diff --git a/target/arm/helper.h b/target/arm/helper.h index 7f0d3b2d84..1e2d7025de 100644 --- a/target/arm/helper.h +++ b/target/arm/helper.h @@ -578,6 +578,13 @@ DEF_HELPER_FLAGS_5(gvec_qrdmlah_s32, TCG_CALL_NO_RWG, DEF_HELPER_FLAGS_5(gvec_qrdmlsh_s32, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(gvec_fcaddh, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(gvec_fcadds, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) +DEF_HELPER_FLAGS_5(gvec_fcaddd, TCG_CALL_NO_RWG, + void, ptr, ptr, ptr, ptr, i32) + #ifdef TARGET_AARCH64 #include "helper-a64.h" #endif diff --git a/target/arm/translate-a64.c b/target/arm/translate-a64.c index e4d2d548ba..efed4fd9d2 100644 --- a/target/arm/translate-a64.c +++ b/target/arm/translate-a64.c @@ -713,6 +713,21 @@ static void gen_gvec_op3_env(DisasContext *s, bool is_q, int rd, is_q ? 16 : 8, vec_full_reg_size(s), 0, fn); } +/* Expand a 3-operand + fpstatus pointer + simd data value operation using + * an out-of-line helper. + */ +static void gen_gvec_op3_fpst(DisasContext *s, bool is_q, int rd, int rn, + int rm, bool is_fp16, int data, + gen_helper_gvec_3_ptr *fn) +{ + TCGv_ptr fpst = get_fpstatus_ptr(is_fp16); + tcg_gen_gvec_3_ptr(vec_full_reg_offset(s, rd), + vec_full_reg_offset(s, rn), + vec_full_reg_offset(s, rm), fpst, + is_q ? 16 : 8, vec_full_reg_size(s), data, fn); + tcg_temp_free_ptr(fpst); +} + /* Set ZF and NF based on a 64 bit result. This is alas fiddlier * than the 32 bit equivalent. */ @@ -10816,7 +10831,7 @@ static void disas_simd_three_reg_same_extra(DisasContext *s, uint32_t insn) int size = extract32(insn, 22, 2); bool u = extract32(insn, 29, 1); bool is_q = extract32(insn, 30, 1); - int feature; + int feature, rot; switch (u * 16 + opcode) { case 0x10: /* SQRDMLAH (vector) */ @@ -10827,6 +10842,16 @@ static void disas_simd_three_reg_same_extra(DisasContext *s, uint32_t insn) } feature = ARM_FEATURE_V8_RDM; break; + case 0xc: /* FCADD, #90 */ + case 0xe: /* FCADD, #270 */ + if (size == 0 + || (size == 1 && !arm_dc_feature(s, ARM_FEATURE_V8_FP16)) + || (size == 3 && !is_q)) { + unallocated_encoding(s); + return; + } + feature = ARM_FEATURE_V8_FCMA; + break; default: unallocated_encoding(s); return; @@ -10866,6 +10891,27 @@ static void disas_simd_three_reg_same_extra(DisasContext *s, uint32_t insn) } return; + case 0xc: /* FCADD, #90 */ + case 0xe: /* FCADD, #270 */ + rot = extract32(opcode, 1, 1); + switch (size) { + case 1: + gen_gvec_op3_fpst(s, is_q, rd, rn, rm, size == 1, rot, + gen_helper_gvec_fcaddh); + break; + case 2: + gen_gvec_op3_fpst(s, is_q, rd, rn, rm, size == 1, rot, + gen_helper_gvec_fcadds); + break; + case 3: + gen_gvec_op3_fpst(s, is_q, rd, rn, rm, size == 1, rot, + gen_helper_gvec_fcaddd); + break; + default: + g_assert_not_reached(); + } + return; + default: g_assert_not_reached(); } diff --git a/target/arm/vec_helper.c b/target/arm/vec_helper.c index 3072df4d77..a868ca6aac 100644 --- a/target/arm/vec_helper.c +++ b/target/arm/vec_helper.c @@ -22,8 +22,21 @@ #include "exec/exec-all.h" #include "exec/helper-proto.h" #include "tcg/tcg-gvec-desc.h" +#include "fpu/softfloat.h" +/* Note that vector data is stored in host-endian 64-bit chunks, + so addressing units smaller than that needs a host-endian fixup. */ +#ifdef HOST_WORDS_BIGENDIAN +#define H1(x) ((x) ^ 7) +#define H2(x) ((x) ^ 3) +#define H4(x) ((x) ^ 1) +#else +#define H1(x) (x) +#define H2(x) (x) +#define H4(x) (x) +#endif + #define SET_QC() env->vfp.xregs[ARM_VFP_FPSCR] |= CPSR_Q static void clear_tail(void *vd, uintptr_t opr_sz, uintptr_t max_sz) @@ -181,3 +194,87 @@ void HELPER(gvec_qrdmlsh_s32)(void *vd, void *vn, void *vm, } clear_tail(d, opr_sz, simd_maxsz(desc)); } + +void HELPER(gvec_fcaddh)(void *vd, void *vn, void *vm, + void *vfpst, uint32_t desc) +{ + uintptr_t opr_sz = simd_oprsz(desc); + float16 *d = vd; + float16 *n = vn; + float16 *m = vm; + float_status *fpst = vfpst; + uint32_t neg_real = extract32(desc, SIMD_DATA_SHIFT, 1); + uint32_t neg_imag = neg_real ^ 1; + uintptr_t i; + + /* Shift boolean to the sign bit so we can xor to negate. */ + neg_real <<= 15; + neg_imag <<= 15; + + for (i = 0; i < opr_sz / 2; i += 2) { + float16 e0 = n[H2(i)]; + float16 e1 = m[H2(i + 1)] ^ neg_imag; + float16 e2 = n[H2(i + 1)]; + float16 e3 = m[H2(i)] ^ neg_real; + + d[H2(i)] = float16_add(e0, e1, fpst); + d[H2(i + 1)] = float16_add(e2, e3, fpst); + } + clear_tail(d, opr_sz, simd_maxsz(desc)); +} + +void HELPER(gvec_fcadds)(void *vd, void *vn, void *vm, + void *vfpst, uint32_t desc) +{ + uintptr_t opr_sz = simd_oprsz(desc); + float32 *d = vd; + float32 *n = vn; + float32 *m = vm; + float_status *fpst = vfpst; + uint32_t neg_real = extract32(desc, SIMD_DATA_SHIFT, 1); + uint32_t neg_imag = neg_real ^ 1; + uintptr_t i; + + /* Shift boolean to the sign bit so we can xor to negate. */ + neg_real <<= 31; + neg_imag <<= 31; + + for (i = 0; i < opr_sz / 4; i += 2) { + float32 e0 = n[H4(i)]; + float32 e1 = m[H4(i + 1)] ^ neg_imag; + float32 e2 = n[H4(i + 1)]; + float32 e3 = m[H4(i)] ^ neg_real; + + d[H4(i)] = float32_add(e0, e1, fpst); + d[H4(i + 1)] = float32_add(e2, e3, fpst); + } + clear_tail(d, opr_sz, simd_maxsz(desc)); +} + +void HELPER(gvec_fcaddd)(void *vd, void *vn, void *vm, + void *vfpst, uint32_t desc) +{ + uintptr_t opr_sz = simd_oprsz(desc); + float64 *d = vd; + float64 *n = vn; + float64 *m = vm; + float_status *fpst = vfpst; + uint64_t neg_real = extract64(desc, SIMD_DATA_SHIFT, 1); + uint64_t neg_imag = neg_real ^ 1; + uintptr_t i; + + /* Shift boolean to the sign bit so we can xor to negate. */ + neg_real <<= 63; + neg_imag <<= 63; + + for (i = 0; i < opr_sz / 8; i += 2) { + float64 e0 = n[i]; + float64 e1 = m[i + 1] ^ neg_imag; + float64 e2 = n[i + 1]; + float64 e3 = m[i] ^ neg_real; + + d[i] = float64_add(e0, e1, fpst); + d[i + 1] = float64_add(e2, e3, fpst); + } + clear_tail(d, opr_sz, simd_maxsz(desc)); +}