ptr [esp], xmm3
+; PROB90: nop # variant = 4
+; PROB90: movups xmm0, xmmword ptr [esp]
+; PROB90: nop # variant = 2
+; PROB90: add esp, 60
+; PROB90: nop # variant = 3
+; PROB90: ret
+; PROB90: nop # variant = 3
+; PROB90: nop # variant = 0

[jvoung (off chromium), 2014/08/15 20:18:11]

Interesting, I guess counted pseudo-insts like FakeUse(?) 

[wala, 2014/08/15 22:19:46]

Yes. That's probably not what we want. Fixed.

+
+; MAXNOPS2-LABEL: mul_v4i32:
+; MAXNOPS2: sub	esp, 60
+; MAXNOPS2: nop	# variant = 4
+; MAXNOPS2: movups	xmmword ptr [esp+32], xmm0
+; MAXNOPS2: nop	# variant = 0
+; MAXNOPS2: nop	# variant = 4
+; MAXNOPS2: movups	xmmword ptr [esp+16], xmm1
+; MAXNOPS2: movups	xmm0, xmmword ptr [esp+32]
+; MAXNOPS2: nop	# variant = 0
+; MAXNOPS2: pshufd	xmm1, xmmword ptr [esp+32], 49
+; MAXNOPS2: nop	# variant = 2
+; MAXNOPS2: pshufd	xmm2, xmmword ptr [esp+16], 49
+; MAXNOPS2: pmuludq	xmm0, xmmword ptr [esp+16]
+; MAXNOPS2: nop	# variant = 0
+; MAXNOPS2: nop	# variant = 3
+; MAXNOPS2: pmuludq	xmm1, xmm2
+; MAXNOPS2: shufps	xmm0, xmm1, 136
+; MAXNOPS2: pshufd	xmm3, xmm0, 216
+; MAXNOPS2: nop	# variant = 3
+; MAXNOPS2: movups	xmmword ptr [esp], xmm3
+; MAXNOPS2: nop	# variant = 0
+; MAXNOPS2: movups	xmm0, xmmword ptr [esp]
+; MAXNOPS2: nop	# variant = 2
+; MAXNOPS2: add	esp, 60
+; MAXNOPS2: nop	# variant = 4
+; MAXNOPS2: ret
+; MAXNOPS2: nop	# variant = 2
+}