cranelift / PR #1388 Infer REX prefix for SIMD `store` an... · git-cranelift

Stream: git-cranelift

Topic: cranelift / PR #1388 Infer REX prefix for SIMD `store` an...

GitHub (Feb 12 2020 at 23:08):

abrown requested iximeow for a review on PR #1388.

GitHub (Feb 12 2020 at 23:08):

abrown opened PR #1388 from infer-rex-for-simd to master:

[x] This has been discussed in issues #1090 and #1379.

[x] A short description of what this does, why it is needed: this demonstrates inferring REX prefixes for SIMD store and vconst instructions; if we think this is the right direction, we have some additional to do to infer REX prefixes for the following recipes:

[x] This PR contains test cases, if meaningful.

[x] A reviewer from the core maintainer team has been assigned for this PR.

GitHub (Feb 12 2020 at 23:09):

abrown edited PR #1388 from infer-rex-for-simd to master:

[x] This has been discussed in issues #1090 and #1379.

[x] A short description of what this does, why it is needed: this demonstrates inferring REX prefixes for SIMD store and vconst instructions; if we think this is the right direction, we have some additional to do to infer REX prefixes for the following recipes:

__enc_32_64__: fstDisp8, fstDisp32, fld, fldDisp8, fldDisp32, fspillSib32, fregspill32, ffillSib32, fregfill32, frmov, furm, fa, fax, f_ib
__enc_32_64_maybe_isap__: r_ib_unsigned_r, fa_ib, fa, r_ib_unsigned_gpr, fcmp, icscc_fpr
__enc_both__: frurm, pfcmp, fa, furm

[x] This PR contains test cases, if meaningful.

[x] A reviewer from the core maintainer team has been assigned for this PR.

GitHub (Feb 13 2020 at 02:47):

iximeow submitted PR Review.

GitHub (Feb 13 2020 at 02:47):

iximeow submitted PR Review.

GitHub (Feb 13 2020 at 02:47):

iximeow created PR Review Comment:

:tada:

GitHub (Feb 13 2020 at 02:47):

iximeow created PR Review Comment:

I think this also needs a || test_input(1, inst, divert, func, is_extended_reg) in the case that the base register is r8-r15? I have admit I'm not sure which order registers are specified but the other part of the recipe this is used in has me suspect. I'm also not sure how well our tests cover getting less-likely encodings like that - I'd not think about it if I didn't just try to do [r11] a few days ago.

GitHub (Feb 13 2020 at 18:34):

abrown updated PR #1388 from infer-rex-for-simd to master:

[x] This has been discussed in issues #1090 and #1379.

[x] A short description of what this does, why it is needed: this demonstrates inferring REX prefixes for SIMD store and vconst instructions; if we think this is the right direction, we have some additional to do to infer REX prefixes for the following recipes:

__enc_32_64__: fstDisp8, fstDisp32, fld, fldDisp8, fldDisp32, fspillSib32, fregspill32, ffillSib32, fregfill32, frmov, furm, fa, fax, f_ib
__enc_32_64_maybe_isap__: r_ib_unsigned_r, fa_ib, fa, r_ib_unsigned_gpr, fcmp, icscc_fpr
__enc_both__: frurm, pfcmp, fa, furm

[x] This PR contains test cases, if meaningful.

[x] A reviewer from the core maintainer team has been assigned for this PR.

GitHub (Feb 13 2020 at 18:48):

abrown updated PR #1388 from infer-rex-for-simd to master:

[x] This has been discussed in issues #1090 and #1379.

[x] A short description of what this does, why it is needed: this demonstrates inferring REX prefixes for SIMD store and vconst instructions; if we think this is the right direction, we have some additional to do to infer REX prefixes for the following recipes:

__enc_32_64__: fstDisp8, fstDisp32, fld, fldDisp8, fldDisp32, fspillSib32, fregspill32, ffillSib32, fregfill32, frmov, furm, fa, fax, f_ib
__enc_32_64_maybe_isap__: r_ib_unsigned_r, fa_ib, fa, r_ib_unsigned_gpr, fcmp, icscc_fpr
__enc_both__: frurm, pfcmp, fa, furm

[x] This PR contains test cases, if meaningful.

[x] A reviewer from the core maintainer team has been assigned for this PR.

GitHub (Feb 13 2020 at 23:08):

iximeow submitted PR Review.

GitHub (Feb 19 2020 at 17:24):

abrown merged PR #1388.

Last updated: Nov 22 2024 at 17:03 UTC