Fix ICE when associating DOT_PROD_EXPR in SLP vectorization.
Prevents an internal compiler error (ICE) during SLP vectorization when handling DOT_PROD_EXPR by avoiding non-binary associatable statements.
This commit fixes an internal compiler error (ICE) that could occur during SLP (Superword Level Parallelism) vectorization when dealing with DOT_PROD_EXPR nodes. The issue arose when non-binary associatable statements were incorrectly passed to vect_slp_linearize_chain. The fix adds a check to prevent this, thus avoiding the ICE. A new test case is added to verify the fix.
In Details
During SLP vectorization, the compiler attempts to identify reduction chains. This patch modifies tree-vect-slp.cc to prevent non-binary associatable statements from being passed to vect_slp_linearize_chain when handling DOT_PROD_EXPR nodes. This avoids an ICE (internal compiler error) reported in PR tree-optimization/125185. A new test case, pr125185.c, is added to gcc.dg/torture/.
For Context
Vectorization is a compiler optimization technique that allows the compiler to perform multiple operations simultaneously using vector instructions. SLP (Superword Level Parallelism) vectorization is a specific type of vectorization that identifies sequences of independent operations that can be combined into a single vector operation. This commit fixes a bug that could cause the compiler to crash during SLP vectorization when dealing with dot product expressions (calculating the sum of the products of corresponding elements of two vectors). The fix ensures that the compiler correctly handles these expressions during optimization, preventing unexpected errors.