aboutsummaryrefslogtreecommitdiff
path: root/intl
diff options
context:
space:
mode:
authorKyrylo Tkachov <kyrylo.tkachov@arm.com>2023-06-21 13:43:26 +0100
committerKyrylo Tkachov <kyrylo.tkachov@arm.com>2023-06-21 13:43:26 +0100
commitb375c5340b55d6e7ee703717831dd27528962570 (patch)
tree5d740c7e8f8ac1f4cb763e4c0298b7980a2a8c48 /intl
parent4d9d207c668e757adefcfe5368d86bcc008fa4db (diff)
downloadgcc-b375c5340b55d6e7ee703717831dd27528962570.zip
gcc-b375c5340b55d6e7ee703717831dd27528962570.tar.gz
gcc-b375c5340b55d6e7ee703717831dd27528962570.tar.bz2
aarch64: Avoid same input and output Z register for gather loads
The architecture recommends that load-gather instructions avoid using the same Z register for the load address and the destination, and the Software Optimization Guides for Arm cores recommend that as well. This means that for code like: svuint64_t food (svbool_t p, uint64_t *in, svint64_t offsets, svuint64_t a) { return svadd_u64_x (p, a, svld1_gather_offset(p, in, offsets)); } we'll want to avoid generating the current: food: ld1d z0.d, p0/z, [x0, z0.d] // Z0 reused as input and output. add z0.d, z1.d, z0.d ret However, we still want to avoid generating extra moves where there were none before, so the tight aarch64-sve-acle.exp tests for load gathers should still pass as they are. This patch implements that recommendation for the load gather patterns by: * duplicating the alternatives * marking the output operand as early clobber * Tying the input Z register operand in the original alternatives to 0 * Penalising the original alternatives with '?' This results in a large-ish patch in terms of diff lines but the new compact syntax (thanks Tamar) makes it quite a readable an regular change. The benchmark numbers on a Neoverse V1 on fprate look okay: diff 503.bwaves_r 0.00% 507.cactuBSSN_r 0.00% 508.namd_r 0.00% 510.parest_r 0.55% 511.povray_r 0.22% 519.lbm_r 0.00% 521.wrf_r 0.00% 526.blender_r 0.00% 527.cam4_r 0.56% 538.imagick_r 0.00% 544.nab_r 0.00% 549.fotonik3d_r 0.00% 554.roms_r 0.00% fprate 0.10% Bootstrapped and tested on aarch64-none-linux-gnu. gcc/ChangeLog: * config/aarch64/aarch64-sve.md (mask_gather_load<mode><v_int_container>): Add alternatives to prefer to avoid same input and output Z register. (mask_gather_load<mode><v_int_container>): Likewise. (*mask_gather_load<mode><v_int_container>_<su>xtw_unpacked): Likewise. (*mask_gather_load<mode><v_int_container>_sxtw): Likewise. (*mask_gather_load<mode><v_int_container>_uxtw): Likewise. (@aarch64_gather_load_<ANY_EXTEND:optab><SVE_4HSI:mode><SVE_4BHI:mode>): Likewise. (@aarch64_gather_load_<ANY_EXTEND:optab><SVE_2HSDI:mode><SVE_2BHSI:mode>): Likewise. (*aarch64_gather_load_<ANY_EXTEND:optab><SVE_2HSDI:mode> <SVE_2BHSI:mode>_<ANY_EXTEND2:su>xtw_unpacked): Likewise. (*aarch64_gather_load_<ANY_EXTEND:optab><SVE_2HSDI:mode> <SVE_2BHSI:mode>_sxtw): Likewise. (*aarch64_gather_load_<ANY_EXTEND:optab><SVE_2HSDI:mode> <SVE_2BHSI:mode>_uxtw): Likewise. (@aarch64_ldff1_gather<mode>): Likewise. (@aarch64_ldff1_gather<mode>): Likewise. (*aarch64_ldff1_gather<mode>_sxtw): Likewise. (*aarch64_ldff1_gather<mode>_uxtw): Likewise. (@aarch64_ldff1_gather_<ANY_EXTEND:optab><VNx4_WIDE:mode> <VNx4_NARROW:mode>): Likewise. (@aarch64_ldff1_gather_<ANY_EXTEND:optab><VNx2_WIDE:mode> <VNx2_NARROW:mode>): Likewise. (*aarch64_ldff1_gather_<ANY_EXTEND:optab><VNx2_WIDE:mode> <VNx2_NARROW:mode>_sxtw): Likewise. (*aarch64_ldff1_gather_<ANY_EXTEND:optab><VNx2_WIDE:mode> <VNx2_NARROW:mode>_uxtw): Likewise. * config/aarch64/aarch64-sve2.md (@aarch64_gather_ldnt<mode>): Likewise. (@aarch64_gather_ldnt_<ANY_EXTEND:optab><SVE_FULL_SDI:mode> <SVE_PARTIAL_I:mode>): Likewise. gcc/testsuite/ChangeLog: * gcc.target/aarch64/sve/gather_earlyclobber.c: New test. * gcc.target/aarch64/sve2/gather_earlyclobber.c: New test.
Diffstat (limited to 'intl')
0 files changed, 0 insertions, 0 deletions