r195367 - Fix a crash in EmitStoreThroughExtVectorComponentLValue for vectors of odd sizes.

[Joey Gouly]

Hi Bill,

Could you pull this into the 3.4 release?

Tom Stellard also wants it:
http://lists.cs.uiuc.edu/pipermail/cfe-commits/Week-of-Mon-20131118/093809.h
tml

Thanks,
Joey

-----Original Message-----
From: cfe-commits-bounces at cs.uiuc.edu
[mailto:cfe-commits-bounces at cs.uiuc.edu] On Behalf Of Joey Gouly
Sent: 21 November 2013 17:09
To: cfe-commits at cs.uiuc.edu
Subject: r195367 - Fix a crash in EmitStoreThroughExtVectorComponentLValue
for vectors of odd sizes.

Author: joey
Date: Thu Nov 21 11:09:05 2013
New Revision: 195367

URL: http://llvm.org/viewvc/llvm-project?rev=195367&view=rev
Log:
Fix a crash in EmitStoreThroughExtVectorComponentLValue for vectors of odd
sizes.

In OpenCL a vector of 3 elements, acts like a vector of four elements.
So for a vector of size 3 the '.hi' and '.odd' accessors, would access
the elements {2, 3} and {1, 3} respectively.
However, in EmitStoreThroughExtVectorComponentLValue we are still operating
on
a vector of size 3, so we should only access {2} and {1}. We do this by
checking
the last element to be accessed, and ignore it if it is out-of-bounds.

EmitLoadOfExtVectorElementLValue doesn't have a similar problem, because it
does
a direct shufflevector with undef, so an out-of-bounds access just gives an
undef
value.

Patch by Anastasia Stulova!



Added:
    cfe/trunk/test/CodeGenOpenCL/vector_odd.cl
Modified:
    cfe/trunk/lib/CodeGen/CGExpr.cpp

Modified: cfe/trunk/lib/CodeGen/CGExpr.cpp
URL:
http://llvm.org/viewvc/llvm-project/cfe/trunk/lib/CodeGen/CGExpr.cpp?rev=195
367&r1=195366&r2=195367&view=diff
============================================================================
==
--- cfe/trunk/lib/CodeGen/CGExpr.cpp (original)
+++ cfe/trunk/lib/CodeGen/CGExpr.cpp Thu Nov 21 11:09:05 2013
@@ -1553,6 +1553,12 @@ void CodeGenFunction::EmitStoreThroughEx
       for (unsigned i = 0; i != NumDstElts; ++i)
         Mask.push_back(Builder.getInt32(i));
 
+      // When the vector size is odd and .odd or .hi is used, the last
element
+      // of the Elts constant array will be one past the size of the
vector.
+      // Ignore the last element here, if it is greater than the mask size.
+      if (getAccessedFieldNo(NumSrcElts - 1, Elts) == Mask.size())
+        NumSrcElts--;
+
       // modify when what gets shuffled in
       for (unsigned i = 0; i != NumSrcElts; ++i)
         Mask[getAccessedFieldNo(i, Elts)] = Builder.getInt32(i+NumDstElts);

Added: cfe/trunk/test/CodeGenOpenCL/vector_odd.cl
URL:
http://llvm.org/viewvc/llvm-project/cfe/trunk/test/CodeGenOpenCL/vector_odd.
cl?rev=195367&view=auto
============================================================================
==
--- cfe/trunk/test/CodeGenOpenCL/vector_odd.cl (added)
+++ cfe/trunk/test/CodeGenOpenCL/vector_odd.cl Thu Nov 21 11:09:05 2013
@@ -0,0 +1,17 @@
+// RUN: %clang_cc1 %s -O0 -emit-llvm -o - | FileCheck %s
+
+typedef unsigned char __attribute__((ext_vector_type(3))) uchar3;
+
+//CHECK: {{%.*}} = shufflevector <3 x i8> {{%.*}}, <3 x i8> <i8 1, i8 1, i8
undef>, <3 x i32> <i32 0, i32 3, i32 2>
+
+kernel void test_odd_vector1 (uchar3 lhs)
+{
+  lhs.odd = 1;
+}
+
+//CHECK: {{%.*}} = shufflevector <3 x i8> {{%.*}}, <3 x i8> <i8 2, i8 2, i8
undef>, <3 x i32> <i32 0, i32 1, i32 3>
+
+kernel void test_odd_vector2 (uchar3 lhs)
+{
+  lhs.hi = 2;
+}


_______________________________________________
cfe-commits mailing list
cfe-commits at cs.uiuc.edu
http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits