r600g: fix check_and_set_bank_swizzle
Need to do full check when not all bank swizzles in the group are forced
(e.g. when trying to merge interp_* group with the next instruction)
Note: This is a candidate for the 7.11 branch.
Signed-off-by: Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
diff --git a/src/gallium/drivers/r600/r600_asm.c b/src/gallium/drivers/r600/r600_asm.c
index 0311b56..604cb60 100644
--- a/src/gallium/drivers/r600/r600_asm.c
+++ b/src/gallium/drivers/r600/r600_asm.c
@@ -696,15 +696,19 @@
{
struct alu_bank_swizzle bs;
int bank_swizzle[5];
- int i, r = 0, forced = 0;
+ int i, r = 0, forced = 1;
boolean scalar_only = bc->chip_class == CAYMAN ? false : true;
int max_slots = bc->chip_class == CAYMAN ? 4 : 5;
for (i = 0; i < max_slots; i++) {
- if (slots[i] && slots[i]->bank_swizzle_force) {
- slots[i]->bank_swizzle = slots[i]->bank_swizzle_force;
- forced = 1;
+ if (slots[i]) {
+ if (slots[i]->bank_swizzle_force) {
+ slots[i]->bank_swizzle = slots[i]->bank_swizzle_force;
+ } else {
+ forced = 0;
+ }
}
+
if (i < 4 && slots[i])
scalar_only = false;
}
@@ -714,7 +718,11 @@
/* Just check every possible combination of bank swizzle.
* Not very efficent, but works on the first try in most of the cases. */
for (i = 0; i < 4; i++)
- bank_swizzle[i] = SQ_ALU_VEC_012;
+ if (!slots[i] || !slots[i]->bank_swizzle_force)
+ bank_swizzle[i] = SQ_ALU_VEC_012;
+ else
+ bank_swizzle[i] = slots[i]->bank_swizzle;
+
bank_swizzle[4] = SQ_ALU_SCL_210;
while(bank_swizzle[4] <= SQ_ALU_SCL_221) {
@@ -751,11 +759,13 @@
bank_swizzle[4]++;
} else {
for (i = 0; i < max_slots; i++) {
- bank_swizzle[i]++;
- if (bank_swizzle[i] <= SQ_ALU_VEC_210)
- break;
- else
- bank_swizzle[i] = SQ_ALU_VEC_012;
+ if (!slots[i] || !slots[i]->bank_swizzle_force) {
+ bank_swizzle[i]++;
+ if (bank_swizzle[i] <= SQ_ALU_VEC_210)
+ break;
+ else
+ bank_swizzle[i] = SQ_ALU_VEC_012;
+ }
}
}
}