Re-apply unroll=1 and 112 channels for gfx950 (#1706)

* Reapply "[SRC] Enable unroll=1 for gfx950 (#1602)" (#1667)
This reverts commit a6972c0d09.

* Reapply "[GRAPH] Increase default nChannels to 112 for gfx950 (#1596)" (#1620)
This reverts commit 1a2eca1756.

[ROCm/rccl commit: 12517a957e]
Этот коммит содержится в:
Nilesh M Negi authored and GitHub committed 2025-05-28 14:58:10 -05:00
1 родитель 22120c6303
Коммит 19ed482121
9 files changed
+123 -41

No files matched your search

+9 -6
Просмотреть файл
@@ -34,17 +34,20 @@ struct ncclKernelMatch {
};
#ifdef ENABLE_COLLTRACE
#define ncclGetKernelIndex(p_comm) ((p_comm)->unroll + ((p_comm)->collTraceEnabled ? 2 : 0))
static ncclKernelMatch const ncclKerns[4] = {
{(void *)ncclDevKernel_Generic, true},
#define ncclGetKernelIndex(p_comm) ((p_comm)->unroll + ((p_comm)->collTraceEnabled ? 3 : 0))
static ncclKernelMatch const ncclKerns[6] = {
{(void *)ncclDevKernel_Generic_1, true},
{(void *)ncclDevKernel_Generic_2, true},
{(void *)ncclDevKernel_Generic_4, true},
{(void *)ncclDevKernelDebug_Generic, true},
{(void *)ncclDevKernelDebug_Generic_1, true},
{(void *)ncclDevKernelDebug_Generic_2, true},
{(void *)ncclDevKernelDebug_Generic_4, true}
};
#else
#define ncclGetKernelIndex(p_comm) ((p_comm)->unroll)
static ncclKernelMatch const ncclKerns[2] = {
{(void*)ncclDevKernel_Generic, true},
static ncclKernelMatch const ncclKerns[3] = {
{(void*)ncclDevKernel_Generic_1, true},
{(void*)ncclDevKernel_Generic_2, true},
{(void*)ncclDevKernel_Generic_4, true}
};
#endif