[llvm-profdata] Refactoring Sample Profile Reader to increase FDO build speed using MD5 as key to Sample Profile map

This is phase 1 of multiple planned improvements on the sample profile loader.

[llvm-profdata] Refactoring Sample Profile Reader to increase FDO build speed using MD5 as key to Sample Profile map

This is phase 1 of multiple planned improvements on the sample profile loader. The major change is to use MD5 hash code ((instead of the function itself) as the key to look up the function offset table and the profiles, which significantly reduce the time it takes to construct the map.

The optimization is based on the fact that many practical sample profiles are using MD5 values for function names to reduce profile size, so we shouldn't need to convert the MD5 to a string and then to a SampleContext and use it as the map's key, because it's extremely slow.

Several changes to note:

(1) For non-CS SampleContext, if it is already MD5 string, the hash value will be its integral value, instead of hashing the MD5 again. In phase 2 this is going to be optimized further using a union to represent MD5 function (without converting it to string) and regular function names.

(2) The SampleProfileMap is a wrapper to *map<uint64_t, FunctionSamples>, while providing interface allowing using SampleContext as key, so that existing code still work. It will check for MD5 collision (unlikely but not too unlikely, since we only takes the lower 64 bits) and handle it to at least guarantee compilation correctness (conflicting old profile is dropped, instead of returning an old profile with inconsistent context). Other code should not try to use MD5 as key to access the map directly, because it will not be able to handle MD5 collision at all. (see exception at (5) )

(3) Any SampleProfileMap::emplace() followed by SampleContext assignment if newly inserted, should be replaced with SampleProfileMap::Create(), which does the same thing.

(4) Previously we ensure an invariant that in SampleProfileMap, the key is equal to the Context of the value, for profile map that is eventually being used for output (as in llvm-profdata/llvm-profgen). Since the key became MD5 hash, only the value keeps the context now, in several places where an intermediate SampleProfileMap is created, each new FunctionSample's context is set immediately after insertion, which is necessary to "remember" the context otherwise irretrievable.

(5) When reading a profile, we cache the MD5 values of all functions, because they are used at least twice (one to index into FuncOffsetTable, the other into SampleProfileMap, more if there are additional sections), in this case the SampleProfileMap is directly accessed with MD5 value so that we don't recalculate it each time (expensive)

Performance impact:
When reading a ~1GB extbinary profile (fixed length MD5, not compressed) with 10 million function names and 2.5 million top level functions (non CS functions, each function has varying nesting level from 0 to 20), this patch improves the function offset table loading time by 20%, and improves full profile read by 5%.

Reviewed By: davidxl, snehasish

Differential Revision: https://reviews.llvm.org/D147740

show more ...

Revert "[llvm-profdata] Refactoring Sample Profile Reader to increase FDO build speed using MD5 as key to Sample Profile map"

This reverts commit 66ba71d913df7f7cd75e92c0c4265932b7c93292.

Addressin

Revert "[llvm-profdata] Refactoring Sample Profile Reader to increase FDO build speed using MD5 as key to Sample Profile map"

This reverts commit 66ba71d913df7f7cd75e92c0c4265932b7c93292.

Addressing issues found by:
https://lab.llvm.org/buildbot/#/builders/245/builds/11732
https://lab.llvm.org/buildbot/#/builders/187/builds/12251
https://lab.llvm.org/buildbot/#/builders/186/builds/11099
https://lab.llvm.org/buildbot/#/builders/182/builds/6976

show more ...

[llvm-profdata] Refactoring Sample Profile Reader to increase FDO build speed using MD5 as key to Sample Profile map

This is phase 1 of multiple planned improvements on the sample profile loader.

[llvm-profdata] Refactoring Sample Profile Reader to increase FDO build speed using MD5 as key to Sample Profile map

This is phase 1 of multiple planned improvements on the sample profile loader. The major change is to use MD5 hash code ((instead of the function itself) as the key to look up the function offset table and the profiles, which significantly reduce the time it takes to construct the map.

The optimization is based on the fact that many practical sample profiles are using MD5 values for function names to reduce profile size, so we shouldn't need to convert the MD5 to a string and then to a SampleContext and use it as the map's key, because it's extremely slow.

Several changes to note:

(1) For non-CS SampleContext, if it is already MD5 string, the hash value will be its integral value, instead of hashing the MD5 again. In phase 2 this is going to be optimized further using a union to represent MD5 function (without converting it to string) and regular function names.

(2) The SampleProfileMap is a wrapper to *map<uint64_t, FunctionSamples>, while providing interface allowing using SampleContext as key, so that existing code still work. It will check for MD5 collision (unlikely but not too unlikely, since we only takes the lower 64 bits) and handle it to at least guarantee compilation correctness (conflicting old profile is dropped, instead of returning an old profile with inconsistent context). Other code should not try to use MD5 as key to access the map directly, because it will not be able to handle MD5 collision at all. (see exception at (5) )

(3) Any SampleProfileMap::emplace() followed by SampleContext assignment if newly inserted, should be replaced with SampleProfileMap::Create(), which does the same thing.

(4) Previously we ensure an invariant that in SampleProfileMap, the key is equal to the Context of the value, for profile map that is eventually being used for output (as in llvm-profdata/llvm-profgen). Since the key became MD5 hash, only the value keeps the context now, in several places where an intermediate SampleProfileMap is created, each new FunctionSample's context is set immediately after insertion, which is necessary to "remember" the context otherwise irretrievable.

(5) When reading a profile, we cache the MD5 values of all functions, because they are used at least twice (one to index into FuncOffsetTable, the other into SampleProfileMap, more if there are additional sections), in this case the SampleProfileMap is directly accessed with MD5 value so that we don't recalculate it each time (expensive)

Performance impact:
When reading a ~1GB extbinary profile (fixed length MD5, not compressed) with 10 million function names and 2.5 million top level functions (non CS functions, each function has varying nesting level from 0 to 20), this patch improves the function offset table loading time by 20%, and improves full profile read by 5%.

Reviewed By: davidxl, snehasish

Differential Revision: https://reviews.llvm.org/D147740

show more ...

Revert "[llvm-profdata] Refactoring Sample Profile Reader to increase FDO build speed using MD5 as key to Sample Profile map"

This reverts commit 12e9c7aaa66b7624b5d7666ce2794d912bf9e4b7.

The commi

Revert "[llvm-profdata] Refactoring Sample Profile Reader to increase FDO build speed using MD5 as key to Sample Profile map"

This reverts commit 12e9c7aaa66b7624b5d7666ce2794d912bf9e4b7.

The commit has broken the buildbot, see comment https://reviews.llvm.org/D147740#4451540

show more ...

[llvm-profdata] Refactoring Sample Profile Reader to increase FDO build speed using MD5 as key to Sample Profile map

This is phase 1 of multiple planned improvements on the sample profile loader.

[llvm-profdata] Refactoring Sample Profile Reader to increase FDO build speed using MD5 as key to Sample Profile map

This is phase 1 of multiple planned improvements on the sample profile loader. The major change is to use MD5 hash code ((instead of the function itself) as the key to look up the function offset table and the profiles, which significantly reduce the time it takes to construct the map.

The optimization is based on the fact that many practical sample profiles are using MD5 values for function names to reduce profile size, so we shouldn't need to convert the MD5 to a string and then to a SampleContext and use it as the map's key, because it's extremely slow.

Several changes to note:

(1) For non-CS SampleContext, if it is already MD5 string, the hash value will be its integral value, instead of hashing the MD5 again. In phase 2 this is going to be optimized further using a union to represent MD5 function (without converting it to string) and regular function names.

(2) The SampleProfileMap is a wrapper to *map<uint64_t, FunctionSamples>, while providing interface allowing using SampleContext as key, so that existing code still work. It will check for MD5 collision (unlikely but not too unlikely, since we only takes the lower 64 bits) and handle it to at least guarantee compilation correctness (conflicting old profile is dropped, instead of returning an old profile with inconsistent context). Other code should not try to use MD5 as key to access the map directly, because it will not be able to handle MD5 collision at all. (see exception at (5) )

(3) Any SampleProfileMap::emplace() followed by SampleContext assignment if newly inserted, should be replaced with SampleProfileMap::Create(), which does the same thing.

(4) Previously we ensure an invariant that in SampleProfileMap, the key is equal to the Context of the value, for profile map that is eventually being used for output (as in llvm-profdata/llvm-profgen). Since the key became MD5 hash, only the value keeps the context now, in several places where an intermediate SampleProfileMap is created, each new FunctionSample's context is set immediately after insertion, which is necessary to "remember" the context otherwise irretrievable.

(5) When reading a profile, we cache the MD5 values of all functions, because they are used at least twice (one to index into FuncOffsetTable, the other into SampleProfileMap, more if there are additional sections), in this case the SampleProfileMap is directly accessed with MD5 value so that we don't recalculate it each time (expensive)

Performance impact:
When reading a ~1GB extbinary profile (fixed length MD5, not compressed) with 10 million function names and 2.5 million top level functions (non CS functions, each function has varying nesting level from 0 to 20), this patch improves the function offset table loading time by 20%, and improves full profile read by 5%.

Reviewed By: davidxl, snehasish

Differential Revision: https://reviews.llvm.org/D147740

show more ...

Revert "[llvm-profdata] Refactoring Sample Profile Reader to increase FDO build speed using MD5 as key to Sample Profile map"

This reverts commit 31af18bccea95fe1ae8aa2c51cf7c8e92a1c208e.

This chan

Revert "[llvm-profdata] Refactoring Sample Profile Reader to increase FDO build speed using MD5 as key to Sample Profile map"

This reverts commit 31af18bccea95fe1ae8aa2c51cf7c8e92a1c208e.

This change is causing build failures on many Windows build bots:

https://lab.llvm.org/buildbot/#/builders/216/builds/22833
https://lab.llvm.org/buildbot/#/builders/123/builds/19602
https://lab.llvm.org/buildbot/#/builders/172/builds/28315
https://lab.llvm.org/buildbot/#/builders/119/builds/13870
https://lab.llvm.org/buildbot/#/builders/233/builds/794
https://lab.llvm.org/buildbot/#/builders/235/builds/387
https://lab.llvm.org/buildbot/#/builders/13/builds/36921
https://lab.llvm.org/buildbot/#/builders/127/builds/50510

show more ...

[llvm-profdata] Refactoring Sample Profile Reader to increase FDO build speed using MD5 as key to Sample Profile map

This is phase 1 of multiple planned improvements on the sample profile loader.

[llvm-profdata] Refactoring Sample Profile Reader to increase FDO build speed using MD5 as key to Sample Profile map

This is phase 1 of multiple planned improvements on the sample profile loader. The major change is to use MD5 hash code ((instead of the function itself) as the key to look up the function offset table and the profiles, which significantly reduce the time it takes to construct the map.

The optimization is based on the fact that many practical sample profiles are using MD5 values for function names to reduce profile size, so we shouldn't need to convert the MD5 to a string and then to a SampleContext and use it as the map's key, because it's extremely slow.

Several changes to note:

(1) For non-CS SampleContext, if it is already MD5 string, the hash value will be its integral value, instead of hashing the MD5 again. In phase 2 this is going to be optimized further using a union to represent MD5 function (without converting it to string) and regular function names.

(2) The SampleProfileMap is a wrapper to *map<uint64_t, FunctionSamples>, while providing interface allowing using SampleContext as key, so that existing code still work. It will check for MD5 collision (unlikely but not too unlikely, since we only takes the lower 64 bits) and handle it to at least guarantee compilation correctness (conflicting old profile is dropped, instead of returning an old profile with inconsistent context). Other code should not try to use MD5 as key to access the map directly, because it will not be able to handle MD5 collision at all. (see exception at (5) )

(3) Any SampleProfileMap::emplace() followed by SampleContext assignment if newly inserted, should be replaced with SampleProfileMap::Create(), which does the same thing.

(4) Previously we ensure an invariant that in SampleProfileMap, the key is equal to the Context of the value, for profile map that is eventually being used for output (as in llvm-profdata/llvm-profgen). Since the key became MD5 hash, only the value keeps the context now, in several places where an intermediate SampleProfileMap is created, each new FunctionSample's context is set immediately after insertion, which is necessary to "remember" the context otherwise irretrievable.

(5) When reading a profile, we cache the MD5 values of all functions, because they are used at least twice (one to index into FuncOffsetTable, the other into SampleProfileMap, more if there are additional sections), in this case the SampleProfileMap is directly accessed with MD5 value so that we don't recalculate it each time (expensive)

Performance impact:
When reading a ~1GB extbinary profile (fixed length MD5, not compressed) with 10 million function names and 2.5 million top level functions (non CS functions, each function has varying nesting level from 0 to 20), this patch improves the function offset table loading time by 20%, and improves full profile read by 5%.

Reviewed By: davidxl, snehasish

Differential Revision: https://reviews.llvm.org/D147740

show more ...

[llvm-profdata] Deprecate Compact Binary Sample Profile Format

Remove support for compact binary sample profile format

Reviewed By: davidxl, wenlei

Differential Revision: https://reviews.llvm.org/

[llvm-profdata] Deprecate Compact Binary Sample Profile Format

Remove support for compact binary sample profile format

Reviewed By: davidxl, wenlei

Differential Revision: https://reviews.llvm.org/D149400

show more ...

The test added in d8690bc requires ZLIB, so only run the test if LLVM_ENABLE_ZLIB is set.

This will fix buildbots that fail because they do not have ZLIB installed such as
https://lab.llvm.org/build

The test added in d8690bc requires ZLIB, so only run the test if LLVM_ENABLE_ZLIB is set.

This will fix buildbots that fail because they do not have ZLIB installed such as
https://lab.llvm.org/buildbot/#/builders/216/builds/18424

show more ...

[llvm-profdata] Add more unit tests to check --output-size-limit feature when compression is enabled

Add unit tests for ExtBinary format with compression. This patch doesn't imply there's a bug in t

[llvm-profdata] Add more unit tests to check --output-size-limit feature when compression is enabled

Add unit tests for ExtBinary format with compression. This patch doesn't imply there's a bug in the existing implementation,
but as a precaution that planned changes to SampleProfileReader may cause it to break.

Reviewed By: snehasish

Differential Revision: https://reviews.llvm.org/D144524

show more ...

[llvm-profdata] Add option to cap profile output size

D139603 (add option to llvm-profdata to reduce output profile size) contains test cases that are not cross-platform. Moving those tests to unit

[llvm-profdata] Add option to cap profile output size

D139603 (add option to llvm-profdata to reduce output profile size) contains test cases that are not cross-platform. Moving those tests to unit test and making sure the feature is callable from llvm library

Reviewed By: snehasish

Differential Revision: https://reviews.llvm.org/D141446

show more ...

[llvm-profdata] Add option to cap profile output size

D139603 (add option to llvm-profdata to reduce output profile size) contains test cases that are not cross-platform. Moving those tests to unit

[llvm-profdata] Add option to cap profile output size

D139603 (add option to llvm-profdata to reduce output profile size) contains test cases that are not cross-platform. Moving those tests to unit test and making sure the feature is callable from llvm library

Reviewed By: snehasish

Differential Revision: https://reviews.llvm.org/D141446

show more ...

Fix to D139603(reverted) - moved size check to unit test so that it is cross-platform

D139603 (add option to llvm-profdata to reduce output profile size) contains test cases that are not cross-platf

Fix to D139603(reverted) - moved size check to unit test so that it is cross-platform

D139603 (add option to llvm-profdata to reduce output profile size) contains test cases that are not cross-platform. Moving those tests to unit test and making sure the feature is callable from llvm library

Reviewed By: snehasish

Differential Revision: https://reviews.llvm.org/D141446

show more ...