RSForEachExpand: Improve getelementptr instruction generation.
(1) The old pattern, when accessing fields of the driver info
structure, was:
%base = gep %DriverInfo *%info, 0, <base array index>
%addr = gep [N x i8*]* %base, 0, <index>
This change folds the above into:
%addr = gep %DriverInfo *%info, 0, <base array index>, <index>
which simplifies both way that we generate the GEP instructions as
well as the resulting bitcode.
(2) In addition, GEPs of input pointers and instep sizes can be
hoisted out of the loop body.
(3) Finally, mark all GEPs into the DriverInfo structure as inbounds,
because we know they are.
Change-Id: Icfabbdb464af8865c2a4623d967bbb63ef711680
diff --git a/tests/README.lit b/tests/README.lit
new file mode 100644
index 0000000..16fa305
--- /dev/null
+++ b/tests/README.lit
@@ -0,0 +1,7 @@
+To run the libbcc lit tests:
+ * Ensure `llvm-rs-as` is built, either by doing a top-level `make
+ checkbuild` or by doing `mm` from frameworks/compile/slang.
+ * Ensure that LLVM and libbcc are built with
+ `FORCE_BUILD_LLVM_COMPONENTS=true`.
+ * Ensure `opt` is built from external/llvm, either by top-level `make
+ checkbuild` or by doing `mm` from external/llvm.
diff --git a/tests/libbcc/getelementptr.ll b/tests/libbcc/getelementptr.ll
new file mode 100644
index 0000000..6f3e175
--- /dev/null
+++ b/tests/libbcc/getelementptr.ll
@@ -0,0 +1,74 @@
+; This checks that RSForEachExpand generates getelementptr
+; instructions into the driver info structure as expected - namely,
+; that they index into the right positions of the structure and that
+; the instructions that are generated are in the loop header.
+
+; RUN: opt -load libbcc.so -foreachexp -S < %s | FileCheck %s
+
+; ModuleID = 'test_getelementptr.bc'
+target datalayout = "e-m:e-i64:64-i128:128-n32:64-S128"
+target triple = "aarch64-none-linux-gnueabi"
+
+; Declarations expected by the expansion pass.
+declare void @_Z14rsGetElementAt13rs_allocationj()
+declare void @_Z14rsGetElementAt13rs_allocationjj()
+declare void @_Z14rsGetElementAt13rs_allocationjjj()
+declare void @_Z14rsSetElementAt13rs_allocationPvj()
+declare void @_Z14rsSetElementAt13rs_allocationPvjj()
+declare void @_Z14rsSetElementAt13rs_allocationPvjjj()
+declare void @_Z25rsGetElementAtYuv_uchar_Y13rs_allocationjj()
+declare void @_Z25rsGetElementAtYuv_uchar_U13rs_allocationjj()
+declare void @_Z25rsGetElementAtYuv_uchar_V13rs_allocationjj()
+
+; Old-style kernel
+define void @root(i32* nocapture %ain, i32* nocapture %out, i32 %x, i32 %y, i32 %z) {
+ ret void
+; CHECK: define void @root.expand(%RsExpandKernelDriverInfoPfx* %p, i32 %x1, i32 %x2, i32 %outstep)
+; CHECK: Begin:
+; CHECK: %instep_addr.gep = getelementptr inbounds %RsExpandKernelDriverInfoPfx, %RsExpandKernelDriverInfoPfx* %p, i32 0, i32 1, i32 0
+; CHECK: load i32, i32* %instep_addr.gep
+; CHECK: %input_buf.gep = getelementptr inbounds %RsExpandKernelDriverInfoPfx, %RsExpandKernelDriverInfoPfx* %p, i32 0, i32 0, i32 0
+; CHECK: load i8*, i8** %input_buf.gep
+; CHECK: %out_buf.gep = getelementptr inbounds %RsExpandKernelDriverInfoPfx, %RsExpandKernelDriverInfoPfx* %p, i32 0, i32 3, i32 0
+; CHECK: load i8*, i8** %out_buf.gep
+; CHECK: %Y.gep = getelementptr inbounds %RsExpandKernelDriverInfoPfx, %RsExpandKernelDriverInfoPfx* %p, i32 0, i32 7, i32 1
+; CHECK: load i32, i32* %Y.gep
+; CHECK: %Z.gep = getelementptr inbounds %RsExpandKernelDriverInfoPfx, %RsExpandKernelDriverInfoPfx* %p, i32 0, i32 7, i32 2
+; CHECK: load i32, i32* %Z.gep
+; CHECK: Loop:
+}
+
+; New style kernel with multiple inputs
+define i32 @foo(i32 %in0, i32 %in1, i32 %x, i32 %y, i32 %z) {
+ ret i32 0
+; CHECK: define void @foo.expand(%RsExpandKernelDriverInfoPfx* %p, i32 %x1, i32 %x2, i32 %outstep)
+; CHECK: Begin:
+; CHECK: %out_buf.gep = getelementptr inbounds %RsExpandKernelDriverInfoPfx, %RsExpandKernelDriverInfoPfx* %p, i32 0, i32 3, i32 0
+; CHECK: load i8*, i8** %out_buf.gep
+; CHECK: %Y.gep = getelementptr inbounds %RsExpandKernelDriverInfoPfx, %RsExpandKernelDriverInfoPfx* %p, i32 0, i32 7, i32 1
+; CHECK: load i32, i32* %Y.gep
+; CHECK: %Z.gep = getelementptr inbounds %RsExpandKernelDriverInfoPfx, %RsExpandKernelDriverInfoPfx* %p, i32 0, i32 7, i32 2
+; CHECK: load i32, i32* %Z.gep
+; CHECK: %instep_addr.gep = getelementptr inbounds %RsExpandKernelDriverInfoPfx, %RsExpandKernelDriverInfoPfx* %p, i32 0, i32 1, i32 0
+; CHECK: load i32, i32* %instep_addr.gep
+; CHECK: %input_buf.gep = getelementptr inbounds %RsExpandKernelDriverInfoPfx, %RsExpandKernelDriverInfoPfx* %p, i32 0, i32 0, i32 0
+; CHECK: load i8*, i8** %input_buf.gep
+; CHECK: %instep_addr.gep1 = getelementptr inbounds %RsExpandKernelDriverInfoPfx, %RsExpandKernelDriverInfoPfx* %p, i32 0, i32 1, i32 1
+; CHECK: load i32, i32* %instep_addr.gep1
+; CHECK: %input_buf.gep3 = getelementptr inbounds %RsExpandKernelDriverInfoPfx, %RsExpandKernelDriverInfoPfx* %p, i32 0, i32 0, i32 1
+; CHECK: load i8*, i8** %input_buf.gep3
+; CHECK: Loop:
+}
+
+!llvm.ident = !{!0}
+!\23pragma = !{!1, !2}
+!\23rs_export_foreach_name = !{!3, !4}
+!\23rs_export_foreach = !{!5, !6}
+
+!0 = !{!"clang version 3.6 "}
+!1 = !{!"version", !"1"}
+!2 = !{!"java_package_name", !"foo"}
+!3 = !{!"root"}
+!4 = !{!"foo"}
+!5 = !{!"91"}
+!6 = !{!"123"}
diff --git a/tests/libbcc/lit.cfg b/tests/libbcc/lit.cfg
index 5b3c749..109a9d7 100644
--- a/tests/libbcc/lit.cfg
+++ b/tests/libbcc/lit.cfg
@@ -24,17 +24,20 @@
# test_source_root: The path where tests are located (default is the test suite
# root).
config.test_source_root = None
-config.test_exec_root = os.path.join(ANDROID_HOST_OUT, 'tests', 'bcinfo')
+config.test_exec_root = os.path.join(ANDROID_HOST_OUT, 'tests', 'libbcc')
-tools_dir = os.path.join(ANDROID_HOST_OUT, 'bin')
+tools_dir = os.pathsep.join([os.path.join(ANDROID_HOST_OUT, 'bin'),
+ os.path.join(ANDROID_HOST_OUT, 'lib64')])
# Based on LLVM's lit.cfg: "For each occurrence of an llvm tool name
# as its own word, replace it with the full path to the build directory
# holding that tool."
for pattern in [r"\bFileCheck\b",
r"\bllvm-rs-as\b",
- r"\bbcinfo\b"]:
- tool_match = re.match(r"^(\\)?((\| )?)\W+b([0-9A-Za-z-_]+)\\b\W*$",
+ r"\bbcinfo\b",
+ r"\bopt\b",
+ r"\blibbcc.so\b"]:
+ tool_match = re.match(r"^(\\)?((\| )?)\W+b([\.0-9A-Za-z-_]+)\\b\W*$",
pattern)
tool_pipe = tool_match.group(2)
tool_name = tool_match.group(4)