On 28/07/2025 13:54, Tobias Burnus wrote:
Hi Andrew,
thanks for all the suggestions and the review!
Andrew Stubbs wrote:
...
+ /* CDNA3: VALU writes VGPR/VCC: v_readlane, v_readfirstlane,
v_cmp,
+ v_add_*i/u, v_sub_*i/u, v_div_*scale - followed by:
+ - VALU reads SGPR as
Hi Andrew,
thanks for all the suggestions and the review!
Andrew Stubbs wrote:
...
+ /* CDNA3: VALU writes VGPR/VCC: v_readlane, v_readfirstlane,
v_cmp,
+ v_add_*i/u, v_sub_*i/u, v_div_*scale - followed by:
+ - VALU reads SGPR as constant requires 1 waite state
+ -
On 28/07/2025 12:06, Tobias Burnus wrote:
Hi Andrew,
Andrew Stubbs wrote:
+static bool
+gcn_v_cmp_insn_p (attr_type type)
+{
+ return type == TYPE_VOPC || type == TYPE_VOP3A;
}
There are many vop3a encoded instructions. I don't understand how this
uniquely identifies v_cmp instructions?
It
Hi Andrew,
Andrew Stubbs wrote:
+static bool
+gcn_v_cmp_insn_p (attr_type type)
+{
+ return type == TYPE_VOPC || type == TYPE_VOP3A;
}
There are many vop3a encoded instructions. I don't understand how this
uniquely identifies v_cmp instructions?
It doesn't - how about an attribute variant?
On 25/07/2025 11:54, Tobias Burnus wrote:
There are still issues with MI300, some which get resolved by adding s_nop.
One case where it is exactly known where the s_nop fixes a fail is for
libgomp.c-c++-common/task-detach-10.c, where libgomp/single.c's
GOMP_single_start() never returns 1, such t
