Commit Graph

4 Commits

Author SHA1 Message Date
Simon Pilgrim 77dcdc2f50 [CostModel][X86] Pre-SSE41 targets can use PMADDWD for sext sub-i16 -> i32
Without SSE41 sext/zext instructions the extensions will be split, meaning that the MUL->PMADDWD fold will split the sext_i32(x) into zext_i32(sext_i16(x))
2021-10-14 12:17:40 +01:00
Simon Pilgrim 8c83bd3bd4 [CostModel][X86] Adjust vXi32 multiply costs if it can be performed using PMADDWD
Update the costs to match the codegen from combineMulToPMADDWD - not only can we use PMADDWD is its zero-extended, but also if its a constant or sign-extended from a vXi16 (which can be replaced with a zero-extension).
2021-09-25 16:28:48 +01:00
Simon Pilgrim f114ef3731 [CostModel][X86] Add generic costs for vXi32 MUL -> v2Xi16 PMADDDW folds
Based off the improved fold in D108522

This should eventually allow us to replace the SLM only cost patterns with generic versions.
2021-09-05 16:08:11 +01:00
Simon Pilgrim 9962ebaee5 [CostModel][X86] Add vXi32 multiply pattern tests
Add tests for vXi32 multiplies where the operands have been extended from vXi8/vXi16
2021-09-05 16:08:11 +01:00