You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@tvm.apache.org by GitBox <gi...@apache.org> on 2022/06/22 05:51:46 UTC
[GitHub] [tvm] crazydemo opened a new pull request, #11820: [Injective Schedule] make injective ops's opt schedule applied to every output tensor
crazydemo opened a new pull request, #11820:
URL: https://github.com/apache/tvm/pull/11820
Modify the schedule for injective ops. All outputs of injective ops should share the same schedule optimization, e.g. outputs of OP split are all supposed to be lowered into parallel for.
```
new ir:
IRModule({GlobalVar(tvmgen_default_fused_split): PrimFunc([placeholder, T_split_sections, T_split_sections]) attrs={"from_legacy_te_schedule": (bool)1, "global_symbol": "tvmgen_default_fused_split", "tir.noalias": (bool)1} {
buffer_realize T_split_sections([0, 1], [0, 64], [0, 56], [0, 56]) {
parallel (ax0.ax1.fused, 0, 64) {
for (ax2, 0, 56) {
for (ax3.outer, 0, 4) {
vectorized (ax3.inner, 0, 16) {
if (tir.likely(((ax3.inner + (ax3.outer*16)) < 56))) {
T_split_sections[floordiv(ax0.ax1.fused, 64), floormod(ax0.ax1.fused, 64), ax2, (ax3.inner + (ax3.outer*16))] = placeholder[floordiv(ax0.ax1.fused, 64), floormod(ax0.ax1.fused, 64), ax2, (ax3.inner + (ax3.outer*16))]
}
}
}
}
}
buffer_realize T_split_sections([0, 1], [0, 64], [0, 56], [0, 56]) {
parallel (ax0.ax1.fused, 0, 64) {
for (ax2, 0, 56) {
for (ax3.outer, 0, 4) {
vectorized (ax3.inner, 0, 16) {
if (tir.likely(((ax3.inner + (ax3.outer*16)) < 56))) {
T_split_sections[floordiv(ax0.ax1.fused, 64), floormod(ax0.ax1.fused, 64), ax2, (ax3.inner + (ax3.outer*16))] = placeholder[floordiv(ax0.ax1.fused, 64), (floormod(ax0.ax1.fused, 64) + 64), ax2, (ax3.inner + (ax3.outer*16))]
}
}
}
}
}
}
}
}
})
old ir:
IRModule({GlobalVar(tvmgen_default_fused_split): PrimFunc([placeholder, T_split_sections, T_split_sections]) attrs={"from_legacy_te_schedule": (bool)1, "global_symbol": "tvmgen_default_fused_split", "tir.noalias": (bool)1} {
buffer_realize T_split_sections([0, 1], [0, 64], [0, 56], [0, 56]) {
parallel (ax0.ax1.fused, 0, 64) {
for (ax2, 0, 56) {
for (ax3.outer, 0, 4) {
vectorized (ax3.inner, 0, 16) {
if (tir.likely(((ax3.inner + (ax3.outer*16)) < 56))) {
T_split_sections[floordiv(ax0.ax1.fused, 64), floormod(ax0.ax1.fused, 64), ax2, (ax3.inner + (ax3.outer*16))] = placeholder[floordiv(ax0.ax1.fused, 64), floormod(ax0.ax1.fused, 64), ax2, (ax3.inner + (ax3.outer*16))]
}
}
}
}
}
buffer_realize T_split_sections([0, 1], [0, 64], [0, 56], [0, 56]) {
for (ax1, 0, 64) {
for (ax2, 0, 56) {
for (ax3, 0, 56) {
T_split_sections[0, ax1, ax2, ax3] = placeholder[0, (ax1 + 64), ax2, ax3]
}
}
}
}
}
}
})
```
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: commits-unsubscribe@tvm.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [tvm] masahi merged pull request #11820: [Injective Schedule] make injective ops's opt schedule applied to every output tensor
Posted by GitBox <gi...@apache.org>.
masahi merged PR #11820:
URL: https://github.com/apache/tvm/pull/11820
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: commits-unsubscribe@tvm.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
[GitHub] [tvm] crazydemo commented on pull request #11820: [Injective Schedule] make injective ops's opt schedule applied to every output tensor
Posted by GitBox <gi...@apache.org>.
crazydemo commented on PR #11820:
URL: https://github.com/apache/tvm/pull/11820#issuecomment-1166728270
@masahi Could you please help review this PR?
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: commits-unsubscribe@tvm.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org