On 2024-11-01 19:47, Richard Earnshaw (lists) wrote:
On 24/10/2024 20:12, Torbjörn SVENSSON wrote:Ok for trunk? -- With the changes in r15-1579-g792f97b44ff, the test_vmul_n_16x8 function does not contain any vdup.16 q* r* instruction with -mfloat-abi=softfp. The differnce between r15-1578-g5185274c76c and r15-1579-g792f97b44ff with -mfloat-abi=softfp for the function is: .global test_vmul_n_16x8 .syntax unified .arm .type test_vmul_n_16x8, %function test_vmul_n_16x8: @ args = 4, pretend = 0, frame = 0 @ frame_needed = 0, uses_anonymous_args = 0 @ link register save eliminated. vmov d16, r0, r1 @ v8hf vmov d17, r2, r3 - ldrh r3, [sp] @ __fp16 - vdup.16 q9, r3 + vld1.16 {d18[], d19[]}, [sp] vmul.f16 q8, q9, q8 vmov r0, r1, d16 @ v8hf vmov r2, r3, d17 bx lr .size test_vmul_n_16x8, .-test_vmul_n_16x8 gcc/testsuite/ChangeLog: * gcc.target/arm/armv8_2-fp16-neon-2.c: Expect 3 vdup.16 q* r* when in arm_hf_eabi else 2. Signed-off-by: Torbjörn SVENSSON <[email protected]>OK.
Pushed as r15-4929-ge3f2db9ef42. Kind regards, Torbjörn
