[Backport rls-v3.7] xe: jit: gemm: disable dynamic quant k-blocking #2772
+4
−4
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
github-actions
bot
added
the
platform:gpu-intel
kealan-barbieri
approved these changes
Feb 27, 2025
|
make test |
skazakov1
approved these changes
Feb 27, 2025
petercad
reviewed
Feb 27, 2025
| {{'E', "gemm", {"F", "O", "S"}, {"T", "N", "N"}}, {-1, -1, {-1, -1, -1}, {-1, -1, -1}, {-1, -1, -1}, {-1, -1, -1}, {16, 16, 1}, "Ixyz"}, "sB64 sB32x2 sB wg 4x8 ca4x2 ks64 af dw nse hi sr sm dm grf256 cr0 sys pab", {8, (LoopType) 0, 256, {(LoopType) 144, (LoopType) 255, (LoopType) 255}, {524288, 262144, 16777216}, {524288, 262144, 16777216}, {32, 16, 64}, {4, 8, 1}, 1, (WGType) 1, 257, 16384, 0, {16, 16, 4}, {false, false, false}}, {'E', 17, {930230, 383972, 0, 0, 0, 0, 1.36662, 2.39816, 6.07666, 16.7056, 0.00930946, 0.00736716, 0.0110739, 1, 1.22963, 1.21426, 6.39235e-14}}}, | ||
| {{'E', "gemm", {"F", "O", "S"}, {"T", "N", "N"}}, {-1, -1, {-1, -1, -1}, {-1, -1, -1}, {-1, -1, -1}, {-1, -1, -1}, {4, 4, 1}, "Ixyz"}, "sB64 sB32x2 sB wg 4x8 ca4x2 ks64 af dw nse hi sr sm dm grf256 cr0 sys pab", {8, (LoopType) 0, 256, {(LoopType) 144, (LoopType) 255, (LoopType) 255}, {524288, 262144, 16777216}, {524288, 262144, 16777216}, {32, 16, 64}, {4, 8, 1}, 1, (WGType) 1, 257, 16384, 0, {4, 4, 4}, {false, false, false}}, {'E', 17, {928487, 383872, 0, 0, 0, 0, 1.36666, 2.39635, 6.08292, 16.706, 0.00967946, 0.0275317, 0.0124194, 0.709751, 1.30644, 1.2369, 3.12589e-14}}}, | ||
| {{'E', "gemm", {"F", "O", "S"}, {"T", "N", "N"}}, {-1, -1, {-1, -1, -1}, {-1, 8, -1}, {-1, -1, -1}, {-1, 8, -1}, {4, 4, 1}, "Iqxy"}, "sB64 sS16 aS wg 2x1x8 ikr af acb sr bk0 bm0 sys pab grf256 rc0", {8, (LoopType) 0, 256, {(LoopType) 0, (LoopType) 1, (LoopType) 2}, {16777216, 8192, 16777216}, {8192, 8192, 16777216}, {16, 8, 64}, {2, 1, 8}, 1, (WGType) 0, 4357, 0, 1024, {4, 4, 4}, {false, false, true}}, {'E', 17, {3.5449e+06, 60571.4, -243099, 15595.1, 0, 0, 1.78243, 2.8889, 2.76679, 6.10171, 0.051381, 0.0216118, 0.0510683, 1, 1.21576, 1.21633, -9.23968e-14}}}, | ||
| {{'E', "gemm", {"F", "O", "S"}, {"T", "N", "N"}}, {-1, -1, {-1, -1, -1}, {-1, 8, -1}, {-1, -1, -1}, {-1, 8, -1}, {4, 4, 1}, "IQxy"}, "sS64 sB32 aB wg 2x1x8 ikr ki64 sys af k64 grf256 acb di sr nch fm pab rc0", {8, (LoopType) 0, 256, {(LoopType) 0, (LoopType) 1, (LoopType) 2}, {8192, 8192, 16777216}, {8192, 8192, 16777216}, {16, 8, 64}, {2, 1, 8}, 1, (WGType) 0, 4357, 0, 1024, {4, 4, 4}, {false, false, true}}, {'W', 1, {128}}}, |
There was a problem hiding this comment.
Fine as is, but the k-blocking factor would disappear after updating the CommonDriverInfo:
Suggested change
| {{'E', "gemm", {"F", "O", "S"}, {"T", "N", "N"}}, {-1, -1, {-1, -1, -1}, {-1, 8, -1}, {-1, -1, -1}, {-1, 8, -1}, {4, 4, 1}, "IQxy"}, "sS64 sB32 aB wg 2x1x8 ikr ki64 sys af k64 grf256 acb di sr nch fm pab rc0", {8, (LoopType) 0, 256, {(LoopType) 0, (LoopType) 1, (LoopType) 2}, {8192, 8192, 16777216}, {8192, 8192, 16777216}, {16, 8, 64}, {2, 1, 8}, 1, (WGType) 0, 4357, 0, 1024, {4, 4, 4}, {false, false, true}}, {'W', 1, {128}}}, | |
| {{'E', "gemm", {"F", "O", "S"}, {"T", "N", "N"}}, {-1, -1, {-1, -1, -1}, {-1, 8, -1}, {-1, -1, -1}, {-1, 8, -1}, {4, 4, 1}, "IQxy"}, "sS64 sB32 aB wg 2x1x8 ikr ki64 sys af k64 grf256 acb di sr nch fm pab rc0 bk0", {8, (LoopType) 0, 256, {(LoopType) 0, (LoopType) 1, (LoopType) 2}, {8192, 8192, 16777216}, {8192, 8192, 16777216}, {16, 8, 64}, {2, 1, 8}, 1, (WGType) 0, 4357, 0, 1024, {4, 4, 4}, {false, false, true}}, {'W', 1, {128}}}, |
There was a problem hiding this comment.
Discussed offline, gemmstone now defaults to bk0 so it doesn't need to be added manually (although it doesn't hurt to include). The updates to this strategy came from updating the CommonDriverInfo, so it's safe to leave, and k-blocking will be disabled, even it it's updated in the future.
petercad
approved these changes
Feb 27, 2025
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Backport of #2731 to rls-v3.7.