summaryrefslogtreecommitdiff
path: root/src/mesa/shader/slang/library/slang_common_builtin.gc
diff options
context:
space:
mode:
authorBrian Paul <brianp@vmware.com>2010-03-29 11:21:26 -0600
committerBrian Paul <brianp@vmware.com>2010-03-29 11:47:38 -0600
commitab0b9f80f45fcab55155b0a16b91a56ff4b3db1e (patch)
treeeaa472ecc0e94a5af28a1021814f063cda1bb967 /src/mesa/shader/slang/library/slang_common_builtin.gc
parent588dd187c45490368354f340f5175d2c4d7cbd3c (diff)
glsl: remove rcp from sqrt()
Per a patch from Marek Olšák, we can simply multiply the incoming value by 1/sqrt(x) instead of using rcp. We're keeping the x==0 check to avoid generating NaN for sqrt(0).
Diffstat (limited to 'src/mesa/shader/slang/library/slang_common_builtin.gc')
-rw-r--r--src/mesa/shader/slang/library/slang_common_builtin.gc14
1 files changed, 4 insertions, 10 deletions
diff --git a/src/mesa/shader/slang/library/slang_common_builtin.gc b/src/mesa/shader/slang/library/slang_common_builtin.gc
index a25ca55bc4..3d464630df 100644
--- a/src/mesa/shader/slang/library/slang_common_builtin.gc
+++ b/src/mesa/shader/slang/library/slang_common_builtin.gc
@@ -605,7 +605,7 @@ float sqrt(const float x)
const float nx = -x;
float r;
__asm float_rsq r, x;
- __asm float_rcp r, r;
+ r = r * x;
__asm vec4_cmp __retVal, nx, r, 0.0;
}
@@ -615,8 +615,7 @@ vec2 sqrt(const vec2 x)
vec2 r;
__asm float_rsq r.x, x.x;
__asm float_rsq r.y, x.y;
- __asm float_rcp r.x, r.x;
- __asm float_rcp r.y, r.y;
+ r = r * x;
__asm vec4_cmp __retVal, nx, r, zero;
}
@@ -627,9 +626,7 @@ vec3 sqrt(const vec3 x)
__asm float_rsq r.x, x.x;
__asm float_rsq r.y, x.y;
__asm float_rsq r.z, x.z;
- __asm float_rcp r.x, r.x;
- __asm float_rcp r.y, r.y;
- __asm float_rcp r.z, r.z;
+ r = r * x;
__asm vec4_cmp __retVal, nx, r, zero;
}
@@ -641,10 +638,7 @@ vec4 sqrt(const vec4 x)
__asm float_rsq r.y, x.y;
__asm float_rsq r.z, x.z;
__asm float_rsq r.w, x.w;
- __asm float_rcp r.x, r.x;
- __asm float_rcp r.y, r.y;
- __asm float_rcp r.z, r.z;
- __asm float_rcp r.w, r.w;
+ r = r * x;
__asm vec4_cmp __retVal, nx, r, zero;
}