Mesa (master): cell: use some SPU intrinsics to get slightly better code in eval_inputs()
Brian Paul
brianp at kemper.freedesktop.org
Mon Feb 16 15:33:27 UTC 2009
Module: Mesa
Branch: master
Commit: c51938afe1626bfccfe38fe2f508bf90e58ca74c
URL: http://cgit.freedesktop.org/mesa/mesa/commit/?id=c51938afe1626bfccfe38fe2f508bf90e58ca74c
Author: Brian Paul <brianp at vmware.com>
Date: Mon Feb 16 08:25:33 2009 -0700
cell: use some SPU intrinsics to get slightly better code in eval_inputs()
Suggested by Jonathan Adamczewski. There may be more places to do this...
---
src/gallium/drivers/cell/spu/spu_tri.c | 11 +++++++----
1 files changed, 7 insertions(+), 4 deletions(-)
diff --git a/src/gallium/drivers/cell/spu/spu_tri.c b/src/gallium/drivers/cell/spu/spu_tri.c
index 04e4584..d727268 100644
--- a/src/gallium/drivers/cell/spu/spu_tri.c
+++ b/src/gallium/drivers/cell/spu/spu_tri.c
@@ -219,10 +219,13 @@ eval_inputs(float x, float y, vector float *fragZ, vector float fragInputs[])
/* linear term */
vector float dadx = setup.coef[attr].dadx;
vector float dady = setup.coef[attr].dady;
- r0 += fragX * splatx(dadx) + fragY * splatx(dady);
- r1 += fragX * splaty(dadx) + fragY * splaty(dady);
- r2 += fragX * splatz(dadx) + fragY * splatz(dady);
- r3 += fragX * splatw(dadx) + fragY * splatw(dady);
+ /* Use SPU intrinsics here to get slightly better code.
+ * originally: r0 += fragX * splatx(dadx) + fragY * splatx(dady);
+ */
+ r0 = spu_madd(fragX, splatx(dadx), spu_madd(fragY, splatx(dady), r0));
+ r1 = spu_madd(fragX, splaty(dadx), spu_madd(fragY, splaty(dady), r1));
+ r2 = spu_madd(fragX, splatz(dadx), spu_madd(fragY, splatz(dady), r2));
+ r3 = spu_madd(fragX, splatw(dadx), spu_madd(fragY, splatw(dady), r3));
if (interp == INTERP_PERSPECTIVE) {
/* perspective term */
r0 *= wInv;
More information about the mesa-commit
mailing list