Skip to content

Mlir rocblas succeeds #1

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 83 commits into from
Jan 19, 2022
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
83 commits
Select commit Hold shift + click to select a range
ea0662a
Update rules_cc
keith Dec 16, 2021
bc32ea5
Internal change
Dec 28, 2021
ea96586
Integrate LLVM at llvm/llvm-project@8414321becdb
Dec 29, 2021
148e5ae
Integrate LLVM at llvm/llvm-project@fdd741dd3181
Dec 29, 2021
a2236fb
Integrate LLVM at llvm/llvm-project@4fedd4be385e
Dec 29, 2021
b45ea14
Integrate LLVM at llvm/llvm-project@6d702a1e6a06
Dec 29, 2021
3ee98e9
Integrate LLVM at llvm/llvm-project@33fc675e1670
Dec 29, 2021
0ff8ddd
Integrate LLVM at llvm/llvm-project@2ce586bc497f
Dec 29, 2021
f9e6305
Integrate LLVM at llvm/llvm-project@505d57486e57
Dec 29, 2021
cbf7fb4
Integrate LLVM at llvm/llvm-project@09f43c107fc7
Dec 30, 2021
51c52a7
Integrate LLVM at llvm/llvm-project@5da6d26896d1
Dec 30, 2021
225c641
Integrate LLVM at llvm/llvm-project@9ebeac8613c6
Dec 30, 2021
b23557d
Integrate LLVM at llvm/llvm-project@128c6ed73b8f
Dec 30, 2021
af83ec7
Integrate LLVM at llvm/llvm-project@7128bb61fb59
Dec 30, 2021
fdbab6f
Integrate LLVM at llvm/llvm-project@0c6979b2d64d
Dec 30, 2021
319d8f3
Integrate LLVM at llvm/llvm-project@4943cda3988a
Dec 30, 2021
9558d7f
Integrate LLVM at llvm/llvm-project@7d659c6ac741
Dec 31, 2021
c2ad498
Integrate LLVM at llvm/llvm-project@41454ab25645
Dec 31, 2021
c56f677
Integrate LLVM at llvm/llvm-project@249a5fb005ea
Dec 31, 2021
5a10618
Integrate LLVM at llvm/llvm-project@5527139302d9
pifon2a Jan 3, 2022
9583059
Integrate LLVM at llvm/llvm-project@127d95544164
pifon2a Jan 3, 2022
6b8428c
Integrate LLVM at llvm/llvm-project@862fffd8231c
pifon2a Jan 3, 2022
f47028d
Merge pull request #86 from keith:ks/update-rules_cc
copybara-github Jan 3, 2022
cdc7f0d
Move RewriteVectorMultiReduction pass to codegen_passes.td.
akuegel Jan 4, 2022
9df1dcf
Move AlignedAllocation pass to codegen_passes.td.
akuegel Jan 4, 2022
cd9b911
Integrate LLVM at llvm/llvm-project@cb9ccd38c55f
pifon2a Jan 4, 2022
70494c0
Integrate LLVM at llvm/llvm-project@4c2aba999e54
pifon2a Jan 4, 2022
1480cb8
Integrate LLVM at llvm/llvm-project@3064dd8ccffc
pifon2a Jan 4, 2022
9f93687
Update Eigen to: https://gitlab.com/libeigen/eigen/-/commit/e939c06b0…
Jan 4, 2022
751f4ef
Integrate LLVM at llvm/llvm-project@11a46b174923
Jan 4, 2022
0c3c60a
Integrate LLVM at llvm/llvm-project@37be74885946
Jan 5, 2022
cb4a44b
Make `tfrt_gpu.blas/dnn/solver.handle` only refer to context instead …
chsigg Jan 5, 2022
e48d5d5
Integrate LLVM at llvm/llvm-project@b5149f4e66a4
pifon2a Jan 5, 2022
2b7b134
[tfrt:jit] Disable specialization if operands have static shapes
ezhulenev Jan 5, 2022
fa3a125
Add a cache for GpuContext and the associated ResourceContext.
chsigg Jan 5, 2022
56a51b7
Flip Identifier to StringAttr
jpienaar Jan 5, 2022
70fd50d
Integrate LLVM at llvm/llvm-project@564bcf9d0243
Jan 6, 2022
5c2c0d3
Wrap `tfrt_gpu.blas/dnn/solver.create` in `tfrt.once` during lowering.
chsigg Jan 6, 2022
f0d0541
Integrate LLVM at llvm/llvm-project@c5965a411c63
pifon2a Jan 6, 2022
8350ac6
Add MNIST test.
chsigg Jan 6, 2022
b496e21
Integrate LLVM at llvm/llvm-project@15c7e3ee1597
Jan 7, 2022
3bf5e37
Integrate LLVM at llvm/llvm-project@75d65293ca83
pifon2a Jan 7, 2022
354bc0f
Fix race condition for the KeepAsyncValuePayloadOnError variant of As…
jing-dong Jan 7, 2022
6ecf55d
Integrate LLVM at llvm/llvm-project@4a661602ef2d
Jan 7, 2022
4d9b657
Integrate LLVM at llvm/llvm-project@f96ab6cc1b20
Jan 7, 2022
58aacf0
Integrate LLVM at llvm/llvm-project@669eb6004670
Jan 8, 2022
859c9d4
Integrate LLVM at llvm/llvm-project@e745507eda27
pifon2a Jan 10, 2022
f57c209
Integrate LLVM at llvm/llvm-project@ad1b8772cf6b
Jan 10, 2022
216e913
[tfrt:jit] NFC: Move jit compilation passes/kernels to a separate bac…
ezhulenev Jan 10, 2022
8997563
Integrate LLVM at llvm/llvm-project@4e77868c7c4b
Jan 10, 2022
50ab57f
[tfrt:jit] Check that symbol lookup succeeded before building Executable
ezhulenev Jan 11, 2022
7dbe5f7
[tfrt:jit] Explicitly register CRunnerUtils symbols with JIT session
ezhulenev Jan 11, 2022
6cf0eab
Integrate LLVM at llvm/llvm-project@a74826d30a2e
Jan 11, 2022
6893009
[tfrt:jit] Move vector.multi_reduction rewrite to TF CPURT pipeline.
pifon2a Jan 11, 2022
2b1dc45
Integrate LLVM at llvm/llvm-project@98d51c2542dc
Jan 11, 2022
72956ec
[tf:tfrt] Forward operands with a correct Tensor metadata, and also c…
ezhulenev Jan 11, 2022
723c79c
[tfrt] Report consistent error messages for invalid AsyncValue::get a…
ezhulenev Jan 11, 2022
0c27d6b
Integrate LLVM at llvm/llvm-project@f77d115cc136
Jan 12, 2022
1ea78c2
Register additional Tensor dialect information.
tpopp Jan 12, 2022
6ef053e
Integrate LLVM at llvm/llvm-project@6c654b519835
Jan 12, 2022
1f7eb64
Integrate LLVM at llvm/llvm-project@c490f8feb71e
Jan 12, 2022
c6da6b0
[tf:tfrt] Always specialize compiled binaries
ezhulenev Jan 12, 2022
1b2af50
[tfrt:jit] Record time it took to compile the executable
ezhulenev Jan 12, 2022
cc7772a
[tfrt:jit] NFC: Rename cpurt* files to jitrt*
ezhulenev Jan 12, 2022
a9b8210
Integrate LLVM at llvm/llvm-project@9a1ce35d7e7f
Jan 13, 2022
a2385f7
Implement cost model to drive determining parallelization granularity.
Jan 13, 2022
3ebd9a1
[tfrt:jit] NFC: Rename c++ namespace to ::tfrt::jitrt
ezhulenev Jan 13, 2022
9e5a326
Fix race condition in tfrt print output.
jing-dong Jan 13, 2022
27cd314
[tfrt:jit] Rename cpurt MLIR dialect to jitrt
ezhulenev Jan 13, 2022
79512c3
Integrate LLVM at llvm/llvm-project@1093b9f2e984
Jan 14, 2022
c921fb9
[tfrt:jit] Add missing operator<< for kernel context type
ezhulenev Jan 15, 2022
36500e4
[tfrt:jit] Remove resolved TODO
ezhulenev Jan 16, 2022
57264e9
Integrate LLVM at llvm/llvm-project@0d3f2fd26940
tpopp Jan 17, 2022
f065adf
Integrate LLVM at llvm/llvm-project@da77db58d762
tpopp Jan 17, 2022
84b400a
Integrate LLVM at llvm/llvm-project@2325f363010d
tpopp Jan 17, 2022
090e8f9
[tfrt:jit] Add jitrt CallingConvention to get the compiled function A…
ezhulenev Jan 18, 2022
7c91ffe
[tfrt:jit] Add a separate MLIR pipeline that runs at specialization time
ezhulenev Jan 18, 2022
8ffb29d
Integrate LLVM at llvm/llvm-project@f4615feaa1fb
tpopp Jan 18, 2022
cb196ce
[tfrt:jit] Add a separate MLIR pipeline that runs at specialization time
chsigg Jan 18, 2022
452646d
[tfrt:jit] Add a separate MLIR pipeline that runs at specialization time
ezhulenev Jan 18, 2022
97e6080
[tfrt:jit] Update rt-to-kernel-function pass to add kernel context ar…
ezhulenev Jan 19, 2022
f206be0
Integrate LLVM at llvm/llvm-project@347d4d7323c4
Jan 19, 2022
b0b89af
rocblas(mlir_tests) passed unit tests
weihanmines Jan 19, 2022
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
55 changes: 40 additions & 15 deletions BUILD
Original file line number Diff line number Diff line change
@@ -1,4 +1,4 @@
load(":build_defs.bzl", "if_google", "if_oss", "tfrt_cc_library")
load(":build_defs.bzl", "if_google", "if_oss", "make_variable", "tfrt_cc_library")

# copybara:uncomment load("//configlang/ncl/build_defs:ncl.bzl", "ncl_test")
load("@bazel_skylib//:bzl_library.bzl", "bzl_library")
Expand Down Expand Up @@ -328,6 +328,19 @@ tfrt_cc_library(
],
)

# Change the maximum reported tracing levels with:
# --//third_party/tf_runtime:TFRT_MAX_TRACING_LEVEL=<value>
make_variable(
name = "TFRT_MAX_TRACING_LEVEL",
build_setting_default = "Verbose",
values = [
"None",
"Default",
"Verbose",
"Debug",
],
)

tfrt_cc_library(
name = "tracing",
srcs = [
Expand All @@ -337,6 +350,8 @@ tfrt_cc_library(
"include/tfrt/tracing/tracing.h",
],
# copybara:uncomment compatible_with = ["//buildenv/target:non_prod"],
defines = ["TFRT_MAX_TRACING_LEVEL=$(TFRT_MAX_TRACING_LEVEL)"],
toolchains = [":TFRT_MAX_TRACING_LEVEL"],
visibility = [":friends"],
deps = [
":support",
Expand All @@ -346,34 +361,42 @@ tfrt_cc_library(

tfrt_cc_library(
name = "simple_tracing_sink",
srcs = [
"lib/tracing/simple_tracing_sink/simple_tracing_sink.cc",
],
hdrs = [
"include/tfrt/tracing/simple_tracing_sink/simple_tracing_sink.h",
],
alwayslink_static_registration_src =
"lib/tracing/simple_tracing_sink/static_registration.cc",
srcs = ["lib/tracing/simple_tracing_sink.cc"],
visibility = [":friends"],
deps = [
":support",
":tracing",
"@llvm-project//llvm:Support",
],
alwayslink = True,
)

tfrt_cc_library(
name = "debug_tracing_sink",
srcs = [
"lib/tracing/debug_tracing_sink.cc",
],
srcs = ["lib/tracing/debug_tracing_sink.cc"],
visibility = [":friends"],
deps = [
":support",
":tracing",
"@llvm-project//llvm:Support",
],
alwayslink = 1,
alwayslink = True,
)

tfrt_cc_library(
name = "nvtx_tracing_sink",
srcs = ["lib/tracing/nvtx_tracing_sink.cc"],
hdrs = ["include/tfrt/tracing/nvtx_tracing_sink.h"],
visibility = [":friends"],
deps = [
":support",
":tracing",
"@llvm-project//llvm:Support",
] + if_google(
["@cuda_headers"],
["@nvtx_headers"],
),
alwayslink = True,
)

tfrt_cc_library(
Expand Down Expand Up @@ -1637,9 +1660,11 @@ tfrt_cc_library(
"@llvm-project//mlir:LinalgOps",
"@llvm-project//mlir:MathDialect",
"@llvm-project//mlir:MemRefDialect",
"@llvm-project//mlir:SCFDialect",
"@llvm-project//mlir:StandardOps",
"@tf_runtime//backends/cpu:cpurt_opdefs",
"@tf_runtime//backends/cpu:rt_opdefs",
"@llvm-project//mlir:VectorOps",
"@tf_runtime//backends/jitrt:jitrt_opdefs",
"@tf_runtime//backends/jitrt:rt_opdefs",
],
)

Expand Down
2 changes: 1 addition & 1 deletion backends/common/lib/compat/eigen/kernels/batch_norm.h
Original file line number Diff line number Diff line change
Expand Up @@ -72,7 +72,7 @@ static AsyncValueRef<Chain> FusedBatchNormV3Impl(
depth_by_one.set(0, depth);

// Reshape and broadcast vectors of [depth] to [rest_size, depth] tensor.
auto to_rest_by_depth = [&](const auto& vec) -> auto {
auto to_rest_by_depth = [&](const auto& vec) {
return vec.reshape(one_by_depth).broadcast(rest_by_one);
};

Expand Down
Loading