mirror of
https://github.com/RPCSX/llvm.git
synced 2024-12-03 01:12:59 +00:00
5918cc57c3
This was originally committed as r255762, but reverted as it broke windows bots. Re-commitiing the exact same patch, as the underlying cause was fixed by r258677. ARMv8.2-A adds 16-bit floating point versions of all existing VFP floating-point instructions. This is an optional extension, so all of these instructions require the FeatureFullFP16 subtarget feature. The assembly for these instructions uses S registers (AArch32 does not have H registers), but the instructions have ".f16" type specifiers rather than ".f32" or ".f64". The top 16 bits of each source register are ignored, and the top 16 bits of the destination register are set to zero. These instructions are mostly the same as the 32- and 64-bit versions, but they use coprocessor 9 rather than 10 and 11. Two new instructions, VMOVX and VINS, have been added to allow packing and extracting two 16-bit floats stored in the top and bottom halves of an S register. New fixup kinds have been added for the PC-relative load and store instructions, but no ELF relocations have been added as they have a range of 512 bytes. Differential Revision: http://reviews.llvm.org/D15038 git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@258678 91177308-0d34-0410-b5e6-96231b3b80d8
187 lines
3.9 KiB
Plaintext
187 lines
3.9 KiB
Plaintext
# RUN: llvm-mc -disassemble -triple armv8a-none-eabi -mattr=+fullfp16 -show-encoding < %s | FileCheck %s
|
|
|
|
# CHECK: vadd.f16 s0, s1, s0
|
|
[0x80,0x09,0x30,0xee]
|
|
|
|
# CHECK: vsub.f16 s0, s1, s0
|
|
[0xc0,0x09,0x30,0xee]
|
|
|
|
# CHECK: vdiv.f16 s0, s1, s0
|
|
[0x80,0x09,0x80,0xee]
|
|
|
|
# CHECK: vmul.f16 s0, s1, s0
|
|
[0x80,0x09,0x20,0xee]
|
|
|
|
# CHECK: vnmul.f16 s0, s1, s0
|
|
[0xc0,0x09,0x20,0xee]
|
|
|
|
# CHECK: vmla.f16 s1, s2, s0
|
|
[0x00,0x09,0x41,0xee]
|
|
|
|
# CHECK: vmls.f16 s1, s2, s0
|
|
[0x40,0x09,0x41,0xee]
|
|
|
|
# CHECK: vnmla.f16 s1, s2, s0
|
|
[0x40,0x09,0x51,0xee]
|
|
|
|
# CHECK: vnmls.f16 s1, s2, s0
|
|
[0x00,0x09,0x51,0xee]
|
|
|
|
# CHECK: vcmp.f16 s0, s1
|
|
[0x60,0x09,0xb4,0xee]
|
|
|
|
# CHECK: vcmp.f16 s2, #0
|
|
[0x40,0x19,0xb5,0xee]
|
|
|
|
# CHECK: vcmpe.f16 s1, s0
|
|
[0xc0,0x09,0xf4,0xee]
|
|
|
|
# CHECK: vcmpe.f16 s0, #0
|
|
[0xc0,0x09,0xb5,0xee]
|
|
|
|
# CHECK: vabs.f16 s0, s0
|
|
[0xc0,0x09,0xb0,0xee]
|
|
|
|
# CHECK: vneg.f16 s0, s0
|
|
[0x40,0x09,0xb1,0xee]
|
|
|
|
# CHECK: vsqrt.f16 s0, s0
|
|
[0xc0,0x09,0xb1,0xee]
|
|
|
|
# CHECK: vcvt.f16.s32 s0, s0
|
|
# CHECK: vcvt.f16.u32 s0, s0
|
|
# CHECK: vcvt.s32.f16 s0, s0
|
|
# CHECK: vcvt.u32.f16 s0, s0
|
|
[0xc0,0x09,0xb8,0xee]
|
|
[0x40,0x09,0xb8,0xee]
|
|
[0xc0,0x09,0xbd,0xee]
|
|
[0xc0,0x09,0xbc,0xee]
|
|
|
|
# CHECK: vcvtr.s32.f16 s0, s1
|
|
# CHECK: vcvtr.u32.f16 s0, s1
|
|
[0x60,0x09,0xbd,0xee]
|
|
[0x60,0x09,0xbc,0xee]
|
|
|
|
# CHECK: vcvt.f16.u32 s0, s0, #20
|
|
# CHECK: vcvt.f16.u16 s0, s0, #1
|
|
# CHECK: vcvt.f16.s32 s1, s1, #20
|
|
# CHECK: vcvt.f16.s16 s17, s17, #1
|
|
# CHECK: vcvt.u32.f16 s12, s12, #20
|
|
# CHECK: vcvt.u16.f16 s28, s28, #1
|
|
# CHECK: vcvt.s32.f16 s1, s1, #20
|
|
# CHECK: vcvt.s16.f16 s17, s17, #1
|
|
[0xc6,0x09,0xbb,0xee]
|
|
[0x67,0x09,0xbb,0xee]
|
|
[0xc6,0x09,0xfa,0xee]
|
|
[0x67,0x89,0xfa,0xee]
|
|
[0xc6,0x69,0xbf,0xee]
|
|
[0x67,0xe9,0xbf,0xee]
|
|
[0xc6,0x09,0xfe,0xee]
|
|
[0x67,0x89,0xfe,0xee]
|
|
|
|
# CHECK: vcvta.s32.f16 s2, s3
|
|
[0xe1,0x19,0xbc,0xfe]
|
|
|
|
# CHECK: vcvtn.s32.f16 s6, s23
|
|
[0xeb,0x39,0xbd,0xfe]
|
|
|
|
# CHECK: vcvtp.s32.f16 s0, s4
|
|
[0xc2,0x09,0xbe,0xfe]
|
|
|
|
# CHECK: vcvtm.s32.f16 s17, s8
|
|
[0xc4,0x89,0xff,0xfe]
|
|
|
|
# CHECK: vcvta.u32.f16 s2, s3
|
|
[0x61,0x19,0xbc,0xfe]
|
|
|
|
# CHECK: vcvtn.u32.f16 s6, s23
|
|
[0x6b,0x39,0xbd,0xfe]
|
|
|
|
# CHECK: vcvtp.u32.f16 s0, s4
|
|
[0x42,0x09,0xbe,0xfe]
|
|
|
|
# CHECK: vcvtm.u32.f16 s17, s8
|
|
[0x44,0x89,0xff,0xfe]
|
|
|
|
# CHECK: vselge.f16 s4, s1, s23
|
|
[0xab,0x29,0x20,0xfe]
|
|
|
|
# CHECK: vselgt.f16 s0, s1, s0
|
|
[0x80,0x09,0x30,0xfe]
|
|
|
|
# CHECK: vseleq.f16 s30, s28, s23
|
|
[0x2b,0xf9,0x0e,0xfe]
|
|
|
|
# CHECK: vselvs.f16 s21, s16, s14
|
|
[0x07,0xa9,0x58,0xfe]
|
|
|
|
# CHECK: vmaxnm.f16 s5, s12, s0
|
|
[0x00,0x29,0xc6,0xfe]
|
|
|
|
# CHECK: vminnm.f16 s0, s0, s12
|
|
[0x46,0x09,0x80,0xfe]
|
|
|
|
# CHECK: vrintz.f16 s3, s24
|
|
[0xcc,0x19,0xf6,0xee]
|
|
|
|
# CHECK: vrintr.f16 s0, s9
|
|
[0x64,0x09,0xb6,0xee]
|
|
|
|
# CHECK: vrintx.f16 s10, s14
|
|
[0x47,0x59,0xb7,0xee]
|
|
|
|
# CHECK: vrinta.f16 s12, s1
|
|
[0x60,0x69,0xb8,0xfe]
|
|
|
|
# CHECK: vrintn.f16 s12, s1
|
|
[0x60,0x69,0xb9,0xfe]
|
|
|
|
# CHECK: vrintp.f16 s12, s1
|
|
[0x60,0x69,0xba,0xfe]
|
|
|
|
# CHECK: vrintm.f16 s12, s1
|
|
[0x60,0x69,0xbb,0xfe]
|
|
|
|
# CHECK: vfma.f16 s2, s7, s4
|
|
[0x82,0x19,0xa3,0xee]
|
|
|
|
# CHECK: vfms.f16 s2, s7, s4
|
|
[0xc2,0x19,0xa3,0xee]
|
|
|
|
# CHECK: vfnma.f16 s2, s7, s4
|
|
[0xc2,0x19,0x93,0xee]
|
|
|
|
# CHECK: vfnms.f16 s2, s7, s4
|
|
[0x82,0x19,0x93,0xee]
|
|
|
|
# CHECK: vmovx.f16 s2, s5
|
|
# CHECK: vins.f16 s2, s5
|
|
[0x62,0x1a,0xb0,0xfe]
|
|
[0xe2,0x1a,0xb0,0xfe]
|
|
|
|
# CHECK: vldr.16 s1, [pc, #6]
|
|
# CHECK: vldr.16 s2, [pc, #510]
|
|
# CHECK: vldr.16 s3, [pc, #-510]
|
|
# CHECK: vldr.16 s4, [r4, #-18]
|
|
[0x03,0x09,0xdf,0xed]
|
|
[0xff,0x19,0x9f,0xed]
|
|
[0xff,0x19,0x5f,0xed]
|
|
[0x09,0x29,0x14,0xed]
|
|
|
|
# CHECK: vstr.16 s1, [pc, #6]
|
|
# CHECK: vstr.16 s2, [pc, #510]
|
|
# CHECK: vstr.16 s3, [pc, #-510]
|
|
# CHECK: vstr.16 s4, [r4, #-18]
|
|
[0x03,0x09,0xcf,0xed]
|
|
[0xff,0x19,0x8f,0xed]
|
|
[0xff,0x19,0x4f,0xed]
|
|
[0x09,0x29,0x04,0xed]
|
|
|
|
# CHECK: vmov.f16 s0, #1.0
|
|
[0x00,0x09,0xb7,0xee]
|
|
|
|
# CHECK: vmov.f16 s1, r2
|
|
# CHECK: vmov.f16 r3, s4
|
|
[0x90,0x29,0x00,0xee]
|
|
[0x10,0x39,0x12,0xee]
|