gallivm: Lower TGSI_OPCODE_MUL to fmul by default
authorTom Stellard <thomas.stellard@amd.com>
Thu, 6 Dec 2012 19:56:21 +0000 (11:56 -0800)
committerTom Stellard <thomas.stellard@amd.com>
Tue, 11 Dec 2012 00:22:37 +0000 (19:22 -0500)
commitffe1794e0c7efc46e7a5056ac222dd081cae4020
tree88c56bf12efd7183e52db334755944b65feb8d93
parent8000e7b4b6e2866b42de19448cbf3412cce1f26c
gallivm: Lower TGSI_OPCODE_MUL to fmul by default

This fixes a number of crashes on r600g due to the fact that
lp_build_mul assumes vector types when optimizing mul to bit shifts.

This bug was uncovered by 0ad1fefd6951aa47ab58a41dc9ee73083cbcf85c
src/gallium/auxiliary/gallivm/lp_bld_tgsi_action.c