[PATCH 0/6] Automatically use vector optabs

Richard Sandiford richard.sandiford@arm.com
Mon Nov 9 16:20:00 GMT 2015


The main goal of this series is to allow functions to be vectorised
simply by defining the associated optab.  At the moment you can get
a scalar square root instruction by defining an md pattern like sqrtdf2.
But if you want to have vectorised sqrt, you need to have a target-
specific C-level built-in function for the vector form of sqrt,
implement TARGET_BUILTIN_VECTORIZED_FUNCTION, and expand the sqrt in
the same way that the target expands other directly-called built-in
functions.  That seems unnecessarily indirect, especially when in
practice those target-specific functions tend to use patterns like
sqrtv2df2 anyway.  It also means GCC has less idea what the vector
function is doing.

The series uses the combined_fn enum and internal functions from
the patches I posted on Saturday to make the vectoriser pick up
vector optabs automatically and represent the result as a vector
internal function call.  For example, we can convert BUILT_IN_SQRT{F,,L}
or a scalar IFN_SQRT into a vector IFN_SQRT if the appropriate sqrtM2
optab is defined.

Tested on x86_64-linux-gnu, aarch64-linux-gnu, arm-linux-gnu and
powerpc64-linux-gnu.

Thanks,
Richard



More information about the Gcc-patches mailing list