i965/fs: Constant-fold immediates in src0 of CMP instructions.

This is like what we do with add/mul, but we also have to flip the
conditional test.

Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>

diff --git a/src/mesa/drivers/dri/i965/brw_eu.c b/src/mesa/drivers/dri/i965/brw_eu.c
index 3b5c4c071e3416f2b5e217661067143bfa00bc73..b59d6b251e6d0ae28801f2160c288a1d5b5af6d6 100644 (file)
--- a/src/mesa/drivers/dri/i965/brw_eu.c
+++ b/src/mesa/drivers/dri/i965/brw_eu.c
@@ -34,6 +34,28 @@
 #include "brw_defines.h"
 #include "brw_eu.h"
 
+/* Returns the corresponding conditional mod for swapping src0 and
+ * src1 in e.g. CMP.
+ */
+uint32_t
+brw_swap_cmod(uint32_t cmod)
+{
+   switch (cmod) {
+   case BRW_CONDITIONAL_Z:
+   case BRW_CONDITIONAL_NZ:
+      return cmod;
+   case BRW_CONDITIONAL_G:
+      return BRW_CONDITIONAL_LE;
+   case BRW_CONDITIONAL_GE:
+      return BRW_CONDITIONAL_L;
+   case BRW_CONDITIONAL_L:
+      return BRW_CONDITIONAL_GE;
+   case BRW_CONDITIONAL_LE:
+      return BRW_CONDITIONAL_G;
+   default:
+      return ~0;
+   }
+}
 
 
 /* How does predicate control work when execution_size != 8?  Do I

diff --git a/src/mesa/drivers/dri/i965/brw_eu.h b/src/mesa/drivers/dri/i965/brw_eu.h
index 2d2ed9de985b94dd86767ed6393c3ff0d2ffa26f..c2e59c1be13da44d153834397b9bb186b880edeb 100644 (file)
--- a/src/mesa/drivers/dri/i965/brw_eu.h
+++ b/src/mesa/drivers/dri/i965/brw_eu.h
@@ -1017,6 +1017,8 @@ void brw_set_src1( struct brw_instruction *insn,
 
 void brw_set_uip_jip(struct brw_compile *p);
 
+uint32_t brw_swap_cmod(uint32_t cmod);
+
 /* brw_optimize.c */
 void brw_optimize(struct brw_compile *p);
 void brw_remove_duplicate_mrf_moves(struct brw_compile *p);

diff --git a/src/mesa/drivers/dri/i965/brw_fs.cpp b/src/mesa/drivers/dri/i965/brw_fs.cpp
index 99cd8f833a7087aaa4447f979dadc8a725b81b19..128bbe93a323788f220da231d8c71aa9dd14f7aa 100644 (file)
--- a/src/mesa/drivers/dri/i965/brw_fs.cpp
+++ b/src/mesa/drivers/dri/i965/brw_fs.cpp
@@ -2938,12 +2938,33 @@ fs_visitor::propagate_constants()
                  progress = true;
               }
               break;
+
            case BRW_OPCODE_CMP:
+              if (i == 1) {
+                 scan_inst->src[i] = inst->src[0];
+                 progress = true;
+              } else if (i == 0 && scan_inst->src[1].file != IMM) {
+                 uint32_t new_cmod;
+
+                 new_cmod = brw_swap_cmod(scan_inst->conditional_mod);
+                 if (new_cmod != ~0u) {
+                    /* Fit this constant in by swapping the operands and
+                     * flipping the test
+                     */
+                    scan_inst->src[0] = scan_inst->src[1];
+                    scan_inst->src[1] = inst->src[0];
+                    scan_inst->conditional_mod = new_cmod;
+                    progress = true;
+                 }
+              }
+              break;
+
            case BRW_OPCODE_SEL:
               if (i == 1) {
                  scan_inst->src[i] = inst->src[0];
                  progress = true;
               }
+              break;
            }
         }