gallivm: Increase lp_build_rsqrt() precision.

Add an iteration step, which makes rqsqrt precision go from 12bits to
24, and fixes RSQ/NRM test case of PSPrecision/VSPrevision DCTs.

There are no uses of this function outside shader translation.

diff --git a/src/gallium/auxiliary/gallivm/lp_bld_arit.c b/src/gallium/auxiliary/gallivm/lp_bld_arit.c
index 06e5debe4a3a512ffebdd4fd0099bb2ae58dc3c7..fce4685cc2d2cd95179c7097f788bb9df84cec60 100644 (file)
--- a/src/gallium/auxiliary/gallivm/lp_bld_arit.c
+++ b/src/gallium/auxiliary/gallivm/lp_bld_arit.c
@@ -1645,7 +1645,7 @@ lp_build_rsqrt(struct lp_build_context *bld,
    assert(type.floating);
 
    if (util_cpu_caps.has_sse && type.width == 32 && type.length == 4) {
-      const unsigned num_iterations = 0;
+      const unsigned num_iterations = 1;
       LLVMValueRef res;
       unsigned i;