abrown opened PR #1248 from swizzle to master:
This implements the SIMD swizzle instruction using the
PADDUSB 0x70 mask + PSHUFBcode sequence proposed by @zeux. This implementation has an extra instruction toMOVthe mask into a register but this could be removed by a future optimization adding additional addressing modes to arithmetic instructions (in this caseuadd_sat).This PR depends on #1246 being merged.
abrown updated PR #1248 from swizzle to master:
This implements the SIMD swizzle instruction using the
PADDUSB 0x70 mask + PSHUFBcode sequence proposed by @zeux. This implementation has an extra instruction toMOVthe mask into a register but this could be removed by a future optimization adding additional addressing modes to arithmetic instructions (in this caseuadd_sat).This PR depends on #1246 being merged.
abrown updated PR #1248 from swizzle to master:
This implements the SIMD swizzle instruction using the
PADDUSB 0x70 mask + PSHUFBcode sequence proposed by @zeux. This implementation has an extra instruction toMOVthe mask into a register but this could be removed by a future optimization adding additional addressing modes to arithmetic instructions (in this caseuadd_sat).This PR depends on #1246 being merged.
abrown edited PR #1248 from swizzle to master:
This implements the SIMD swizzle instruction using the
PADDUSB 0x70 mask + PSHUFBcode sequence proposed by @zeux. This implementation has an extra instruction toMOVthe mask into a register but this could be removed by a future optimization adding additional addressing modes to arithmetic instructions (in this caseuadd_sat).
This PR depends on #1246 being merged[edit: this is now merged]
sunfishcode submitted PR Review.
abrown merged PR #1248.
Last updated: Dec 06 2025 at 06:05 UTC