diff options
| author | Jon Roelofs <jonathan_roelofs@apple.com> | 2023-08-09 15:23:13 -0700 |
|---|---|---|
| committer | Jon Roelofs <jonathan_roelofs@apple.com> | 2023-08-09 15:24:31 -0700 |
| commit | 9d6db32640d727a1b51fd6b1f9c90ca7258cdb17 (patch) | |
| tree | a7d057acee91caa57ba87b85b594cec3701f4ac9 /llvm/lib | |
| parent | fe04baf1f79c52c89808f3ec41743205227c022d (diff) | |
| download | llvm-9d6db32640d727a1b51fd6b1f9c90ca7258cdb17.zip llvm-9d6db32640d727a1b51fd6b1f9c90ca7258cdb17.tar.gz llvm-9d6db32640d727a1b51fd6b1f9c90ca7258cdb17.tar.bz2 | |
Remove a reference to rdar://4653682 from X86/README.txt
The radar in question has long since been closed, and the issue the README
documented appears to have been fixed (at least as of clang 16, if not
earlier):
https://clang.godbolt.org/z/7KT5nnoTo
Diffstat (limited to 'llvm/lib')
| -rw-r--r-- | llvm/lib/Target/X86/README.txt | 46 |
1 files changed, 0 insertions, 46 deletions
diff --git a/llvm/lib/Target/X86/README.txt b/llvm/lib/Target/X86/README.txt index c06a7b1..717124b 100644 --- a/llvm/lib/Target/X86/README.txt +++ b/llvm/lib/Target/X86/README.txt @@ -1006,52 +1006,6 @@ couple more movls by putting 4(%esp) into %eax instead of %ecx. //===---------------------------------------------------------------------===// -See rdar://4653682. - -From flops: - -LBB1_15: # bb310 - cvtss2sd LCPI1_0, %xmm1 - addsd %xmm1, %xmm0 - movsd 176(%esp), %xmm2 - mulsd %xmm0, %xmm2 - movapd %xmm2, %xmm3 - mulsd %xmm3, %xmm3 - movapd %xmm3, %xmm4 - mulsd LCPI1_23, %xmm4 - addsd LCPI1_24, %xmm4 - mulsd %xmm3, %xmm4 - addsd LCPI1_25, %xmm4 - mulsd %xmm3, %xmm4 - addsd LCPI1_26, %xmm4 - mulsd %xmm3, %xmm4 - addsd LCPI1_27, %xmm4 - mulsd %xmm3, %xmm4 - addsd LCPI1_28, %xmm4 - mulsd %xmm3, %xmm4 - addsd %xmm1, %xmm4 - mulsd %xmm2, %xmm4 - movsd 152(%esp), %xmm1 - addsd %xmm4, %xmm1 - movsd %xmm1, 152(%esp) - incl %eax - cmpl %eax, %esi - jge LBB1_15 # bb310 -LBB1_16: # bb358.loopexit - movsd 152(%esp), %xmm0 - addsd %xmm0, %xmm0 - addsd LCPI1_22, %xmm0 - movsd %xmm0, 152(%esp) - -Rather than spilling the result of the last addsd in the loop, we should have -insert a copy to split the interval (one for the duration of the loop, one -extending to the fall through). The register pressure in the loop isn't high -enough to warrant the spill. - -Also check why xmm7 is not used at all in the function. - -//===---------------------------------------------------------------------===// - Take the following: target datalayout = "e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-f32:32:32-f64:32:64-v64:64:64-v128:128:128-a0:0:64-f80:128:128-S128" |