diff --git a/llvm/test/Transforms/InstCombine/signmask-of-sext-vs-of-shl-of-zext.ll b/llvm/test/Transforms/InstCombine/signmask-of-sext-vs-of-shl-of-zext.ll new file mode 100644 index 000000000000..f529023e2a0d --- /dev/null +++ b/llvm/test/Transforms/InstCombine/signmask-of-sext-vs-of-shl-of-zext.ll @@ -0,0 +1,174 @@ +; NOTE: Assertions have been autogenerated by utils/update_test_checks.py +; RUN: opt < %s -instcombine -S | FileCheck %s + +; If we zero-extend some value, and then immediately left-shift-out all the new +; sign bits, and apply a mask to keep only the sign bit (which is the original +; sign bit from before zero-extension), we might as well just sign-extend +; and apply the same signmask. + +declare void @use32(i32) + +; Basic pattern + +define i32 @t0(i16 %x) { +; CHECK-LABEL: @t0( +; CHECK-NEXT: [[I0:%.*]] = zext i16 [[X:%.*]] to i32 +; CHECK-NEXT: [[I1:%.*]] = shl nuw i32 [[I0]], 16 +; CHECK-NEXT: [[R:%.*]] = and i32 [[I1]], -2147483648 +; CHECK-NEXT: ret i32 [[R]] +; + %i0 = zext i16 %x to i32 + %i1 = shl i32 %i0, 16 + %r = and i32 %i1, -2147483648 + ret i32 %r +} +define i32 @t1(i8 %x) { +; CHECK-LABEL: @t1( +; CHECK-NEXT: [[I0:%.*]] = zext i8 [[X:%.*]] to i32 +; CHECK-NEXT: [[I1:%.*]] = shl nuw i32 [[I0]], 24 +; CHECK-NEXT: [[R:%.*]] = and i32 [[I1]], -2147483648 +; CHECK-NEXT: ret i32 [[R]] +; + %i0 = zext i8 %x to i32 + %i1 = shl i32 %i0, 24 + %r = and i32 %i1, -2147483648 + ret i32 %r +} + +; Some negative tests + +define i32 @n2(i16 %x) { +; CHECK-LABEL: @n2( +; CHECK-NEXT: ret i32 0 +; + %i0 = zext i16 %x to i32 + %i1 = shl i32 %i0, 15 ; undershifting + %r = and i32 %i1, -2147483648 + ret i32 %r +} +define i32 @n3(i16 %x) { +; CHECK-LABEL: @n3( +; CHECK-NEXT: [[I0:%.*]] = zext i16 [[X:%.*]] to i32 +; CHECK-NEXT: [[I1:%.*]] = shl i32 [[I0]], 17 +; CHECK-NEXT: [[R:%.*]] = and i32 [[I1]], -2147483648 +; CHECK-NEXT: ret i32 [[R]] +; + %i0 = zext i16 %x to i32 + %i1 = shl i32 %i0, 17 ; overshifting + %r = and i32 %i1, -2147483648 + ret i32 %r +} +define i32 @n4(i16 %x) { +; CHECK-LABEL: @n4( +; CHECK-NEXT: [[I0:%.*]] = zext i16 [[X:%.*]] to i32 +; CHECK-NEXT: [[I1:%.*]] = shl nuw i32 [[I0]], 16 +; CHECK-NEXT: [[R:%.*]] = and i32 [[I1]], -1073741824 +; CHECK-NEXT: ret i32 [[R]] +; + %i0 = zext i16 %x to i32 + %i1 = shl i32 %i0, 16 + %r = and i32 %i1, 3221225472 ; not a sign bit + ret i32 %r +} + +; Extra-use tests + +define i32 @t5(i16 %x) { +; CHECK-LABEL: @t5( +; CHECK-NEXT: [[I0:%.*]] = zext i16 [[X:%.*]] to i32 +; CHECK-NEXT: call void @use32(i32 [[I0]]) +; CHECK-NEXT: [[I1:%.*]] = shl nuw i32 [[I0]], 16 +; CHECK-NEXT: [[R:%.*]] = and i32 [[I1]], -2147483648 +; CHECK-NEXT: ret i32 [[R]] +; + %i0 = zext i16 %x to i32 + call void @use32(i32 %i0) + %i1 = shl i32 %i0, 16 + %r = and i32 %i1, -2147483648 + ret i32 %r +} +define i32 @n6(i16 %x) { +; CHECK-LABEL: @n6( +; CHECK-NEXT: [[I0:%.*]] = zext i16 [[X:%.*]] to i32 +; CHECK-NEXT: [[I1:%.*]] = shl nuw i32 [[I0]], 16 +; CHECK-NEXT: call void @use32(i32 [[I1]]) +; CHECK-NEXT: [[R:%.*]] = and i32 [[I1]], -2147483648 +; CHECK-NEXT: ret i32 [[R]] +; + %i0 = zext i16 %x to i32 + %i1 = shl i32 %i0, 16 ; not one-use + call void @use32(i32 %i1) + %r = and i32 %i1, -2147483648 + ret i32 %r +} +define i32 @n7(i16 %x) { +; CHECK-LABEL: @n7( +; CHECK-NEXT: [[I0:%.*]] = zext i16 [[X:%.*]] to i32 +; CHECK-NEXT: call void @use32(i32 [[I0]]) +; CHECK-NEXT: [[I1:%.*]] = shl nuw i32 [[I0]], 16 +; CHECK-NEXT: call void @use32(i32 [[I1]]) +; CHECK-NEXT: [[R:%.*]] = and i32 [[I1]], -2147483648 +; CHECK-NEXT: ret i32 [[R]] +; + %i0 = zext i16 %x to i32 ; not one-use + call void @use32(i32 %i0) + %i1 = shl i32 %i0, 16 ; not one-use + call void @use32(i32 %i1) + %r = and i32 %i1, -2147483648 + ret i32 %r +} + +; Some vector tests + +define <2 x i32> @t8(<2 x i16> %x) { +; CHECK-LABEL: @t8( +; CHECK-NEXT: [[I0:%.*]] = zext <2 x i16> [[X:%.*]] to <2 x i32> +; CHECK-NEXT: [[I1:%.*]] = shl nuw <2 x i32> [[I0]], +; CHECK-NEXT: [[R:%.*]] = and <2 x i32> [[I1]], +; CHECK-NEXT: ret <2 x i32> [[R]] +; + %i0 = zext <2 x i16> %x to <2 x i32> + %i1 = shl <2 x i32> %i0, + %r = and <2 x i32> %i1, + ret <2 x i32> %r +} +define <2 x i32> @t9(<2 x i16> %x) { +; CHECK-LABEL: @t9( +; CHECK-NEXT: [[I0:%.*]] = zext <2 x i16> [[X:%.*]] to <2 x i32> +; CHECK-NEXT: [[I1:%.*]] = shl <2 x i32> [[I0]], +; CHECK-NEXT: [[R:%.*]] = and <2 x i32> [[I1]], +; CHECK-NEXT: ret <2 x i32> [[R]] +; + %i0 = zext <2 x i16> %x to <2 x i32> + %i1 = shl <2 x i32> %i0, + %r = and <2 x i32> %i1, + ; Here undef can be propagated into the mask. + ret <2 x i32> %r +} +define <2 x i32> @t10(<2 x i16> %x) { +; CHECK-LABEL: @t10( +; CHECK-NEXT: [[I0:%.*]] = zext <2 x i16> [[X:%.*]] to <2 x i32> +; CHECK-NEXT: [[I1:%.*]] = shl nuw <2 x i32> [[I0]], +; CHECK-NEXT: [[R:%.*]] = and <2 x i32> [[I1]], +; CHECK-NEXT: ret <2 x i32> [[R]] +; + %i0 = zext <2 x i16> %x to <2 x i32> + %i1 = shl <2 x i32> %i0, + %r = and <2 x i32> %i1, + ; CAREFUL! We can't keep undef mask here, since high bits are no longer zero, + ; we must sanitize it to 0. + ret <2 x i32> %r +} +define <2 x i32> @t11(<2 x i16> %x) { +; CHECK-LABEL: @t11( +; CHECK-NEXT: [[I0:%.*]] = zext <2 x i16> [[X:%.*]] to <2 x i32> +; CHECK-NEXT: [[I1:%.*]] = shl <2 x i32> [[I0]], +; CHECK-NEXT: [[R:%.*]] = and <2 x i32> [[I1]], +; CHECK-NEXT: ret <2 x i32> [[R]] +; + %i0 = zext <2 x i16> %x to <2 x i32> + %i1 = shl <2 x i32> %i0, + %r = and <2 x i32> %i1, + ; Here undef mask is fine. + ret <2 x i32> %r +}