Add vector insert/extract instructions to Subzero bitcode reader.

tests_lit/reader_tests/vector.ll

+; Tests insertelement and extractelement vector instructions.
+
+
+; RUN: llvm-as < %s | pnacl-freeze \
+; RUN:              | %llvm2ice -notranslate -verbose=inst -build-on-read \

[jvoung (off chromium), 2014/09/02 21:06:46]

I don't remember what this % was for, but is this % before the llvm2ice
necessary anymore, after some of the lit.cfg changes?

[Jim Stichnoth, 2014/09/02 21:23:32]

I think it's still necessary, as it's not part of LLVM_BIN_PATH.

[Karl, 2014/09/02 22:45:11]

I agree with Jim.

+; RUN:                -allow-pnacl-reader-error-recovery \
+; RUN:              | FileCheck %s
+
+define void @ExtractV4xi1(<4 x i1> %v, i32 %i) {
+  %e = extractelement <4 x i1> %v, i32 %i

[jvoung (off chromium), 2014/09/02 21:06:46]

Technically, PNaCl only accepts constant indices for insert/extract so that they
are known to be in bounds.

[Jim Stichnoth, 2014/09/02 21:23:32]

I don't think Karl's code supports constants yet. :)

[Karl, 2014/09/02 22:45:11]

Adding comments to clarify (both in code and tests).

[jvoung (off chromium), 2014/09/03 16:23:30]

Ah right...

+  ret void
+}
+
+; CHECK:      define void @ExtractV4xi1(<4 x i1> %__0, i32 %__1) {
+; CHECK-NEXT: __0:
+; CHECK-NEXT:   %__2 = extractelement <4 x i1> %__0, i32 %__1
+; CHECK-NEXT:   ret void
+; CHECK-NEXT: }
+
+define void @ExtractV8xi1(<8 x i1> %v, i32 %i) {
+  %e = extractelement <8 x i1> %v, i32 %i
+  ret void
+}
+
+; CHECK-NEXT: define void @ExtractV8xi1(<8 x i1> %__0, i32 %__1) {
+; CHECK-NEXT: __0:
+; CHECK-NEXT:   %__2 = extractelement <8 x i1> %__0, i32 %__1
+; CHECK-NEXT:   ret void
+; CHECK-NEXT: }
+
+define void @ExtractV16xi1(<16 x i1> %v, i32 %i) {
+  %e = extractelement <16 x i1> %v, i32 %i
+  ret void
+}
+
+; CHECK-NEXT: define void @ExtractV16xi1(<16 x i1> %__0, i32 %__1) {
+; CHECK-NEXT: __0:
+; CHECK-NEXT:   %__2 = extractelement <16 x i1> %__0, i32 %__1
+; CHECK-NEXT:   ret void
+; CHECK-NEXT: }
+
+define void @ExtractV16xi8(<16 x i8> %v, i32 %i) {
+  %e = extractelement <16 x i8> %v, i32 %i
+  ret void
+}
+
+; CHECK-NEXT: define void @ExtractV16xi8(<16 x i8> %__0, i32 %__1) {
+; CHECK-NEXT: __0:
+; CHECK-NEXT:   %__2 = extractelement <16 x i8> %__0, i32 %__1
+; CHECK-NEXT:   ret void
+; CHECK-NEXT: }
+
+define void @ExtractV8xi16(<8 x i16> %v, i32 %i) {
+  %e = extractelement <8 x i16> %v, i32 %i
+  ret void
+}
+
+; CHECK-NEXT: define void @ExtractV8xi16(<8 x i16> %__0, i32 %__1) {
+; CHECK-NEXT: __0:
+; CHECK-NEXT:   %__2 = extractelement <8 x i16> %__0, i32 %__1
+; CHECK-NEXT:   ret void
+; CHECK-NEXT: }
+
+define i32 @ExtractV4xi32(<4 x i32> %v, i32 %i) {
+  %e = extractelement <4 x i32> %v, i32 %i
+  ret i32 %e
+}
+
+; CHECK-NEXT: define i32 @ExtractV4xi32(<4 x i32> %__0, i32 %__1) {
+; CHECK-NEXT: __0:
+; CHECK-NEXT:   %__2 = extractelement <4 x i32> %__0, i32 %__1
+; CHECK-NEXT:   ret i32 %__2
+; CHECK-NEXT: }
+
+define float @ExtractV4xfloat(<4 x float> %v, i32 %i) {
+  %e = extractelement <4 x float> %v, i32 %i
+  ret float %e
+}
+
+; CHECK-NEXT: define float @ExtractV4xfloat(<4 x float> %__0, i32 %__1) {
+; CHECK-NEXT: __0:
+; CHECK-NEXT:   %__2 = extractelement <4 x float> %__0, i32 %__1
+; CHECK-NEXT:   ret float %__2
+; CHECK-NEXT: }
+
+define <4 x i1> @InsertV4xi1(<4 x i1> %v, i32 %pe, i32 %i) {
+  %e = trunc i32 %pe to i1
+  %r = insertelement <4 x i1> %v, i1 %e, i32 %i
+  ret <4 x i1> %r
+}
+
+; CHECK-NEXT: define <4 x i1> @InsertV4xi1(<4 x i1> %__0, i32 %__1, i32 %__2) {
+; CHECK-NEXT: __0:
+; CHECK-NEXT:   %__3 = trunc i32 %__1 to i1
+; CHECK-NEXT:   %__4 = insertelement <4 x i1> %__0, i1 %__3, i32 %__2
+; CHECK-NEXT:   ret i1 %__4
+; CHECK-NEXT: }
+
+define <8 x i1> @InsertV8xi1(<8 x i1> %v, i32 %pe, i32 %i) {
+  %e = trunc i32 %pe to i1
+  %r = insertelement <8 x i1> %v, i1 %e, i32 %i
+  ret <8 x i1> %r
+}
+
+; CHECK-NEXT: define <8 x i1> @InsertV8xi1(<8 x i1> %__0, i32 %__1, i32 %__2) {
+; CHECK-NEXT: __0:
+; CHECK-NEXT:   %__3 = trunc i32 %__1 to i1
+; CHECK-NEXT:   %__4 = insertelement <8 x i1> %__0, i1 %__3, i32 %__2
+; CHECK-NEXT:   ret i1 %__4
+; CHECK-NEXT: }
+
+define <16 x i1> @InsertV16xi1(<16 x i1> %v, i32 %pe, i32 %i) {
+  %e = trunc i32 %pe to i1
+  %r = insertelement <16 x i1> %v, i1 %e, i32 %i
+  ret <16 x i1> %r
+}
+
+; CHECK-NEXT: define <16 x i1> @InsertV16xi1(<16 x i1> %__0, i32 %__1, i32 %__2) {
+; CHECK-NEXT: __0:
+; CHECK-NEXT:   %__3 = trunc i32 %__1 to i1
+; CHECK-NEXT:   %__4 = insertelement <16 x i1> %__0, i1 %__3, i32 %__2
+; CHECK-NEXT:   ret i1 %__4
+; CHECK-NEXT: }
+
+define <16 x i8> @InsertV16xi8(<16 x i8> %v, i32 %pe, i32 %i) {
+  %e = trunc i32 %pe to i8
+  %r = insertelement <16 x i8> %v, i8 %e, i32 %i
+  ret <16 x i8> %r
+}
+
+; CHECK-NEXT: define <16 x i8> @InsertV16xi8(<16 x i8> %__0, i32 %__1, i32 %__2) {
+; CHECK-NEXT: __0:
+; CHECK-NEXT:   %__3 = trunc i32 %__1 to i8
+; CHECK-NEXT:   %__4 = insertelement <16 x i8> %__0, i8 %__3, i32 %__2
+; CHECK-NEXT:   ret i8 %__4
+; CHECK-NEXT: }
+
+define <8 x i16> @InsertV8xi16(<8 x i16> %v, i32 %pe, i32 %i) {
+  %e = trunc i32 %pe to i16
+  %r = insertelement <8 x i16> %v, i16 %e, i32 %i
+  ret <8 x i16> %r
+}
+
+; CHECK-NEXT: define <8 x i16> @InsertV8xi16(<8 x i16> %__0, i32 %__1, i32 %__2) {
+; CHECK-NEXT: __0:
+; CHECK-NEXT:   %__3 = trunc i32 %__1 to i16
+; CHECK-NEXT:   %__4 = insertelement <8 x i16> %__0, i16 %__3, i32 %__2
+; CHECK-NEXT:   ret i16 %__4
+; CHECK-NEXT: }
+
+define <4 x i32> @InsertV16xi32(<4 x i32> %v, i32 %e, i32 %i) {
+  %r = insertelement <4 x i32> %v, i32 %e, i32 %i
+  ret <4 x i32> %r
+}
+
+; CHECK-NEXT: define <4 x i32> @InsertV16xi32(<4 x i32> %__0, i32 %__1, i32 %__2) {
+; CHECK-NEXT: __0:
+; CHECK-NEXT:   %__3 = insertelement <4 x i32> %__0, i32 %__1, i32 %__2
+; CHECK-NEXT:   ret i32 %__3
+; CHECK-NEXT: }
+
+define <4 x float> @InsertV16xfloat(<4 x float> %v, float %e, i32 %i) {
+  %r = insertelement <4 x float> %v, float %e, i32 %i
+  ret <4 x float> %r
+}
+
+; CHECK-NEXT: define <4 x float> @InsertV16xfloat(<4 x float> %__0, float %__1, i32 %__2) {
+; CHECK-NEXT: __0:
+; CHECK-NEXT:   %__3 = insertelement <4 x float> %__0, float %__1, i32 %__2
+; CHECK-NEXT:   ret float %__3
+; CHECK-NEXT: }