Add load and store instructions to Subzero bitcode reader.

tests_lit/reader_tests/load.ll

+; Test if we can read load instructions.
+
+; RUN: llvm-as < %s | pnacl-freeze \
+; RUN:              | %llvm2ice -notranslate -verbose=inst -build-on-read \
+; RUN:                -allow-pnacl-reader-error-recovery \
+; RUN:              | FileCheck %s
+
+define i32 @load_i8(i32 %addr) {
+entry:
+  %addr_i8 = inttoptr i32 %addr to i8*
+  %v = load i8* %addr_i8, align 1
+  %r = sext i8 %v to i32
+  ret i32 %r
+
+; CHECK:      __0:
+; CHECK-NEXT:   %__1 = load i8* %__0, align 1
+; CHECK-NEXT:   %__2 = sext i8 %__1 to i32
+; CHECK-NEXT:   ret i32 %__2
+}
+
+define i32 @load_i16(i32 %addr) {
+entry:
+  %addr_i16 = inttoptr i32 %addr to i16*
+  %v = load i16* %addr_i16, align 1
+  %r = sext i16 %v to i32
+  ret i32 %r
+
+; CHECK:      __0:
+; CHECK-NEXT:   %__1 = load i16* %__0, align 1
+; CHECK-NEXT:   %__2 = sext i16 %__1 to i32
+; CHECK-NEXT:   ret i32 %__2
+}
+
+define i32 @load_i32(i32 %addr) {
+entry:
+  %addr_i32 = inttoptr i32 %addr to i32*
+  %v = load i32* %addr_i32, align 1
+  ret i32 %v
+
+; CHECK:      __0:
+; CHECK-NEXT:   %__1 = load i32* %__0, align 1
+; CHECK-NEXT:   ret i32 %__1
+}
+
+define i64 @load_i64(i32 %addr) {
+entry:
+  %addr_i64 = inttoptr i32 %addr to i64*
+  %v = load i64* %addr_i64, align 1
+  ret i64 %v
+
+; CHECK:      __0:
+; CHECK-NEXT:   %__1 = load i64* %__0, align 1
+; CHECK-NEXT:   ret i64 %__1
+}
+
+define float @load_float_a1(i32 %addr) {
+entry:
+  %addr_float = inttoptr i32 %addr to float*
+  %v = load float* %addr_float, align 1
+  ret float %v
+
+; TODO(kschimpf) Fix load alignment in ICE to allow non-default.
+
+; CHECK:      __0:
+; CHECK-NEXT:   %__1 = load float* %__0, align 4
+; CHECK-NEXT:   ret float %__1
+}
+
+
+define float @load_float_a4(i32 %addr) {
+entry:
+  %addr_float = inttoptr i32 %addr to float*
+  %v = load float* %addr_float, align 4
+  ret float %v
+
+; CHECK:       __0:
+; CHECK-NEXT:   %__1 = load float* %__0, align 4
+; CHECK-NEXT:   ret float %__1
+}
+
+define double @load_double_a1(i32 %addr) {
+entry:
+  %addr_double = inttoptr i32 %addr to double*
+  %v = load double* %addr_double, align 1
+  ret double %v
+
+; TODO(kschimpf) Fix load alignment in ICE to allow non-default.
+
+; CHECK:      __0:
+; CHECK-NEXT:   %__1 = load double* %__0, align 8
+; CHECK-NEXT:   ret double %__1
+}
+
+
+define double @load_double_a8(i32 %addr) {
+entry:
+  %addr_double = inttoptr i32 %addr to double*
+  %v = load double* %addr_double, align 8
+  ret double %v
+
+; CHECK:      __0:
+; CHECK-NEXT:   %__1 = load double* %__0, align 8
+; CHECK-NEXT:   ret double %__1
+}
+
+define <16 x i8> @load_v16xI8(i32 %addr) {
+entry:
+  %addr_v16xI8 = inttoptr i32 %addr to <16 x i8>*
+  %v = load <16 x i8>* %addr_v16xI8, align 1
+  ret <16 x i8> %v
+
+; CHECK:      __0:
+; CHECK-NEXT:   %__1 = load <16 x i8>* %__0, align 1
+; CHECK-NEXT:   ret <16 x i8> %__1
+}
+
+define <8 x i16> @load_v8xI16(i32 %addr) {
+entry:
+  %addr_v8xI16 = inttoptr i32 %addr to <8 x i16>*
+  %v = load <8 x i16>* %addr_v8xI16, align 2
+  ret <8 x i16> %v
+
+; CHECK:      __0:
+; CHECK-NEXT:   %__1 = load <8 x i16>* %__0, align 2
+; CHECK-NEXT:   ret <8 x i16> %__1
+}
+
+define <4 x i32> @load_v4xI32(i32 %addr) {
+entry:
+  %addr_v4xI32 = inttoptr i32 %addr to <4 x i32>*
+  %v = load <4 x i32>* %addr_v4xI32, align 4
+  ret <4 x i32> %v
+
+; CHECK:      __0:
+; CHECK-NEXT:   %__1 = load <4 x i32>* %__0, align 4
+; CHECK-NEXT:   ret <4 x i32> %__1
+}
+
+define <4 x float> @load_v4xFloat(i32 %addr) {
+entry:
+  %addr_v4xFloat = inttoptr i32 %addr to <4 x float>*
+  %v = load <4 x float>* %addr_v4xFloat, align 4
+  ret <4 x float> %v
+
+; CHECK:      __0:
+; CHECK-NEXT:   %__1 = load <4 x float>* %__0, align 4
+; CHECK-NEXT:   ret <4 x float> %__1
+}
+