r600: use min_dx10/max_dx10 instead of min/max
I believe this is the safe thing to do, especially ever since the driver
actually generates NaNs for muls too.
The ISA docs are not very helpful here, however the dx10 versions will pick
a non-nan result over a NaN one (this is also the ieee754 behavior), whereas
the non-dx10 ones will pick the NaN (verified by newly changed piglit
isinf-and-isnan test).
Other "modern" drivers will most likely do the same.
This was shown to make some difference for bug 103544, albeit it is not
required to fix it.
Reviewed-by: Dave Airlie <airlied@redhat.com>
diff --git a/src/gallium/drivers/r600/r600_shader.c b/src/gallium/drivers/r600/r600_shader.c
index 0fa2a1f..805b3b6 100644
--- a/src/gallium/drivers/r600/r600_shader.c
+++ b/src/gallium/drivers/r600/r600_shader.c
@@ -9175,8 +9175,9 @@
[TGSI_OPCODE_DP3] = { ALU_OP2_DOT4_IEEE, tgsi_dp},
[TGSI_OPCODE_DP4] = { ALU_OP2_DOT4_IEEE, tgsi_dp},
[TGSI_OPCODE_DST] = { ALU_OP0_NOP, tgsi_opdst},
- [TGSI_OPCODE_MIN] = { ALU_OP2_MIN, tgsi_op2},
- [TGSI_OPCODE_MAX] = { ALU_OP2_MAX, tgsi_op2},
+ /* MIN_DX10 returns non-nan result if one src is NaN, MIN returns NaN */
+ [TGSI_OPCODE_MIN] = { ALU_OP2_MIN_DX10, tgsi_op2},
+ [TGSI_OPCODE_MAX] = { ALU_OP2_MAX_DX10, tgsi_op2},
[TGSI_OPCODE_SLT] = { ALU_OP2_SETGT, tgsi_op2_swap},
[TGSI_OPCODE_SGE] = { ALU_OP2_SETGE, tgsi_op2},
[TGSI_OPCODE_MAD] = { ALU_OP3_MULADD_IEEE, tgsi_op3},
@@ -9373,8 +9374,8 @@
[TGSI_OPCODE_DP3] = { ALU_OP2_DOT4_IEEE, tgsi_dp},
[TGSI_OPCODE_DP4] = { ALU_OP2_DOT4_IEEE, tgsi_dp},
[TGSI_OPCODE_DST] = { ALU_OP0_NOP, tgsi_opdst},
- [TGSI_OPCODE_MIN] = { ALU_OP2_MIN, tgsi_op2},
- [TGSI_OPCODE_MAX] = { ALU_OP2_MAX, tgsi_op2},
+ [TGSI_OPCODE_MIN] = { ALU_OP2_MIN_DX10, tgsi_op2},
+ [TGSI_OPCODE_MAX] = { ALU_OP2_MAX_DX10, tgsi_op2},
[TGSI_OPCODE_SLT] = { ALU_OP2_SETGT, tgsi_op2_swap},
[TGSI_OPCODE_SGE] = { ALU_OP2_SETGE, tgsi_op2},
[TGSI_OPCODE_MAD] = { ALU_OP3_MULADD_IEEE, tgsi_op3},
@@ -9596,8 +9597,8 @@
[TGSI_OPCODE_DP3] = { ALU_OP2_DOT4_IEEE, tgsi_dp},
[TGSI_OPCODE_DP4] = { ALU_OP2_DOT4_IEEE, tgsi_dp},
[TGSI_OPCODE_DST] = { ALU_OP0_NOP, tgsi_opdst},
- [TGSI_OPCODE_MIN] = { ALU_OP2_MIN, tgsi_op2},
- [TGSI_OPCODE_MAX] = { ALU_OP2_MAX, tgsi_op2},
+ [TGSI_OPCODE_MIN] = { ALU_OP2_MIN_DX10, tgsi_op2},
+ [TGSI_OPCODE_MAX] = { ALU_OP2_MAX_DX10, tgsi_op2},
[TGSI_OPCODE_SLT] = { ALU_OP2_SETGT, tgsi_op2_swap},
[TGSI_OPCODE_SGE] = { ALU_OP2_SETGE, tgsi_op2},
[TGSI_OPCODE_MAD] = { ALU_OP3_MULADD_IEEE, tgsi_op3},
diff --git a/src/gallium/drivers/r600/sb/sb_expr.cpp b/src/gallium/drivers/r600/sb/sb_expr.cpp
index 3dd3a48..7a5d62c 100644
--- a/src/gallium/drivers/r600/sb/sb_expr.cpp
+++ b/src/gallium/drivers/r600/sb/sb_expr.cpp
@@ -753,7 +753,9 @@
n.bc.src[0].abs == n.bc.src[1].abs) {
switch (n.bc.op) {
case ALU_OP2_MIN: // (MIN x, x) => (MOV x)
+ case ALU_OP2_MIN_DX10:
case ALU_OP2_MAX:
+ case ALU_OP2_MAX_DX10:
convert_to_mov(n, v0, n.bc.src[0].neg, n.bc.src[0].abs);
return fold_alu_op1(n);
case ALU_OP2_ADD: // (ADD x, x) => (MUL x, 2)