From: Sagar Ghuge <sagar.ghuge@intel.com>
Date: Mon, 22 Jul 2019 23:30:56 +0000 (-0700)
Subject: nir: Optimize umod lowering
X-Git-Url: https://git.libre-soc.org/?a=commitdiff_plain;h=d5992ab134f6d3d8d8a885be50fe2f696fe0354e;p=mesa.git

nir: Optimize umod lowering

We don't have calculate final quotient in order to calculate unsigned
modulo result.  Once we are done with error correction we have partial
result which can be used to find out modulo operation result

Signed-off-by: Sagar Ghuge <sagar.ghuge@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
---

diff --git a/src/compiler/nir/nir_lower_idiv.c b/src/compiler/nir/nir_lower_idiv.c
index 96e9412c80e..c59a3eb8b3d 100644
--- a/src/compiler/nir/nir_lower_idiv.c
+++ b/src/compiler/nir/nir_lower_idiv.c
@@ -39,7 +39,7 @@
 static bool
 convert_instr(nir_builder *bld, nir_alu_instr *alu)
 {
-   nir_ssa_def *numer, *denom, *af, *bf, *a, *b, *q, *r;
+   nir_ssa_def *numer, *denom, *af, *bf, *a, *b, *q, *r, *rt;
    nir_op op = alu->op;
    bool is_signed;
 
@@ -97,35 +97,33 @@ convert_instr(nir_builder *bld, nir_alu_instr *alu)
    /* correction: if modulus >= divisor, add 1 */
    r = nir_imul(bld, q, b);
    r = nir_isub(bld, a, r);
+   rt = nir_uge(bld, r, b);
 
-   r = nir_uge(bld, r, b);
-   r = nir_b2i32(bld, r);
-
-   q = nir_iadd(bld, q, r);
-   if (is_signed)  {
-      /* fix the sign: */
-      r = nir_ixor(bld, numer, denom);
-      r = nir_ilt(bld, r, nir_imm_int(bld, 0));
-      b = nir_ineg(bld, q);
-      q = nir_bcsel(bld, r, b, q);
-
-      if (op == nir_op_imod || op == nir_op_irem) {
-         q = nir_imul(bld, q, denom);
-         q = nir_isub(bld, numer, q);
-         if (op == nir_op_imod) {
-            q = nir_bcsel(bld, nir_ieq(bld, q, nir_imm_int(bld, 0)),
-                          nir_imm_int(bld, 0),
-                          nir_bcsel(bld, r, nir_iadd(bld, q, denom), q));
+   if (op == nir_op_umod) {
+      q = nir_bcsel(bld, rt, nir_isub(bld, r, b), r);
+   } else {
+      r = nir_b2i32(bld, rt);
+
+      q = nir_iadd(bld, q, r);
+      if (is_signed)  {
+         /* fix the sign: */
+         r = nir_ixor(bld, numer, denom);
+         r = nir_ilt(bld, r, nir_imm_int(bld, 0));
+         b = nir_ineg(bld, q);
+         q = nir_bcsel(bld, r, b, q);
+
+         if (op == nir_op_imod || op == nir_op_irem) {
+            q = nir_imul(bld, q, denom);
+            q = nir_isub(bld, numer, q);
+            if (op == nir_op_imod) {
+               q = nir_bcsel(bld, nir_ieq(bld, q, nir_imm_int(bld, 0)),
+                             nir_imm_int(bld, 0),
+                             nir_bcsel(bld, r, nir_iadd(bld, q, denom), q));
+            }
          }
       }
    }
 
-   if (op == nir_op_umod) {
-      /* division result in q */
-      r = nir_imul(bld, q, b);
-      q = nir_isub(bld, a, r);
-   }
-
    assert(alu->dest.dest.is_ssa);
    nir_ssa_def_rewrite_uses(&alu->dest.dest.ssa, nir_src_for_ssa(q));