summaryrefslogtreecommitdiff
path: root/src/gallium/drivers/r600/r600_asm.c
diff options
context:
space:
mode:
Diffstat (limited to 'src/gallium/drivers/r600/r600_asm.c')
-rw-r--r--src/gallium/drivers/r600/r600_asm.c56
1 files changed, 47 insertions, 9 deletions
diff --git a/src/gallium/drivers/r600/r600_asm.c b/src/gallium/drivers/r600/r600_asm.c
index 066fb67aba4..5ef611993ff 100644
--- a/src/gallium/drivers/r600/r600_asm.c
+++ b/src/gallium/drivers/r600/r600_asm.c
@@ -475,6 +475,23 @@ static int is_alu_mova_inst(struct r600_bytecode *bc, struct r600_bytecode_alu *
}
}
+static int alu_uses_rel(struct r600_bytecode *bc, struct r600_bytecode_alu *alu)
+{
+ unsigned num_src = r600_bytecode_get_num_operands(bc, alu);
+ unsigned src;
+
+ if (alu->dst.rel) {
+ return 1;
+ }
+
+ for (src = 0; src < num_src; ++src) {
+ if (alu->src[src].rel) {
+ return 1;
+ }
+ }
+ return 0;
+}
+
static int is_opcode_in_range(unsigned opcode, unsigned min, unsigned max)
{
return min <= opcode && opcode <= max;
@@ -1047,20 +1064,24 @@ static int merge_inst_groups(struct r600_bytecode *bc, struct r600_bytecode_alu
if (r)
return r;
+// fprintf(stderr, "*** Begin List ***\n");
for (i = 0; i < max_slots; ++i) {
if (prev[i]) {
+// fprintf(stderr, "prev[%i] = %x\n", i, prev[i]->inst);
if (prev[i]->pred_sel)
return 0;
if (is_alu_once_inst(bc, prev[i]))
return 0;
}
if (slots[i]) {
+// fprintf(stderr, "slots[%i] = %x\n", i, slots[i]->inst);
if (slots[i]->pred_sel)
return 0;
if (is_alu_once_inst(bc, slots[i]))
return 0;
}
}
+// fprintf(stderr, "*** End List ***\n");
for (i = 0; i < max_slots; ++i) {
struct r600_bytecode_alu *alu;
@@ -1075,10 +1096,22 @@ static int merge_inst_groups(struct r600_bytecode *bc, struct r600_bytecode_alu
if (r600_bytecode_alu_nliterals(bc, prev[i], prev_literal, &prev_nliteral))
return 0;
if (is_alu_mova_inst(bc, prev[i])) {
- if (have_rel)
+// fprintf(stderr, "Prev[%u] is mova\n", i);
+ if (have_rel) {
+// fprintf(stderr, "slots contins a rel, so cannot merge\n");
return 0;
+ }
have_mova = 1;
}
+
+ if (alu_uses_rel(bc, prev[i])) {
+// fprintf(stderr, "Prev[%u] has rel\n", i);
+ if (have_mova) {
+ return 0;
+ }
+ have_rel = 1;
+ }
+
num_once_inst += is_alu_once_inst(bc, prev[i]);
}
if (slots[i] && r600_bytecode_alu_nliterals(bc, slots[i], literal, &nliteral))
@@ -1126,21 +1159,25 @@ static int merge_inst_groups(struct r600_bytecode *bc, struct r600_bytecode_alu
if (is_nop_inst(bc, alu))
return 0;
- /* Let's check dst gpr. */
- if (alu->dst.rel) {
- if (have_mova)
+ if (is_alu_mova_inst(bc, alu)) {
+// fprintf(stderr, "Slots[%u] is mova\n", i);
+ if (have_rel) {
return 0;
+ }
+ have_mova = 1;
+ }
+
+ if (alu_uses_rel(bc, alu)) {
+// fprintf(stderr, "Slots[%u] has rel\n", i);
+ if (have_mova) {
+ return 0;
+ }
have_rel = 1;
}
/* Let's check source gprs */
num_src = r600_bytecode_get_num_operands(bc, alu);
for (src = 0; src < num_src; ++src) {
- if (alu->src[src].rel) {
- if (have_mova)
- return 0;
- have_rel = 1;
- }
/* Constants don't matter. */
if (!is_gpr(alu->src[src].sel))
@@ -1169,6 +1206,7 @@ static int merge_inst_groups(struct r600_bytecode *bc, struct r600_bytecode_alu
return 0;
/* looks like everything worked out right, apply the changes */
+// fprintf(stderr, "Can merge\n");
/* undo adding previus literals */
bc->cf_last->ndw -= align(prev_nliteral, 2);