Add support for ARM half-precision conversion instructions.

This commit is contained in:
Catherine Moore 2008-11-18 15:45:05 +00:00
parent dc80fd5cfc
commit 8e79c3df51
19 changed files with 458 additions and 57 deletions

View file

@ -65,6 +65,7 @@
#define FPU_VFP_EXT_V3 0x01000000 /* VFPv3 insns. */
#define FPU_NEON_EXT_V1 0x00800000 /* Neon (SIMD) insns. */
#define FPU_VFP_EXT_D32 0x00400000 /* Registers D16-D31. */
#define FPU_NEON_FP16 0x00200000 /* Half-precision extensions. */
/* Architectures are the sum of the base and extensions. The ARM ARM (rev E)
defines the following: ARMv3, ARMv3M, ARMv4xM, ARMv4, ARMv4TxM, ARMv4T,
@ -139,6 +140,8 @@
#define FPU_ARCH_NEON_V1 ARM_FEATURE (0, FPU_NEON_EXT_V1)
#define FPU_ARCH_VFP_V3_PLUS_NEON_V1 \
ARM_FEATURE (0, FPU_VFP_V3 | FPU_NEON_EXT_V1)
#define FPU_ARCH_NEON_FP16 \
ARM_FEATURE (0, FPU_VFP_V3 | FPU_NEON_EXT_V1 | FPU_NEON_FP16)
#define FPU_ARCH_VFP_HARD ARM_FEATURE (0, FPU_VFP_HARD)
#define FPU_ARCH_ENDIAN_PURE ARM_FEATURE (0, FPU_ENDIAN_PURE)