CELL: add codegen for logic op, color mask

- rtasm_ppc_spe.c, rtasm_ppc_spe.h: added a new macro function
  "spe_load_uint" for loading and splatting unsigned integers
  in a register; it will use "ila" for values 18 bits or less,
  "ilh" for word values that are symmetric across halfwords,
  "ilhu" for values that have zeroes in their bottom halfwords,
  or "ilhu" followed by "iohl" for general 32-bit values.

  Of the 15 color masks of interest, 4 are 18 bits or less,
  2 are symmetric across halfwords, 3 are zero in the bottom
  halfword, and 6 require two instructions to load.

- cell_gen_fragment.c: added full codegen for logic op and
  color mask.

2 files changed, 26 insertions, 1 deletions

diff --git a/src/gallium/auxiliary/rtasm/rtasm_ppc_spe.c b/src/gallium/auxiliary/rtasm/rtasm_ppc_spe.c
index 12e0826fb9b..f60bfba3f51 100644
--- a/src/gallium/auxiliary/rtasm/rtasm_ppc_spe.c
+++ b/src/gallium/auxiliary/rtasm/rtasm_ppc_spe.c
@@ -592,11 +592,32 @@ spe_load_int(struct spe_function *p, unsigned rT, int i)
    }
 }
 
+void spe_load_uint(struct spe_function *p, unsigned rT, unsigned int ui)
+{
+   /* If the whole value is in the lower 18 bits, use ila, which
+    * doesn't sign-extend.  Otherwise, if the two halfwords of
+    * the constant are identical, use ilh.  Otherwise, we have
+    * to use ilhu followed by iohl.
+    */
+   if ((ui & 0xfffc0000) == ui) {
+      spe_ila(p, rT, ui);
+   }
+   else if ((ui >> 16) == (ui & 0xffff)) {
+      spe_ilh(p, rT, ui & 0xffff);
+   }
+   else {
+      spe_ilhu(p, rT, ui >> 16);
+      if (ui & 0xffff)
+         spe_iohl(p, rT, ui & 0xffff);
+   }
+}
+
 
 void
 spe_splat(struct spe_function *p, unsigned rT, unsigned rA)
 {
-   spe_ila(p, rT, 66051);
+   /* Duplicate bytes 0, 1, 2, and 3 across the whole register */
+   spe_ila(p, rT, 0x00010203);
    spe_shufb(p, rT, rA, rA, rT);
 }
 
diff --git a/src/gallium/auxiliary/rtasm/rtasm_ppc_spe.h b/src/gallium/auxiliary/rtasm/rtasm_ppc_spe.h
index 4ef05ea27d1..09400b3fb2a 100644
--- a/src/gallium/auxiliary/rtasm/rtasm_ppc_spe.h
+++ b/src/gallium/auxiliary/rtasm/rtasm_ppc_spe.h
@@ -302,6 +302,10 @@ spe_load_float(struct spe_function *p, unsigned rT, float x);
 extern void
 spe_load_int(struct spe_function *p, unsigned rT, int i);
 
+/** Load/splat immediate unsigned int into rT. */
+extern void
+spe_load_uint(struct spe_function *p, unsigned rT, unsigned int ui);
+
 /** Replicate word 0 of rA across rT. */
 extern void
 spe_splat(struct spe_function *p, unsigned rT, unsigned rA);